Résumé
The Warm Water Volume (WWV), a proxy for the equatorial Pacific heat content, is the most widely used oceanic precursor of the El Nino Southern Oscillation (ENSO). The standard interpretation of this lead relation in the context of the recharge oscillator theory is that anomalous easterlies during, e.g. La Nina, favour a slow recharge of the equatorial band that will later favour a transition to El Nino. Here we demonstrate that WWV only works as the best ENSO predictor during boreal spring, i.e. during ENSO onset, in both observations and CMIP5 models. At longer lead times, the heat content in the western Pacific (WWVW) is the best ENSO predictor, as initially formulated in the recharge oscillator theory. Using idealised and realistic experiments with a linear continuously stratified ocean model, and a comprehensive wave decomposition method, we demonstrate that spring WWV mostly reflects the fast Kelvin wave response to wind anomalies early in the year, rather than the longer-term influence of winds during the previous year. WWV is hence not an adequate index of the slow recharge invoked in the recharge oscillator. The WWVW evolution before spring is dominated by forced Rossby waves, with a smaller contribution from the western boundary reflection. WWVW can be approximated from the integral of equatorial wind stress over the previous similar to 10months, thus involving a longer-term time scale than WWV main time scale (similar to 3months). We hence recommend using WWVW rather than WWV as an index for the slow recharge before the spring predictability barrier.