Do asymmetries in ENSO predictability arise from different recharged states?

Abstract

Prospects for El Niño–Southern Oscillation (ENSO) predictability at long lead-times lie in the subsurface oceanic memory along the equatorial Pacific. Long considered a reliable precursor to ENSO, the oceanic heat content in springtime, often referred to as the recharge-discharge, is considered the most promising indicator of an ENSO event to come. In this study, we utilize January initialized hindcasts from the North American Multi-model Ensemble (NMME) over 1982–2010 to confront the hypothesis that the springtime recharge is a skillful predictor of ENSO the following winter. We find that the NMME ensemble mean predictions for the springtime recharge are highly skilled, even at a 10-months lead. Overall, as an independent predictor of ENSO, the springtime recharge-discharge tips the scale towards like-sign ENSO, but the spread of ENSO outcomes remains large. In both observations and the NMME predictions, recharged (discharged) states rarely evolve into La Niña (El Niño) events, yet an ENSO-neutral state is as likely to occur after a preconditioned state as is a like-sign ENSO event, particularly in observations. However, more often than in observations, the initialized predictions follow springtime recharged, neutral, and discharged states with El Niño, ENSO-neutral, and La Niña events, respectively, indicating that the NMME underestimates the uncertainty in nature. Predictions from initially recharged and discharged states also produce comparable signal-to-noise ratios in December ENSO predictions over the hindcast period. Therefore, in the realistic forecast setting considered, neither a recharged nor a discharged state produces a more predictable ENSO outcome, which is at odds with conclusions from recent predictability studies.