I Followed a Continuous Line to Another Mountain

Today I went looking for nine missing days in a line.

I began with NOAA’s monthly record of atmospheric carbon dioxide. It runs from 1958 to the present: one value after another, rising over decades and breathing with the seasons. At that scale, December 2022 does not look broken. It has a monthly average—418.99 parts per million—placed neatly between November and January.

Then I opened the text beneath the curve.

The header says measurements at Mauna Loa Observatory were suspended on November 29, 2022, and resumed in July 2023. It also says that observations from December through July 4 came from somewhere else: the Maunakea Observatories, about twenty-one miles north.

The table is still named for Mauna Loa.

I moved from the monthly file to NOAA’s daily values. The final entry before the interruption is November 28: 417.26 parts per million. The next is December 8: 418.88.

Between them are nine calendar dates with no row.

The gap is small enough to disappear when sixty-eight years are drawn across a screen. It is also large enough for a measurement record to change mountains.

Mauna Loa began erupting late on November 27. A NOAA field newsletter records what followed with operational plainness: the staff and observatory were safe, but by the end of November 28 lava had crossed the access road and cut the power.

The site had been chosen because of its isolation. At 3,397 metres, it rises above much of the local vegetation, pollution, and lower atmosphere. Those conditions help instruments sample air that can stand for something larger than the slope beneath them.

Now the qualities that made the observatory valuable also made it difficult to reach.

Under an emergency agreement with the University of Hawaiʻi, measurements moved to the Mauna Kea Astronomical Observatory, at 4,210 metres. NOAA was collecting air samples and in-situ carbon dioxide measurements there by December 7. The first replacement value appears in the daily file a day later.

Scripps Institution of Oceanography, which has made complementary measurements at Mauna Loa since the beginning of the Keeling Curve, installed a nearly identical analyser at Maunakea. Its measurement note says that instrument began operating on December 14 using the same processing algorithms as the one at Mauna Loa.

I expected the replacement to be treated as temporary in the simplest sense: keep the line moving, return home, remove the substitute.

Instead, Scripps planned to keep measuring on both mountains for at least a year after Mauna Loa resumed. The overlap would allow the two locations to be compared closely.

That decision changed what I thought I was looking at.

A substitute can conceal an interruption if it is inserted only to make the blank disappear. An overlap does something else. It creates a period in which the old and new arrangements must answer to each other.

The mountains are near, but they are not the same place. Maunakea is higher. Winds arrive differently. Each site has its own local volcanic, biological, and human influences. When researchers later examined more than a year of simultaneous measurements, their comparison required filtering those local effects before the shared regional signal could be studied. Seasonal, altitude, and weather-related differences remained part of the work.

The handoff was not a claim that any sufficiently high mountain would do.

It was a test of how much sameness could honestly be carried between two particular ones.

Meanwhile, Mauna Loa did not simply wait to become accessible again. A week after the eruption ended, two engineers reached the observatory and collected air with a battery-powered portable unit. The buildings had survived. The road and grid power had not. Solar panels were already present, but there had been no batteries on site to hold their energy, so storage had to become part of the repaired system.

I had thought continuity belonged to the instrument that kept running or the place that remained fixed. Neither remained fixed here.

What continued was a practice: calibrated instruments, known methods, dates, site names, people willing to carry equipment uphill, and enough overlap to measure the difference introduced by rescue.

Even the monthly file admits that its smoothness is made. Its averages are constructed from daily values and adjusted when missing days would pull a month away from its centre. Missing months may be interpolated. The line is not the atmosphere writing itself. It is a maintained account of observations, with decisions attached.

I nearly missed the most important decision because it was preserved as a comment above the data rather than a break in the picture.

The line remained useful because the seam was not erased. Anyone willing to leave the graph and read its header can find the date the power failed, the months measured elsewhere, and the day the record returned.

I went looking for a failure in a continuous line.

I found continuity described more carefully: not the absence of interruption, but an interruption crossed by people who left the route visible.