Earlier-year range scaling reduced high tide flood interval scores at 83 US gauges in 2022-2025 but covered only 59% of counts

We tested whether errors from earlier years could improve annual high tide flood day ranges. We used a verified campaign table with stable-threshold gauges and NOAA outlooks. Each test year used only earlier years to set a common scale. The test has 307 station-years at 83 US gauges in 2022-2025. Mean 68%-target interval score fell from 15.43 to 13.86 days. The equal-year score change was -1.56 days. Scores fell in all four years. Coverage rose from 50.2% to 59.3%, below the 68% target. Mean range width rose from 4.46 to 6.81 days. A conditional year bootstrap gave limits of -2.56 to -0.56 days for the score change. Four test years are too few for a strong test of future skill. The fair-sign probability for four gains is 0.0625 under independent year signs. This is a retrospective test with revised data. Dependence, inherited threshold checks and a short record limit the result. Code, input data, source hashes, all forecasts, a separate scalar audit and a figure are included.

high tide flooding · interval score · chronological test · range calibration · NOAA · uncertainty