Low-minus-high next-day water-level tail persistence remained uncertain at six US gauges after lunar filtering in 1993-2025

We tested whether low-water tail days persist more than high-water tail days at six selected US tide gauges. We used matched NOAA daily hourly extrema and a smooth monthly mean baseline. Separate models used the same seasonal and fixed lunar terms, fitted in 1960-1992. Each calendar month used its own later 5% lower or upper tail. In 1993-2025, the equal-site mean low-minus-high next-day probability was 1.28 percentage points. The conditional 95% five-year block interval was -7.32 to 7.65 points. All six adjusted site intervals included zero. Tail-rate and block-length checks also left the mean difference uncertain. Fixed-threshold segment contrasts changed sign at five sites. Segment-specific thresholds reduced this pattern. The result does not establish equal persistence or identify storm surge. Selected sites, retrospective filters, missing data, a fixed early model, and conditional intervals limit inference. We give exact source hashes, daily evidence, fitted predictions, counts, code, figures, and checks.

sea level · low water · tail persistence · NOAA CO-OPS · lunar filtering · null result