Interannual Variations in the Southern Hemisphere Circulation

An analysis has been made of variations in the atmospheric circulation with periods of more than 50 days occurring in the Southern Hemisphere between 1980 and 1986, using analyses produced by the European Centre for Medium Range Weather Forecasts (ECMWF), mainly at the 500 hPa level. A smoothed mean annual cycle was removed from the data to define anomalies which were then filtered to retain variations with periods greater than 50 days. The >50 day period variations account for 20%–30% of the variance of daily 500 hPa geopotential anomalies at midlatitudes, up to 50% over Antarctica and more than 70% in the tropics. The total variance was found to be more uniform at high latitudes over the southern oceans than reported in previous studies, possibly reflecting the more comprehensive dataset used in the ECMWF analyses. The first three empirical orthogonal functions (EOFs) of the geopotential height anomalies amounted for 40% of the low-pass filtered fields. The first is a zonal field with the sign of the anomalies reversing near 60°S and a three-wave pattern superimposed at midlatitudes. The second and third have nearly equal eigenvalues and resemble two interleaved wave trains at middle and high latitudes. The first EOF is seen to be related to barotropic variations in the zonally averaged westerly wind which are preceded a few days earlier by changes in the poleward eddy momentum flux. The concomitant changes in energy integrals and conversions averaged over the hemisphere are small. While time series of the EOF coefficients show persistent anomalies of up to 4 months duration, the correlations between the series are small and add nothing to the predictability resulting from persistence over the 3 week period in which their autocorrelations remain positive. Plots of both the trajectories of the daily analyses and the distribution of persistent events in phase space, projected on to the planes of the principal axes of variation, showed little or no indication of concentration about multiple equilibria. Cluster analysis also failed to compress the original distribution to a small number of clusters unless some totally uncorrelated states were grouped together. Taken overall, these results are not encouraging to those seeking to develop long-range forecasting schemes for the Southern Hemisphere. Even with synoptic scale variations removed, the resulting anomaly patterns show great variety and little evidence for preferred paths between them. Apparently, on this time scale at least, the atmosphere does not conform to the predictions of nonlinear theory or to the behavior of simple nonlinear models.