show large interannual variations, as well as some fluctuations on decadal time scales. The interannual variations of the South Pacific anticyclone are associated with the Southern Oscillation, and they are larger than those of the anticyclones over the other two SH ocean basins. The index used for the strength of the southern westerlies is also a measure of the high-latitude mode described earlier. It shows pronounced interannual and longer-term variations.
Allan and Haylock (1993) have used the J91 MSLP data set to identify the SH circulation variations associated with the long-term winter rainfall decrease over southwestern Australia. In the Australasian region, they describe variations of MSLP on decadal time scales as well as a longer-time-scale trend, both of which are associated with the regional rainfall variations.
Several indices of the large-scale surface-pressure field in the SH have been used to investigate variations of the SH circulation (Troup, 1965; Trenberth, 1976; Pittock, 1984). The index most frequently used is the Southern Oscillation Index (SOI), the difference between normalized mean sea-level pressure anomalies for Tahiti and those for Darwin. This index represents the magnitude of the pressure oscillation between the eastern Pacific Ocean and Indian Ocean regions associated with the Southern Oscillation. Pittock (1984) defined the Trans-Polar Index (TPI) as the pressure anomaly at Hobart (43°S, 147°E) minus that at Stanley (52°S, 58°W), on the opposite side of the hemisphere. The TPI represents variations of the amplitude of zonal wave-number-one asymmetries in the pressure field. It probably captures variations in the high-latitude mode to some extent, because of the difference between the latitudes of the two stations. Pittock (1984) has shown that the two dominant modes of common interannual variations of rainfall in Australia, New Zealand, South Africa, and South America are associated with variations of the SOI and the TPI, and that there is little relationship between these two indices. Pittock considered only 30 years of data, 1931 to 1960, and did not directly examine interdecadal variations of these indices and hence of the SH circulation. However, in commenting on several other studies that have considered relatively long time series of the SOI, he noted that there were significant differences in the relationship between the SOI and Australian rainfall between the periods 1930 to 1950 and 1950 to 1980, with a much weaker relationship in the 1930-to-1950 period.
Several recent studies (Hamilton and Garcia, 1986; Elliott and Angell, 1988; Allan et al., 1991; Quinn and Neal, 1992) have examined decadal variability of the Southern Oscillation; it is covered in some detail in Cane et al. (1995), Keeling and Whorf (1995), Rasmusson et al. (1995), and Trenberth and Hurrell (1995), all in this volume. It is sufficient to say here that decadal variations in the Southern Oscillation must be associated with significant decadal variations in the SH circulation.
Time series of the SOI and the TPI since 1896 are shown in Figure 5. There is negligible correlation of the interannual variations between these two time series over the past century, confirming the conclusion of Pittock (1984) that they represent independent modes of variation of the SH circulation. The SOI and TPI show large interannual fluctuations as well as variations on longer time scales. The TPI exhibits decadal variations that are a larger fraction of its interannual variability than does the SOI. The decadal time-scale variations of these two indices must be associated with variations of the SH circulation on these time scales and, almost certainly, with regional climate fluctuations as well.
It is not possible to use the available surface meteorological observations for the SH to describe variations at century or longer time scales in the SH. The only source of data at these longer time scales is proxy data from tree rings or other sources from a very few sites in the SH. Cook et al. (1991) and Villalba et al. (1989) have used tree rings to reconstruct temperature variations at some SH sites, while Lough and Fritts (1990) have used tree rings to identify