On the Evolution of the Southern Oscillation

Abstract

The evolution of the Southern Oscillation (SO) is examined in the time domain by computing lagged cross correlations between sea level pressures at Darwin and sea level or surface pressures at selected stations. Also, in the Northern Hemisphere, the historical and U.S. Navy sea level pressure analyses are used. All monthly time series are low-pass filtered to retain periodicities greater than 20 months in order to highlight the interannual fluctuations which are primarily associated with the SO. A detailed analysis of the post-1941 period results in plotted maps of the phase (lead or lag) and magnitude of the maximum cross correlations with Darwin, in a manner analogous to a broadband coherence and phase spectrum. The relationships within the SO are further examined, where possible, back to 1882 using time series of running decadal cross covariances. The dominant pattern reveals the two poles of the traditional standing oscillation or seesaw of the SO, with centers of opposite sign over Indonesia and the central South Pacific Ocean. But there are significant phase variations within each center and clear indications that changes over the South Pacific lead the opposite changes in the Indonesian pole by 1?2 seasons. Largest leads of three seasons begin near New Zealand but quickly spread over the subtropics of both hemispheres in the Pacific. Typically 1?3 seasons later, opposite anomalies begin over the Indian region and progress east and southeast into the western Pacific. Significant positive and negative lagged correlations occur only in the New Zealand area. For the post-1950 period, which is the basis for most recent analyses of El Niño?SO events, the SO was dominated by a three-six year quasi-periodicity which leads to ambiguity in interpreting phase relationships. The pattern of leads and lags is consistent with a progression of anomalies from southeast Australia across New Zealand and into the Pacific about two years later. The progression is not very regular, often occurring in discrete jumps. Moreover, it requires reinterpretation of the negative correlations as positive correlations that are half a period (π radians) out of phase. Eastward propagation is likely to be exaggerated by the implied cyclicity imposed by analyses in the frequency domain. Over the longer term (1882?1984) the ambiguity is lessened and the two poles are seen to be more distinct. Systematic leads are still apparent over the subtropics of the Pacific but the evidence for an eastward-propagating component extending from Australia across the southwest Pacific is not consistent throughout the record. The results show that caution must be exercised in interpreting the post-1950 period as representative of the long-term mean behavior. The importance of the tendency for changes over the South Pacific to lead the SO lies in the probable role of associated processes in setting up tropical sea surface temperature anomalies, especially during the onset stage of El Niño events. The South Pacific Convergence Zone (SPCZ) is regarded as a key feature, and the physical mechanisms likely to be important are discussed. It is noted that the SO and El Niño events do not always coincide. Tropical Pacific sea surface temperatures can be anomalously warm without a change in the SO, apparently provided that the SPCZ is not involved to any extent. However, global-scale atmospheric teleconnections are primarily associated with the SO.