Quantifying the Predictability of ENSO Complexity Using a Statistically Accurate Multiscale Stochastic Model and Information Theory

Abstract

An information-theoretic framework is developed to assess the predictability of ENSO complexity, which is a central problem in contemporary meteorology with large societal impacts. The information theory advances a unique way to quantify the forecast uncertainty and allows to distinguish the predictability limit of different ENSO events. One key step in applying the framework to compute the information gain representing the predictability is to build a statistically accurate dynamical model. To this end, a recently developed multiscale stochastic model, which succeeds in capturing both the large-scale dynamics and many crucial statistical properties of the observed ENSO complexity, is incorporated into the information-theoretic framework. It is shown that different ENSO events possess very distinct predictability limits. In addition to the ensemble mean, the ensemble spread also has remarkable contributions to the predictability. While the information theory indicates that predicting the onset of the eastern Pacific El Niños is challenging, it reveals a universal tendency to convert strong predictability to skillful forecast for predicting many central Pacific El Niños about two years in advance. In addition, strong predictability is found for the La Niña events, corresponding to the effective discharge process. In the climate change scenario with the strengthening of the background Walker circulation, the predictability of sea surface temperature in central Pacific has a significant response with a notable increase in summer and fall. Finally, the Gaussian approximation is shown to be accurate in computing the information gain, which facilitates the use of more sophisticated models to study the ENSO predictability.