Climate & Energy · Saturday, 22 August 2026

01 · Briefing · what happened

Washington wants fewer futures on the table. New work says each one runs hotter.

Climate & Energy 7 min 23 sources

The US moved to strike its hottest emissions scenario from federal planning, in the same week a study found each unit of carbon may warm the planet about a quarter more than assumed.

2.25C

warming per doubling of carbon dioxide

new study's central estimate, against the IPCC's 1.8C

360 TWh

clean power China turned away

in six months, enough to run Mexico for a year

47%

rise in Europe's hydro plus solar

since the 2022 drought summer, same ten systems

11 points

how far UK MPs underestimated public support

across four climate policies, in a survey of 100

At a glance

The US government proposed striking its hottest emissions scenario from federal planning, calling results built on it implausible.
The change would reach the assessment's figures, captions and regional summaries, plus federal rules, adaptation plans and vulnerability assessments.
The document agrees the hot case was drawn to test models, not to predict; the fight is over whether a test case is still worth planning against.
It is a proposal, not a rule, and it declines to endorse the replacement scenarios being built for the next IPCC report either.
A separate study, judging models against satellite heat readings, put warming per doubling of carbon dioxide at 2.25C against the IPCC's 1.8C.
Europe's hydro output fell to its lowest summer on record; solar covered the gap and lifted the combined total 47% above 2022.
Grid connections, not demand, are the brake: China turned away 360 terawatt-hours of clean power in six months, up 49%.
Land-use emissions have fallen about 23% since the late 1990s, while fossil-fuel and cement emissions hit a record 38.1bn tonnes.

Forces in play

Federal science rules Building↑

Washington moved to bar its hottest emissions scenario from federal planning, and declined to bless the replacements either.

Evidence of heat High↑

Record European sea temperatures were judged virtually impossible without human warming, and Lake Powell hit a record low of 1,072.87 metres.

Connection backlog High→

China turned away 360 terawatt-hours of clean power in six months; India's new tenders fell from 40.6 GW to 4.7 GW as wires lagged.

Clean supply relief Easing↑

Europe's solar produced 17% more than usual through the heatwaves and covered the lowest hydro summer on record.

In play US Office of Science and Technology Policy — proposed striking the hot scenario from federal planning Climate modellers — reweighted models on satellite heat data, raising the warming estimate PJM — largest US grid operator, retrying a faster route onto the grid Rural electric co-ops — pressing to repeal the gas power plant emissions rule

How it unfolded

Originally the hot pathway is drawn to mark the top of a range for climate-model experiments
Since its results spread through federal assessments as ordinary planning futures
This week Washington proposes ruling those results implausible for planning
Same week a study puts warming per doubling of carbon dioxide about a quarter higher
Next whether the amendment is adopted, and whether federal documents actually change

Where this points

Watch whether the amendment is adopted as written and then shows up downstream - in flood maps, adaptation plans and cost-benefit analyses - or whether it stays a paper determination that changes no federal document.

Full briefing

The scenario Washington wants off the planning table

The US government has proposed amending its own Fifth National Climate Assessment [1]. The change would rule that results built on the highest emissions pathways must not be used as expected, baseline, business-as-usual or policy-relevant futures [1]. Those pathways, labelled RCP8.5 and SSP5-8.5, describe a world of sustained coal growth and heavy fossil dependence. The amendment calls their results “implausible scenario outputs” [1].

The reach is wide. It covers the assessment’s figures, tables, captions, maps and regional summaries [1]. It also covers federal regulatory analysis, vulnerability assessments and adaptation planning [1]. Any affected result would have to carry an implausibility caveat in the same sentence as the impact it reports [1]. The stated basis is Executive Order 14303, “Restoring Gold Standard Science”, which tells agencies to be transparent about how likely a scenario is [1].

Two honest caveats. This is a proposal, not a finished rule [1]. And it is not a preference for newer scenarios. The document also declines to bless the replacements being built for the next report of the IPCC, the UN’s climate science panel [1]. Every pathway must be judged on its own evidence, it says [1].

The document also concedes the point its critics will press. High-forcing scenarios “may remain useful for model intercomparison, sensitivity testing, and exploratory research”, it says - just not as expectations [1]. That is the whole argument in one line. Nobody disputes the hot case was drawn as a test. The fight is over whether a test case is still worth planning against.

The other half of the sum got bigger

Warming is a product of two things. How much carbon we add, and how much the planet heats per unit added. Washington moved on the first. New work moved the second, the other way.

A study reported this week reassessed which climate models to trust [2]. Rather than judge models on past surface temperatures, the team compared them against satellite readings of heat entering and leaving the atmosphere [2]. That is the more fundamental measure. The models that matched it best project more warming. Their estimate for a doubling of carbon dioxide is 1.9C to 2.6C, centred on 2.25C [2]. The IPCC’s last report gave 1.2C to 2.4C, centred on 1.8C [2]. Roughly a quarter higher. The lead researcher says the result needs confirming from other sources before anyone is confident [2]. Treat it as one strong claim, not a settled number.

The observed record kept adding this week too. World Weather Attribution, a group that assesses warming’s role in specific events using peer-reviewed methods, found human-caused warming was the main driver of record European sea temperatures [3]. It puts about 2C of added warming in the Mediterranean and 1.4C around western Europe [3]. Peak temperatures there and along the Bay of Biscay would have been “virtually impossible” without it [3]. Marine heatwaves covered 90% of the Bay of Biscay and Iberian coast, and 80% of the western Mediterranean [4]. Without human warming, those areas would be roughly half as large [4].

On land, Lake Powell hit a record low of 1,072.87 metres on 15 August [6]. It has not filled since the 1980s [6]. About 40 million people draw some or all of their water from the Colorado River [6]. Separately, Stockholm University’s Gustaf Hugelius warned that heatwaves and wildfires are pushing permafrost regions nearer irreversible thresholds [5]. That frozen ground holds around 1,500 gigatons of carbon, three times all living vegetation on Earth [5].

What is actually getting built

Europe’s summer showed the argument in miniature. Hydro output across ten reporting systems fell to 31.65 TWh between 1 June and 15 August [7]. That is the lowest of seven summers, and 6% under the drought year of 2022 [7]. Solar covered the gap and more. Across the same systems it rose from 45.27 TWh in 2022 to 84.44 TWh this year [7]. Hydro and solar combined were up 47% [7]. The panels were not built because the rivers ran low; the two simply arrived together [7]. During the heatwaves themselves, Europe’s solar produced 17% more than usual, because heat brings clear skies [8].

Batteries carried that power into the evening [8]. Europe installed 36 GWh in 2025, up 48% on the year before [8][9]. Germany led with 6.6 GWh, Britain came second with 5.2 GWh, and Ukraine entered the top five at 2.9 GWh, about five times its previous year [9]. A home battery elsewhere in Europe trims a bill, Ukraine’s solar association chairman noted; in Ukraine it keeps the lights on [9].

The connection queue, not the appetite, is now the brake. India has over 150 GW of renewables under construction, yet new tendering has collapsed [10]. It awarded 40.6 GW in 2024-25, then 14.7 GW, then just 4.7 GW by 10 August this year [10]. Wires are the reason. At affected substations, projects on temporary connections lose 30-50% of their output during the sunny hours [10]. China turned away 360 TWh of clean power in the first half of 2026, up 49% on the year - enough to run Mexico for a year [11]. Analysts call the waste structural, not a passing squeeze [11].

America is trying process fixes. PJM, the largest US grid operator, is trying again to open a faster route onto the grid [12]. A new generator could plug into the spare room on an existing plant’s connection, skipping a full study [12]. Its first attempt drew eight applications and two approvals since 2023 [12]. Duke Energy filed a plan for 18.5 GW of new solar, 4.5 GW of nuclear and 14 GW of gas across the Carolinas by 2041 [13]. And despite federal hostility, this year’s US additions are still mostly solar and storage [14]. Solar was the largest single-year addition of any generating technology worldwide last year, on IEA figures [22].

Rules pulled three ways

Rural electric co-ops publicly pressed the administration to repeal the rule limiting emissions from gas power plants [15]. Their argument is data centres. A 1,000 MW site draws close to that around the clock, and carbon capture, they say, is not commercially ready [15]. Pulling the other way, an appeals court upheld the EPA’s designation of two “forever chemicals” as hazardous, keeping their makers liable for future cleanup [16]. And the Science Based Targets initiative, which certifies corporate climate goals, published version 2.0 of its net-zero standard [17]. It removes the option of buying offsets in bulk just before a target date [17]. Reaction split between “more practical” and “weaker” [17].

In Britain, Carbon Brief factchecked a Conservative report arguing that cheaper power means more gas and nuclear and fewer renewables [18]. Its finding: the report’s own alternative pathway ends with less electrification and an extra 524m tonnes of carbon dioxide by 2050 [18].

What the deciders cannot see

Three findings this week point the same way. A survey of 100 sitting UK MPs found they underestimated public support for climate policies by an average of 11 percentage points [19]. They also overestimated how divided voters are. MPs who thought support was lower were less willing to speak for a pricing policy. Showing them the real polling had “limited and uncertain” corrective effects [19].

In Ottawa, 317 people chose from real house listings. The homes they picked emitted a median 6.93 tonnes a year, against 1.6 tonnes for the lowest-emission options in front of them [20]. Greener homes were chosen no more often than chance. Afterwards, 54% said energy efficiency played no part [20]. The listings never showed it.

London published results from its expanded low-emission zone, where older vehicles pay 12.50 pounds a day to drive [21]. Nitrogen dioxide is down 27% in outer London and 54% in the centre; fine particles are down 31% [21]. In deprived areas, the number of people exposed to illegal pollution levels fell 80% [21].

The number that quietly improved

Emissions from land-use change - deforestation, peat loss, forest degradation - averaged 5bn tonnes of carbon dioxide a year over 2015-24 [23]. That is about 23% below the 1995-2004 average [23]. Preliminary data puts 2025 lower still, at 4.1bn tonnes [23]. Fossil-fuel and cement emissions went the other way, reaching a record 38.1bn tonnes [23]. One of those numbers is genuinely bending. It is not the big one.

02 · Lesson · why it matters

Built as a test, argued about as a prediction

Every field draws a deliberately extreme case to test its designs against, and the people who inherit it mistake the boundary for a forecast.

How it works

A field needs to test its designs against something extreme
So it draws a deliberate boundary case
The case gets published; the reason it was drawn does not travel with it
Inheritors read the boundary as an expectation
Critics then attack it for being a poor expectation
Both sides stop using it for the job it does well

The twist

The extreme case was never a guess about the future; it was a measuring stick. Once it leaves the hands that made it, both its users and its critics argue about a prediction nobody made.

Where you've seen this

Crash testing

cars are hit at a set speed nobody expects to drive at, and that number sets the whole design

Bank stress tests

the scenario is drawn deliberately bleak, then reported as what the bank thinks is coming

Building codes

a roof is sized for a wind the town may never see, and the sizing is not a weather forecast

Fire drills

a building is emptied for an event nobody predicts, and the drill still decides where the exits go

The catch

A boundary is cheap to keep, but it is not evidence the world will reach it. Reading a test case as a forecast overstates the odds; deleting it hides them.

Full lesson

A line drawn to be at the edge

Somebody, once, had to decide how hot to make the hottest climate scenario.

They were not guessing at the future. They were building test conditions. If you want to know whether a climate model behaves sensibly, you have to run it across a range wide enough to include the far corners. So the range needs a top. The top was drawn where it was because it marked the upper edge of what the scenario literature contained, not because anyone expected the world to go there.

That is what a measuring stick is. It has a far end so you can measure things near the far end.

The reason does not travel with the number

Here is the part that matters. When that scenario was published, the number travelled. The reason it had been drawn did not.

A number is portable. It fits in a chart, a caption, a headline, a planning document. The sentence explaining what job the number was built for is longer, duller, and gets cut for space. Within a few years the number is being used by people who never met the ones who drew it. They have no way to know it was a test condition, not an expectation.

This is not carelessness. It is what happens to any tool that outlives the room it was made in. The tool arrives; the manual does not.

Then both sides inherit the same mistake

Watch what follows, because it is symmetrical and almost nobody notices.

Some people begin citing the extreme case as what is coming. That is the first error: reading a boundary as a bet. Others, seeing that, attack the number for being an implausible forecast. That is the second error, and it is the same error. You cannot rebut a forecast that was never made.

Both camps are now arguing about a prediction. Neither is using the number for the job it does well, which was never predicting anything. The measuring stick has become a flag, and once it is a flag, the only two available moves are to wave it or to burn it.

The shelf, and who is allowed to put things on it

There is a structure under this, and it is worth seeing plainly.

Nobody chooses a scenario in the moment they make a decision. They pick from a shelf that was stocked earlier, by people they will never meet, whose choices about what belongs there now pose as the natural set of options. Removing something from that shelf is a far more powerful act than arguing against it, because after removal there is nothing left to argue with. It simply is not among the things one considers.

This is true whoever does the removing, and for whatever reason. The shelf is the thing to look at.

You are already downstream of it

Scenario numbers do not stay in academic papers. They set the height of a sea wall and the size of a culvert. They set the wind a roof must hold, the flood line your insurer prices from, the cooling a hospital installs.

You will never see the choice. You will only ever see its output: a premium, a building code, a road that floods or does not. The distance between the room where the range was set and the street where it lands is enormous. Nowhere along it does anyone stop to ask whether the number still means what it meant.

What this does not settle

None of this tells you how hot the world will get. That is a separate question, answered with evidence, and the evidence keeps moving in both directions at once.

What it offers is a smaller and more useful habit: before arguing about a number, ask what job it was built to do. A boundary case is cheap to keep and tells you something real about the edges of a system. It is also not a claim about what is likely. Someone who quotes it as a forecast has overstated it. Someone who deletes it for failing as a forecast has removed a tool that was never failing.

Both have made the same move. It is remarkably easy to make, and from inside either camp it looks like the obvious correction to the other.

03 · Lab · your turn

The Planner's Shelf

Rehearse giving four planning jobs the right scenario line, and feel how the same number is correct for one job and misleading for the rest.

04 · Hope · carry this

A lifetime ago nobody could measure the heat entering and leaving the whole planet. We can now, and the argument has moved on to which range we plan against.