Two-month-old mean sea level gave uncertain mean high-water rate forecast gains at six US gauges in 1991-2025

We tested whether local monthly mean sea level from two months earlier improves monthly high-water-day rate forecasts at six US gauges. We used fixed NOAA records, early calendar-month 95th percentiles, and fits that used only earlier years. There were 2,426 held-out station-months in 1991-2025. The equal-site mean reduction in monthly rate mean squared error was 3.30%, with a conditional 95% five-year block interval of -0.01% to 6.68%. Adding mean level after recent high-water rates gave 0.99%, with an interval of -1.22% to 3.17%. Planned 90th-percentile and three-month-lag checks gave positive mean gains. Thus evidence depends on the target, lag, and reference model. Statistical thresholds, selected gauges, revised data, a simple tide model, and conditional intervals limit the result. This is not a test against NOAA's operational outlook.

sea level · high water · forecast · NOAA · conditional uncertainty