5 Discussion and conclusions
In this study we assessed extratropical Southern Hemisphere zonally asymmetric circulation in austral spring. For this purpose, we derived two complex indices using Complex Empirical Orthogonal Functions and used to characterise both amplitude and phase of planetary waves.
The cEOF1 represents the variability of the zonal wave 1 in the stratosphere and is closely related to stratospheric variability such as anomalies in Total Column Ozone. Otherwise, this complex EOF is not related with SST variability and continental precipitation in the Southern Hemisphere. On the other hand, the cEOF2 represents a wave-3 pattern with maximum magnitude in the Pacific sector, that is an alternative representation of the PSA1 and PSA2 patterns (Mo and Paegle 2001). The 90º cEOF2 can be identified with the PSA1 and the 0º cEOF2 with the PSA2. While the cEOF2 variability is related to surface impacts, the cEOF1 surface influence is almost negligible. For instance, precipitation anomalies in South America associated with the 90º cEOF2 show a clear ENSO-like impact, with positive anomalies in South-eastern South America, negative anomalies in Southern Brazil and positive anomalies in central Chile for positive 90º cEOF2 phase.
Variability patterns that arise from cEOF methodology describe the zonally asymmetric springtime extratropical SH circulation, reproducing previous features such as the variability related to PSAs or A-SAM.
Since the spatial fields obtained from both components of cEOF2, which resemble PSA patterns, are in quadrature by construction, the cEOF methodology allows to derive, for the first time to our knowledge, a joint PSA index from the resulted amplitude and phase. These patterns are not forced to be orthogonal to other modes of circulation, like they are in standard EOF methodology. This allows us to show for example, that the 90º cEOF2, corresponding to PSA1 variability, is closely associated with the SAM in the troposphere. Previous research in the SAM–PSA relationship had the issue that the SAM and the PSA patterns are not independently derived and so the correlation between these indices had to be zero by construction (e.g. Yu et al. 2015).
Most studies on the relationship between ENSO and SAM rely on correlations between an ENSO index and the SAM index Cai, Sullivan, and Cowan (2011) or between the SAM index and other variables associated with tropical convection, such as OLR or tropical SSTS (e.g. Carvalho, Jones, and Ambrizzi 2005). However, Campitelli, Díaz, and Vera (2022a) showed that the correlation between ENSO and SAM is almost completely explained by the asymmetric component of the SAM. In this work we show that the asymmetric component of the SAM can be identified with the PSA1. Therefore, the correlation between ENSO and SAM in SON is predominantly the correlation between ENSO and PSA1, at least in SON. This sheds new light into the previous literature, as it cannot be assumed that a high correlation between ENSO and SAM indexes indicates a relationship between ENSO and zonally symmetric variability.
Further investigation is necessary to determine the connection between the symmetric component of the SAM and the PSA. It is possible that the PSA may force a zonally symmetric response (or vice versa) via wave-zonal mean flow interactions (Kim and Lee 2004), or that this correlation is simply a statistical artefact resulting from the EOF methodology used to define the SAM and the fact that the spatial structure of the PSA projects onto the spatial structure of the symmetric SAM.
Irving and Simmonds (2016) argued that there is some disagreement in the literature of whether the phase of the PSA pattern is affected by the location of tropical SST anomalies. With the methodology used in this study, we were able to show not only that the cEOF2 tends to be in the positive or negative 90º phase (~PSA1) when the ENSO region is warm or cold, respectively, but also that central Pacific SST anomalies tend to align the cEOF towards the negative 0º phase and eastern Pacific SST anomalies tend to align it towards positive 0º phase. When ENSO phase is neutral, the cEOF2 is still as active, but with no preferred phase. The latter agrees with the results of Cai and Watterson (2002), who showed that the CSIRO Model can develop PSA-like variability even in the absence of ENSO forcing (i.e. with a climatological run), but that the variability of one of the PSA modes was enhanced when adding the ENSO signal. The sensitivity of the phase of the PSAs to the location of the tropical SST anomalies was also seen by Ciasto, Simpkins, and England (2015), who detected similar Rossby wave patterns associated with central Pacific and eastern Pacific SST anomalies but with a change in phase.
The method used in this study has similarities to the one used by Goyal et al. (2022) as they construct an index of amplitude and phase of zonal wave 3-like variability by combining the two leading EOFs of meridional wind anomalies. The patterns obtained by them bear high resemblance with cEOF2. Although a detailed comparison is out of scope for this paper, the cEOF analysis has the advantage of constructing the indices based on patterns that are exactly in quadrature by construction.
The methodology proposed in this study allows for a deeper understanding of the zonally asymmetric springtime extratropical SH circulation such as a better description of PSA like variability using a unique complex index and the understanding of relationship between PSAs and ENSO or SAM variability. Further work should extend this analysis to other seasons and further study the relationship between the cEOF2 and the SAM.