Autumn DMI gave no clear added next-year ENSO forecast gain in 25 winters from 2000 to 2024

We tested whether autumn Indian Ocean Dipole Mode Index (DMI) adds next-year winter Nino 3.4 forecast information beyond autumn and summer Nino 3.4. We used fixed HadISST1.1 DMI and CPC ERSSTv5 indices. Each fit used only earlier complete target seasons from 1950. There were 25 test winters, starting in 2000-2024. Adding DMI gave -2.59% mean squared-error reduction. The conditional 95% three-year block interval was -21.83% to +17.54%. RMSE rose from 1.132 to 1.147 degrees C. Correlation rose from 0.031 to 0.258, but a larger cold mean error offset lower centered error. Planned checks for separate DMI slopes, unequal ENSO responses, trend, recent training and target season all had intervals that included zero. The test gives no clear added gain under these models. It does not rule out useful ocean information or identify a physical cause. A short test, revised products, simple models and conditional uncertainty limit the result. The package gives exact monthly evidence, forecasts, source hashes, code, draws, a figure and independent numerical checks.

ENSO · Indian Ocean Dipole · long-lead forecast · Nino 3.4 · time split · forecast bias · null result