Adaptive filtering and prediction of the Southern Oscillation index

Abstract

Singular spectrum analysis (SSA), a variant of principal component analysis, is applied to a time series of the Southern Oscillation index (SOI). The analysis filters out variability unrelated to the Southern Oscillation and separates the high‐frequency, 2‐ to 3‐year variability, including the quasi‐biennial oscillation, from the lower‐frequency 4‐ to 6‐year El Niño cycle. The maximum entropy method (MEM) is applied to forecasting the prefiltered SOI. Prediction based on MEM‐associated autoregressive models has useful skill for 30–36 months. A 1993–1994 La Niña event is predicted based on data through February 1992.