East-west Pacific temperature contrast cut six-month volume forecast error by 18.76% at 2000-2023 origins

We tested whether Nino 3 minus Nino 4 temperature anomalies improve forecasts of basin warm water volume change. We used fixed NOAA GODAS and CPC ERSSTv5 monthly records. Annual expanding fits used only completed earlier targets. Adding current and six-month past temperature contrast to a state, history and absolute-temperature model reduced six-month mean squared error by 18.76% at 288 monthly origins in 2000-2023. RMSE fell from 0.84251 to 0.75940 in units of 10^14 cubic metres. The conditional 95% three-year block interval for MSE reduction was 7.55% to 28.23%. One- and five-year block limits were positive. The gain was 25.05% in 2000-2011 and 8.30% in 2012-2023; the later interval included zero. Three- and twelve-month checks did not resolve gains. A check added after the main result retained the six-month gain with one-month state history. Revised data, few ENSO cycles, model choice and conditional uncertainty limit inference. This result does not establish a physical cause or operational skill. The package includes original code, aggregate scores, uncertainty draws, source hashes, a separate numerical audit and a figure.

ENSO · warm water volume · spatial contrast · forecast benchmark · GODAS · ocean recharge