Predicting the State of the Southern Oscillation Using Principal Oscillation Pattern Analysis

Abstract

Principal oscillation pattern (POP) analysis is a diagnostic technique for deriving the space-time characteristics of a dataset objectively. A multiyear dataset of monthly mean sea level pressure (SLP) in the area 15°S to 40°S is examined with the POP technique. In the low-frequency band one physically significant pair of patterns is identified, which is clearly associated with the Southern Oscillation (SO). According to the POP analysis, the 50 may be described as a damped oscillatory sequence of patterns ??P1?P2?-P1?-P2?P1? having a time scale of two to three years. The first pattern, P1, representative of the ?peak? phase of ENSO, exhibits a dipole with anomalies of opposite sign over the central and eastern Pacific and over the Indian Ocean/Australian sector. The second, P2, pattern is dominated by an anomaly in the SPCZ region and describes an intermediate, or ?onset? phase. The time coefficients of the two patterns, P1 and P2, may be interpreted as a bivariate index of the SO. Generalizing the original diagnostic concept, the POP framework is used to predict this index and the traditional univariate SO index. The POP prediction scheme is tested in a series of hindcast experiments. The scheme turns out to be skillful for a lead time of two to three seasons. In terms of a correlation skill score, the POP model is better than persistence and a conventional ARMA model in hindcasting the traditional SO index.