A Stochastic Bayesian Approach to Identify the Dynamical Regimes of ENSO

Statistical inverse modeling is used to explore and quantify the relative likelihood of the different dynamic regimes exhibited within an intermediate coupled model of the tropical Pacific. This is accomplished by systematically searching model parameter space for the settings that enable the model to reproduce the observed variance, skewness, kurtosis, and decorrelation times of the Nino3 index of the past 150 years. One objective of this exercise is to relate particular features of ENSO behavior to the specification of model parameters including aspects of the climatological mean state such as the mean thermocline depth and surface wind forcing. In particular we find that the manifestation of a positive skewness in histograms of modeled SSTs is most strongly related to the specification of the mean thermocline depth. The model configuration that most resembles the statistical characteristics of the observed ENSO is a system with regular self-sustaining (non-damped) oscillations in which ENSO irregularity is entirely governed by atmospheric noise. Emphasis will be placed on the use of this statistical inverse modeling technique to shed light on the processes governing ENSO and its use in interpreting changes in ENSO behavior through time.