Does the coupling of the mesospheric semiannual oscillation with the quasi-biennial oscillation provide predictability of Antarctic sudden stratospheric warmings?

During September 2019 there was a sudden stratospheric warming over Antarctica, which brought disruption to the usually stable winter vortex. The mesospheric winds reversed and temperatures in the stratosphere rose by over 50 K. Whilst this was only the second SSW in the Southern Hemisphere (SH), the other having occurred in 2002, its Northern counterpart experiences about six per decade. Currently, an amplification of atmospheric waves during winter is thought to trigger SSWs. Our understanding, however, remains incomplete, especially with regards to its occurrence in the SH. Here, we investigate the 5 interaction of two equatorial atmospheric modes, the Quasi Biennial Oscillation (QBO) and the Semiannual Oscillation (SAO) during the SH winters of 2019 and 2002. Using MERRA-2 reanalysis data we find that the two modes interact at low latitudes during their easterly phases in the early winter, forming a zero wind line that stretches from the lower stratosphere into the mesosphere. This influences the meridional wave guide, resulting in easterly momentum being deposited in the mesosphere throughout the polar winter, reducing the magnitude of the westerly winds. As the winter progresses these features descend 10 into the stratosphere, until SSW conditions are reached. We find similar behaviour in two other years leading to delayed dynamical disruptions later in the spring. The timing and magnitude of the SAO and the extent of the upper stratospheric easterly QBO signal, that results in the SAO-QBO interaction, was found to be unique in these years, when compared to the years with a similar QBO phase. We propose that this early winter behaviour may be a key physical process in decelerating the mesospheric winds which may precondition the Southern atmosphere for a SSW. Thus the early winter equatorial upper 15 stratosphere-mesosphere together with the polar mesosphere may provide critical early clues to an imminent SH SSW.

Northern Hemisphere SSWs. Their modelling study found that SSWs were only reproduced well when the flow in the equatorial upper stratosphere was constrained, simulating the two atmospheric modes in this region, the SAO and the QBO. Similar results were previously presented by Pascoe et al. (2006): In a troposphere-stratosphere-mesosphere global circulation model with forced QBO and SAO like variability, the timing of the NH mid winter warming advanced by about one month.
Whilst many studies have investigated the troposphere for answers to the questions raised by SSWs, we are now beginning 100 to see suggestions that we should also be looking at the upper atmosphere to understand the drivers of SSWs. The works of Pascoe et al. (2006) and Gray et al. (2020) discussed above, draws attention to the upper atmosphere in the formation of a SSW, with a focus on the NH.
Here, we report on the interaction of the QBO and SAO in the Southern Hemisphere during the winters of 2002 and 2019 (the two years with SH SSWs), based on reanalysis data. Both years exhibited clear easterly QBO conditions during the polar 105 winter, leading to a comparison with other easterly QBO years in the SH. We find an interaction between the QBO and SAO during the austral winters of 2019 and 2002 that is unique in its timing and extent. Coinciding with the SAO-QBO interaction is an intensification of atmospheric wave propagation, which deposit easterly momentum in the upper atmosphere throughout the two winters, leading to a disturbed SH polar stratosphere and the observed SSW events. MERRA-2 has a horizontal resolution of 0.5 • × 0.625 • with 42 vertical levels from the surface to 0.01 hPa (Gelaro et al., 2017). To investigate the connections between the SAO, QBO and SSW we used the four-times-daily zonal wind, geopotential 115 height and temperature information of MERRA-2, averaged into daily means. We focus on the vertical pressure range of 550 to 0.1 hPa and the austral winter (June-July-August-September-October, JJASO). Streams (Bosilovich et al., 2016). The first three covered the periods 1980-1991 (stream 100), 1992-2000 (stream 200) and 2001-2010 (stream 300), and the final stream from 2011-present (stream 400). Each stream had initial conditions derived from MERRA with a subsequent single year spin-up period, details of the process can be In later analysis, were results from several years are averaged, the averaging is based on the streams. All years presented here were also analysed individually. The stream analysis ensures that decadal variability of the SAO and QBO interactions are not lost in a large average average, as the initial conditions change across the streams in MERRA2.

Semiannual Oscillation
The SAO is known to have a period of six months, but it has appreciable inter-annual variability (Smith et al., 2020). Here, we focus our investigation on the easterly SAO maxima that occurs close to 1 hPa during the Southern Hemisphere winter. At 1 hPa MERRA-2 has been found to represent the easterly SAO in qualitative agreement with satellite derived winds (Kawatani et al., 2020), giving confidence that the SAO representation is reasonably realistic, particularly for the changes from westerly to easterly phases, and their propagation.

Quasi-biennial Oscillation
To analyse SAO and QBO interactions in the upper stratosphere, we focus on years with easterly QBO (eQBO) phase specifically in the equatorial upper stratosphere in the MERRA-2 zonal mean zonal wind. Analogous to Rao et al. (2020), we take the QBO phase at the 10 hPa pressure level, which Rao et al. have shown to provide predictability in the SH SSW cases.
QBO structure and dynamics in MERRA-2 reanalysis is discussed in detail by Coy et al. (2016), who conclude that MERRA-2 135 displays a realistic QBO behaviour in zonal winds. We verified this by contrasting to sonde observations of zonal wind from Singapore and found the two to be consistent, as expected (Coy et al., 2016).  To contrast the two SSW years to others with similar large scale equatorial flow conditions, other years with equivalent, i.e. eQBO phase, conditions during the austral winter months were analysed. The eQBO years in the MERRA-2 period were 1980, 1983, 1988, 1990, 1993, 1995, 1997, 1999, 2004, 2006, 2008, 2011, 2014, and 2017. In later analysis, the easterly QBO years have been split by model stream: i.e. 1980, 1983, 1985 and 1990 are stream 100; 1993, 1995, 1997 and 1999 are stream 200; 2004, 2006 and 2008 are stream 300; and finally, 2011, 2014 are stream 400. 145 We note that all years were analysed individually as well as in groups based on streams.
The years 1988 and 2017 are left out of these groups as their dynamics were found to be unique, all experienced mesospheric wind reversals in October. These years were thus analysed individually, and will be discussed separately from the other eQBO years. In 2017 the polar vortex has been reported to have experienced a disruption due to enhanced planetary wave activity throughout winter (Klekociuk et al., 2020). This lead to a smaller than average spring ozone hole (Klekociuk et al., 2020). 150 There have also been reports of an SSW occurrence in 1988 (Kanzawa and Kawaguchi, 1990), but to our knowledge this has not been verified subsequently.

Wave propagation
We calculate the Eliassen-Palm flux (EP flux) from MERRA2 fields to visualise wave propagation and momentum deposit.
The EP flux is a vector in the meridional plane and its direction and magnitude portray the relative importance of the eddy heat The upward (F y ) and meridional (F φ ) components of EP flux are: where f is the Coriolis parameter, dθ/dp is the change of potential temperature θ with respect to pressure p. u and v are the zonal and meridional winds, respectively and a is the radius of the Earth. Overbar denotes a mean and indicates deviation from the mean of the parameter in question.
The divergence of the EP flux indicates when and where momentum is being deposited. The convergence (negative val-  we now proceed to investigate other potential occurrences of this type of coupling during SH winter months. SAO is known to occur regularly, but with appreciable inter-annual variability (Smith et al., 2020). The QBO on the other hand has an average period of 28 months.

Remaining eQBO years
Here, we present similar analysis to Figures 3 and 6, however, instead of individual years, the averaging is now based on the MERRA-2 streams (Bosilovich et al., 2016) described in section 2.1. Note that all years were initially analysed individuallythe stream averages were found to be representative of the individual years, and no SSW like behaviour was observed for the 245 individual years. Figure 9 shows the zonal wind, EP flux and EP flux convergence averaged over June 15-21, July 6-12, August 10-16 and August 31 -September 6, averaged for the years 1980, 1983, 1985 and 1990. In Figures 9a and b, the SAO wind pattern is noticeable, but not as distinct as before, above 1 hPa. Figures 9c and d   The 1980s and 1990s were characterised by having both a smaller QBO and SAO which only occasionally interacted in late winter. Whilst the 2000s had a smaller QBO and larger SAO, which interacted but did not produce a zero wind line similar to the SSW years. The 2010s had a large eQBO, with a smaller SAO which did interact in July, however, no SSW was produced 275 in September. This suggests that not only may the particular phases be important for preconditioning the area for a SSW, but there vertical and poleward extents (and thus any mechanisms influencing these) seem to also be a factor.

Discussion
The  The SAO-QBO interaction is not unique to 2019 and 2002 and was found to happen during other easterly QBO years. However, apart from the two SSW-like years of 1988 and 2017, the timing and extent of the zero wind line was not found to occur in these other years. We suggest that this may be a reflection of variations not only in QBO but also in the amplitude and descent pattern of the SAO, the latter of which, to our knowledge, are not well understood (see e.g. Moss et al., 2016;Kawatani et al., 2020). 315 We propose that this early winter behaviour may be a key physical process in decelerating the mesospheric winds, which may precondition the atmosphere for a SSW. It may also help explain why SSWs are less common in the Southern Hemisphere: the early and large SAO-QBO interaction is dependent on both the QBO being in the correct phase, and the SAO appears to need