A fixed 13.66-day fit reduced daily midpoint error more in summer at five of six US gauges in 1993-2025

We tested if the use of a daily water-level midpoint leaves a seasonal lunar signal. We fitted one 13.66079-day term on 1960-1992 and tested it on 1993-2025 at six US tide gauges. We removed each observed calendar month mean. The same fixed fit reduced squared error more in June-August than in December-February at five gauges. The summer-minus-winter gains were 17.27 percentage points at San Francisco and 13.81 at Seattle. Simultaneous six-site conditional 95% intervals were 12.07 to 21.99 and 11.05 to 16.71 points. Sewells Point had no clear difference. Nearby 13-, 14-, and 15-day terms gave little gain at the two Pacific sites. Results remained positive in both parts of the test record there. This is a seasonal extension of a published campaign method check. The midpoint retains tides and is not a storm-surge measure. Selected sites, fixed fits, retrospective month centering, missing data, and block assumptions limit inference.

sea level · daily midpoint · lunar cycle · season · NOAA CO-OPS · method test