Karoly (1989b), 500 hPa height in most other studies), differences in time filters used, or seasonal stratification. The main data sets used have been the Australian SH analyses, available since 1972 (Karoly, 1989b; Rogers and van Loon, 1982; MW) and the European Centre for Medium-range Weather Forecasts (ECMWF) analyses (K88 and many intraseasonal studies), which are available only for 1979 and later but are likely to be of higher quality than the Australian analyses.
In the following discussions, the typical modes of interannual variations are described first, using results from Karoly (1989b). A rotated PCA has been carried out on monthly anomalies of 300 hPa height using 15 years of Australian analyses (1973-1987). Three different types of horizontal structure have been identified for the low-frequency variations in the SH. The leading modes are approximately zonally symmetric, occur in both summer and winter, and explain about a quarter of the variance. They are associated with opposite departures of height for middle compared with high latitudes and for the tropics compared with middle latitudes. The other stable modes are hemispheric-scale wave-like variations, either zonal wave trains around the hemisphere or regional meridional wave trains, both of which seem to be less important than the symmetric modes. All the modes have an equivalent-barotropic vertical structure in the extratropics, and generally are more horizontally extensive than the typical modes of low-frequency variation in the Northern Hemisphere. The equivalent-barotropic vertical structure of low-frequency variations in the SH was identified using radiosonde data by Szeredi and Karoly (1987a).
The two leading modes from the rotated PCA of monthly anomalies of 300 hPa height are mainly zonally symmetric, representing opposite height variations over different latitude bands. These modes are found in both summer and winter. One mode represents a north-south out-of-phase variation of height between Antarctica and middle latitudes; it will be referred to hereafter as the high-latitude mode. The horizontal structure of this mode is shown in Figure 1a for winter, using correlations of the time series of PC scores for this mode with monthly SH height anomalies. The proportion of the total monthly variance explained by this mode is about 12 percent. Although it is primarily zonally symmetric, there are substantial asymmetries, with larger correlations and gradients over the Indian Ocean and south of Australia. The middle latitudes show evidence of zonal asymmetry resembling a wave-number-three pattern. This high-latitude mode has an equivalent-barotropic vertical structure with very similar variations in both 500 hPa height and MSLP.
This mode of low-frequency variation in the SH has
been identified by SK using station data; by Rogers and van Loon (1982), MW, and K88 using shorter time periods of analyses; by Shiotani (1990), Kidson (1988a), and Trenberth (1979) using zonal-mean fields; and by Trenberth and Christy (1985) using zonal-mean surface pressure. SK found that it was the second mode of monthly variations of the station data, explaining a smaller fraction of variance than