Construction of low dimensional models of El Niño-southern oscillation using empirical orthogonal functions

Author: Roulston, Mark Stephen

Year: 2000

Degree: Dissertation (Ph.D.)

Advisor: Yung, Yuk L.

Committee Member: Unknown, Unknown

Option: Planetary Sciences

DOI: 10.7907/32fe-ag39

Abstract

The main theme of this thesis is the construction and analysis of low dimensional dynamical models of El Nino-Southern Oscillation (ENSO). Low dimensional models of ENSO have been constructed [Val86, WF96, Jin97b] but all the models were built on ad hoc assumptions and simplifications. The low dimensional models in this thesis were constructed from a higher dimensional intermediate coupled model (ICM) by Galerkin projection on truncated basis sets of empirical orthogonal functions (EOFs), a method that has previously been applied to atmospheric models and models of turbulence [Sel95, HLB96]. This method makes no simplications to the physics of the ICM but instead projects it into a lower dimensional subspace of its full state space that has been empirically shown to contain the maximum amount of variance of any subspace of the same dimension.

Analysis of the reduced models shows that the variability of the full model can be explained by three types of mode; I. self-sustaining. II. linearly stable but nonlinearly coupled to a self-sustaining mode, III. linearly stable and excited by noise. When driven by noise the stable modes can couple to the primary modes, and this provides an additional pathway for noise to perturb these leading modes.

The implication is that the behavior of the model without stochastic forcing can be well reproduced by the low dimensional models obtained by projection onto the leading EOFs but that many more EOFs are required to properly model the response to stochastic forcing.

Experiments with the full model also indicate that the model is not efficient at transferring power from high (intraseasonal) frequencies to low (interannual) frequencies. In particular the impact of a Madden-Julian type oscillation in the 30-60 day frequency range on the interannual ENSO variability is negligble compared to the impact of the interannual component of atmospheric variability.

Files