A Mechanism From 1859

In 1859 John Tyndall built an apparatus to measure how different gases absorb infrared radiation. He found that nitrogen and oxygen, which make up almost all of the atmosphere, are essentially transparent to it, while water vapour, carbon dioxide and methane absorb it strongly.

In 1896 Svante Arrhenius calculated by hand what doubling atmospheric carbon dioxide would do to surface temperature, and got a figure in the range that modern models produce.

The mechanism is nineteenth-century laboratory physics, and it is worth stating plainly because it is often described as though it were a modern inference from statistics. Sunlight arrives mostly as visible light and passes through the atmosphere. The warmed surface radiates heat back as infrared. Molecules with the right vibrational modes absorb that infrared and re-emit it in all directions, including downward. The result is that the surface is warmer than it would otherwise be: by about 33 degrees, without which the Earth would average roughly minus 18 and be a ball of ice.

The greenhouse effect is not the controversial part. The controversial part is how much warming follows from adding more gas, and that turns on feedbacks.

Doubling carbon dioxide on its own produces about one degree. What makes the answer uncertain is what the warming then does. Warmer air holds more water vapour, which is itself a greenhouse gas: a strong positive feedback. Melting ice exposes darker surfaces that absorb more sunlight: another. Clouds do both: they reflect sunlight, cooling, and trap infrared, warming, and which dominates depends on their type and altitude.

Climate sensitivity is the name for the total long-run warming from a doubling, and its range has been stubborn. The 1979 Charney report gave 1.5 to 4.5 degrees. Decades of work narrowed it barely at all until recently, when combining multiple lines of evidence brought the likely range to roughly 2.5 to 4. Clouds remain the largest single source of the remaining spread.

The evidence that this is happening now comes from several directions, and the ice cores are the most direct.

Snow falls on Antarctica and Greenland, traps air between the flakes, and is compressed into ice as more falls on top. The bubbles preserve a sample of the atmosphere from the moment the ice sealed, and the oxygen isotope ratio in the ice itself records the temperature when the snow fell. Cores from Antarctica reach back over 800,000 years.

Across eight glacial cycles, temperature and carbon dioxide track each other closely, and current carbon dioxide is far above anything in that record.

There is a standard objection to this, and it is a good one that repays understanding properly.

At the end of past ice ages, temperature rose before carbon dioxide did, by several hundred years. If carbon dioxide were driving the warming, this looks backwards.

The sequence is: Milankovitch cycles, slow variations in the Earth's orbit and tilt, change how sunlight is distributed across the seasons and latitudes, which starts a deglaciation. Warming oceans then release dissolved carbon dioxide, which traps more heat, which warms further. Roughly 90% of the eventual warming comes after the carbon dioxide rise.

A feedback lagging its trigger is what a feedback does. And it says nothing about what happens when the gas is added first and the orbital forcing is not doing anything, which is the current situation.

There is also a signature that distinguishes greenhouse warming from a brighter Sun, and it is worth knowing because it settles the question cleanly. If the Sun were the cause, the whole atmosphere would warm. What is observed is that the surface and lower atmosphere are warming while the stratosphere is cooling, which is exactly what you get when heat is trapped below rather than added from above. Solar output has in fact been flat or slightly declining over the period of the strongest warming.

Finally, the newest branch, which reframed a question that had seemed unanswerable.

For years the correct response to "was this heatwave caused by climate change?" was that no single event can be attributed to a long-run trend. Attribution science changed the question. Run large ensembles of climate models twice (once with observed emissions, once with a counterfactual atmosphere) and compare how often an event of that severity occurs in each.

The output is a change in probability rather than a yes or no: this heatwave was thirty times more likely, this rainfall was ten per cent more intense. That is a statement about a distribution rather than about a cause, which is what makes it answerable, and it is a good demonstration that reframing a question is sometimes what turns it into science.