The CO₂ record is missing 59 weeks, and 35 of them are four outages

When the Mauna Loa record goes blank, how long does it stay blank? Each row is one run of consecutive missing weeks in the weekly CO₂ series, 1958 to 2001: 22 runs in all. Most are a single week. A handful are the kind of outage that matters, and one of them, in early 1964, lasted four months.

Interactive chart available when JavaScript loads.

One outage dwarfs the rest

Each point is one run, placed at its year; its height is the number of weeks lost. The longest run spans 18 weeks, from 25 January to 23 May 1964. Nineteen of the 22 runs fall before 1970; after 1970 the record loses a week in 1976, four weeks in 1984, and one in 1985, and then nothing.

Interactive chart available when JavaScript loads.

Most runs are one week; most missing weeks are not

Count runs and the picture is benign: 63.6% of runs are a single missing week, and the median run is 1 week. Count weeks and it flips. The four runs of four weeks or more hold 59.3% of all missing weeks, so the typical blank in the record belongs to an outage, not a skipped reading.

Interactive chart available when JavaScript loads.

The air did not move while nobody was watching

Each point compares the last reading before a gap with the first after it. The points hug the diagonal: across the 22 gaps, the reading after averaged -0.04 ppm different from the reading before, and the largest jump in either direction was 2.2 ppm. Even the four-month gap closed with CO₂ 2.2 ppm higher, which is one spring's seasonal climb.

Interactive chart available when JavaScript loads.

Takeaway

Missing weeks are a 1960s problem and mostly a 1964 problem: 59 blanks in 2,284 weekly slots, with the four longest outages holding three in five of them. Readings on either side of every gap sit within a seasonal step of each other, so the holes hide no surprises, only the exact shape of a few springs.

Limits

A run is consecutive blank rows in this series, not a documented instrument failure. The values before and after a gap bound it but do not fill it; a 1964 spring peak is simply not in the record.