Separate spring change slopes by winter sign gave no clear RONI forecast gain in 1981-2024

We tested whether separate spring change slopes after negative and nonnegative winters improve forecasts of the relative ENSO index, RONI. We used the NOAA CPC ERSSTv5 table and 44 target winters from 1981 to 2024. Each fit used earlier issue years only. Adding a winter sign indicator and its product with MAM minus DJF to a MAM level and change model increased mean squared error by 1.73%. Conditional 95% five-year block limits for error reduction were -6.47% to +2.24%. Error also rose in the two target halves, with a trend control, and when three large warm targets were excluded from scoring. A last-30-year fit increased error by 5.14%. The main test gives no clear gain for this response form. It does not establish symmetric ENSO dynamics or real-time forecast skill. One revised index, future reference-period information, rounding, 44 test winters and conditional uncertainty limit the result. Code, fixed numerical inputs, every forecast, all planned scores, audit checks and a figure are included.

ENSO · RONI · warm cold asymmetry · seasonal forecast · null result