A Pacific SST contrast reduced reconstruction error against a six-month Nino 3.4 model at three island gauges in 2000-2024

We tested whether the east-west pattern of tropical Pacific sea surface temperature (SST) adds information about monthly relative sea level beyond Nino 3.4. We used PSMSL records at Kwajalein, Apra Harbor, and Malakal-B. Each test year in 2000-2024 used only earlier sea-level targets from 1975 onward for model fitting. The baseline had month means, a linear trend, two three-month Nino 3.4 means, and their absolute values. The added model had two matching means of Nino 4 minus Nino 3. Across 878 accepted gauge months, the equal-site mean reduction in squared error was 7.76%. Its conditional 95% three-year block interval was 1.78% to 12.65%. All three site point estimates were positive, but all site intervals included zero. With previous-month local sea level in both models, the gain was 1.91% (-0.42% to 3.96%). A later exploratory check with a longer Nino 3.4 history also had an interval that included zero. Thus, the SST contrast added information against the specified baseline, but broader added value remains uncertain. These are reconstructions that use observed target-month SST, not lead forecasts. Three selected gauges, 25 test years, revised data, datum history, and uncertainty conditional on saved errors limit inference. We provide exact source extracts, hashes, predictions, all checks, code, figures, and a successful offline repeat.

ENSO · Nino 3.4 · SST contrast · sea level · Pacific islands · PSMSL · reconstruction · block bootstrap · baseline sensitivity · reproducibility