It has been a busy few months here at THB, including the recent rollout of several new data dashboards (with more to come!):

• Global Tropical Cyclones (with Ryan Maue)

• US Extreme Weather and Climate Change Dashboard

• Decarbonization Around the World

Please explore them — and let me know suggestions for improvements or if anything does not appear to be working like it should. They are work in progress and will be consistently improved and expanded.

On the new decarbonization page I included a tab with per capita carbon dioxide emissions for regions and countries around the world. Below, you can see that the world’s per capita emissions have remained just about constant over the past 60 years, while per capita emissions of both the U.S. and Europe have dropped significantly.

In the coming weeks, I’ll be digging into the per capita emissions data more deeply with an analysis of what it implies for climate scenarios and projections — there are some important surprises there, so stay tuned.

Going into the weekend I thought I’d share my summer book pile and invite your recommendations for good reads in the comments.

Leave a comment

Read more

Subscribe now

I’ve just released a new data dashboard — Decarbonization Around the World.

The site allows you to explore rates of decarbonization for countries and regions around the world. Decarbonization is a reduction in the ratio of carbon dioxide emissions to GDP. Every climate mitigation policy seeks to accelerate the rate of decarbonization — when that rate exceeds the rate of economic growth, emissions go down. Approaching net zero this century requires a significant acceleration in decarbonization.

The image below shows decarbonization for the world, U.S., and EU from 1965 to 2025 — and no signs of any acceleration in the era of climate policy.

The site also allows you to get a sense of the scale of the decarbonization challenge that each country faces. The figure at the top of this post shows that for the U.S. to achieve a 50% reduction in carbon dioxide emissions (from fossil fuels) — assuming growing energy demand at 0.3% per year — would require the equivalent of more than 1,500 new 1 GW nuclear power plants (about 1.2 coming online every week, starting now) or more than 1.3 million new 3 MW wind turbines (more than 1,000 per week).

If you assume that all the waste heat from combustion can be forgone (some can, not all of course), then these numbers are reduced to 0.7 nuclear plants or over 500 wind turbines per week — still a huge number.

You can adjust any of these variables to explore different scenarios.

You can download figures and data. I encourage you to share what you find on your favorite social media platform and to tag me when you do. Click on the button below to start exploring.

Decarbonization Around the World

Check it out — it is work in progress, so suggestions, critique, and advice are all welcome!

Leave a comment

A quick request — please click that “❤️ Like” button before you go.

Appreciated!

Subscribe now

The pattern is very familiar: With every extreme event comes claims — from journalists, scientists, and politicians — that we can reduce the impacts of those events by quickly cutting our emissions of carbon dioxide from the burning of fossil fuels.

Here are a few examples just in the last week with respect to the wildfires across southern France:

• “[I]f we want to reduce our fire risk, the first thing we have to do is reduce our emissions;”

• “To counter this escalating threat from wildfires, societies need to re-double efforts to mitigate the effects of extreme weather and reduce carbon emissions towards net zero;”

• “The science on the cause is unequivocal . . . What needs to be done is equally clear: leave coal, oil and gas behind faster, scale up renewables and protect people where the impacts are already hitting hardest.”

The logic is attractive: emissions cause warming, warming causes fires, so cutting emissions stops the fires.

Today’s post, as in a recent THB analysis of France’s heat waves, takes such policy causality claims seriously, and asks: What effect would cutting emissions — all the way to net zero and beyond — have on wildfires in France?

This post makes three arguments that I suspect, as a set, many will want to disagree with — and I hope that we can discuss them in the comments:

• First, the fire weather climate in southern France has changed. Fire weather days (defined below) have become much more common in recent decades;

• Second, actual fire on the landscape in South France measured as area burned has not increased, and arguably has decreased, over that same period;

• Third, compared to today, aggressive mitigation is not expected to have discernible effects on the number of fire weather days this century, and perhaps much longer. Further, the benefits of mitigation are a function of faith in models not evidence from the real world.

This post considers each in turn.

A note on scope. This analysis covers Southwest France — the Landes region, home to the largest fires of recent years — and the Mediterranean coast where most of France’s big fires have historically burned. It makes no claim about the whole planet, or even the whole country.

Point one: fire weather has changed in Southern France

Fire scientists use the Fire Weather Index, or FWI — a number combining temperature, humidity, wind, and rainfall — to assess a day’s favorability for fire. A hot, dry, windy day has a high score; a cool, damp, still day has a low score. The FWI measures weather, not fire.

The figure above shows the FWI for Mediterranean France for 1980 to 2024 from the Copernicus Climate Data Store’s European Climate Data Explorer — the monthly reanalysis Fire Weather Index for the Mediterranean region.1 This product computes the FWI from the ERA5 reanalysis, which reconstructs past weather from measured conditions.

In the data I downloaded from Copernicus, I used five months — May through September — and averaged them within each year to give one fire-season FWI value per year. You can see in the figure that the index has a lot of variability, but it is also unmistakable that there is an upwards trend over this time series.

Issues of attribution go beyond today’s post, but this evidence shows that weather conditions conducive to fire in Mediterranean France have increased over the past 45 years.

Point two: but the area actually burned did not rise

If fire weather conditions, you might expect the area burned to also increase. In France it has not.

The figures above show from two different data sources — MODIS satellite observations (2002–2024) and MODIS + VIIRS (2012-2026) — trends in area burned across France. Note that the Mediterranean region accounts for about 70% of the nation’s total area burned. The orange bar for 2026 is area burned year to date.

From 2002 to 2024 the total area burned decreased and there is no trend from 2012 to present.

How can fire weather worsen while area burned holds steady or falls?

Weather is just one ingredient for fire. Whether or not a landscape burns depends on other factors as well: how much fuel has accumulated, whether an ignition occurs (in France, humans cause about 95% of ignitions), and how effectively crews suppress fires. France invested heavily in fire prevention and firefighting, and much of Europe watched burned area fall even as the climate warmed, largely through land management and fire suppression.

Thus two things are true: a clear upward trend in fire-favorable weather and a flat-to-declining trend in actual fire on the landscape. They don’t contradict each other — they simply show that the climate does not equal the outcome.

If I had a dollar for every time this past week I saw someone posting online trends in fires in France or area burned as an argument for or against detection of changes in climate I’d be able to buy a house in Mediterranean France — with AC!

Point three: emissions reductions are not a fire risk control knob

Let’s set aside all that complexity — fuel, ignition, suppression, the divergence between climate trends and fire trends — and grant the strongest reasonable version of the mitigation argument:

Suppose we care only about fire-weather risk, and let’s assume further that it is tightly associated with cumulative carbon emissions. If we cut emissions to achieve net zero (and beyond), how much less fire-weather risk would southwest France face by the end of the century?

A 2025 study in npj Natural Hazards — Lanet et al. — gives us information we need to engage this thought experiment with numbers. That paper projected how often summer days would cross the “extreme fire-weather day” threshold — the level that days exceeded only 10% of the time during the 1991–2020 baseline period.2

The study looked at RCP2.6 and RCP4.5 (yes, RCP8.5 is in there also, but I have ignored it in this analysis — I am modeling good practice!).

In the figure below I have graphed what Lanet et al. found for the increase in the probability of extreme weather days for RCP2.6 (blue dot, typically interpreted as being consistent with hitting a 2C target) and RCP4.5 (yellow dot) versus each scenario’s cumulative carbon dioxide emissions — and drew a straight line between them. I next added the two new CMIP7 scenarios that bracket a current/pledged policy range — MEDIUM (orange diamond) and MEDIUM-LOW (green diamond) — based on their cumulative carbon dioxide emissions.

That straight line allows me to estimate changing probability of extreme fire days under the methods of Lamet et al. for the new scenarios that bracket where CMIP7 thinks the world might be currently headed.3 I also calculated the midpoint of the two CMIP7 scenarios. All this is summarized in the table below the figure.

In the table, I converted each scenario’s fire-weather probability into something more tangible: the expected number of extreme fire-weather days in a typical July or August at the end of the century, averaged over 30 years.

A present-day baseline has about 3 extreme days per month. Based on these probabilities, expected variability is about +/- 3 days per month.4 Under the higher-emissions baseline cases, under the methods of Lanet et al, an average August could see 6 to 7-and-a-half such days — about double today.

So the conclusion of Lanet et al. is a strong response of extreme fire weather days to accumulating carbon dioxide — which is perfect for an analysis of the effects of mitigation on their projections of fire risk.

Let’s next introduce strong mitigation — that envisioned under RCP2.6, which reaches net-zero in the 2080s and goes net-negative after that. Under that scenario, at the end of the century July and August would average about 3 extreme days per month — essentially the same as today.5

Mitigation, under these assumptions, provides benefits in preventing modeled risks from increasing, but mitigation does not reduce risk from today. That means if the fires in France today are a problem today, then mitigation has no potential for reducing risks from today for generations at least, if ever.

Let me say that again — Under assumptions most favorable to the fire weather impacts of mitigation,6 end-of-century fire weather resembles the fire weather we already live with.

The CMIP7 midpoint projects 10.8 extreme fire days in July and August, averaged over 2070-2100, while RCP2.6 projects 5.9 — about five fewer days. In the modeled results, mitigation does cut risks relative to the baseline.

So the answer to the question “does cutting carbon lower fire danger in the Landes region of France?” has two defensible answers from a single study depending on the metric: no and yes.

• No, not with respect to today’s level of risks,

• Yes, with respect to a difference between two alternative, modeled futures.

This means that the mitigation benefit — the difference from the path not taken — is unobservable. Strong mitigation has an effect between the modeled RCP2.6 world and a modeled imagined higher-emissions world. But the quantification of benefits exist only in model land — not in the real world.

The yes or no answer depends not only on which baseline is assumed but on which variable you measure. Lanet et al. track two fire-danger indices that, under the same low-emissions pathway, move in opposite directions: the daily Fire Weather Index eases (an extreme day’s odds fall from 0.12 to 0.08 in August), while the seasonal hot-dry index climbs (from 0.03 to 0.15). One risk metric falls under mitigation while another rises — from the same paper.

None of this means that mitigation would not influence future risk of fire weather, but it does mean that the benefits of mitigation only exist insofar as we trust a model to accurately represent the world and our ability to accurately project where the world is headed under alternative pathways.

Were we to introduce in our analysis more variables, more scenarios, more models, more assumptions, more baselines — the scope of plausibly defensible perspectives would multiply. We have entered the garden of forking paths.7

Let’s illustrate this with numbers: If we assume a baseline of RCP4.5, then a signal of mitigation by following RCP2.6 emerges 38 years after the scenarios diverge — which would be sometime well into the second half of this century.8 Assume a baseline that approaches RCP2.6, and the time of emergence stretches to infinity. There are many defensible positions on where the world might be headed as a baseline scenario that fall in between RCP2.6 and RCP4.5 to satisfy every proponent or opponent of mitigation.

Science cannot resolve these perspectives.

The real-world impact of mitigation on fire risk in this case is thus a function of what future one believes we are headed for — which can not be resolved empirically. This is not an argument saying that mitigation can have no benefits for reducing fire risk, but rather, an explanation why our perspectives on what that benefit might be rests on beliefs and faith, not data and science.

Where this leaves us

To their credit, the Lamet et al. study acknowledges that adaptation to fire risk is not just necessary, but essential. Unlike those quoted at the top of this post, they do not treat emissions reductions as a fire weather control knob:

Adapting to the escalating risk of wildfires necessitates a comprehensive approach that addresses the three fundamental components of risk: danger (also referred as hazard), exposure, and vulnerability. Mitigating the hazard involves decreasing the likelihood and potential severity of wildfire events. This can be achieved through public education campaigns that promote fire-safe behaviours among residents, tourists, and professionals. Fuel management practices, such as prescribed burning and mechanical thinning, have proven effective in reducing wildfire severity. Exposure can be curtailed by regulating the expansion of human settlements and infrastructure into fire-prone areas, particularly the wildland–urban interface (WUI). The WUI, where human development meets undeveloped wildland, has been identified as a zone of heightened wildfire risk. Efforts to limit development in these areas can reduce the number of structures and lives at risk. However, it is important to recognise that even without further expansion, existing exposure within the WUI remains substantial. Moreover, many peri-urban areas—often overlapping with the WUI—provide essential affordable housing, underscoring the need to balance wildfire risk reduction with broader social and economic considerations. Finally, reducing vulnerability entails enhancing the resilience of communities and systems to withstand and recover from wildfire events.

This post does not argue against accelerated decarbonization. It argues that justifying decarbonization as a fire-risk reduction policy — as so many are about France this month — misleads the public and distracts from policies that can actually make a difference.

For accelerated decarbonization to gain wide public support, its benefits must be both near-term and perceptible.

Even granting the strongest reasonable assumptions, strong mitigation leaves end-of-century fire weather in Southwest France looking much like today’s — thus, the promised risk reduction cannot be delivered, even in highly simplified models.

Further, the benefits of mitigation for fire weather risk — not actual fire on the landscape — are entirely a function and magnitude of assumptions made about counterfactual futures. That doesn’t make benefits unreal, but it does make them intangible and determined by faith in models and beliefs about the future.

There are many better reasons for accelerating decarbonization of our economies than reducing fire risk in France. If those good reasons for accelerating decarbonization have a potential side benefit of contributing to lower fire risks, even if not completely quantifiable, so much the better.

My advice to climate advocates: Lead with the better arguments. Don’t preach weather control. That is bad science and a false promise.

Mitigation matters, justifications for it should matter too.

A quick request — please click that “❤️ Like” button before you go.

Appreciated!

THB exists because of its free and paid subscribers. Please consider joining the community or an upgrade. Thank you!

Subscribe now

Paid subscribers can download an Excel file below with all sources, data, and calculations in this post.

Read more

Subscribe now

Back in 1999, I published a paper in the Bulletin of the American Meteorological Society with Ken Kunkel and Stan Changnon titled “Temporal Fluctuations in Weather and Climate Extremes That Cause Economic and Human Health Impacts: A Review”. The paper looked at long-term trends in extreme weather in the United States based on the understanding that to understand whether climate has changed in ways that matter to people, you have to examine long-term, high-quality records of variable associated with the extremes that actually cause harm — and you have to try to distinguish any observed trend from the noise of internal variability.

That paper included a time series of the frequency and intensity of heat waves across the United States which was adopted by the U.S. Environmental Protection Agency (EPA) as an official climate indicator, featured on the EPA’s Climate Change Indicators website.

That resource no longer exists. In December 2025, the EPA removed at least 80 climate-related webpages, including the Climate Change Indicators section that housed data on heat waves, coastal flooding, wildfires, snowfall, and droughts. Whatever you think about the politics, the removal took a widely used, taxpayer-funded reference — built on official data and years of careful analysis — offline.

So I build a new and much better extreme weather and climate change dashboard, which I am releasing today.

The THB US Extreme Weather and Climate Change Dashboard

The THB US Extreme Weather and Climate Change Dashboard follows the Intergovernmental Panel on Climate Change’s (IPCC) framework for detecting a change in climate in the context of internal variability. The dashboard tracks 32 variables across 7 types of extremes: heat waves, tornadoes, flooding, drought, winter storms, wildfire, and hurricanes.

THB Extreme Weather Dashboard

A few principles guide the project:

It relies on official data. The underlying records come from authoritative government and scientific sources. The dashboard presents the observational record based on the best available data.

It respects the limits of the data. The site presents data only for the full range judged of sufficient quality for trend analysis. Users are empowered to explore the data over different time windows. Data and figures can be easily downloaded.

It places the data into the context of the IPCC AR6. The dashboard follows the IPCC framework for detection and attribution and explains how the information presented is or is not consistent with the most recent assessment of the IPCC Working Group 1.

Detecting a change in climate demands long records, attention to data quality, and an understanding of historical variability. The dashboard empowers users to look at the data themselves.

Use this page to:

• learn the facts on trends in extreme weather in the U.S.,

• hold the media accountable,

• find graphs and data to share online,

• elevate your game in debates about climate,

• and to jst explore some wonderful data from outstanding scientists and important organizations about this wonderful, complex world we share.

Long-overdue

From one perspective, this dashboard updates our 1999 paper in which Ken Kunkel, Stan Changnon, and I sought to integrate data and research on trends in the extremes with the greatest societal impacts. The dashboard takes that same approach across seven phenomena and 32 variables, using the records we have now instead of the records we had then.

That paper took a few years to complete — Using AI tools, I built this one in less than a month (again, thanks to Ryan Maue for the education!).

I am surprised that no such dashboard exists. The U.S. National Climate Assessment, with far more resources than I have available, could have created an ongoing, browsable set of extreme-weather indicators that anyone can explore. So could any number of research groups and agencies. Yet as far as I can tell, no one has.

So I built it myself.

An invitation

This project remains a work in progress. I plan to add variables, refine sources, fix mistakes, and improve the presentation in collaboration with the readers of THB. If you spot an error, know of a better data source, or think a phenomenon deserves an indicator it lacks, tell me.

Next up — a global dashboard!

Suggestions welcome!

Leave a comment

Spread the word on the new dashboard by clicking that “❤️ Like” button.

Appreciated!

PS. Check out the new landing page for all of my work.

Subscribe now

The image above shows global tropical cyclone landfalls from 1980 through 2020 — via the IBTrACS global best-track data — and plots the difference between the two ENSO phases. Blue indicates coasts where more storms make landfall during the La Niña phase and red shows coasts where El Niño has more landfalls — the darker the shading the wider the gap between phases — using NOAA Climate Prediction Center ONI v5 ENSO index.

For the Caribbean, Central America, and U.S. Gulf and East coasts, the La Niña phase has about twice as many landfalls as during El Niño — 2.1× across the basin and 1.4× along the U.S. coastline. The Northwest Pacific — including the Philippines, China, Japan, and Korea — and the eastern Pacific toward Mexico and the South Pacific islands have more landfalls during the El Niño phase.

El Niño means fewer Atlantic hurricanes. But worldwide, El Niño years have more tropical cyclone activity. Today’s post explores what the emerging El Niño means for tropical cyclone activity and what it implies for U.S. damage potential.

1. What happens in the Atlantic

William Gray1, of Colorado State University, established the core dynamical understandings in the 1980s: During El Niño, warm water in the central and eastern equatorial Pacific drives an atmospheric response that strengthens vertical wind shear over the tropical Atlantic, and shear makes it difficult for developing storms to grow stronger. During La Niña the pattern flips: shear is reduced and the Atlantic main development region is more active.

Accumulated Cyclone Energy (ACE) is a metric that integrates tropical cyclone frequency and intensity, and the figure below shows that ACE values in the Atlantic are generally higher during the La Niña phase than El Niño — but that does not completely rule out the possibility of a very active El Niño year or an inactive La Niña year.

From 1950–2025, the median North Atlantic season had ~129 units of ACE in La Niña years versus ~79 in El Niño years — a ratio of ~1.6×. Neutral years fall in between, at ~97. For major hurricanes La Niña seasons average ~3.6 annually and El Niño seasons average ~2.0.

The figure above shows ACE for 1950 to 2025, shaded by ENSO phase. Most El Niño Atlantic seasons fall below the long-term ACE median of 99.4, but not all do, and vice versa for La Niña.

2. U.S. damage: updating our 1999 analysis

In 1999, Chris Landsea and I published a paper in the Bulletin of the American Meteorological Society, “La Niña, El Niño, and Atlantic Hurricane Damages in the United States.” We assessed U.S. hurricane losses normalized to 1997 values according to ENSO phase and we found a strong relationship — La Niña years had much more normalized losses.

The figure above updates our analysis through 2025, and shows that our original findings have stood the test of time. In fact, the relationship is even stronger than that which we documented in our initial analysis.

• The median La Niña year has ~$12 billion in damage versus ~$2 billion for the median El Niño year — a factor of six difference.

• La Niña years have losses above $10 billion about 52% of the time versus 29% for El Niño years, a 1.8× gap, about what we reported in 1999.

ENSO shifts the odds without removing the possibility of extreme losses. Also, about 40% of years are neutral, also with large variability in losses. A very large loss year can strike in any ENSO phase when a major hurricane meets an exposed coast.

3. The global picture

Summing all six tropical cyclone basins, the median global season generates more ACE in El Niño years than in La Niña years — about 29% higher. El Niño does not merely relocate the planet’s storm energy; it raises the total — led by the Western Pacific, which is responsible for ~40% of global ACE across years.

An El Niño therefore has both a strong Atlantic signal and a strong Pacific signal, but in different directions.

Our 1999 paper covered only U.S. losses, but it turns out those losses are a big part of the global story. The figure below shows that global insured losses are significantly influenced by whether or not the U.S. mainland experiences a major hurricane that hits a populated region.2

Significantly, at the global level, El Niño has more tropical cyclone activity but global economic losses are generally higher in La Niña years.

Where storms hit matters more than how many or how strong. The Pacific generates most of the world’s cyclone energy, but landfalls there cross coastlines with far lower insured values than the U.S. Gulf and Atlantic coasts. So global losses follow the Atlantic ENSO signal, not the global pattern.

The figure above shows that separating years by whether a major hurricane struck the U.S. coast reveals a huge insured loss difference: U.S. major hurricane landfall years have a median insured loss of ~$119 billion; non-landfall years have about ~$47 billion — about a 2.5× difference.

The five costliest years — 2017 (Harvey, Irma, Maria), 2024 (Helene, Milton), 2022 (Ian), 2011, and 2021 (Ida) — all featured major U.S. landfalls, and four of the five were La Niña years.

4. What to expect for 2026?

NOAA has issued an El Niño Advisory, and forecasters expect the event to strengthen into winter, perhaps into a very significant El Niño event.

The Atlantic hurricane season thus far has been mostly dormant: Through 27 July, the Atlantic has produced 2 named storms, no hurricanes, and about 5.3 units of ACE — about 35% the typical year-to-date.

The rest of the world is well above the long-term average with global ACE at about 365 units year-to-date, about 29% above the 1980–present average — and growing fast. A quiet Atlantic and busy Pacific is exactly what the data record would lead us to expect.

If El Niño deepens through the August–October Atlantic hurricane season peak, as expected, history suggest that odds shift toward a below-average U.S. damage season with about a two-to-one chance of a sub-$10-billion loss year. A strong El Niño could see very low U.S. losses — as in 1982 and 1997, years with significant El Niño events.

In contrast, 2004 — an El Niño year, but with 4 hurricane landfalls in Florida (and 3 of them major) causing more than $100 billion in normalized losses — showed that an exposed coast can experience a catastrophic year in any ENSO phase. ENSO changes the odds, but does not eliminate risk.

For decision-makers who act on probabilistic information — reinsurers and commodity brokers among them — a developing strong El Niño has a signal in the Atlantic basin with economic value. However, for anyone on an exposed coastline, in any basin, don’t let your guard down, a big one could hit in any year.

Data sources and references: North Atlantic and global tropical cyclone activity from NHC/CPHC HURDAT2 and JTWC best tracks, via global-tropical-cyclones.com (through 27 July 2026). Global landfall maps from IBTrACS v04r00 (NOAA NCEI), all basins, 1980–2020. Normalized U.S. hurricane damage from Weinkle et al. (2018), extended through 2025 (2024 dollars), building on Pielke and Landsea (1998) and (1999). Global insured catastrophe losses from Swiss Re sigma and Munich Re NatCatSERVICE annual reviews (nominal USD, 2010–2025). ENSO phase by August–September–October Oceanic Niño Index (ONI v5). ENSO status from the NOAA Climate Prediction Center; 2026 Northwest Pacific context from Munich Re. References: Pielke, R. A., Jr., and C. W. Landsea, 1999: La Niña, El Niño, and Atlantic Hurricane Damages in the United States. Bulletin of the American Meteorological Society, 80, 2027–2033. Camargo, S. J., and A. H. Sobel, 2005: Western North Pacific Tropical Cyclone Intensity and ENSO. Journal of Climate, 18, 2996–3006.

Want more from THB on this topic?

Data appendix below

Read more

Subscribe now

In early February 2020, four virologists — Kristian Andersen, Eddie Holmes, Robert Garry, and Andrew Rambaut — told each other and Anthony Fauci — then director of the National Institute of Allergy and Infectious Diseases (NIAID) and member of President Trump’s just-created Coronavirus Task force — privately that a lab origin for SARS-CoV-2 looked plausible.

Weeks later the same group published a paper — “The Proximal Origin of SARS-CoV-2,” hereafter Proximal Origin — declaring flatly that “we do not believe that any type of laboratory-based scenario is plausible.”

That paper became the authoritative justification for treating the lab-leak hypothesis as a conspiracy theory and settled science. Fauci and NIH director Francis Collins promoted Proximal Origin from the White House to quash debate over COVID_19 origins, without ever disclosing that they helped set the paper in motion.

Here at THB, I’ve previously discussed the email trail that shows this gap between private doubt and public certainty forming in the first two weeks of February 2020, before anyone drafted the paper.

What happened after publication remained murkier — did the authors’ doubts fade as real evidence came in, or persist while the group kept defending the paper in public?

This week, Senator Rand Paul (R-KY) released two new batches of the group’s Slack messages — first batch (AM drop) and a second batch (PM drop} — spanning April 2020 through September 2021, with incredible details of how doubt never faded for the Proximal Origin gang, despite their public expressions to the contrary.

Whatever the origins of COVID-19, we know with certainty that there was a concerted effort to cover-up the possibility that COVID-19 may have emerged from a research-related incident. Let’s dive in . . .

A quick primer for anyone catching up

SARS-CoV-2’s spike protein contains a furin cleavage site, a short genetic insertion that helps the virus enter human cells. No other close relative found in nature carries one at that exact spot, which made it a focal point of the origins debate from the start.

Scientists on both sides agree the insertion looks unusual: they disagree about whether evolution might produce it, or if instead it required a lab to intervene.

Proximal Origin argued for the former and dismissed the latter.

“[Baric] would be really in the cross-hairs without [Proximal Origin]”

The details discussed below come from this week’s Slack release, and add significant new context and details.

In a May 1, 2020 exchange, Andersen tells Holmes about a conversation he had with Ralph Baric, who then ran one of the world’s most prominent coronavirus labs at the University of North Carolina. Baric’s group specializes in the exact kind of research at the center of the origins debate: taking coronaviruses found in bats and modifying them in the lab to test whether they can infect human cells.

On what Andersen calls a “White House committee call” in “early January” 2020 — one of the government-convened scientific advisory calls set up in the earliest days of the outbreak, before most of the world grasped its scale — Andersen had predicted two things about the new virus: that its receptor-binding domain (the part of the spike protein that grips onto human cells) would bind unusually well to human receptors, and that its furin cleavage site would turn out biologically active.

Baric brushed him off. Andersen recounts it this way: “He essentially said that was all bullshit and I couldn’t possibly know” (Andersen, Slack messages, PM drop, May 1, 2020). Four months later, after research had confirmed both predictions, Andersen commented to Holmes about Baric’s dismissal: “Well.”

That exchange quickly shifted from Andersen’s victory lap to comments far more significant: Holmes tells Andersen that Proximal Origin “did him a MASSIVE favour” — “him” meaning Baric — and adds:

“He would be really in the cross-hairs without it” (Holmes, Slack messages, PM drop, May 1, 2020).

Andersen agrees, and mentions that a co-author on one of Baric’s papers — a 2015 study Baric ran with the Wuhan Institute of Virology that spliced a bat-virus spike protein onto a lab-adapted coronavirus backbone specifically to test its ability to infect human cells (Menachery et al., Nature Medicine, 2015) — had emailed him the day before to thank him for writing Proximal Origin (Andersen, Slack messages, PM drop, May 1, 2020).

In other words, the authors understood in real time that their published conclusion — that no lab could plausibly have made this virus — doubled as protection for a specific colleague whose own published techniques might have otherwise come under greater scrutiny. Further, the authors were being lauded by colleagues who might have been in “the cross-hairs” for writing the paper.

Five days later, on May 6, 2020, the group faced a version of that same scrutiny itself. Holmes learned that The Daily Telegraph, a British newspaper, intended to publish a story claiming he worked “in the pay of China” (Holmes, Slack messages, PM drop, May 6, 2020) — implying a financial motive behind his public dismissal of the lab-leak theory. Holmes told the others he’d received no such money. But he added a line that undercuts the group’s public position more than it defends him:

“the irony is that we were some of the very first people to actually raise the lab escape issue, although I obviously can’t say that” (Holmes, Slack messages, PM drop, May 6, 2020).

Andersen followed with agreement on that point:

“we were the first to take this seriously — and so we did. And our conclusions are solid and, importantly, stick 100% to the science” (Andersen, Slack messages, PM drop, May 6, 2020).

In public they were taking credit for dismissing the notion that COVID-19 may have come from a research-related incident, while in private they were taking credit for being among the first to assert its plausibility.

May–June 2020, just after Proximal Origin, doubt continues

By May 12, 2020, eight weeks after publication, Holmes asks Andersen and Garry for their

“latest thinking about whether any lab manipulation would involve cell culture or live animal passage” (Holmes, Slack messages, PM drop, May 12, 2020).

Andersen answers:

“I’m starting to worry about cell culture again because of some of the things we’re noticing in the cell culture sequencing experiments” (Andersen, Slack messages, PM drop, May 12, 2020).

He lists three concerns:

• the furin site repeatedly appears and disappears depending on tissue-culture conditions;

• a bovine coronavirus picked up an identical-length furin insertion through tissue culture alone;

• and SARS-CoV-2 itself kept acquiring similar short insertions elsewhere in the spike, but only in culture (Andersen, Slack messages, PM drop, May 12, 2020).

A month later, on June 11, the doubt sharpens. Andersen tells the group:

“some smoking(ish) guns starting to appear that I’m not comfortable with,” and points out that the furin insertion sits exactly where earlier lab experiments on a related coronavirus inserted one experimentally (Andersen, Slack messages, AM drop, June 11, 2020).

In the same message he goes further (emphasis in original):

“we really don’t have any hard evidence one way or the other... we also can’t rule out that somebody actually put it in there. That’s obviously not to say that somebody did, but we can’t rule it out” (Andersen, Slack messages, AM drop, June 11, 2020).

Then, about Proximal Origin’s central conclusion:

“Our paper was pretty strong in saying ‘there’s no way’ — I have less confidence in that statement at this stage” (Andersen, Slack messages, AM drop, June 11, 2020).

The lead author of the world’s most influential lab-leak rebuttal privately walked back its headline claim, just three months after publication, while the paper was being used in public to argue that the COVID-19 origin was settled science.

The same thread has Andersen noting that the paper mistakenly described risky low-containment coronavirus culturing as happening “all over the world,” when in actual fact it concentrated almost exclusively at the Wuhan Institute of Virology (Andersen, Slack messages, AM drop, June 11, 2020).

By the end of the conversation the four scientists traded private odds on lab origin:

• Holmes at 80/20 against,

• Andersen at 70/30,

• Holmes later revising to 90/10 (Andersen and Holmes, Slack messages, AM drop, June 11, 2020).

Months after publicly ruling out a lab origin, the authors debated it privately as a real and significant possibility.

Other scientists raised the same red flags

The new messages also show the group fielding legitimate technical pushback from outside their circle — echoing comments they had made amongst themselves — and dismissed those critiques rather than engaging.

On May 22, 2020, a BBC correspondent forwards a detailed critique from an anonymous molecular biologist:

“It is the acquisition en masse of these 4 key new residues and nothing in flanking sequences that raises the distinct impression that someone deliberately inserted these 4 residues in one step... there is absolutely none — all the residues on each side are identical — this is what is so concerning.” (unnamed critic, quoted in Andersen, Slack messages, PM drop, May 22, 2020).

He asks whether investigators should independently examine WIV’s records. Andersen’s private reply: “’Molecular biologist’ — my arse... I think we know who this is,” (Andersen, Slack messages, PM drop, May 22, 2020). He decides it doesn’t deserve an answer.

Three days later a second scientist raises a parallel, more technical point:

“The furin cleavage site insert has the unusual codon usage for the RR dimer of CGG-CGG... most laboratory use optimized codon primers and kits use CGG routinely; including in primers in published papers from the Wuhan Institute of Virology.” (unnamed critic, quoted in Andersen, Slack messages, PM drop, May 25, 2020).

Andersen calls part of it “pure bs” and moves on (Andersen, Slack messages, PM drop, May 25, 2020). Both comments anticipate arguments the group makes to each other in private within weeks.

September 2021: reverse-engineering the DEFUSE proposal

More consequential material comes from September 2021, after DRASTIC and The Intercept published a leaked 2018 grant proposal — DEFUSE — that EcoHealth Alliance submitted with the Wuhan Institute of Virology and Ralph Baric’s UNC lab. The proposal, written before the pandemic, expressed an intent to insert furin cleavage sites into bat coronaviruses.

Holmes’s reaction, on September 21, does not mince words:

“We still have no evidence for SC2 in that lab and they must have sequenced it first. But it is very striking that a bat-derived CoV with an FCS turns up in a city that was on a grant proposing to muck around with FCSs in bat-derived CoVs” (Holmes, Slack messages, AM drop, September 21, 2021).

Holmes calls the central coincidence of the whole lab-leak hypothesis “very striking” in private, eighteen months after publicly ruling one out.

Once DEFUSE became public, Andersen, Holmes, Garry, and Rambaut reconvened to check whether its methods could explain SARS-CoV-2’s furin site.

In the messages, Andersen lays out a formal work plan to explore that possibility: spike-protein alignments, flagging the grant’s described sites, codon-usage analysis to determine whether the insertion looks designed or natural (Andersen, Slack messages, AM drop, September 21, 2021).

He and Rambaut debate cloning strategy and which codons a lab would plausibly pick if engineering the site deliberately (Andersen and Rambaut, Slack messages, AM drop, September 21, 2021).

Nearly a year and a half after publishing a paper ruling a lab origin implausible, its authors privately ran a forensic check on whether people they knew, using a grant they knew about, had built the virus.

The clearest sign of what drove this exercise comes from Rambaut himself, at the start of that morning’s exchange:

“It really isn’t beyond plausibility that someone went ahead and tried some of this stuff. We just need to re-convince ourselves it is coincidence” (Rambaut, Slack messages, AM drop, September 21, 2021).

This doesn’t sound like a scientist following evidence wherever it leads.

What follows undercuts one of the strongest public claims against a lab origin. Proximal Origin and its authors have repeatedly pointed to the furin insertion sitting “out of frame” as evidence against deliberate design — a lab-made insertion, the argument goes, would look cleaner.

In this exchange, Rambaut and Andersen work through how a lab could produce exactly that messy, “natural-looking” signature on purpose. Rambaut poses a key question:

“Key question is — would anyone chose to put the FCS we see in SC2 in and put it out of frame”? (Rambaut, Slack messages, AM drop, September 21, 2021),

He then provides one answer:

“The alternative is that it was inserted in frame and then mutated to make it look out of frame” (Rambaut, Slack messages, AM drop, September 21, 2021).

Andersen adds a second, more technical route:

“if the use restriction enzymes and there would be an easy way to insert an FCS with PCR utilizing REs already present in the sequence, then it could be a way to do it. Not the optimal way, for sure, but sometimes ease trumps perfection” (Andersen, Slack messages, AM drop, September 21, 2021).

None of this shows SARS-CoV-2 came from a lab.

But it does show that the Proximal Origin authors, eighteen months after publication, treated “out of frame” — one of the specific technical points used publicly to rule out engineering — as something a real lab could plausibly produce using ordinary methods.

Right after Andersen proposes an investigation to explore if the virus’ furin site matched what DEFUSE’s methods would have produced, Rambaut writes:

“And stay off email.”

Then:

“Or carefully curate some emails for the FOI records” (Rambaut, Slack messages, AM drop, September 21, 2021)

These suggest intent to keep their discussions out of any official record, using Slack and Zoom instead of email because email carries more exposure to public-records requests.

What Andersen told Congress was completely different

Andersen later testified under oath before the U.S. Congress. At the July 11, 2023 House Select Subcommittee hearing, he told members:

“By the time we published our final version of Proximal Origin, however, I no longer believed that a ‘culturing’ scenario was plausible” (Andersen testimony, House Select Subcommittee on the Coronavirus Pandemic, July 11, 2023).

He described the zoonosis conclusion as something that

“has only been further supported by additional evidence and studies” ever since, and added that “currently, no such evidence exists” that would justify revisiting that conclusion (Andersen testimony, House Select Subcommittee on the Coronavirus Pandemic, July 11, 2023).

We now know that these statements fail to accurately convey the doubt in Proximal Origin that he had expressed to his co-authors after it was published.

Why this matters

We already knew that Proximal Origin’s conclusions didn’t match what its authors privately believed when they wrote it. These two new batches of Slack messages reveal that after the paper was published, the authors still did not believe what they had written.

When outside scientists raised the same concerns through legitimate channels, the group mocked them in private and ignored them in public, even though their critiques anticipated arguments the group made to itself. And when the group finally investigated whether their own colleagues’ grant-funded research could explain the virus’s key feature, one author’s instinct led him to keep that investigation off any record public-records law could reach.

None of this proves SARS-CoV-2 came from a lab. It does show that the public, and eventually Congress, were badly misled by a small group of scientists in coordination with officials of the U.S. government.

The effect was to reduce the chances that there would have been any high-level investigation of COVID-19 origins — as there should have been — compromising our ability to ever know the details of what might have been one of the most significant failures of science institutions in history.

Comments and discussion welcome!

Leave a comment

More from THB on Proximal Origin

Subscribe now

We have just released a new website that is one-stop shopping for official data on tropical cyclones and climate. The project is a collaboration with Ryan Maue, who is teaching this old dog some new tricks. If you do not follow or subscribe to Ryan’s fantastic weather site, you should (check these out).

The site is Global Tropical Cyclones and can be reached by clicking the button below.

Global Tropical Cyclones

Global Tropical Cyclones is the first of what will be a growing number of data dashboards available here at THB — which you can access from the menu bar at the top under THB Data Dashboards.

Global Tropical Cyclones tracks storm counts by intensity, Accumulated Cyclone Energy (ACE), and more across all six ocean basins and globally, drawing on the operational best-track data from NHC, CPHC, and JTWC via IBTrACS.

The site updates daily with year-to-date totals compared against 1980–present averages, and offers storm rankings, annual rankings, basin-by-basin running totals back to 1966. The site also includes a dedicated climate change section putting current activity in historical context, drawing upon the IPCC AR6 and GFDL.

The site includes a daily briefing on tropical cyclone activity year-to-date and a monthly briefing with the latest on hurricanes and climate change.

It's built for anyone who wants the underlying numbers on tropical cyclone activity rather than headline-driven claims about them. Check it out!

Comments and suggestions welcome!

Leave a comment

Subscribe now

Consider how prominent voices in the climate discussion describe the link between fossil fuels and extreme heat:

• World Weather Attribution claims that scientists can now “predict how heat waves will grow more frequent and intense unless emissions are cut drastically.”

• A climate scientist, David Faranda, argued that to prevent severe early-season heat across India and Pakistan: “The only sustainable solution is to stop burning fossil fuels and reduce emissions.”

• UN Secretary-General António Guterres, in his Call to Action on Extreme Heat, agreed: “To tackle the root cause, countries must phase out fossil fuels — fast and fairly.”

Such claims are ubiquitous: fossil-fuel emissions are framed as a control knob for extreme weather, and turning the knob down will reduce the incidence of weather extremes. In fact, much of climate advocacy is grounded in claims that climate change causes undesirable weather and the solution therefore must be to quickly transition off of fossil fuels.

Today’s post takes such claims seriously and explores the science behind them in the context of European heat waves.

Could decarbonization actually increase European heat?

Recall the causal structure from Part 1. Atmospheric circulation generates a heat wave and dry soils, brightened skies, warmer upwind seas, and city surfaces make it hotter. Greenhouse gases raise the energy in the climate system but do not cause the weather. Cutting carbon dioxide therefore acts on one input only, and slowly: soil moisture, jet-stream patterns, and the heat already stored in the Mediterranean and the Atlantic do not respond to an emissions cut the way a room responds to a thermostat.

Part 1 also discussed the hypothesis that cleaner air drove a large part of Europe’s rapid summer warming since 1980: as sulfate pollution fell, more sunlight reached the surface, adding about half a degree of regional warming (Schumacher et al. 2024) and possibly reshaping circulation patterns toward conditions more favorable for heat waves.

If that hypothesis is correct, then reducing the burning of fossil fuels, and thus carbon dioxide emissions, would also contribute in the near term to increased heat waves in Europe: Stopping the burning of coal, oil, and gas also reduces particulate pollution. Those aerosols mask some warming by reflecting sunlight. Globally, aerosol cooling offsets a substantial share of greenhouse warming; removing the aerosols unmasks that warming. In some projections, the near-term warming from reducing aerosols rivals the warming avoided by cutting the carbon dioxide itself (Palazzo Corner et al. 2023).

The point here is not against cutting carbon dioxide emissions. It is against claiming that reducing those emissions is a solution to Europe’s heat waves as cutting emissions might not deliver what is promised.

When would we detect the effects of reducing emissions?

Let’s assume that the world achieves net-zero carbon dioxide emissions: When would we detect the effects on climate?

The scientific literature provides a clear answer: not soon. Tebaldi and Friedlingstein (2013) found that even under an assumption of strong mitigation, a change in the global mean temperature trajectory might not be detectable until later this century (e.g., “up to 60 y[ears] for RCP4.5 vs. RCP2.6”). At the regional scale the time of emergence of the impacts of mitigation on climate would necessarily be longer than at the global scale. For extreme events, longer still.

For European heat waves under strong mitigation policies: When would European heat wave frequency stop climbing, and when might they turn down?

A sensitivity analysis of a modeled climate under net-zero carbon dioxide emissions — and no other changes to forcings, including aerosols — finds that heat extreme frequency over Northern Hemisphere land might be reduced by 40% over a century following reaching net-zero (Cassidy et al. 2023):

Land-based hot temperature extreme frequencies are projected to decrease as much as 40% within the century after net zero CO2 emissions, with largest reductions over land in the Northern Hemisphere. Reduction in land-based local monthly temperature extreme frequency within only a century after CO2 emissions cessation is a promising result, as heat extremes can have severe infrastructural impacts and can lead to loss of life.

We can take these idealized numbers and perform a quantitative thought experiment. The figure below shows the same data that I shared in Part 1 — European heat wave days have increased since 1950 — as the dark blue bars.

I assume that the increasing trend of 2.6 more heat wave days per year continues until 2060, when I assume that the world hits net zero. At that point I assume a reduction in the number of heat wave days of 40% over the following 100 years (to 2160), following Cassidy et al. 2023. For the entire projected time series of 2019 to 2160 I have included random variability of the same magnitude as in the data from 1950 to 2018.

The thought experiment illustrates clearly why reducing emissions — even to net zero by 2060 — is not a “solution” to heat waves. In this idealized example, mitigation does have benefits in the form of reducing heat wave days over the century after reaching net zero. Yet, at the same time there would be no detectable reduction across the lifetimes of anyone alive today. Even in 2160 the number of European heat wave days (~20 per year) would be similar to that of the 2020s (~20 per year). And in reality, net-zero by 2060 is a questionable assumption.

This gets to a core challenge of framing mitigation as weather control — the time scale of promised benefits, even in the best of cases, is far, far longer than the time frame for the costs of action. Aligning costs and benefits closely in time makes good political sense, in most any policy context — but it also means leaving behind promises of better weather through climate policy. I do wonder if climate advocates are too vested in such promises to ever leave them behind.

This thought experiment does not include the possibility that unmasked aerosols might act to increase heat wave days following net zero or that internal variability is larger than suggested by the period covered by the original time series. Other literature suggests that the climate system response to achieving net-zero is more complex than just a reversal of changes leading up to net zero (see, e.g., Smith et al. 2025 and King et al. 2025).

Such possibilities mean that the thought experiment here might be unrealistic in assuming that carbon dioxide emissions reduction is actually a “control knob” for heat wave days. But even under this scenario, reducing carbon dioxide emissions does not fundamentally change the nature of the heat wave problem facing Europe.

So what would accelerated decarbonization actually do?

Getting to net zero still makes sense even though it is not a solution to heat waves or other extreme weather events: Achieving net zero would halt the growth of the greenhouse gas contribution to the Earth’s energy imbalance. On that basis alone, seeking to accelerate decarbonization makes good sense — something I’ve argued for decades.

Of course, emissions reductions necessarily will take place in the broader context of costs and benefits, a topic that goes far beyond the scope of today’s post (but have a look at The Climate Fix for much more!)

The fact that carbon dioxide is not an extreme weather control knob does not weaken the case for reducing emissions. But it does mean that many climate advocates are justifying their calls for net zero based on a false promise — that emissions reductions will improve the weather.

As I argued in Part 1, there are far better justifications for accelerating decarbonization than trying to convince people that climate change causes bad weather. Selling emissions cuts as a solution to bad weather not only misstates the science; it makes a policy promise that can’t be kept.

Comments welcomed!

Leave a comment

Subscribe now

Back in April, THB broke the news to the world that the only three official reference scenarios used for the past 15 years in much of climate research, assessment and policy had been retired. The CMIP71 conclusion that RCP8.5, SSP5-8.5, and SSP3-7.0 were “implausible” was huge news.

Today, I share more big news about how official climate scenarios continue to evolve.

CMIP7 has just published the final versions of its new scenarios as they have emerged from integrated assessment models (IAMs), updating the placeholders that accompanied their April announcement. The final versions — once again — dramatically reduces the cumulative carbon dioxide emissions from fossil fuels from the April placeholders.

The table below shows just how large these changes are in gigatons (Gt) of carbon dioxide emitted from the burning of fossil fuels to 2100 — both in terms of the highest cumulative emissions across the scenario set (in the first bullet below) and in terms of the scenarios designated as reference or current policy (second bullet below):

• SSP5-8.5, 7,380—> CMIP7 HIGH 1.0, 4,349—> CMIP7 HIGH FINAL, 3,438

• SSP5-8.5, 7,380—> CMIP7 MEDIUM 1.0, 3,098—> CMIP7 MEDIUM FINAL, 2,528

The bar chart below shows just how large the reduction of projected emissions has been.

You can see in the table above that each integrated assessment model associated with each new scenario has been identified, however, the ability to share the details of each scenario awaits the publication of those IAM runs and associated papers. ScenarioMIP explains that this embargo holds until September 1:

The data can be used only for scientific purposes aimed at the production of peer-reviewed publications. The posting of the scenario data and/or analyses based on the data on social media or other dissemination (in the public or the press) is thus strictly prohibited.

I will of course respect this embargo and this post — and any other discussions of the new scenarios here at THB — will rely only on the publicly available data archive accompanying Van Vuuren et al. 2026.

Based on the publicly available data accompanying the CMIP7 paper describing the new scenarios, I can provide an analysis of the new scenarios based on Pielke et al. 2022.

The figure below shows the annual trajectory of emissions for the retired reference scenario SSP5-8.5, along with several of the new CMIP7 scenarios. CMIP7 HIGH is intended not as a reference, but as an exploratory scenario assuming increased greenhouse gas emissions, “resulting from deep political, technological, and structural deviation from current trends.”

The figure below — based on Pielke et al. 2022 — shows how the new CMIP7 scenarios compare to the AR5/SSP scenario database in terms of cumulative carbon dioxide emissions and implied 2100 global average temperature. Note that these temperature estimates are from the emulator used by Van Vuuren et al. 2026, and have yet to be run in earth system models, so are preliminary.2

For the MEDIUM scenario, CMIP7 explains:

The Medium emission scenario is a benchmark that shows the consequences of the current policy situation (as of 2025) and trends continuing over the century. In the scenario, it is assumed that policy effort is continued at the current level. As for other scenarios, also this scenario should not be considered as a “most likely” scenario.

For the MEDIUM-to-LOW scenario CMIP7 explains:

The Medium to Low emission scenario describes a future in which strengthened mitigation efforts are delayed, falling short of the levels required to meet the objectives of the Paris Agreement. . . . The underlying logic of the scenario is that political constraints limit rapid near-term action; however, these lessen over time as the magnitude of observed climate impacts increases and mitigation costs decline with technological progress.

While MEDIUM is the only scenario framed as “current policy” in the CMIP7 set — at 3C in 2100 — it sits above other “current policy projections,” which you can see in the table below. In addition, the MEDIUM-to-LOW scenario at 2.3C in 2100 is just about at the lower end of the projections that assume that national pledges are met.

Thus, the range from MEDIUM-to-LOW to MEDIUM — with medians of 2.3C to 3.0C in 2100 — can serve usefully as the equivalent of a baseline or reference range. The HIGH scenario, at ~3.4C in 2100, is explicitly designed as an exploratory scenario to illustrate a world in which policies are adopted that increase emissions. CMIP7 explains:

Clearly, this scenario is not a “business-as-usual” scenario nor the no-policy reference scenario for the other scenarios.

This point needs to be emphasized. HIGH is not a baseline or reference scenario and should not be treated as such. I’m sure this warning will be ignored by some.

What is the significance of the new emissions scenarios? Some thoughts:

• The massive scale-back of projected future carbon dioxide emissions means, all else equal, less future changes in climate;

• As a consequence, there will surely be enormous political pressures placed on the climate science community to create projections with greater changes in climate, based on factors other than carbon dioxide emissions;

• If so, then we should expect significant challenges to scientific understandings that have long been stable, such as related to the carbon cycle, climate sensitivity, aerosols, and other greenhouse gases;

• Any such evolution of understandings may well be scientifically justifiable and legitimate, but a political overlay will create challenges;

• Meantime, tens of thousands of studies reliant on the implausible scenarios need to be reconsidered carefully, as the new reference range undershoots the 4.5 scenarios that were often presented — and misrepresented — as climate policy success;

• The range from MEDIUM-to-LOW to MEDIUM spans ~4.1–5.7 W/m², so the current and pledged policies range now sits below what RCP4.5 was presented as achieving with aggressive policy;

• None of this affects the huge magnitude of the challenge of accelerating the decarbonization of the global economy, which is not affected by scenarios;

• Climate research, assessment, advocacy, policy, and politics all need to catch up to the altered view of the climate future.

However things evolve, climate science and policy are now in a new era.

Comments and discussion welcomed!

Leave a comment

Paid subscribers can find sources and data for the figures and tables in the Excel file below.

Read more

Subscribe now

The image at the top of this post is real — not AI. It shows the Earth as a thermostat with the message: “You Control Climate Change.” The 2006 PR campaign was developed by the European Commission — in the words of European Commission President José Manuel Barroso — to make:

”. . . clear to which extent we all are responsible for climate change and what individuals can and need to do to limit this threat.”

In the twenty years since that campaign, the way that we discuss climate change has changed significantly. Back then “climate change” referred to an outcome, specifically, a change in the statistics of weather variables over time, consistent with the definition of the Intergovernmental Panel on Climate Change (IPCC).

Increasingly, “climate change” has morphed into an all-purpose causal explanation for extreme weather events.

For instance, over the weekend the Financial Times — in an otherwise very good article on increasing heat waves in Europe and the UK — explained that:

“[C]limate change drives more extreme weather”

Saying that climate change drives extreme weather is the equivalent of going to your doctor and having them tell you that your body’s abnormal high temperature is being driven by a fever.

Well, yes . . . but most of us would want to know what is causing the fever and then, based on an accurate understanding of causality, what might be done to address the illness causing the fever.

Nowadays, discussions of climate and climate policy are often simplified into a linear, causal chain that goes directly from fossil fuels to extreme weather.

As an explanation for the causality of undesirable weather, this causal chain might suffice for conversations with your neighbors —

— but if taken too seriously, the framework implies that we can modulate weather by reducing consumption of fossil fuels. When that framing pits adaptation against mitigation, for example, it can lead to poor policies.

This logic has been on display in recent weeks in the context of extreme heat across Europe, such as in reporting by the BBC:

Far from attacking the root causes of global-warming, [French climate] activists said, recourse to la clim [air conditioning] was merely attenuating the effects of global-warming.

And by making those effects more bearable, it distracted from the essential fight against the causes.

Not only that, but air-conditioning is often criticised by environmentalists for aggravating climate change.

This is because it requires electricity to run - and though most of France’s electricity comes from nuclear power, elsewhere it means more fossil-fuels being burned.

It is not just the media.

The simple, linear explanation for weather events and their impacts has found its way into the peer-reviewed literature. The image below illustrates what has been called “end-to-end” attribution that purports to be able to quantify how “climate change” causes both weather events and their impacts.

Does climate change cause heat waves or make them more common or more intense?

That answer then leads to an important question from the standpoint of policy: If “climate change” is not a cause of European heat waves or their intensification, then what is?

Today’s post, Part 1 of 2, takes this question on, and the answer is far more complicated and interesting than the simple story illustrated above.

The climate has changed

The increasing number of heat wave days and their intensification across Europe is well documented. The figure below shows that the number of heat wave days across Europe has risen steadily since 1950, at a rate of about 2.6 days per decade.1

Europe’s heat waves have also become more intense. The magnitude of each year's worst heat wave, from the Oliveira et al. (2022) dataset, has increased since the 1970s, as shown below.

Why have European heat waves become more frequent and intense?

The increase in heat wave frequency and intensity represents a change in climate, but to assert that “climate change” is the cause of the increase is, at best, completely circular. Let’s dig into the scientific literature for a more satisfactory and accurate set of explanations for the observed trends.

Let’s start with the IPCC. Hot extremes at the global level, including heat waves, are the extreme weather phenomena for which the IPCC has the most confidence in detected changes and their attribution to human causes.

However, the IPCC provides more nuance when discussing heat waves at the regional level. The IPCC Sixth Assessment states that trends in extreme heat observed in a particular region are influenced — sometimes modified, sometimes amplified — by factors such as soil moisture, land use, aerosols, and internal modes of climate variability:

The effect of enhanced greenhouse gas concentrations on extreme temperatures is moderated or amplified at the regional scale by regional processes such as soil moisture or snow/ice-albedo feedbacks, by regional forcing from land use and land-cover changes, or aerosol concentrations, and decadal and multi-decadal natural variability.

The IPCC is careful to add that its confidence in these regional and dynamical factors is lower than for overall global trends.

I consider the attribution of causes of increased European heat wave frequency and intensity in terms of three bins2:

• Bin 1: the direct effect of increasing concentrations of greenhouse gases, particularly carbon dioxide, in the atmosphere due to human activity, notably the burning of fossil fuels;

• Bin 2: the indirect effects of increasing concentrations of greenhouse gases;

• Bin 3: factors other than greenhouse gases.

The literature reflects the greatest confidence in the role of Bin 1 and there are areas of uncertainty, conflicting claims, and dueling hypotheses when it comes to Bins 2 and 3.

Let’s take a closer look.

What is a heat wave?

A heat wave is a weather event — a stalled weather pattern, or more technically, a high-pressure “block” or ridge that settles over a particular region. Under it, skies stay clear, air sinks and compresses (which warms it), and the sun heats the ground. This is basic meteorology, and it’s the proximate cause of every major heat wave in Europe or elsewhere. Different regional atmospheric blocking patterns produce different regional heating, which is why no two heat waves are quite alike (Sousa et al. 2018; Liu et al. 2020).

So the first thing to understand is that heat waves are made by circulation — the movement or stalling of air. Greenhouse gases do not create a heat wave. A weather pattern does.

Bin 1: The effects of increasing concentrations of greenhouse gases

Every surface — land, ocean, air — constantly radiates heat outward as infrared. Carbon dioxide, and other GHGs, absorb part of that outgoing infrared and re-emit it in all directions, including back down, which is the essence of the “greenhouse effect.” More downward infrared radiation means that the Earth’s surface and the lower atmosphere warm — and this dynamic has been directly measured over Europe. This heating is what is meant by the phrase “global warming.”

Since 1971 the Earth has accumulated roughly 380 zettajoules of excess heat, and the overwhelming majority — about 89% — has gone into the ocean, about 6% into the land, about 4% has melted ice, and only ~1% has gone into the atmosphere itself (von Schuckmann et al. 2023). So the single biggest reservoir of the energy that greenhouse gases have trapped is the global oceans.

Of the ocean’s share, about a tenth has gone below 2,000 meters, into the deep ocean, where it is sequestered on timescales of centuries and does little to influence weather. That deep heat matters for sea level.

That extra energy in the climate system heats the land, the near-surface air, and — critically — the oceans and the source regions upwind of Europe. So when a stalled high-pressure pattern occurs, it does so in a climate system that has more energy than it did decades ago. That means that normal, meteorological extreme heat events will — all else equal — be hotter than they used to be.

The first order greenhouse effect changes how much energy is in the system. It does not directly alter atmospheric circulation.

Bins 2 and 3: What else is going on?

In much of western and central Europe, summer temperatures have risen about three times faster than the global average since 1980 — faster than expected from the direct effect of the Earth’s changing energy balance. Western European heat extremes have increased at roughly 3.4°C for every 1°C of global warming, the fastest rate of any comparable region on Earth — outside the poles.

A recent decomposition attributes a substantial share of this excess warming — the part above the global rate — to changes in atmospheric circulation rather than to the direct energy-budget effect alone.

Philipona et al. (2009) identified the disparity early on:

Mainland Europe’s temperature rise of about 1°C since the 1980s is considerably larger than expected from anthropogenic greenhouse warming.

Explanations in the literature emphasize three other factors contributing to the increasing frequency and intensity of European heat waves:

Dry soils. When soil has moisture in it, the sun’s energy goes partly into evaporating that water, which cools the surface. When the soil dries out, that cooling stops, and the sun’s energy goes straight into heating the air instead. This is a large effect. Modeling studies that eliminate the soil-moisture feedback conclude that more than 70% of the extra hot-day warming in Central Europe disappears in the model projections without this feedback. Drought and heat feed each other: dry ground makes the air hotter, and hot air dries the ground further (Fischer et al. 2007; Dirmeyer et al. 2021).

Cleaner air. Starting around 1980, Europe’s air pollution was considerably reduced (Yang et al. 2020). The tiny particles in that pollution — from coal smoke and industry — used to reflect sunlight back to space, partly via clouds that were brighter and more reflective. As the pollution fell, more sunlight reached the ground (Philipona et al. 2009), with a direct warming effect on the surface, separate from greenhouse gases. Schumacher et al. 2024 estimate the resulting warming effect at about half a degree Celsius, a meaningful part of the observed warming in Europe.

Warmer seas and land upwind. The air that flows into Europe during a heat wave comes from somewhere — often over the Mediterranean, the Atlantic, or the hotter landscapes of Iberia and North Africa. All of those are warmer than they used to be, in part because the same greenhouse and aerosol changes have been heating them for decades. So the same wind that would have brought hot air decades ago now brings hotter air (Vautard et al. 2020).

Complexities. Here is where things get complicated — Drier soils and warmer seas and land are influenced by greenhouse gas warming. So there are second order processes at work.

At the same time, cleaner air is not directly influenced by greenhouse gases. There are also hypotheses that suggest that circulation changes that lead to increased blocking, and thus more heat waves, are a consequence of the drop in aerosols. So here as well there are second order processes in play.

Several studies now attribute the observed weakening of the summer jet stream — which makes weather patterns stall more often and longer — primarily to the drop in aerosols (Dong et al. 2022; Chemke & Coumou 2024; Kang et al. 2024). A 2026 study goes further, arguing that the accelerated western European warming is, to a large extent, a response to changes in aerosol forcing. If accurate, the cleanup of air pollution is warming Europe in two ways: directly, by letting in more sun, and indirectly, by making the stalling patterns that cause heat waves more common.

Both increasing greenhouse gases and reducted pollution are human effects on climate, but they operate very differently. Lumping them together as “climate change” subtracts from our understanding.

Modes of variability. An even further complication is that not all of a detected trend might indicate a change in climate — it could reflect the consequences of different modes of internal climate variability. European summer heat is affected by patterns in the ocean and atmosphere that redistribute heat without requiring any change in external forcing.3

That said, the IPCC AR6 is explicit that the heat wave trend in Europe cannot be explained without human influences:

“[D]espite strong internal variability, observed trends in European mean and extreme temperatures cannot be explained without accounting for anthropogenic factors. Before the 1980s, warming by greenhouse gases was partly offset by anthropogenic aerosol emissions. Reduced aerosol influence in the recent decades has led to an observable positive trend in shortwave radiation."

The IPCC AR6 further notes that the greenhouse effect on temperature extremes is moderated or amplified regionally by, among other things, “decadal and multi-decadal natural variability.”

How much of the recent trend in European heat extremes is due to internal variability of the climate system rather than a forced change is not fully understood, and the IPCC AR6 is also explicit on this point (emphasis added):

“Anthropogenic forcing (e.g., increases in greenhouse gas concentrations) directly affects thermodynamic variables, including overall increases in high temperatures and atmospheric evaporative demand, and regional changes in atmospheric moisture, which intensify heatwaves, droughts and heavy precipitation events when they occur (high confidence). Dynamic processes are often indirect responses to thermodynamic changes, are strongly affected by internal climate variability, and are also less well understood. As such, there is low confidence in how dynamic changes affect the location and magnitude of extreme events in a warming climate.”

The component of the trends in European heat waves resulting from changes to atmospheric circulation — that is, dynamic processes — whether forced by greenhouse gases, driven by the reduction in aerosols, a consequence of internal variability, or the interactions among all these, is where attribution is uncertain.

Urban heat islands. Heat waves are hotter in cities: Pavement and buildings warm during the day and release that heat slowly at night, less greenery means less cooling from evaporation, and narrow streets trap heat. The result is the urban heat island. During the 2003 heat wave, a modelling study of the West Midlands found that roughly half of the heat-related deaths were attributable to the urban heat island rather than to the regional heat, and across 85 European cities the urban effect raises the risk of death during extreme heat by about 45%. Cities dominate the human toll of European heat waves, because that is where the people are: about three-quarters of Europeans live in urban areas.

Complexity matters

I am unaware of any robust quantification of how much of Europe’s heat wave trend comes from greenhouse gases versus aerosols versus soil moisture versus circulation changes versus modes of variability versus cities — much less the complex inter-relationships among these various factors. What I do know is that the simple story that climate change causes heat waves is unsatisfactory.

The figure below — which better represents the real complexity — illustrates a far more nuanced set of factors that contribute to the documented trends in European heat wave frequency and intensity.

The politics of the climate issue have favored a simple story that collapses all of the complexity in the figure above into the phrase “climate change” which is typically and incorrectly oversimplified into a synonym for carbon dioxide emissions from fossil fuels. Political advocacy runs on simple descriptions of cause and effect. However, effective policy making depends on accurate understanding of causes and effects.

Let me be perfectly clear — Nothing in this post is an argument against cutting emissions. In fact, as I've often argued, cutting carbon dioxide emissions is a matter of long-run risk reduction, and on that basis alone, makes good sense. Reducing emissions is not about modulating the weather — and certainly is not a thermostat. There are far better justifications for accelerating decarbonization than trying to convince people that climate change causes bad weather.

Today’s post sets the stage for Part 2, where I explore what effects efforts to accelerate decarbonization of the global economy might have and on what time scales for future European heat waves.

Comments and discussion welcomed!

Leave a comment

Subscribe now

The retirement of implausible baseline climate scenarios is now starting to have real-world consequences as decision makers realize they’ve been making decisions that have had unnecessary real-world, negative impacts on their constituents and clients.

In a letter shared 1 July 2026 with local communities across New Zealand (reproduced below), the country’s Minister of Climate Change, Simon Watts, has asked public officials to reconsider using RCP8.5 and other retired scenarios in their planning decisions.

Councils, the letter states, have used these implausible emissions scenarios in a way that,

“results in unnecessary costs being imposed on ratepayers, businesses, and communities.”

The letter also instructs councils that they “should not automatically” continue to treat these scenarios

“as the central or most likely future when making decisions that affect infrastructure investment, land use, property rights, or rates.”

A first hand look at what the impact of RCP8.5 looks like in New Zealand can be seen in Managed Out, documentary on “managed retreat” from the coasts filmed before the minister's letter.

On the Kapiti Coast, residents describe hazard notations stamped onto their property files that are based on a consultant's report built on futures projected under RCP8.5. One local resident complained:

"They're using RCP 8.5 which is the most extreme outcome. House prices are on the way down, insurance prices are sky-rocketing, because they have put notations on all of our LIM reports [Land Information Memorandum, which assesses hazards] which uses the extreme modelling."

A former Wellington city councilor in the film argues that the problem lies in:

"the use of the IPCC 8.5 model that is supposed to be used only for stress-testing and not for policy-making."

These residents are experiencing the unintended consequences of the use of a scenario that scientists built originally for exploratory earth system modeling to inform global climate negotiations turning up in the price of their homes.

Using RCP8.5 to predict coastal hazards in New Zealand — or anywhere else — and thus serve as the basis for home pricing and insurance is simply a misapplication of science (that may have had valid uses in research in other contexts).

How things went cattywampus

A baseline scenario — also called business as usual or a reference scenario — is according to the IPCC, a “scenario used as starting or reference point for a comparison between two or more scenarios.”

Climate policy analyses, including cost-benefit studies, typically start with a baseline scenario as a representation of where-we-are-headed and then compare that to other scenarios in which policy is introduced within the scenario. Such scenarios are typically called policy (or mitigation) scenarios.

Such a comparison, in principle, allows decision makers to evaluate the expected effects of implementing a policy versus not implementing that policy.

Based on the IPCC’s official AR6 baseline scenarios — RCP8.5, SSP5-8.5, SSP3-7.0 — New Zealand councils had drawn hazard lines across property titles, adjusted infrastructure plans, and shifted insurance and lending exposure, with real economic consequences for New Zealanders.

Thanks to the correct decision by the CMIP7 scenario committee, we now know that all of these decisions and consequences have been based on a scenario now officially recognized as implausible.

In a recent interview with Alistair Harding in New Zealand, Detlef van Vuuren — the Dutch scientist who helped to lead the development of the 2011 RCP scenarios (including RCP8.5) as well as the new 2026 CMIP7 scenarios that officially retired RCP8.5 — was asked if it was ever a fair description of RCP8.5 and the like as business-as-usual (BAU) or a no-mitigation future.

His answer:

“No, that was always a wrong way to describe it.”

And yet, as has been carefully documented here at THB, the set of three baseline scenarios has — for fifteen years — been characterized as a BAU or baseline or reference scenarios throughout climate research and policy. That designation trickled down to land use decisions in coastal New Zealand.

Van Vuuren now says that we all got that wrong from the start.

What happened?

Confusion from the beginning

In the original 2011 paper announcing the RCPs to the world, Van Vuuren et al. 2011 (VV11) describes the four RCP scenarios in a way that has been lost to history:

• one very low mitigation scenario (RCP2.6),

• and three that its Table 4 classifies as baselines:

  • RCP4.5 a "very low baseline,"

  • RCP6.0 a "medium baseline,"

  • and RCP8.5 a "very high baseline emission scenario."

You read that right — originally, there were supposed to be three baseline scenarios, spanning a very wide range, not just RCP8.5.

As baselines, RCP4.5, 6.0, and 8.5 were meant to represent the range of baselines found across the literature on scenarios. RCP8.5 was at the upper extreme of this range and RCP4.5 was at the lower extreme — both were equally valid as baselines.

At the same time, RCP4.5 and RCP6.0 were simultaneously also supposed to represent policy/mitigation scenarios, which are not baselines. They were thus intended to fulfill two conflicting roles in representing the literature, as shown in the table below, from VV11.

But when it came time to select the specific scenarios from the IAM literature to characterize the RCPs, the scenarios actually chosen to fill the 4.5 and 6.0 slots were policy scenarios, not baseline scenarios. Thus the RCP4.5 and 6.0 scenarios had their own baselines, but they were well above 4.5 and 6.0 levels of radiative forcing. Only RCP8.5 was from an IAM baseline — a world with no climate policy at all.

You can see this in the figure below from Pielke and Ritchie 2021. Confusion is the only predictable result and that is exactly what happened.

For anyone who took the RCP pathways as they were constructed and presented, rather than as the broader range they were meant to represent as described in VV11 — the set read as a single baseline and three policy scenarios.

The only unambiguous no-policy reference in the framework was RCP8.5.

The scenario community was at the time well aware of the confusion this might cause. For instance, the image below is a slide from a presentation given by Richard Moss, another leader of the RCP development process, warning that RCP8.5 was not a reference scenario for the three other RCPs — another detail lost to history.

Despite the warning, tens of thousands of papers have been published that pair RCP8.5 as a reference scenario with the other RCPs as policy scenarios (typically RCP4.5 and RCP2.6). In particular RCP4.5 was often treated as representative of the consequences of aggressive climate policy,1 when under the RCP framework, it should have been treated just like RCP8.5 — as a baseline.

Without correction from people who knew better — which includes the many people who have participated in scenario development the entire time — the misinterpretation of RCP8.5 as the only baseline scenario in the set took off and came to dominate how the RCPs (and SSPs) have been used for the past 15 years in both science and policy.

As one unnamed Dutch scientist told De Volkskrant in 2020:

“You simply do not realize that the RCPs can start a life of their own.”

Scenarios were flawed — which is why a wide range matters

It is now widely understood that the RCP and SSP scenarios were flawed from the start, based on reliance on a narrow set of assumptions — the details of which I won’t revisit today. However, the story here is much more complicated than just that.

The figure below shows how the official scenario baseline range of CMIP and IPCC changed from VV11 (original RCP) to CMIP5 — to IPCC AR5 (2014) — to CMIP6 and AR6 (2016-2021) — to CMIP7 (2026).

Two things to highlight:

• Over more than a decade, from the development of the RCPs through IPCC AR6, the baseline scenario range became both more extreme and narrower;

• The original VV11 range largely (but not entirely) spans the new CMIP7 range.2 Were the CMIP7 update being done off of VV11, the changes would fairly be considered a slight readjustment, not the massive change that you can see from AR6 to CMIP7.

The reason that the VV11 range largely lines up with CMIP7 is not because the individual scenarios held up.

The ranges largely overlap because VV11 included a very wide range of baselines, reflecting fundamental and irreducible uncertainties and ignorance — two characteristics of our understandings of the real world that scenario planning is, in principle, designed to help us grapple with.

A wide range of assumptions is what would have allowed the 2011 framework stay useful even though RCP8.5 later collapsed.

The lesson here is that the value of a scenario set lies in the breadth of assumptions it spans, not in the correctness of any single pathway. Incorporating a range of assumptions into scenarios is the hedge against the flaws of the details of any individual scenario, because the builders cannot know in advance which of their assumptions will hold up and which will not.

Even though the implausible baselines have been retired — appropriately — it seems that the importance of a diversity of assumptions across scenarios has not been appreciated by the CMIP7 scenario creators.

Correcting course is just starting

Returning to New Zealand and the RCR Van Vuuren interview — he expresses sympathy for those in the country whose fortunes have been compromised because of the unwitting use of RCP8.5 as a baseline scenario for current land use planning and regulation (emphasis added):

If I was owning a house somewhere and there is a risk in 2150, I think it [RCP8.5] is useful information. But I would expect that my mortgage being out 30 years or so, that it would be completely unfair that the value of my property would be so much determined by something that is uncertain and way out.

Van Vuuren suggests that, from his perspective, the RCP scenarios were never really intended to be used in local decision making:

The way that I normally use these scenarios is in climate negotiations. It is quite interesting that even these rather abstract scenarios have such consequences as well in daily decisions . . .

Interesting, indeed.

The mistakes made in New Zealand in using RCP8.5 as a scenario depicting an expected future, a forecast even, is understandable as a cascade of confusion and conflicting intent for the role of scenarios in climate research. However that cascade has had material consequences for real people in real places, reaching far beyond New Zealand.

With the RCPs, the scenario community pulled four scenarios from a literature of many thousands. That literature spanned an enormous range of possible futures, and appropriately so, as the future is a fuzzy place. The small group of four scenarios was asked to do double duty — to represent futures that we might be headed to without any consideration of climate policy, and futures that required considerable climate policy interventions. The intent was to “jump start” earth system modeling, not to serve as the basis for decisions all around the world spanning the local to the global scale.

Along the way, as mistakes were made, no one spoke up or pointed out the unfolding errors of scenario design and application.

We now know — and have known since 2017 thanks to Ritchie and Dowlatabadi — that the single anointed baseline was fatally flawed. Out in the real world (as opposed to model land) RCP8.5 was variously interpreted as a projection of our most likely future or even as a forecast and thus wired into consequential decisions.

Had the climate community recommended that scenario planning be based on a wide range of baselines, from 4.5 to 8.5, some may still have chosen the most extreme of the set but others would have had the chance to consider other baselines. By considering that wide range, the recalibration of that range would have likely been less consequential than the huge change in expectations that many are now experiencing.

Comments welcomed!

Leave a comment

Subscribe now

The 2026 Energy Institute Statistical Review of World Energy — successor to the old BP Statistical Review, and an authoritative accounting of global energy data — has just been released with data through 2025. As I have done here at THB for years, in today’s post I update THB analyses of where the “energy transition” actually stands.

Regular readers will recognize the approach, which is based on my 2010 book The Climate Fix. At the bottom of this post you can find a methods appendix with details on the calculations, and below that paid THB subscribers will find an Excel spreadsheet with the sources and data from each graph and table. As always, this is work in progress — Caveat lector!

There is lots to discuss, so let’s get to it . . .

Carbon-free energy is growing — slowly

Let’s start with the good news for those promoting a transition away from fossil fuels. The carbon-free share of global energy consumption — nuclear, hydro, and other carbon-free generation technologies — reached its highest level ever.

That share now sits near 14% and has climbed steadily over the past decade, but slowly with respect to aggressive targets of climate policies: net-zero means going from 14% to 100%. At the pace of the past ten years, that would take centuries.

Total energy demand continues to increase, and fossil fuels have continued to meet the majority of that growth — with global fossil fuel consumption still growing as shown below.

Global fossil fuel consumption reached a new high in 2025 — about 518 exajoules. Net-zero is further away than ever. It is essential to understand that retiring fossil fuels reduces carbon dioxide emissions; simply deploying carbon-free energy does not.

The math of a 2050 net-zero target

What would reverse that fossil fuel line and move it towards zero?

Consider the arithmetic of a net-zero-by-2050 target, measured from 2025

Reaching zero by 2050 requires retiring ~21 exajoules of fossil energy every year, starting now. That annual reduction exceeds the total energy consumption of most countries on Earth. And every year the line fails to bend down, the required rate in the remaining years grows larger, just as a matter of math.

Replacing 21 EJ of fossil energy per year, and retiring an equal amount of fossil supply alongside it, means building the equivalent of about one 1.75-gigawatt nuclear plant every day from now until 2050 — roughly 420 plants a year. Measured in wind turbines instead, at 3 megawatts and a 0.30 capacity factor, that comes to about 2,000 turbines a day, every day, for 25 years.

I’ve published this analysis for years, and net-zero isn’t getting any closer — Net-zero 2050 is infeasible, by any practical standard, but that does not mean that the aspiration of net-zero is impossible.

Consider that if the net-zero target is 2075, then the implied pace of retirement and deployment is cut in half: about one nuclear plant every two days, or 1,000 wind turbines a day. As I’ve previously noted, if the world can maintain its historical pace of decarbonization (a big ask!) then the world would hit net-zero around the 2080s.

The pie chart above shows the change in total energy consumption from 2024 to 2025. You can see that carbon-free energy made up almost half the year-on-year additions. Fossil fuels still won out.

Below, the left panel extends the comparison of carbon-free to FF year-to-year change in global energy consumption back to 2010. Fossil consumption rose in every year except the pandemic year of 2020, and was outpaced by carbon-free in only one other year.

The right panel shows what is implied by a net-zero 2050 target — on average, every year to 2050: fossil energy decreases by about 21 EJ per year while carbon-free energy increases by nearly 28 EJ per year to cover both the retirements and continued demand growth

Reasonable people can argue over the exact required amounts of annual carbon-free deployment or fossil fuel retirement — it is huge regardless of the exact methods or assumptions.

Where did fossil fuel consumption increase and decrease 2024-2025?

From 2024 to 2025, the United States and China led all countries in added fossil fuel consumption, at +1.6 and +1.1 exajoules respectively — together more than the next dozen countries combined.

After the U.S. and China is a set of fast-growing developing economies: Indonesia, Türkiye, Bangladesh, and others added smaller but steady increments of fossil fuels as their energy demand continues to climb.

The countries that cut the most fossil fuel — Japan, Russia, South Korea, Colombia — trimmed only a few tenths of an exajoule each, an order of magnitude smaller than the increases at the top of the other chart.

The additions dwarf the subtractions: the largest reduction, Japan’s, would not offset even a fifth of the U.S. increase. As long as the countries adding fossil energy add far more than the countries cutting it subtract, worldwide fossil fuel consumption keeps climbing.

Decarbonization: so far, a long, straight line

The figure below shows the carbon intensity of the global economy 1992-2025 — CO₂ per unit of GDP.

The global economy has decarbonized steadily since well before climate policy existed — back to at least the 1960s.

Two things follow.

• First, global climate policy does not drive decarbonization as the decrease in carbon intensity of the global economy long pre-dates the climate movement (which started as coordinated global policy in 1992 with the Rio Earth Summit); economies growing wealthier and using more energy more productively have driven that trend.

• Second, and less comfortably: since the climate-policy era began in 1992 the background rate shows no acceleration. Hitting deep-decarbonization targets requires this straight line to bend down. It has not done so.

Some economies cut their carbon intensity by about two-thirds since 1990 — China, the United Kingdom, Germany. Others moved the wrong way and grew more carbon-intensive, led by Iran. These differences track different starting points, energy mixes, and stages of development — not the presence or absence of climate ambition.

What actually drives decarbonization?

The figure below shows a simple Kaya decomposition of global decarbonization: Carbon intensity of GDP equals the product of two factors: how much energy the economy uses per dollar (energy intensity), and how much CO₂ that energy emits (carbon intensity of energy).

Decomposing the decline shows how much each factor has contributed to global decarbonization.

The green line — the economy becoming less energy intensive per dollar of output — accounts for almost the entire decline in CO₂ per unit of GDP. The orange line — the carbon intensity of the energy that is consumed — has barely moved in 35 years, despite the impressive growth in carbon-free energy.

How much faster can decarbonization go? I will take that up in future posts.

The decarbonization rankings tables

The first table ranks the G20 economies and the world by how much their carbon intensity of GDP — carbon dioxide emitted per unit of economic output — changed from 2024 to 2025.

India and Argentina led, each cutting emissions per dollar of output by nearly 6%, followed by China and Saudi Arabia. Most major economies decarbonized, which is common: economies tend to achieve more output per unit of energy as they grow and modernize. However, others, including the United States, saw emissions growth outpace economic growth.

The world as a whole decarbonized by about 2.3% — a fairly typical rate. Compare that to deep-decarbonization targets for 2050 which imply a sustained global rate of decarbonization closer to 8-9% per year or more.

A single year, however, tells us little, since annual figures vary with weather, fuel prices, and economic cycles. The second table takes the longer view, ranking the same economies by their change in carbon intensity across the full Paris-era decade, from 2015 to 2025.

Here the leaders are the United Kingdom, Germany, and China, each having cut carbon intensity by a third or more over ten years. Only Indonesia moved the wrong way over the decade, its emissions outrunning even its rapid economic expansion. The world decarbonized by about 19% over ten years, or a little over 2% annually .

The third table asks: did the 2015 Paris Agreement bend the curve?

It compares each economy’s annual rate of decarbonization in the decade before Paris (2005–2015) against the decade after (2015–2025), and shows the difference.

A handful of economies accelerated sharply, though the largest movers — Saudi Arabia and Brazil — did so for reasons based on their own energy and economics circumstances, rather than efforts to comply with the Paris Agreement.

Global decarbonization changed little: from ~1.9% per year before Paris to ~2.1% after. China, which dominates global emissions, actually decarbonized more slowly after Paris than before, offsetting gains elsewhere.

A change in the decarbonization rate is not sufficient to prove policy causation on its own, but the lack of meaningful change at the global level is hard to square with the frequently-expressed claims that climate policy has already achieved notable success (such as we’ve been hearing related to the retirement of RCP8.5, but I digress).

Bottom line

The 2026 update is fully consistent with the pattern I have documented here year after year. Carbon-free energy is growing. Decarbonization continues its long, slow, pre-policy downward trend. Global fossil fuel consumption continues to increase. The carbon intensity of energy has barely budged, drifting downward ever so slowly. The gap between the required rate of change to hit aggressive decarbonization targets and the observed rate in the real world widens every year.

If we are to achieve deep decarbonization this century, the lesson to take is that we should start thinking about some different options for how to get there. What we have been doing isn’t working.

Comments, critique, debate, discussion — all welcome at THB!

Leave a comment

Subscribe now

Methods Appendix

Energy and emissions data. All energy consumption and carbon dioxide emissions figures come from the 2026 Energy Institute Statistical Review of World Energy, which reports data through 2025. Energy uses exajoules (EJ); emissions use million tonnes of CO₂ from energy. Non-fossil energy (nuclear, hydro, and other renewables) follows the Physical Energy Content method the Statistical Review adopted in its 2025 edition, which counts non-combustible generation at its actual energy content rather than scaling it up to a fossil-fuel input equivalent. Earlier editions used the substitution (fossil-fuel-equivalent) method; the change lowered the reported carbon-free share by several points, as discussed in the post. This basis moves the level of the carbon-free share but not the direction of the trends.

Definitions. “Fossil fuels” means oil, natural gas, and coal. “Carbon-free” means nuclear, hydroelectric, and other renewables (renewables excluding hydro), so no source double-counts. “Total energy” sums fossil and carbon-free.

Carbon-free shares. The total-energy carbon-free share divides carbon-free energy by total primary energy, from the consumption-in-EJ tables.

Decarbonization and carbon intensity. Carbon intensity of GDP divides CO₂ from energy by real GDP. GDP uses World Bank GDP, PPP (constant 2021 international dollars), indicator NY.GDP.MKTP.PP.KD, which runs through 2024. Where a 2025 value applies, 2025 GDP extends from 2024 using country real-GDP growth estimates from the IMF World Economic Outlook, October 2025; these 2025 figures remain estimates, not final actuals, and carry that label. The figures index each series to a base year (1992 = 100 for the world curve; 1990 = 100 for the country series), so only ratios of GDP enter and currency units cancel.

Kaya decomposition (Figure 9). Carbon intensity of GDP equals the product of two factors: energy intensity of GDP (primary energy ÷ GDP) and carbon intensity of energy (CO₂ ÷ primary energy). Each factor and their product index to 1990 = 100. By construction the product line equals the carbon-intensity-of-GDP series in Figure 7 rebased to 1990.

Carbon intensity change table (2024–2025). For each G20 economy and the world, the change in carbon intensity of GDP comes directly from the two growth rates as (1 + CO₂ growth) ÷ (1 + GDP growth) − 1, using EI emissions growth 2024→2025 and IMF October 2025 real-GDP growth. This approach avoids constructing absolute 2025 GDP levels. The ranking runs from the largest reduction (most negative) to the largest increase.

Country selection. The top-20 emitters and the G20 come from 2024 CO₂ emissions in the Statistical Review; supranational aggregates such as the European Union drop out of country rankings.

Approximations. No values in the figures or table approximate or interpolate anything except the labeled 2025 GDP estimates described above. All emissions and energy values match the 2026 Statistical Review as reported.

Sources and data file below.

Read more

Subscribe now

Note: Originally published 2 Oct 2023

Over the weekend, so-called “video-assisted referees” (or VAR) made a major mistake in the Liverpool-Tottenham match that denied Liverpool a goal, which proved to be the difference in the match. It was not the first error made by VAR — Sky Sports has documented 14 instances of VAR errors in the Premier League over just the past 2 seasons, for which an apology was later issued, as shown in the graphic below.

In my book The Edge, I made the case for a light touch with technology-assisted refereeing. Below I have excerpted that argument. At the bottom, THB paid subscribers can download a PDF of the full chapter, which is titled Hacking the Athlete and the Games.

Enjoy!

Share

Part of the application of technology to refereeing games is about reducing the uncertainty in judgments, but another part is about legitimacy; that is, the general acceptance of referee judgments and the overall integrity of compe­tition. Technology, it turns out, presents challenges to legitimacy and the key to preserving legitimacy. Let me explain.

In general, people—especially the sports-viewing public—un­derstand uncertainties just fine. Studies of public understanding of probabilities related to weather forecasts indicate that, when it comes to the weather, the public actually has an appreciation for probabilistic information, even when the information is provided with no mention of the possibility that the forecast might be wrong. It turns out that we each have enough experience with weather fore­casts to be able to develop a sense of their accuracy.

Such research on the public understanding of science suggests that we should be cautious about assuming how the public might react to information in other contexts as well. Now let’s apply this thinking to sports. Consider that most sports broadcasts are accom­panied by a wealth of statistical information, some of which is very sophisticated. The role of technology-assisted refereeing in sports is primarily about legitimacy—the better that referees understand what just happened on the field of play, the more likely they are to make the right calls, which in turn means that the players and the fans are more likely to regard the outcome of a competition as fair, and a system that delivers fair outcomes is more likely to be regarded by athletes and fans as legitimate.

Consider two shots on goal in the World Cup that occurred forty-four years apart. In the 1966 World Cup final, England was awarded a goal in extra time against West Germany when the ball may or may not have crossed the goal line after ricocheting straight down off the crossbar. Video evidence remains inconclusive (though one analysis conducted thirty years later by engineers at Oxford University found that the ball was six centimeters short of fully crossing the line). Nonetheless, the goal was given and be­came part of football lore. Uncertainty still lingers over whether the ball actually crossed the goal line, and a half century later, some Germany supporters still feel wronged.

In contrast, the persistence of uncertainty today is often not possible. In the 2010 World Cup, England’s Frank Lampard had an obvious goal disallowed, also against Germany, this time in the quarterfinals. Like its doppelganger generations before, Lampard’s shot ricocheted downward off the cross bar. In this case, however, the ball undoubtedly crossed the goal line, according to the video replays. Although Germans with long memories called it payback for 1966, the English felt aggrieved, and most neutral observers found the incident problematic: one of the most important events to occur on the pitch was missed by the officials but seen clearly by millions around the world. There was no hiding behind fuzzy black-and-white television imagery. The wrong decision was there for ev­eryone to see.

It is this difference between what the referee can detect on the pitch and what television viewers can see at home—with ultra-slow-motion replays in high definition—that has led to the intro­duction of technological aids to assist referees. Equipped with this high-tech assistance, human referees are now faced with about the same level of uncertainty that remote spectators (the fans watching their TVs) must deal with. Referees’ uncertainty has been realigned with spectators’ uncertainty.

But sports have stopped short of trying to reduce uncertainty in every decision that referees make. It’s just not practical or afford­able—or perhaps even necessary from a legitimacy standpoint—to do so. In soccer, goal-line technologies are used to detect whether balls cross the goal line, not to detect whether balls go out of bounds or who should have subsequent possession. In tennis, the Hawk-Eye system is used sparingly, with players granted a limited number of challenges to deploy it in a match. In professional cricket and bas­ketball, video replay as an officiating aid is also limited to certain situations.

With the introduction of technology, uncertainty does not go away. During the 2013 French Open, Ukrainian tennis player Sergiy Stakhovsky received much attention (and a fine from officials) for taking a picture of a ball mark on the clay with his iPhone and then tweeting it to document a dispute over a ball that was ruled out. When Stakhovsky had done the same thing in Munich a few weeks earlier, several of his colleagues on the professional tennis tour tweeted back to dispute his complaint. The introduction of tech­nology has not eliminated line call disputes, and social media has created new challenges for officials. One can only guess what John McEnroe, the famously controversial and colorful tennis star of the 1970s and 1980s, might have done with an iPhone and a Twitter account.

However, technology can create incentives that help human ref­erees perform better. A 2011 study by David Hamermesh and col­leagues in the American Economic Review looked at more than 3.5 million pitches in Major League Baseball games from 2004 to 2008 to assess whether umpires displayed biases in how they ruled balls and strikes. The study found a small but significant bias in how umpires ruled depending on whether the umpire’s “race” was the same or different than the pitcher or batter. When the umpire was being evaluated by a computerized system (or performed before an exceptionally large crowd or in an important game), however, the bias went away. The presence of the technology helped umpires overcome potential biases when they knew that they would be eval­uated objectively.

Technology also creates new problems. In February 2016, Col­orado State University was hosting Boise State in basketball. The teams had played a close game that went to overtime. With 0.8 sec­onds left, Boise State had the ball in its half with the score tied at 84. Boise State inbounded the ball to James Webb III, who took two quick steps and, before the buzzer, launched a one-handed shot as he was falling out of bounds. Miraculously, the ball went in. Boise State won, right?

Nope. The referees believed that the game clock had not started on time, so they went to the video replay. The video replay technol­ogy allowed the referees to compare when the clock started, when it should have started, and how long it took for Webb to get the shot off. The referees determined, based on the video replay, that Webb took about 1.2 seconds to get his shot off, more than the 0.8 sec­onds remaining. The shot was waved off, and the game went into a second overtime, from which CSU emerged victorious.

It did not take long for TV viewers and reporters to perform their own timing of the final play to discover that Webb’s shot was launched in only 0.6 seconds, not the 1.2 seconds claimed by the referees. What had happened? The software used by the replay technology had a bug—it counted the elapsed time twice as fast as what really occurred. The Mountain West Conference, in which both teams played, issued a statement explaining that Boise State should have won the game. Under NCAA rules, protests of such mistakes are not allowed, so Colorado State kept the victory. The conference announced that changes would be made to the rules governing the use of video replays and the episode would be used as a future teach­ing tool. Technology doesn’t necessarily get rid of uncertainty—it may just move it from one setting to another.

The improved alignment of refereeing decisions and what the public observes on television through the introduction of technol­ogy is to be applauded. When such decisions fall out of alignment—whether less or more precise than what the public demands—sport faces a legitimacy crisis that then necessitates innovation.

For instance, sometimes we look to technology to do the impos­sible. The struggles that the NFL has faced in defining a “catch” illus­trate this point. Here is an experiment that you can try right now. Pick up a ball (or any tossable object). Throw the ball to your nearest companion and ask him or her to catch it. Now answer this: Did he or she catch your throw? I bet that neither one of you will have any difficulty answering either yes or no. Understanding the catching of a ball doesn’t require nuance, subtlety, or litigious interpretation. That used to be the case in the NFL. But not anymore.

As recently as 1996, the NFL needed only 109 words to define what it meant to “catch” a football.* This is the number of words that appear in the prior paragraph. By 2015, the NFL definition of a “catch” had swollen to almost 600 words, and the definition of a catch had become an endless point of controversy among fans, play­ers, officials, and the media.

Defining a catch became difficult after the introduction of high-definition slow-motion instant replay. Today’s television technology allows viewers to have a far better look at the act of catching a ball than referees could ever hope for on the field. Using instant replay can help referees have the same view as everyone else, but it also forces the NFL to define more precisely what a catch actually is. It was clever when the folksy TV commentator and former coach John Madden used to explain that a player getting one knee down in­bounds while making a catch was equivalent to having two feet come down in bounds. Madden wasn’t invoking any real rule, just help­ing the viewer to make sense of referee judgments. High-definition TV and instant replay has consigned folksy interpretations of the rules to the era of vacuum tubes and low-definition broadcasts.

Today, referees need to be able to address what happens when the player bobbles the ball, or the ball touches the ground while it is being firmly held, or when the ball is jarred loose and hits the ground just after a player crosses the goal line for a touchdown. Ev­eryone watching can see when these things occur, often in excruci­ating detail. Then the definition of a “catch” gets tough. The act of catching a ball is not, it turns out, a discrete event like a ball crossing a goal line. Whatever a catch is, it is ultimately a judgment made by a referee. Judging if a ball was caught or not is more like judging if pass interference occurred (“pass interference” is called when the defender makes prohibited contact with a would-be receiver while a pass is in the air) than it is like determining if a ball crossed a goal line. The closer we look at the making of a catch, the more contin­gencies we see, which has led to more considerations being identi­fied that referees must be aware of. Hence, the word count in NFL rules for what it means to catch a ball has increased by 600 percent.

The best way for the NFL to deal with challenges over the defini­tion of catching the ball is to accept that it can’t be precisely defined. The NFL should simply go back to letting referees decide if a ball was caught or not, based on the sort of general guidance provided in the old 109-word definition. Such an approach is not all that radical, as the NFL routinely determines that some decisions are not review­able, such as whether a field goal that passes over the upright is good or not. In that case, technology could certainly be used, but a deci­sion has been made not to go there. And football has survived. Using technology in sport in a smart way means giving up the idea that its use can eliminate uncertainty—and that means that sometimes it is better to figure out how to live with uncertainties rather than try to completely eliminate them.

Let’s consider one more example that illustrates this last point. In baseball, pitches are judged to be strikes or balls by a human um­pire. In 2015, FiveThirtyEight, a website owned by ESPN, published an evaluation of umpire accuracy in MLB. It found that the average MLB umpire is 86 percent accurate in judging balls and strikes. In a game that involves 280 pitches (for both teams), the umpire might make 40 mistakes. That is a lot of mistakes.

The technology exists to standardize the calling of balls and strikes in baseball, and the technology has been tested in the mi­nor leagues. There are arguments for and against adopting the tech­nology. An argument for adopting it is that it would standardize the definition and the strike zone and apply the standard impartially to each player. However, some observers argue that umpires and their judgment are a key part of the game. Wherever you come out on this issue, at its core is the notion of uncertainty and judg­ment. Is umpire error (14 percent on average) part of the game of baseball? Or is umpire error something to be reduced and, ideally, eliminated? In 2003, Arizona Diamondbacks pitcher Curt Schilling answered this question by taking a baseball bat to one of the cam­eras of an early computerized pitch-calling system. Your answer to this question gets to the core of what you think baseball is as a sport and the role that technology has in it.

Put me down on the side of keeping the imperfect umpires in place. Why? I can think of at least three reasons. First, for better or worse, human umpires are part of the tradition of baseball. Even if computers are more accurate, deploying them completely changes that tradition, which is an important norm of the sport. Second, umpire uncertainty is also essential to the game of baseball. Re­move the umps and you have a different game. Finally, if baseball is, as Pittsburgh Pirate first baseman Willie Stargell once said, “a reflection of life,” then uncertainty in the application of the rules deserves a prominent place in the game.

Comments, discussion, debate — all welcomed!

Leave a comment

Share

Thanks for reading! Below, THB paid subscribers can download a PDF of Chapter 7 of The Edge — Hacking the Athlete and the Games.

Subscribe now

Read more

Subscribe now

A disaster loss normalization estimates what direct economic losses related to an extreme event of the past would occur if that same event happened under today’s economic conditions — today’s infrastructure, property, and wealth. Today, I present a new normalization of Europe’s weather and climate catastrophe losses for 1990 to 2024, updating an analysis I first shared here at THB three years ago, covering 1995 to 2019.

The European Environment Agency (EEA) has since updated and reorganized its underlying catastrophe loss dataset. That increases the time series by ten years, includes the catastrophic 2021 floods in Germany and Belgium, and provides a broader country grouping (the full EEA-38, which adds the Western Balkans).1

Countries included in the analysis are shown below.

Here is the top-line conclusion:

Once you account for economic growth, the normalized cost of weather and climate extremes in Europe has not increased over 1990–2024. The overall trend is flat.

This post shares the data, methods, and the significance of the analysis. At the bottom of the post is a methods appendix, a bibliography, and for THB paid subscribers an Excel spreadsheet with all sources, data, and calculations.2

Inflation-adjusted losses are increasing

Economic losses from weather- and climate-related events across the 38 EEA countries over 1990–2024 (in inflation-adjusted 2024 Euros) have increased, as shown in the figure below.

The time series shows a sharp increase over more than three decades. Taken at face value, a casual observer would conclude European disasters are becoming dramatically more expensive, with a growing impact on the continent’s economy.

That conclusion — as expressed in the quote at the top of this post — would be wrong.

Normalization: accounting for economic growth

Over these 35 years the combined economy of the 38 countries more than doubled in real, inflation-adjusted terms — accompanied by an increase in buildings, infrastructure, and wealth, and therefore far more stuff exposed to any storm or flood.

All else equal, the euro cost of a fixed set of physical events must increase simply because there is more wealth exposed to extreme events.

A normalization corrects for changing exposure and vulnerability. The simplest version scales each year’s loss by the ratio of a base year’s economic output (GDP, in this case 2024) to that year’s output:

Normalized loss in year t = Reported loss in year t × ( GDP in 2024 ÷ GDP in year t )

There are in the literature far more sophisticated methodological approaches to normalization. However, the United Nations Sendai Framework on Disaster Risk Reduction has adopted losses as a percent of GDP as an official indicator of progress in disaster risk reduction.

The figure below shows a GDP-based normalization of the time series shown above, adjusted to a constant 2024 level — the dramatic upwards trend in the time series disappears.

The early 1990s — modest in raw terms — become some of the most damaging years on record once exposure is held constant. The 1999 windstorms Lothar and Martin, the 2002 central-European floods, and the 1990 storm season rival or exceed the recent large-loss years.

At ~0.25% of GDP, 2021 was the most economically significant loss year since 2002.. However, normalized losses in 1990 and 1999 were even greater.

A robustness check: does the method matter?

Above, I normalized the aggregate continental losses based on aggregate GDP — scaling Europe’s total losses by Europe’s total GDP growth. But these countries did not grow at the same rate. Poland, Romania, and the Baltics expanded several times faster than Germany or France over this period. If a disproportionate share of early losses fell in fast-growing economies, normalizing the bloc as a whole could distort the picture.

As a robustness check I also normalize losses reported for each country by its own GDP growth and then sum the results. A comparison of the two methods is shown below, and there is very little difference.

Variation within Europe

There are some interesting details at the country level.

First, in aggregated terms, normalized losses are heavily concentrated in the largest economies. Germany, France, Italy, and Spain together account for ~58% of all normalized European losses since 1990; the top eight account for ~80%.

The concentration of normalized losses in a small number of countries largely reflects economic size — big economies post big absolute losses because they are big, not rocket science.

More interesting is the overall economic burden: how large are each country’s losses relative to its own economy? The answer is shown below.

In the context of individual country’s GDP, it is the smaller, more exposed economies that bear the heaviest relative load. Slovenia tops the list at ~1.0% of GDP over 1990-2024 — driven heavily by its catastrophic 2023 floods — followed by Bosnia and Herzegovina and Serbia, all well above the ~0.12% EEA average.

In contrast, Germany, France, Italy, and Spain, which dominate the pie, sit in the middle of the relative ranking.

Why normalization matters

Normalization of Europe’s weather and climate losses helps to answer a narrow question:

Are weather and climate extracting a growing economic toll from Europe over time, once accounting for the fact that economies are growing, resulting in increased exposure to loss?

The evidence presented above shows that the answer is no — after normalization by GDP the resulting time series has no trend. That means that direct economic losses from weather and climate events have scaled proportionately with the region’s growing economy.

It is important to emphasize that this result says nothing about changes in climate. To detect changes in climate variables, always look at weather and climate data — not economic data.

The finding of no trend in normalized losses across the EEA-38 is fully consistent with peer-reviewed literature (see the bibliography below) and with what the IPCC concludes about the difficulty of detecting (and attributing) long-term trends in most categories of extreme weather.

Regardless how climate varies and changes, future losses from extremes will be largely a function of where and how we build, and policies put into place in recognition of the ever-present risks of extreme events.

Subscribe now

Methods Appendix Thb Eea38 Normalization

202KB ∙ PDF file

Download

Download

Comments, critique, debate, discussion — All welcome!

Leave a comment

BIBLIOGRAPHY

Literature review

Pielke, R. (2021). Economic ‘normalisation’ of disaster losses 1998–2020: a literature review and assessment. Environmental Hazards, 20(2), 93–111. https://doi.org/10.1080/17477891.2020.1800440

Surveys 54 peer-reviewed normalization studies published 1998–2020 across regions, countries, and hazard types. Concludes there is little evidence that any part of the documented increase in global economic losses on climate timescales is attributable to human-caused climate change, with increases instead explained by growth in exposure, wealth, and development. Prepared as a contribution to the IPCC AR6 assessment of this literature (however, it was not cited in the AR6, nor were almost all of the studies in the review).

European normalization studies

Barredo, J. I. (2009). Normalised flood losses in Europe: 1970–2006. Natural Hazards and Earth System Sciences, 9(1), 97–104. https://doi.org/10.5194/nhess-9-97-2009

The foundational normalization of European flood losses. Finds no significant trend in normalized flood losses across Europe once losses are adjusted for changes in population, wealth, and inflation. The methodology underpins the European Environment Agency’s normalized flood-loss indicator.

Barredo, J. I. (2010). No upward trend in normalised windstorm losses in Europe: 1970–2008. Natural Hazards and Earth System Sciences, 10(1), 97–104. https://doi.org/10.5194/nhess-10-97-2010

The companion windstorm study. Finds no upward trend in normalized European windstorm losses over 1970–2008, consistent with the absence of a robust trend in European storminess.

Paprotny, D., Sebastian, A., Morales-Nápoles, O., & Jonkman, S. N. (2018). Trends in flood losses in Europe over the past 150 years. Nature Communications, 9, 1985. https://doi.org/10.1038/s41467-018-04253-1

The most temporally and spatially comprehensive European flood-loss normalization to date, covering 37 European countries back to 1870 using a gridded reconstruction of flood exposure (the HANZE database). After correcting for exposure change, finds increases in inundated area and persons affected since 1870 but a substantial decline in flood fatalities and, in recent decades, in financial losses per year — while cautioning that underreporting of smaller floods affects observed trends.

Formetta, G., & Feyen, L. (2019). Empirical evidence of declining global vulnerability to climate-related hazards. Global Environmental Change, 57, 101920. https://doi.org/10.1016/j.gloenvcha.2019.05.004

Though global in scope, this widely-cited vulnerability analysis is a standard reference in the European normalization literature, documenting declining human and economic vulnerability to climate-related hazards as income rises — the mechanism that offsets rising exposure.

Spinoni, J., Formetta, G., Mentaschi, L., Forzieri, G., & Feyen, L. (2020). Global warming and windstorm impacts in the EU: JRC PESETA IV project – Task 13. Publications Office of the European Union, Luxembourg. https://doi.org/10.2760/039014

EU-wide assessment of windstorm impacts. Reports absolute losses highest in Germany, France, Italy, and the UK, but losses relative to GDP roughly double the EU average in Bulgaria, Estonia, and elevated in the Baltics and Slovenia — the same relative-burden inversion seen when losses are scaled to national economies. Notes no robust trend in European windstorms.

Barredo, J. I., Saurí, D., & Llasat, M. C. (2012). Assessing trends in insured losses from floods in Spain 1971–2008. Natural Hazards and Earth System Sciences, 12(5), 1723–1729. https://doi.org/10.5194/nhess-12-1723-2012

Normalizes insured flood losses for Spain and finds no upward trend once socioeconomic change is accounted for.

Underlying exposure and loss databases

Paprotny, D., Morales-Nápoles, O., & Jonkman, S. N. (2018). HANZE: a pan-European database of exposure to natural hazards and damaging historical floods since 1870. Earth System Science Data, 10(1), 565–581. https://doi.org/10.5194/essd-10-565-2018

The Historical Analysis of Natural HaZards in Europe (HANZE) exposure database, providing the gridded population, asset, and land-use reconstruction that enables exposure-based flood normalization across Europe.

Paprotny, D., et al. (2024). HANZE v2.1: an improved database of flood impacts in Europe from 1870 to 2020. Earth System Science Data, 16, 5145–5170. https://doi.org/10.5194/essd-16-5145-2024

The updated HANZE flood-impact database covering 42 European countries and 2,521 events (1870–2020), compiled from more than 800 sources. Documents the substantial effect of impact-data completeness on estimated normalized-loss trends.

Daniell, J. E., Khazai, B., Wenzel, F., & Vervaeck, A. (2011). The CATDAT damaging earthquakes database. Natural Hazards and Earth System Sciences, 11(8), 2235–2251. https://doi.org/10.5194/nhess-11-2235-2011

Describes CATDAT (Risklayer GmbH), the catastrophe-loss database underlying the European Environment Agency’s economic-loss dashboards used in this analysis. Originally developed for earthquakes, CATDAT now spans weather- and climate-related hazards.

THB on normalization

• Part 1: What is a “normalization”?

• Part 2: Normalized disaster losses in Europe 1995 to 2019

• Part 3: Normalized Hurricane Losses in the United States, 1900 to 2021

• Part 4: How would we know if disasters are becoming more costly due to climate change?

• Part 5: Climate change and disaster losses

Read more

Subscribe now

June was another busy month here at THB. I published a three-part series on nuclear energy, THB Deep Dive Report #2 on ENSO, Hurricanes, and Global Disasters, an update on the ongoing economic fallout of the Iran war, a report on new research on IPCC bias, explored the implications of Europe’s aversion to AC, and much more.

Below are links and capsule summaries of every post. In the capsule summaries I highlight the material available only to paid subscribers.

Five of My Favorite Papers and the Stories Behind Them (June 4)

A paywall-free share of a post originally written for paid subscribers. Five of my favorite papers that I’ve published from over thirty years of research — not my most cited, but (some of) my favorites. The backstories range from the 1997 Red River flood survey that shaped my thinking on forecasts and decision making, to a tour of reclaimed coal-mining lakes in Germany, to the Caster Semenya arbitration in Lausanne and the bad science used in that case, to the paper that pulled me back into climate research after several years away.

Partisan Politics Trumps Peer Review (June 4)

This post critiques OMB’s proposed rule reshaping federal science funding across NSF, NIH, NASA, DOE, and more, through the framework of The Honest Broker. The post identifies five core problems: political appointees trumping peer review, the embedding of unclear normative standards into grant criteria, a regulatory foundation built on partisan advocacy documents rather than the science policy literature, heavy restrictions on foreign-collaboration, and a final-say structure that substitutes partisan acceptability for democratic accountability. The post closes with a specific action agenda — and a warning to the scientific community about the risks of fighting partisan fire with more partisan fire.

Does the IPCC Exaggerate Climate Science? (June 8)

A potentially important new preprint by Galiani et al. on how the IPCC process amplifies scientific claims beyond what is supported in the literature. They scored ~114,000 matched claim pairs across all six IPCC Assessment Reports and ~116,000 newspaper articles using three independent LLMs. The headline finding: at every stage, in every Assessment Report, claims shift systematically toward the more severe end of the underlying scientific ranges. It’s a preprint, so the findings are preliminary, but it’s the first systematic attempt I’m aware of to empirically quantify the amplification of climate science from report to Summary for Policymakers to press coverage.

The New Republic of Science (June 11)

This post shares my reflections from a Vienna workshop on current disruptions in science policy. Framed through Michael Polanyi’s 1962 essay “The Republic of Science,” the post identifies six tensions straining the postwar model of self-governing science: The post argues that the Trump administration’s actions, including DOGE cuts and the proposed OMB rule, are real and harmful, but are symptoms of these deeper structural tensions rather than their cause, and the Biden administration’s pathological politicization of science is also part of this longer term dynamic. The post closes with three design principles for renegotiating science’s social contract.

Can America Build Nuclear Again? Part 1 (June 15)

The first installment of a three-part series on commercial nuclear power. Part 1 makes the case — via a cross-national comparison of France, Japan, South Korea, and China — that U.S. nuclear cost escalation was never technologically inevitable. Overnight construction costs rose from roughly $1,000/kW in the 1960s to nearly $11,000/kW for plants completed after Three Mile Island; other countries didn’t experience this sort of cost escalation. The divergence reflects policy choices, not anything intrinsic to nuclear power.

Can America Build Nuclear Again? Part 2 (June 17)

Part 2 connects U.S. cost escalation to seven federal mechanisms: full fee recovery, mid-construction backfitting authority, 1960s-era quality assurance mandates, NEPA requirements locked in by Calvert Cliffs (1971), decommissioning funding mandates, Price-Anderson insurance, and IRA credit-timing asymmetry — together producing a capital cost differential of roughly 15–20× between new nuclear and new gas. The post also traces how the 1977 abolition of the Joint Committee on Atomic Energy left nuclear policy split across eleven congressional committees, with none currently equipped to consider nuclear energy comprehensively.

Can America Build Nuclear Again? Part 3 (June 19)

The final installment reviews eleven policy options drawn from academic research, government reports, and nuclear policy organizations — from enforcing ADVANCE Act licensing timelines and codifying design finality, to volume licensing and a federal government orderbook of five to ten reactors, to long-term investment in domestic nuclear manufacturing and UK-style Contracts for Difference. Includes a 13-tab XLS workbook with all data, sources, and cost estimates across the three parts for paid subscribers.

The Economic Costs of the Iran War: Update and Scorecard (June 22)

An update and retrospective on my April 2 Iran War cost projections, which held up well against actual outcomes. Through mid-June, the average U.S. household had experienced approximately $555 in elevated fuel and fertilizer costs, with total per-household costs (including government expenditures) running from ~$775 to ~$1,313 depending on assumptions — consistent with an independent Moody’s Analytics estimate of ~$1,000. Full sourced spreadsheet included for all subscribers.

ENSO, U.S. Hurricanes, and Global Disasters (June 24) — THB Deep Dive Report #2

This THB Deep Dive report updates two long-standing lines of research. First, the original 1999 Pielke & Landsea finding that U.S. hurricane damage varies strongly with ENSO phase holds up — the La Niña-to-El Niño ratio of median normalized CONUS hurricane damage was 4.0× in the original 1925–1997 sample and 7.5× over 1998–2025. Second, global disaster losses over 1990–2024 show a strong La Niña signal: median global losses run ~$270 billion in La Niña years versus ~$155 billion in El Niño years. As of late June, NOAA carried an El Niño Advisory, with a 63% probability of a “very strong” El Niño this fall — implying below-average base-rate expectations for 2026 U.S. hurricane losses, though tail risk remains. The full report is available to paid subscribers.

Europe’s Deadly Aversion to Air Conditioning (June 25)

Europe pairs the lowest air-conditioning penetration among rich regions (~19% household AC, versus ~76% in North America and 90%+ in Japan) with the world’s highest per-capita heat-death rate — roughly 50,000 to 68,000 heat deaths a year across 2022–2024. Using Barreca et al.’s estimate that AC reduces the risk of dying on an extreme heat day by about 75%, the post models a counterfactual: estimating what North American AC penetration in Europe would imply for heat mortality. The result: avoiding of ~26,000 European heat deaths in a summer like 2022; near-universal coverage would avoid ~35,000. Heat deaths concentrate overwhelmingly among the elderly, so targeted AC access in care homes and the residences of older adults could capture most of the benefit. The post traces the policy levers — EU and national rules on insulation, permitting, and energy use — that actively discourage AC adoption, and connects the issue to a broader argument (from a new preprint with Matthew Burgess, Patrick Brown, and Matthew Kahn) that climate advocacy has long underrated adaptation relative to mitigation. Includes a sourced Excel workbook for paid subscribers.

Another Pillar of Climate Advocacy Collapses (June 28)

This post follows up the ProPublica investigation revealing that BP helped conceive of and arguably ghostwrite the influential 2004 Pacala-Socolow “stabilization wedges” paper, which claimed humanity already possessed the technology needed to solve climate change for 50 years. The post traces the scientific debate the Wedges paper was designed to shut down — Hoffert, Caldeira, and colleagues had argued in 2002 that existing technology was nowhere near sufficient — and walks through how BP CEO John Browne suggested the paper’s central “wedges” metaphor and how Princeton researchers acknowledged the paper likely wouldn’t have existed without BP funding, none of which was disclosed. The post also revisits my 2008 Nature commentary “Dangerous Assumptions” (with Tom Wigley and Christopher Green), which challenged the Wedges conclusions. I also discuss the subsequent multi-year campaign by the Center for American Progress and others to delegitimize anyone who questioned Wedges orthodoxy. Includes a bonus PDF of the stabilization-wedges chapter from The Climate Fix (2010) for paid subscribers.

Comments, suggestions, requests — All welcome!

Leave a comment

Before you go: If you would like these month-in-review posts to continue, please let me know by clicking that “❤️ Like” button. Based on traffic in past monthly summaries, I’m not sure they add much value. Thanks!

Subscribe now

Subscribe now

A blockbuster ProPublica investigation published last week revealed that a fossil fuel company helped shape and even ghostwrite one of the most influential climate papers ever published.1

That 2004 paper — Stabilization Wedges: Solving the Climate Problem for the Next 50 Years with Current Technologies by Stephen Pacala and Robert Socolow of Princeton University — argued that humanity already possessed all the technology needed to solve the climate problem well into the 21st century.

We didn’t then and we still don’t today.

This post traces the story of the “Wedges” paper, including the scientific debate over technological readiness that the Wedges paper sought to end, the political and economic motivations behind the paper, the philanthropic machinery built on top of it, and the ruthless campaign to silence those who questioned the notion that we had in hand all the technologies needed to solve climate change.

Those efforts set back efforts to accelerate decarbonization for a generation.

Grab a coffee and settle in . . .

On one side: Successful climate mitigation requires innovation

In 2002, Martin Hoffert, Ken Caldeira, and eighteen co-authors — including Christopher Green and Tom Wigley — published a paper in Science titled “Advanced Technology Paths to Global Climate Stability: Energy for a Greenhouse Planet.”

They argued that stabilizing CO₂ concentrations in the atmosphere was fundamentally an energy problem. Meeting mid-century power requirements without CO₂ emissions would demand several times what the world then derived from fossil fuels, even with efficiency gains. They explained that the technologies then available were not nearly up to meeting that challenge.

Hoffert and colleagues argued that the risks associated with climate change were serious enough that false reassurance about technological readiness could lead to flawed policy by redirecting attention and investment away from the innovation most urgently required.

Not long after, Ted Nordhaus and Michael Shellenberger — who would go on to found The Breakthrough Institute in 2007 — reached a closely related conclusion from the direction of political strategy. Their 2004 essay “The Death of Environmentalism” argued that the climate movement had trapped itself in a politics of limits and regulatory incrementalism incapable of building a coalition needed for large-scale policy action.

Their 2007 book Break Through made the case for large-scale public investment in clean energy R&D — which they characterized in terms of a new Apollo program. Time magazine named them “Heroes of the Environment” in 2008, calling Break Through “prescient” for its prediction that climate policy should focus on making clean energy cheap rather than making fossil fuels expensive.

On the other side: we already had all the needed technologies

In 2004, Princeton academics Stephen Pacala and Robert Socolow published “Stabilization Wedges: Solving the Climate Problem for the Next 50 Years with Current Technologies” in Science.

Each wedge in the Pacala-Socolow framework represented a strategy that, if deployed and scaled up over 50 years, would avoid the emission of one billion tons of carbon per year by 2054. A full portfolio of seven wedges would together hold global emissions flat at 2004 levels rather than allowing them to follow the projected upward trajectory, thereby stabilizing atmospheric CO₂ concentration levels below a doubling of the pre-industrial baseline. At the time, climate policy focused on stabilizing concentrations, not temperature targets or net-zero emissions.

Their paper explained:

Wedges can be achieved from energy efficiency, from the decarbonization of the supply of electricity and fuels (by means of fuel shifting, carbon capture and storage, nuclear energy, and renewable energy), and from biological storage in forests and agricultural soils. Below, we discuss 15 different examples of options that are already deployed at an industrial scale and that could be scaled up further to produce at least one wedge.

They argued that achieving implementation their wedge strategy would constitute climate policy success, and that doing so was possible with technologies already in existence:

“Humanity already possesses the fundamental scientific, technical, and industrial know-how to solve the carbon and climate problem for the next half-century.”

Three clarifications about what they were arguing:

• First, the paper was about stabilization of atmospheric concentrations of carbon dioxide, not net zero. Net zero emissions was a concept that entered climate policy discourse only much later.2

• Second, the timescale was 50 years, not a few decades. Each wedge represented a half-century deployment challenge, not anything that could succeed over years or even a decade.

• Third, the paper’s ambition was limited. Seven wedges held concentrations at roughly 500 ppm. Wedge advocates later inflated this into claims about achieving 450 ppm or even 350 ppm stabilization — targets the paper never addressed.

Al Gore featured the Wedges framework in his 2006 documentary An Inconvenient Truth, presenting it as proof that we already had all the technology we needed. The problem, as Gore characterized it, was simply a matter of political will.

The paper has more than 4,500 citations in the scientific literature. It was cited in assessment reports. It showed up in college classrooms for a generation of students. Someone made it into a board game.

Much more significantly than any academic impact, in the climate advocacy world, Wedges became the intellectual centerpiece of lobbying, philanthropy, and the basis for efforts — largely successful for a time — to shut down any discussion of the framework’s weaknesses and shortfalls.

The political function of a scientific claim

Pacala later disclosed the paper’s political motivation. In 2008, he explained that its purpose was,

. . . narrow and simple — we wanted to stop the Bush administration from what we saw as a strategy to stall action on global warming by claiming that we lacked the technology to tackle it.

I saw it as an unhealthy collusion between the scientific community who believed that there was a serious problem and a political movement that didn’t. I wanted that to stop and the paper for me was surprisingly effective at doing that.

Whatever shortfalls in the Bush administration’s broader posture on climate, its emphasis on energy R&D was on target. Pacala was acknowledging that the Wedges paper’s core scientific claim — that existing technology sufficed — was constructed to win a political argument with policy makers that he disagreed with, using the authority of science to do it.

What neither the scientific community nor the public knew at the time — and what ProPublica’s investigation exposes — was that the Wedges paper was motivated not just by politics but by hidden funding and influence of a major fossil fuel company, BP.

The problem here is not that a fossil fuel company sponsored research, but that executives played a major role in conceiving of and ghost writing the Wedges paper, and none of this was disclosed. In my view, if the reporting by ProPublica holds up, Science should retract the paper.3

BP funded Princeton’s Carbon Mitigation Initiative to pursue research into addressing climate change without abandoning fossil fuels. Until now,the BP role was hidden, and included collaborating on paper drafts as they were passed back and forth with the Princeton researchers.

In fact, BP chief executive John Browne suggested the paper’s central metaphor — “wedges” — during a London meeting at which Pacala pitched the framework. ProPublica provides amazing detail (emphasis added):

Pacala, whom a colleague described as an expert “pitchman,” presented his chart of ideas: Use oil and gas more efficiently. Replace coal-fired power plants. Reduce emissions, ultimately, by capturing them and burying them underground. Each action, he said, would take “slices” out of the total amount of future carbon pollution.

Browne listened attentively. The straightforward framework made a complex problem seem manageable. But the “slices” terminology confused him. “They’re kind of wedges, aren’t they?” Pacala recalls him saying.

“We’re like, ‘Yeah, whatever you want,’” Pacala remembers thinking. “‘You’re paying the bills, buddy.’”

Coincident with the publication of the Wedges paper in Science, Browne published an endorsement of the paper in Foreign Affairs, citing it as evidence that oil and gas had “sustainable futures.”

Neither the funding nor the ghost authorship was disclosed in the coordinated communications campaign.

Socolow told ProPublica the paper would likely never have existed without BP’s funding. BP held a direct financial stake in the answer to the question the paper was purportedly answering.

As Caldeira and Hoffert had written two years before the Wedges paper appeared, the technologies that the paper focused on were not mature, and the paper obscured rather than resolved the technology challenge.

In 2013, Davis, Cao, Caldeira, and Hoffert formalized that critique in “Rethinking Wedges” in Environmental Research Letters:

“Pacala and Socolow gave us a way to believe that the energy-carbon-climate problem was manageable. An unfortunate consequence of their paper was to make the solution seem easy.”

The Wedges quickly moved to the center of the climate movement

The Wedges claim that we had all the technology we needed became the intellectual foundation for a major coordinated philanthropic enterprise focused on a promoting one perspective on climate policy while silencing all others.

Thanks to the work of Matthew Nisbet, then at American University, we have a picture of how this movement developed. Nisbet documented the dynamics in his important 2011 report, Climate Shift, which examined the financial resources, strategies, and communication choices of the dominant organizations advocating for action on climate change.

Nisbet explains that in 2007, the Packard, Doris Duke, Energy, Joyce, Oak, and Hewlett Foundations commissioned a report” Design to Win: Philanthropy’s Role in the Fight Against Global Warming. The report treated the Wedges perspective on available technologies as fact.

The report explained that for climate advocates, the top priority was political and regulatory — block fossil fuels, promote efficiency and renewables, win the political fight. R&D and innovation were treated as secondary at best to the political and regulatory fight — or worse, as cover for opposition to climate policies.

The Hewlett, Packard, and McKnight Foundations then launched ClimateWorks Foundation with more than $1.1 billion in committed funding to implement the Design to Win strategy.

Questioning whether the Wedges analysis was correct or discussing the need for technological innovation risked serious professional repercussions, as well-funded and powerful climate advocates of the era were quick to punish anyone deviating from orthodoxy.

Of course, I was not shy in my critiques of Wedges, my informed belief that Hoffert, Caldeira, Wigley and colleagues were correct, and in promotion of the need for broad-spectrum technological innovation in energy production and consumption if deep decarbonization were ever to be possible.

“Dangerous Assumptions” and the furious response

My 2008 Nature commentary (with Tom Wigley and Christopher Green) — “Dangerous Assumptions” — built directly on the 2002 Hoffert et al. analysis. We made the case that all of the IPCC’s baseline emissions scenarios,4 — including those with lower cumulative emissions and those with very high cumulative emissions5 — had embedded enormous assumed rates of spontaneous decarbonization. That is, an assumption that future progress in energy and carbon intensity reductions would occur at a rate far beyond that observed historically, as if magically on its own, without directed policy.

The implication that we drew was fully consistent with Hoffert et al. and directly contrary to the Wedges analysis: enormous advances in energy technology were required. Our paper was well received and widely discussed:

• Nature ran a companion editorial signaling the commentary’s significance.

• Vaclav Smil said “I largely agree, but I fear that the situation is even worse than the authors imply.”

• Marty Hoffert called it a “bombshell.”

• Carnegie’s Christopher Field said: “Given recent trends, it is hard to see how, without a massive increase in investment, the requisite number of relevant technologies will be mature and available when we need them.”

Our paper, and its acceptance by leaders in the field was a direct challenge to the Wedges framework and thus much of the climate advocacy movement, to which billionaires and foundations were supporting with over a billion dollars of support.

The Center for American Progress (CAP), a progressive political advocacy group, put their senior fellow Joe Romm in charge of delegitimizing our paper and me in particular. Romm — a prolific and combative blogger who at the time was widely infuential — called the paper “pointless and misleading if not outright dangerous.”

He labeled it:

“the standard ‘Technology, technology, blah, blah, blah’ delayer message developed by Frank Luntz and perfected by Bush/Lomborg/Gingrich.”

A typical strategy of delegitimization at the time was to associate anyone who deviated from the climate catechism to Republicans — which was a particularly effective technique of delegitimization in professional circles, as the scientific and academic community increasingly associated itself with Democrats, the political left, and the climate movement.

Romm devoted well over a hundred posts at CAP to attacking my work (figure above).6 He instructed journalists not to quote me in their coverage, and most listened and for many years, some still do. Based on Romm’s fatwa, in 2012 former New York Times reporter Andy Revkin likened me to Voldemort.7 Romm refused to debate the substance, saying he had no interest in elevating the arguments.

As the Breakthrough Institute documented: Romm regularly advised his readers not to read the articles and books he criticized, and often acknowledged proudly that he had not read them himself.

Nisbet’s Climate Shift report, which documented some of these dynamics in 2011, received the same treatment. Romm broke the publication embargo before the report’s release and attacked it as “an error-riddled, self-contradictory, demonstrably false report.” One academic reviewer who had seen the draft was so disturbed by Romm’s characterization that he returned his honorarium and withdrew his name.

It may be difficult in 2026 to understand the power wielded by the Center for American Progress and its attack dog Joe Romm almost 20 years ago, but when he ordered climate advocates to jump, they jumped. The Breakthrough Institute appropriately characterized Romm’s campaign of personal and professional destruction as Climate McCarthyism.

Over a decade ago, Romm largely disappeared from the climate issue, leaving CAP to pursue other interests, including working with Michael Mann at the University of Pennsylvania.

A full accounting of the harm inflicted on climate discourse, science, and policy by CAP, Romm, and their fellow travelers is still to be written — and from someone other than me, having been one of their primary targets,

What the record shows

Both science and policy are self-correcting, even if it takes too long sometimes. With the advantage of hindsight, we now know incontrovertibly that not only was the Wedges paper just wrong, but it was also deeply compromised from the start to advance a political and economic agenda.

Energy system evolution that has occurred over the past two decades has depended on technological innovation. For instance, solar and wind power at current costs required decades of R&D, manufacturing scale-up, and policy-supported deployment to achieve incredible reductions in costs. Battery storage continues to advance as an area of intense innovation. Grid integration at high renewable penetrations requires ongoing technical development. Hard-to-abate sectors — steel, cement, shipping, aviation, industrial heat — still lack affordable, deployable low-carbon solutions at scale.

The need for continued technological innovation remains. The IEA and the IPCC now treat it as a necessity, not an option. Ironically, this was the same position that got researchers labeled “deniers” and “delayers” and subjected to coordinated campaigns of professional delegitimization and outright harassment.

Good science and better policy are winning out. That is good news.

Lessons of Wedges

The wedges episode holds some lessons.

Scientific claims designed to serve predetermined political conclusions carry risks even when — or maybe especially when — the underlying cause is broadly viewed to be legitimate or virtuous. The Wedges paper’s political purpose did not make climate change less real or less serious. But it did produce a paper whose analytical honesty suffered for its political ambitions, and that paper then became a tool for suppressing legitimate inquiry that over the course of time proved correct.

Philanthropy should do a better job supporting honest brokering — clarifying and expanding policy choice, not simply picking one approach to promote above all others and attacking others as political enemies. Political advocacy without a grounding in policy effectiveness is just pathological politics.

Here is what I wrote back in 2011, which has stood the test of time:

What explains the adherence to bad ideas in the form of bad policy? I'm not entirely sure but it just so happens that groups such as the Center for American Progress have been funded under the Design to Win strategy to spread its message. Apparently that includes a healthy dose of efforts to delegitmize alternative points of view and to poison what otherwise might be characterized as a healthy public debate over policy options. To the extent that these efforts succeed, climate policy and the broader environmental movement suffer.

Despite the policy set-back engineered by climate activists, the work of developing, scaling, and deploying the full suite of low-carbon technologies the world requires continues.

Comments, critique, debate, discussion — all welcome at THB!

Leave a comment

Before you go, a small request with big impact

Please click that “❤️ Like” button. Thanks!

Bonus! For THB paid subscribers below is a PDF of Chapter 2 of The Climate Fix (2010) with my critique of the stabilization wedges and documentation of its political purpose.

Read more

Subscribe now

Europe is warming: ~0.5°C per decade since the 1980s. Figure 1 shows the continent’s annual count of days since 1950 with “strong heat stress” — a feels-like temperature of 32°C or higher. The trend was flat into the 1980s, then rose sharply: 2022 through 2024 rank as the highest on record.

Figure 1. Annual number of days with at least 'strong heat stress' (UTCI ≥ 32°C), 1950–2024, for Europe as defined by WMO Region VI (geographic Europe plus Greenland and parts of the Middle East and Caucasus — broader than the EU figures used elsewhere). Digitized from Copernicus / ESOTC 2024, Fig. S4.3 (ERA5-HEAT).

Heat mortality has also been increasing in Europe: Summer 2022 saw an estimated ~68,000 heat-related deaths, 2023, ~50,800; 2024, ~62,800. The WHO European Region reports that heat mortality is up by about 30 percent over two decades.

This post shows that Europe largely chooses these deaths through a long resistance to a 1902 invention that the rest of the rich world treats as an incredible benefit of modern technology — air conditioning.

This post on the human toll of Europe’s aversion to airconditioning was motivated by three essays:

• Make Europe Cool Again by Kevin Kohler details how French and Swiss rules deter installed AC.

• How Europe Became the World Champion of Heat Deaths by Maarten Boudry traces the aversion to a deeper hostility toward energy.

• Air conditioning: saving lives and accelerating net-zero by Ed Hezlet and Lauren Gilbert explain how U.K policies “all but ban” air conditioning, despite its benefits.

I have discussed heat mortality in several posts here at THB:

• In Public Health and Climate Change, I walked through the Sheridan et al. (2021) data showing U.S. heat mortality falling dramatically across nearly every region over fifty years even as heat waves grew more frequent.

• In Truth Bombs, I showed how a consistent vertical scale reveals that cold still kills far more Europeans than heat — a point that Bjorn Lomborg has made for years.

In today’s post I ask and answer a straightforward question:

I provide my methods, data, and sources in an Excel file at the bottom of the post.

Richest, coolest, deadliest

In his excellent post, Kohler well frames the issue: A continent that enjoys both comparative wealth and, by latitude, fewer hot days than most inhabited regions nonetheless records the world’s highest per capita heat-death rate. Age explains some of it — but the United States and Japan also have aging populations and yet have far fewer heat-related deaths.

A more important difference: air conditioning: European household penetration sits near ~19%, versus ~76% in North America and more than 90% in Japan.

Figure 2. Europe pairs the lowest AC penetration among rich regions with the highest relative heat-death rate. Per-capita heat-death rates from Zhao et al. (2021); penetration from IEA (2023). Framing after Kohler (2025).

The continental average hides a wide spread in heat deaths, death rates and AC pentration, shown in Table 1 below.

Sources: Continental totals reflect the restated series in Ballester / Janoš et al. (2025). Country figures use the original Ballester et al. (2023) breakdown. AC shares: IEA (2023) and national surveys.

The counterfactual

Here I use the results of Barreca and colleagues who tracked the same American states over time, before and after air conditioning spread, with the local climate held constant.1 That approach enables them to conclude that increased AC penetration caused the drop in the risk of dying on an extremely hot day by about three-quarters. I use that 75 percent as my central figure, and also test lower — 65 percent — and higher — 85 percent — values to assess the sensitivity of conlcusions to that value.

The math is simple. Today’s heat deaths reflect today’s level of AC coverage. Raise the coverage, and a share of those deaths are eliminated — in proportion to how protective AC is and how many more households gain it.2

Table 2 and Figure 3 show the results.

Counterfactual scales observed deaths by population-weighted residual risk, assuming AC removes ~75% (of per-household heat-death risk). Applied to a 2022-scale summer.

Figure 3. AC penetration of 40% avoids several thousand deaths; matching North America’s 76% avoids ~26,000; near-universal coverage avoids ~35,000 deaths.

If Europe had AC penetration similar to North-America in the summer of 2022, then the continent would have avoided ~26,000 heat deaths in a summer like 2022 (range ~22,000–~31,000). Near-universal coverage is ~35,000. Even a 40 percent floor — below today’s Spanish or Italian levels — would save something like 6,000 to 8,000 lives a year.

Age matters

Heat deaths concentrate almost entirely among the old. Ballester et al. (2023) find the over-80s alone account for about two-thirds of European heat deaths, and the over-65s for about 93 percent.

The risk climbs steadily with age: Masselot et al. (2023) looked at 854 European cities and measured how much a stretch of extreme heat raises a person's odds of dying compared with a normal day. For a 65-year-old, extreme heat lifts those odds by about 21 percent; for an 85-year-old, by 27 percent. Extreme cold is harsher still, raising the odds by 21 percent at 65 and 36 percent at 85.3

Figure 4 shows both the concentration and the age gradient.

Figure 4. Heat deaths cluster in the oldest age bands, and relative risk climbs with age. Sources: Shares from Ballester et al. (2023); relative-risk values from Masselot et al. (2023), heat at the 99th and cold at the 1st temperature percentile versus the minimum-mortality point.

Based on studies from the U.S. and Canada, the typical European heat victim is an older adult, often over 80, with cardiovascular or respiratory disease, living alone in poorly insulated housing. Most die indoors, at home.

• New York City’s health department, reviewing medical-examiner records of heat-stress deaths, found that 89 percent of decedents had at least one chronic health condition, most commonly cardiovascular disease.

• British Columbia’s coroners service, reviewing the 619 deaths from the 2021 heat dome, found that 98 percent occurred indoors, 67 percent of the dead were 70 or older, 56 percent lived alone, and 93 percent had no air conditioning.

• Barreca and colleagues study across the U.S. found that AC’s protective effect fell most heavily on the very groups most exposed — infants and the elderly.

In Europe, over-65s account for about 93 percent of heat deaths, so a focus on cooling that group first would capture nearly the entire avoidable death toll. Equipping care homes, hospital wards, and the homes of old folks would deliver most of the benefit in Table 2 at a small share of the cost of providing AC everywhere. Europe does not need air conditioning in every home or office to save the many thousands of lives.

Where lives can be saved

The cooler northern countries combine large populations, lower heat-death rates, and very thin AC coverage, so they supply a surprisingly large share of the avoidable deaths — it is not just the Mediterranean. Germany — at 3% household AC, the U.K. — at 5%, and France — at 25%, each show potential for more avoidable deaths than hotter Greece, with ~66% AC penetration.

Figure 5. Avoidable deaths under a US-style 76% scenario. Green marks deaths averted; grey marks the residual. The low-AC north — Germany, France, the UK — supplies much of the gain alongside Italy and Spain.

The countries with leaders that most loudly champion passive cooling and energy restraint — notably Germany, the U.K. and France — have the most potential for saving lives through expanding AC.

The policies that block AC

Europe high heat mortality risk and lower penetration of AC reflect policies that actively deter cooling. Some of these policies are at the EU level, and some are at the national or sub-national level.

In his post, Kohler describes these polices as reflective of an energy-degrowth ideology — one that treats every kilowatt-hour as a vice rather than a virtue for improving human lives:

These regulations have not happened by accident, but they come from an ideology that emphasizes energy degrowth as the only viable solution to climate change. What that means in practice is that Europe has heavily prioritized insulation and passive cooling. In contrast, active cooling through an AC, even if running on clean energy, has been disincentivized because it requires energy.

Table 3 lists the main levers and what reversal would take.

Sources: EU Regulation 2024/573; Energy Performance of Buildings Directive; Spanish and Italian 2022 thermostat decrees. The Geneva permit details (Art. 22B; roughly four housing permits a year) and the France DPE example draw on Kohler (2025), who reports both firsthand.

What about Europe’s commitment to climate goals?

Setting aside the benefits of saving lives, expanded penetration of AC would not necessarily undo the continent’s commitment to achieving net zero.

Air-conditioning demand peaks during the day and evening: during the 2025 heat waves the IEA found that France, with little AC, saw evening electricity peaks about 25 percent above its off-season average, while air-conditioned New York ran 90 percent higher. Higher, but not impossibly so. France already runs a grid that is largely decarbonized, it would simply need more of what it already has. And air conditioning demand matches well with the time of day when solar energy is most readily available.

Consider also that electrifying cars and home heating is projected to raise European electricity demand by ~1,500 terawatt-hours a year by 2050, lifting total demand by some 40 percent — what the EU has formally committed to under Fit-for-55 and REPowerEU. Increasing demand for electricity to power AC in the context of this huge buildout seems very doable.

The larger problem is not technology or cost, but the fact that among many, cooling technologies have taken on a moral framing as a vice.

The bigger issue: adaptation optimism

The issue of AC in Europe illustrate a broader argument that Matthew Burgess, Patrick Brown, Matthew Kahn, and I make in a new preprint, The Economics of Climate Adaptation Optimism.

Climate advocates have long disparaged adaptation — they cast it as an avoidable cost of failed mitigation, dismiss its effectiveness, and treat it as an obstacle to emissions cuts. We argue that this framing inverts reality.

Adaptation, especially through economic development, drives most climate-sensitive outcomes on the space and time scales that affect people living today. Crop yields, affluence, and death rates from climate hazards have improved across almost all regions over recent decades, despite climate change, because development and adaptation pulled them in beneficial directions faster than any growth in impacts from changes in climate.

Adaptation costs and benefits are typically close in time and space. Mitigation, by contrast, spreads benefits across regions and generations, and can even erode adaptive possibilities. Adaptation and mitigation are not trade-offs. Europe and its approach to AC might just be the poster child for getting this wrong.

Before you go, a small request: If you think AC is cool, please click that “❤️ Like” button. More likes means more readers of THB. Thanks!

Comments, critique, questions — all welcome!

Leave a comment

Bibliography

Ballester, J., Quijal-Zamorano, M., Méndez Turrubiates, R. F., et al. (2023). “Heat-related mortality in Europe during the summer of 2022.” Nature Medicine 29(7), 1857–1866. https://www.nature.com/articles/s41591-023-02419-z

Barreca, A., Clay, K., Deschênes, O., Greenstone, M., & Shapiro, J. S. (2016). “Adapting to Climate Change: The Remarkable Decline in the US Temperature-Mortality Relationship over the Twentieth Century.” Journal of Political Economy 124(1), 105–159. https://www.journals.uchicago.edu/doi/10.1086/684582

Boudry, M. (2026). “How Europe Became the World Champion of Heat Deaths.” Quillette, 24 June. https://quillette.com/2026/06/24/how-europe-became-the-world-champion-of-heat-deaths/

Burgess, M., Brown, P. T., Kahn, M. E., & Pielke, R. Jr. (2026). “The Economics of Climate Adaptation Optimism.” Preprint, SocArXiv. https://osf.io/preprints/socarxiv/m7tqu_v1

Copernicus Climate Change Service / WMO (2025). European State of the Climate 2024 (ESOTC 2024). DOI: 10.24381/14j9-s541. https://climate.copernicus.eu/esotc/2024

European Union (2024). Regulation (EU) 2024/573 on fluorinated greenhouse gases (F-Gas Regulation). Official Journal of the EU. https://climate.ec.europa.eu/eu-action/fluorinated-greenhouse-gases/f-gas-legislation_en

Financial Times (2026). “To cool or not to cool: French politicians draw battle lines over aircon.” FT, 23 June. https://www.ft.com/content/to-cool-or-not-to-cool-french-politicians-draw-battle-lines-over-aircon

Gallo, E., et al. (2024). “Heat-related mortality in Europe during 2023 and the role of adaptation in protecting health.” Nature Medicine 30, 3101–3105. https://www.nature.com/articles/s41591-024-03186-1

García-León, D., Masselot, P., Mistry, M. N., Gasparrini, A., et al. (2024). “Temperature-related mortality burden and projected change in 1368 European regions: a modelling study.” The Lancet Public Health 9, e644–e653. https://www.thelancet.com/journals/lanpub/article/PIIS2468-2667(24)00179-8/fulltext

Hezlet, E., & Gilbert, L. (2025). "Air conditioning: saving lives and accelerating net-zero." British Progress. https://britishprogress.org/reports/air-conditioning-saving-lives-and-accelerating-net-zero

International Energy Agency (2023). “Share of population living in a hot climate, 2022, and penetration of air conditioners, 2000–2022.” IEA, Paris. https://www.iea.org/data-and-statistics/charts/share-of-population-living-in-a-hot-climate-2022-and-penetration-of-air-conditioners-2000-2022

Janoš, T., Quijal-Zamorano, M., Shartova, N., …Ballester, J. (2025). “Heat-related mortality in Europe during 2024 and health emergency forecasting to reduce preventable deaths.” Nature Medicine 31(12), 4065–4074. (Restated 2022–2024 continental totals.) https://www.nature.com/articles/s41591-025-03954-7

Kohler, K. (2025). “Make Europe Cool Again.” Machinocene, 4 July.

Lomborg, B. (2021). “Climate change coverage ignores heavy impact of heat on cold deaths.” USA Today, 25 June. https://eu.usatoday.com/story/opinion/2021/06/25/climate-change-science-media-spread-fears-ignore-possible-positives/7738009002/

Lomborg, B. “The Heresy of Heat and Cold Deaths.” lomborg.com. https://lomborg.com/heresy-heat-and-cold-deaths

Masselot, P., Mistry, M., Vanoli, J., et al. (2023). “Excess mortality attributed to heat and cold: a health impact assessment study in 854 cities in Europe.” The Lancet Planetary Health 7, e271–e281. https://www.thelancet.com/journals/lanplh/article/PIIS2542-5196(23)00023-2/fulltext

Pielke, R. Jr. (2023). “Public Health and Climate Change.” The Honest Broker, 24 April. https://rogerpielkejr.substack.com/p/public-health-and-climate-change

Pielke, R. Jr. (2023). “Truth Bombs.” The Honest Broker, 2 August. https://rogerpielkejr.substack.com/p/truth-bombs

Sheridan, S. C., Dixon, P. G., Kalkstein, A. J., & Allen, M. J. (2021). “Recent Trends in Heat-Related Mortality in the United States: An Update Through 2018.” Weather, Climate, and Society 13(1), 95–106. https://journals.ametsoc.org/view/journals/wcas/13/1/wcas-d-20-0083.1.xml

World Health Organization, Regional Office for Europe (2024). Heat claims more than 175 000 lives annually in the WHO European Region." (Source for the ~30% increase over two decades.) https://www.who.int/europe/news/item/01-08-2024-statement--heat-claims-more-than-175-000-lives-annually-in-the-who-european-region--with-numbers-set-to-soar

World Resources Institute (2025). “A Warming Europe and the Air Conditioning Dilemma.” WRI Insights. https://www.wri.org/insights/europes-heat-and-air-conditioning-dilemma

Zhao, Q., et al. (2021). “Global, regional, and national burden of mortality associated with non-optimal ambient temperatures from 2000 to 2019: a three-stage modelling study.” The Lancet Planetary Health 5, e415–e425. https://www.thelancet.com/journals/lanplh/article/PIIS2542-5196(21)00081-4/fulltext

Read more

Subscribe now

ENSO, U.S. Hurricanes, and Global Disasters — Executive Summary

The El Niño-Southern Oscillation (ENSO) drives the largest year-to-year climate variability on Earth. This deep dive updates my past analyses of the relationship of (a) ENSO and U.S. hurricane damage and (b) ENSO and global disaster losses.

First, more than 25 years ago Chris Landsea and I published a paper showing that that U.S. hurricane damage varies with ENSO: La Niña years in general have larger U.S. hurricane losses than El Niño years. The update shows that the original 1999 finding holds with almost 30 years of additional data. The La Niña-to-El Niño ratio of median normalized CONUS damage was 4.0× in the 1925-1997 period covered by the original paper; in the 1998-2025 period the ratio was 7.5×.

Second, global disaster losses over 1990-2024 show a strong La Niña signal at the aggregate level: median global losses total `$270 billion in La Niña years versus ~$155 billion in El Niño years (2024 dollars). U.S. hurricane losses likely dominate the global aggregate in any year with a major U.S. landfall. Global weather-disaster death rates (annual median) are ~1.3x greater during La Niña than El Niño, dominated by a handful of catastrophic events.

Third, as of 22 June 2026, NOAA carries an El Niño Advisory: El Niño conditions are present and expected to strengthen into Northern Hemisphere winter 2026-27. Forecasters predict a 63% chance of a “very strong” El Niño (SST anomalies exceeding 2.0°C) developing this fall. The base-rate expectation for the 2026 Atlantic hurricane season favors below-average U.S. losses, but tail risk remains.

The policy implications are clear: ENSO information has economic value for decision-makers who can act on probabilistic knowledge. ENSO shifts where and when the risks of extremes occur. This report quantifies some of the better known consequences of this important mode of climate variability.

The full report can be found below, for paid THB subscribers.

Comments welcomed! I invite your recommendations for future Deep Dive Reports.

Leave a comment

If you like the idea of occasional THB Deep Dive Reports —> Please let me know by clicking that “❤️ Like.” Thanks!

Read more

Subscribe now

With JD Vance meeting with his Iranian counterparts in Switzerland as I write this, I thought it would be a good time to look back at my April 2 analysis of Iran War costs — both to update them and see how my projections have fared thus far. You can find a spreadsheet with source data and updated calculations at the bottom of this post.

The three THB scenarios captured the costs

My April 2 post included three scenarios for Iran War costs looking to the end of June — central, +25%, -25% — and with a caveat: “the +/-25% scenarios are not upper or lower bounds.” The +/-25% was just a guessimate on my part, but it worked out well to date. As you can see in the figure above, the projection envelope encompassed the evolution of summed consumer costs to date.

Let’s next look at the elements that comprise the consumer costs: fuel and fertilizer.

As you can see in the figures below, with respect to gasoline and diesel, the projections did well, though gasoline prices exceeded the +25% upper bound for the entire month of May: the scenario implied a peak average price of $4.28/gallon while the actual peak was $4.56 on May 21. Diesel rose, to a peak of $5.69 in early April (it neared that level again in mid-May) versus $5.87 in the +25% projection.

As of yesterday, June 21, 2026, the AAA national average is $3.938/gallon for regular gasoline and $5.040/gallon for diesel, meaning that prices remained well elevated from pre-war conditions and the costs borne by American consumers will continue to mount.

Through June 19, cumulative extra household fuel spending is ~$543 ($334 gasoline, $209 diesel) — up from $92 estimated through April 1. Projected to June 30 by holding prices flat (as in the original method), the total reaches ~$592.

This is broadly consistent with independent estimates of energy-driven household cost: Moody’s Analytics put it at ~$450 per household as of late May, with gasoline accounting for “roughly half” — in the same range as my $334 gasoline-only figure for that period.

The figure below shows the relationship of weekly WTI crude oil prices and average U.S. gasoline pricess, which has a strong relationship.

On average, $3/gallon gasoline is associated with WTI at ~$60/barrel and $4/gallon with ~$90-95 a barrel.

Fertilizer costs remain elevated

While fuel prices are off their highs since the war started, fertilizer prices have kept rising, as you can see i the figure below. DTN’s survey for the fourth full week of May put anhydrous ammonia at a record $1,118 per short ton — up from $1,035 on April 1 and 29% above the $865 February baseline, capping 14 straight weeks of higher prices — DTN’s longest such run since it began tracking the index in 2008 — before finally easing in late May. Urea has pulled back from its mid-April peak but at $823/ton is still 35% above baseline; urea ammonium nitrate (UAN is also substantially pricier — UAN32 ($585/ton, +26%) and UAN28 ($530/ton, +29%). All eight of DTN’s tracked fertilizers remain higher than they were a year earlier.

An American Farm Bureau Federation survey from mid-April found about 70% of farmers said they couldn’t afford all the fertilizer they needed for spring planting — though exposure split sharply by region. American Farm Bureau Federation president Zippy Duvall noted that most Midwest corn and soybean growers had locked in prices before the war, while a majority of Southern cotton and rice farmers had not.

The April 2 post undershot this one. Working from a rougher method than the fuel calculations, that post’s “current prices held flat” scenario projected cumulative spring nitrogen costs of about $700 million by June 30. Rerunning the original spreadsheet’s own methodology, cumulative extra nitrogen spending had already reached roughly $1.5 billion by June 19, more than double the original’s June 30 estimate with 11 days of the application season still to come. Some of that gap is already realized and some of it reflects farmers like the Midwest corn growers above who locked in prices and won’t feel the increase until 2027.

On a per-household basis this works out to ~$11 per household by June 19, up from ~$1.60 through April 1. That increase is real but is not a direct expense to households, unlike gasoline and diesel. Increased fertilizer costs first impact ~1.9 million U.S. farm operations that purchase nitrogen each spring.

For a corn grower applying 180-200 lb of nitrogen per acre, University of Illinois economists calculate the war-driven price increase adds about $28-31/acre if buying anhydrous ammonia and $49-54/acre if buying urea — a real, current-season cost for an exposed operation. The impact on 132 million U.S. households will arrive later and more diffusely, through the increased prices of corn, soybeans, and the meat and dairy raised on them.

What the war has cost the taxpayer

The original post extrapolated the CSIS day-12 estimate of $16.5 billion forward at the same rate to get roughly $35 billion by April 1. That overshot the government’s own accounting: Pentagon Comptroller Jules Hurst testified the cumulative cost was $25 billion by April 29 — CSIS’s follow-up found the sustained daily rate fell sharply after the first dozen days, from $1.4B/day to roughly $500M/day, as spending front-loaded into the opening strikes. Hurst updated the figure to $29 billion in May 12 testimony, the latest official number, which doesn’t include the cost of repairing damaged bases in Kuwait and Bahrain.

Independent analysts think that’s a significant undercount, and suggest direct costs of $50-100 billion, two to three times the official figure. Harvard’s Linda Bilmes, who built a widely-cited methodology for estimating the Iraq/Afghanistan wars’ long-run costs, has said she’s “certain” the eventual total will reach $1 trillion once veterans’ disability and healthcare costs are folded in — costs the administration does not consider in its tabulation. My April 2 projection looks pretty good, based on published estimates available at the time.

Bottom line: what does this add up to per household?

Converting the federal cost to a per-household basis and adding that to the $555 per household in fuel and fertilizer costs tracked above, and the combined per-household bill through mid-June runs from ~$775 (using the Pentagon’s own number) to ~$1,313 (using the high end of independent estimates) — versus $92 through April 1 in the original post.

For a rough cross-check: Moody’s Analytics estimated the combined cost to U.S. consumers and taxpayers at about $132 billion as of mid-June — about $1,000 per household on the same basis — which is fully consistent with my analysis, despite using a different (top-down) methodology.

Ongoing, increasing costs of the Iran war borne by the American consumer do not require a restart of kinetic war: Shipping companies and insurers still have to rebuild the trade volumes and confidence lost over four months of closures and blockade — a process GasBuddy analysts don't expect to finish before 2027, even if the current round of talks succeeds. Veterans will draw care and disability payments for decades, farmers will continue to pay higher prices for fertilizer, the federal government will continue to incur costs of the conflict.

Together, the costs of the Iran war documented here should be viewed as an opening tally that is still accumulating. The Iran war will have economic consequences for American consumers, farmers, and taxpayers that will long outlast whatever date future historians settle on as the war's end.

Comments and critique are welcome!

Leave a comment

The supporting spreadsheet — full daily price series, scenario-evaluation tables, and sourcing for every figure — is available below to all subscribers. This is work in progress.

Caveat Lector!

Iran War Cost Data Update 21jun2026

63.3KB ∙ XLSX file

Download

Download

Methods Appendix (updated June 2026)

Timeline. April 8: two-week ceasefire announced. April 13: U.S. imposes a naval blockade on Iran after the Islamabad talks fail. Prices peak the third week of May. June 17: The U.S. and Iran sign a memorandum of understanding. June 18: full text released; gasoline falls below $4/gallon for the first time since March.

Fuel — data and sourcing. Daily gasoline and diesel prices from April 2 through June 19 are reconstructed by linear interpolation between sourced checkpoints. Gasoline checkpoints: AAA, April 9 ($4.16); AAA, April 16 ($4.09); Rigzone/AAA, April 17 ($4.076); US News/AAA, May 1 ($4.39); Al Jazeera, May 12 ($4.50); AAA, May 21 peak ($4.56); EIA weekly update, June 1/8/15 ($4.305/$4.146/$4.052); AAA, June 18 ($3.999, held flat through June 19, the latest confirmed reading). Diesel checkpoints: TIME/AAA, April 6 ($5.62); Overdrive/EIA, week of April 6 peak (~$5.64); WRAL/AAA, ~May 1 ($5.57); Rigzone/AAA, May 21 (yesterday $5.652, week ago $5.667, month ago $5.511); EIA weekly update, June 1/8/15 ($5.350/$5.210/$5.059). The companion spreadsheet flags every daily value as “Confirmed” (a sourced checkpoint), “Held flat (latest verified),” or “Interpolated.”

Cumulative extra cost uses the original formula — (price − Feb 27 baseline) × daily consumption — at a constant consumption rate matching the rate implied by the original spreadsheet’s own Feb 27–April 1 data (≈8.70 Mbpd gasoline, ≈3.90 Mbpd diesel), consistent with EIA’s reported range of 360-380 million gallons/day for U.S. gasoline demand. No seasonal summer-driving increase is layered on top, which likely makes this a conservative (slightly low) estimate for May-June relative to the original’s own stated seasonal range.

Fertilizer — data and sourcing. Weekly purchase-volume assumptions (urea 226,452 short tons/week; anhydrous 108,890; UAN28 159,449; UAN32 186,015) were back-solved to exactly reproduce the original spreadsheet’s April 1 cumulative figure and are otherwise unchanged from the original post (USDA/IFA spring application volume of ~270,000 metric tons N/week, same product-mix shares). Post-April-1 price checkpoints are from DTN’s weekly retail fertilizer survey: April 15 (urea $847, anhydrous $1,088, UAN28 $513, UAN32 $572) and the fourth full week of May, published June 10 (urea $823, anhydrous $1,118, UAN28 $530, UAN32 $585) — the most current DTN data available, held flat through June 19 pending a more recent reading.

Federal cost. Official figures are Pentagon Comptroller Jules Hurst’s testimony to the House and Senate Armed Services Committees, April 29 and the House Appropriations defense subcommittee, May 12. CSIS’s cost-estimate updates, reported via The Intercept, supply the daily-rate breakdown. Independent estimates and Bilmes quotes are from ms.now/MSNBC, June 19 and Common Dreams, April 15. The Moody’s Analytics cross-check figure is reported in NPR’s June 17 retrospective. Household conversions throughout use 132 million U.S. households, unchanged from the original post.

This week I participated in an event at AEI — How Will the US Unleash Commercial Nuclear Power? — which you can view here. In parallel, I am publishing a three-part series on U.S. nuclear power. You can find Part 1 here and Part 2 here. Today, in the final installment of the series, I review some of the policy proposals that leading analysts of U.S. federal nuclear policy have proposed to address the issues that I discuss in the series. At the bottom of this post, for THB paid subscribers, I share a substantial XLS file with all data, sources, and links of the series. As always, this is work in progress — Caveat Lector!

Subscribe now

Part 1 of this THB series argued via a cross-national comparison that nuclear cost escalation does not follow inevitably from the technology. Part 2 argued that U.S. cost escalation traces to federal policy choices that had both intended and unintended consequences.

This post reviews options from nuclear policy experts consistent with those posts — drawn from academic research, government reports, and policy organizations. Source citations accompany each option. The literature on nuclear energy policy is vast, and this review is far from comprehensive. Let me know in the comments what good stuff I’ve missed and you would recommend to add constructively to the conversation. Honest brokering is a group activity!

Political Landscape

Currently, nuclear energy is enjoying favorable bipartisan policy tailwinds — a rarity in 2026. For example, in 2024, the ADVANCE Act (Pub. L. 118-67) passed the Senate 88–2 and the House 393–13. Executive Order 14300 (May 2025) directed the Nuclear Regulatory Commission (NRC) to accelerate licensing and reform risk-based regulation. Microsoft, Amazon, and Google have signed nuclear power purchase agreements. The American Nuclear Society has documented 350+ nuclear bills introduced and 60+ enacted across 45 states in 2025.

These tailwinds reflect an alignment of interests: Republicans cite grid resilience and strategic competition with China; Democrats seek a proven 24/7 carbon-free source at scale; industrial policy advocates in both parties see a domestic manufacturing opportunity; and technology companies, facing growing power demand, see nuclear as a natural fit

Combined with growing public support for nuclear, the moment may be ripe for Congress to consider nuclear energy policy more comprehensively than it has in a half century. This post reviews some of the policy options that Congress might consider, drawn from the recommendations in the “Source basis” boxes within each section below.1

Institutional Reform: A Joint Committee on Nuclear Energy (JCNE)

Since it was terminated in 1977, no formal proposal for a nuclear-specific successor to Congress’ Joint Committee on Atomic Energy (JCAE) has appeared in the congressional record. The closest precedent was when the 9/11 Commission (2004) recommended a joint intelligence committee using the JCAE as its model. Congress declined to implement that recommendation, but the notion of a joint committee as a means of empowering Congress to take a comprehensive view of a specific policy area remains a powerful tool in the legislative organizational toolbox.

One option in support of a nuclear forward agenda: Congress creates a new Joint Committee on Nuclear Energy (JCNE) — say, nine senators and nine representatives, evenly divided between the parties — holding authorizing and oversight jurisdiction over civilian nuclear energy policy. As Part 2 discussed, a tension between promotion and regulation contributed to the 1977 abolition of the JCAE. In contrast, a new JCNE would carry purely legislative and oversight authority.

A new JCNE might be given explicit statutory jurisdiction over NEPA as applied to NRC-licensed facilities. No single committee in either chamber currently holds this authority. NEPA jurisdiction is distributed across Environment and Public Works, Energy and Commerce, and others, with none carrying a specific mandate for nuclear — the legacy of a 1971 court ruling that Congress has not revisited. A JCNE would provides a legislative institutional home for the more specific policy options discussed below nd do so without having so many cooks in the kitchen.

Let’s next take a close look at policy options recommended by nuclear policy experts that would address the issues discussed in Parts 1 and 2.

Near-Term Options (1–3 Years)

Reform 1: Enforce ADVANCE Act Licensing Timelines

The ADVANCE Act set 18-month review windows for standard light-water reactors and 24 months for first-of-a-kind designs. The NRC has typically treated such timelines as aspirational; the Congressional Research Service (CRS) documented a six-year average for combined license applications.

Option: amend the Atomic Energy Act to make timelines enforceable, with a deemed-approved mechanism — subject to explicit safety findings — if the NRC misses statutory deadlines.

Expected impact: $300-$800/kW in avoided financing costs at 7% (details for all estimates can be found in the XLS at the bottom of the post). Timeline: 2-3 years post-enactment.

Reform 2: Codify a Nuclear-Specific NEPA Pathway

As discussed in Part 2, the Environmental Impact Statement (EIS) required for each nuclear license traces to Calvert Cliffs (1971), which since then has never been subject to a congressional vote. The Nuclear Innovation Alliance identifies NEPA requirements as among the highest-priority reform targets.

Option: amend NEPA to allow a completed Generic EIS for a certified standard design to satisfy the “detailed statement” requirement at qualifying sites, with a site-specific supplement for seismic, hydrological, and population factors. Combined with Reform 6 below, this could reduce this aspect of review by more than two years — resulting in approximately $600–900/kW in avoided interest costs.

Reform 3: Codify Design Finality

Lovering et al. (2016) document that post-Three Mile Island backfitting produced a 2.8× cost premium on plants then under construction. Executive Order 14300 directed the NRC to set stringent mid-construction thresholds — legislation would codify that change beyond any single administration.

Option: amend the Atomic Energy Act to require an “imminent and substantial hazard” finding before imposing safety-significant design changes once construction begins.

Expected impact: eliminates mid-construction backfitting risk, saving approximately $680–1,130/kW over a ten-year construction period.

Reform 4: Expand DOE Loan Guarantees to $100 Billion

The single biggest reason nuclear power costs so much to build is not steel and concrete — it's the cost of borrowing. A nuclear plant takes ten or more years to build before it generates a single dollar of revenue, and investors who wait that long demand a high return to compensate for the risk. The OECD's Nuclear Energy Agency found that financing comprises the largest component of the overall costs of building new nuclear. The federal government already has a tool to address this: Title XVII of the Energy Policy Act of 2005 lets the Department of Energy guarantee nuclear construction loans — the government co-signs, so lenders charge lower interest — but the current ceiling is $18.5 billion, and DOE's risk calculations are skewed by basing them heavily on Vogtle's cost overruns and the cancelled 1980s projects rather than the far lower failure rates oberved in South Korea, China, and the early years of France’s nuclear build out.

One option: raise the ceiling to $100 billion and require DOE to weigh the full international track record, which would bring interest rates on nuclear financing down from the current 8–10% to something closer to 5–6% — reducing the cost of nuclear electricity by roughly $20–30 per megawatt-hour, enough in most U.S. markets to make new nuclear more competitive with natural gas without any other subsidy.

Reform 5: Update Regulatory Delay Insurance

Building a nuclear plant means betting that regulators won't cause delays that escalate costs — a real risk, since NRC review timelines have run long for decades. Back in 2005, Congress created a form of insurance for this risk: if regulatory delays cost a developer money, the federal government would cover part of the loss, codified at 42 U.S.C. § 16014. But Congress only made this insurance available to the first six advanced reactor projects, and that list hasn't been updated in twenty years — so any newer project is out of luck.

The Nuclear Innovation Alliance recommends fixing this by opening the insurance to every new reactor design, dropping the extra paperwork step where DOE has to approve each project's eligibility, and making sure enough money is actually set aside to cover the full pipeline of projects that might need it. Combined with making NRC's own licensing deadlines binding (Reform 1), this option sends a clear signal to investors: the government will back up its promise of regulatory predictability with real money if it falls short — exactly the kind of assurance that makes private lenders willing to finance a nuclear project in the first place.

Medium-Term Options (3–10 Years)

Reform 6: Volume Licensing for Standardized Designs

Under current policy, even if a reactor design has already been approved once, building it again at a new site means going through nearly the same lengthy review from scratch. The ADVANCE Act created a fix for this called volume licensing, which lets a single approved design get reused across multiple sites without repeating the full review each time, though the detailed rules for how this will actually work aren't expected until late 2026. South Korea built 16 nearly identical reactors this way, and the Department of Energy estimates that once a company builds the fifth through tenth copies of a standardized design, each one costs 25–40% less than the first ones.

One option: require regulators to finish the detailed volume-licensing rules within 12 months instead of leaving the timeline open-ended, and presume that any project built within that design's already-approved footprint can skip a brand-new environmental review if one has already been done for that design.

Reform 7: A Federal Government Orderbook (5–10 Units)

New nuclear faces a supply chain chicken-and-egg problem: A reactor depends upon giant forged steel components, specialized pumps, and control systems that require manufacturers to invest millions in equipment and expertise. But manufacturers won't make that investment without firm orders lined up, and utilities won't place firm orders without knowing the parts will actually be available and affordable. The Department of Energy has identified nuclear reactor supply chains as one of the biggest obstacles to building plants at scale. Idaho National Laboratory estimates that locking in orders for just five reactors of the same design could cut costs per reactor by 15–25%, simply by giving manufacturers the certainty to invest.

One option: have the federal government become that first reliable customer, committing to buy nuclear power for its own facilities — national laboratories, military bases, data centers running federal systems — across 5 to 10 reactors of a single design. In the private sector, tech companies like Microsoft, Amazon, and Google have already signed long-term deals to buy nuclear power for their own data centers, proving that a big, reliable buyer can give the industry the confidence to scale up. A federal commitment would work alongside those private deals, not replace them.

Reform 8: Require Front-End Engineering Maturity (70–90%)

One of the reasons nuclear projects experience cost escalation is that construction often starts before the design is actually finished. Researchers studying cost overruns and the Department of Energy both point to this issue: at the Vogtle plant in Georgia, crews building the physical pieces of the reactor were working from blueprints that engineers kept revising even after construction had begun. The result was that 40–80% of the components had to be reworked or rebuilt to match updated specs — contributing to why the project ran behind schedule.

One option to fix this going forward: require that a nuclear project's engineering be 70–90% complete before regulators issue a construction permit, which is already the standard practice in industries like petrochemical plants and LNG terminals that build similarly complex facilities. Making this the rule for nuclear construction could save an estimated $1,000–2,500 per kilowatt of capacity and shave two to four years off the typical construction timeline — simply by making sure plans are essentially complete before breaking ground.

Reform 9: National Nuclear Construction Center

Choi et al. (2009) document how KOPEC institutionalized construction knowledge across consecutive South Korean reactor programs. The 30,000 workers trained during Vogtle’s construction scattered when the project closed; the next project team starts largely from scratch.

Option: fund a DOE construction center — $100–200 million annually — at a national laboratory to maintain and distribute construction playbooks, QA protocols, module specifications, and schedule management tools. The Vogtle 3-to-4 improvement (approximately 30% efficiency gain, 20% cost reduction within a two-unit program) illustrates cost reductions through learning.

Long-Term Options (10+ Years)

Reform 10: Domestic Nuclear Manufacturing Investment

Liu et al. (2025) trace China’s cost reduction from approximately $6,150/kW to $1,540/kW primarily to domesticating supply chains.

Option: authorize a Defense Production Act-style manufacturing investment program ($3–5 billion over ten years) for domestic firms building nuclear-grade manufacturing capacity. Expected impact: 10–20% unit construction cost reduction for a mature fleet along with reduced supply chain geopolitical risk.

Reform 11: Contracts for Difference

A nuclear plant operates for 60 years, but nobody can predict what electricity will sell for that far into the future — and that uncertainty makes investors nervous about putting up the billions of dollars a plant costs to build. The OECD's Nuclear Energy Agency has found that one of the most effective ways governments make nuclear investment attractive is by removing that uncertainty by guaranteeing the price of its electricity over the long run. The United Kingdom did this in 2013, setting up a system where the government guarantees a minimum price for nuclear electricity for 35 years, giving investors the certainty they need. The U.S. has something similar but much weaker: the Inflation Reduction Act's tax credit for clean electricity, which only lasts ten years — a fraction of a nuclear plant's lifespan.

One option: authorize federal agencies to offer the same kind of long-term price guarantee the UK uses, for nuclear projects that meet certain cost targets. Paired with expanded loan guarantees (Reform 4), this kind of price certainty could lower the cost of nuclear electricity by an estimated $35–50 per megawatt-hour.

States: A Coalition of the Willing

As discussed in Part 2, sixteen states have laws on the books that specifically restrict or ban new nuclear construction — and trying to override those state laws from Washington would mean lengthy court battles that could drag on for years with no guarantee of winning.

The good news is that none of the federal options above require every state to get on board; a coalition of the willing states would be sufficient. That coalition is already forming: six states have repealed their nuclear bans since 2016 — Wisconsin, Kentucky, Montana, West Virginia, Illinois, and New Jersey — and four more (California, Massachusetts, Minnesota, and Vermont) currently have active efforts underway to do the same.

The federal government can support and accelerate this trend. When DOE chooses where to put its own facilities under the federal orderbook (Reform 7), it can favor states that have already cleared their nuclear barriers. Loan guarantees and long-term price contracts (Reform 4 and Reform 11) can likewise give priority to projects in states with friendlier rules.

Federal progress on a permanent nuclear waste storage site would have an automatic bonus effect: several states, including Minnesota and Oregon, wrote their bans to expire once the federal government solves the waste problem. If Congress adopts a significant nuclear forward agenda, we should expect other states to follow along. Everyone needs more reliable, secure, affordable energy.

Closing Note

Nuclear construction costs in the United States have escalated while other countries’ costs did not escalate to the same degree.. Part 1 showed that this divergence is not a function of the technology — nuclear plants built elsewhere, and built here decades ago, did not carry today’s escalated costs.

In the United States, nuclear energy production remained flat this century while natural gas generation has tripled. Part 2 traced the gap to seven specific federal mechanisms, compounded by an institutional blind spot: when Congress abolished the Joint Committee on Atomic Energy in 1977, it left nuclear policy split across many committees with no single one responsible for a comprehensive view of nuclear energy policy.

This concluding third post answers the question: What might be done?

The options above were not invented for this series — they draw on the recommendations of the experts who study and implement nuclear construction engineering, economics, and policy.2

The DOE Liftoff Report projects that a serial construction program following this path could bring the overnight cost of a tenth U.S. reactor down to $4,000–6,000 per kilowatt — still above what Korea and China have achieved, but competitive with natural gas on a full system basis, particularly considering (possible) carbon costs and fossil fuel price volatility.

Of course, a nuclear forward agenda would represent a political commitment requiring a coalition sustained for decades across two parties, multiple Congresses, and presidencies. Should U.S. leaders choose to organize in support of a nuclear forward agenda, many policy options are ready and available. Build baby build!

Subscribe now

Before you go, one request — If you learned something in this series, let me know by clicking that “❤️ Like” button. Your feedback matters. Thanks!

Comments, critique, other views are all welcome!

Leave a comment

This series is the result of a substantial research effort over the past few months. All data underlying the figures, full references to papers, reports, legislation, and grey literature, and the bases and calculations for every estimate can be downloaded in the 13-tab XLS file available below to THB's paid subscribers.

Read more