March western Pacific water volume gave higher ENSO forecast error than basin volume in a 2000-2024 test

We tested whether March warm water volume improves November-January Nino 3.4 forecasts beyond March sea surface temperature and its change since December. We used public PMEL GODAS and CPC ERSSTv5 snapshots, with earlier-year fits and 25 test issues in 2000-2024. Western Pacific volume raised mean squared error by 37.8% against the surface baseline. The conditional 95% three-year block interval for error reduction was -104.4% to +6.1%. Basin volume changed error by less than 1%. The western model had 36.9% higher error than the basin model, with a conditional interval of 8.2% to 95.3%. A common time trend made this secondary difference uncertain. The primary test therefore gives no clear added forecast gain beyond recent surface conditions. This result applies to simple linear models and revised products. It does not reject the recharge process or establish future forecast skill. The package contains raw snapshots, exact hashes, code, annual evidence, all forecast scores, block draws, an algorithm audit, and a figure.

ENSO · warm water volume · spring forecasts · chronological validation · GODAS · null result