Three Interruptions

The graph is the jagged one that climbs from left to right, the one that gets put on a screen in every argument about the climate. The teeth along it are the northern hemisphere breathing: forests and crops leaf out through the northern spring and pull carbon dioxide out of the air, then die back in autumn and hand it over again. Under the teeth the line goes up, every year, without exception, since March 1958.

It exists because a small number of people kept one instrument switched on near the top of a volcano, and it has stopped three times.

Where the air is clean

The Mauna Loa Baseline Observatory sits on the north flank of the volcano at 3,397 metres, or 11,135 feet. The altitude is the entire reason it is there. The site pushes up through the marine temperature inversion that sits over that part of the Pacific, the layer that separates the messier air below from the free troposphere above it. Sampling above the inversion means sampling air that has not just come off a city, a forest or a field. NOAA picked the spot for that, and the instruments have been reading there since the 1950s.

The first monthly mean in the record is March 1958. It is 315.71 parts per million.

What the teeth are

Most of the world's land sits north of the equator, so the northern growing season dominates the whole atmosphere and shows up as a wobble in a reading taken in the middle of the Pacific. The swing is not small. Take the highest and lowest monthly means in a single year and subtract one from the other: in 1959 the gap was 4.96 parts per million, and in 2025 it was 6.14. Every one of those swings is the land in the northern half of the planet taking in carbon over one summer and releasing it over one winter, and it happens every year whatever else is going on.

What the line does

The most recent monthly mean in the file, for July 2026, is 429.12 parts per million. That is 113.41 more than the first reading, a rise of about 36 per cent in sixty eight years.

The more useful number is the slope. Average the monthly means into annual means and the arithmetic is easy to follow. The annual mean for 1959 was 315.98 parts per million and for 1969 it was 324.62, a rise of 8.64 across the decade, or 0.86 a year. The annual mean for 2015 was 401.01 and for 2025 it was 427.35, a rise of 26.34, or 2.63 a year. Three times the rate.

The thresholds tell the same story from the other end. The annual mean first went above 350 parts per million in 1988. It took another twenty seven years to reach 400, in 2015. It took ten more to reach 425, in 2025. The line is not simply rising, it is steepening, and that is a thing you can only know because somebody measured the same air in the same way for six decades rather than measuring it well once.

What a monthly mean is

There are 821 monthly means in the file. Each is built from daily means, and the file says how many days went into each one, which is the part of the record that never gets quoted. The seven monthly figures published so far for 2026 rest on between 17 and 26 days of measurement each. Weather closes the site, instruments need calibration, and the volcano occasionally has opinions. A monthly average is not a month of continuous readings, and the honest version of the record says so in a column.

Three interruptions

The first was money. In the spring of 1964 sampling at Mauna Loa was interrupted by funding problems, six years into a record that had already shown the concentration climbing.

The second was the volcano. An eruption in 1984 left another small gap.

The third was the volcano again, and it was the serious one. Mauna Loa erupted in late November 2022. The observatory lost power about five hours after the eruption began, and the lava crossed the access road. NOAA's own data file records the outcome in a note at the top: measurements from Mauna Loa Observatory were suspended as of 29 November 2022 and resumed in July 2023, and everything from December 2022 to 4 July 2023 was measured instead at a site among the Maunakea Observatories, roughly 21 miles north.

Getting back was slow. Road crews cut a temporary route across the hardened flow, and construction was completed on 26 March 2026. The observatory remains closed to the public.

What that episode shows is what a long record is actually for. The value of the thing is not any single reading. It is that every reading was taken the same way, in the same air, so that a number from 1961 and a number from 2021 can be put next to each other and the difference between them means something. A gap is expensive because it breaks that comparison. So when the lava cut the road, the measurements were not simply suspended; they were moved twenty one miles north and up another mountain, and the file marks every one of those months so that nobody reads them as Mauna Loa numbers. That is the work. It is unglamorous and it is the entire basis of the graph.

Three stops in sixty eight years, then. One budget and two volcanoes.

Who pays for it

NOAA runs the observatory. The Scripps measurements, which are the ones that produced the curve and which run alongside NOAA's own, are currently supported by Schmidt Sciences with supplemental support from Earth Networks.

That is worth sitting with for a moment. The longest continuous instrumental record of what the atmosphere of this planet is made of is not a permanent public utility. It is a set of line items that somebody has to keep renewing, in the same way that somebody had to keep renewing them in 1964.

In November 1997 Keeling stood outside the observatory next to John Chin of NOAA while a plaque was unveiled naming the building after him. The plaque carries an engraving of the curve itself, running from 1958 to somewhere around 1995, which at that point was most of it.

The file NOAA publishes today says, in a comment line near the top, that the data from March 1958 through April 1974 were obtained by C. David Keeling of the Scripps Institution of Oceanography. That file was last regenerated on 5 August 2026.

Sources

• NOAA Global Monitoring Laboratory, "Monthly mean CO2 at Mauna Loa Observatory" (the first monthly mean of 315.71 parts per million in March 1958 and the July 2026 value of 429.12; the annual seasonal ranges of 4.96 in 1959 and 6.14 in 2025 and the average annual rises of 0.86 and 2.63 parts per million, all computed from the monthly means in this file; the header note that measurements were suspended as of 29 November 2022, resumed in July 2023, and were taken from Maunakea for the months between; the note that data from March 1958 through April 1974 came from C. David Keeling of Scripps; and the file creation date of 5 August 2026).
• NOAA Global Monitoring Laboratory, "Trends in Atmospheric Carbon Dioxide" (the published trend series and the documentation behind the monthly mean file).
• NOAA, "Mauna Loa Baseline Observatory" (the elevation of 3,397 metres or 11,135 feet on the north flank; the marine temperature inversion and why sampling above it gives clean free tropospheric air; monitoring since the 1950s; the eruption cutting road access; the completion of the temporary access road on 26 March 2026; and the absence of public access).
• EarthSky, "Historic Keeling Curve data offline after Mauna Loa eruption" (the brief funding problem in 1964 and the 1984 eruption each leaving small gaps in the record, and the monitoring equipment losing power about five hours after the November 2022 eruption began).
• Scripps Institution of Oceanography, "The Keeling Curve" (the Scripps measurements being currently supported by Schmidt Sciences with supplemental support from Earth Networks).
• United States Geological Survey, "The Mauna Loa carbon dioxide record: lessons for long-term Earth observations" (the record treated as a case study in what long term Earth observation requires to survive).