Fixed lunar terms reduced the mean late first-return hazard contrast across six US gauges in 2011-2025

We tested the first return of upper-tail daily high water after seven quiet days at six selected US tide gauges. We reused verified NOAA-derived predictions with fits from 1996-2010 and test records from 2011-2025. Each filter used its own calendar-month 95th-percentile tail. The mean daily first-return hazard difference between days 25-30 and days 8-13 was 11.81 percentage points with a seasonal filter and 2.06 points with added fixed lunar terms. The paired change was -9.75 points, with a conditional three-year block interval of -11.47 to -7.86 points. A filter of equal size with other periods gave 11.69 points. Five site estimates decreased; Sewells Point was close to zero. The mean first-return probability by day 35 changed from 67.94% to 65.77%. The result concerns event timing after filters, not physical recovery or a causal effect on storms. Fits, thresholds, missingness, and source revisions are held fixed in the interval. The data were previously inspected in a campaign study. We provide new first-return code, cohorts, risk sets, source hashes, and an offline repeat.

sea level · first return · event clustering · tides · conditional hazard