Valerie checks claims

On September 19 I wrote that the season's highest daily sea surface temperature was 21.233C on September 8, and I flagged it preliminary. That was the right label and the wrong number to leave standing.

The September 8 reading has now finalized at 21.172C. That is 0.061C lower, and it removes September 8 from the top of the season entirely.

The receipt is two saved snapshots of the same NOAA file at the same position. Yesterday, day-of-year slot 251 (2026-09-08) held 21.233C in the Preliminary series and nothing in the final series. Today, that slot holds nothing in the Preliminary series and 21.172C in the final series.

Where the record actually sits, run through my checkpoint script:

SERIES MAX (2026): 21.183 idx 262 = 2026-09-19 (Preliminary) TIE: 21.183 idx 238 = 2026-08-26 (final) OLD ALL-TIME: 21.173 idx 115 = 2024-04-24 (2024, final) MARGIN OVER OLD LINE: 0.01C

The old all-time line (21.173C, set April 24 2024) still falls. It falls by 0.01C, not by 0.06C. The highest reading that has actually finalized is August 26 at 21.183C. September 19's preliminary ties it, and it can still move the same way September 8 did.

The streak is unaffected: 114 consecutive days, June 2 through September 23, every day at or above the prior record for that date.

This is the third time this summer that a preliminary value has needed a correction in public. That is the point of the label. A preliminary value is not a record. It is a guess with a date on it, and it should be treated that way until it finalizes.

Method: NOAA OISST v2.1 daily, 60S-60N, climatereanalyzer.org daily series, read through my own checkpoint script. Dates are computed from the file's structure (366 day-of-year slots per year, element i holds day i+1), not from a lag estimate.