In this study, the NCAR Thermosphere Ionosphere Mesosphere Electrodynamics General Circulation Model (TIME-GCM) is used to examine short-term changes in the global structure of the migrating and s = 1 nonmigrating semidiurnal tides in a model run approximating the 2002 Southern Hemisphere stratospheric sudden warming. The TIME-GCM is found to resolve an s = 1 nonmigrating semidiurnal tide with latitudinal structure similar to that of prior studies and with global structure in the MLT region strongly correlated with Southern Hemisphere planetary wave 1 (PW1) wave events, prior to equinox. Comparison with a control run with no planetary wave activity, Eliassen-Palm flux, and correlation analysis of the s = 1 semidiurnal tide structure suggest that the wave propagates from the source region in the winter hemisphere stratosphere into both the summer and winter lower thermospheres. The amplitudes of the migrating semidiurnal tide in the Southern Hemisphere are anticorrelated with the nonmigrating tide, suggesting significant energy loss due to nonmigrating tidal generation, a result not previously seen. The amplitude of the s = 1 nonmigrating semidiurnal tide from TIME-GCM is also compared with data from the South Pole meteor radar system. While the observations show bursts of nonmigrating tidal activity correlated to PW1 activity of comparable amplitudes to the model results, these events do not always appear at the same times as those in the model. This suggests that while the processes generating the nonmigrating tide in TIME-GCM are realistic, the TIME-GCM run itself displays differences in time evolution compared to the observations during this time.
 Atmospheric tides are global-scale oscillations with periods that are harmonics of a solar day, which dominate the dynamics of the Mesosphere and Lower Thermosphere (MLT) region. The tides may be further subdivided into migrating and nonmigrating tides on the basis of their zonal phase velocity: migrating tides have periods and zonal wave numbers such that they are Sun-synchronous, while nonmigrating tides do not. The migrating tides tend to dominate in the MLT region, with the migrating diurnal tide largest in low-latitude horizontal wind fields, and the migrating semidiurnal tide largest in the midlatitude to high-latitude horizontal wind fields. Excitation sources for different migrating and nonmigrating tidal components vary, but include absorption of solar radiation and latent heat release in the lower atmosphere, or generation through nonlinear interactions between the dominant migrating tides and stationary planetary waves.
 Despite the dominance of the migrating tides, nonmigrating tides can also be significant, and even exceed the amplitudes of the migrating tides under certain conditions. One nonmigrating tidal component that has attracted considerable interest in recent years is the westward propagating semidiurnal tide with zonal wave number 1 (s = 1). The s = 1 semidiurnal tide was first observed in mesospheric horizontal wind fields at the South Pole via optical measurements of OH emissions by Hernandez et al.  during austral winter, who noted the lack of a corresponding oscillation in simultaneously measured temperatures. The westward zonal wave number 1 was inferred from the phase progression of the observed semidiurnal oscillation, and was deemed to be consistent with prior predictions that large-scale wave activity at the poles could only be attributed to zonal wave number 1. Longer-term meteor radar observations from the South Pole by Forbes et al.  and Forbes et al. [1999a] found a large s = 1 semidiurnal tide occurring during the summer months, with meridional wind amplitudes in excess of 15 m/s, exhibiting day-to-day variability on scales longer than 4 days. Further observations by Riggin et al.  from McMurdo (77.8°S, 166.67°E) and Halley (75.8°S, 26.4°E) revealed that the s = 1 semidiurnal tide was also present in high-latitude regions away from the South Pole, occurring primarily as bursts during the austral summer solstice, while the migrating semidiurnal tide dominated during other time periods. Multiple mechanisms have been proposed for the generation of this component, which include nonlinear interaction between the migrating semidiurnal tide (s = 2) and the planetary wave with zonal wave number 1 (PW1), an in situ excitation, or excitation from a lower atmospheric source [Portnyagin et al., 1998].
 Current understanding suggests that the nonlinear interaction mechanism appears to be the most probable. Angelats i Coll and Forbes  were able to generate the observed summertime high-latitude enhancements of the s = 1 semidiurnal tide in a spectral model, as a byproduct of nonlinear interactions between the migrating semidiurnal tide and a prescribed PW1 profile in the winter hemisphere stratosphere and lower mesosphere. The study also found that the meridional structure of the resulting nonmigrating tidal component was mainly composed of the meridional wind expansion functions corresponding to the first and second symmetric Hough modes, accounting for peaks at both poles, and multiple smaller peaks (2 and 4 respectively) at low latitudes to midlatitudes. Further dedicated general circulation and linear response model studies by Yamashita et al.  reproduced the observed annual amplitude oscillation, finding that the s = 1 semidiurnal tide generated in the winter hemisphere propagated upward and across the equator into the summer hemisphere high latitudes. It was also noted that the global distribution of the s = 1 semidiurnal tide in the lower thermosphere showed a strong sensitivity to the background zonal mean zonal wind profile, and that more realistic zonal mean zonal wind distributions were needed in future modeling studies.
 Observational evidence for the nonlinear interaction mechanism was found by Smith et al. , who found a correlation between summertime and fall semidiurnal tidal amplitudes at Esrange (68°N, 21°E) and PW1 activity in the high-latitude Southern Hemisphere stratosphere. It has also been suggested that the interaction region between PW1 and the migrating semidiurnal tide may not be limited to the winter hemisphere: Hibbins et al.  found a modulation of the semidiurnal tide at Halley by the Quasi-Biennial Oscillation (QBO) during times at which the s = 1 semidiurnal tide was dominant, and suggested that the s = 1 semidiurnal tide could be more effectively excited during periods of eastward QBO, which would allow PW1 to propagate from the winter to the summer hemisphere.
 The primary motivations for this study include examining the short-term variability of the s = 1 semidiurnal tide, as well as understanding the sources and effects of semidiurnal tidal variability associated with the generation of this component. This requires the resolution of realistic background atmospheric conditions, including the zonal mean zonal winds, and PW1 structure, which have been identified in the aforementioned studies as playing an important role in the generation and propagation of the s = 1 semidiurnal tide. At the present time, satellite observations of the atmospheric tides are limited by the long time periods (around 60 days for SABER) required to unambiguously resolve the tides from stationary planetary waves, due to the slow local time precession rates of most satellite orbits. Additionally, most observational studies examining the short-term variability of the migrating and nonmigrating semidiurnal tides utilized ground-based observations unable to resolve tidal variability and structure on a global scale. To address these restrictions, we employ a general circulation model for this study, in order to examine semidiurnal tidal variability during a specific time of interest that the model has been known to resolve in a realistic fashion.
Liu and Roble  noted an increase in the semidiurnal s = 1 component at high latitudes during their simulation of a generic stratospheric sudden warming with a coupled GCM. This has been attributed to the fact that in addition to being responsible for the generation of the s = 1 semidiurnal tide, PW1 is also one of the planetary waves commonly associated with the generation of stratospheric sudden warmings. Using the NCAR Thermosphere Ionosphere Mesosphere Electrodynamics General Circulation Model (TIME-GCM), Liu and Roble  were able to produce a realistic simulation of the 2002 Southern Hemisphere stratospheric sudden warming. In the period leading up to the major warming, they found the occurrence of multiple PW1 events, serving to weaken and reverse the westerly mean wind jet in the stratosphere, while altering the background atmosphere to allow wave breaking at successively lower altitudes. The presence of these recurring PW1 events are likely a manifestation of the quasi-10-day wave found by Palo et al.  in their SABER analyses. The strong PW1 activity during the time leading up to the stratospheric warming makes it interesting in the context of nonmigrating tidal generation and studies of migrating tidal variability. The ability to examine model results made in the presence of a quasi-realistic background atmosphere is also attractive in understanding changes in the global structure, generation, and propagation of the s = 1 semidiurnal tide.
 In this study, the behavior of the migrating and s = 1 semidiurnal tides, as well as PW1 were examined from a realistic simulation of the 2002 Southern Hemisphere stratospheric warming performed using the NCAR TIME-GCM. The model results were then compared to observations taken during the same period from the South Pole meteor radar system. The model, radar system, and analysis methodology are briefly described in the following sections.
2.1. TIME-GCM Stratospheric Sudden Warming Run
 TIME-GCM is a nonlinear general circulation model spanning from a lower boundary near 30 km in the stratosphere up to approximately 500 km in the thermosphere. For this study, we analyze the results from the model run performed by Liu and Roble  in their study of the 2002 Southern Hemisphere stratospheric sudden warming. In this run, the tropospheric sources of the migrating diurnal and migrating semidiurnal tides were accounted for by using forcing from the linear mechanistic Global Scale Wave Model (GSWM) at the lower boundary. Nonmigrating tides were not explicitly forced at the lower boundary, but were generated self-consistently by the model, in order to quantify the effect of extratropospheric nonmigrating tidal sources. Upward propagating planetary waves were forced at the lower boundary using daily NCEP Reanalysis values. TIME-GCM also accounts for forcing in the thermospheric high latitudes from auroral precipitation using geomagnetic indices as a proxy. These were held constant during the run, with F10.7 set to 150. A detailed discussion on the tidal effects of this mechanism is given by Hagan and Roble .
 This model run covers early June through early November 2002 (days 160–310), and covers the major warming itself, as well as the wave events leading up to the major warming. Figure 1 shows the daily zonal mean zonal wind amplitudes at 82.5°S in the model run. The reversal of the stratospheric and mesospheric jets are clearly visible, as are the effects of the multiple smaller PW1 events occurring prior to the major warming on day 268, which are manifested in the mean winds as sudden decreases (increases) in the stratospheric and mesospheric (lower thermospheric) westerly jet.
 Hourly global snapshots of relevant atmospheric fields were output, from which the daily amplitudes (As) and phases (ϕs) of the semidiurnal tides and PW1 were derived using a linear least-squares fit. Unless otherwise specified, a sliding window of 4 days was used for analysis of the semidiurnal tides, and 6 days for PW1. The basis function used for the semidiurnal tidal fits included a zonal mean component (denoted as ), as well as semidiurnal tides (frequency given in terms of the Earth's rotation rate Ω) with zonal wave numbers from 1 to 4 in both eastward and westward directions:
 The basis function for the planetary wave fits included the zonal mean, as well as planetary waves 1 through 4:
 While the relatively short averaging window used for PW1 will result in aliasing with longer-period planetary waves, it was deemed necessary in order to examine the variability of the stationary planetary wave and the tide on shorter time scales. The longer-period planetary waves are manifested as part of the tidal and PW1 fluctuations. A separate control run was also performed without the NCEP lower boundary forcing, in order to better quantify the effect of planetary waves on the semidiurnal tides.
2.2. South Pole Meteor Radar
 The South Pole meteor radar system provides hourly measurements of meridional winds at around 95 km in viewing directions of 0, 90, 180, and 270° longitude [Lau et al., 2006]. This allows the unambiguous determination of eastward or westward propagating zonal wave number 1, using a basis function of the form:
 The frequency ω was set to 0 for the PW1 fits, and to 2Ω for the semidiurnal tidal fits. These basis functions were used to derive the semidiurnal amplitudes and phases using a least-squares fitting routine, which also provides an estimate of fitting errors through the calculated covariances. PW1 was fit separately from the semidiurnal components. A sliding window of 6 days has been used for fitting both the nonmigrating semidiurnal tide and PW1. The increased sample length provides sufficient spectral resolution to avoid spillover into the semidiurnal fits from Lamb waves with periods around 10 and 13 h, which are known to occur in the winter months [Forbes et al., 1999b]. A fit is performed daily using the aforementioned 6 day sliding window, unless the number of missing data points exceeds 25%.
3. Model Results
Figure 2 shows the meridional wind field amplitudes of the s = 1 semidiurnal tide as a function of latitude and altitude on day 203, representative of conditions before the major warming; results are shown from both the TIME-GCM run with planetary wave forcing as well as the control run without such forcing. Prior to the major warming on day 268, the meridional structure of the s = 1 semidiurnal tide shows four to five peaks spanning both hemispheres in the planetary wave forcing run. This latitudinal structure is similar to that found by Angelats i Coll and Forbes  in their modeling studies of the s = 1 semidiurnal tide, though the two middle TIME-GCM nodes around 95–100 km appear to be shifted southward in latitude compared to the previous study. Significant s = 1 semidiurnal tidal amplitudes are also seen below 100 km, in the southern high-latitude MLT region, which are stronger than those in the northern high latitudes in the same height range. This causes a tilt downward and southward found in the TIME-GCM nonmigrating tidal structure, though this asymmetry appears to decrease as the model progresses toward austral summer. In contrast to the planetary wave forcing runs, the control runs show minimal amplitudes (less than 4 m/s) for the s = 1 semidiurnal tide below about 110 km. From this comparison it is clear that the s = 1 semidiurnal tidal response at low latitudes to midlatitudes, below 110 km, and the boreal summertime peak between 50 and 80°N in the lower thermosphere are driven through interaction with PW1.
Figure 3 shows the s = 1 semidiurnal tidal meridional wind field amplitudes on day 301, representative of conditions after the major warming, for both the planetary wave forcing and control runs. A similar situation is seen, with the s = 1 response below 110 km and at low latitudes to midlatitudes occurring only in the model run with planetary waves. In contrast to the prestratwarming run, the s = 1 amplitudes below 100 km now appear to be larger in the northern high latitudes, as the atmosphere enters boreal winter (and austral summer) conditions. The Southern Hemisphere peak is visible as a bulge around 110 km extending from the South Pole to about 50°S.
 Above 110 km, two enhancements can be seen around the North and South Poles of both model runs (Figures 2 and 3). Similar thermospheric features attributed to aurorally driven circulation were found in TIME-GCM migrating diurnal tide and stationary planetary wave 1 fields by Hagan and Roble , who found that variability in the aforementioned enhancements were tied to changes in geomagnetic indices, as well as changes in photoionization rates due to seasonal variations in the solar heating profile. In our model runs, the high-latitude enhancements vary slowly on a seasonal scale, starting out stronger in the northern high latitudes during the boreal summer before gradually becoming stronger in the southern high latitudes as the model progresses toward austral summer. Combined with the fact that the geomagnetic indices in our model runs were held constant throughout the entire period, this strongly supports the idea that the high-latitude enhancements resolved above 110 km are aurorally driven circulation. It should be noted that both the aurorally driven circulation and the s = 1 semidiurnal tide will be stronger in the summer hemisphere, though for different reasons. As previously mentioned however, the effects of the auroral source on the s = 1 semidiurnal tide are confined primarily to the high-latitude regions above 110 km, and cannot explain the s = 1 semidiurnal tidal features at other latitudes and below 110 km.
Figure 4 illustrates the time evolution of meridional wind fields of the s = 1 semidiurnal tide at 100 and 110 km, PW1 at 80 km (generally representative of PW1 activity in the stratosphere and mesosphere), and the migrating semidiurnal tide at 100 km. Figure 5 shows the time evolution of the s = 1 semidiurnal tide as a function of altitude and time near the nonmigrating tidal peaks at 87.5°S and 52.5°N, as well as PW1 at 87.5°S. The temporal variability in the peaks of the s = 1 semidiurnal tide in both hemispheres appear to be highly correlated in time prior to around day 245, with wave events showing a near simultaneous global response at these altitudes. The altitude structure of the s = 1 semidiurnal tide near the peaks shown in Figures 5a and 5b are also relatively consistent up till around day 230, with the nonmigrating tidal peaks occurring around 105 km and 115 km, respectively. Successive PW1 wave events occur in the Southern Hemisphere in the period prior to the major warming (Figures 4c and 5c), as was noted by Liu and Roble . During this time, PW1 is confined mainly to the southern midlatitudes to high latitudes. This is due to the zero wind line formed by westward equatorial zonal mean zonal winds in the stratosphere, which inhibits cross-equatorial propagation of PW1 in the stratosphere and lower mesosphere, as suggested by Hibbins et al. .
 It is interesting to note the changes in nonmigrating tidal structure that result from the changes in the zonal mean zonal winds in the 30 or so days prior to the major warming. From Figure 1, it can be seen that the westward mean winds above roughly 100 km change direction on day 234, creating a broad region of eastward mean winds between the stratosphere and the lower thermosphere, up till the reversal of the mean winds in the mesosphere on day 255. Two strong PW1 events occur during this time, around days 235 and 244, with corresponding nonmigrating tidal events occurring around the same time. From Figure 5a, it can be seen that the southern high-latitude events that occur during these 2 days peak at much lower altitudes (peaking around approximately 90 km) compared to the prior winter hemisphere events. The eastward winds during this time create favorable conditions for the upward propagation of westward waves, as well as PW1. This results in amplification of PW1, and increased wave breaking at progressively lower altitudes [Liu and Roble, 2005]. The increased eddy dissipation in the southern high latitudes that results from this process may explain the altered structure of the nonmigrating tide in this region, as well as the penetration of PW1 into higher altitudes.
 Following the s = 1 semidiurnal tidal event occurring around day 244, there appears to be a change in the meridional structure of the s = 1 semidiurnal tide with the two peaks at 100 km in the Southern Hemisphere low latitudes and midlatitudes in Figure 4a (and the peaks in the northern low latitudes to midlatitudes at 110 km in Figure 4b) becoming much less prominent in the subsequent nonmigrating tidal event on day 255. The high-latitude peak near the South Pole at 110 km (Figure 4b) also begins to strengthen during this period, first becoming larger compared to the northern midlatitude to high-latitude peaks in the wave event on day 255, and subsequently transitioning to higher altitudes after day 270 (Figure 5a). The nonmigrating tidal peaks in the northern midlatitudes to high latitudes at 110 km also begin to decrease in amplitude at the same time. In addition to the increased PW1 activity in the Northern Hemisphere, PW1 activity in the southern stratosphere is decreasing during this time, because of the reversal of the zonal mean zonal wind jets in the stratosphere/mesosphere during this time (Figure 1).
 A clearer picture of the relations between the various semidiurnal tides and PW1 can be seen in Figure 6, which shows the amplitude perturbations (scaled by the mean values over the model run) of PW1 and the s = 1 semidiurnal tide at selected latitudes (located near the peaks seen in Figure 4) as a function of time. Prior to the major warming, a strong correlation is seen between the variation of the s = 1 semidiurnal tide at the selected latitudes and PW1. Additionally, the migrating semidiurnal tide in the Southern Hemisphere decreases in amplitude during these wave events (Figure 4d), indicative of energy loss, possibly through a nonlinear interaction with PW1. However, this relation appears to break down following the major warming, and may be due to the transition in the model from austral winter to equinox, and subsequently summer conditions. After this transition both the PW1 and the migrating semidiurnal tide weaken in the Southern Hemisphere, while strengthening in the Northern Hemisphere. Subsequent s = 1 semidiurnal tide events appear to be correlated to PW1 events in the Northern Hemisphere high latitudes, as can be seen around days 270 and 293 in Figure 4.
 The Eliassen-Palm flux (Fϕ and Fz) for the s = 1 semidiurnal tide in the planetary wave forcing run were computed as:
 Here, Eliassen-Palm flux divergence for the s = 1 semidiurnal tide has been scaled by density and is in units of pressure to better illustrate wave activity at the lower altitudes of the model, where PW1 (and the presumed sources of the s = 1 semidiurnal tide) are located. It is expected that source regions will be characterized by strong Eliassen-Palm flux divergence, while the flux directions will illustrate the propagation of wave energy away from these sources. Figures 7 and 8 show the Eliassen-Palm flux direction and flux divergence computed for the s = 1 semidiurnal tide on days 203 and 301, respectively representative of prestratwarm (austral winter) and poststratwarm (transitioning to boreal winter) conditions. In Figure 7, the strongest regions of divergence are seen in the Southern Hemisphere stratosphere. The flux vectors immediately surrounding these regions point upward and equatorward for most regions between 30 and 60°S, constant with cross-equatorial propagation from a winter hemisphere source. At latitudes poleward of 60°S surrounding the source region and at heights stretching up to around 100 km, the direction of the flux vectors is primarily upward and converging toward the South Pole. This suggests that the s = 1 semidiurnal tidal activity occurring in the winter high latitudes is driven at least in part by the same source as the summertime peak.
 The highly active areas of Eliassen-Palm flux divergence in the stratosphere are roughly coincident with regions and times where PW1 is strong, again suggesting that PW1 events in the winter hemisphere serve as the source for most of the s = 1 semidiurnal tidal activity in the planetary wave forcing run. Divergence and downward EP flux are also resolved in the winter midlatitude to high-latitude MLT region in both the pre and post warming periods, indicating the presence of a second source in the lower thermosphere midlatitudes to high latitudes. Two possibilities for this source are downward propagation from auroral forcing in the lower thermosphere, or nonlinear interaction between PW1 and the migrating semidiurnal tide in the lower thermosphere. The downward winter EP flux is also resolved in the control run, where only the auroral forcing is present, though it is not as strong as in the planetary wave forcing run. This suggests that both auroral forcing and planetary wave activity contribute to the lower thermospheric source in the planetary wave forcing run. However, given the relatively low amplitudes of PW1 in the lower thermosphere compared to PW1 amplitudes in the stratosphere before the major warming, as well as the comparisons between the model runs with and without planetary waves (Figures 2 and 3), this lower thermospheric source for the s = 1 semidiurnal tide is unlikely to be as important as the source from PW1/migrating semidiurnal tide interactions in the stratosphere.
 Changes in wave sources and propagation following the major warming and equinox transition can be seen in Figure 8, for day 301. As the model continues past equinox, the regions of strong convergence and divergence in the Southern Hemisphere weaken, while regions of convergence and divergence in the Northern Hemisphere begin to strengthen. During this transition period, wave activity can propagate from either hemisphere, though the flux divergence (and PW1 amplitudes in the stratosphere and lower mesosphere) are lower than during the boreal summer portion of the model run.
 Further evidence for the potential source region for the s = 1 semidiurnal tide prior to the stratospheric sudden warming can be seen in Figure 9, which shows the correlation coefficients for PW1 amplitudes at each model grid point correlated to the variation of the s = 1 semidiurnal tide near the Northern Hemisphere peak at 52°N, 114 km altitude. This is similar to the correlation analysis performed by Smith et al.  using PW1 values from SABER, though in that study the proxy for the s = 1 semidiurnal tide was limited to the variation of the semidiurnal tide at Esrange, with no wave number determination being possible because of the limited spatial sampling. The confidence level is computed in a manner analogous to that in the work by Smith et al. , except here the number of independent data points are scaled by a factor of 6 to account for the 6 day averaging window used to compute PW1. Areas where the confidence level is less than 90% are shaded. Assuming zero lag time, correlation in excess of 0.7 is found in an area below 75 km and between roughly 42 and 60°S. Combined with the strong PW1 activity in the area, as well as the strong Eliassen-Palm flux divergence shown previously in Figure 7, these results point to this area as the source region for the s = 1 nonmigrating semidiurnal tide prior to the major warming.
 Similar correlation analysis performed for the period after the major warming, using the s = 1 semidiurnal tide at 85°S, 108 km shows reduced correlation coefficients overall. However, because of the shorter duration of the model run after the warming, the confidence levels for the postwarming correlation analysis do not meet the 90% confidence threshold and are not shown here.
Figure 10 shows the correlation coefficients of the time variations of the s = 1 semidiurnal tide at each model grid point correlated with the integrated power of PW1 at all model grid points below 80 km altitude, prior to the stratospheric sudden warming. The results show an area where the variation of the s = 1 semidiurnal tide is highly correlated to the global variation of PW1, extending from the Southern Hemisphere MLT high latitudes across the equator into the lower thermosphere of the Northern Hemisphere. This region of high correlation is also seen in the line plots of amplitude perturbation shown previously in Figure 6. These results suggest the existence of a waveguide that is channeling wave energy from the Southern Hemisphere high-latitude source region into the lower thermosphere of the Northern Hemisphere during most of the model run. This is consistent with previous hypotheses concerning the origin of the s = 1 semidiurnal tide, and may explain why nonmigrating tidal amplitudes are stronger in the Southern Hemisphere at lower altitudes, but become larger in the Northern Hemisphere at higher altitudes in the lower thermosphere.
 The amplitudes of the s = 1 semidiurnal tide in region of stronger correlation in the northern mesosphere poleward of 30° are very small (less than 1 m/s) during this period. As such, the results in this area are not believed to be significant. Similar correlation analysis performed for the period after the major warming display a much lower correlation and do not show the potential waveguide seen in Figure 10. However, the confidence levels for the postwarming correlation analysis again do not meet the 90% confidence threshold and are not shown here.
4. Radar Comparison
Figure 11 shows the s = 1 semidiurnal tide and PW1 derived from South Pole meteor radar data, as well as similar values taken from the nearest TIME-GCM grid point. While correlation between the timing of wave events in the model and the observations is generally rather poor, there are still some points worth noting. Both the nonmigrating tide and PW1 in the radar data display intermittent wave events with amplitudes similar to that seen in the TIME-GCM results. There also appears to be some correlation between nonmigrating tide and PW1 wave events in the radar data around days 213, 220, 226, 258, and 274. This correlation between the two waves in the Southern Hemisphere is also seen in TIME-GCM, as mentioned previously. While Smith et al.  did not find any correlation between PW1 wave events and the s = 1 semidiurnal tide in the same hemisphere during their Northern Hemisphere winter observations, Baumgaertner et al.  did find such a relation in Southern Hemisphere winter observations. It is possible that the interaction region in the Northern Hemisphere is further poleward than the observations by Smith et al.  at Esrange (68°N, 21°E). However, a definitive result cannot be reached without a longer boreal winter model run. It is also worth noting that the summertime amplitude peaks in the model occur at altitudes above the meteor radar observational domain, an issue also encountered in previous modeling efforts by Angelats i Coll and Forbes  and in the linear response model utilized by Yamashita et al. . The correlation analysis by Smith et al.  found that potential source regions for the s = 1 semidiurnal tide often extended below 30 km altitude. It is possible that the higher altitudes of the summertime amplitude peaks in TIME-GCM may be the result of the model being unable to resolve the entire interaction region, because of the model lower boundary being around 30 km. This would result in the model amplitudes peaking at a higher altitude than observed.
 One notable exception to this strong correlation between the PW1 and the s = 1 semidiurnal tide observed by the South Pole meteor radar is the large nonmigrating tidal event observed on day 285. Possible explanations for this discrepancy include changes in PW1 structure after propagating above the interaction region in the mesosphere and upper stratosphere, or the possibility that the nonmigrating tidal event was generated in the Northern Hemisphere, because of increasing PW1 amplitudes in that hemisphere with the seasonal transition to boreal winter.
5. Conclusions and Future Work
 The short-term variability of the s = 1 semidiurnal tide around the 2002 Southern Hemisphere stratospheric sudden warming in TIME-GCM has been found to be strongly dependent upon the PW1 events prior to the major warming. Amplitudes of the migrating semidiurnal tide in the Southern Hemisphere display a strong anticorrelation to PW1 and nonmigrating tidal events, consistent with nonmigrating tidal generation via migrating tide/PW1 interactions. Comparisons with a control run without planetary waves show that the global s = 1 response below 110 km is likely the result of this nonlinear interaction. Above 110 km, this nonlinearly forced component interferes with aurorally forced components, which are mostly confined to the very high latitudes around the poles, and vary on a longer seasonal scale. EP flux analysis reveals that the s = 1 semidiurnal tide propagates upward from a source in the winter stratosphere, both across the equator, as well as into the winter high latitudes.
 The mean wind reversal in the lower thermosphere prior to the major warming appears to alter the vertical structure of the nonmigrating tide in the winter high latitudes, while the summer hemisphere nonmigrating tidal peaks reflect the amplification of PW1 prior to the major warming. The nonmigrating tidal peaks at midlatitudes and low latitudes decrease in a wave event occurring roughly 10 days thereafter, during which, the southern high-latitude peak becomes stronger than the northern high-latitude peak for the first time. This is coincident with the time when PW1 begins to occur in the Northern Hemisphere, and the PW1 events in the southern high latitudes are decreasing because of the wind reversals occurring in the mesosphere and upper stratosphere. Subsequent nonmigrating tidal events occurring around and after equinox become larger in the southern high latitudes above 110 km, and smaller in the northern high latitudes.
 The summertime amplitude enhancement is reproduced in the Northern Hemisphere lower thermosphere around 115 km, however our results also show amplitude enhancements in the Southern (winter) Hemisphere around 105 km, with slightly smaller amplitudes. A longer model run will be needed to ascertain the strength of these enhancements relative to enhancements in the Southern Hemisphere summer at meteor radar altitudes, given the fact that the model results show the tidal summertime enhancements peaking above the meteor radar domain. Additionally, it will be interesting to compare simultaneous observations of the s = 1 semidiurnal tide in both hemispheres to better understand the structure of these high-latitude enhancements as a function of altitude over short time scales. Observations from multiple ground sites at high latitudes in both hemispheres will likely be required for such an effort.
 Observations of the s = 1 semidiurnal tide and PW1 in South Pole meteor radar data show that wave events of these two components are correlated prior to the major warming, suggesting again that the nonmigrating tidal events in the winter hemisphere are the result of PW1/migrating semidiurnal tide interactions in the same hemisphere. However, comparisons with TIME-GCM results show that wave events in the model and the data are not identical. While this raises questions on how accurately TIME-GCM can reproduce real-world events when forced with NCEP reanalysis data, the qualitative features of PW1 and nonmigrating tidal behavior (e.g., prewarming correlation between the nonmigrating tide and PW1) are similar, suggesting that the physical processes generating the s = 1 semidiurnal tide in TIME-GCM are realistic. This is also interesting in the context of prior observational studies which showed correlation between the wintertime s = 1 semidiurnal tide and PW1 in the Southern Hemisphere [Baumgaertner et al., 2006], but not the Northern Hemisphere [Smith et al., 2007], and remains a possible topic for future model runs.
 The authors wish to acknowledge Jeffery Forbes and Jeffery Thayer of the University of Colorado for their guidance in this project, as well as Hiroyuki Iimura for his assistance in obtaining the raw South Pole meteor radar data and Qian Wu of NCAR for reviewing this manuscript, along with three anonymous reviewers. The first author also thanks Chihoko Yamashita, Jason Reimuller, and Xiaoli Zhang of the University of Colorado for support and helpful suggestions. This research was supported by National Science Foundation grant ATM-0228026, part of the CEDAR program under the direction of Scott Palo.