Lunar filtering left more upper-tail concurrence within than across coasts at six US gauges in 2011-2025

We tested upper-tail date concurrence after a lunar filter at six US tide gauges. We used NOAA hourly daily maxima minus monthly mean sea level. Prior models used 1996-2010 for fitting. We tested 2011-2025 with a separate 95th percentile for each site, month and filter. The mean joint-date rate was 1.039% for seven same-coast pairs and 0.298% for eight opposite-coast pairs. After subtraction of a calendar-month marginal-rate reference, the within-minus-across contrast was 0.741 (0.512 to 0.843) percentage points, with a 95% three-year block interval. Thresholds were refitted in each sample. This result shows a regional pattern after filtering. It does not identify storms or establish independent coasts. These are selected harbor sites and reused data. Only 15 test years, missing data, fixed fitted terms and the block assumptions limit inference. The package gives source hashes, daily evidence, code, all draws, a figure and an independent audit.

sea level · upper tails · spatial dependence · lunar filter · NOAA CO-OPS