Stratospheric gravity wave characteristics and seasonal variations observed by lidar at the South Pole and Rothera, Antarctica

Authors


Abstract

[1] We present an observational study of stratospheric gravity wave spectra and seasonal variations of potential energy density at the South Pole (90°S) and Rothera (67.5°S, 68.0°W), Antarctica. The gravity wave spectra are derived from the atmospheric relative density perturbation in the altitude range of 30–45 km measured by an iron Boltzmann lidar. The ground-relative wave characteristics obtained at each location are comparable, with an annual mean vertical wavelength of ∼4.1 km, vertical phase speed of ∼0.7 m s−1, and period of ∼104 min. Approximately 44% of the observed waves show an upward phase progression while the rest display a downward phase progression in ground-based reference for both locations. Gravity wave potential energy density (GW-EP) at Rothera is ∼4 times higher than the South Pole in winter but is comparable in summer. Clear seasonal variations of GW-EP are observed at Rothera with the winter average being 6 times larger than that of summer. The seasonal variations of GW-EP at the South Pole are significantly smaller than those at Rothera. The absence of seasonal variations in wave sources and critical level filtering at the South Pole is likely to be responsible for the nearly constant GW-EP. The minimum critical level filtering in winter at Rothera is likely to be a main cause for the winter enhanced GW-EP, as this would allow more orography-generated waves to reach the 30 to 45 km range. The stratospheric jet streams may also contribute to the winter enhancement at Rothera.

1. Introduction

[2] Atmospheric gravity waves are a key element in defining large-scale circulation, thermal structure, chemical constituent distributions, and atmospheric variability from troposphere to thermosphere [Fritts and Alexander, 2003]. Poor representation of gravity waves in modern climate models is a leading source of uncertainty. Thus, improving the characterization of gravity wave spectra, propagation, dissipation, and source distributions is a very challenging and urgent issue for the atmospheric community. Numerous gravity wave observational techniques have been developed, and among them lidars provide high spatial/temporal resolutions and long-period observations to help address the issue. Chanin and Hauchecorne [1981] pioneered the lidar study of gravity waves. Following this, gravity waves were studied by lidars in the stratosphere and mesosphere at various midlatitude locations, e.g., Observatoire de Haute Provence (44°N, 6°E) and Biscarosse (44°N, 1°W) [Wilson et al., 1991a, 1991b], Urbana (40°N, 88°W) [Gardner et al., 1989; Senft and Gardner, 1991], and also in the Arctic [e.g., Whiteway and Carswell, 1994; Gerrard et al., 2000]. Lidar studies of gravity waves in Antarctica are rare. Collins and Gardner [1995] pioneered a gravity wave study at the South Pole using a Na lidar to measure gravity waves in the Na density between 80 and 105 km, and in polar stratospheric clouds between 12 and 30 km. Observations of Antarctic gravity waves were also made by balloon radiosonde, satellite UARS/MLS, all-sky imager, MF radar, and GPS radio occultation [Pfenninger et al., 1999; Tsuda et al., 2000; Hertzog et al., 2007, 2008; Wu and Jiang, 2002; Espy et al., 2004, 2006; Hibbins et al., 2007; Baumgaertner and McDonald, 2007]. However, most of these studies were either below 30 km or above 80 km, except two very recent reports by Innis et al. [2008] at Davis (68.6°S, 78°E) and Chu et al. [2009] at the South Pole and Rothera describing lidar studies of the correlation between polar mesospheric clouds and summer gravity waves in the 30 to 45 km range. Therefore, it is very valuable to obtain the missing piece of information for Antarctic gravity waves in between 30 and 80 km.

[3] The University of Illinois Fe Boltzmann lidar was operated at the South Pole (90°S) and then at Rothera (67.5°S, 68.0°W), Antarctica. Year-round atmospheric data at altitudes between 10 and 110 km were collected at both sites. Because this lidar functions as a Rayleigh lidar from 30 to 70 km and can provide atmospheric density data with high signal-to-noise ratios from 30 to 45 km, these data sets provide a unique opportunity to study stratospheric gravity waves and their seasonal variations at these two geophysically distinct locations. The Pole is on the Antarctic plateau while Rothera is on the western side of the Antarctic Peninsula surrounded by mountains and the Southern Ocean. Different wave features might be observed at these two sites because of different wave sources and the difference in their locations relative to the polar vortex that results in different propagation conditions. These observations therefore provide an opportunity to investigate wave sources, spectra, and propagation. In this paper we use the year-round data sets to characterize the stratospheric gravity wave spectrum, potential energy density and the seasonal variations in the altitude range between 30 and 45 km. Possible causes for the observed differences between the two sites are discussed in terms of potential differences in wave sources and critical level filtering.

2. Observations

[4] Gravity wave observations were made at the South Pole (90°S) from December 1999 to October 2001 and at Rothera (67.5°S, 68.0°W) from December 2002 to March 2005 using an Fe Boltzmann/Rayleigh lidar developed at the University of Illinois [Chu et al., 2002]. Gravity waves are manifested as fluctuations in the atmospheric number density derived from the Rayleigh scattered signals. Pure Rayleigh signals exist from ∼30 km (above aerosol layers) to 70 km (below the Fe fluorescence region). The signal-to-noise ratio (SNR) of lidar data decreases with height. For a 5-min integration and 192-m vertical resolution, the Fe Boltzmann lidar SNRs change from 62.6 at 30 km to 0.5 at 70 km. The altitude range of 30 to 45 km was chosen for its high SNR and also for comparison to the work of Gerrard et al. [1998].

[5] Lidar observations covered all 12 months throughout a year under both day and night conditions at the South Pole and Rothera. However, our data quality control removed the data in September and October for the South Pole because of the low Rayleigh counts and weather problems. Any data sets of less than 3 h were also excluded because of the use of a 3-h window in estimating the background density profile (see section 3). Hence, this gravity wave investigation covers 10 months of the year at the South Pole but all 12 months for Rothera, resulting in 618 and 507 observation hours respectively. Another criterion implemented to qualify the data as hours with gravity wave occurrence was that the RMS uncertainty of pure mean square relative density perturbation in the 3-h window had to be smaller than 50% of the perturbation itself [Chu et al., 2009]. The total numbers of gravity wave occurrence hours are 507 h at the South Pole and 471 h at Rothera. Among these hours, 300 h and 318 h occur during summer (November to February), equaling 77% and 90% of the summer observations for the South Pole and Rothera, respectively. In winter (May to August), gravity waves occur during 150 h and 90 h, equating to 100% of the winter observations at both locations. The monthly and diurnal distributions of gravity wave occurrence hours are illustrated in Figure 1, where Rothera local midnight is 0430 UT. The larger database for wave examination during summer was primarily due to increased personnel and improved weather conditions. The reduction in the fraction of data meeting the wave criteria during summer and around noon at Rothera, is due to the additional solar background and the accompanying reduction in the signal-to-noise ratio.

Figure 1.

Histograms of the gravity wave occurrence hours versus month and universal time for the South Pole and Rothera. See text for details of lidar observations and examinations.

3. Gravity Wave Spectral Analysis and Statistical Studies

[6] We characterize stratospheric gravity waves with an atmospheric relative density perturbation r(z, t). The derivation of r(z, t) from the lidar-obtained relative atmospheric density closely follows the approach taken by Gardner et al. [1989], Hostetler and Gardner [1990], and Gerrard et al. [2004]. Key points are summarized below. The r(z, t) is defined as:

equation image

where ρ(zR, t) is the perturbed total atmospheric density at an altitude of zR = 40 km for normalization purposes [Gardner et al., 1989]. The quantity directly measured by the lidar is the perturbed total atmospheric density, ρ(z, t), so we must estimate the background atmospheric density, ρ0(z, t), in order to derive the gravity wave-induced perturbation, ρ′(z, t) ≡ ρ(z, t) − ρ0(z, t). The background density ρ0(z, t) for gravity waves to propagate through is the atmosphere with the presence of tidal and planetary waves. These long-period waves should be preserved in the background estimation so that the obtained perturbation ρ′(z, t) is purely due to small-scale gravity waves. In this study ρ0(z, t) is derived for each 3-h sliding window in the following steps similar to those of Hostetler and Gardner [1990] and Gerrard et al. [2004].

[7] 1. The perturbed relative density profiles ρ(z, t)/ρ(zR, t) in a 3-h window are averaged to remove waves with shorter temporal periods, while preserving the longer temporal-period waves as part of the 3-h averaged profile. According to Kennedy [1980], such an average is equivalent to a sinc filter with a 3-dB cutoff at 4.97 h for this 3-h window.

[8] 2. A 5th order polynomial fit with respect to altitude is applied to this profile to simulate the trend of the unperturbed background and long vertical wavelengths [Shibata et al., 1986; Mitchell et al., 1990]. This fit is used to detrend the data for the subsequent filtering operation. By subtracting the fit from the 3-h averaged profile, the residual obtained is a detrended 3-h averaged profile.

[9] 3. The residual profile is filtered to remove vertical variations with wavelengths shorter than 2 km. This reduces the high-frequency photon noise but retain long-period waves with vertical wavelengths longer than 2 km.

[10] 4. The 5th order polynomial trend is then added back to this filtered (detrended) residual to form the background relative density, ρ0(z, t)/ρ(zR, t), for this 3-h window.

[11] This background relative density is therefore the sum of the unperturbed background density and perturbations induced by long-period waves (>5 h), such as tides and planetary waves. By subtracting ρ0(z, t)/ρ(zR, t) from each 5-min ρ(z, t)/ρ(zR, t), these long-period waves are removed so the obtained relative density perturbation r(z, t) given by equation (1) contains only the short temporal-period waves. To reduce photon noise in the perturbation profile, r(z, t) is low-pass filtered to remove variations with vertical wavelengths shorter than 2 km and temporal periods shorter than 60 min. In the following spectral analysis, we only extract waves with periods between 1 and 3 h and vertical wavelengths between 2 and 15 km (window length) that dominate the relative density perturbation. Nevertheless, the potential energy density derived in section 4 will include contributions from waves with periods of ∼1–6 h and vertical wavelengths of ∼2–30 km (double the window length), as pointed out by Senft and Gardner [1991] and Gerrard et al. [2004].

[12] The resultant 2-D relative density perturbation r(z, t) in the 3-h and 15-km window is then analyzed with a 2-D fast Fourier transform (FFT) method to reveal the wave spectrum. The power spectral density (PSD) for the data in Figure 2a is shown in Figure 2b. To identify the dominant waves, the PSD is converted to area-preserving form (Figure 2c), i.e., multiplying the PSD by the corresponding wave number and frequency following the approaches taken by Smith et al. [1987] and Tsuda et al. [2000]. The highest 1–3 peaks above the noise floor are then picked as the dominant waves. The vertical wavelengths λz and periods T of these waves are recorded and the corresponding vertical phase speeds are calculated as cz = λz/T. As demonstrated by Fukao et al. [1985] and Wilson et al. [1991a], the 2-D FFT spectral analysis allows us to distinguish gravity wave propagation directions. Using the data shown in Figure 2 as an example, the two spectral peaks with positive periods correspond to downward phase progression waves while the peak with a negative period represents upward phase progression. It is worth pointing out that what is obtained from the lidar data is the ground-relative (apparent) frequency, not the intrinsic frequency.

Figure 2.

An example of spectral analysis (17 March 2004 at Rothera). (a) Sequence plot of relative density perturbation profiles. Each profile represents a 5-min data integration and it is shifted by 1% to the right progressively. (b) Power spectral density of the relative density perturbations shown in Figure 2a. (c) Area-preserving form of Figure 2b. Positive period represents downward phase progression.

[13] The above spectral analysis method is applied to each 3-h and 15-km data. The number of waves picked from each 3-h window is not fixed but varied between 1 and 3 peaks per data window. In Table 1 we summarize the total numbers of waves identified from the 3-h data windows and corresponding numbers of ground-relative downward and upward phase progression waves. The fractions of upward and downward phase-progression waves are calculated for the entire seasons as well as for the summer and winter seasons separately. Statistically, the downward phase-progression waves (positive frequencies) are 59% in summer but 51% in winter at the South Pole, and 52% in summer but 62% in winter at Rothera. The overall fractions of downward phase-progression waves in ground-based reference are 56% for both locations. Ground-based lidars observe the ground-relative frequency and phase speed, not intrinsic frequency or intrinsic phase speed. As pointed out by Fritts and Chou [1987] and Fritts and Alexander [2003], when the mean background wind is comparable to or larger than the intrinsic phase speed of the wave motion, the ground-relative frequency observed by the lidar can have an opposite sign of the intrinsic frequency, thus confusing the vertical directionality of the wave. Because of the lack of background wind and horizontal wave number information, we cannot firmly state the fractions of upward and downward propagating waves. Nevertheless, the ground-relative (apparent) phase progression information is still very valuable for comparison with model simulations. Wilson et al. [1991a] reported 65–80% of apparent downward phase progression for 30–50 km at 44°N, much larger than our results. This difference could be due to the different background wind conditions or the additional sources above our observational heights.

Table 1. Gravity Wave Propagation Direction in the 30–45 km at the South Pole and Rothera
 Number of WavesDownward Phase ProgressionUpward Phase ProgressionDownward Phase to Total Wave Ratio
South Pole
   Total47126420756%
   Summer31818713159%
   Winter91464451%
Rothera
   Total50127922256%
   Summer30315914452%
   Winter150935762%

[14] We performed a statistical study on the spectra of both the upward and downward phase-progression waves identified above. Histograms of vertical wavelengths, vertical phase speeds, and periods of the dominant waves derived from the area-preserving form of each 3-h window are plotted in Figure 3 for both locations. The means, standard deviations, and standard errors of these spectral parameters are summarized in Tables 2 and 3 for the South Pole and Rothera, respectively. The overall statistics of gravity wave spectra are comparable between the South Pole and Rothera: the annual mean gravity wave vertical wavelengths are ∼4.1 km, the vertical phase speeds ∼0.7 m s−1, and the periods ∼104 min. There are no significant variations between the summer and winter seasons at either location considering the large standard deviations. Our observed vertical wavelengths are shorter than the previous results of 8–12 km observed by lidar between 30 and 45 km at Sondrestrom, Greenland (67°N, 310°E) [Gerrard et al., 1998], and 8–10 km observed between 25 and 45 km by lidar at Eureka (80°N, 86°W) [Whiteway and Carswell, 1994]. Gravity waves between 35 and 50 km observed by lidar at Urbana (40°N, 88°W) showed two peaks of vertical wavelength at 2–4 km and 7–10 km [Gardner et al., 1989]. Our histograms (Figure 3) also show a peak at ∼2–4 km and then some longer wavelength waves that are comparable to Urbana results. The 3-km vertical wavelength between 18 and 35 km observed by CHAMP/GPS in Antarctica (60°S–90°S) [Baumgaertner and McDonald, 2007] is shorter than our results in the 30 to 45 km range. These differences could be due to the different observational filters of the respective measurement techniques. Our observed vertical phase speed of ∼0.7 m s−1 is at the high end of previous observations: 0.01 to 0.85 m s−1 for Urbana and 0.55 to 0.83 m s−1 for Sondrestrom [Gardner et al., 1989; Gerrard et al., 1998]. The observed period of ∼104 min in this study is slightly shorter but comparable to the periods of 120–180 min observed at Sondrestrom by Gerrard et al. [1998].

Figure 3.

Spectra of the upward (darker histogram) and downward (lighter histogram) phase-progression gravity waves in the 30 to 45 km altitude range at (a) the South Pole and (b) Rothera: (top) entire season, (middle) winter (May to August), and (bottom) summer (November to February); (left) vertical wavelength, (middle) vertical phase speed, and (right) period.

Table 2. Gravity Wave Spectrum in the 30–45 km at the South Pole for Downward and Upward Phase Progression Waves
 Mean ± Standard Deviation (Standard Error)
Entire SeasonWinter (May to Aug)Summer (Nov to Feb)
Vertical wavelength (km)
   Downward4.1 ± 2.2 (0.13)4.8 ± 2.6 (0.27)3.7 ± 1.8 (0.14)
   Upward3.6 ± 1.8 (0.12)3.8 ± 1.9 (0.25)3.6 ± 1.9 (0.16)
Vertical phase speed (m s−1)
   Downward0.71 ± 0.41 (0.02)0.77 ± 0.48 (0.05)0.66 ± 0.34 (0.03)
   Upward0.66 ± 0.40 (0.03)0.66 ± 0.42 (0.06)0.68 ± 0.42 (0.04)
Period (min)
   Downward102 ± 26 (1.5)111 ± 28 (2.9)97 ± 23 (1.9)
   Upward97 ± 25 (1.6)108 ± 28 (3.7)93 ± 22 (1.8)
Table 3. Gravity Wave Spectrum in the 30–45 km at Rothera for Downward and Upward Phase Progression Waves
 Mean ± Standard Deviation (Standard Error)
Entire SeasonWinter (May to Aug)Summer (Nov to Feb)
Vertical wavelength (km)
   Downward4.4 ± 2.4 (0.15)5.3 ± 3.0 (0.44)4.1 ± 2.3 (0.15)
   Upward4.3 ± 2.4 (0.17)5.6 ± 3.3 (0.50)3.6 ± 1.7 (0.15)
Vertical phase speed (m s−1)
   Downward0.73 ± 0.42 (0.03)0.80 ± 0.53 (0.08)0.70 ± 0.39 (0.03)
   Upward0.74 ± 0.44 (0.03)0.88 ± 0.57 (0.09)0.65 ± 0.32 (0.03)
Period (min)
   Downward107 ± 28 (1.7)120 ± 28 (4.2)103 ± 27 (2.0)
   Upward101 ± 26 (1.8)113 ± 30 (4.5)98 ± 24 (2.1)

4. Gravity Wave Potential Energy Density and Its Seasonal Variations

[15] The mean potential energy density per unit mass equation image in the 2-D window of 3 h and 15 km is defined as:

equation image

where 〈r2pure = 〈(r(z, t))2〉 − equation image is the pure mean square (MS) relative density perturbation calculated for the 3-h and 30–45-km window with the positive bias due to photon noise, equation image, removed [Chu et al., 2009]. The gravitational acceleration is given by g, and N is the Brunt–Väisälä frequency as defined in [Lighthill, 1978]:

equation image

where cs is the speed of sound and Cp is the specific heat at constant pressure. We take the first half of equation (3) to calculate the N2 using the lidar-measured atmospheric number density: ρ0 is the background density of the 3-h window as described above, and 0 is the background density difference between two altitude levels (dz = 1000 m). The sound speed, cs, for an ideal gas is given by cs = equation image, where R is the universal gas constant, μ is the atmosphere molecular weight, R* is the gas constant of dry air (287.05 J kg−1 K−1), Cp (1005 J kg−1 K−1) and Cv (718 J kg−1 K−1) are the specific heat at constant pressure and constant volume respectively, and T is the atmospheric temperature taken from the United Kingdom Met Office (UKMO) stratospheric assimilated data. The N2 is calculated for each altitude and then (g/N)2 is averaged over the 30 to 45 km range to obtain the mean equation image. By scaling 〈r2pure with equation image, a mean potential energy density equation image is obtained for the 3-h and 30–45 km window using equation (2). Our approach described here is comparable to that of Wilson et al. [1991b] who used an averaged N2 to scale the integrated power spectral density of density perturbation in selected spectrum range to derive the mean potential energy density.

[16] RMS relative density perturbations (r)rms = equation image are calculated for each 3-h data window and the statistics of the results are given in Table 4. It is striking that Rothera RMS perturbations display a significant increase (over 2 times) from summer to winter while the South Pole RMS perturbations are nearly flat through the year. To illustrate the seasonal variations of RMS perturbations, we take the weekly means of these 3-h perturbations and then plot the weekly means against month in Figure 4a. Despite the large fluctuations, it is clear from Figure 4a that RMS perturbations are comparable between the South Pole and Rothera in summer but in winter Rothera RMS perturbations are much larger than those at the South Pole. The mean N2 in each 2-D window of 30–45 km and 3 h is computed using equation (3) and then the weekly means of N2 are plotted versus month in Figure 4b for both the South Pole and Rothera. From the RMS perturbations and N2 obtained above, the mean gravity wave potential energy density per unit mass for each 3-h and 15-km window (GW-equation image) is derived using equation (2) and the statistics of the results are summarized in Table 5. The potential energy density is similar between summer and winter at the South Pole while at Rothera, the potential energy density increases by about 6 times from summer to winter. The statistics also shows that the potential energy densities are comparable between the South Pole and Rothera in summer but in winter the Rothera potential energy density is about 4 times that of the South Pole. Weekly means of GW-equation image versus month are plotted in Figure 4c. The winter enhancement of GW-equation image is apparent at Rothera despite the large fluctuations.

Figure 4.

The seasonal variations of weekly mean (a) RMS relative density perturbation, (b) N2, and (c) GW-equation image for the South Pole (red/gray) and Rothera (blue/black). The solid curves are the annual plus semiannual harmonic fittings given by equation (4). Error bars are the uncertainty of RMS relative density perturbation in Figure 4a, the standard deviations of N2 in Figure 4b, and the uncertainty of GW-equation image in Figure 4c.

Table 4. Statistics of RMS Relative Density Perturbation at the South Pole and Rothera
 South Pole Mean ± Standard DeviationRothera Mean ± Standard Deviation
Entire Season0.43 ± 0.17 (%)0.52 ± 0.32 (%)
Winter (from May to Aug)0.45 ± 0.12 (%)0.95 ± 0.30 (%)
Summer (from Nov to Feb)0.42 ± 0.18 (%)0.37 ± 0.15 (%)
The ratio of winter to summer1.072.51
Table 5. Statistics of GW-equation image at the South Pole and Rothera
 South Pole Mean ± Standard DeviationRothera Mean ± Standard Deviation
Entire season2.70 ± 2.11 (J kg−1)4.21 ± 5.53 (J kg−1)
Winter (from May to Aug)2.73 ± 1.91 (J kg−1)10.97 ± 8.19 (J kg−1)
Summer (from Nov to Feb)2.71 ± 2.18 (J kg−1)1.80 ± 1.56 (J kg−1)
The ratio of winter to summer1.006.61

[17] To help quantify the seasonal variations for both locations, the harmonic fits of “annual + semiannual + mean” given by equation (4) are applied to the weekly means of RMS perturbation, N2, and GW-equation image shown in Figure 4:

equation image

where A12 and A6 are amplitudes of annual (12 months) and semiannual (6 months) oscillations, Φ12 and Φ6 are the corresponding phases, and A0 is the annual mean. The obtained fitting parameters are summarized in Table 6. The RMS perturbation and GW-equation image at the South Pole are nearly constant and do not exhibit significant seasonal variations, although N2 does show a weak seasonal dependence (>50% correlation). With balloon radiosonde data from 1993 to 1996, Pfenninger et al. [1999] reported a seasonal variation of gravity wave potential energy density in the 10 to 35 km range at the South Pole. They showed that potential energies vary between about 0.5 J kg−1 to 2.5 J kg−1 with a clear maximum in November from 11 to 35 km and a minimum in February from 11 to 21 km. The weak seasonal variations in the 30 to 45 km range, coupled with the large week-to-week variability, preclude our comparison to the observations by Pfenninger et al. [1999].

Table 6. Fitting Curve Coefficients for RMS Relative Density Perturbations, N2, and Gravity Wave Potential Energy Density at the South Pole and Rothera
 RMS (%)N2 (× 10−5 s−2)GW-equation image (J kg−1)
South PoleRotheraSouth PoleRotheraSouth PoleRothera
A120.04 ± 0.030.34 ± 0.033.54 ± 1.315.70 ± 0.970.54 ± 0.334.97 ± 0.83
Φ12 (day)278 ± 28197 ± 6200 ± 32180 ± 12310 ± 28196 ± 11
A60.05 ± 0.030.08 ± 0.035.13 ± 1.831.59 ± 1.080.26 ± 0.291.23 ± 0.89
Φ6 (day)94 ± 1598 ± 12272 ± 8195 ± 19122 ± 35101 ± 21
A00.47 ± 0.020.69 ± 0.0246.7 ± 1.249.6 ± 0.762.83 ± 0.226.51 ± 0.63
Correlation coefficient29.9%89.1%59.1%72.4%32.9%74.5%

[18] The annual oscillations dominate the seasonal variations of RMS perturbation and GW-equation image at Rothera, with maxima around the day of year 196 to 197. The semiannual oscillation amplitudes are about one fourth those of the annual oscillations. Therefore, Rothera exhibits a clear seasonal trend with the maximum in winter and the minimum in summer. The GW-equation image seasonal variations observed at Rothera are comparable to observations made by others in Antarctica. MLS measurements showed the wave variance to be maximum in winter and minimum in summer around Drake Passage (50°S to 70°S latitude and 290°E to 315°E longitude) at ∼28 km and ∼43 km [Wu and Jiang, 2002]. Baumgaertner and McDonald [2007] reported Antarctic GW-equation image seasonal variations from ground to 35 km using CHAMP/GPS occultation temperature data. They took an average of the GW-equation image over the entire Antarctic (60°S to 90°S) and showed a peak of GW-equation image in August between 30 and 35 km. This is comparable to our results at Rothera, but different from those at the South Pole. The results of Hertzog et al. [2008] show that the wave momentum fluxes are generally larger from September to October than during December to January, but they exhibit a decrease in momentum flux as one approaches the pole. Similarly, gravity wave observations made with a MF radar at Rothera revealed a peak wind variance in winter in the mesosphere [Hibbins et al., 2007]. Finally, lidar observations at Sondrestrom, Greenland (67°N, 310°E) exhibited the wave RMS perturbation in the 30 to 45 km range maximizing in winter and minimizing in summer [Gerrard et al., 2000], very similar to the Rothera trend.

5. Discussion

[19] We discuss the possible causes for the observed different seasonal variations of GW-equation image at Rothera and the South Pole. There is an obvious difference in topography between the two sites: Rothera is surrounded by ocean and mountains while the South Pole is in the middle of the Antarctic plateau. In addition, Rothera is closer to the winter storm track that passes through the Drake Passage. Therefore, Rothera is expected to have much stronger gravity wave sources than the South Pole. Baumgaertner and McDonald [2007] observed strong GW-equation image between 18 and 27 km around the Antarctic Peninsula (∼60°W) and the transantarctic mountains (∼160°E) in the yearly average. Similarly, Hertzog et al. [2007, 2008] and Ern et al. [2004] also observed enhanced gravity wave activity below 30 km over the Antarctic Peninsula in winter. Alexander and Teitelbaum [2007] reported the propagation of a mountain wave from the troposphere to the stratosphere (above 30 km) over the Antarctic Peninsula. Similar mountain-wave propagation events were observed in the Vorcore campaign and simulated in the Weather Research and Forecast (WRF) model [Plougonven et al., 2008]. Espy et al. [2006] observed a winter gravity wave momentum flux in the mesosphere that was four times larger at Rothera than at Halley station, which is well inland of the winter storm track and has a featureless topography like that of the South Pole. These results suggested that enhanced gravity wave sources could cause strong gravity wave activity near Rothera during the winter.

[20] Our results show the GW-equation image to be nearly equal during the summer at Rothera and the South Pole, but that it increases a factor of ∼6 at Rothera during the winter. This is consistent with enhanced wave sources playing a greater role at Rothera during the winter in accordance with earlier studies. The wind component perpendicular to the Antarctic Peninsula derived from UKMO at a height of ∼3 km over sea level, the approximate mountain height near Rothera, would allow us to account in both the wind amplitude and direction for the generation of orographic gravity waves. A correlation study between GW-equation image and the perpendicular wind component shows no significant correlation during the summer but a linear correlation coefficient of 65% (significant at the 99% level) during the winter at Rothera (Figure 5). This would indicate that orographic gravity waves are generated more during the winter at Rothera. However, while there may be a seasonal variation in the strength of orographically generated waves throughout the year, one must also consider the effects of critical level filtering on the observed GW-equation image.

Figure 5.

The correlation between daily mean background wind component perpendicular to the Antarctic Peninsula at ∼3 km and GW-equation image at Rothera. Error bars are the uncertainty of GW-equation image. Squares, circles, and asterisks represent data in winter (May–August), summer (November–February), and spring/fall, respectively.

[21] Gravity waves can be filtered out by the background wind when the gravity wave horizontal phase velocity is equal to the background wind velocity [Lindzen, 1981; Whiteway and Duck, 1996; Espy et al., 2004, 2006]. Seasonal variations of this critical level filtering effect can significantly influence the GW-equation image reaching the 30–45 km range. We examine the critical level filtering of both mountain waves and freely propagating waves. The critical level filtering effects on freely propagating gravity waves are examined with the wind-blocking diagram following Taylor et al. [1993]. We launch gravity waves from the mountain height with 0–40 m s−1 ground-based horizontal phase speed [Garcia and Solomon, 1985]. Assuming that the gravity wave propagation direction is homogeneously distributed from 0 to 360 degree and interpolating the UKMO wind data to fixed altitudes, we calculate the probability of upward propagating gravity waves reaching our observational range, i.e., the transmission of gravity waves through the lower atmosphere. The results of such probability versus altitude and month are illustrated in Figure 6b for Rothera. The mean probability of reaching 30–45 km (Figure 6a) is between 70% and 80% for summer months but around 55% during winter. This indicates that freely propagating gravity waves can contribute to our observed GW-equation image through a year but their chances to reach the observational heights are more in summer than in winter at Rothera.

Figure 6.

(a) The mean probability for freely propagating gravity waves reaching our observational heights between 30 and 45 km. (b) Seasonal variations of the probability for freely propagating gravity waves to reach each altitude at Rothera in 2003 and 2004, assuming that gravity wave propagation directions are distributed homogeneously from 0 to 360 degrees, and the phase speeds are 0–40 m s−1. A 15-day Hamming window is used to smooth the contour. (c) The mean probability similar to Figure 6a but for mountain waves with phase speeds of 0–1 m s−1. (d) Seasonal variations of weekly mean wind rotations from 3 to 30 km at Rothera. Weekly means are calculated with the UK Met Office data from December 2002 to March 2005. Error bars are ±1σ. The solid curve is the annual plus semiannual harmonic fit.

[22] With similar wind-blocking diagram we calculate the mean probability of mountain waves to reach the 30–45 km range. This is done by launching waves with phase speeds of 0–1 m s−1. The obtained results in Figure 6c indicate that mountains waves have higher probability to reach our observational height in winter (∼50%) than in summer (nearly zero). Following Baumgaertner and McDonald [2007], we derived the wind rotation using UKMO wind data, which is defined as the maximum wind direction change from the mountain height (∼3 km) to 30 km. The weekly averaged wind rotations obtained at Rothera are plotted in Figure 6d. Significant wind rotations (>90°) occur in summer with significantly smaller wind rotations in winter. As suggested by Whiteway and Duck [1996] and Baumgaertner and McDonald [2007], wind rotations >90° in the lower atmosphere imply critical level filtering of the mountain waves. Figure 6d indicates that mountain waves have very little chance of reaching our observational heights in summer at Rothera but much higher probabilities in winter, consistent with the wind-blocking diagram results. Hence, even a constant source of mountain waves would have a larger contribution to our observed GW-equation image in winter than in summer at Rothera. No clear seasonal variations of the critical level filtering are found in either the wind rotation or the blocking diagram for the South Pole. It is likely that neither the critical level filtering nor the wave source has significant seasonal variations at the South Pole, which is consistent with the nearly constant GW-equation image observed through a year.

[23] The seasonal variations of the critical level filtering effects at Rothera may help explain our observations. Although orographic sources at Rothera are much stronger than at the South Pole, the strong critical level filtering effects in summer remove most mountain waves but allow most freely propagating waves to reach the 30 to 45 km range. As the freely propagating waves are not significantly filtered at either Rothera or the South Pole, there is a comparable GW-equation image between the two sites in summer. However, in winter most mountain waves at Rothera can propagate upward to reach our observational heights, leading to much higher winter GW-equation image at Rothera than at the South Pole. This situation also enables us to observe the positive correlation between the surface perpendicular wind component and the GW-equation image in winter at Rothera, and the large fluctuations in winter GW-equation image. Although freely propagating waves are more heavily filtered during the winter at Rothera, one cannot exclude the possibility that sources of these waves increase at Rothera while remaining relatively constant at the South Pole, since the measurements of GW-equation image do not distinguish the two types of waves.

[24] The wintertime polar vortex may contribute to the GW-equation image enhancement in winter at Rothera. To examine the polar vortex effects, the polar vortex edge position relative to Rothera is calculated using National Centers for Environmental Prediction reanalysis data, and the results are shown in Figure 7 along with the GW-equation image. No statistically significant correlation exists between vortex position and GW-equation image at Rothera; however, all of the large GW-equation image (>10 J kg−1) events are observed between +5 and −10 degrees from the vortex edge, and no large GW-equation image is observed outside 10 degrees. This result is consistent with the previous CHAMP/GPS observations of the gravity wave enhancement within ±10 degrees from the vortex edge [Baumgaertner and McDonald, 2007]. Whiteway et al. [1997] suggested two reasons of gravity wave enhancement by polar vortex: a reduction of the critical level filtering effects and the Doppler shift of vertical wavelength. Gravity wave generation by the polar vortex through the geostrophic adjustment has also been suggested [Yoshiki et al., 2004; O'Sullivan and Dunkerton, 1995; Zhang, 2004].

Figure 7.

Correlations between the GW-equation image and the locations of polar vortex edge relative to Rothera. Zero degree means that polar vortex edge is at Rothera, negative degrees mean that Rothera is inside the vortex toward the pole, and positive degrees mean that Rothera is outside the vortex. Each data point of GW-equation image is daily mean, and each data point of polar vortex position is calculated from daily National Centers for Environmental Prediction reanalysis data.

[25] The spectral analysis results in section 3 show ∼44% waves having apparent upward phase progression at both the South Pole and Rothera. This percentage is much larger than the previous observations like those of Wilson et al. [1991a]. It is necessary to point out that if the background wind flowing over the mountains varies in time, it may cause the mountain waves appearing as upward (or downward) phase progression in the perturbation profiles at Rothera, although the waves in this case are propagating upwards. However, we observed similar percentage at the South Pole where the mountain waves are believed to be absent. In addition, Pfenninger et al. [1999] did not find correlations between tropospheric and stratospheric gravity wave spectra at the South Pole, and 20% of their observed waves in the 16 to 23 km region propagated downward. They suggested that there were wave sources above their observed altitudes, resulting in gravity wave spectrum varying with altitude. At Syowa (69°S, 39°E), Yoshiki et al. [2004] observed from radiosonde data that ∼43% waves in the range of 15–25 km propagated downward in winter while ∼12% downward propagating waves in summer. Considering all these results, we speculate that some of our observed waves with apparent upward phase progression might represent downward energy propagation. Since our observational range at Rothera is located near the edge of the polar vortex, this stratospheric jet stream may have contributed to the downward propagating waves, as suggested by Sato and Yoshiki [2008] for Syowa. Regarding the summer case at Rothera and for the South Pole case, other sources or waves reflected by the region of evanescence above our observational heights could be responsible. Such downward propagating waves and causes deserve further investigation but are beyond the scope of this work.

6. Conclusions

[26] We present the first observational study of stratospheric gravity wave spectra, potential energy density, and seasonal variations through an entire year in the 30 to 45 km region at Antarctica with an Fe Boltzmann/Rayleigh lidar data. Our data analysis approach reliably extracts gravity wave spectra with vertical wavelengths of 2–15 km and periods of 60–180 min. The obtained ground-relative gravity wave spectra are comparable between the South Pole and Rothera: the annual mean gravity wave vertical wavelength is ∼4.1 km, the vertical phase speed ∼0.7 m s−1, and the period ∼104 min. The observed ground-based gravity waves are dominated by waves with downward phase progression but ∼44% waves show upward phase progression at both sites. However, because of Doppler shift effects, these ground-relative upward phase speeds do not necessarily represent downward energy propagation. The observed GW-equation image in the 30 to 45 km region are comparable between Rothera and the South Pole in summer (∼1.8–2.7 J kg−1), but the winter GW-equation image at Rothera (∼11 J kg−1) is about 4 times larger than that at the South Pole. The observed Rothera GW-equation image shows a significant seasonal variation with the winter average being 6 times larger than the summer values. The seasonal variations of GW-equation image at the South Pole are much weaker than those at Rothera, with nearly constant values (∼2.7 J kg−1) through the year.

[27] Our investigations suggest that the critical level filtering of orographic gravity waves may have played a key role in the GW-equation image seasonal variations at Rothera. Even though the wave sources are expected to be much stronger at Rothera, the strong critical level filtering effect in summer removes most mountain waves, resulting in comparable GW-equation image between these two locations. The minimum critical level filtering in winter allows most orographic gravity waves to propagate upward and reach the 30–45 km range, leading to the winter wave enhancement at Rothera. The polar vortex may also contribute to the Rothera GW-equation image enhancement in winter. The absence of clear seasonal variations in wave sources and critical level filtering is likely responsible for the significantly small seasonal variations of GW-equation image observed at the South Pole. We speculate that the polar vortex may act as a source for downward propagating waves in the wintertime at Rothera, and waves reflected from the region of evanescence above our observational heights might introduce waves propagating downward in summer or at the South Pole. The downward propagating waves and the South Pole seasonal variations certainly deserve further investigation with more observational data and gravity wave modeling.

Acknowledgments

[28] We sincerely acknowledge Jan C. Diettrich and the staff of the British Antarctic Survey for their operation of the lidar at Rothera. We also acknowledge two winter-over scientists, Ashraf El Dakrouri and John Bird, and the staff of Amundsen-Scott South Pole Station for their help and support in collecting data at the South Pole. We appreciate the valuable discussion with Dave Fritts. Rothera lidar project was supported by the U.K. Natural Environmental Research Council and by the United States National Science Foundation. C.Y., X.C., and W.H. were partially supported by NSF grants ATM-0602334, ATM-0645584, ATM-0632501, and ATM-0545353. C.Y. sincerely acknowledges the generous support of NCAR Newkirk Graduate Fellowship. H.L.L's effort is in part supported by the Office of Naval Research (N00014-07-C0209). National Center for Atmospheric Research is sponsored by the National Science Foundation.

Ancillary