The Chain of Evidence

Almost everyone holds a settled view about this, and almost nobody has read the papers it rests on. This is the case that the warming is ours, rebuilt from the original literature — link by link, in the order the questions were actually asked, from experiments anyone can check. Where the science is genuinely contested, or where a scientist got it wrong, it says so.

How to read this

The terms of the argument

Most argument about climate is really an argument about authority. One side counts scientists; the other counts the ways a count of scientists can mislead. Nobody in that exchange is discussing anything that was ever measured, and nobody comes out of it knowing more than they went in — including the people who end up agreeing. So there are no consensus figures in what follows. It goes back to the measurements.

The route is slower. It follows the chain of reasoning in the order it was actually built, from 1824 to now, and at each link states the question somebody asked, the measurement they made to answer it, and what that measurement did and did not establish. Most of the links are ordinary laboratory physics of the sort that gets done in an undergraduate practical. Several are experiments you could, given a weekend and some kit, do yourself.

The claim being defended is narrow and specific:

Not that every heatwave is our fault. Not that the models are perfect. Not that any particular policy follows. Just that.

Three promises about how it is argued. First, no bad-faith moves: no strawmanning the opposing case, no quoting a scientist out of context, no pretending a contested thing is settled. Second, wherever a piece of the science is genuinely weak, disputed, or was simply got wrong, it is marked with an orange bar like the ones below, and those boxes are not decoration — several of them contain objections that were correct and that took decades to resolve. Third, every factual claim points to the original paper, numbered, at the end. None of it has to be taken on trust.

1824

Why isn’t the Earth frozen solid?

Joseph Fourier · Paris

Fourier was a mathematician working on how heat moves through solids — the work that gives us Fourier series, still the backbone of signal processing. Around 1824 he turned the same accounting on the planet, and noticed something that did not add up.¹

This is a sum you can do on the back of an envelope, and everything downstream hangs off it. Sunlight arrives at about 1,361 watts per square metre. The Earth intercepts a disc but radiates from a sphere, which divides that by four. About 29 per cent is reflected straight back by cloud, ice and desert. So the planet absorbs roughly 240 watts per square metre.

For the temperature to hold steady, it must radiate that same 240 W/m² back to space. The Stefan–Boltzmann law tells you what temperature a body needs to radiate a given amount, and 240 W/m² corresponds to about 255 kelvin, or −18 °C.

The observed average surface temperature is about 15 °C. There is a 33-degree gap. Something is keeping the surface much warmer than the raw energy balance permits, and there is no third heat source — geothermal heat is about a ten-thousandth of the solar input.

Fourier’s suggestion was that the atmosphere is asymmetric: broadly transparent to incoming sunlight, but not transparent to the heat the warm ground radiates back up. Heat gets in more easily than it gets out. He compared it, loosely, to a box with a glass lid.

1856

Which gases actually stop heat?

Eunice Newton Foote · John Tyndall · New York and London

Fourier left a hole and two people, working separately and with very different resources, started to fill it.

In 1856 Eunice Foote, an American amateur scientist and campaigner for women’s suffrage, filled glass cylinders with different gases, put a thermometer in each, left them in the sun and wrote down which got hottest.² The carbon dioxide cylinder got hotter than the one of ordinary air, and stayed hot longer once moved into the shade. She wrote that an atmosphere with more of that gas would give the Earth a higher temperature — a remarkable inference for 1856, published as a two-page note and then largely forgotten for a century and a half.

Tyndall’s tube

John Tyndall was an Irish physicist at the Royal Institution in London, a mountaineer and glaciologist by inclination, and he wanted to know what stopped glaciers from being permanent. In 1859 he built an instrument specifically to answer the question Fourier had left open.³

The apparatus is simple. A brass tube, about 1.2 metres long, sealed at each end with rock salt (which is transparent to infrared, unlike glass). At one end, a source of heat: a copper cube of boiling water, giving off pure long-wave infrared, no visible light at all. At the other, a thermopile wired into a galvanometer — a device that turns a temperature difference into an electric current and deflects a needle. Tyndall used a differential arrangement, with a second heat source on the other side of the thermopile, so that the needle sat at zero unless the gas in the tube absorbed something. Any absorption showed up as a deflection.

He pumped the tube down to vacuum, took a reading, then admitted one gas at a time. The results were sharp:

• Oxygen, nitrogen, hydrogen and dry air: essentially transparent. The needle barely moved. And these make up 99.9 per cent of the atmosphere.
• Water vapour: strongly absorbing. The single largest contributor.
• Carbon dioxide (which he called carbonic acid): strongly absorbing, far out of proportion to its tiny abundance.
• Methane, nitrous oxide, ozone, coal gas: all strong absorbers.

Tyndall understood immediately what he had found. He wrote that water vapour is a blanket more necessary to English vegetation than clothing is to a person, and that removing it for a single summer night would wipe out every plant a frost could kill. He also grasped that changing the amount of these gases would change the climate, and suggested it as the cause of the ice ages.

Why it works, from first principles

The reason the composition splits so cleanly is molecular, and it is the sort of thing taught in a first-year chemistry course.

A molecule absorbs infrared radiation only if the vibration excited by the photon changes the molecule’s electric dipole moment — its separation of positive and negative charge. Nitrogen (N₂) and oxygen (O₂) are two identical atoms joined symmetrically. Stretch that bond and the molecule remains perfectly symmetrical; the dipole moment stays at zero. There is nothing for the infrared photon to grab hold of. That is why the bulk of the atmosphere is transparent.

Carbon dioxide has three atoms in a line, O=C=O. Its symmetric stretch is invisible to infrared for the same reason. But it has two other modes that are not symmetrical: the bending mode, where the molecule flexes out of line, and the asymmetric stretch, where one bond lengthens as the other shortens. Both create a fluctuating dipole. Both absorb. The bending mode absorbs at a wavelength of about 15 micrometres — and that is precisely where a surface at around 15 °C radiates most of its heat. It is an unlucky coincidence of chemistry that CO₂’s strongest absorption band sits in the middle of the Earth’s outgoing thermal spectrum.

None of this is contested by anyone. It is the basis of the infrared gas analysers used in every hospital anaesthetics machine, every brewery, and the CO₂ sensor in a modern building’s ventilation system. If CO₂ did not absorb infrared strongly and specifically, those instruments would not work.

1896

How much warming, per doubling?

Svante Arrhenius · Arvid Högbom · Stockholm

Knowing that CO₂ absorbs infrared tells you the direction of the effect but not its size. Arrhenius, a Swedish physical chemist who would later win the Nobel Prize for his work on electrolytes, set out to put a number on it — and, characteristically for the period, did the entire calculation by hand.⁴

His method was clever and cheap. He did not have a spectrometer good enough to measure atmospheric absorption directly across all the relevant wavelengths, so he used the Moon. Samuel Langley had measured lunar infrared arriving at the ground at various angles above the horizon. Low in the sky, that radiation passes through much more air — a longer path, more absorbing gas. By comparing readings at different lunar altitudes, Arrhenius could work out how much absorption a given quantity of CO₂ and water vapour produced. He then divided the globe into latitude bands, allowed for seasons, and included one feedback: warmer air holds more water vapour, and water vapour is itself a strong absorber.

He got, for a doubling of CO₂, a warming of 5 to 6 °C. The modern best estimate is 3 °C, likely between 2.5 and 4.⁵ He was high, but he was within a factor of two, using pencil, paper, lunar observations and roughly a year of arithmetic.

He also produced the logarithmic relationship that still holds: each doubling of CO₂ produces roughly the same increment of warming. Going from 280 to 560 ppm has about the same effect as going from 560 to 1,120. Which means the effect does not run away, and also that it does not stop.

The colleague who asked where the carbon would come from

Arrhenius did not set out to study coal. He was working alongside Arvid Högbom, a geologist studying the natural carbon cycle — volcanoes out, rock weathering in. Högbom had, more or less as a curiosity, added up how much CO₂ the world’s coal-burning was putting into the air, and found it was roughly comparable to the volcanic flux. Arrhenius picked this up and calculated that at 1896 rates of coal consumption it would take about three thousand years to double atmospheric CO₂.

He was wrong, badly, and in the direction that matters. He had no way to foresee the twentieth century — oil, the internal combustion engine, aviation, a global population going from 1.6 to 8 billion. We are on course to double it in roughly 250 years, not 3,000. Arrhenius, a Swede, thought a warmer climate would be rather pleasant and would help feed a growing population. He was not raising an alarm; he was answering an interesting question about ice ages.

1900

The objection that held for forty years

Knut Ångström · Uppsala

Four years after Arrhenius published, a Swedish physicist demolished him. The demolition was accepted, and the subject went quiet for the better part of four decades. The objection was intelligent, made in good faith, and it is still the commonest argument raised against the greenhouse effect today, so the reasons it eventually failed matter.

Knut Ångström had his assistant, Herr Koch, pass infrared through a tube of CO₂, then reduce the amount by a third and measure again.⁶ Absorption dropped by only around 0.4 per cent. Ångström concluded that the effect was saturated: the band was already opaque, adding CO₂ changed nothing, and in any case water vapour absorbed at overlapping wavelengths and swamped it. It looked decisive. Arrhenius disputed it but lost the argument, and by the 1910s the CO₂ theory of climate was regarded as a dead end. It stayed dead until the 1940s.

Why saturation is a real effect and still the wrong conclusion

Ångström was not wrong about the measurement. He was wrong about what it implied, for three reasons, and it took better instruments and better physics to see them.

One: the band has wings. Low-resolution instruments in 1900 saw the 15-micrometre band as a single smear. It is actually thousands of individual spectral lines, each with a sharp centre and a broad, weak skirt. The centre saturates almost immediately — Ångström was right about that. The wings do not. Adding CO₂ widens the effective band outwards into wavelengths that were previously letting heat through. That is the physical origin of Arrhenius’s logarithm.

Two: pressure matters. Absorption lines are broadened by molecular collisions, so their width depends on pressure. A laboratory tube at ground-level pressure tells you nothing about what happens at 10 kilometres, where the pressure is a fifth as much and the lines are narrow and far from saturated. You cannot extrapolate a bench measurement to a column of atmosphere that spans a factor of a thousand in pressure.

Three, and this is the one that actually matters: the surface is the wrong place to be looking.

Consider what happens to a photon of 15-micrometre infrared emitted by the ground. It travels a few metres and is absorbed by a CO₂ molecule. That molecule collides with a nitrogen molecule long before it can re-radiate, so the energy becomes ordinary heat in the air. That parcel of air, being warm, radiates in all directions, including up. The photon that goes up is absorbed again a few metres higher. And again. The energy random-walks upward, absorbed and re-emitted, through the whole atmosphere.

Eventually it reaches an altitude where the air above is thin enough that a photon can escape to space without being caught again. Call that the emission height. It is from there, not from the ground, that the planet actually radiates to space at CO₂’s wavelengths.

Now add more CO₂. The air becomes more opaque, so the escape point moves higher. And in the troposphere, temperature falls with altitude — about 6.5 °C per kilometre, which is why mountains have snow on them. A higher emission height means a colder emission height. And a colder body radiates less, by Stefan–Boltzmann.

So the planet is now absorbing the same sunlight but radiating less heat to space. Energy accumulates. The whole column, including the surface, warms until the emission height is back to the temperature it needs to be to balance the books again.

So the greenhouse effect does not work by trapping heat at the surface. It works by raising the altitude from which the planet radiates into colder air. Saturation at the surface is real, and irrelevant. As long as there is a lapse rate — as long as it gets colder with height — adding an absorbing gas warms the surface. Nothing about this saturates until the atmosphere runs out of altitude.

1938

Is it actually happening yet?

Guy Stewart Callendar · London

Callendar was a British steam engineer, an expert on the thermodynamics of steam turbines, and climate was his hobby. In 1938 he read a paper to the Royal Meteorological Society that nobody much liked.⁹

He did the tedious work nobody else would. He collected temperature records from 147 weather stations around the world, corrected and averaged them by hand, and found warming of around 0.3 °C over the preceding fifty years. He then trawled through nineteenth-century chemical measurements of atmospheric CO₂, threw out the ones taken in cities or near vegetation, and concluded that concentration had risen from roughly 290 ppm to around 310 ppm. He estimated the resulting warming and found it matched what the thermometers showed.

The meeting’s discussion, which is printed after the paper and makes fairly brutal reading, was largely dismissive. Sir George Simpson, then Director of the Met Office, said it was impossible to believe such a complex system could be reduced to so simple a cause, and that the warming was probably just a natural fluctuation.

1957

Won’t the ocean just soak it up?

Roger Revelle · Hans Suess · Scripps, California

Through the 1950s the standard reply to Callendar was straightforward oceanography. The oceans hold about fifty times as much carbon as the atmosphere. CO₂ dissolves readily in seawater. Whatever we emit will therefore end up in the sea within a few decades, and the atmospheric concentration will barely move. This was not a fringe view; it was the mainstream one, and it was based on real chemistry.

Revelle and Suess worked out the answer in 1957 and it surprised them.¹⁰ The problem is that seawater is a buffered solution, not plain water.

When CO₂ dissolves it forms carbonic acid, which dissociates into bicarbonate and hydrogen ions. Those hydrogen ions then react with the carbonate ions already present in seawater, converting them to more bicarbonate. Carbonate is the ion that lets seawater keep absorbing CO₂. Every molecule you add uses some up. The water becomes progressively less able to accept the next molecule.

The quantitative result — now called the Revelle factor — is that to raise the total dissolved carbon in surface seawater by one per cent, you have to raise the CO₂ pressure above it by about ten per cent. The ocean is roughly ten times less absorbent than its sheer volume suggests.

There is a second constraint, which is mixing. Only the top hundred metres or so of ocean exchanges gas with the air on short timescales. Getting carbon into the deep ocean requires the overturning circulation, which takes centuries to a millennium. So the effective sink available on a human timescale is a thin skin of water, not the whole ocean.

Revelle wrote the line that gets quoted: human beings are running a large-scale geophysical experiment of a kind that could not have happened in the past and cannot be repeated. He and Suess did not know how it would turn out. But they had established that the ocean would not quietly absorb the problem, and that made the atmospheric concentration worth measuring accurately.

1958

Is CO₂ really rising?

Charles David Keeling · Mauna Loa, Hawaii and the South Pole

Every measurement of atmospheric CO₂ before 1958 was contaminated. Take a sample in a town and you catch chimney smoke. Take it in a field and you catch the plants, which strip CO₂ out of the air by day and pump it back at night. Nineteenth-century values scattered from 250 to 550 ppm and nobody could say which were real.

Charles Keeling, then a young postdoctoral chemist, cared about measurement in a way that was close to obsessive, and it is that obsessiveness rather than any clever hypothesis that makes him important. He built a manometric calibration system of extraordinary precision and made a decision that was as much about geography as chemistry: measure where there is nothing.¹¹

He chose the top of Mauna Loa, 3,400 metres up a Hawaiian volcano in the middle of the Pacific, thousands of kilometres from any continent, with a second station at the South Pole. Air arriving there has spent weeks over open ocean. Sampling only happens when the wind is coming downslope off the free atmosphere rather than upslope off the volcano’s own vents, and the raw records are screened for volcanic contamination, which shows up as obvious short spikes.

He nearly did not get to continue. Funding was cut in the early 1960s and again several times afterwards; the record survived on Keeling’s stubbornness more than on institutional enthusiasm. It has now run without a break since March 1958 and is the longest direct atmospheric measurement of anything.

Figure 1 · The Keeling Curve Annual mean CO₂ at Mauna Loa Observatory, 1958 to 2026. The 2026 value is the Met Office forecast of 429.4 ± 0.6 ppm; May 2025 saw the first monthly average above 430 ppm. Inset shows the seasonal cycle at expanded scale. Data: Scripps CO₂ Program and NOAA Global Monitoring Laboratory.¹²

The wobble is the quality control

The first thing Keeling saw, within two years, was not the rise. It was a sawtooth. CO₂ climbs from October to May and falls from May to October, by about 6 ppm.

This is the northern hemisphere’s forests breathing. Most of the world’s land is north of the equator; in spring the leaves come out and strip carbon from the air, and in autumn they drop and rot and give it back. The southern hemisphere, being mostly ocean, has a much smaller cycle in the opposite phase.

The wobble does something for the record that has nothing to do with climate: it proves the instrument is honest. An apparatus that could detect the seasonal respiration of the world’s vegetation, in the right phase, with the right amplitude, at the top of a Pacific volcano, is an apparatus that can be trusted to detect a trend. If Keeling had reported only a smooth rise, you would be entitled to wonder about drift in his equipment. He reported a signal that no drift could fake.

And the trend was there. 315 ppm in 1958. 429 ppm now, and still climbing at rather more than 2 ppm each year. The rise is measured identically by NOAA on the same mountain with independent equipment, at more than a hundred other stations from Alaska to the South Pole, from aircraft, and from satellites. Nobody disputes it.

1955

Whose carbon is it?

Hans Suess · Charles Keeling · Ralph Keeling · the isotope evidence

Everything so far establishes that CO₂ warms the planet, and that CO₂ is rising. Neither shows that we are the cause. The atmosphere is criss-crossed by enormous natural carbon flows — plants and soils exchange something like 120 gigatonnes of carbon a year with the air, the ocean around 80. Human emissions are about 10. It is entirely reasonable to ask why a small addition to a large churn should be assumed to be the culprit.

It does. Four independent lines of evidence, each of which would be sufficient on its own, and none of which relies on a computer model.

1. Radiocarbon: fossil fuel is dead carbon

Cosmic rays hitting the upper atmosphere continuously produce carbon-14, a radioactive isotope with a half-life of 5,730 years. Living things take it up while alive and stop when they die, after which it decays — the basis of radiocarbon dating.

Coal, oil and gas are the remains of organisms that died tens or hundreds of millions of years ago. That is tens of thousands of half-lives. Fossil fuel contains no carbon-14 at all. It is radiocarbon-dead.

So if you burn fossil fuel and add the CO₂ to the air, the proportion of carbon-14 in the atmosphere must fall — not because any is being removed, but because it is being diluted with dead carbon. Hans Suess measured exactly this in 1955, by comparing the radiocarbon in wood grown in the industrial era with wood grown before it.¹³ The dilution is now called the Suess effect and it is measured in tree rings around the world.

Volcanic CO₂ is also radiocarbon-dead, so this test alone does not distinguish fossil fuel from volcanoes. The other three do.

2. Carbon-13: plants prefer the light isotope

Carbon comes in two stable isotopes, carbon-12 and the slightly heavier carbon-13, at a natural ratio of about 99 to 1. The enzyme plants use to fix carbon during photosynthesis, RuBisCO, is very slightly fussy: it grabs the lighter carbon-12 marginally more readily. Over the life of a plant this discriminates measurably. Plant tissue is depleted in carbon-13 relative to the air it grew in.

Coal and oil are ancient plant matter, so they carry that same depleted signature. Ocean carbon and volcanic carbon do not — they are isotopically much heavier.

So there is a discriminating test. If the extra CO₂ is coming out of the ocean, or out of volcanoes, the atmosphere’s carbon-13 ratio should stay flat or rise. If it is coming from burning plant-derived fossil carbon, the ratio must fall, and it must fall in step with the concentration rising.

Keeling began making these isotope measurements in the 1970s and the record has been continued and extended ever since.¹⁴

Figure 2 · The fingerprint As CO₂ concentration rises (blue), the proportion of carbon-13 in that CO₂ falls (red), from about −7.5 to about −8.6 parts per thousand relative to the standard. That is the signature of carbon that was once living plant tissue. Ocean outgassing or volcanic emission would push the red line the other way. Data: Scripps CO₂ Program isotope record.¹⁴

The two lines cross. That crossing is the most direct piece of evidence in this entire document, and it involves no model, no simulation and no assumption — just a mass spectrometer and a long series of air samples.

3. Oxygen: burning consumes it

Ralph Keeling, Charles Keeling’s son, spent years developing an interferometric technique precise enough to measure changes in atmospheric oxygen, which is hard because oxygen is 21 per cent of the atmosphere and the changes are parts per million against that huge background.¹⁵

The logic is elementary. Combustion consumes oxygen and produces carbon dioxide, in a ratio fixed by the chemistry of the fuel — roughly 1.4 molecules of O₂ per molecule of CO₂ for oil, 1.17 for coal, 2 for natural gas.

Atmospheric oxygen is falling. It is falling at the rate combustion requires, and the ratio of O₂ lost to CO₂ gained matches the known mix of fuels being burned.

This test separates the candidates completely. If the ocean were outgassing CO₂, oxygen would not fall — a warming ocean releases dissolved oxygen too, so O₂ would rise. If volcanoes were the source, oxygen would be unaffected. Only combustion draws down oxygen in the observed proportion. The oxygen measurement is also what allows the split between the land sink and the ocean sink to be calculated, because photosynthesis and combustion have opposite oxygen signatures.

4. The arithmetic: we emit twice what stays

The simplest argument of the four needs no isotopes at all.

We know roughly how much fossil fuel is burned each year, because it is bought, sold, taxed and reported. Coal tonnages, oil barrels, gas volumes — commercial data, collected for reasons that have nothing to do with climate. Currently that is about 37 gigatonnes of CO₂ a year from fossil fuels and cement, plus another 4 or so from deforestation and land use change.¹⁶

Now measure how much CO₂ the atmosphere actually gained. It is a little under half of what we emitted.

That single fact settles the direction of the flow. Nature as a whole is absorbing carbon, not emitting it — roughly half of what we put out is being taken up by oceans and vegetation each year. The natural world is a net sink. It cannot simultaneously be the source of the increase. For nature to be causing the rise, it would have to be emitting the increase and we would have to find somewhere else for our own 41 gigatonnes a year to be disappearing to, and no such sink exists.

Cumulatively since 1850 we have emitted somewhere near 2,500 gigatonnes of CO₂; the atmosphere holds about 1,100 gigatonnes more than it did. The books balance with room to spare.

1978

Where does it go, and for how long?

Urs Siegenthaler · Hans Oeschger · Bern

Urs Siegenthaler was a Swiss physicist at Bern, working in Hans Oeschger’s laboratory, and his contribution is the least famous and among the most useful: he made the carbon cycle quantitative.

Siegenthaler and Oeschger built what is called a box-diffusion model.¹⁹ The atmosphere is one box, the surface ocean another, and the deep ocean is treated as a column in which carbon spreads downward by turbulent diffusion. The clever part is that they did not tune the mixing rate to fit CO₂ data. They calibrated it independently, using the distribution of natural radiocarbon and of the bomb-test radiocarbon injected into the atmosphere by nuclear weapons testing in the 1950s and 60s. That gave them a physically constrained rate at which the ocean can swallow anything, checked against a tracer with nothing to do with the question.

With that in hand they could project future concentrations from emission scenarios. Their conclusions have held up: roughly half of what we emit stays airborne, and getting rid of the rest takes a very long time.

The distinction that causes most of the confusion

You will often hear that CO₂ only lasts about five years in the atmosphere, and this is used to argue that our emissions cannot accumulate. The figure is correct and the conclusion is wrong, because two different quantities are being confused.

Residence time is how long an individual molecule stays airborne before being absorbed by a leaf or the sea. That is indeed about five years, because the natural exchange fluxes are huge.

Adjustment time is how long it takes the total amount in the atmosphere to return to normal after you add a slug of it. That is centuries to millennia.

The analogy that makes it obvious: imagine a bath with the tap running hard and the plug out, sitting at a steady level. The residence time of a water molecule is short — it goes straight down the drain. Now open the tap a little wider. The level rises and stays risen, even though each individual molecule is still passing through in seconds. What sets the new level is the imbalance between inflow and outflow, not how long any molecule lingers.

For CO₂ the removal is also slow in stages: the surface ocean equilibrates in years to decades, the deep ocean over centuries, and the final removal by chemical weathering of silicate rocks takes hundreds of thousands of years. Modern work finds that something like 20 to 35 per cent of a CO₂ pulse is still in the atmosphere a thousand years later.²⁰ This is, in practice, effectively permanent on any timescale a person cares about.

1980

What was it doing before we arrived?

Oeschger · Neftel · Stauffer · Lorius · Jouzel · Raynaud · Petit · Siegenthaler · Lüthi

The Mauna Loa record starts in 1958. Callendar’s nineteenth-century chemistry is unreliable. So the obvious challenge is: how do you know 429 ppm is unusual? Perhaps CO₂ swings about like this all the time and we simply have not been watching long enough.

The answer comes from an ordinary question about snow.

How an ice core works

On the Antarctic and Greenland ice sheets, snow falls and never melts. Each year’s fall is buried by the next. Under pressure the snow compacts into firn, a granular half-way stage still connected to the surface by open pores, and then — at somewhere between 50 and 120 metres depth, depending on temperature and snowfall rate — the pores pinch off and seal. From that moment the air inside is a sealed sample of the atmosphere, held in bubbles, under a growing weight of ice.

Drill a cylinder down through the sheet and you recover a continuous stratigraphy, youngest at the top. At Vostok in East Antarctica that means 3,600 metres of ice. At EPICA Dome C, 3,200 metres reaching back 800,000 years.

To read the CO₂, you cut a section, put it in an evacuated chamber, crush or mill it to release the bubbles, and pass the liberated gas through a gas chromatograph or a laser spectrometer. Hans Oeschger’s group at Bern developed the crushing technique; Andreas Neftel and Bernhard Stauffer published the first ice-core CO₂ record covering the last glacial period in 1982.²¹

To read the temperature, you use the water itself. Willi Dansgaard had established in the 1960s that the ratio of heavy to light isotopes in precipitation depends on the temperature at which the water condensed — deuterium against ordinary hydrogen, oxygen-18 against oxygen-16.²² Water containing the heavier isotope condenses out first, so the colder the air when the snow formed, the more depleted the vapour left behind. It is a thermometer, calibrated against modern snow and against borehole temperature profiles, and it is written into the ice.

The dating comes from counting annual layers where accumulation is high enough, from ice-flow models where it is not, and from tie-points: volcanic ash and sulphate spikes from eruptions of known date, and the beryllium-10 signature of geomagnetic reversals. Records from different cores, different continents, drilled by different teams with different equipment, are matched against each other.

What the ice says

Claude Lorius, Jean Jouzel and Dominique Raynaud led the French effort at Vostok, and Jean-Robert Petit published the 420,000-year record in 1999.²³ Urs Siegenthaler extended it to 650,000 years using EPICA Dome C in 2005, tying the ice record back to the carbon-cycle work he had begun three decades earlier.²⁴ Dieter Lüthi reached 800,000 years in 2008.²⁵

Figure 3 · Eight hundred thousand years Composite CO₂ record from the Vostok and EPICA Dome C Antarctic ice cores, simplified. Across eight full glacial cycles the concentration never fell below about 172 ppm nor rose above about 300 ppm. The red line is the industrial era at the same scale. Data: Petit et al. 1999, Siegenthaler et al. 2005, Lüthi et al. 2008.²³²⁴²⁵

For eight complete ice age cycles, CO₂ oscillates between roughly 180 ppm in the depths of glacials and roughly 280 to 300 ppm in the warm interglacials. It never leaves that band. Then, in about 170 years, it goes to 429 — a rise larger than the entire glacial-to-interglacial swing, at something like a hundred times the speed.

It also runs in lockstep with the temperature reconstructed from the same ice. Cold periods have low CO₂; warm periods have high CO₂. That relationship, over eight cycles, is not a coincidence, and we will come to what causes what in a moment.

The Lag

The best objection there is

Milanković · Hays, Imbrie & Shackleton · Caillon · Shakun · Parrenin

If you look closely at the ice cores at the ends of ice ages, you find something awkward: Antarctic temperature appears to start rising before CO₂ does, by several centuries.

This is a real observation. It was published by mainstream glaciologists in mainstream journals. And it is used, reasonably, to argue that the causation runs the other way — that warming releases CO₂ from the ocean rather than CO₂ causing warming, and that our current rise might therefore be an effect rather than a cause.

It deserves a proper answer rather than a dismissal, because half of it is correct.

What actually triggers an ice age ending

Nobody has ever claimed CO₂ starts a deglaciation. The trigger is orbital. Milutin Milanković, a Serbian engineer, worked out in the 1920s and 30s that the Earth’s orbit changes shape on a 100,000-year cycle, its axial tilt varies on a 41,000-year cycle, and the direction of that tilt precesses on cycles of 19,000 and 23,000 years. None of these changes the total sunlight the Earth receives by much; what they change is its distribution — how much falls on northern high latitudes in summer, which governs whether the previous winter’s snow survives.

James Hays, John Imbrie and Nicholas Shackleton confirmed those exact periodicities in deep-sea sediment cores in 1976, in a paper subtitled “pacemaker of the ice ages”.²⁷ The orbital cycles are the pacemaker. They are not, however, remotely strong enough on their own to produce a 5 or 6 °C global temperature change. Something has to amplify them.

What the amplifier is

The sequence that emerges is this. Orbital change warms the northern high latitudes slightly. Ice sheets begin to retreat, exposing dark ground and reducing reflectivity, which absorbs more sunlight. Ocean circulation shifts. The Southern Ocean, where deep water rich in old carbon comes to the surface, begins to release CO₂, and warming water holds less dissolved gas. CO₂ rises. That warms the planet globally, which melts more ice, which releases more CO₂.

So CO₂ is a feedback, not a trigger. That is completely consistent with it also being a cause. A feedback that amplifies is causing part of the outcome. And the arithmetic requires it: without the CO₂ and methane changes, the orbital forcing plus ice-albedo feedback cannot produce the observed magnitude of glacial-interglacial temperature change. The ice cores are, in fact, one of the strongest independent constraints we have on how sensitive the climate is to CO₂ — which is exactly how Lorius, Jouzel, Raynaud and Hansen used them in 1990.²⁸

Three things that sharpen this

The lag applies to Antarctica, not the globe. Jeremy Shakun and colleagues assembled 80 temperature proxy records from around the world for the last deglaciation and found that while Antarctic temperature does lead CO₂, global average temperature lags it.²⁹ The Antarctic warms first for reasons specific to southern ocean circulation; the rest of the world follows CO₂. That is precisely the pattern the amplifier account predicts.

The lag is smaller than first reported. The widely quoted figure comes from Nicolas Caillon in 2003: 800 ± 200 years, for one particular termination.³⁰ Frédéric Parrenin reanalysed the problem in 2013, using nitrogen-15 isotopes to pin down the gas-age/ice-age offset far more tightly. He found the lag at the last deglaciation was essentially zero: −10 ± 160 years.³¹ Much of the apparent lag had been an artefact of the dating offset flagged in the previous chapter.

The analogy does not transfer. Even if the lag were the full 800 years, it would tell you about a world where orbital changes drive slow ocean outgassing over millennia. It would say nothing about a world in which a species digs up fossil carbon and burns it in two centuries. In the ice ages, temperature rose and the ocean gave up CO₂. Today, CO₂ is rising and the ocean is absorbing it, which we know from its falling pH. The direction of the ocean flux is the opposite. Whatever is happening now, it is not the deglacial mechanism.

1967

Doing the sum properly

Syukuro Manabe · Richard Wetherald · Princeton

Arrhenius had done the calculation with radiation only. But the lower atmosphere does not move heat mainly by radiation — it moves it by convection, warm air rising. Any model that leaves convection out will give the wrong answer, and everyone knew it.

Syukuro Manabe, a Japanese meteorologist working in the United States, built a one-dimensional radiative-convective model with Richard Wetherald.³² It divided a column of atmosphere into layers and calculated the infrared transfer between them, with one added rule: if the temperature profile ever became steeper than the rate at which rising air cools, convection kicked in and flattened it back. They also made a key assumption: that relative humidity, rather than absolute humidity, stays roughly constant as the world warms — which is what observation shows, because warmer air both holds and contains more water vapour.

Their answer for a doubling of CO₂ was 2.4 °C. Manabe shared the 2021 Nobel Prize in Physics for this line of work.

What matters more than the number is the prediction that came with it. Their model said that as the surface and troposphere warmed, the stratosphere would cool. This is not intuitive and it is not what any other proposed cause of warming would produce, and we will use it in two chapters’ time.

The reason is the emission-height argument again. More CO₂ makes the lower atmosphere more opaque, so less infrared reaches the stratosphere from below. But the stratosphere’s own CO₂ radiates to space just as efficiently as before — in fact more so, since there is more of it. Less in, same or more out, so it cools. Meanwhile the troposphere warms. A sharply divided vertical signature.

Two years later, in 1979, the US National Academy of Sciences convened a panel under the meteorologist Jule Charney to assess where the science stood.³³ They concluded that a doubling of CO₂ would produce warming of 1.5 to 4.5 °C, with 3 °C most likely. That range has narrowed only modestly in forty-five years of subsequent work — the current assessment is a best estimate of 3 °C, likely 2.5 to 4.⁵ You can read that as a failure to make progress, or as a very early estimate that has proved remarkably robust. Both readings have something to them.

1981

A prediction, and its scorecard

James Hansen · NASA Goddard Institute for Space Studies

A theory earns its keep by predicting something before it is observed. Hansen’s 1981 paper in Science is the clearest instance of it in this field, and it named dates.³⁴

Writing in 1981, when the global temperature record still showed the post-1940 flat spell and warming was not yet obvious, Hansen and colleagues predicted:

• Warming would emerge from the noise of natural variability around the end of the twentieth century. It did, in the 1990s.
• The warming would be greatest at high northern latitudes. The Arctic is warming roughly three to four times the global rate.
• Sea level would rise, both from thermal expansion and from ice loss. It has, at an accelerating rate.
• The Northwest Passage would open. It did, first in 2007.
• Drought-prone regions would be created in North America and central Asia. Broadly borne out, though this is the weakest of the five.

They also gave a temperature projection. Over the four decades since, actual warming has tracked that 1981 projection closely.

The 1988 testimony, and being honest about it

In June 1988 Hansen told a US Senate committee that he was 99 per cent certain the warming was real and not natural variation. That summer was hot and the testimony made the front pages. The accompanying paper gave three scenarios: A, with emissions growing exponentially; B, with growth slowing; and C, with emissions stabilising.³⁵

Fingerprint

Sun or greenhouse?

Manabe · Santer · Harries · Feldman · the satellite era

Everything so far establishes that the extra CO₂ is ours, and that CO₂ warms the planet. Neither of those rules out the obvious alternative: that the warming actually being observed is driven by the Sun, or by some internal wobble of the climate system, and the CO₂ rise is running alongside it as a coincidence.

That is a scientific question with a scientific answer, because the two candidate causes leave different marks. More sunlight and more greenhouse gas both warm the surface, but they do it by different routes, and the differences are measurable.

The vertical signature

The clearest discriminator is the one Manabe and Wetherald predicted in 1967, before the satellites existed to test it.

Extra sunlight heats the whole atmospheric column, top to bottom. More sunlight arriving means more energy deposited at every level, including the stratosphere. The stratosphere should warm along with everything else.

Extra greenhouse gas does something quite different. It blocks infrared from below while radiating just as freely to space from above. The stratosphere should cool while the troposphere warms.

Figure 4 · The vertical fingerprint Predicted and observed change in atmospheric temperature with altitude. The stratosphere has cooled measurably since satellite monitoring began in 1979, while the troposphere has warmed — the greenhouse pattern, not the solar one. Note that ozone depletion also cools the lower stratosphere; the upper stratospheric cooling is dominated by CO₂.³⁷³⁸

Satellites and weather balloons have measured this since 1979. The stratosphere has cooled substantially. The troposphere has warmed. Ben Santer and colleagues published the first formal detection of this pattern in 1996 and it has been repeatedly confirmed as the records lengthened.³⁷

The rest of the fingerprint

Two rows of that table deserve expanding, because they are direct measurements of the greenhouse mechanism itself rather than of its consequences.

The notch, seen from space. John Harries and colleagues at Imperial College compared the infrared spectrum leaving the Earth, measured by a satellite instrument in 1970, with the spectrum measured by a different instrument in 1997.³⁹ The 1997 spectrum has a deeper bite taken out of it at the wavelengths CO₂ and methane absorb. That is the mechanism itself, observed from orbit: less heat escaping, at exactly the wavelengths laboratory spectroscopy predicts.

The same thing, seen from the ground. Daniel Feldman and colleagues pointed spectrometers upward at two American sites for eleven years, measuring the infrared arriving at the ground and subtracting the contributions of water vapour, cloud and everything else to isolate CO₂’s alone.⁴⁰ It increased by about 0.2 watts per square metre per decade, which is what line-by-line radiative transfer predicts from the measured CO₂ rise. The surface forcing by CO₂ has now been directly observed, not inferred.

And the last row is decisive against internal variability. If the warming were just heat being shuffled around within the climate system — the ocean releasing stored heat into the air, say — then the ocean would be cooling as the atmosphere warmed. It is not. The ocean has taken up more than 90 per cent of the extra energy, and its heat content is rising steadily, tracked now by some 4,000 autonomous Argo floats.⁴¹ Atmosphere and ocean are gaining energy at the same time. That means the planet as a whole is gaining energy, and the imbalance must therefore be at the top of the atmosphere. Internal redistribution cannot do that. Satellites measure the imbalance directly at around 1 watt per square metre.⁴²

Rivals

Could it be something else?

Lockwood · Svensmark · Kirkby · Muller

Alternative explanations have been proposed seriously and tested seriously. Here is where the main ones stand, including the two where the sceptical side scored real points.

The Sun

Total solar irradiance has been measured from satellites since 1978, which removes the need to argue about proxies. It varies by about 0.1 per cent over the eleven-year sunspot cycle and shows no upward trend; if anything solar activity has declined slightly since the 1980s. Mike Lockwood and Claus Fröhlich showed that since about 1985 the solar trend has been in the opposite direction to the temperature trend.⁴⁴ Estimated solar forcing since 1750 is a few hundredths of a watt per square metre against a total anthropogenic forcing of about 2.7.⁵ The Sun did contribute to early twentieth-century warming. It cannot account for what has happened since.

Cosmic rays and clouds

Henrik Svensmark proposed that galactic cosmic rays seed cloud formation, that the solar magnetic field modulates how many reach us, and that this amplifies a small solar signal into a large climate one. It was a good hypothesis with a plausible mechanism, and CERN built an experiment to test it. The CLOUD chamber found that ionisation does affect the nucleation of aerosol particles, so the first step is real. But the effect on the number of particles that go on to grow large enough to seed a cloud droplet is small.⁴⁵ And there is no trend in cosmic ray flux over the period of warming, so even a strong mechanism would have nothing to work with. The idea was tested at CERN and does not carry the weight required of it.

Urban heat islands and the temperature record

This one has a good ending. Cities are warmer than countryside, many weather stations are in or near cities, and stations get moved, replaced and rebuilt around. The suspicion that the surface temperature record was contaminated was reasonable.

In 2010 Richard Muller, a Berkeley physicist and a public critic of climate science’s statistical practices, set up the Berkeley Earth Surface Temperature project to redo the analysis from scratch, with different methods, a larger station set of some 39,000 stations, and open code. Part of the funding came from foundations associated with the Koch brothers. The stated aim was to find out whether the existing records could be trusted.

The result reproduced the existing records almost exactly.⁴⁶ Urban heat island effects were real but too small to matter for the global average, because urban stations are a small minority and the corrections were already being applied. Muller published an article titled “The conversion of a climate change sceptic” and Berkeley Earth is now one of the standard global datasets. This is what the process is supposed to look like: a critic funded to check the work, checking it, and reporting what he found.

Volcanoes, and the ocean, and internal variability

Covered above: volcanic CO₂ is around one per cent of ours and volcanic aerosols cool rather than warm; the ocean is a net sink with falling pH; and internal variability cannot explain a system that is gaining energy overall. Internal variability is very real over years and decades — it is most of why 2023 to 2025 were so hot and why the early 2010s looked flat — but it moves heat around, it does not create it.

Unknowns

What is genuinely uncertain

Where the field is arguing with itself

Everything above is settled to the standard applied in any other branch of physics. What follows is not, and pretending otherwise would undermine the whole document. These are real disagreements between competent people, and some of them are large.

Climate sensitivity is still a range, not a number

The equilibrium warming from a doubling of CO₂ is assessed at 3 °C, likely 2.5 to 4, very likely 2 to 5.⁵ That has narrowed a little since Charney’s 1.5 to 4.5 in 1979. Most of the narrowing came from a 2020 assessment that combined process understanding, the historical record and palaeoclimate constraints instead of relying on models alone.⁴⁷ It is still nearly a factor of two from end to end, which is the difference between a serious problem and a severe one.

Clouds

Clouds both reflect sunlight away, which cools, and trap infrared, which warms. Which effect dominates depends on cloud height, thickness, particle size and location, all of which change as the world warms in ways that are hard to model — individual clouds are far smaller than a model grid cell, so they must be parameterised rather than simulated. The net cloud feedback is assessed as positive, but with an uncertainty range that still spans from nearly zero to strongly amplifying, and it remains the largest single contributor to the spread in sensitivity estimates.

Aerosols

Sulphate and other pollution particles reflect sunlight and change cloud properties, and they cool. Total aerosol forcing is estimated at around −1.1 W/m², but with a range wide enough to be worth stating: roughly −2.0 to −0.6.⁵ That is the single biggest uncertainty in the total human forcing, because a strongly cooling aerosol effect would require a strongly warming greenhouse effect to produce the temperature actually observed — high aerosols imply high sensitivity, and vice versa. The two uncertainties are coupled.

Some models run too hot

A subset of the latest generation of climate models have sensitivities above 5 °C, higher than the palaeoclimate and historical evidence supports, generally because of changes in how they handle clouds. Hausfather and colleagues argued in Nature that these should not simply be averaged in with the rest, and that projections should be weighted by how well models reproduce observed warming.⁴⁸ This is a criticism from inside the field, it is correct, and it means that raw multi-model averages of the most recent generation overstate expected warming.

The last three years are not fully explained

2023, 2024 and 2025 were warmer than a straightforward reading of the forcing plus El Niño would have predicted, and 2024 exceeded 1.5 °C above the pre-industrial baseline for the first time in an individual year. Candidate explanations include the 2020 regulations that cut sulphur in shipping fuel, removing a cooling aerosol effect over the oceans; declining Chinese air pollution; a measured drop in low cloud cover; the water vapour injected into the stratosphere by the Hunga Tonga eruption; and the solar cycle peak. There is no settled attribution and there is active argument about whether it represents an acceleration or a temporary excursion. It is an open problem and it is being treated as one.

Regional projections and extreme weather attribution

Confidence varies enormously by claim, and this is where public discussion is loosest. Attribution of heatwaves to warming is on very firm ground — it is close to arithmetic, since shifting a distribution warmer makes hot extremes far more likely. Heavy rainfall is well supported by the physics of atmospheric moisture. But attribution of drought, of tropical cyclone frequency, of tornado activity, and of regional changes in rainfall patterns is much weaker, and some claimed links do not survive scrutiny. When a specific weather event is attributed to climate change, the honest question is which of those categories it falls into.

Scandals

The scandals, honestly

Mann · McIntyre · the CRU emails

Two episodes have done more than anything else to make people distrust this field. Both involved real misconduct or real error, and both are routinely misdescribed by both sides.

The hockey stick

Michael Mann, Raymond Bradley and Malcolm Hughes published temperature reconstructions for the past six centuries and then the past millennium, using tree rings, corals and ice cores.⁴⁹ The resulting curve — a long gentle decline followed by a sharp twentieth-century rise — became the emblem of the whole subject, put in the IPCC’s 2001 summary and reproduced everywhere.

Stephen McIntyre and Ross McKitrick argued that the principal component method used had a bias: because the data were centred on the twentieth-century mean rather than the full-record mean, the algorithm preferentially selected series with hockey-stick shapes.⁵⁰

They were right about the statistical flaw. The US National Research Council reviewed the whole controversy in 2006. It found the criticism of the centring convention valid, and the reconstruction’s uncertainty ranges understated, particularly before 1600.⁵¹ It also concluded that recent decades being the warmest in 400 years was well supported, while claims about the full thousand years carried much less confidence than had been presented.

And the conclusion survived anyway. Redoing the analysis with the corrected method changes the curve only slightly. More importantly, more than a dozen subsequent reconstructions by independent groups using different proxies and different statistics find broadly the same shape, most comprehensively the PAGES 2k consortium’s global synthesis.⁵²

The hockey stick is also not load-bearing. Nothing in the argument of this document depends on it. If the Medieval Warm Period had been as warm as today globally — the strongest version of the sceptical claim — it would show that the climate is capable of large natural swings, which nobody denies, and it would not touch the isotope evidence, the mass balance, the spectroscopy or the fingerprints. The hockey stick was made into the emblem of the field by its defenders as much as by its critics, and that was a mistake by both.

Climategate

In November 2009, shortly before the Copenhagen summit, someone released thousands of emails taken from the Climatic Research Unit at the University of East Anglia. Some of them read very badly: discussion of a “trick” to handle a divergence between tree-ring and thermometer records, discussion of keeping papers out of an IPCC report, and discussion of deleting emails subject to Freedom of Information requests.

Eight separate investigations followed, including by the House of Commons Science and Technology Committee, a panel under Lord Oxburgh, and an independent review under Sir Muir Russell. Their findings were consistent: no evidence of scientific fraud, no evidence that the temperature records had been manipulated, and the scientific conclusions unaffected.⁵³ The “trick” referred to a data-presentation technique that had been described openly in the published literature.

But the reviews also found real fault. The Muir Russell review criticised a consistent failure of openness, and the Information Commissioner’s Office noted that FOI requests had been handled improperly. That is not nothing. A field that is asking the public to act on its findings has an obligation to share its data and code, and in 2009 parts of it were not doing so. The practice has improved substantially since, partly as a direct result.

Tests

What would break the argument

The falsification conditions

A position with no specifiable disproof is not a scientific position. So here is the list. If any of the following were established, the argument in this document would be in serious trouble.

• The stratosphere warming alongside the surface. That would point to the Sun or to something else entirely, and would contradict the emission-height mechanism directly.
• Atmospheric carbon-13 rising as CO₂ rises, rather than falling. That would mean the extra carbon is not plant-derived and therefore not fossil.
• Atmospheric oxygen staying flat while CO₂ rises. That would rule out combustion as the source.
• Ocean pH rising as atmospheric CO₂ rises. That would mean the ocean is the source rather than a sink.
• Ocean heat content falling while the surface warmed. That would mean heat is being redistributed rather than accumulating, and the system is not gaining energy.
• Total solar irradiance trending upward in step with temperature since 1980. The satellites say it is not, and those satellites are the same class of instrument used for everything else.
• Ice cores from multiple sites showing 400+ ppm in the past 800,000 years without accompanying warmth. That would break the palaeo relationship.
• No spectral notch deepening in the outgoing infrared, and no increase in downward infrared at CO₂’s wavelengths at the surface.
• Warming stopping for twenty-five to thirty years while CO₂ continued to rise. Note that fifteen years does not qualify — the early 2010s slowdown was within the range internal variability produces, and ocean heat content kept climbing throughout it. Thirty years would be a different matter.

None of these has happened. All of them are things we are actively measuring, with instruments that would show them if they occurred.

The short version

The chain, stated once, end to end

The Earth should be about −18 °C from the sunlight it absorbs; it is about 15 °C. Fourier, 1824

The gap is caused by trace gases that absorb infrared while nitrogen and oxygen do not. Measured on a bench with a brass tube and a thermopile. Tyndall, 1859

Adding more of those gases raises the altitude from which the planet radiates to space; that altitude is colder; a colder radiator emits less; so the surface must warm. Saturation at the surface is irrelevant. Arrhenius 1896, Ångström’s objection 1900, resolved by Plass 1956

The ocean cannot absorb our emissions quickly, because carbonate buffering makes seawater roughly ten times less absorbent than its volume implies. Revelle & Suess, 1957

CO₂ has risen from 278 to 429 ppm, measured continuously in clean air for sixty-eight years, confirmed at more than a hundred stations worldwide. Keeling, 1958 to now

That extra carbon is radiocarbon-dead, depleted in carbon-13, accompanied by falling atmospheric oxygen in exactly combustion’s proportion, and amounts to about half of what we are documented to have burned. It is ours. Suess 1955, Keeling 1979, R. Keeling 1992, Global Carbon Project

About half of each year’s emissions stays airborne, for centuries. Siegenthaler & Oeschger, 1978

For 800,000 years CO₂ stayed between 172 and 300 ppm, tracking temperature through eight ice ages. We have exceeded that band by 43 per cent in under two centuries. Neftel & Oeschger 1982, Petit 1999, Siegenthaler 2005, Lüthi 2008

The warming that has followed has the greenhouse fingerprint and not the solar one: stratosphere cooling, Arctic amplification, nights and winters warming fastest, a deepening notch in the infrared leaving the planet, and more infrared arriving at the ground at CO₂’s exact wavelengths. Manabe & Wetherald 1967, Santer 1996, Harries 2001, Feldman 2015

And the planet is gaining energy overall, at about a watt per square metre, with the ocean holding more than 90 per cent of it. That cannot be internal reshuffling. Argo, CERES

Each of those links was established separately, usually by someone asking a question narrow enough to answer with an instrument. The persuasive force is not in any one of them. It is that they were built independently, by people who mostly were not trying to prove a thesis, and they fit.

The original papers

Full citations so you can go and read the primary sources rather than anyone’s summary of them, including this one. Most of the historical papers are out of copyright and freely available; the modern ones mostly have open-access versions. Where a DOI is given it links directly to the paper.

1. 1Fourier, J. B. J. (1824). “Remarques générales sur les températures du globe terrestre et des espaces planétaires.” Annales de Chimie et de Physique, 27, 136–167. Extended as “Mémoire sur les températures du globe terrestre et des espaces planétaires”, Mémoires de l’Académie Royale des Sciences, 7 (1827), 569–604. English translations by Burgess (1837) and by Pierrehumbert (2004) are online.
2. 2Foote, E. (1856). “Circumstances affecting the heat of the sun’s rays.” The American Journal of Science and Arts, 22, 382–383.
3. 3Tyndall, J. (1861). “On the absorption and radiation of heat by gases and vapours, and on the physical connexion of radiation, absorption, and conduction.” Philosophical Transactions of the Royal Society of London, 151, 1–36. The Bakerian Lecture; the experiments date from 1859, first reported in Proceedings of the Royal Society, 10, 37–39.
4. 4Arrhenius, S. (1896). “On the influence of carbonic acid in the air upon the temperature of the ground.” Philosophical Magazine and Journal of Science, Series 5, 41, 237–276. doi.org/10.1080/14786449608620846
5. 5IPCC (2021). Climate Change 2021: The Physical Science Basis. Working Group I contribution to the Sixth Assessment Report. Cambridge University Press. Sensitivity: Chapter 7 and Table 7.13. Forcings: Chapter 7, Figure 7.6. Attribution: Chapter 3, Section 3.3. ipcc.ch/report/ar6/wg1
6. 6Ångström, K. (1900). “Über die Bedeutung des Wasserdampfes und der Kohlensäure bei der Absorption der Erdatmosphäre.” Annalen der Physik, 308(12), 720–732.
7. 7Plass, G. N. (1956). “The carbon dioxide theory of climatic change.” Tellus, 8(2), 140–154.
8. 8Gordon, I. E. et al. (2022). “The HITRAN2020 molecular spectroscopic database.” Journal of Quantitative Spectroscopy and Radiative Transfer, 277, 107949. The laboratory line data underpinning all modern radiative transfer calculation. hitran.org
9. 9Callendar, G. S. (1938). “The artificial production of carbon dioxide and its influence on temperature.” Quarterly Journal of the Royal Meteorological Society, 64(275), 223–240. Worth reading with the printed discussion that follows it.
10. 10Revelle, R. & Suess, H. E. (1957). “Carbon dioxide exchange between atmosphere and ocean and the question of an increase of atmospheric CO₂ during the past decades.” Tellus, 9(1), 18–27.
11. 11Keeling, C. D. (1960). “The concentration and isotopic abundances of carbon dioxide in the atmosphere.” Tellus, 12(2), 200–203. See also Keeling, C. D. (1998), “Rewards and penalties of monitoring the Earth”, Annual Review of Energy and the Environment, 23, 25–82, for his own account of keeping the record alive.
12. 12Scripps CO₂ Program, scrippsco2.ucsd.edu, and NOAA Global Monitoring Laboratory, gml.noaa.gov/ccgg/trends. Both publish the full daily and monthly data. 2026 forecast: Betts, R. A. et al., Met Office annual Mauna Loa CO₂ forecast.
13. 13Suess, H. E. (1955). “Radiocarbon concentration in modern wood.” Science, 122(3166), 415–417.
14. 14Keeling, C. D. et al. (2005). “Atmospheric CO₂ and ¹³CO₂ exchange with the terrestrial biosphere and oceans from 1978 to 2000.” In A History of Atmospheric CO₂ and its Effects on Plants, Animals, and Ecosystems, Springer. Ongoing data at scrippsco2.ucsd.edu. See also Graven, H. et al. (2017), Geoscientific Model Development, 10, 4405–4417, for the compiled isotope histories.
15. 15Keeling, R. F. & Shertz, S. R. (1992). “Seasonal and interannual variations in atmospheric oxygen and implications for the global carbon cycle.” Nature, 358, 723–727.
16. 16Friedlingstein, P. et al. “Global Carbon Budget” (published annually). Earth System Science Data. The full emissions, sink and airborne-fraction accounting, with data and code released openly each year. globalcarbonbudget.org
17. 17Gerlach, T. (2011). “Volcanic versus anthropogenic carbon dioxide.” Eos, 92(24), 201–202. See also Burton, M. R., Sawyer, G. M. & Granieri, D. (2013), “Deep carbon emissions from volcanoes”, Reviews in Mineralogy and Geochemistry, 75, 323–354.
18. 18Bates, N. R. et al. (2014). “A time-series view of changing surface ocean chemistry due to ocean uptake of anthropogenic CO₂ and ocean acidification.” Oceanography, 27(1), 126–141. Covers the Bermuda, Hawaii, Canary and other long-running stations.
19. 19Siegenthaler, U. & Oeschger, H. (1978). “Predicting future atmospheric carbon dioxide levels.” Science, 199(4327), 388–395. See also Siegenthaler, U. & Sarmiento, J. L. (1993), “Atmospheric carbon dioxide and the ocean”, Nature, 365, 119–125.
20. 20Archer, D. et al. (2009). “Atmospheric lifetime of fossil fuel carbon dioxide.” Annual Review of Earth and Planetary Sciences, 37, 117–134.
21. 21Neftel, A., Oeschger, H., Schwander, J., Stauffer, B. & Zumbrunn, R. (1982). “Ice core sample measurements give atmospheric CO₂ content during the past 40,000 yr.” Nature, 295, 220–223.
22. 22Dansgaard, W. (1964). “Stable isotopes in precipitation.” Tellus, 16(4), 436–468. The basis of ice-core palaeothermometry.
23. 23Petit, J. R. et al. (1999). “Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica.” Nature, 399, 429–436. doi.org/10.1038/20859
24. 24Siegenthaler, U. et al. (2005). “Stable carbon cycle–climate relationship during the late Pleistocene.” Science, 310(5752), 1313–1317. The EPICA Dome C record to 650,000 years.
25. 25Lüthi, D. et al. (2008). “High-resolution carbon dioxide concentration record 650,000–800,000 years before present.” Nature, 453, 379–382.
26. 26MacFarling Meure, C. et al. (2006). “Law Dome CO₂, CH₄ and N₂O ice core records extended to 2000 years BP.” Geophysical Research Letters, 33, L14810. See also Etheridge, D. M. et al. (1996), Journal of Geophysical Research, 101(D2), 4115–4128.
27. 27Hays, J. D., Imbrie, J. & Shackleton, N. J. (1976). “Variations in the Earth’s orbit: pacemaker of the ice ages.” Science, 194(4270), 1121–1132.
28. 28Lorius, C., Jouzel, J., Raynaud, D., Hansen, J. & Le Treut, H. (1990). “The ice-core record: climate sensitivity and future greenhouse warming.” Nature, 347, 139–145.
29. 29Shakun, J. D. et al. (2012). “Global warming preceded by increasing carbon dioxide concentrations during the last deglaciation.” Nature, 484, 49–54.
30. 30Caillon, N. et al. (2003). “Timing of atmospheric CO₂ and Antarctic temperature changes across Termination III.” Science, 299(5613), 1728–1731. The source of the widely quoted 800-year lag.
31. 31Parrenin, F. et al. (2013). “Synchronous change of atmospheric CO₂ and Antarctic temperature during the last deglacial warming.” Science, 339(6123), 1060–1063.
32. 32Manabe, S. & Wetherald, R. T. (1967). “Thermal equilibrium of the atmosphere with a given distribution of relative humidity.” Journal of the Atmospheric Sciences, 24(3), 241–259.
33. 33Charney, J. G. et al. (1979). Carbon Dioxide and Climate: A Scientific Assessment. National Academy of Sciences, Washington DC. Twenty-two pages, and the range it gives has barely moved since.
34. 34Hansen, J. et al. (1981). “Climate impact of increasing atmospheric carbon dioxide.” Science, 213(4511), 957–966. doi.org/10.1126/science.213.4511.957
35. 35Hansen, J. et al. (1988). “Global climate changes as forecast by Goddard Institute for Space Studies three-dimensional model.” Journal of Geophysical Research, 93(D8), 9341–9364.
36. 36Hausfather, Z., Drake, H. F., Abbott, T. & Schmidt, G. A. (2020). “Evaluating the performance of past climate model projections.” Geophysical Research Letters, 47(1), e2019GL085378.
37. 37Santer, B. D. et al. (1996). “A search for human influences on the thermal structure of the atmosphere.” Nature, 382, 39–46.
38. 38Steiner, A. K. et al. (2020). “Observed temperature changes in the troposphere and stratosphere from 1979 to 2018.” Journal of Climate, 33(19), 8165–8194.
39. 39Harries, J. E., Brindley, H. E., Sagoo, P. J. & Bantges, R. J. (2001). “Increases in greenhouse forcing inferred from the outgoing longwave radiation spectra of the Earth in 1970 and 1997.” Nature, 410, 355–357.
40. 40Feldman, D. R. et al. (2015). “Observational determination of surface radiative forcing by CO₂ from 2000 to 2010.” Nature, 519, 339–343.
41. 41von Schuckmann, K. et al. (2023). “Heat stored in the Earth system 1960–2020: where does the energy go?” Earth System Science Data, 15, 1675–1709.
42. 42Loeb, N. G. et al. (2021). “Satellite and ocean data reveal marked increase in Earth’s heating rate.” Geophysical Research Letters, 48(13), e2021GL093047.
43. 43Vose, R. S., Easterling, D. R. & Gleason, B. (2005). “Maximum and minimum temperature trends for the globe: an update through 2004.” Geophysical Research Letters, 32, L23822.
44. 44Lockwood, M. & Fröhlich, C. (2007). “Recent oppositely directed trends in solar climate forcings and the global mean surface air temperature.” Proceedings of the Royal Society A, 463(2086), 2447–2460.
45. 45Kirkby, J. et al. (2011). “Role of sulphuric acid, ammonia and galactic cosmic rays in atmospheric aerosol nucleation.” Nature, 476, 429–433. And Dunne, E. M. et al. (2016), “Global atmospheric particle formation from CERN CLOUD measurements”, Science, 354(6316), 1119–1124.
46. 46Rohde, R. et al. (2013). “A new estimate of the average Earth surface land temperature spanning 1753 to 2011.” Geoinformatics & Geostatistics: An Overview, 1:1. Data, code and methods at berkeleyearth.org.
47. 47Sherwood, S. C. et al. (2020). “An assessment of Earth’s climate sensitivity using multiple lines of evidence.” Reviews of Geophysics, 58(4), e2019RG000678. The paper that narrowed the Charney range.
48. 48Hausfather, Z., Marvel, K., Schmidt, G. A., Nielsen-Gammon, J. W. & Zelinka, M. (2022). “Climate simulations: recognize the ‘hot model’ problem.” Nature, 605, 26–29.
49. 49Mann, M. E., Bradley, R. S. & Hughes, M. K. (1998). “Global-scale temperature patterns and climate forcing over the past six centuries.” Nature, 392, 779–787. And (1999), “Northern hemisphere temperatures during the past millennium”, Geophysical Research Letters, 26(6), 759–762.
50. 50McIntyre, S. & McKitrick, R. (2005). “Hockey sticks, principal components, and spurious significance.” Geophysical Research Letters, 32, L03710. See also their 2003 paper in Energy & Environment, 14(6), 751–771.
51. 51National Research Council (2006). Surface Temperature Reconstructions for the Last 2,000 Years. National Academies Press, Washington DC. The independent adjudication of the hockey stick dispute.
52. 52PAGES 2k Consortium (2019). “Consistent multidecadal variability in global temperature reconstructions and simulations over the Common Era.” Nature Geoscience, 12, 643–649.
53. 53Russell, M. et al. (2010). The Independent Climate Change E-mails Review. And House of Commons Science and Technology Committee (2010), The disclosure of climate data from the Climatic Research Unit at the University of East Anglia, Eighth Report of Session 2009–10, HC 387-I. And Oxburgh, R. et al. (2010), Report of the International Panel set up by the University of East Anglia.

Changes

1. 1.4March 2026Revised and expanded.
2. 1.0March 2026First published.

How to cite

Allen, P. (2026). The Chain of Evidence (version 1.4). Occam’s Shaving Kit. https://occamsshavingkit.uk/papers/the-chain-of-evidence/