That carbon dioxide traps heat had been known since the nineteenth century: Eunice Foote in 1856 and John Tyndall in 1859 measured the gas's thermal behaviour, and Svante Arrhenius calculated in 1896 that increasing it would raise temperature. What was missing was data on what the atmosphere was actually doing. Until the mid-1950s measurements were scattered, inconsistent and contaminated by local pollution, and many scientists assumed the oceans would absorb any excess.

Charles David Keeling argued that a clean reference measurement could only be made mid-ocean and high up. In 1958 he began continuous measurement at the Mauna Loa Observatory in Hawaii, at 3,400 metres. The two patterns that emerged in the first years are the same today: a seasonal oscillation as northern hemisphere plants leaf out and die back — a breathing at planetary scale — and, beneath it, a baseline shifting a little higher every year.

The record began at 315 ppm in 1958 and passed 420 ppm in the 2020s. Combined with ice cores, the line points to a level unseen in any natural period of the last 800,000 years. The Keeling curve is the single most reproduced dataset in climate science and underlies every policy argument from the Paris Agreement to carbon pricing. The programme itself came close to ending several times through budget cuts: that the value of continuous measurement only appears once it has run long enough is a recurring problem of science policy.