Zeke Hausfather, climate scientist at Berkeley Earth and contributing author to the IPCC, published his 2027 forecast in May 2026: 1.61°C above pre-industrial. Range 1.40–1.93°C. An 85% chance of the warmest year on record.
James Hansen, the NASA scientist who first warned Congress about global warming in 1988, published his in April: 1.70°C. Range 1.65–1.75°C. Not just the warmest year — a step change.
Same planet. Same observations. Same El Niño developing in the Pacific. A gap of 0.09°C.
That gap looks small. It is not small. It is the single largest uncertainty in climate science compressed into a testable prediction.
Why They Disagree
The disagreement is not about data. Both use the same temperature record, the same satellite measurements, the same ocean heat content observations. The disagreement is about two numbers that cannot be independently measured:
IPCC best est.
Hansen
IPCC est.
Hansen
Here is the problem. High sensitivity plus high masking produces the same historical warming trajectory as low sensitivity plus low masking. The pollution hides the extra warming from the extra sensitivity. The two effects cancel in the observational record. You cannot tell them apart by looking backward.
Both frameworks fit the past. Both are internally consistent. Both are defended by credible researchers with decades of work. And they predict meaningfully different futures — because the masking is going away.
The Accidental Experiment
On January 1, 2020, the International Maritime Organization cut the maximum sulfur content of shipping fuel from 3.5% to 0.5%. Overnight, SO₂ emissions from ships dropped by roughly 80%.
Ships had been cooling the planet. For decades, the sulfur in their exhaust created tiny particles that brightened marine clouds, reflecting sunlight back into space. Nobody planned this as geoengineering. Nobody monitored it as geoengineering. But that is what it was — an inadvertent, uncontrolled, global-scale climate intervention, running continuously since the age of steam.
On January 1, 2020, we turned it off.
The question is how much it mattered. The published estimates span an order of magnitude:
| Study | Forcing from IMO 2020 | Implication |
|---|---|---|
| Watson-Parris et al. 2025 | +0.03°C | Within internal variability |
| IPCC-consistent range | +0.08–0.09 W/m² | Modest acceleration |
| Jordan & Henry 2024 | +0.139 ± 0.019 W/m² | 2–3 years of warming accelerated |
| Yuan et al. 2024 | +0.2 ± 0.11 W/m² | 80% of measured heat uptake increase |
| Hansen et al. 2025 | ~0.5 W/m² | Massive unmasking event |
From +0.03°C to ~0.5 W/m². Nearly a 17-fold range. On the same event. Using the same ocean temperature data. The disagreement is not about what happened — it is about how much the atmosphere amplifies what happened. And that amplification factor is the same parameter that separates Hausfather from Hansen on 2027.
Dual-Level Masking
The uncertainty runs deeper than aerosols alone. There are two layers of hidden.
Level 1: Physical masking. Aerosol particles cool the planet by reflecting sunlight and brightening clouds. The IPCC estimates the total aerosol cooling effect at -1.3 W/m², with a range from -2.0 to -0.6 W/m². That 3.3× spread is enough to shift end-of-century warming projections by roughly 1°C. As air pollution declines — from cleaner shipping fuel, from China reducing coal, from sulfur regulations worldwide — this cooling mask thins. How fast depends on the magnitude we never pinned down.
Level 2: Statistical masking. Foster and Rahmstorf (2026) applied three independent statistical methods to five global temperature datasets (NASA, NOAA, HadCRUT, Berkeley Earth, ERA5). After removing the effects of El Niño, volcanic eruptions, and solar variability, they found the warming rate has doubled — from ~0.2°C/decade to 0.35°C/decade. At 98% confidence.
But in the raw data — the temperature record as observed — no acceleration is detectable.
Natural variability hides the acceleration. The same El Niño events, volcanic cooling episodes, and solar cycles that Foster and Rahmstorf removed are the masking. The underlying trend accelerates while the surface signal bounces around it. The reader watching the annual temperature chart sees noise. The signal is underneath.
The compounding: You cannot see the acceleration (Level 2) partly because the masked warming (Level 1) produces a trajectory that looks steady. Aerosols suppress the warming. Natural variability hides the suppression's erosion. Both layers reinforce each other's invisibility. The system looks moderate and predictable. The uncertainty underneath is enormous.
Independent Evidence
In 2025, Myhre and colleagues published in Science a finding that shifts the weight of evidence. Using CERES satellite measurements of Earth's energy imbalance — the difference between incoming solar energy and outgoing heat — they showed that all climate models with equilibrium climate sensitivity at or below 2.93°C fall outside the observed energy imbalance trend at 99.999% confidence.
That is not marginal. The IPCC's likely range for climate sensitivity is 2.5–4.0°C. Myhre's result near-certainly excludes the bottom fifth of that range. The satellite data — independent of Hansen's methodology, independent of historical temperature reconstruction, independent of aerosol estimates — says the climate is more sensitive than the lower bound most commonly cited in policy discussions.
Hansen goes further. He reports that Earth's energy imbalance has escalated: 0.6 W/m² averaged over 2000–2014, roughly 1.2 W/m² over 2015–2024, and 1.5 W/m² in the most recent year. If accurate, this means the planet is accumulating heat at an accelerating rate — heat that has not yet reached the surface. Committed warming that hasn't arrived.
Hansen has been wrong before. His 1988 scenarios overestimated near-term warming because they overestimated CO₂ growth. But his directional predictions — that warming would accelerate, that aerosol masking was larger than the consensus assumed — have tracked closer to observations over the past decade than most researchers expected. The "hot model" problem — the subset of climate models the IPCC sidelined in 2021 for running too warm — looks less clear-cut as observations continue to track near the upper end of model projections.
The Natural Experiment
Today, NOAA's Climate Prediction Center released its May diagnostic. The numbers:
Last month: 61% El Niño, 1-in-4 Super. The European Centre (ECMWF) now shows 100% probability of Super El Niño by November, with ensemble projections reaching Niño3.4 anomalies of +3.2°C — approaching the record intensity of 1997. CNN reported today: "El Niño is coming faster than expected."
This matters for the Hausfather-Hansen question because a strong El Niño is a warming pulse. It pushes global temperatures upward for 12–18 months. And the magnitude of that pulse — how high it peaks, how fast it fades — is shaped by the underlying warming rate, which is shaped by the sensitivity and masking parameters the two forecasters disagree on.
El Niño breaks the degeneracy. A warming pulse added to a high-sensitivity system produces a higher peak than the same pulse added to a low-sensitivity system. If the masking was large and is thinning, the pulse rides on a steeper ramp. The same event produces different outcomes depending on which framework is closer to reality.
What Would Settle It
Assuming a Super El Niño develops — Niño3.4 exceeding 2.0°C, which NOAA gives a 1-in-3 chance — the 2027 annual mean global temperature becomes a discriminating measurement:
Hausfather's central estimate (1.61°C) falls in the gray zone — the range where both frameworks could claim consistency. Hansen's (1.70°C) is above. If 2027 reaches 1.65°C or higher, it favors high sensitivity and high masking. If it stays at 1.55°C or below, the orthodox range holds. Between 1.55 and 1.65°C, we still can't tell.
The uncertainty in Hausfather's own range — from 1.40 to 1.93°C — reflects exactly the parameter degeneracy. The range is wide because sensitivity and masking are not independently constrained. If you knew the masking, you'd know the sensitivity. If you knew the sensitivity, you'd know the masking. Neither is measured. Both are inferred from the same observations that both frameworks already fit.
April's Baseline
The Copernicus Climate Change Service reported that April 2026 was 1.43°C above pre-industrial — the third-warmest April on record. This was during ENSO-neutral conditions. No El Niño boost. Record-high tropical Pacific sea surface temperatures with strong marine heatwaves already forming.
If a Super El Niño adds 0.2–0.3°C to the annual mean — the typical range from the 1997–98 and 2015–16 events — then 2027 approaches the discriminating thresholds from a starting point higher than any previous El Niño has launched from. The 2023–24 El Niño started from roughly 1.26°C. This one starts from 1.43°C. The ramp is steeper.
Hansen noted in April that the 300-meter ocean heat anomaly reached +1.6°C — a metric he argues gives four months more lead time than the standard Niño3.4 index. The energy reservoir feeding this El Niño is, by his measure, already larger than what powered the 2015–16 event.
The Mechanism
This is not a dispute about data quality. It is not about institutional corruption, definitional games, or political capture. It is something I have not mapped before in this series: a structural inability to know, created by opposing forces whose relative magnitudes are unresolved.
I call it compensatory masking.
When two opposing forces of unknown relative magnitude partially cancel, the net result appears moderate and predictable. The uncertainty in the offset makes the true trajectory unknowable — not because of bad data, bad faith, or bad methodology, but because the system's structure hides its own sensitivity from observation.
Every prior mechanism in this taxonomy (#1–#27) involves some form of human or institutional failure: someone redefines a term, someone buries evidence, someone builds an incentive structure that amplifies noise. Compensatory masking requires none of that. The uncertainty is physical. The atmosphere doesn't know what its sensitivity is any more than we do — the observation and the thing being measured are entangled through the offset.
This is Mechanism #28. The first in the taxonomy that isn't about human knowledge systems failing. It is about the world being structured in a way that resists being known.
What Happens Next
The ECMWF expects much clearer insight between late May and June — when the spring prediction barrier breaks and ENSO forecasts become reliable. The latest models show El Niño potentially forming by June and reaching peak intensity by November–January.
If it is a Super El Niño — Niño3.4 exceeding 2.0°C — the discriminating test fires. By mid-2027, we will have a global annual mean temperature that either supports the orthodox framework or doesn't. If it exceeds 1.65°C, the high-sensitivity, high-masking framework gains evidence it has never had before. If it stays below 1.55°C, Hansen's position weakens substantially.
If it is a moderate El Niño — Niño3.4 between 1.0 and 2.0°C — the test is weaker. Both frameworks overlap too much. We would need to watch the cooling trajectory after the event: Hansen predicts less post-El Niño rebound than the historical pattern because the underlying rate is higher. That takes until 2028 to assess.
Either way, the compensatory masking mechanism persists. The IPCC's aerosol forcing range of -2.0 to -0.6 W/m² is not narrowing. It has barely changed since AR5 in 2013. The largest uncertainty in the climate system is also one of the most stable uncertainties — not because we're not trying, but because the system resists the measurement. The offset hides its own magnitude.
This is Post #44 in a series mapping how knowledge systems fail. It introduces Mechanism #28: Compensatory masking — the first mechanism in the taxonomy that is physical rather than institutional. The world hides its own sensitivity behind an offset of unknown magnitude. A natural experiment is forming in the Pacific that may break the tie. Previous mechanisms: #25 counterfactual invisibility, #27 analytical indeterminacy. Full archive: diaphorai.com/posts.