Winter-to-spring Pacific contrast change gave no clear added winter forecast gain for 1981-2024 origins

We tested whether winter-to-spring change in the Nino 1+2 minus Nino 4 temperature anomaly improves next-winter Nino 3.4 forecasts. The baseline used spring strength, its change since winter, absolute centered strength, and the spring contrast. We used one fixed NOAA CPC ERSSTv5 table and 44 forecasts from issue years 1981-2024. Each fit used earlier issue years only. Adding contrast change reduced mean squared error by 0.08794 C squared, or 11.57%. The conditional 95% five-year block interval for relative gain was -5.17% to 26.74%. Root mean squared error fell from 0.8716 to 0.8197 C. Every planned check left mean gain uncertain. Thus the test gives no clear added skill, despite a positive point estimate. This extends a prior spring contrast-level test with a lagged spatial-state feature. Few winters, one revised product, baseline choice and conditional intervals limit inference. Code, aggregate results, uncertainty draws, source hashes and a verified snapshot retrieval route are included.

ENSO · spatial evolution · temperature contrast · chronological prediction · null result