Climate Warming Predictions Have Been Revised Downward

Accuracy Assessment: ✅ Largely True

Climate warming predictions for 2100 have shifted substantially downward over the last decade. The high-end “business-as-usual” scenario — RCP8.5 / SSP5-8.5, which assumed a roughly five-fold increase in coal use and a tripling of CO₂ emissions by 2100 — has been formally retired in the next generation of IPCC scenarios (CMIP7) as implausible¹. In 2015, projections under existing policies pointed to around 3.6°C of warming by 2100. As of the November 2025 Climate Action Tracker update, that figure is ~2.6°C². The IEA’s 2024 World Energy Outlook puts current-policy warming at 2.4°C³; the UN Environment Programme’s 2024 Emissions Gap Report puts it at 2.6–3.1°C⁴. The new CMIP7 medium illustrative scenario lands at 2.8°C (5–95% range 2.1–3.7°C)¹.

Two important framing points are needed up front. First, the headline “~1°C reduction since 2015” is partly real progress and partly methodological correction. Hausfather, Peters and Forster — the same authors who flagged RCP8.5 as implausible in 2020 — themselves estimate that only about 0.7°C of the apparent reduction reflects genuine policy and technology progress; the rest reflects revisions to overly high earlier baselines¹. Second, the causal story that climate policy “bent the curve” is contested by serious mainstream-credentialed critics. Roger Pielke Jr (and Burgess, Ritchie, Tol) have published peer-reviewed work arguing RCP8.5 was implausible from construction — built on a coal-revival pathway and population assumptions that were never consistent with available evidence — and that “global decarbonization has not accelerated, as would be necessary for climate policy to have had a discernible effect to date”⁵. The article below presents both readings.

The IPCC’s Sixth Assessment Report (AR6, 2021) also narrowed equilibrium climate sensitivity — how much warming you get from doubling CO₂ — from the long-standing 1.5–4.5°C range used since 1979 to a likely range of 2.5–4°C with a best estimate of 3°C⁶⁷. That cuts off both the most-catastrophic and the most-reassuring tails of older projections.

A ~2.7°C world is not benign. Mora et al. 2017 projected that, depending on emissions, between ~48% and ~74% of the global population would face at least 20 days per year of deadly heat-humidity combinations by 2100; the UN page derives a ~3.25 billion figure for ~3°C warming on this basis⁸. Other published projections at ~3°C include near-total loss of tropical coral reefs at 2°C and beyond, sharp falls in crop yields above 1°C, ~½m of sea-level rise by 2100 with multi-metre commitments on multi-century timescales, and IPCC-assessed risks of “abrupt and irreversible” changes in fragile systems becoming “high” above 3°C⁸. Crucially, the most recent comprehensive tipping-point synthesis (Armstrong McKay et al. 2022, Science) concludes that five tipping points become possible already between 1.5°C and 2°C — including West Antarctic Ice Sheet collapse, low-latitude coral die-off, and abrupt permafrost thaw⁹. This means a 2.7°C trajectory does not just produce the “linear” impacts in the impact tables; it commits the Earth system to multi-millennial sea-level rise, possible AMOC weakening, and possible Amazon dieback as part of cascading feedbacks not captured in 2100 surface-temperature numbers.

It is also necessary to note that global emissions are still rising, not falling. The IEA’s Coal 2025 report finds that global coal demand reached a record 8,805 Mt in 2024 and is projected at a further-record 8,845 Mt in 2025¹⁰. Fossil CO₂ emissions hit a new all-time high in 2025¹¹. “Plateauing at record levels” is a more accurate summary than “peaked and declining.”

Major scientific counter-theories that attribute modern warming primarily to non-CO₂ causes — Svensmark’s cosmic-ray/cloud hypothesis, large solar-irradiance variation, urban heat island contamination of the temperature record — have not held up to direct experimental and observational testing. The CERN CLOUD experiments found cosmic-ray nucleation insufficient to drive modern climate change¹²; satellite-era total solar irradiance has trended slightly downward while temperatures rose¹³; and the Berkeley Earth Surface Temperature project (Wickham et al. 2011) — using a station network five times larger than NOAA’s, with rural sites separated from urban — found that urban heat island contamination of the global trend is “nearly negligible” and slightly negative if anything (rural sites warmed marginally faster than urban)¹⁴. Two further commonly-cited counter-theories — the Happer & Wijngaarden CO₂-saturation argument, and Lindzen’s iris hypothesis — are addressed briefly in Sub-Claim 5.

Key Claims at a Glance

Background: Why the Predictions Changed

Climate projections rest on two pillars: emissions scenarios (how much CO₂ humans will emit) and climate sensitivity (how much the planet warms per unit of CO₂). Both have moved in the last fifteen years — but in opposite directions, with a net downward shift in central 2100 warming.

Two things changed between 2015 and 2025:

1. Coal use plateaued at record levels and clean-energy costs collapsed. RCP8.5 required a five-fold increase in coal use to 2100, with assumptions like converting coal to liquid fuels once oil ran out. Solar PV, batteries and EVs all undercut that trajectory. However, “plateaued” does not mean “fell”: global coal demand reached a new all-time high of 8,805 Mt in 2024 and is projected at a further-record 8,845 Mt in 2025, with IEA forecasting demand to stay near that level through 2030¹⁰. Fossil CO₂ emissions also set a new record in 2025¹¹. The world is no longer tracking RCP8.5 but it is not yet on a declining emissions trajectory either.
2. Climate sensitivity estimates got more confident, not more extreme. IPCC AR6 used multiple independent lines of evidence (instrumental record, paleoclimate, process-level cloud physics, emergent constraints) to converge on a likely range of 2.5–4°C, narrower than the 1.5–4.5°C range that had been used since the 1979 Charney report⁶⁷.

The net effect: central 2100 warming projections under existing policies have fallen by about 1°C in the published numbers since 2015². As noted in the assessment above, Hausfather, Peters and Forster estimate that the genuine policy/technology share of this reduction is closer to ~0.7°C, with the remainder reflecting baseline correction and (per CAT itself) methodological changes such as the extension of China’s emissions pathway¹. Pielke and colleagues go further and argue that almost none of the apparent improvement reflects climate-policy efficacy, because the original baseline was implausible from construction⁵.

Claim Breakdown

1. “The RCP8.5 ‘business-as-usual’ worst case has been retired as implausible”

✅ True — formally retired in CMIP7

RCP8.5 (Representative Concentration Pathway 8.5 W/m² radiative forcing by 2100) and its CMIP6 successor SSP5-8.5 were designed as 90th-percentile baseline scenarios — a deliberately high-end stress test, not a central expectation¹. They became widely cited in the academic literature and the media as “business-as-usual” because they were the scenarios climate modellers ran most extensively. Hausfather and Peters’ 2020 Nature commentary — written by two mainstream climate scientists, not contrarians — argued explicitly that this usage was “misleading” and that outcomes like RCP8.5 had become “increasingly implausible with every passing year as clean energy costs fell and coal use plateaued”¹.

In the new CMIP7 generation of scenarios published in van Vuuren et al. 2026, RCP8.5/SSP5-8.5 has been formally retired. The justification cited by the authors is that the scenario “has become implausible, based on trends in the costs of renewables, the emergence of climate policy and recent emission trends”¹. The replacement “medium illustrative scenario” projects ~2.8°C of warming by 2100 (5–95% range 2.1–3.7°C). A new “high” scenario explores a world in which current policies are rolled back, producing ~3.3°C (range 2.5–4.4°C) — still well below RCP8.5’s nominal ~5°C¹.

This is a substantial revision, but it should not be over-interpreted. Hausfather, Peters and Forster explicitly note that:

• The scenario change reflects technology and policy progress, not the disappearance of climate risk¹.
• The IPCC WGII report found that risks across all five “Reasons for Concern” have risen for a given level of warming since AR5¹ — so a “lower-warming” scenario does not straightforwardly imply “lower damage”.
• Even under the new medium scenario, ~2% of model runs still produce ≥4°C of warming¹.

The causal narrative is genuinely contested. Roger Pielke Jr’s direct response to the CMIP7 retirement argues the opposite of the climate-policy-success framing⁵:

“RCP8.5 does not provide a physically consistent worst case BAU trajectory that warrants continued emphasis in scientific research… RCP8.5, and other extreme scenarios, were never plausible.”

Pielke (with Burgess, Ritchie and Tol) identifies three structural problems baked into RCP8.5 from construction: (i) an implausible coal-revival pathway out of line with available coal reserves and economics⁵; (ii) outdated GDP and population assumptions that have since proven wrong⁵; (iii) overly pessimistic clean-energy cost trajectories⁵. On this reading, the retirement of RCP8.5 is not “policy success” but a long-overdue correction the scenario community resisted for over a decade.

Both readings are defensible. The retirement itself is fact; the cause of the retirement is the contested part.

Verdict: ✅ True (on retirement) — RCP8.5 has been formally retired in CMIP7. The causal narrative — whether due to policy success or always-implausible baseline assumptions — remains 🟡 Contested.

2. “Latest projections under stated policies cluster around 2.4–2.7°C by 2100”

✅ True — four major independent assessments now sit in this range

These projections are independent in methodology — IEA uses an energy-system model, CAT uses an integrated NDC tracker, UNEP synthesises across published models, and CMIP7 uses coupled climate-carbon-cycle modelling — yet they converge in a tight band of roughly 2.4–2.8°C with a higher tail running up to ~3.1–3.5°C if policy slips.

Three important caveats:

1. “Stated policies” is not “implemented policies.” UNEP 2024 finds that, even if every 2030 NDC is fully delivered, global emissions in 2030 would still be roughly 70–85% above the level required for a 1.5°C pathway²⁴. Major emitters (including the G20 collectively) are not on track for their own current 2030 commitments. The 2.4–2.6°C numbers above assume governments deliver on what they have announced — a non-trivial assumption.
2. “Current policies” is not “guaranteed outcome.” All these projections assume policies stay in place and clean-energy costs continue declining. A roll-back (the CMIP7 “high” scenario) bumps projections back up toward ~3.3°C¹.
3. The 1.5°C Paris target is effectively gone. UNEP’s 2024 report estimates a 100% chance of exceeding 1.5°C under continuation of current policies, a 97% chance of exceeding 2°C, and a 37% chance of exceeding 3°C by 2100⁴.

Verdict: ✅ True — current-policy projections cluster around 2.4–2.8°C with a higher tail to ~3.1°C, conditional on the assumption that announced policies are actually implemented.

3. “IPCC AR6 narrowed equilibrium climate sensitivity to 2.5–4°C”

✅ True — confirmed in the AR6 WG1 report itself

Equilibrium climate sensitivity (ECS) is the long-run global mean surface temperature increase per doubling of atmospheric CO₂. It is one of the two most important numbers in climate science (the other being the carbon budget).

From IPCC AR6 WG1, Chapter 7 (Forster et al. 2021), the executive summary states verbatim:

“Based on multiple lines of evidence the best estimate of ECS is 3°C, the likely range is 2.5°C to 4°C, and the very likely range is 2°C to 5°C. It is virtually certain that ECS is larger than 1.5°C.”⁶

This is a meaningful narrowing. Every IPCC report from 1990 (FAR) through 2013 (AR5) used a “likely range” of 1.5–4.5°C for ECS — a range that traces back to the 1979 Charney report, in which the central estimate was 3°C ± 1.5°C arrived at essentially by averaging two competing model estimates and adding subjective uncertainty⁷. AR6 cut both ends of this range:

The narrowing was achieved by combining four independent lines of evidence — process-level understanding (especially of cloud feedbacks), the instrumental record, paleoclimate reconstructions, and “emergent constraints” from model ensembles — which the IPCC noted showed “a high level of agreement”⁶.

Two nuances worth flagging:

• The lower tail moved more than the upper tail. AR6 effectively ruled out ECS below 2°C with high confidence (a “lukewarmer” position popular in some sceptical circles). The upper tail (5°C) is now “medium confidence” rather than being ruled out⁶.
• Some CMIP6 climate models run hotter than the AR6 assessed range. Of 27 CMIP6 models, 10 had ECS exceeding 4.5°C⁷. AR6 explicitly down-weighted these “hot models” when producing its policy-relevant warming projections — a methodological break from prior reports, which leaned more heavily on the model ensemble mean⁷.

Verdict: ✅ True — AR6 narrowed the ECS likely range from 1.5–4.5°C to 2.5–4°C, with a best estimate of 3°C, based on convergence of four independent evidence streams.

4. “A ~2.7°C world entails large, serious, but non-apocalyptic impacts”

✅ Largely True — serious but not civilisation-ending under most physical-impact estimates

Most published impact assessments are framed around discrete temperature thresholds (1.5°C, 2°C, 3°C, 4°C). A ~2.7°C world sits closer to the 3°C threshold. Summarising published projections specifically for 3°C of warming⁸:

Four qualifications matter:

1. These are mostly large but bounded impacts, not human extinction. The UN-cited literature consistently describes 3°C as a world with “devastating” and possibly irreversible regional outcomes (collapse of Greenland ice on multi-millennial timescales; loss of nearly all tropical reefs; mass migration pressure from coastal inundation) — but not as a thermodynamically uninhabitable planet. The “4°C may not be adaptable” warning from the World Bank applies above the current 2.4–2.8°C central range⁸.
2. Adaptation capacity differs sharply by country. Tropical countries — with lower per-capita incomes and more pre-existing heat exposure — bear most of the projected impact. This is not adequately captured by a single global temperature number.
3. The “3.25 billion lethal heat” figure has methodological caveats. It derives from Mora et al. 2017 (Nature Climate Change), which projected ~48% to ~74% of the global population would face ≥20 days/year of deadly heat-humidity combinations by 2100 depending on the emissions scenario. The exact wet-bulb-temperature threshold for “deadly” has been refined and debated since — Vecellio et al. 2023 (PNAS) revised the empirical human-tolerance limit downward based on chamber experiments. The 3.25 bn figure should be read as “billions, not millions” rather than as a precise count.
4. Tipping-point fat-tail risk is not captured by linear-impact tables. The most authoritative recent synthesis — Armstrong McKay et al. 2022, published in Science — identifies 16 potential tipping elements and concludes that five become possible already between 1.5°C and 2°C: West Antarctic Ice Sheet collapse, Greenland Ice Sheet collapse, Labrador Sea / Subpolar Gyre convection collapse, low-latitude coral reef die-off, and abrupt boreal permafrost thaw⁹. At ~2.7°C the probability of triggering these — and pushing several others (East Antarctic subglacial basins, Amazon dieback, AMOC weakening) into the “possible” range — rises substantially. Most cross over from millennial-timescale slow tipping to faster regime changes. The impact tables above do not include the cascading-feedback scenario in which crossing one tipping point (e.g. AMOC weakening) makes another (e.g. Amazon dieback) more likely. Recent (2025) cascade modelling indicates such interactions can produce non-linear amplification of regional damages.

So the direction of impact is severe, the scale is unevenly distributed, and the framing as “civilisational collapse” is not what the mainstream impact literature concludes for ~2.7°C — but the fat-tail risks from tipping-point cascades and feedback amplification mean that the actual Earth-system commitment is meaningfully worse than the central impact projections suggest.

Verdict: ✅ Largely True — a 2.7°C world is projected to cause large, geographically uneven, partially irreversible damage. Central estimates do not support apocalyptic framings, but fat-tail tipping-point risks make “non-apocalyptic” a fragile claim that depends on the cascading-feedback question.

5. “Cosmic-ray, solar-variability, and urban heat island counter-theories do not explain modern warming”

❌ Mostly False — each of the three main scientific counter-theories has been tested and falsified as the primary driver

Three scientific (as opposed to political) alternative explanations for modern warming have been seriously proposed and tested:

5a. The Svensmark cosmic-ray / cloud nucleation hypothesis

Henrik Svensmark (Danish National Space Institute) proposed in 1997 that galactic cosmic rays seed cloud condensation nuclei. The chain of causation is: solar magnetic field weakens → more cosmic rays reach Earth → more low cloud → more reflected sunlight → cooling (and vice-versa for warming)¹².

The CERN CLOUD (Cosmics Leaving OUtdoor Droplets) experiment was specifically built to test this. Dunne et al. 2016 reported, after a decade of CLOUD experiments, that while cosmic-ray ionisation does contribute to aerosol nucleation, “this process is insufficient to attribute the present climate modifications to the fluctuations of the cosmic rays intensity modulated by changes in the solar activity and Earth magnetosphere”¹². The Wikipedia summary, drawing on Lockwood et al. and other independent analyses, notes: “Scientists have generally not found the published work of Svensmark et al. persuasive… The cloud-cosmic ray suggestion increasingly fails to match observations”¹².

The hypothesis is scientifically respectable — Svensmark continues to publish in peer-reviewed journals — but the empirical case for it as a major driver of modern warming has weakened, not strengthened, in the two decades since the CLOUD experiments began.

5b. Large solar irradiance variation

The hypothesis that the Sun, not CO₂, is responsible for modern warming requires either a large positive trend in total solar irradiance (TSI) over recent decades, or an indirect amplification mechanism.

The empirical record cuts directly against this:

• The IPCC AR6 estimates the change in the Sun’s brightness between pre-industrial and 2019 contributed about 0.06 W/m² of radiative forcing — capable of producing roughly 0.01°C of warming (~1% of observed industrial-era warming)¹³.
• A 2024 PNAS study (Coddington et al.) reconciling proxy and satellite TSI measurements found a negative TSI trend of −0.15 W/m² per decade over the satellite era (1980–2023)¹³. Solar input has fallen slightly while temperatures rose substantially.
• For comparison, the effective radiative forcing from CO₂ alone increased by 1.23 W/m² between 1978 and 2023 — twenty times the entire estimated industrial-era solar contribution¹³.

5c. Urban heat island contamination of the temperature record

The argument is that surface temperature stations are increasingly surrounded by built environments, artificially inflating the warming trend. This is associated with US blogger Anthony Watts and the “surfacestations” project.

This is the only one of the three with significant ongoing scientific debate at the local level — UHI effects on individual stations are real and well-documented. But the question of whether they bias the global trend has been independently tested, most decisively by the Berkeley Earth Surface Temperature (BEST) project, an independent study initiated by physicist Richard Muller (previously a public sceptic of mainstream temperature reconstructions) and partly funded by the Koch Foundation. BEST examined a station network roughly five times larger than the NOAA/GHCN-M set (~39,000 vs ~7,000 stations), and the specific UHI test paper by Wickham, Curry, Muller, Rohde and colleagues (2011) compared the global trend computed from all stations to the trend computed from only “very-rural” sites identified using MODIS satellite land-cover data. The result: the urban heat island effect on the global trend is “nearly negligible” and in fact slightly negative — rural-only sites showed marginally more warming than the full dataset (median 1.08°C/century vs 0.98°C/century)¹⁴. NASA, NOAA, the UK Met Office’s HadCRUT, and the Berkeley Earth datasets all agree to within ~0.05°C on the global trend.

5d. Other counter-theories (briefly)

Several other scientific counter-theories have peer-reviewed presence but were not addressed at length above because they propose smaller modifications to mainstream climate science rather than full alternative explanations of post-1950 warming:

• CO₂ saturation / Happer & van Wijngaarden. William Happer (Princeton emeritus) and Wijngaarden have argued from line-by-line HITRAN spectroscopy that the radiative effect of doubling CO₂ is small because key absorption bands are already saturated. The mainstream rebuttal is that “saturation at the surface” does not equal “saturation as a forcing”, because doubling CO₂ shifts the effective emission altitude upward to colder regions of the atmosphere, increasing the imbalance regardless of surface-level saturation. The Happer-Wijngaarden numbers, taken on their face, would imply an ECS of roughly 1.4–1.6°C — below the AR6 “virtually certain” lower bound of 1.5°C. This is a contested but live debate; it has not displaced the AR6 multi-line consensus on ECS.
• Lindzen iris hypothesis. Richard Lindzen (MIT emeritus) proposed in 2001 that high-altitude tropical cirrus cloud cover contracts as the tropics warm, increasing outgoing longwave radiation and providing a strong negative feedback. The original 2001 form was rejected by most cloud-feedback studies. A revised “iris” was partially revived by Mauritsen and Stevens (2015) showing some negative cirrus feedback consistent with observations, but the effect is too small to reduce ECS below the AR6 likely range.
• Multidecadal natural variability (PDO/AMO). The role of ocean oscillations (Pacific Decadal Oscillation, Atlantic Multidecadal Oscillation) in modulating short-term temperature trends is fully mainstream. A stronger “skeptical” version — that these oscillations account for a major share of post-1950 warming — is not supported by the attribution literature, because the underlying forcings (CO₂, solar, aerosols) have been quantified independently and CO₂ dominates the post-1950 budget.

Why these counter-theories matter — and why their failure matters

The point is not that climate science is settled in every detail — IPCC AR6 still gives ~5°C of warming above 3°C ECS only “medium confidence” rather than ruling it out, and cloud feedbacks remain the largest source of model uncertainty⁶. The point is that the specific scientific alternatives to anthropogenic CO₂ as the dominant driver of post-1950 warming have each been directly tested and falsified as primary explanations:

• Cosmic-ray cloud seeding: tested in CLOUD, found insufficient.
• Solar variability: TSI trended slightly down while temperatures rose.
• Urban heat island bias: ruled out as a major contributor to the global trend by BEST.

A coherent scientific counter-explanation would require either resurrecting one of these (against the experimental evidence) or proposing a new mechanism. No such mechanism with comparable empirical support exists in the peer-reviewed literature as of 2026.

Verdict: ❌ Mostly False — each of the three main scientific counter-theories has been directly tested and fails to explain modern warming. CO₂ remains the best-supported explanation for the post-1950 trend.

Summary Table

Overall: ✅ Largely True — Climate warming predictions for 2100 have been revised downward by roughly 1°C in the published numbers since 2015, from ~3.6°C to ~2.6°C under current policies. About 0.7°C of that reduction reflects genuine progress on clean-energy costs and coal use (which has plateaued, but at record-high levels — global emissions are still rising slightly each year). The remainder is partly methodological correction and partly the retirement of an unrealistic high-end baseline that was widely misused as a “business as usual” expectation. Climate sensitivity has also been tightened, ruling out both the lowest and highest tails. A ~2.7°C world still entails large, geographically uneven, partially irreversible damage — and tipping-point cascades (AMOC, Amazon, Greenland) make the actual Earth-system commitment meaningfully worse than the linear impact tables suggest. The major scientific alternatives to anthropogenic CO₂ as the dominant driver have each been directly tested and fail. The honest summary: the worst-case scenarios have been retired, but the central case is still well above the Paris targets, and the 1.5°C goal is effectively gone.

References

Climate Action Tracker — November 2025 update screenshot

IEA World Energy Outlook 2024 executive summary screenshot

UNEP Emissions Gap Report 2024 screenshot

IPCC AR6 WG1 Chapter 7 (ECS) screenshot