This work presents the first lidar observations of a Quasi‐Biennial Oscillation (QBO) in the interannual variations of stratospheric gravity wave potential energy density (*E*_{pm} in 30–50 km) at McMurdo (77.84°S, 166.67°E), Antarctica. This paper also reports the first identification of QBO signals in the distance between McMurdo and the polar vortex edge. Midwinter stratospheric gravity wave activity is stronger during the QBO easterly phase when the June polar vortex expands and the polar night jet shifts equatorward. During the QBO westerly phase, gravity wave activity is weaker when the polar vortex contracts and the polar night jet moves poleward. Nine years of lidar data (2011–2019) exhibit the mean *E*_{pm} winter maxima being ~43% higher during QBO easterly than westerly. The June polar vortex edge at 45 km altitude moves equatorward/poleward during QBO easterly/westerly phases with ~8° latitude differences (39.7°S vs. 47.7°S) as revealed in 21 years of MERRA‐2 data (1999–2019). We hypothesize that an equatorward shifted polar vortex corresponds to less critical level filtering of gravity waves and thus higher *E*_{pm} at McMurdo. The critical level filtering is characterized by wind rotation angle (WRA), and we find a linear correlation between the WRA and *E*_{pm} interannual variations. The results suggest that the QBO is likely controlling the interannual variations of the *E*_{pm} winter maxima over McMurdo via the critical level filtering. This observationally based study lays the groundwork for a rigorous numerical study that will provide robust statistics to better understand the mechanisms that link the tropical QBO to extratropical waves.

*E*_{pm} in QBO easterly winters

Atmospheric waves, such as gravity, tidal, and planetary waves, transport momentum and energy both vertically and horizontally, making significant contributions to the complex coupling among stratified layers from the troposphere to the thermosphere, between the Northern and Southern Hemispheres, and crossing over the polar, middle, and equatorial regions (e.g., Andrews et al., 1987; Becker, 2012, 2017; Forbes, 1995; Smith, 2012a, 2012b). Insufficient and inaccurate representation of gravity waves remains one of the leading errors in most general circulation models and chemical climate models (e.g., Alexander et al., 2010; Garcia et al., 2014, 2017; Geller et al., 2013; Kim et al., 2003; Liu, 2019; McLandress et al., 2006; Richter et al., 2010), because, due to their small scales and wide spectra, it is still very challenging to characterize gravity waves with observations and provide the physical basis for wave parameterizations and simulations. The McMurdo lidar campaign conducted by the University of Colorado Boulder was designed to help address this challenge by making high‐resolution, full diurnal cycle, and all‐year‐round observations at a very high southern latitude of ~78°S (Chu, Huang, et al., 2011; Chu, Yu, et al., 2011). This ongoing campaign has collected data for the last decade since late 2010. The lidar data have revealed McMurdo as a geophysical hot spot of atmospheric waves (e.g., Chen et al., 2013, 2016; Chen & Chu, 2017; Chu, Huang, et al., 2011; Chu, Yu, et al., 2011; Chu et al., 2018; Lu et al., 2015, 2017; Zhao et al., 2017), and the discovery of persistent gravity waves in the mesosphere and lower thermosphere (MLT) (Chen et al., 2016) have inspired systematic studies of gravity waves from the thermosphere to the stratosphere in Antarctica.

Much has been learned about short‐term gravity wave variations over McMurdo. For example, lidar observations reveal thermosphere‐ionosphere Fe (TIFe) layers with clear gravity wave signatures (periods ~1.5 hr) to at least 170 km in altitude (Chu, Yu, et al., 2011; Chu et al., 2016; Chu & Yu, 2017). Such 1.5–2 hr gravity waves have vertical wavelengths varying from ~ 20 km to over 70 km into the thermosphere (Chen, 2016; Chu, Yu, et al., 2011; Chu et al., 2012). Gravity waves with periods of 3–10 hr dominate the temperature perturbations in the MLT in such paramount ways (Chen et al., 2016; Chen & Chu, 2017) that no tidal signatures can be seen without averaging over several years of lidar data (Fong et al., 2014). Consequently, Chen et al. (2016) named these MLT waves the “persistent gravity waves.” Case studies of the persistent waves point their sources to the stratosphere (Chen et al., 2013). Studies of stratospheric gravity waves using 5 years of Fe lidar data from 2011 to 2015 exhibit a repeated seasonal pattern with summer minima and winter maxima in the gravity wave potential energy mass density (*E*_{pm}) between 30 and 50 km (Chu et al., 2018) and the seasonal variations in the wave periods and vertical wavelengths of stratospheric dominant gravity waves (Zhao et al., 2017). Chu et al. (2018) have found that the summer‐winter asymmetry in the *E*_{pm} variations are mainly caused by the critical level filtering, while the *E*_{pm} variations during winter months are positively correlated with the near‐surface and stratospheric winds. The derived horizontal wavelengths (~400 km) of dominant gravity waves in the stratosphere (Zhao et al., 2017) are significantly shorter than those (~2,000 km) of the persistent gravity waves in the MLT (Chen et al., 2013; Chen & Chu, 2017), leading to the conclusion that dominant gravity waves in the stratosphere are starkly different from the large‐scale persistent gravity waves in the MLT. Combining the McMurdo lidar observations (Chen et al., 2013, 2016; Chen & Chu, 2017; Chu et al., 2018; Zhao et al., 2017) with Kühlungsborn Mechanistic general Circulation Model (KMCM) numerical simulations (Becker & Vadas, 2018; Vadas & Becker, 2018) and updating the original theory of secondary gravity wave generation (Vadas et al., 2003), Vadas et al. (2018) proposed a “primary‐secondary” wave coupling picture in Antarctica to reconcile the observed large differences between the MLT and the stratospheric gravity waves. That is, the downslope winds blowing down from the East Antarctic plateau onto the Ross Sea generate orographic gravity waves with considerably large amplitudes. As these orographic gravity waves dissipate near the stratopause or in the lower mesosphere, momentum deposition results in localized and intermittent body forces. Such body forces excite the secondary gravity waves that possess scales several times larger than the primary orographic gravity waves (Vadas et al., 2018). These large‐scale secondary gravity waves can propagate and penetrate to the lower thermosphere, forming the persistent gravity waves. The secondary gravity wave signatures predicted by the theory in Vadas et al. (2018) were successfully identified in the McMurdo lidar data (Vadas et al., 2018) and in the KMCM simulations (Vadas & Becker, 2018).

Despite the above picture being quite complete in terms of short‐term variations and vertical coupling by gravity waves in Antarctica, there remain important unanswered questions concerning long‐term variations and horizontal coupling. For example, we still do not know what factors drive the interannual variations in the observed gravity wave *E*_{pm}. Chu et al. (2018) have noted from the 5 years of data that the stratospheric *E*_{pm} shows larger winter values in 2012 and 2015 than in the other three years (2011, 2013, and 2014) and suggested that the critical level filtering may play a role; however, the causes and mechanisms remained unresolved. Now we have learned that 2012 and 2015 correspond to the easterly phase of the equatorial Quasi‐Biennial Oscillation (QBO), while the other three years are in the westerly phase. Is the equatorial QBO responsible for the interannual variations of polar gravity waves? Signatures of the QBO at high latitudes have been observed since the 1960s following the discovery of the tropical oscillation (see a review by Anstey & Shepherd, 2014), and Holton and Tan (1980) proposed a mechanism, often referred to as the Holton‐Tan effect, to explain the dynamical coupling between low and high latitudes. It is known that QBO can affect the polar wind and temperature fields including the Antarctic (e.g., Baldwin & Dunkerton, 1998; Ford et al., 2009; Lu et al., 2019; Yamashita et al., 2018) and that background winds impact gravity waves substantially (e.g., Becker, 2012; Fritts & Alexander, 2003); therefore, it is plausible that the tropical QBO affects the very high latitude gravity wave activity. This clue leads us to investigate the interannual variations of *E*_{pm} and their correlation with the QBO as Part 3 of our studies of stratospheric gravity waves at McMurdo, following Part 1 (dominant gravity wave characteristics and their seasonal variations) in Zhao et al. (2017) and Part 2 (gravity wave potential energy densities and their seasonal variations) in Chu et al. (2018). With the aid of 21 years (1999–2019) of reanalysis data from the Modern Era Retrospective Analysis of Research and Applications, Version 2 (MERRA‐2; Bosilovich et al., 2015), we find in this study that the Antarctic polar vortex edge exhibits an equatorward shift corresponding to the enhanced gravity wave *E*_{pm} in QBO easterly June. Based on the results, we will propose an explanation for how the equatorial QBO may impact the polar gravity waves. This observational study lays the groundwork for future numerical studies that will provide more robust statistics to better understand the mechanisms that link the tropical QBO to extratropical gravity waves.

The McMurdo lidar observation campaigns have been conducted by the University of Colorado Boulder at Arrival Heights Observatory via collaboration between the United States Antarctic Program (USAP) and the Antarctica New Zealand (AntNZ) for nearly a decade. The first Fe Boltzmann lidar (Chu et al., 2002) was installed in the AntNZ Arrival Heights building and started the data collection in December 2010 (Chu, Huang, et al., 2011; Chu, Yu, et al., 2011). Seven years later the second lidar, a STAR Na Doppler lidar (Smith & Chu, 2015), was added to the same lidar lab, and simultaneous Fe and Na lidar data have been collected since January 2018. Nine years of Fe lidar data are chosen for this study because of their longer records than the Na lidar. Atmospheric temperature profiles are derived using the Rayleigh integration technique in the pure molecular scattering region (Hauchecorne & Chanin, 1980) and using the Fe Boltzmann technique in the meteoric Fe layer region (Chu et al., 2002; Gelbwachs, 1994). The raw photon counts of Fe lidar were collected in resolutions of 1 min and 48 m. To achieve sufficient signal‐to‐noise ratios (SNRs), we take the same methodology as in Chu et al. (2018); that is, the Rayleigh temperature data were retrieved with a temporal integration window of 2 hr and a spatial binning window of 0.96 km, and an oversampling method was applied to have steps of 1 hr and 0.96 km.

Over the course of 9 years (2011–2019), more than 9,000 hr of Fe lidar data were collected from Arrival Heights Observatory. Comprehensive data screening similar to that in Chu et al. (2018) and Zhao et al. (2017) removed data that have durations less than 6 hr or large temporal gaps (≥4 hr), or low SNRs (Zhao et al., 2017). Consequently, 7,143 hr of data qualify for this study. Statistics of the qualified lidar data are tabulated in Table 1. Qualified data further went through a data division process as described in Zhao et al. (2017) to create data segments between 6 and 12 hr in duration. As pointed out by Chu et al. (2018), segmenting the data in this way enriches statistical samples while preserving the persistent gravity waves with periods of 3–10 hr that are the dominant features in the temperature variations at McMurdo (Chen et al., 2016; Zhao et al., 2017). There are 646 total segments, and their distributions are listed in Table 2. All months, except for June 2018 and November 2016, have acquired qualified data samples for gravity wave studies. After the data screening and division process, the actual data used in this study total to ~6,941 hr, and their distributions of segment lengths and hours are summarized in Table 3.

About 70% of the segments have durations longer than 10 hr, and the short duration segments (6 hr) account for only ~5%. The data segmenting and the subsequent processing with background subtraction and spectral filtering, as outlined in Chu et al. (2018), indicates that gravity waves with periods of 3–9 hr and vertical wavelengths of 2–20 km are the major contributors to the derived gravity wave *E*_{pm}, while gravity waves with periods of 9–11 hr and vertical wavelengths of 20–30 km contribute partial energy to the *E*_{pm}. The 3 dB cutoff period and vertical wavelength are 11 hr and 30 km for the temporal and spatial filters used in the data processing. Such filters are set to preserve the persistent gravity wave spectra (Chen et al., 2013, 2016; Chen & Chu, 2017) as fully as possible while significantly removing semidiurnal tides, diurnal tides, and planetary wave components with periods over 1 day and vertical wavelengths over 40 km that are dominant in the stratosphere (Lu et al., 2013, 2017). As studied in Part 1 (Zhao et al., 2017), the dominant gravity waves in the stratosphere have similar periods (3–10 hr) as the persistent gravity waves in the MLT but with much shorter vertical wavelengths (3–16 km), and the mean wave periods and vertical wavelengths in the stratosphere vary from ~4.5 hr and ~5.5 km in summer to ~6 hr and ~8.5 km in winter. Detailed analysis of how the data processing and segmenting procedure affects gravity wave spectra and how the aforementioned spectral ranges are reached can be found in section 2.2.1 of Chu et al. (2018).

Two main physical quantities used in this study are the gravity wave potential energy per unit mass (*E*_{pm}, potential energy mass density) and potential energy per unit volume (*E*_{pv}, potential energy volume density). We use the same method as Chu et al. (2018) to calculate the altitude profile of *E*_{pm}(*Z*)*g* = 9.7 m/s^{2} corresponding to the stratosphere, *z* and *t* respectively represent altitude and time, *T*_{Bkg} is the monthly mean background temperature, *n*_{p} is the number of temperature perturbation profiles within the segment time span. Here *N*(*z*) is the buoyancy frequency calculated from the segment temporal mean temperatures *T*_{0}(*Z*) through*C*_{p} = 1,004 J/K/kg is the specific heat of dry air at constant pressure. We also compute the altitude profile of *E*_{pv}(*z*), for each observational segment, by multiplying *E*_{pm}(*z*) with the background atmospheric density *ρ*_{0}(*z*):

The procedures of deriving *E*_{pm} and *E*_{pv} from temperature profiles are quite involved, as the key steps include (1) deriving gravity wave‐induced temperature perturbations from the raw temperature data, and (2) accurately estimating the energy of gravity waves from the derived perturbations (Chu et al., 2018). The major challenges are how to retain gravity wave‐induced perturbations while removing the influences from other atmospheric waves (such as tides and planetary waves) and how to remove the contributions from noise variance to the total energy density. Chu et al. (2018) have done extensive investigations into these issues and provided thorough considerations and step‐by‐step procedures on the derivations. We closely followed the procedures outlined in section 2.2 of Chu et al. (2018), especially the background subtraction, spectral filtering, and applying the spectral proportion method to extract the potential energy density induced by gravity waves. Interested readers may refer to that article for more details.

The NASA MERRA‐2 reanalysis data (Bosilovich et al., 2015; Gelaro et al., 2017) provide the background temperature, zonal and meridional winds from the surface to the top boundary of 0.01 hPa (~80 km) for characterizing the QBO index, polar vortex edge distance and wind strength, and wind rotation angle (WRA). MERRA‐2 produces a realistic QBO in the zonal wind, mean meridional circulation, and ozone over the 1980–2015 time period (Coy et al., 2016). This work uses reanalysis data from 1999 to 2019 due to improved data quality following the addition of AMSU‐A and AMSU‐B upper air radiances to the data assimilation stream in 1999 (see Figure 8 in Fujiwara et al., 2017, where AMSU stands for Advanced Microwave Sounding Unit). The inclusion of Aura Microwave Limb Sounder temperatures in August 2004 further improved the data quality above 5 hPa (40 km). The temporal, horizontal, and vertical resolutions of the MERRA‐2 data used in this work are 6 hr, 0.5° latitude by 0.625° longitude, and 1–2 km, respectively.

The distance from McMurdo Station to the polar vortex edge is derived from the MERRA‐2 reanalysis data. The method to demark the Antarctic vortex edge is described by Harvey et al. (2002) and briefly described here. First, the scalar field “*Q*” (see their Figure 1) is calculated; *Q* is a relative measure of strain versus rotation in the horizontal flow; it is negative where rotation dominates and is positive in regions of strong horizontal wind shear. Areas of negative *Q* are located inside the polar vortex, while positive *Q* is located at the vortex edge on the flanks of the polar night jet (PNJ) stream. The vortex edge demarcation algorithm first integrates *Q* around stream function contours starting in the center of the vortex and incrementing outward. Neighboring stream function contours in which line‐integrated *Q* changes sign are saved as vortex edge “candidates.” The actual vortex edge is the candidate with the fastest line‐integrated wind speed. Illustrations of vortex edge location and the corresponding wind field and PNJ will be given in sections 4 and 5 (Figures 4c and 6–8). In the middle and upper stratosphere, this vortex demarcation algorithm is slightly more conservative but in good agreement (Meek et al., 2017) with the widely used “Nash” algorithm that is based on horizontal gradients in potential vorticity (Nash et al., 1996). In general, this method yields a vortex edge that is located along the poleward flank of the PNJ stream. Thus, in this work, the distance between McMurdo and the edge of the vortex can also be interpreted as the distance between McMurdo and the PNJ.

We first explore interannual variations in 9 years of lidar data on gravity wave potential energy density averaged over the altitude range of 30 to 50 km and within each qualified data segment. Figure 1 shows the resulting 646 data points of altitude mean *E*_{pm} and *E*_{pv} versus time from 1 January 2011 to 31 December 2019. Two notable features are (1) the large fluctuations of *E*_{pm} and *E*_{pv} from one segment to another and (2) the repeated seasonal cycle showing summer minima and winter maxima. Both features are consistent with the findings studied in Chu et al. (2018) who used the first 5 years of lidar data. The goal of this work is to explain the interannual variability in the winter maxima.

To characterize the seasonal and interannual variations, the red lines in Figures 1a and 1c show nine individual harmonic fittings of annual plus semiannual oscillations, similar to that in Chu et al. (2018), for *E*_{pm} and *E*_{pv}, respectively:*A*_{0} is the annual mean, *A*_{12} and *φ*_{12} are the amplitude and phase of annual oscillation, and *A*_{6} and *φ*_{6} are the amplitude and phase of semiannual oscillation. The fitting parameters and their errors are listed in Table 4. The red curves in Figures 1b and 1d represent the uniform fittings to all years. The annual means *A*_{0} in 2012 and 2015 are larger than the other seven years, and the amplitudes *A*_{12} of the annual variations in 2012, 2015, and 2018 are larger than in the other six years. The accumulated effects of larger annual means and larger annual amplitudes in the years with a QBO easterly phase are reflected in the fitted *E*_{pm} curves (see Figure 1a) that show higher midwinter maxima in 2012, 2015, and 2018 than the other years.

To quantify the interannual variations of the *E*_{pm} winter maxima occurring around the beginning of July, we take the monthly means in June and July from the nine individual annual fittings for comparison. The mean of all 18 monthly means is 6.20 ± 0.04 J/kg, while the maximum and minimum monthly means are 9.23 ± 0.05 J/kg in July 2015 and 4.75 ± 0.03 J/kg July 2013, respectively. Thus, the interannual variation from the maximum to the minimum is ~72.3%, which is computed as (maximum − minimum)/mean of 18 monthly means. The mean *E*_{pm} of the three winters (2012, 2015, and 2018) with a QBO easterly phase is 8.12 ± 0.03 J/kg, while the mean of the rest six winters in the QBO westerly phase is 5.23 ± 0.05 J/kg. Hence, the *E*_{pm} winter maxima are ~43% higher in the easterly phase than in the westerly phase of the QBO (the definition of QBO phases will be given in section 4). To inspect whether the interannual variations of *N*^{2} are sufficient to account for the observed *E*_{pm} variations, the harmonic fittings given by Equation 4 are applied to nine individual years of the buoyancy frequency square *N*^{2} in Figure 2a and to all years in Figure 2b. The fitting parameters and their errors are listed in Table 6, and the fitted results are plotted as red curves in Figure 2. The interannual variations of *N*^{2} winter maxima are less than 7%, which is insufficient to explain the *E*_{pm} interannual variations.

We now derive the mean annual cycle in *E*_{pm}, *E*_{pv}, and *N*^{2} in order to account for and subtract the annual cycle to isolate QBO signatures in section 4. Figure 3 shows 9 years of data binned every 5 days to form composite annual data sets of *E*_{pm}, *E*_{pv}, and *N*^{2}, respectively. Harmonic fittings of the annual plus semiannual oscillations given by Equation 4 are applied to the composite years for all three parameters. The fitting parameters and their errors are listed in Tables 4, 5, and 6 for *E*_{pm}, *E*_{pv}, and *N*^{2}, respectively, while the fitted results are shown as red curves in Figure 3. The 9‐year results are consistent with the *E*_{pm} and *N*^{2} composite annual variations derived from 5 years of data (2011–2015) shown in Chu et al. (2018). Winter maxima occurring at the beginning of July are evident in all three parameters.

The QBO was originally discovered as the alternation between the downward propagating easterly and westerly zonal wind regimes in the equatorial stratosphere (~16–50 km) with variable periods from 22 to 34 months (see a review by Baldwin et al., 2001). Singapore radiosondes have been used to characterize the QBO index since 1980, and the zonal‐mean zonal winds of MERRA‐2 at the equator give nearly identical results as the Singapore sondes (

We investigate the observed interannual variations of winter *E*_{pm} for possible signatures of the QBO. A QBO index, defined as the MERRA‐2 zonal‐mean zonal winds at the equator (0° latitude) averaged over the 40 and 50 hPa pressure levels, is plotted versus the 9 years of lidar‐observed *E*_{pm} data and harmonic fittings in Figure 4a. According to this QBO index, 2012, 2015, and 2018 midwinters are classified as QBOe, while 2011, 2013, 2014, 2017, and 2019 are classified as QBOw. The year 2016 is abnormal as stated above—although the QBO index at 40–50 hPa is negative, the majority of pressure levels classify the 2016 southern winter into the westerly phase (positive index). The Antarctic winters of 2012 and 2015 both occur in the QBO easterly phase and correspond to high *E*_{pm} peaks. Although the winter maximum in 2018 is not as high as the maxima in 2012 and 2015, likely due to the lack of lidar data in June 2018, it is still larger than the surrounding years like 2016, 2017, and 2019. The overall mean *E*_{pm} in June and July over three QBOe years (2012, 2015, and 2018) are ~43% higher than the overall mean during the QBOw years (the other six years), as computed in section 3. This *E*_{pm} interannual variation has a period of roughly 3 years, rather than the 28 month quasi‐biennial period (Baldwin et al., 2001). A wavelet analysis of the QBO index at 25 hPa in Lu et al. (2019) shows that the QBO period has evolved from ~24 to ~36 months between 2003 and 2018, which is consistent with the longer QBO period in the last decade of lidar observations.

To quantify the correlation between the gravity wave *E*_{pm} and QBO index, we compute the *E*_{pm} anomaly, wherein seasonal variations are removed. We use the harmonic fittings of nine individual years to represent the yearly *E*_{pm} values and the composite‐mean fit in Figure 3a as the mean annual cycle of *E*_{pm} and then subtract the mean annual cycle from the nine individual harmonic fittings to derive the *E*_{pm} anomalies. The derived *E*_{pm} anomalies of two winter months (June and July) are plotted as diamonds in Figure 4a, except for 2016 and June 2018 that are plotted as squares. These results show that the positive/negative anomalies align well with the negative/positive QBO index (easterly/westerly zonal winds at the equator), except during the abnormal year 2016. All positive *E*_{pm} anomalies occur in the three QBO easterly years (2012, 2015, and 2018). In Figure 4b we plot the monthly mean *E*_{pm} anomalies versus the monthly mean QBO index for June and July. Because 2016 is an abnormal year and 2018 lacks June data, Figure 4b excludes 2016 data and 2018 June data. Based on the remaining 15 data points, the linear correlation coefficient between the *E*_{pm} anomaly and the QBO index at 40–50 hPa is −0.766 with a confidence level of 99.9%. This statistically significant correlation between the gravity wave *E*_{pm} and QBO phase supports that the stratospheric gravity wave activity is stronger in the QBO easterly winters than in the westerly winters, at least during the last decade. The observed interannual variations of *E*_{pm} are likely due to the modulation by the QBO signals.

We now investigate whether the *E*_{pm} enhancement in the QBO easterly winters is related to the polar vortex changes and whether the Antarctic polar vortex is also affected by the equatorial QBO signals. This work is motivated by the findings shown in Figure 10 of Chu et al. (2018), namely, that the maximum *E*_{pm} values occur in June and July when McMurdo is deep inside the polar vortex. At the time of publication of Chu et al. (2018) the correlation of *E*_{pm} with McMurdo distance from the vortex edge was disputed as a coincidence with the peak of *E*_{pm} seasonal variations. Now in light of evidence of the QBO in the interannual variations of *E*_{pm}, it may not be a coincidence anymore. The shortest distance from McMurdo to the polar vortex edge is calculated on each day using the MERRA‐2 reanalysis data as described in section 2.3 and is termed “vortex edge distance.” Negative/positive vortex edge distances mean that McMurdo is inside/outside the polar vortex. Vortex edge distances are negative during the Antarctic winter due to McMurdo being located deep inside the core of the vortex. The 9‐year time series of vortex edge distance at 45 km is plotted against the time series of *E*_{pm} in Figure 4c. Visual inspection of these two time series suggests that winters with lower *E*_{pm} maxima, such as in 2013 and 2016, correspond to the shorter vortex distances, that is, less negative values that occur when the vortex edge is closer to the McMurdo Station, while winters with high *E*_{pm} maxima, such as in 2012, 2015 and 2018, correspond to the longer (more negative) vortex edge distances when the vortex edge is farther away from McMurdo Station. Figure 4d shows the mean *E*_{pm} anomaly in June versus the mean vortex edge distance in June. A linear fit gives a correlation coefficient of −0.52 with a confidence level of 85% when all 9 years of lidar data are included. If June 2018 data is excluded because of the lack of lidar observations in that month, then the correlation coefficient becomes −0.75 with a confidence level of 96.7%. Figure 4d indicates that with or without June 2018 data the negative correlation between *E*_{pm} anomalies and polar vortex edge distances is clear. The June 2018 fitting data is likely an outlier, which was very likely biased toward smaller values due to the lack of observational data at a time when *E*_{pm} is usually the highest. While a longer data record is needed, Figures 4c and 4d indicate a possible QBO signal in the polar vortex edge location and a correlation between the *E*_{pm} anomaly and the vortex edge distance.

To further investigate the QBO signals in the polar vortex location, we plot the time series of 21 years (1999–2019) of vortex edge distance at 45 km in the upper portion of Figure 5a. The vortex edge distance in June (red color) is farther (more negative edge distances) in the QBOe (negative QBO winds) and closer (less negative edge distances) in the QBOw (positive QBO winds). Three different QBO indices are shown in the bottom portion of Figures 5a–5c. The new QBO index in Figure 5a is defined as the zonal‐mean zonal wind at the equator averaged over pressure levels of 20, 30, 40, and 50 hPa, while the QBO indices in Figures 5b and 5c are the zonal winds averaged over pressure levels of 20 and 30 hPa and over pressure levels of 40 and 50 hPa, respectively. As mentioned earlier, different QBO indices defined at different pressure levels can alter the QBO easterly and westerly classifications for some June months. For example, 2000, 2009, 2011, 2014, 2015, 2016, and 2017 flip signs between the 20–30 hPa QBO index of Figure 5b and the 40–50 hPa QBO index of Figure 5c. At the same time, 2001, 2003, 2005, 2007, 2010, 2012, and 2018 remain in the QBO easterly phase, while 1999, 2002, 2004, 2006, 2008, 2013, and 2019 remain in the QBO westerly phase under all three QBO indices. The monthly means of vortex edge distance in June at 45 km altitude are plotted against the June means of three different QBO indices in Figures 5d–5f, respectively. All three figures show positive correlations between the vortex edge distance and the QBO indices, that is, more negative vortex edge distances (McMurdo is inside and farther from the vortex edge) correspond to more negative (easterly) QBO winds. Likewise, less negative vortex edge distances (McMurdo is inside but closer to the vortex edge) correspond to more positive (westerly) QBO winds. The linear correlation coefficients are 0.846, 0.768, and 0.652 with confidence levels of 100%, 99.995%, and 99.863%, respectively. Over this 21‐year period, the QBO signals in the polar vortex edge location are robust and have statistical significance. The new QBO index given in Figure 5a has the highest correlation with the vortex edge distance. This new index is the integration (i.e., the mean) of all four pressure levels of 20, 30, 40, and 50 hPa, reflecting the downward progression of the equatorial zonal wind better than other two indices. This analysis convincingly shows that the polar vortex edge moves equatorward in the QBO easterly winters and moves poleward in the QBO westerly winters.

Winters (June) with consistent easterly (2001, 2003, 2005, 2007, 2010, 2012, and 2018) or westerly (1999, 2002, 2004, 2006, 2008, 2013, and 2019) classifications (they do not flip signs for different QBO indices) are termed the “stable years” (stable winters) of QBOe or QBOw, while the rest (2000, 2009, 2011, 2014, 2015, 2016, and 2017) are called “transition years.” The set including both stable and transition years is referred to as the “all years” set. Using the vertically integrated QBO index given in Figure 5a, we classify the “all years” set of QBO easterly winters as (2000, 2001, 2003, 2005, 2007, 2009, 2010, 2012, 2014, 2015, 2017, and 2018) and the “all years” set of QBO westerly winters as (1999, 2002, 2004, 2006, 2008, 2011, 2013, 2016, and 2019). We now average the vortex edge distance and mean wind speed along the vortex edge within these two groups. Two trials are made—one averages over the stable years of QBOe and QBOw, and another averages over the “all years” sets for QBOe and QBOw. The “stable years” results are compared in Figures 6a and 6b, and the “all years” results are in Figures 6c and 6d for the vortex edge at 45 km altitude. It is clear in both cases that the polar vortex edge stays farther away from McMurdo in the QBO easterly years than in the westerly years during the midwinter, and the differences in vortex edge distance reach the maximum in June with a time span of about 2 months (from mid‐May to mid‐July). The differences in the means of the stable winters (about 1,200 km) are larger than those in the means of “all years” winters (about 900 km), as expected. The differences in the mean winds along the vortex edge are relatively small, but in general the mean winds in the QBOe are weaker than in the QBOw by about 10–20 m/s in June. Note that the QBO differences seen in vortex edge distance and vortex edge wind speed are seasonally dependent.

To further investigate the polar vortex changes, three panels of polar vortex maps in the Southern Hemisphere (SH) are plotted in Figure 7. All panels are derived from MERRA‐2 reanalysis data from 1999 to 2019 for June near 50 km altitude. The left panel shows the average June wind speed (in color) and Antarctic vortex edge (thick black contour) for QBO easterly years. The thin white contours are the vortex edge for the individual easterly years. The middle panel is the June average for QBO westerly years. The right panel is the difference (QBOe − QBOw) in wind speed (blue/red colors) and the mean vortex edge for easterly years (solid black) and for westerly years (dashed black). This right panel shows that the zonal wind over McMurdo (marked by a gray filled circle) is ~5–10 m/s faster in the QBOe than in the QBOw at 50 km. The blue shades from 40–70°S latitudes and the red shades equatorward of 40°S in the right panel of Figure 7 reflect an equatorward shift in the PNJ during QBO easterly years. In terms of the vortex edge, the June mean vortex edge is located at 39.7°S latitude during QBO easterly years and at 47.7°S latitude during QBO westerly years (right panel) when all years of QBOe and QBOw are included. Removing two sudden stratospheric warming (SSW) years (2002 and 2019) from QBOw, the mean vortex edge in June is located at 39.7°S and 47.2°S for QBOe and QBOw, respectively, making only a 6% difference. The difference of 7.5–8°S latitude between the QBOe and QBOw years is statistically significant and it confirms that the QBO signals in the vortex edge distance shown in Figures 5 and 6 are robust.

Our results indicate that the QBO signals in the Antarctic polar vortex edge location are largest in June. This finding prompts us to examine the PNJ in June in order to understand the environment surrounding the polar vortex. We average the MERRA‐2 zonal‐mean zonal winds (U‐bar) over the stable years of QBOe and QBOw, respectively, for June in Figures 8a1 and 8a2, and take the differences between them in Figure 8a3. This differentiation method is similar to that of Baldwin and Dunkerton (1998), Hitchman and Huesmann (2009), and Lu et al. (2019). As MERRA‐2 is a pressure level‐based model, the data are converted to pressure heights (z) using the equation *z* = *H* · log(*p*_{o}/*p*), where the scale height (*H*) is set to 7 km and *p*_{o} = 1,000 hPa. On these plots, the PNJ core is defined as the maximum zonal wind velocity (eastward) in the latitude‐altitude plane and is overlaid for each altitude on Figures 8a1 and 8a2, as well as its location reiterated and value displayed in Figures 8a4 and 8a5, respectively.

From the Figures 8a1 and 8a2, we can see that the maximum U‐bar occurs around 50 km altitude at −39° and −46° latitude for the QBOe and QBOw, respectively, showing that the PNJ core location moves equatorward in a QBO easterly phase as compared to a QBO westerly phase by ~7° latitude. Figure 8a3 displays the different responses of different regions to the changing QBO phase. Figure 8a4 shows that the maximum wind location varies with altitude, reaching equatorwardmost near the stratopause and lower mesosphere (~45–60 km) and slowly moving poleward with decreasing altitude to around −60° latitude in the lower stratosphere (~20–30 km). The equatorward shift during QBOe can again be seen here, but no major differences can be discerned in zonal wind velocity for any altitude between QBO phases in Figure 8a5. The shift in PNJ location seen in Figures 8a1–8a4 is a clear reflection of the vortex expansion during QBOe demonstrated above. During the QBO easterly phase, the polar vortex expands, pushing the PNJ toward the equator. It is worth mentioning that Yamashita et al. (2018) reported a seemingly opposite result in July that the QBO westerly phase corresponds to an equatorward vortex shift. However, further examination shows that a QBO index defined at 10–20 hPa in Yamashita et al. (2018) is almost opposite to the QBO phase classifications shown in our Figure 5a. Therefore, their results are basically consistent with our findings in June.

The easterly anomalies (QBOe − QBOw) in Figure 8a3 show a triple‐cell structure from the equator to the South Pole around the stratopause. The June zonal‐mean zonal winds in the QBOe are stronger in the equatorial to subtropical region (0–40°S), weaker in the midlatitude to high‐latitude region (45–70°S), and slightly stronger in the polar region (75–90°S) than those in the QBOw between 40 and 70 km altitude. Such pattern around the stratopause is similar to the dipole‐cell structure poleward of 30°S in August reported by Lu et al. (2019). This pattern also supports a finding of the northern‐southern hemispheric asymmetry in the QBO easterly anomalies of zonal‐mean zonal winds in Lu et al. (2019). The Northern Hemisphere QBO easterly anomalies in Figure 2 (D3) of Lu et al. (2019) reveal a two‐cell structure from the equator to the North Pole around the stratopause in January, in contrast to the triple‐cell structure shown here. Our tests have revealed that this triple‐cell pattern around the stratopause is robust in June, insensitive to including or removing years from the data set. Figure 8a3 shows negligible differences between QBOe and QBOw below 30 km in the polar region in June, which agrees with the August result of Baldwin and Dunkerton (1998). Columns 2–4 of Figure 8 are included to demonstrate that variability exists between years, but the QBO phase‐related behavior is still well represented by the “stable” years used in Column 1. The three displayed QBO easterly winters (2012, 2015, and 2018) have higher June and July *E*_{pm} values than other years and correspond to the equatorward shifted PNJ as compared to the adjacent QBO westerly winters (2013, 2016, and 2019). It can also be seen that the trends demonstrated in each row of plots are similar from column to column.

Why does an equatorward expanding vortex (in QBOe) correspond to enhanced *E*_{pm}? There may be several factors contributing to this correlation with critical level filtering as one of the major factors. The expanding vortex likely corresponds to a quieter polar region characterized by horizontal wind profiles being less variable with altitude. Quieter winters likely have less critical level filtering, thereby higher probability for gravity waves to reach the upper stratosphere. To evaluate this hypothesis, we investigate the interannual variations of WRA and the potential correlation to those of *E*_{pm}. WRA is a measure of the critical level filtering as discussed in Chu et al. (2018) and Yamashita et al. (2009). Following the denotation in Chu et al. (2018) we compute WRA near McMurdo as the maximum wind direction change from 11 to 30 km. Nine MERRA‐2 grid points near McMurdo within the spatial box of 77–78°S latitude by 166.25–167.5°E longitude are used in this computation. We first average the horizontal winds within the nine grid points for each altitude then trace the wind direction changes from the lowest to the highest altitudes under the assumption that the wind direction changes between two adjacent altitude bins (~1 km interval) do not exceed 180° and finally calculate the angle between the two outmost flanks of wind directions as the WRA. Figure 8b3 shows that the upper stratosphere and lower mesosphere zonal winds are weaker in QBOe (2012) than in QBOw (2013) from 60°S to 80°S, while the lower stratospheric and tropospheric zonal winds are stronger in QBOe than in QBOw, as can be seen from the positive cell around 75°S at 7 km. As a result, the WRA calculated from MERRA‐2 data is smaller and less variable in 2012 than in 2013 (see Figure 9a). Smaller and less variable WRAs in midwinter 2012 indicate less critical level filtering, thereby a higher probability for gravity waves to reach 30–50 km, leading to higher *E*_{pm} in 2012 than in 2013.

To investigate this relation further, the seven‐point running smoothed gravity wave *E*_{pm} is plotted in Figure 9a (red curves) to compare with the WRA (blue curves). The gaps in Nov 2016 and June 2018 are due to the lack of qualified lidar data. The anticorrelation between the *E*_{pm} and WRA is considered to be the main cause of the summer‐winter asymmetry in the seasonal variations of *E*_{pm} as studied in Chu et al. (2018). What we care about here is whether the interannual variations of WRA can account for the interannual variations of *E*_{pm} in the midwinter. For this purpose, we take the June means of the *E*_{pm} and WRA curves in Figure 9a, respectively, and plot their correlation in Figure 9b. The standard deviation of the WRA in June is plotted against the June mean *E*_{pm} in Figure 9c. Negative correlations of −0.82 and −0.76 are obtained with the confidence levels of 99% and 97% in Figures 9b and 9c, respectively. The standard deviation of WRA is used here to represent the stability of the background winds, which may be affected by the relative location between McMurdo and the vortex edge. When the polar vortex edge moves equatorward, thus McMurdo being well inside the polar vortex, the background winds are likely more stable, thereby lower standard deviation of WRA and less critical level filtering are expected. Plotting the WRA standard deviation in June against the vortex edge distance in June in Figure 9d, we do see a positive correlation of ~0.6, although the confidence level (91%) is less than 95%. The results above support the hypothesis that the interannual variations of *E*_{pm} is mainly determined by the interannual variations of WRA that is strongly influenced by the polar vortex location in the midwinter. The equatorward expansion of the polar vortex in the QBO easterly phase leads to quieter winters and more stable background winds that correspond to less critical level filtering and thus higher *E*_{pm} in the upper stratosphere. Lidar observations in Alaska also show significant interannual variations in gravity wave *E*_{pm} averaged over three winter months (December, January, and February) from 40 to 50 km (Triplett et al., 2017). The highest gravity wave activity occurred during quiet winters when no SSW occurred, and the lowest gravity wave activity in the disturbed winters when major SSWs occurred (Triplett et al., 2017). These Arctic results are basically consistent with our findings of higher *E*_{pm} occurring in quieter Antarctic winters (QBOe).

There are additional long‐term variations that affect the polar wind and temperature fields, thus influencing the interannual variations of gravity waves and polar vortex. Here we briefly discuss two major factors—the solar cycle and the El Niño–Southern Oscillation (ENSO). The Holton‐Tan effect, based on the Northern Hemisphere studies (Holton & Tan, 1988), emphasizes that the effective waveguide for upward and equatorward propagating planetary‐scale waves is altered by the tropical winds. The zonal wind structure in the QBO easterly phase focuses more wave activity toward the pole, where planetary waves converge and slow down the zonal‐mean flow. Hence, the Arctic polar winter vortex tends to be disturbed and weak when the QBO is easterly but undisturbed and strong when the QBO is westerly. Baldwin et al. (2001) point out that the QBO acts as predicted by Holton and Tan (1980) only in years with low solar activity but appears to reverse its behavior during years with high solar activity (Labitzke, 1987; Labitzke & van Loon, 1988; van Loon & Labitzke, 1994). Given the interhemispheric asymmetry in the QBO easterly anomalies (section 5.1), it is necessary to examine the response of QBO signals to solar activity in the SH. Analyzing the 21‐year time series in Figure 5a, 1999–2003 corresponds to a solar maximum, while 2006–2011 corresponds to a solar minimum; however, the equatorward shift of the SH polar vortex in the QBOe occurs during both the solar maximum and minimum periods. To the first order, the robust QBO signals in the latitudes of vortex edge and PNJ are not affected by solar activity.

ENSO has been shown to affect the latitude of the June PNJ (Yamashita et al., 2018). The ENSO index used in that work was calculated from the 5 month running mean data of the Nino 3.4 index (NINO3.4) (*E*_{pm}, we note a possible link between a particularly strong ENSO year and the year of our highest *E*_{pm} measurement (2015). However, there is no individual year correlation between ENSO and *E*_{pm} in the last decade. For example, 2019 corresponds to the second highest ENSO index among the nine years of 2011–2019 but one of the lowest *E*_{pm} winter maxima, while 2012 and 2018 correspond to the second and third highest *E*_{pm}, but the ENSO indices in these two years are lower than that in 2019. Therefore, it is most likely that the interannual variations of the gravity wave *E*_{pm} are mainly controlled by the QBO signals, not the ENSO effects, although the possible influence of ENSO on the highest *E*_{pm} year is noticeable and a longer data record is needed to draw definitive conclusions.

The benefit of using a longer data record can be demonstrated in the following student *t* tests. From the statistics point of view, the linear correlations obtained in Figures 4b, 4d, 5d–5f, and 9b–9d are the sample correlation coefficients that are calculated from *N* samples. To draw conclusions about populations, the student *t* test for the hypothesis of population correlation *ρ* = 0 is given by*r* is the sample correlation coefficient and *n* is the effective degree of freedom. Here *n* is estimated, following Zülicke and Peters (2010), as the ratio of the observational time length over the persistent time of geophysical processes. In the case of this work, a typical QBO period is 28 months, and the observation time lengths are 108 and 252 months (i.e., 9 and 21 years) for lidar and MERRA‐2 data sets, respectively. Assuming one QBO event is statistically independent from the other, the effective degree of freedom *n* is 3.86 and 9 for lidar observations and MERRA‐2 data, respectively. Substitute *n* values into Equation 5, we compute the *t* values and confidence levels and compare them with the sample correlations in Table 7. In this table, *r* is the sample correlation coefficient, *t* is the student *t* test value, and *P*_{r} and *P*_{t} are the confidence levels of the sample and population correlations, calculated with *N* and *n*, respectively. The *t* values in Table 7 are far from 0, rejecting the hypothesis of *ρ* = 0. The confidence levels of population correlations are smaller than those of the corresponding sample correlations, but they are still convincing probabilities. The student *t* test results support the conclusions of the QBO signals in the gravity wave *E*_{pm} and the polar vortex edge location. Figures 5d–5f have the confidence levels well above 95% due to much larger effective degree of freedom (*n* = 9), while the probabilities for other figures are slightly below 95% due to the much smaller *n* = 3.86. If a longer data record enables *n* = 6.4 (e.g., 15 years of lidar observations), then the confidence level will be improved above 95%.

Other possible factors, such as tropospheric and stratospheric wave sources and the Doppler shifting effects as discussed in Chu et al. (2018), may also affect the interannual variations of gravity wave *E*_{pm} if these factors have long‐term changes. Moreover, we make a few speculations here based on several interesting results from recent studies. (1) If the polar vortex acts as a barrier so that gravity waves cannot penetrate the vortex edge and propagate poleward from the midtropical/subtropical regions, the equatorward expansion of polar vortex edge will include larger areas within the polar vortex during QBOe than during QBOw, thereby more wave sources may be included, leading to higher chances for more waves to reach McMurdo. (2) Competing with this point is the finding made by Sato et al. (2009) that gravity waves propagating upward tend to focus on the axis of PNJ in the Antarctic winter season, and the focusing effect is largely dependent on the meridional gradient of the background zonal wind velocity (Hindley et al., 2019; Shibuya & Sato, 2019). The contraction of vortex area during QBOw will increase the meridional gradient of the zonal wind inside the polar vortex; thus, the focusing effects of GWs will become stronger during QBOw than during QBOe. The shorter distance from the PNJ to McMurdo and the more focusing of GWs into the PNJ may suggest higher *E*_{pm} over McMurdo during QBOw than during QBOe, seemingly controversial with the observations. (3) However, this gravity wave focusing effect may turn out to be supportive of the observations if we consider the case from another perspective. Results from recent high‐resolution model simulations (Becker & Vadas, 2018; Vadas & Becker, 2018; Watanabe et al., 2006) indicate strong orographic gravity wave sources near McMurdo, which are attributed to the eastward downslope winds blowing down the steep slopes of the East Antarctic Plateau and Trans‐Antarctic Mountains (altitudes over 4 km) onto the west coast of Ross sea. While some of these waves propagate to and focus into the axis of PNJ, the rest of these waves can propagate directly to McMurdo and other high‐latitude locations (Vadas & Becker, 2018). If we consider that the midwinter PNJ location is far away from McMurdo (even in QBOw), the gravity waves focused into the axis of PNJ may not contribute much to the gravity waves over McMurdo. Then the less focusing effect during QBOe means more chance for gravity waves to propagate to McMurdo from wave sources and thus possibly higher *E*_{pm} during QBOe as the lidar observed. Quantitative analyses of these factors, including the solar cycle and ENSO effects, and their contributions to the enhanced *E*_{pm} in the QBO easterly phase are beyond the scope of this work but will be addressed in future modeling studies.

Using 9 years of lidar observations (2011–2019) from McMurdo/Arrival Heights (77.84°S, 166.67°E), Antarctica, and 21 years of MERRA‐2 reanalysis data (1999–2019), we present the first observations of potential QBO signals in the stratospheric gravity wave potential energy mass density (*E*_{pm}) at McMurdo and in the Antarctic polar vortex location near the stratopause. Midwinter stratospheric gravity wave activity is stronger during the QBO easterly years when the June polar vortex expands and the PNJ shifts equatorward. During the QBO westerly years, gravity wave activity is weaker when the polar vortex contracts and the PNJ moves poleward. The lidar‐observed interannual variations of gravity wave *E*_{pm} winter maxima over McMurdo are most likely controlled by the equatorial QBO signals via modulating the polar vortex location and thus the critical level filtering of gravity waves in the Antarctic. The ENSO may play a minor role in modulating the *E*_{pm} interannual variations. The QBO signals in the polar vortex edge and PNJ during June months are robust and appear not to be affected by the solar cycle and ENSO to the first order.

In addition to the repeated seasonal pattern of summer minima and winter maxima, qualified lidar data of ~7,000 hr also exhibit interannual variations of gravity wave *E*_{pm} in the stratosphere 30–50 km. That is, the mean *E*_{pm} winter maxima in 2012, 2015, and 2018 (during QBOe) are ~43% higher than those in the other years (during QBOw). Using the 9‐year composite‐mean harmonic fit to represent the mean annual cycle, the McMurdo lidar observations reveal the positive/negative gravity wave *E*_{pm} anomalies during QBO easterly/westerly winters. The correlation of the *E*_{pm} anomaly with a QBO index averaged over the 40 and 50 hPa pressure levels reaches a coefficient of −0.766 with a confidence level of 99.9%, confirming a statistically significant correlation between the gravity wave *E*_{pm} and the QBO phase in the last decade. The positive *E*_{pm} anomalies occur when McMurdo is well inside the polar vortex core and the distance to the vortex edge is large; both of these aspects tend to occur during the QBO easterly winters.

Using the 21 years of MERRA‐2 reanalysis data (1999–2019), we find that the June polar vortex edge near the stratopause moves equatorward/poleward during QBO easterly/westerly years. The correlation between the negative vortex edge distance from McMurdo at 45 km and a QBO index vertically averaged over 20, 30, 40, and 50 hPa pressure levels reaches a correlation coefficient of 0.846 with a confidence level of nearly 100%. The mean distance from McMurdo to the vortex edge, averaged over the QBO easterly years, reaches over 4,000 km, while the distance averaged over the QBO westerly years is about 3,000 km in June at 45 km. The mean wind along the vortex edge in June is weaker by 10–20 m/s in the QBO easterly than in the westerly years. Such vortex edge distance corresponds to the mean vortex edge near the stratopause (~50 km) located at 39.7°S and 47.7°S, respectively, for the QBO easterly and westerly phases. This latitude difference of ~8° is significant as it equates to a distance of nearly 1,000 km. The QBO signal in the polar vortex edge location is robust with statistical significance.

The equatorward shift of the PNJ core around the stratopause in the QBO easterly years and the poleward shift in the QBO westerly years are clear reflections of the vortex expansion/contraction during QBO easterly/westerly phases. The polar vortex edge is located along the poleward flank of the PNJ stream. During the QBO easterly phase, the polar vortex expands, pushing the PNJ toward the equator. The equatorward expansion of the polar vortex leads to quieter winters and more stable background winds during QBOe than QBOw and corresponds to less critical level filtering and thus higher *E*_{pm} in the upper stratosphere in the QBO easterly phase. The critical level filtering is characterized by the WRA, and our results support the hypothesis that the interannual variations of *E*_{pm} is mainly determined by the interannual variations of WRA, that is, the critical level filtering effects. WRA is likely modulated by the polar vortex location in the midwinter.

Why the SH polar vortex moves equatorward in QBOe, why the triple‐cell structures are formed around the stratopause in the SH, and why gravity wave *E*_{pm} at McMurdo are enhanced in the QBOe when the polar vortex and PNJ move equatorward, are beyond the scope of this study. These intriguing questions deserve future investigations with the combination of theories, models, and long‐term observations. The underlying mechanisms may well reflect the asymmetry between two hemispheres and the complex atmospheric coupling from the equatorial to the polar regions. Furthermore, the QBO modulation of stratospheric gravity waves may impact the vertical wave coupling via modulating the generation of secondary gravity waves (Vadas et al., 2018) and wave dissipation altitudes (Vadas & Becker, 2019), which will affect persistent gravity waves in the MLT (Chen et al., 2016; Chu et al., 2018). It is likely to find QBO signals in the MLT, and such questions deserve future investigation using the lidar data collected from McMurdo. The observationally based study presented in this article lays the groundwork for future modeling studies that will test the observed relationships and proposed mechanisms in long‐term climate simulations. Other future work will also investigate the roles of solar activity and ENSO as well as their entanglement with the QBO, polar vortex, and gravity waves.

We gratefully acknowledge the graduate students and research scientists who made contributions to the McMurdo lidar campaign, including winter‐over lidar scientists Zhibin Yu (2011), Brendan Roberts (2012), Weichun Fong (2013), Cao Chen (2014), Jian Zhao (2015), Ian Barry (2016), Zhengyu Hua (2017), Dongming Chang (2018), Zimu Li and Ian Geraghty (2019), Xianxin Li and Cissi Lin (2020), and summer scientists Wentao Huang, Zhangjun Wang, John A. Smith, Xian Lu, Muzhou Lu, Clare Miller, and Jackson Jandreau. We are grateful to Richard Dean, Nikolas Sinkola, Adam Godfrey, Mark Murphy, Hue Tran, and Kate McKenzie for their engineering help and support. We thank Erich Becker and Sharon L. Vadas for intriguing discussions on orographic gravity waves and critical level filtering. We are indebted to Vladimir Papitashvili, Chester S. Gardner, and Adrian J. McDonald for their guidance and collaboration. We sincerely appreciate the staff of the United States Antarctic Program, McMurdo Station, Antarctica, New Zealand, and Scott Base for their superb support over the years. We gratefully acknowledge MERRA‐2 data provided by the Global Modeling and Assimilation Office (GMAO) at NASA Goddard Space Flight Center through the NASA GES DISC online archive. This project was supported by the National Science Foundation (NSF) Grants OPP‐0839091, OPP‐1246405, and OPP‐1443726. V. L. H. was supported by NASA HGI Grant NNX17AB80G and NASA HGR Grants 80NSSC18K1046 and 80NSSC19K0834. Xian Lu's research was supported by NSF Grants AGS‐1705448 (CEDAR), OPP‐1705450, and CAREER‐1753214.

The data shown in this work can be downloaded online (from