Partner gauge history reduced next-day high-water error by 4.78% across six Atlantic links in 2011-2025

We tested whether partner gauge history improves next-day daily maximum estimates beyond local history and fixed lunar cycles. Models used four Atlantic NOAA gauges, with fits in 1996-2010 and tests in 2011-2025 on six directed links. Adding the previous two partner daily maxima reduced mean squared error by an equal-link mean of 4.78%. A conditional 95% three-year block interval was 4.16% to 5.54%. Seven local lags gave a mean gain of 4.95%. Older partner lags 8 and 9 gave 0.32%. Boston history gave almost no gain at The Battery. The results show lead information under a simple model on retained data support. They do not establish a physical cause or an operational warning gain. Fixed retrospective sources, selected complete months, a limited tide baseline, and 15 test years limit inference. The package includes daily inputs, exact source hashes, all predictions and block draws, code, and source and numerical checks.

water level · tide gauge · lagged prediction · Atlantic coast · lunar cycles