Other regional gauges cut annual high tide flood error by 24% in a 69-gauge reconstruction for 2019-2025

We tested whether annual flood outlook errors contain shared regional information. We used a published NOAA verification table and kept 69 gauges with records in each year from 2019 to 2025. We added the median error at other gauges in the same region and year to each outlook midpoint. Mean absolute error fell from 2.91 to 2.20 days over 483 gauge-years. The conditional 95% interval for the error change was -1.29 to -0.25 days. The regional correction had lower error in all seven years. It also had lower error than a national correction and a previous-year regional correction. This extends an earlier in-sample variance test with a test that leaves each target gauge out. It shows shared regional information in the errors. It does not identify a physical cause or show forecast skill. Peer counts use observations from the same completed year. Seven years, selected gauges, revised data and fixed-estimate uncertainty limit the result.

high tide flooding · regional coherence · forecast error · leave-one-gauge-out · reconstruction