Other Pacific gauges gave a small monthly sea-level forecast gain that depended on local history in 1990-2025

We tested whether past readings at other gauges improve monthly sea-level forecasts at San Diego, San Francisco and Seattle. We used a fixed PSMSL record and fit each test year from earlier years. In 1990-2025, adding the other two gauges to a model with one local lag reduced the equal-site mean squared error by 3.09%. The conditional 95% five-year block interval was 0.14% to 5.86%. Each site had 411 scored months. Seattle had the largest gain, 6.81%. With three local lags, the mean gain was 2.21%, with an interval of -0.82% to 5.06%. Shorter training and an extra month of data delay also left the mean gain uncertain. Thus the small primary gain does not establish a stable general forecast benefit. Shared forcing can explain useful past-gauge information, but this test cannot identify wave travel or cause. Three selected sites, revised records, data access delays, simple models and conditional uncertainty limit the result. Exact source series, hashes, code, forecasts, block draws and a figure are included.

sea level · Pacific · tide gauges · monthly forecasts · local history · conditional uncertainty