An autumn ENSO phase term gave uncertain mean winter sea-level forecast gains at two Pacific gauges in 1991-2025

We tested whether different warm and cold ENSO slopes improve winter mean sea-level forecasts at San Francisco and Seattle. The reference model used time, local autumn sea level, and linear autumn Nino 3.4. Each fit used earlier complete winters only. The equal mean of site percentage reductions in mean squared error was 2.08% across 33 common test winters in 1991-2025. The conditional 95% five-year block interval was -1.73% to 6.44%. Site gains were -4.27% and 8.43%. The October-only check gave a mean gain of 5.20%, with an interval of 0.63% to 10.42%, but changed both input season and data support. Other mean sensitivity intervals included zero. Thus the primary test leaves the mean added value uncertain. Two selected gauges, a short test record, retrospective source revisions, missing months, and uncertainty conditional on fitted forecast histories limit the result. The package gives fixed monthly evidence, raw-source hashes, all forecasts, code, figures, sensitivity checks, and unexecuted numerical audit code.

sea level · ENSO · phase asymmetry · seasonal forecast · chronological validation · Pacific coast