The vertical coupling of the stratosphere-mesosphere system through quasi-stationary and traveling planetary waves during the major sudden stratospheric warming (SSW) in the Arctic winter of 2003/2004 has been studied using three types of data. The UK Met Office (UKMO) assimilated data set was used to examine the features of the global-scale planetary disturbances present in the winter stratosphere of the Northern Hemisphere. Sounding the Atmosphere using Broadband Emission Radiometry (SABER) satellite measurements were used as well for extracting the stationary planetary waves in the zonal and meridional winds of the stratosphere and mesosphere. Radar measurements at eight stations, four of them situated at high latitudes (63–69°N) and the other four at midlatitudes (52–55°N) were used to determine planetary waves in the mesosphere-lower thermosphere (MLT). The basic results show that prior to the SSW, the stratosphere-mesosphere system was dominated by an upward and westward propagating ∼16-day wave detected simultaneously in the UKMO and MLT zonal and meridional wind data. After the onset of the SSW, longer-period (∼22–24 days) oscillations were observed in the zonal and meridional MLT winds. These likely include the upward propagation of stationary planetary waves from below and in situ generation of disturbances by the dissipation and breaking of gravity waves filtered by stratospheric winds.
 The dynamics of the middle atmosphere in winter are known to be dominated by planetary waves of large amplitudes. The most important are quasi-stationary Rossby waves, which propagate upward from the troposphere and are very strong but quite variable during winter. Other planetary waves are the traveling normal modes, also known as free modes. These waves correspond to the natural modes of variability of the Earth's atmosphere; commonly observed modes are those with periods around 2, 5, 10, and 16 days. Normally, these waves do not transport much momentum but they can interact with other waves or with the zonal mean flow. The wave-wave interactions may play an important role in the dynamics of the middle atmosphere and certainly contribute significantly to the variability of the population of atmospheric waves at these heights.
 The interaction of the planetary waves and the zonal mean flow is known to be the major driver of winter stratospheric dynamics [Andrews et al., 1987]. Classical studies showed that the zonal mean flow affects the planetary wave propagation by changing the refractive index [Charney and Drazin, 1961]. Time-varying or dissipating planetary waves interact with the zonal mean flow and can alter it dramatically, as happens in sudden stratospheric warmings (SSW). The key mechanism for generating SSW, originally proposed by Matsuno  and now widely accepted, is related to the growth of upward propagating transient planetary waves and their interaction with the zonal mean flow. The interaction decelerates and/or reverses the eastward winter winds and also induces a downward circulation in the stratosphere, causing adiabatic heating. There may also be an upward circulation in the mesosphere, causing adiabatic cooling [Liu and Roble, 2002]. The time period preceding the onset of a SSW is usually characterized by high wave activity in the stratosphere during which more than one type of planetary wave may be present.
 The present work is focused on two main topics: (1) a detailed investigation of the types of planetary wave observed in the stratosphere during the 2003/2004 Northern Hemisphere winter (1 October 2003 through 30 April 2004) as revealed by the analysis of UK Met Office (UKMO) assimilated data and by Sounding the Atmosphere using Broadband Emission Radiometry (SABER) satellite measurements; and (2) an investigation of the vertical coupling of the stratosphere-mesosphere system associated with quasi-stationary and traveling planetary waves before and during the major SSW in the Arctic winter of 2003/2004.
2. Observations and Data Analysis
2.1. UK Met Office Data
 The UKMO data set is used to examine the features of the global planetary disturbances present in the winter of 2003/2004 stratosphere of the Northern Hemisphere (NH). This data set is a result of assimilation of in situ and remotely sensed data into a numerical forecast model of the stratosphere and troposphere. A description of the original data assimilation system can be found in the work of Swinbank and O'Neill ; the new 3D-variational system is described by Swinbank and Ortland . The outputs of the assimilation are global fields of daily temperature, geopotential heights, and wind components at pressure levels from the surface up to 0.1 hPa. The daily data fields have global coverage with 2.5° and 3.75° steps in latitude and longitude, respectively. The UKMO data well represent the global features of stratospheric dynamics and have been used by many researchers to study dynamical events in the stratosphere including planetary waves [e.g., Fedulina et al., 2004; Chshyolkova et al., 2006] and SSWs [Cho et al., 2004]. The analysis in this paper focuses on planetary waves in the zonal and meridional winds of the NH stratosphere.
 The UKMO zonal and meridional wind data at four pressure levels (30, 10, 1, and 0.3 hPa) are used to explore the features of the global planetary waves in the NH stratosphere. Since the occurrence of SSW is related to the growth of planetary waves in the winter stratosphere, we investigate both the stationary Rossby planetary waves (SPWs) that propagate upward from the troposphere and the traveling planetary waves that are also present in the UKMO data. To determine the predominant periods of the planetary waves, we use a spectral analysis method that is a two-dimensional analogue of the Lomb-Scargle periodogram method [Lomb, 1975; Scargle, 1982] based on a least-squares fitting procedure applied to the entire time series. The planetary waves with periods between 5 and 30 days and with zonal wave numbers up to 3 are studied. The main purpose is not only to define the predominant periods of the wave components that contribute to the variability of the atmospheric fields but also to isolate and to study them in detail. To extract the waves from the data and to determine their amplitudes and phases, we use again a least-squares fitting procedure, but this time it is applied to the time segment twice the length of the longest period under investigation. Then this segment is moved through the time series with steps of 1 day in order to obtain daily values of the wave amplitudes and phases. The planetary wave structures (SPWs and traveling planetary waves together) are viewed in the zonal and meridional wind data for the four pressure levels.
2.2. SABER Data
 Vertical profiles of temperature and geopotential height data obtained by the SABER instrument onboard the TIMED satellite were employed to explore temporal behavior and spatial structures of SPW1 and SPW2 at heights from the stratosphere to lower thermosphere. The SABER latitude coverage extends from about 52° in one hemisphere to 83° in the other. About every 60 days, the latitude ranges flip as the spacecraft yaws to keep the instrument on the anti-Sun side of the spacecraft. In this way, high-latitude data are available only in 60-day segments, with no information for the 60 days preceding or following. The TIMED orbit precesses to cover 12 h of local time in each 60-day yaw period, as ascending and descending data together give almost 24 h of local time sampling. The data of the 1.06 version were analyzed in this study (http://saber.gats-inc.com).
 The daily parameters of the SPWs from 1 October 2003 to 30 April 2004 were determined using data from a 30-day window surrounding the date. Data from profiles of temperature and geopotential height were binned into regular latitude-longitude cells at 7 selected log-pressure height with a width of 1.5 km. The centers of the cells create a regular grid with a step of 5° in latitude from 30°N to 80°N and 15° in longitude. The log-pressure levels range from about 26 km to about 105 km height. In this way we obtained a monthly mean daily time series of temperature and geopotential height for each cell. We note that the time coverage at each cell was less than 16 LT hours and greater than 12 LT hours. Only the monthly mean values were calculated for each cell. It is worth noting that the amplitudes of semidiurnal oscillations were significantly reduced by this procedure (more than 1 order of magnitude for each cell). The monthly means were then fit with s = 1 and s = 2 zonal harmonics to obtain height versus latitude distributions of the parameters of SPW1 and SPW2.
 The parameters of the SPWs are potentially contaminated by the nonmigrating diurnal tides. Employing the known distribution of local time hours for each cell and substituting an analytical representation of the tides (proportional to amplitude × cos (full tidal phase)) into the analysis, we readily find that the migrating tide has no significant influence on the results but the nonmigrating tides can affect the results. Specifically, SPW1 is contaminated by the westward propagating diurnal tide with zonal wave number 2 and by the zonally symmetric tide (s = 0). SPW2 is contaminated by eastward zonal wave number 1 and westward zonal wave number 3 diurnal tides [see also Forbes et al., 2002; Talaat and Lieberman, 1999]. In the main about 15% of the tidal amplitudes can be transferred to the SPW amplitudes in temperature and geopotential by our procedure. Forbes et al.  and Oberheide et al.  presented the distributions of the amplitudes and phases of the wind components for the nonmigrating diurnal westward and eastward propagating tides with zonal wave numbers up to 4 above 85 km height (these tides can be ignored below 85 km). Taking the observed values of the wind amplitudes and assuming a π/2 shift in the phase at latitudes higher than 30°N, we estimated the magnitude of the errors in the SPW winds from the nonmigrating tides. The distortion in the analysis due to the tidal contamination was found to be of the same magnitude or less than errors due to, for example, the impact of noise in the data averaging over space and time.
 The zonal mean zonal wind was calculated as a gradient wind [Fleming et al., 1990] and used in linearized momentum equations to determine the winds associated with SPWs. The analysis ignores vertical winds and external sources or damping of the SPWs. On the basis of the SABER accuracy estimates and our analysis, we estimate the errors as about 0.5 K for temperature; about 150–200 gpm for the geopotential height at latitudes lower than 60°N, growing to 300 gpm at 80°N; about 2 m/s for the SPW1 and 3–4 m/s for the SPW2 for the wind at latitudes lower than 70°N and 4–6 m/s at higher latitudes.
2.3. MLT Neutral Wind Measurements
 We use zonal and meridional wind data from four radars situated at high latitudes between 63°N and 69°N (Table 1) and four radars at middle latitudes between 52°N and 55°N (Table 2). Three of these radars are commercially produced SKiYMET all-sky VHF systems (Esrange, Andenes, and Yellowknife), two are medium frequency (MF) radars (Poker Flat and Saskatoon), two are meteor radars (beam system) without height information (Obninsk and Castle Eaton), and one is a MST radar working in meteor mode (Kuehlungsborn). Some details about these radars and their routine data analysis procedures can be found in the work of Mukhtarov et al. .
Table 1. Geographic Locations of the High-Latitude Radars, the Type of the Instrument, and the Height Range of Available Measurements
Height Range (km)
50–108(only data for the range of 60–90 km are used)
Table 2. Geographic Locations of the Midlatitude Radars, the Type of the Instrument, and the Height Range of Available Measurements
Height Range (km)
no height resolution
MST (meteor mode)
85 and 94
no height resolution
 The basic information used in this study consists of daily estimates of the prevailing zonal wind at each of the height intervals. The daily zonal winds are obtained by applying a linear least-squares fitting algorithm to time segments of 1-day duration; a superposition of mean wind and 24-, 12-, and 8-h harmonic components was fitted. The segment was then incremented through the zonal and meridional wind time series in steps of 1 day in order to yield daily spaced values of the prevailing wind. The prevailing zonal and meridional wind and tidal characteristics are calculated at each step only if at least 16 valid hourly data points are present in the 1-day time segment [Crary and Forbes, 1983]. The daily values of the zonal and meridional winds are used later for studying the planetary waves observed in the MLT region.
 The planetary waves are transient phenomena and they are studied by a wavelet transform method. The wavelet analysis presented here employs the continuous Morlet wavelet, which consists of a cosine wave modulated by a Gaussian envelope. The nondimensional frequency, which gives the number of oscillations within the wavelet itself, is set to six to satisfy the wavelet admissibility condition [Torrence and Compo, 1998]. The Morlet wavelet was selected because of its simplicity and convenience in investigating wave-like events observed in the neutral wind of the MLT region [Pancheva and Mukhtarov, 2000]. The localization characteristics in time and frequency space of the Morlet wavelet used in this study are as follows: the time localization, or the so called “cone of influence”, is defined as a time interval which contributes to the wavelet coefficient at a given instant to.In our case the influence cone is t∈[t0 − a, t0 + a], where a is the wavelet scale and 1.03 a = T, where T is the Fourier period. Likewise, the localization characteristics of the wavelet in frequency space give local information about the function in frequency range: ω∈ [ − ; + ]. The frequency resolution can be increased by using a wavelet with intrinsic frequency higher than 6, albeit at the expense of decreased time resolution.
 In order to study the longitudinal structure of any periodicities simultaneously present in two or more time series we use two approaches: (1) a least-squares best fit procedure for defining the wave amplitudes and phases; and (2) a cross-wavelet analysis [Torrence and Compo, 1998; Grinsted et al., 2004], where the cross-wavelet power serves as an indication for the strength of the oscillations coexisting in both time series, and the argument describes the phase difference between them. We note that the phase difference is only reliable near the maximum of the cross-wavelet power where the phases are relatively stable.
 It is worth noting that in the MLT region we cannot divide the planetary waves found in the radar measurements into SPWs and traveling planetary waves. The reason is that the number of the radars is far from being sufficient for applying a two-dimensional (time-longitude) analogue of the Lomb-Scargle periodogram method used with the UKMO data. As a result of this, some disturbances found in the MLT wind could be affected by superposed motions with similar period but with different zonal structure. The SABER measurements, however, give some information about the SPWs in the zonal and meridional wind for the height range between 80 and 105 km. We note that such information is absent for latitudes higher than 52°N for the period of time before and during the onset of the SSW (November–December 2003) because of the 60-day gap in the SABER measurements.
3. Planetary Waves in the Stratosphere
 The Arctic winter of 2003/2004 was characterized by a major SSW that took place in December/January. This event was particularly remarkable because of an extended period of nearly 2 months of polar vortex disruption and a reversed direction of the zonal mean flow observed in the middle and lower stratosphere [Manney et al., 2005]. The vertical and latitudinal structure of the large-scale temperature and zonal wind anomalies associated with this major SSW has been investigated in detail by Mukhtarov et al. . Here the SSW is only briefly described in order to put the planetary wave activity in the context of the SSW development.
Figure 1, top, shows the altitude-time cross section of the zonal mean temperature, while Figure 1, bottom, shows the zonal mean zonal wind at 60°N. During normal winter conditions, the stratosphere is dominated by a cold pole and a strong eastward zonal mean zonal wind. A large temperature anomaly composed of three warm pulses could be distinguished between day numbers 60 (end of November 2003) and 100 (beginning of January 2004). The rapid deceleration of the zonal mean wind (Figure 1, bottom) started at the time of the first temperature pulse; the wind reversed to westward (breakdown of the polar vortex) at the time of the second pulse, around day number 80 (∼20 December). The reversal of the zonal mean zonal wind, however, did not penetrate lower than 40 km height. The third temperature pulse (Figure 1, top) developed around days 90–100 and was accompanied by another reversal of the mean zonal flow. This time the reversal penetrated deep into the stratosphere, reaching ∼25 km (∼30 hPa). The zonal mean zonal wind became westward at 10 hPa and 60°N around day 95, thus satisfying the WMO criteria for a major SSW (this is an event where both the zonal mean temperature gradient and zonal mean winds at 10 hPa reverse sign poleward of 60°). After day ∼100 the stratosphere rapidly cooled and the westward mean circulation was replaced by a normal winter eastward circulation in the upper stratosphere (above 10 hPa, i.e., ∼32 km). In the middle and lower stratosphere, however, the disruption of the polar vortex persisted until middle of February (day number 135–140, i.e., ∼15 February).
 Observations of planetary waves in the stratosphere are commonly viewed using the geopotential height because of its relationship to wave activity [Andrews et al., 1987]. In this study we instead use horizontal winds for several reasons.
 1. The radar data for the upper mesosphere consist of wind measurements. Since these data include neither continuous temperature profiles nor a horizontal span of wind measurements, it is not possible to get even a rough estimate of the geopotential from them. However, because of the geostrophic relation, one can relate the wave spectrum in the geopotential height to that of the meridional wind.
 2. The background zonal wind in the stratosphere has a strong effect on the propagation of planetary waves, as represented by the refractive index [Charney and Drazin, 1961]. This remains important as the waves propagate from the stratosphere to the mesosphere. In addition, gravity waves become an important part of the dynamical activity. The gravity waves are most sensitive to the local horizontal winds (zonal mean plus that due to all planetary waves). Gravity waves can be filtered by the winds and also interact with the wind in such a way as to damp and/or generate planetary-scale waves [Smith, 1996].
3.1. Spectral Analysis
 To determine the main periods of zonally eastward and westward propagating waves from the UKMO data, we use a two-dimensional analogue of the Lomb-Scargle periodogram method based on a least-squares fitting procedure applied to the entire 7 month time series. The planetary waves with periods between 5 and 30 days and zonal wave numbers up to 3 are studied here. Figure 2a shows the latitudinal amplitude spectra of the eastward propagating waves present in the zonal (middle column) and meridional (right column) wind at 1 hPa pressure level. A similar analysis for the geopotential height is also shown (Figure 2a, left); note the close correspondence between that and the spectra for meridional wind. This is a result of the geostrophic approximation, which relates the zonal gradient of geopotential (which in turn is related to the zonal wave number and amplitude) to the meridional wind. Figure 2a, top, shows the spectra for eastward waves with zonal wave number s = 1; Figure 2a, middle, shows the spectra for s = 2; and Figure 2a, bottom, shows the spectra for s = 3. A careful inspection of the spectra at each row (each zonal wave number) reveals that the spectra of the geopotential height, zonal, and meridional winds have very similar periods. They indicate four prevailing periods visible at all zonal wave numbers: ∼30, 22–24, 15–17, and 11–12 days. The strongest peaks are those for waves with zonal wave number 1. There are three latitudinal amplifications of the s = 1 wave in the zonal wind: the main amplification is centered at ∼40–50°N, and the other two are near the North Pole and ∼20°N. The spectra of the geopotential height and meridional wind have a single latitudinal maximum centered at ∼60–70°N for the geopotential height and near the North Pole for the meridional wind. The strength of the spectral peaks for planetary waves with zonal wave number 2 and 3 are at least a factor of 2 weaker than those for s = 1. On the average, the zonal wind spectra for s = 2 and s = 3 show two latitudinal maxima, centered at ∼60°N and ∼30°N. The spectra of the geopotential height and meridional wind again are dominated by a single latitudinal maximum centered at ∼50°N for both parameters.
Figure 2b is analogous to Figure 2a, but it is for the westward propagating waves. In this case the spectral peaks for some wave numbers do not emerge so clearly. However, we can distinguish maxima at similar periods to those for the eastward propagating waves, i.e., ∼30, ∼22–25, 15–17, and ∼12 days, but in this case the ∼22–25-day peak is the strongest one. Again there are two latitudinal amplifications of the waves in the zonal wind, centered at ∼60°N and ∼30°N, while the spectra of the geopotential height and meridional wind have one latitudinal maximum centered at ∼50–60°N for both parameters.
 To facilitate the comparison between the strength of the spectral peaks for the eastward and westward propagating waves, the same scales have been used for Figures 2a and 2b. It is evident that in most cases the eastward peaks are stronger than the westward ones. This is particularly valid for the ∼30-day peaks for s = 1 and all peaks for s = 3. The ∼22–24-day peaks for s = 1 westward propagating waves are, however, stronger than the analogous ones for the eastward propagating waves.
 The spectral analysis also revealed clear evidence of zonally symmetric planetary waves (s = 0) with very large amplitude present in all of the UKMO data except the meridional wind (a result of the geostrophic approximation), for the Arctic winter of 2003/2004. Figure 2c shows the latitudinal amplitude spectra of the zonally symmetric waves (s = 0) in the zonal wind (Figure 2c, right) and in the geopotential height (Figure 2c, left) at 1 hPa pressure level. Both spectra are very similar and indicate the same prevailing periods as those for the eastward and westward propagating waves: ∼30, ∼22–24, 15–17, and 11–12 days. Again there are two latitudinal amplifications of these periods in the zonal wind spectrum: the main amplification is centered around 50–60°N, and the secondary one is centered around 20–30°N. Analogous to them there are two maxima in the geopotential height spectrum centered near the North Pole and ∼40°N. The presence of global-scale zonally symmetric waves over the entire NH high stratosphere during the Arctic winter of 2003/2004 has been studied by Pancheva et al. ; characteristics of these waves (amplitudes and phases) will not be shown in this study.
3.2. Planetary Wave Characteristics
 It was mentioned before that the main purpose of this study is not only to define the predominant periods of the wave components that contribute to the variability of the atmospheric fields but also to isolate and study these waves in detail. To extract the waves from the data or to determine their amplitudes and phases, we use again a least-squares fitting procedure, but this time it is applied to a time segment twice the length of the longest period under investigation. Then this segment is moved through the time series with steps of 1 day in order to obtain daily values of the wave amplitudes and phases.
 The spectral analysis indicated a strong peak at ∼30 days; however, as the peak is situated just at the end of the interval studied, its exact period cannot be determined. This oscillation will not be extracted from the data and will not be discussed further. The three other oscillations revealed by the spectral analysis with periods of ∼22–23, 16, and 11 days and zonal wave numbers up to 3 are extracted from the UKMO zonal and meridional wind data. The zonally symmetric waves with the same periods as those of the zonally traveling waves are included in the decomposition of the zonal wind data (recall that these waves are absent in the meridional wind data). It is well known that the strongest planetary waves in the winter stratosphere are the SPWs; they should be included in the decomposition as well. We perform a linear two-dimensional (time-longitude) least-squares fitting including traveling waves with the periods listed above and with zonal wave numbers up to 3, together with the first three modes of the SPWs (i.e., SPWs with zonal wave numbers 1, 2, and 3). All oscillations are extracted simultaneously in order to avoid a possible distortion of the usually weaker traveling waves by the stronger SPWs. The length of the time segment used for performing the least-squares fitting procedure is 45 days. We note that in order to cover the entire interval of time between 1 October 2003 and 30 April 2004 with daily values of the wave parameters we perform the analysis on an extended interval by adding 22 days to each side (9 September 2003 to 22 May 2004). The waves at a given latitude and height (log-pressure height) are extracted by using the following expression:
where t is time in days counted from 1 October 2003, l is longitude in degrees, s is zonal wave number, Tj are the investigated wave periods, i.e., T1 = 23 days, T2 = 16 days, and T3 = 11 days, and R is the residual from the least-squares best-fit procedure including oscillations with periods less than 11 days and the noise of the data. From (1) the zonal phase speed is cpx = = (in degrees per day). Positive wave numbers (s > 0) are eastward, and negative ones are westward propagating waves. The first three terms on the right-hand side of (1) are as follows: (1) the mean wind V0, (2) traveling waves (including zonally symmetric waves) with amplitudes Vsj and phases ϕsj, and (3) stationary waves with amplitudes Ws and phases ψs. In the above expression the phase of the traveling wave (including zonally symmetric waves as well) is defined as the time of the wave maximum at zero longitude, while the phase of the stationary wave is as the longitude of the wave maximum. The phase of the traveling wave is presented in degrees as 360t/Tj.
3.2.1. Stationary Planetary Waves
Figure 3 shows the latitude-time cross sections of the amplitudes and phases of the SPW1 (Figure 3, left) and SPW2 (Figure 3, right) in the UKMO zonal wind data at four pressure levels: 30, 10, 1, and 0.3 hPa (the latitude range of 0–80°N is studied). The result for the SPW3 is not shown here because this wave is significantly weaker (maximum amplitude ∼6–7 m/s) than the first two stationary modes. The zonal wind maxima of SPW1 are centered near 40–50°N and near the North Pole. SPW1 reaches amplitudes up to 50 m/s, and its first maximum is observed between days 60–70. This first amplification apparently plays a key role in causing the major SSW. Most probably the rapid growth of the stationary wave is the reason for the rapid decrease of the winter jet after day 70 (see Figure 1, bottom). The second maximum is observed between days 130–140 and is seen mainly in the upper stratosphere (above 10 hPa pressure level). It is related to the rapid recovery of the upper stratosphere after day 100–110 when it again becomes cold and with normal eastward circulation. The phases of the SPW1 at the four pressure levels for 50°N are shown in Figure 3, bottom left. The westward phase tilt with height indicates that this is a vertically upward propagating wave with quite a large vertical wavelength, ∼ 80–90 km.
Figure 3, right, shows the amplitudes and phases of SPW2. This wave amplifies around and after the maximum of the SSW and reaches amplitudes up to 22 m/s in the high-latitude upper stratosphere. The phases of SPW2 at 70°N, shown in the Figure 3, bottom right, indicate that this wave is vertically upward propagating and, similarly to SPW1, it has a large vertical wavelength. There is also clearly outlined eastward phase drift of this wave with time beginning in late December or early January.
Figure 4 is similar to Figure 3 but for SPW1 and SPW2 in the meridional wind. The wave pattern of s = 1 is similar to that in the zonal wind except that the meridional wind maxima are situated near the North Pole and reach amplitudes of up to 55 m/s. SPW2 also amplifies around the time of the SSW; the meridional wind maximum is centered near 50–60°N in the upper stratosphere. Both SPW1 and SPW2 are upward propagating waves with large vertical wavelengths. Similar to SPW2 in the zonal wind, an eastward phase drift with time is present in the phases of the meridional wind SPW2 (Figure 4, bottom right).
 The SPWs for both wind components were calculated from the SABER data as well by using a procedure described in section 2.2. The results (not shown here) revealed that not only the wave pattern but also the absolute values of the wave amplitudes calculated from the SABER are very similar to those calculated from the UKMO data (Figures 3 and 4). This comparison serves as a test for the correctness of the calculation used to convert the SABER temperature and geopotential data to horizontal winds.
3.2.2. Traveling Planetary Waves
 It was mentioned before that all oscillations with periods ∼22–23, 16, and 11 days and zonal wave numbers up to 3 have been extracted from the UKMO zonal and meridional data. In the present paper only the parameters of the westward 23-day and eastward and westward 16-day waves with zonal wave number 1 (referred to as 23-day W1 and 16-day E1 and 16-day W1 waves) will be presented. We focus on these because similar oscillations were found and analysed in the MLT region. The relationship between the stratosphere and mesosphere oscillations will be discussed in detail in section 5.
 The strongest westward traveling wave present in the spectra of the zonal and meridional wind shown in Figure 2b is that with period ∼22–24 days. Figure 5 shows the latitude-time cross sections of the amplitudes and phases of the 23-day W1 wave in the zonal (Figure 5, left) and meridional (Figure 5, right) UKMO wind data. The results are again for the four pressure levels: 30, 10, 1, and 0.3 hPa and for the latitudinal range 0–80°N. The 23-day W1 wave amplitude patterns in both zonal and meridional winds are very similar; the wave appears and amplifies rapidly around days 70–80, and its trail can be traced out to the end of March, i.e., day 180. The 23-day wave zonal wind maximum is centered near 50°N, while the meridional wind maximum is near the North Pole. The 23-day W1 wave reaches amplitudes of 24 m/s in the zonal wind and 26 m/s in the meridional wind. Figure 5, bottom, shows the wave phases at the four pressure levels at selected latitudes where the waves are strong, i.e., 50°N for zonal wind and 70°N for meridional wind. The phase patterns in both zonal and meridional winds are again very similar; they indicate that the wave is vertically upward propagating (downward phase tilt with height) with a very large vertical wavelength, ∼130–150 km (calculated from the mean vertical phase gradient).
Figure 6 shows the latitude-time cross sections of the amplitudes and phases of the 16-day E1 wave present in the zonal (Figure 6, left) and meridional (Figure 6, right) UKMO wind data. This wave amplifies in both wind components before the onset of the SSW, or between days from 40 to 90, and most probably it contributes to the breakdown of the polar vortex. The 16-day E1 wave reaches amplitudes up to 22 m/s in the zonal and 24 m/s in the meridional wind. The wave phases are shown in the bottom row of plots and for both components they indicate that the 16-day E1 wave is downward propagating (upward phase tilt with height) with a vertical wavelength of ∼75 km. We note that some, but not all, of the other eastward traveling waves extracted from the UKMO wind data also revealed downward propagation.
Figure 7 shows the latitude-time cross sections of the amplitudes and phases of the 16-day W1 wave present in the zonal (Figure 7, left) and meridional (Figure 7, right) UKMO wind data. This wave amplifies in both wind components before and during the onset of the SSW, or between days 70 and 120. In the zonal wind there is a burst of significant amplification of this wave centered near day 55, which spreads to lower latitudes. The 16-day W1 wave is slightly weaker than the 16-day E1 wave reaching amplitudes ∼17 and 22 m/s, for the zonal and meridional winds, respectively, but its duration, particularly in the zonal wind, is longer than that of the 16-day E1 wave. Figure 7, bottom, again shows the wave phases at the four pressure levels and for latitudes where the wave is strong, i.e., 50°N for zonal wind and 70°N for meridional wind. Both wind components indicate that the 16-day W1 wave is upward propagating (downward phase progression) with vertical wavelength ∼50–55 km, calculated from the mean vertical phase gradient when the phases are relatively stable (between days 70–100).
4. Planetary Waves in the MLT Region
 As noted above, the planetary waves in the MLT region were studied using radar measurements at eight stations: four at high latitudes (63–69°N, see Table 1) and four at midlatitudes (52–55°N, see Table 2). Since the analysis of the planetary waves in the stratosphere indicated their latitudinal dependence, we studied the planetary wave activity in the MLT region separately for high and midlatitudes.
4.1. Planetary Waves in the MLT Zonal Wind
 Preliminary analysis of the zonal wind data reveals that the strongest wave observed before the onset of the SSW had a period of 15–17 days. At most of the stations its amplitude maximizes near 88–93 km height. Figure 8 shows the wavelet amplitude spectra (left) calculated in the period range from 3 to 30 days for the zonal wind at ∼88 km height measured at 3 high-latitude stations. The data from the fourth high-latitude station, Poker Flat, also hint at long-period variability before the onset of the SSW but, for altitudes higher than 82 km, they have many gaps and the wavelet transform cannot be applied to them (it requires only short-term gaps that can be interpolated). All spectra indicate a peak near 16–17 days, which maximizes near days 60–70. The detailed analysis of this oscillation revealed that it is an upward propagating wave with vertical wavelength ∼50–52 km. Figure 8, top right, shows the 17-day filtered zonal wind data at Esrange in order to demonstrate the downward phase progression of this wave. The zonal wave number is estimated from the longitudinal distribution of the wave phases which are calculated by a least-squares best fit approach. Only the data from 1 November to 31 December were used for this purpose. Figure 8, bottom right, shows the longitudinal distribution of the 16–17-day wave phases; it indicates that the observed 16–17-day wave in the zonal wind is a westward propagating wave with a zonal wave number close to 1.
Figure 9 shows the wavelet spectra calculated from the zonal wind measured at the midlatitude stations. Recall that the meteor radars at Obninsk and Castle Eaton operate without height finding so the results represent a sampling of winds in the MLT region with weighting reflecting the vertical distribution of meteor echoes. This distribution is strongly peaked at heights near 90–93 km. At least some indication of a peak with period 15–17 days can be distinguished in all four stations; however, the oscillations at Kuhlungsborn and Castle Eaton have period that decrease with time from 16–17 days to 12–13 days. The oscillation is observed at all midlatitude stations between days 50–110, or before and during the onset of the SSW. The phases are found again by a least-squares best fit procedure, but in this case the data from the period of time between 15 November and 15 January were only analyzed. The longitudinal distribution of the phases revealed that, similarly to the high latitudes, the 15–17-day oscillation is a westward propagating wave with a zonal wave number close to 1 (Figure 9, bottom).
 The analysis of the zonal MLT wind also indicates the presence of a longer-period oscillation, ∼23–24 days, which can be distinguished at all stations except Saskatoon (Figure 9). This oscillation is particularly strong at high latitudes. Figure 10 shows the amplitude wavelet spectra calculated from the zonal wind measured at 81–82 km height at the four high-latitude stations. A strong 23–24-day peak is clearly evident in all spectra and it maximizes near days 100–120, or during the SSW. We note that the timing of this oscillation in the Poker Flat analysis could be influenced by the long gap after day 126.
 The detailed study of the 23–24-day oscillation indicated the following features:
 1. The oscillation is very strong near 70–80 km and above 94 km and minimizes near 88–90 km. This feature is clearly visible in the data from Andenes and Esrange, where the measurements extend up to 97–98 km. Figure 11, left, shows the amplitude wavelet spectra calculated at Andenes for 3 heights: 82, 91, and 98 km. To facilitate the comparison, the scales are the same at all altitudes. It is clearly visible that this oscillation is strong at 82 and 98 km and very weak at 91 km.
 2. The investigation of the longitudinal distribution of the phases, made by the cross-wavelet analysis applied to all data except those at Saskatoon, did not reveal a definite direction of propagation. The data, however, hint at the presence of eastward propagation of this long-term disturbance which can be found only at lower heights (∼80 km) of the high-latitude stations; it appears first at western stations and later at eastern ones. Figure 11, right, shows this feature; the 23–24-day oscillation appears first at Poker Flat and Yellowknife and later at Andenes and Esrange.
4.2. Planetary Waves in the MLT Meridional Wind
 The planetary wave activity in the meridional MLT wind will also be studied separately for high and midlatitude stations. It was particularly strong during the period of the SSW.
 Similarly to the zonal wind, a 15–16-day oscillation is observed in the meridional MLT wind. The preliminary analysis showed that at high latitudes, particularly before the onset of the SSW, this oscillation amplified at altitudes around 80–85 km height. Figure 12, left, shows the wavelet spectra calculated from the meridional wind measured at 81–82 km height for all four high-latitude stations. The wave peak maximizes around days 80–90 or just before the onset of the SSW. Figure 12, top right, shows the 15-day filtered meridional wind data at Andenes in order to demonstrate the downward phase progression of this wave. The average vertical wavelength of the 15–16-day wave calculated from the mean vertical phase gradient is ∼55 km. The longitudinal distribution of the 15–16-day wave phases (calculated by a least-squares best fit approach applied to the data for the period of time between 1 December and 31 January) revealed that it is a westward propagating wave with zonal wave number very close to −1 (Figure 12, bottom right).
Figure 13 shows the wavelet spectra calculated from the meridional MLT wind measured at the midlatitude stations. A strong peak with period close to 15 days is clearly visible at all stations. It maximizes near day 90 at all stations except Saskatoon (the maximum there is slightly later, near day 100), or just during the onset of the SSW. The longitudinal distribution of the phases (obtained in the same way as those for the high-latitude stations) revealed that, similarly to the high latitudes, the 15–16-day oscillation is a westward propagating wave with zonal wave number very close to −1 (Figure 13, bottom).
 The analysis of the meridional MLT wind, similar to the zonal wind, reveals the presence of a longer-period oscillation, ∼22–25 days. This oscillation is observed mainly at high latitudes and maximizes at the upper altitudes measured. Figure 14 shows the amplitude wavelet spectra calculated from the meridional wind at ∼94 km for all stations except Poker Flat, where the spectrum is for 82 km height (we recall that because of many gaps the wavelet spectra cannot be calculated for upper levels). The oscillation is clearly outlined in all spectra; the prevailing periods are clustered around 22–23 days, except Poker Flat where its period is ∼25 days.
 An investigation of the longitudinal distribution of the 22–25-day wave phases was made by a cross-wavelet analysis. It was applied to the both pairs of stations: Andenes-Poker Flat for altitude 82 km and Esrange-Yellowknife for 94 km in order to see if the zonal structure of this disturbance changes with height. The cross-wavelet analysis revealed that for the lower heights (∼80 km) there is a signature for westward propagation of the ∼23-day disturbance with instantaneous zonal wave number between −0.6 and −0.7 (or closer to −1 than to 0), while for the upper levels the zonal wave number is near −0.5, i.e., the zonal phase progression does not give a definite solution for the zonal wave number. We note that the obtained result for the instantaneous zonal wave number at height of ∼80 km (between −0.6 and −0.5) means that there is a signature for westward propagation, but most probably this disturbance is affected by another oscillation present concurrently with the first one with a similar period but different zonal structure; this could be SPW or even E1 component (but not westward propagating components).
5. Discussion and Summary
 The main focus of this paper is to study the types of planetary wave observed in the stratosphere and mesosphere during the Arctic winter between 1 October 2003 and 30 April 2004 and the vertical coupling of the stratosphere-mesosphere system through the quasi-stationary and traveling planetary waves before and during the major SSW that occurred in December/January. Three intrinsically different types of data are used for this study. One is daily global synoptic analyses for the stratosphere made by UKMO, which incorporate satellite and in situ observations. The second is satellite data from SABER for the entire middle atmosphere. They have three limitations: (1) latitudinal coverage is limited, particularly in high northern latitudes during midwinter, (2) local time sampling is slow so there is some contamination of the analyzed waves by nonmigrating tides in the MLT region, and (3) the basic data need to be converted from temperature and geopotential to horizontal winds. However, the SABER measurements extend up into the MLT and are therefore valuable for determining stationary waves in the upper part of the middle atmosphere. The third type of data is horizontal wind measurements from eight ground-based radars; their distribution is four at high latitudes (63–69°N, Table 1) and the other four at midlatitudes (52–55°N, Table 2). The three different types of data require different analyses, as described in this paper.
 The first step was to search for planetary waves in the UKMO stratosphere data using a two-dimensional analogue of the Lomb-Scargle periodogram method based on least-squares fitting procedure; waves with periods between 5 and 30 days and zonal wave numbers up to 3 were studied. The analysis indicated that four prevailing periods are present in spectra for all zonal wave numbers of both eastward and westward traveling waves (Figures 2a and 2b). These periods are ∼30, 22–24, 15–17, and 11–12 days. It was shown also that there are usually two latitudinal amplifications of these periods in the zonal wind spectra centered near ∼50–60°N and ∼20–30°N, while those in the meridional wind are strong at high latitudes. It is worth noting that in most cases the eastward peaks were stronger than westward ones and that this is quite unusual for the winter dynamics of the NH.
 The next step in our analysis was to extract from the UKMO zonal and meridional winds all waves with zonal wave numbers up to 3 that were either stationary or were traveling and had periods identified by the spectral analysis (except ∼30 days). Only the characteristics of SPW1 and SPW2 were shown in this study (Figures 3 and 4) because SPW3 was significantly weaker than the first two modes (in both zonal and meridional wind). The wave patterns of SPW1 and SPW2 in the zonal wind were similar to those in the meridional wind. SPW1 had two maxima during the winter; the first centered at days 60–70 and the second near days 130–140. The first amplification played a key role in generating the major SSW, while the second one was seen mainly in the upper stratosphere (above 10 hPa pressure level) and was related to the rapid recovery of the upper stratosphere after the onset of the major SSW. The SABER satellite data were analysed as well in order to extract the zonal and meridional wind SPW1 and SPW2 in the stratosphere and mesosphere. The comparison between the UKMO and the SABER SPWs indicated a high degree of similarity (not shown in this study) supporting the correctness of the horizontal wind calculation procedure.
 This study focuses in detail on three traveling oscillations that were prominent in both the stratosphere and the MLT region. These are ∼16-day E1, ∼16-day W1, and ∼23-day W1 waves.
 The 16-day E1 wave (Figure 6) amplified in both wind components before the onset of the SSW and reached large amplitudes of ∼22–24 m/s. The phase analysis indicated that it was a downward propagating wave (upward phase progression) with a vertical wavelength of ∼75 km. It is worth mentioning that Merzlyakov and Pancheva  investigated the short-period waves (∼1.5–5 days) observed in the SABER and radar wind measurements during February of the same winter season and also found eastward and downward propagating ∼2- and 4-day waves. From this, we speculate that the presence of eastward traveling planetary waves with upward phase (downward energy) progression during the Arctic wintertime is probably not such a rare event; these waves deserve to be studied in detail. The investigation of the downward planetary wave coupling in the middle atmosphere is very important because it has been suggested as a possible solar variability-climate link [Arnold and Robinson, 1998; Lean and Rind, 2001; Ruzmaikin et al., 2004].
 The 16-day W1 wave (Figure 7) amplified in both wind components before and during the onset of the SSW and reached amplitudes of ∼17–22 m/s, slightly smaller than those of the 16-day E1 wave. Comparing the 16-day E1 and W1 waves we note that while the main amplification of the E1 mode was centered at days 40–60, that of the W1 wave was later, just before and during the onset of the SSW. The phase analysis of the 16-day W1 wave indicated that it was an upward (downward phase progression) propagating wave with a vertical wavelength of ∼50–55 km.
 A strong 15–17-day wave was found in the zonal (Figures 8 and 9) and meridional (Figures 12 and 13) MLT winds measured at high and midlatitudes. The wave amplifications were observed near days 50–90 in the zonal wind and later, near days 80–110, in meridional wind. The latter amplification was just before and during the SSW. The phase analysis indicated that this was an upward propagating wave (downward phase progression) in both wind components with a vertical wavelength of ∼50–55 km.
 The zonal wave number analysis of the 16-day MLT wave in both wind components revealed that this is a westward propagating wave with zonal wave number close to –1. This result was obtained by the longitudinal distribution of the wave phases measured at 4 stations. We note that the disturbances found in the MLT wind could be affected by superposed motions with similar period but with different zonal structure. However, we cannot separate them because the number of the radars is far from being sufficient for applying a two-dimensional (time-longitude) spectral analysis. Therefore, the result for the zonal wave number of the ∼16-day MLT wave means that the W1 component has predominant effect on the 16-day disturbance observed in the MLT region, or most probably its origin is the stratosphere upward propagating ∼16-day W1 wave. Additional evidence for the predominant effect of the stratosphere W1 component is that both waves are upward propagating, have similar vertical wavelength (∼50–55 km) and amplify almost simultaneously.
 A longer-period, ∼22–24-days, oscillation dominated the dynamics of the MLT region particularly during the onset of the SSW. Taking all aspects of this oscillation together, it did not fit clearly with any simple interpretation, so we will spend some time here discussing its characteristics and what they imply about its origin.
 In both wind components the ∼22–24-day oscillation was an upward propagating disturbance, but its longitudinal and vertical characteristics indicated some differences between the zonal and meridional winds. In the zonal wind the oscillation was very strong near 70–80 km and above 94 km height and minimized near 88–90 km height. The longitudinal distribution of the phases did not reveal a definite direction of propagation. We could determine only that there was a signature of eastward propagation of the disturbance at lower heights (∼80 km): it appeared first at western stations and later at eastern ones. Therefore, we cannot establish a clear relationship between this long-period oscillation composed of only two cycles and the ∼23-day W1 wave found in the stratosphere.
Figure 15, left, shows the altitude-time cross sections of the zonal winds at the locations of Poker Flat (bottom) and Esrange (top) during the winter of 2003/2004. The UKMO stratospheric zonal wind data for the geographical locations of Poker Flat and Esrange and levels between 30 hPa and 0.3 hPa are used for the lower parts of the panels; radar data are used above. Both plots indicate that the two large disturbances (dark features) observed in the MLT region up to 90 km just after the onset of the SSW are upward propagation of the analogous disturbances in the upper stratosphere. During this period the SPW2 in the stratosphere was particularly strong while the SPW1 was growing toward its second maximum (Figure 3), hence the ∼22–24-day MLT disturbance could be affected by the SPW2 if it penetrates to the mesosphere levels.
 The contribution of the mesospheric SPW1 to the ∼22–24-day disturbance in the zonal MLT wind cannot be ruled out because this wave is found to be very strong in the MLT region. Figure 16 shows the latitude-time cross sections of the SPW1 (top left) and SPW2 (bottom left) calculated from the SABER derived zonal winds. In spite of the 60-day gap between days 46 and 106 for latitudes higher than 52°N, the amplification of the SPW1 after the onset of SSW, particularly at 81 km height, is clearly visible. However, there is no clear indication of an amplification of the SPW2 in the MLT region after the onset of the SSW.
 It is worth mentioning that the ∼22–24-day disturbance in the zonal MLT wind could also be affected by the ∼23-day vertically upward propagating zonally symmetric (s = 0) wave found in the UKMO data. Pancheva et al.  showed that there was a strong burst of this wave (Figure 3, top, of Pancheva et al.) observed at the time when the ∼22–24-day wave is present in the zonal MLT wind. A possible effect of the stratospheric ∼23-day (s = 0) wave on the ∼22–24-day MLT disturbance could have partly caused the observed difference between the vertical and longitudinal characteristics of this disturbance in the zonal and meridional MLT wind components (we recall that the ∼23-day zonally symmetric wave is present only the UKMO zonal wind).
Section 4 showed that the onset of the SSW was accompanied by a strong ∼22–24-day oscillation in the meridional MLT wind as well. This disturbance had a different vertical structure than that in the zonal wind; it amplified at upper levels. The zonal wave number estimation again did not give a definite solution; only a signature of westward propagation at altitudes around 80–85 km was found. The speed of its longitudinal phase progression, however, gradually decreased with increasing the height. The direction of propagation of this disturbance resembles that of the ∼23-day W1 wave found in the stratosphere. However, the two oscillations have different vertical wavelengths; the stratosphere ∼23-day W1 wave has very large wavelength, ∼130–150 km, while that in the MLT meridional wind has a shorter wavelength of about ∼85–90 km. We have to have in mind also that the temporal changes in the background conditions for vertical propagation of the planetary waves just before and during the onset of the SSW could affect the penetration of the ∼23-day W1 wave in the mesosphere. The structure of the wave could change with altitude because of changes in the background atmosphere or to damping by gravity waves. Because of all of these aspects, we cannot confirm from the meridional wind data that the ∼23-day wave in the MLT is an upward extension of that in the stratosphere.
 The amplification of the meridional ∼22–24-day oscillation in the MLT region during the SSW could also be related to the SPWs in the mesosphere. Figure 16 shows the latitude-time cross sections of SPW1 (top right) and SPW2 (bottom right) in the SABER MLT derived meridional wind. SPW1 indicates significant amplification particularly at 81 km after the onset of the SSW. There is a signature of strengthening of the SPW2 at 81 and 94 km as well. Therefore, the meridional ∼22–24-day oscillation observed in the MLT region is most probably forced by the SPWs at least for altitudes up to 80–85 km. The significant amplitude growth of the ∼22–24-day oscillation above 90 km height, well documented at Andenes and Esrange, hints at possible contribution of filtered gravity waves.
 As it was mentioned before, the ∼22–24-day oscillation in the zonal MLT wind amplified significantly above 90 km height. A careful inspection of Figure 15, top left (the zonal wind at the location of Esrange) indicates disturbances with strong eastward wind at the uppermost level (97 km) almost coinciding with the reversal of the circulation in the stratosphere and lower mesosphere. Figure 15, top right, shows the altitude-time cross section of the zonal wind measured at Esrange only between days 80 and 160, while Figure 15, bottom right, shows the zonal mean zonal wind at 70°N in the stratosphere between the altitudes 10 and ∼52 km height. The zero line of the wind in the stratosphere is marked by a thick dashed line to emphasize the winter jet reversals. It is known that the deceleration and reversal of the winter eastward jet in the stratosphere also changes the filtering of the gravity waves and allows eastward propagating gravity waves from the lower atmosphere to penetrate and deposit their momentum into the MLT region [Dunkerton and Butchart, 1984]. In this way the previously dominant westward forcing of the MLT region is replaced by eastward forcing that increases with height when the winter eastward jet is reversed to the westward in the stratosphere. The arrows in Figure 15, top right, show two strong disturbances observed in the upper MLT region that coincide with the reversal of the zonal mean circulation in the stratosphere (Figure 15, bottom). This coincidence strongly suggests that the eastward perturbations in the upper MLT region are generated by the change in the filtering in the stratosphere that now allows eastward gravity waves to penetrate to the MLT.
 In conclusion, we note that investigating the planetary waves in coupling the stratosphere-mesosphere system during the major SSW in the winter of 2003/2004 we found that prior to the SSW the stratosphere-mesosphere was dominated mainly by an upward and westward propagating ∼16-day wave detected simultaneously in the UKMO and MLT zonal and meridional wind data. After the onset of the SSW, longer-period (∼22–24 days) oscillations were observed in the zonal and meridional MLT winds that likely were made up of a combination the upward propagation of SPWs (up to 80–85 km) and in situ generation by the dissipation or breaking of gravity waves (above altitudes of 90 km) that were filtered by the winds in the stratosphere.