ENSO Predictability From Combined Wyrtki and Hasselmann Memory in a Cyclostationary Linear Inverse Model

Abstract Low‐order models can improve forecasts of the El Niño‐Southern Oscillation (ENSO) by representing its core dynamics within simple frameworks. The extended nonlinear recharge oscillator (XRO) has demonstrated high skill by integrating fundamental recharge oscillator dynamics and interactions with other modes of global sea surface temperature (SST) variability. Linear Inverse Models (LIMs) are a related approach, but they have often shown lower due to insufficient use of global SST information and redundant predictors. Here we introduce a cyclostationary LIM (CSLIM) inspired by the XRO, trained on equatorial Pacific sea surface height as a proxy for upper‐ocean heat content (“Wyrtki memory”) and leading global SST modes (“Hasselmann memory”). This “Wyrtki‐CSLIM” achieves retrospective Niño3.4 forecast skill out to 15 months, rivaling the XRO in capturing the timing and amplitude of major ENSO events. Both approaches outperform the current operational LIM, underscoring the importance of global SST variability for extended forecasts.