Item

Abstract

Principal Oscillation Pattern (POP) analysis, which is a technique for
deriving the space-time characteristics of a complex system, is used for
analyzing and predicting the Southern Oscillation.

Pairs of patterns are derived from the POP analysis of the Southern Hemisphere
sea level pressure field and from the associated correlation pattern analysis
of other atmospheric and oceanic parameters. The two vectors, that are
combined from these pairs of patterns, form basis vectors of a two-dimensional
space. It is shown that the dynamic process of the Southern Oscillation can be
described as a damped oscillation in this reduced space. The oscillatory
feature is characterized by the eastward migration of the ascending branch of
the Walker Circulation and of the anomalies in the sea level and upper layer
temperature.

The coefficient time series of the two basis vectors in the reduced space may
be understood as a bivariate index of the Southern Oscillation, whlch
describes both, the "peak" phase and the "onset" phase of the phenomenon,
Generalizing the original concept, the POP framework is used to predict this
index and the traditional univariate SO index.

In a series of hindcast experiments the POP prediction scheme is tested. It
turns out to be skillful for a lead time of about two seasons. In terms of the
hindcast correlation the POP forecast scores better than persistence and a
conventional ARMA model in forecasting the traditional so index.