The public conversation about climate change operates on a foundation of stated certainty. Ninety-seven per cent of scientists agree. The science is settled. We have decades to act before catastrophe. Trillions of dollars in policy decisions — carbon taxes, energy transitions, industrial restructuring — flow from this stated certainty. But what happens when you actually examine the foundation? Each layer, when interrogated, reveals uncertainty that rarely surfaces in public discussion.
The temperature record
Start with what should be the simplest question. How much has the planet actually warmed? The answer requires comparing temperatures from before instruments were standardised to temperatures measured today with electronic precision. The world’s longest continuous temperature record is the Central England Temperature series, running back to 1659.
Looking at the actual Met Office data, March temperatures in Central England from 1900 to 1907 ranged from 3.7°C in 1900 to 7.1°C in 1903 — a spread of more than three degrees within a single decade. Compare that to 2000 through 2007, when the range was 5.1°C to 7.6°C. The warm end has shifted by roughly half a degree. The cold end has shifted by about 1.4°C. So there is a signal of warming there.
But consider what produced those early numbers. Mercury thermometers, read by hand, often by clergymen or amateur observers at varying times of day. The accuracy of those instruments was at best plus or minus half a degree Celsius, and often worse depending on exposure, placement, and the diligence of the reader. From 1659 to 1772 the record relies on scattered observations of unknown precision. From 1772 to 1876, single-station data with variance that had to be artificially adjusted to make it comparable across the series. From 1876 onwards, multiple stations were averaged together — different stations, different exposures, different microclimates. From 1974, the Met Office applied further corrections to account for urban warming around the recording sites.
Each correction has a justification. But each correction is also a choice. The signal being extracted — roughly 1°C of warming over 150 years — is not much larger than the cumulative uncertainty introduced by those corrections, the original instrument error, and the changes in methodology.
The Kremsmünster Abbey station in Austria is held up as the gold standard. It is an operational weather station that has never been moved from its original location, with records stretching back to 1762. Even here, comparability across time is not given. The old thermometers were not properly shielded from direct sunlight — they were placed on north-facing walls or windows. In summer, the sun warmed them artificially, making readings appear higher than the actual ambient temperature. When researchers compared the historic methods to modern instrument screens using parallel measurements at Kremsmünster, they found the old readings were biased about 0.4°C too warm in summer. That bias was subtracted from the historic readings to produce a comparable record.
Note what that means. Even at the world’s most consistent observation location, you cannot simply read the thermometer from 1880 and the thermometer from 2020 and compare them. The data must be corrected. The corrections are reasonable. They are also necessarily uncertain. And the warming signal sits within the band of that uncertainty.
The solar variable
The basic physics of climate is not complicated at the highest level. Energy arrives from the sun. Some is reflected back to space. Some is absorbed by the surface and atmosphere. The atmosphere holds some of that absorbed energy through greenhouse effects. The balance of these processes determines temperature.
Public discussion of climate change focuses overwhelmingly on the greenhouse component, specifically on human additions of CO2 and methane. The solar component is treated as roughly constant. But it is not.
Solar output varies on multiple timescales. There is the well-known 11-year sunspot cycle. There is a 22-year magnetic cycle. There are longer modulations — the Maunder Minimum in the late 1600s, when sunspots nearly vanished and Europe experienced the depths of the Little Ice Age, is the most famous example. And there are still longer cycles we only partially understand, reconstructed through proxies like radiocarbon in tree rings and beryllium-10 in ice cores.
Direct, precise measurement of total solar irradiance from satellites only began in the late 1970s. Sunspot records become reasonably systematic from around 1850, but the methods of counting and the observers themselves varied, and there are documented disagreements about sunspot numbers in the late 1800s. Before 1850, you are reconstructing solar activity from proxies, each with its own uncertainties about deposition rates, atmospheric mixing, and geomagnetic effects.
Here is the analogy that matters. Imagine a pot of water on a stove with a flame underneath. You are measuring the water temperature over time. The temperature changes. Did the flame change? Did you add insulation around the pot? You cannot answer the question without knowing both variables. Greenhouse gases function like insulation — they slow the rate at which absorbed solar energy escapes back to space. But the heat itself comes from the flame. If you do not know precisely what the flame was doing over the period you are studying, you cannot isolate the insulation’s contribution.
This is where the conversation breaks down. Climate models do incorporate estimates of solar variation. But those estimates carry uncertainties of their own, and the uncertainty in solar input over 150 years is not negligible compared to the temperature signal being explained. To claim with high confidence that human emissions are responsible for most of recent warming, you need to know how much the sun has contributed. The honest answer is that we know less about that than the confidence statements suggest.
There are researchers studying this. Reconstructions of solar activity over the Holocene — the last roughly 12,000 years — have produced varying conclusions. One recent study found no compelling evidence for long-term variations in solar activity beyond the well-known cycles. Other reconstructions suggest more substantial variation. The point is not that one camp is right. The point is that researchers in this specific area express more uncertainty than the public conversation acknowledges.
The Cook 97 per cent study
The figure that anchors the public consensus narrative — 97 per cent of scientists agree — traces primarily to a 2013 paper by John Cook and colleagues. The study is worth examining in detail because the headline diverges sharply from the methodology.
Cook and his team analysed the abstracts of 11,944 papers published between 1991 and 2011 that mentioned global warming or global climate change. They rated each abstract for whether it endorsed the position that humans are causing global warming. The result they publicised was that 97.1 per cent of papers expressing a position endorsed the consensus.
The key qualifier is “expressing a position”. About two-thirds of the papers were rated as taking no clear position on the cause of warming at all. The 97 per cent figure comes from the remaining third. So the actual statement supported by the data is closer to this. Of the minority of climate-related papers that took a clear position on causation, 97 per cent endorsed human causation.
That is a different claim than 97 per cent of climate scientists agree humans cause warming, but the latter is what entered public discourse.
The study also invited authors to self-rate their own papers. About 1,200 of the original authors responded — a small fraction of the total — and 97.2 per cent of those who responded endorsed the consensus. Critics noted that response rates of this size are subject to selection bias. Authors who hold the consensus view may be more likely to respond to a survey from a consensus-focused project.
Economist Richard Tol published a critique arguing the methodology lacked basic quality controls. Researcher José Duarte argued the selection process included papers it should have excluded — papers on social science, public opinion, and policy that were not actually scientific assessments of climate causation. He called for retraction. The journal declined.
Beyond Cook, other consensus surveys exist, with various methodologies and various definitions of “scientist”. Some count anyone with a PhD in any science field. Some restrict to publishing climatologists. The numbers shift considerably depending on the definition. Hans von Storch and Dennis Bray’s repeated surveys of climate scientists, for example, have found high agreement that warming is occurring and that humans contribute, alongside considerably wider variation on climate sensitivity, the magnitude of human contribution, and the reliability of long-range projections.
The pattern is consistent. The more narrowly you define expert and the more specifically you ask about contested aspects of the science, the lower the agreement. The headlines simplify all of this into a number that sounds airtight.
The expertise gradient
There is an observable inverse pattern in confidence about climate science. The general public, primed by media coverage, tends to be either highly confident in the catastrophe narrative or highly dismissive of it. Among scientists, the picture is more nuanced.
The signatories to broad consensus statements often include researchers from adjacent fields — biologists, engineers, physicists, geologists — who have read the IPCC summaries and accepted them. This is not dishonest. It is normal scientific practice to defer to specialists in fields you do not actively research. But it does mean that a consensus of scientists can include thousands of people who have not personally interrogated the temperature corrections, the solar reconstructions, the model assumptions, or the data adjustments.
The researchers who have spent careers in those specific weeds — paleoclimatology, atmospheric physics, instrumental record analysis — tend to express more uncertainty about specifics. Some are very confident about human contribution to recent warming. Some are quite uncertain about magnitudes, feedbacks, and projection accuracy. Some are publicly sceptical of major claims.
Many in the more sceptical camp do not speak publicly. The professional incentives are significant. Funding bodies, university administrations, and peer review panels are dominated by people who accept the mainstream consensus. A researcher who publicly challenges core claims faces real career risks — denied grants, hostile reviews, social ostracism within the field. This does not mean a hidden majority disagrees with the consensus. It does mean the public expression of scientific opinion is filtered by incentives that do not favour visible dissent.
Confirmation bias and funding
Every researcher, in every field, faces choices about how to handle ambiguous data. Which measurements to include. Which outliers to exclude. Which corrections to apply. Which models to use. Which results to highlight in abstracts. None of these choices are mechanical. All of them involve judgment, and all judgment is shaped by prior beliefs.
This is not a unique problem for climate science. It is a feature of all science. But it becomes particularly important when the data is messy, the signal is small relative to the noise, and the political stakes are high. Climate research has all three of these features.
The funding incentive deserves direct attention. Climate research has expanded enormously over the past three decades because governments and foundations consider it urgent. The urgency depends on the conclusion that human emissions are driving dangerous warming. A research programme that concluded current warming is largely natural and will reverse with solar cycles would not attract the same level of funding. This does not mean researchers consciously bias their findings to keep grants flowing. It means the field, in aggregate, faces a structural pull toward findings that support the importance of the field.
This is true of medical research on diseases, defence research on threats, and education research on interventions. It is a normal feature of research economics, not a unique conspiracy. But it does mean the conclusions of any well-funded field deserve interrogation, not deference.
What scepticism does not imply
Examining the cracks in the climate consensus does not lead automatically to denial that warming is occurring. The basic physics of CO2 absorbing and re-emitting infrared radiation is real, demonstrated in laboratories, and not seriously contested. Atmospheric CO2 has risen substantially since pre-industrial times, primarily from human activity. By multiple measurement methods, average temperatures have risen since the 19th century. Ice has retreated in many places. These are real observations, not invented ones.
What is contested, or should be, is the magnitude, the attribution, and the projection. How much of the observed warming is human-caused versus naturally driven? The honest answer involves wider uncertainty bands than public messaging suggests. How sensitive is the climate to additional CO2? Estimates range significantly, and the high-sensitivity scenarios that drive catastrophic projections are not as well-supported as moderate-sensitivity scenarios. How accurate are projections of future warming and impacts? The track record of climate model projections over the past 30 years is mixed, with many running warmer than observed.
The asymmetry in public discussion is the problem. Statements about basic physics get extended seamlessly into statements about catastrophic projections that depend on contested assumptions about feedbacks, sensitivities, and emissions pathways. Each link in the chain is presented with the certainty appropriate to the strongest link.
The policy question
The reason any of this matters beyond academic interest is that policy decisions are being made on the assumption that the scientific case is closed. Trillions of dollars are being directed toward energy transitions, carbon pricing schemes, restrictions on industries, and changes to land use. These are real costs borne by real economies and real people.
The question worth asking is not whether climate change is real. It is whether the certainty being claimed justifies the policy responses being demanded. Policy under uncertainty looks different than policy under certainty. It involves more hedging, more adaptive strategies, more attention to costs of being wrong in either direction, and more openness to evidence that might shift the picture.
The public conversation rarely allows for this. To question the certainty is treated as denying the phenomenon. To raise methodological concerns is treated as bad faith. The discourse has been structured to make legitimate scepticism socially costly, which is exactly what you would expect in a field where the funding and the politics are aligned with one conclusion.
None of what has been examined here is secret. The temperature corrections are documented. The solar reconstruction uncertainties are in the published literature. The Cook study methodology is openly described. The expertise gradient in confidence is observable to anyone who reads the actual scientific debate rather than press releases.
What is notable is how rarely any of it surfaces in public conversation. The narrative communicated to the public is dramatically more certain than the underlying science supports. That gap, between stated certainty and actual certainty, is the legitimate concern, separate from any view about whether warming is happening or what to do about it.
Asking questions about that gap is not denial. It is the basic operation of a critical mind encountering claims that warrant scrutiny precisely because the stakes are so high.
Where do you sit on this — convinced by the consensus, sceptical of the certainty, or somewhere in between? Have a read of the underlying data and tell us in the comments below.
This article was written by AI, briefed to report the facts, hopefully without some of the bias people bring to the job 🙂
Leave a Reply