Towards Probabilistic Multivariate ENSO Monitoring

A probabilistic approach to describing El Nino-Southern Oscillation (ENSO), based on consideration of the signal-to-noise ratio for the coupled ocean-atmosphere ENSO state, is presented. The ENSO signal is estimated using an ensemble of historical atmospheric model simulations forced by observed sea surface temperatures and sea ice during 1980-2016. The noise is estimated from departures of individual model realizations from their ensemble average when subjected to identical forcing. It is found that this atmospheric noise effect is substantial and yields considerable uncertainty in detecting the true coupled ENSO mode. This uncertainty exceeds analysis errors by an order of magnitude. Greater atmospheric noise is found to prevail during El Nino than La Nina, suggesting that the intensity of the monitored ENSO state during El Nino is prone to greater misattribution. Our results demonstrate that a deterministic state estimate of ENSO conditions may not be representative of the true real-time coupled ENSO mode.