Characteristics of atmospheric gravity wave activity in the polar regions revealed by GPS radio occultation data with CHAMP

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113,, doi: /2007jd008938, 2008 Characteristics of atmospheric gravity wave activity in the polar regions revealed by GPS radio occultation data with CHAMP Hayato Hei, 1 Toshitaka Tsuda, 1 and Toshihiko Hirooka 2 Received 10 May 2007; revised 17 October 2007; accepted 4 December 2007; published 27 February [1] Using GPS radio occultation data during , we studied the climatological behavior of atmospheric gravity waves in the polar stratosphere. We calculated temperature fluctuations with vertical wavelengths shorter than 7 km and then determined the wave potential energy, E p, every month in a longitude-latitude cell of between 12 km and 33 km. In the Arctic region (50 90 N), E p shows an annual variation with maximum in winter, consistent with the zonal mean horizontal wind, V, and the Eliassen-Palm (E-P) flux, F z. The large F z values indicate higher planetary wave activity, resulting in distortion of the polar vortex. The unbalanced flow can then excite gravity waves through geostrophic adjustment. In the Antarctic region (50 90 S), E p gradually increases during winter and reaches its maximum in spring before decreasing rapidly. The time derivative of V coincides with the E p peak and the horizontal distribution of E p has a similar structure to V, suggesting that the E p enhancement is closely related to the decay of the polar vortex. During major warming events over the Arctic, the divergence of E-P flux, DF, was enhanced, coinciding with large E p. In the Antarctic, DF strongly correlates with E p in spring. Gravity waves seem to be effectively generated through planetary wave transience and/or breaking. Orographic generation of gravity waves seems to be important in limited areas only, such as Scandinavia and the Antarctic Peninsula, showing that it is less important than the polar night jet in determining the climatological behavior of gravity waves. Citation: Hei, H., T. Tsuda, and T. Hirooka (2008), Characteristics of atmospheric gravity wave activity in the polar regions revealed by GPS radio occultation data with CHAMP, J. Geophys. Res., 113,, doi: /2007jd Introduction 1 Research Institute for Sustainable Humanosphere, Kyoto University, Uji, Japan. 2 Department of Earth Planetary Sciences, Kyushu University, Fukuoka, Japan. Copyright 2008 by the American Geophysical Union /08/2007JD [2] Atmospheric gravity waves are generally excited in the lower atmosphere and propagate upward, transporting momentum and energy from the troposphere to the middle atmosphere. The gravity waves eventually dissipate because of a variety of unstable processes, such as wave-mean flow interactions, critical level interaction and convective/dynamical instabilities. When wave breaking occurs, wave energy and momentum are deposited into the background mean winds, producing a drag force. Although the individual gravity waves may appear small in the lower atmosphere, the integrated effects of wave drag force become large enough to accelerate or decelerate the mean winds (general circulation) in the middle atmosphere. [3] Theoretical, numerical and observational studies have advanced our understanding of gravity waves on many aspects, such as the wave generation mechanisms, the gravity wave spectrum in terms of frequency, vertical and horizontal wave numbers, wave-mean flow/wave-wave interaction and instability processes, and so on. Recent studies have also significantly expanded our understanding of gravity wave influences on the large-scale circulation and thermal and constituent structures of the middle atmosphere. Advanced study has also improved parameterizations of gravity wave effects which are enabling ever more realistic descriptions of gravity wave forcing in large-scale models [e.g., Fritts and Alexander, 2003]. Many studies have clarified the atmospheric gravity wave sources based on extensive theoretical, modeling and observational efforts, which includes meteorological events, such as a typhoon, severe storms etc., interaction of the surface winds with topography, cumulous convection in the tropics and geostrophic adjustment of unbalanced flows in the vicinity of jet streams and frontal and wave-wave interactions. Thus the behavior of gravity waves has greatly been clarified in the last decades. There remain, nevertheless, a number of areas in which further progress is needed in refining our understanding of and our ability to describe and predict gravity wave influences in the middle atmosphere. [4] There are a number of observations of small-scale temperature and horizontal wind structures caused by gravity waves. The detailed behavior of gravity waves has been observed with direct measurements (balloon-borne radiosondes, rocket soundings, etc.) and ground-based remote sensing (radars, lidars, etc). Radiosonde observations have provided information on short vertical wavelength gravity 1of15

2 waves in the lower stratosphere below about 30 km [e.g., Tsuda et al., 1994]. Using a global data set of meteorological rocket soundings, the seasonal and geographical variability of stratospheric gravity waves was studied [Hirota, 1984; Hirota and Niki, 1985]. Kitamura and Hirota [1989] examined the routine radiosonde data observed at northern midlatitudes and showed a latitudinal and seasonal variation in gravity wave energy. Allen and Vincent [1995] also showed latitudinal variations in gravity wave in the Southern Hemisphere. [5] Various atmospheric radars, such as MST, MF and meteor radars, etc, have been employed to observe gravity wave wind perturbations in the lower stratosphere, the mesosphere and lower thermosphere. These include longterm observations at single sites from which seasonal and interannual variations are determined. For example, the MU radar at midlatitude has revealed both detailed and climatological characteristics of the gravity waves. From the long-term observations with the MU radar, Murayama et al. [1994] clarified a clear annual cycle of the gravity wave activity in the lower stratosphere with maximum in winter and minimum in summer. Lidar observations also showed a seasonal cycle at midlatitudes [e.g., Whiteway and Carswell, 1995]. [6] Because global model studies increasingly recognize the need to describe gravity wave effects via parameterization, efforts to observe and understand the climatology of gravity waves are an area of active research. If we could describe the climatological variations in gravity wave occurrence along with the specific wave characteristics for a given set of meteorological conditions, the mechanisms for gravity wave generation would be understood and their sources fully characterized. However, it is not easy to observe a global morphology of gravity wave activity. Although the horizontal distribution of gravity waves was analyzed with routine radiosonde observations as described above, the study was limited mostly in the northern middle latitudes over continents. [7] There are not so many satellite observations of gravity waves except Wu and Waters [1996] and McLandress et al. [2000] with UARS/MLS, and Eckermann and Preusse [1999] with CRISTA, which studied a global morphology of the gravity wave activity by using moderate height resolution data. Recent GPS radio occultation (RO) measurements can provide temperature profiles in the troposphere and lower stratosphere with height resolution and measurement accuracy comparable to radiosondes. Therefore this data set is suitable for clarifying the global morphology of gravity waves. By using GPS RO data obtained by the GPS Meteorology (GPS/MET) project, Tsuda et al. [2000] examined the global distribution of wave energy and showed the correlation between wave energy and tropical convection. GPS RO provides a unique chance to study the climatology of gravity waves. In particular, the GPS RO data are very useful in the tropics and polar regions where routine radiosonde stations are sparse. [8] Using GPS RO temperature profiles obtained with the Challenging Minisatellite Payload (CHAMP), Ratnam et al. [2004] reported an enhancement of gravity wave energy during a vortex distortion due to the stratospheric sudden warming event in 2002 over the Antarctica. Baumgaertner and McDonald [2007] studied climatology of gravity waves over the Antarctica, and described the dependence of the wave energy on season, topography and altitude. In particular, the localization of wave activity over the Antarctic peninsula and trans Antarctic mountains was presented, which implied the importance of orographic effects in generating gravity waves. [9] In our study, we also employ the CHAMP GPS RO data, collected in , and investigate climatology of gravity wave energy in the high-latitude regions (higher than 50 ) in both the Northern and Southern Hemispheres. 2. Data Analysis 2.1. Database [10] The German CHAMP satellite was launched with a Russian COSMOS rocket on 15 July 2000 into an almost circular, near-polar (inclination = 87 ) orbit with an initial altitude of 454 km. By employing the GPS RO technique, CHAMP provides temperature profiles in the troposphere and lower stratosphere. In this study we use level-3 version 005 GPS RO data from May 2001 to December 2005, which are processed at Geo Forschungs Zentrum (GFZ) Potsdam [Wickert et al., 2001]. The vertical resolution of the temperature profiles ranges from 0.5 km in the lower troposphere to 1.4 km in the stratosphere and the horizontal resolution along the path is about a few hundred km. GFZ provides about globally distributed vertical profile of temperature data every month over the altitude range of km. [11] As well as the temperature profiles observed by CHAMP GPS RO measurements, we used mean wind, surface wind (10 m above the surface), geopotential height and temperature fields based on numerical weather prediction model in European Centre for Medium-Range Weather Forecasts (ECMWF). From May 2001 to August 2002, the 40 Year reanalysis (ERA-40) provides the meteorological parameters four times daily, and we also refer to the monthly mean data. After September 2002, we used the objective analysis data. For these data sets all variables are determined at the gridded coordinates every in longitude and latitude, and in the height range of hpa. ERA-40 and the objective analysis data have 23 and 21 pressure levels, respectively. [12] In addition, in order to examine the relationship between gravity wave and planetary wave activities we calculated the Eliassen-Palm (E-P) flux [e.g., Andrews et al., 1987] from the four dimensional global objective analysis data with grid spacing at 23 levels from 1000 to 0.4 hpa, produced by the Japan Meteorological Agency (JMA). Here, the E-P flux is considered to be a measure of the planetary wave activity. For investigation of the global topography, we refer to the elevation data set analyzed by NOAA with the grid resolution of in longitude and latitude, corresponding to 60 m 60 m Gravity Wave Energy, E p [13] The gravity wave energy E is chosen in this study as the measure of wave activity, and it is defined as follows: E ¼ 1=2 u 02 þ v 02 þ w 02 þ 1=2ðg=NÞ 2 ðt 0 =T 0 Þ 2 ¼ E k þ E p 2of15

3 where u 0, v 0, w 0 are the perturbation components of the zonal, meridional and vertical wind velocity, respectively, and T 0 is temperature fluctuation. T 0 is the background temperature, N is the Brunt-Väisälä frequency and g is the gravitational acceleration. Note that E can be separated into the kinetic (E k ) and potential energy (E p ) per unit mass, respectively. By assuming a linear theory of gravity waves, u 0, v 0, w 0 and T 0 are all coupled to each other through polarization relations, and the ratio E k /E p is predicted to be a constant, ranging from 5/3 to 2.0 [e.g., VanZandt, 1985]. Thus gravity wave activity can be investigated from temperature observations only. In this study, we analyze E p from the GPS RO temperature profiles Analysis of E p With GPS RO Data [14] In this section we describe the data analysis procedure of the gravity wave potential energy E p. Because we are interested in the climatological characteristics of gravity wave activity, we have analyzed monthly mean E p from May 2001 to December Then we need to determine the area (cell) which can normally include the statistically significant number of data points. Considering the data rate of the CHAMP GPS RO results, we have selected the cell size for our analysis as in longitude and latitude. [15] We have counted all available GPS RO data with CHAMP from May 2001 to December The data number in one month steadily increased after satellite launch, and reached about 4000/month in March Then, the number is almost constant at from March 2002 to December 2005, with a mean of As we have separated the globe into 324 cells, we expect more than 10 GPS RO data in each cell, provided the data distribution is homogeneous. However, because of the high inclination angle of the CHAMP satellite orbit, we have more data at higher latitudes. The number of data in one month is as small as 150 over the equator, but, it ranges from 200 to 300 at higher latitudes (50 80 N and S). However, the data rate decreases to about 100 in the highest latitudes (80 N/S to the pole). Note that the data rate per unit area greatly increases at higher latitudes, which is beneficial for our study in the polar regions. As the longitude distribution of the data number is nearly homogeneous, on average we have data points in each cell at 50 80, except at where there are about 5 data points. [16] Next, we check the mean tropopause height. Sharp temperature variations near the tropopause could contaminate the analysis of E p, because we are unable to clearly separate the temperature variation due to the tropopause structure and the perturbations caused by waves. The tropopause is located at around 8 10 km at N and S, and hemispheric differences do not appear to be significant. So, we set the height range for the E p analysis above 12 km in order to avoid the effects of the tropopause structure. [17] In calculating E p we need to separate temperature profiles into T 0 and T 0. We obtained all available GPS RO data in an individual cell in one month. The mean temperature profiles are calculated in each cell, then they are lowpass filtered by using a 4-km vertical running mean, which is defined as T 0. We obtained T 0 by subtracting T 0 from individual GPS RO profiles. We then applied the Fast Fourier Transformation (FFT) at km, and extracted T 0 with vertical wavelengths shorter than 7 km. [18] We describe here the analysis procedure of the vertical wave number spectrum with FFT. First, we removed the mean and a linear trend along altitude by using a linear least squares method. When there are missing values, we linearly interpolate the data. In order to suppress the effects of the edge of the height range, a windowing function (a half cosine function) is applied for the data at a height near the bottom (12 km) and top (33 km) of the entire height range. Practically, a cosine function is multiplied with the data at the km and km altitude, whose height width corresponds to 1/10 of the entire height range (21 km). Then, FFT was applied along altitude. [19] After removing the wave number components of 0, 1 (21 km in wavelength) and 2 (10.5 km), the inverse FFT was applied backward to reproduce T 0 having wave numbers larger than 3 (shorter than 7 km in wavelength). We obtained all E p data in the cell with in longitude and latitude, and the monthly mean E p are calculated at km, km and km. [20] Because the GPS RO temperature profiles are retrieved from the active limb sounding by assuming spherical symmetry, it smears out details of the horizontal structure within a sample volume (about 300 km along the GPS radio raypath) [e.g., Tsuda and Hocke, 2002]. Therefore the GPS RO results cannot resolve gravity waves with short horizontal wavelengths. 3. Characteristics of E p in the Arctic Region 3.1. Horizontal Distribution of E p [21] Figure 1 shows the distribution of the monthly mean E p in 2002 averaged at km in the Northern Hemisphere. In this study we focus on the behavior of E p in the polar region (higher than 50 N), but it may be useful to describe the general latitude distribution of E p here. Because of the active convection in the tropics E p appears generally large at low latitudes [Tsuda et al., 2000]. It is also noteworthy that the tropical tropopause is located at about km at low latitudes; therefore the sharp temperature structure around the tropopause could contaminate E p in our analysis. So, we do not show in Figure 1 the results of E p at latitudes lower than 35 N in order to focus on the E p distribution in the polar region. [22] In the Arctic region, E p is enhanced from November to March, with the largest E p values are seen in winter months (December February). We can recognize in Figure 1 localization of E p, such that the large E p values exceeding 4 are seen around west Eurasia (0 40 E, N) in January and December, middle Eurasia ( E, N) in January and Greenland (20 60 W) in January. On the other hand, small E p value is seen around western North America (50 80 N, W), over the Pacific Ocean ( E, N) and over the North Pole (80 90 N) Year-to-Year Variations of E p in [23] Figure 2 shows the monthly variations of the zonal mean E p at N at km, km and km. The mean E p in winter and summer are estimated in each season, and these values together with the winter/ 3of15

4 Figure 1. Distribution of monthly mean E p from the CHAMP RO data averaged at km in the Arctic region in summer ratio of E p are summarized in Table 1. In all of the three height ranges the monthly variations of E p are similar, having maximum in winter and minimum in summer. Note that in December 2001 and December 2003 E p is enhanced at all heights. At km and km, the winter peak of E p was narrower than that at km. [24] At km, the mean E p is the largest (2.78 ), which is about twice as large as the values at km and km ( ). The mean E p in winter is the largest at km (4.56 ), and the smallest at km (2.15 ). The mean E p in summer is large (1.48 ) at km, while at km and km E p was smaller and comparable (0.79 ). The ratio of mean E p in winter to that in summer is large and comparable at km and km ( ) and smallest at km (2.72). This result shows that the contrast of E p between summer and 4of15

5 Figure 2. Monthly variation of zonal mean E p at N in three altitude regions: km, km, and km. winter is large at km and km and is smaller at km. [25] Figure 3 shows the monthly variation of the zonal mean E p at km in the four latitude regions: N, N, N and N. We use E p in the entire height region of the analysis, i.e., km, because the time variation of E p is similar in all height regions. The annual mean E p value and the E p averaged in winter and summer months are shown in Table 2. The ratio of mean E p between summer and winter is also shown in Table 2. [26] In all latitude regions, E p shows a similar annual cycle with a maximum in winter. Note that gravity waves are enhanced largely in December 2001 at N, N, and N. The peaks of E p appear differently, for example, between winter in 2001/2002 and 2004/2005 such that a single intense peak existed in the former period, while a broad enhancement with three E p peaks is seen in the latter. The peak of E p mostly coincided between the different latitude regions. However, a time lag sometimes occurred. For example, a large peak was simultaneously recognized in December 2001 at N, N and Figure 3. Monthly variation of zonal mean E p at km in four latitude regions:50 60 N, N, N, and N N, but the peak occurred in January 2002 at N. Another example is the peak in March 2003 at N, which did not coincide with the peak at other latitude regions. [27] The 5 year mean E p values are similar at N, N and N ranging from 1.85 to 1.90, but it is smaller (1.39 ) at N. The mean E p in winter months is larger at N and N ( ), slightly smaller (2.89 ) at N, and the smallest at N (2.32 ). The mean E p in summer months shows a tendency for the value to gradually increase from 0.74 at N to 1.17 at N. The ratio of mean E p between winter and summer is the largest at N (3.37) and smallest N (2.47 ). This result shows that the contrast of the mean E p between Table 1. Zonal Mean E p at N in the Arctic Region a Altitude Range Annual Mean E p, E p in Winter, E p in Summer, Winter/Summer Ratio km km km a Annual mean E p, mean E p in winter (DJF), mean E p in summer (JJA) and the ratio of mean E p in winter to that in summer. Table 2. Zonal Mean E p at km in the Arctic Region a Latitude Range Annual Mean E p, E p in Winter, E p in Summer, Winter/Summer Ratio N N N N a Annual mean E p, mean E p in winter (DJF), mean E p in summer (JJA) and the ratio of mean E p in winter to that in summer. 5of15

6 Figure 4. Time variation of zonal mean E p at km and N (first panel), V averaged zonally at hpa in altitude and N in latitude (second panel), A p at 10 hpa and N (third panel), A 0 p at 10 hpa and N (fourth panel), F z at N and 10 hpa (fifth panel), absolute value of DF at N and 10 hpa (sixth panel), and mean V s averaged zonally and at N (seventh panel). Asterisks show the mean value of each variable in winter (DJF). winter and summer is the largest at N and smallest at N Comparison of E p With Mean Horizontal Wind, Planetary Wave Activity, and Surface Wind [28] Time variations of zonal mean E p are compared with mean horizontal wind amplitude, the planetary wave amplitude, the planetary wave amplitude fluctuation, the vertical component of E-P flux, the divergence of E-P flux and the mean horizontal surface wind. E p is averaged over km at N. We have calculated the zonal mean horizontal winds, V, at N in the height range between 250 and 7 hpa. The planetary wave amplitude is estimated daily by using the geopotential height, f. We have analyzed the zonal variations of f by means of FFT, and extracted the planetary wave amplitude corresponding to wave numbers 1 and 2, then we defined the daily planetary wave intensity, A p as the root mean square of the two components as follows; A p ¼ PW 2 1 þ PW 2 2 where PW 1 and PW 2 are the planetary wave amplitudes corresponding to the wave number 1 and 2. Then, we have further smoothed A p using a 7-day running mean, and defined this as ha p i. We have calculated A 0 p = A p ha p i, and regarded it as the magnitude of the short-term fluctuating components of planetary waves. We also refer to the vertical component of E-P flux F z, determined from the four dimensional global objective analysis data produced by JMA. The divergence of E-P flux F, defined as rf, is also analyzed. Here, we plot absolute values of rf/r(df), where r is density. F z is a measure of planetary wave activity, whereas DF is that of rapid changes of planetary wave activity due to planetary wave transience and/or breaking. The wave transience is produced by wave growth and/or damping in the course of planetary wave propagation. The planetary wave breaking intermittently occurs near an equatorial critical surface where the phase velocity of the wave coincides with the zonal wind speed, i.e., the surf zone [McIntyre and Palmer, 1983]. The horizontal wind amplitude 10 m above the surface, V s is averaged zonally at N. [29] Figure 4 shows the time variations of E p together with the variations of other variables. Asterisks in Figure 4 show the mean values of individual parameters in winter months. We also analyzed the cross correlation function between E p and other reference variables and investigated the time lag between the parameters (results not shown here). [30] E p shows a clear annual cycle, with a broad peak from November to March. Annual variations of V and A p correlate well with E p with coefficients of 0.83 and 0.90, 0 and a small time lag. A p correlates with E p (correlation coefficient = 0.77). F z correlated well with E p without time lag (correlation coefficient = 0.86). Some peaks of DF match with those of E p (correlation coefficient = 0.68). Time variations of V s and E p are similar (correlation coefficient = 0.79). [31] The mean values in winter are estimated for all variables. In Table 3 we summarize these variables in each year that are normalized relative to the values in 2001, which can be used to test year-to-year variability. In Table 3, we can recognize that E p is large in winter in , while in other seasons E p are comparable. V, A p and A 0 p have larger values in and in winter than those in F z has large value in 2001 and in other year the F z values are comparable. V s has no clear differences between years. Year-to-year variation of F z is consistent with E p. [32] The cross correlation coefficient between E p and V is as high as 0.79, which might suggest gravity wave gener- Table 3. Ratio of the Mean Value of the Parameters in the Arctic Normalized by Mean Value in 2001 Parameter E p V A p A p F z DF V s of15

7 Figure 5. (left) Five year mean distribution of E p from 2001 to 2005 in December, January and February and (right) the topographical distribution in the Arctic region. ation by topography. However, the cross correlation function was calculated for the zonal mean variables, which does not certify the relative locations of these disturbances. [33] In order to examine the possibility of gravity wave generation by topography (orographic effects) in the polar region (>50 N), we check the horizontal distribution of E p, and compare it with topography. Figure 5 shows the 5 year averaged distribution of E p in winter months during and the topography distribution. In Figure 5, large E p values are not recognized over most of the large mountainous regions, such as the Rocky Mountains. Some enhancement of E p is seen around Europe (Scandianvia). However, over the flat area like central Eurasia E p is also large. So this result suggests that gravity waves are not mainly generated by topography at high latitudes (>50 N). [34] Radiosonde results at the Northern Hemisphere indicated that E p became larger over Scandinavia in winter, and this enhancement was explained as the effect by local topography [Yoshiki and Sato, 2000]. This result is contrary to our results. However, radiosonde observation sites are limited around Europe (Scandinavia) and Greenland only. Our study covering the entire Northern Hemisphere has suggested that the topography does not seem to be the largest generation source of gravity waves in the Arctic region (>50 N). [35] Figure 6 shows the distribution of E p at km, 8 at 50 7 hpa and V in January 2004 in the Arctic region. The gravity wave enhancement occurred in central Europe (0 20 E and 50 N) and middle Eurasia at E and 60 N. In winter the structure of E p is similar to that of V, that is, E p is enhanced along an ellipse over Eurasia at 50 N and E. This distribution is similar to V and 8. [36] On the basis of our analysis, we investigate below three possibilities for the generation mechanism of gravity waves: [37] 1. The first possibility is planetary wave activity and geostrophic adjustment. Near the polar night jet, planetary waves could become active; then, the amplification of planetary waves distorts the polar vortex. Moreover, rapid changes of planetary wave activity due to planetary wave transience and/or breaking lead to rapid changes of such distortion. In general, strong planetary wave activity would be accompanied with its rapid change, so that both the mechanisms could not be absolutely separated each other. Under such conditions, unbalanced flow of the polar vortex would be brought about, which can excite gravity waves through geostrophic adjustment. The annual cycle of F z is consistent with that of E p. The good correlation between E p and DF also suggests that planetary wave transience and/or breaking are effective in generating gravity waves. Figure 6. Distribution of (left) E p at km in January 2004 in the Arctic region, (middle) 8 at 50 7 hpa in January 2004 and (right) V at 50 7 hpa. 7of15

8 Figure 7. Distribution of monthly mean E p from the CHAMP RO data averaged at km in Antarctica in [38] 2. The second possibility is orographic generation of gravity waves. Planetary waves are generally generated by topography and the thermal contrast difference between sea and land. Because a good correlation is seen between E p and A p, we consider the possibility that gravity waves are generated simultaneously with planetary waves by the same excitation source, i.e., topography. However, the horizontal scales are considerably different between planetary waves and gravity waves; therefore the effectiveness of topography in exciting these waves may also be different. Moreover, we do not find that gravity waves are localized over mountain ranges. It is unlikely, therefore, that gravity waves and planetary waves are generated simultaneously. [39] 3. The third possibility is wave-mean flow interaction. Gravity waves are generated in the troposphere and the wave propagation is controlled by the mean horizontal 8of15

9 Figure 8. Monthly variation of zonal mean E p at S in three altitude regions: km, km and km. winds. Because we cannot determine the phase velocity of individual gravity wave events, it is hard to examine the critical level interaction of the gravity waves with the background winds. Provided the gravity waves are generated in a random manner, such that the phase velocity and propagation direction are uniformly distributed, the interannual variations of E p could appear similar to that of the background wind. [40] Our study has clarified that the gravity waves in the Arctic region (50 90 N) in winter are most likely to be generated by geostrophic adjustment in association with planetary wave transience and/or breaking under the strong activity of planetary waves. However, we cannot clearly deny a possibility that the wave-mean flow interaction is responsible for the annual and interannual variations of E p. 4. Characteristics of E p in the Antarctic Region 4.1. Horizontal Distribution of E p [41] Figure 7 shows the distribution of the monthly mean E p in 2003 averaged at km in the Antarctic region. Similar to the Arctic region, we focus on the E p characteristics at high latitudes (50 90 S), so the results at low latitudes (<35 S) are not illustrated. In the polar region in the Southern Hemisphere, E p started to increase in June 2003, became considerably large in August and September, reached the maximum in October, then it suddenly decreased in November. [42] The horizontal distribution of E p showed an oval shape in August October, and the large E p values were localized around the coast line of the Antarctic Continent and over the ocean outside the continent. Large E p values exceeding 4 were seen east of the Antarctic Peninsula in September, and around the Drake Passage between the Antarctic Peninsula and the Southern American Continent (50 70 S, W) in November. Some enhanced E p were also seen over the Amery Ice shelf ( E, S) in September October. We discuss in a later section the possibility of orographic generation of gravity waves Year-to-Year Variations of E p in [43] Figure 8 is the monthly variation of the zonal mean E p at S at km, km and km. The time variations of E p are similar at km and km showing a clear annual variation. The annual variation is characterized by a sharp peak in spring (September October), associated with a very rapid decrease in November. E p is also enhanced from July to September, which is in general consistent with the winter enhancement of E p seen in the Arctic region. It is noteworthy that such tendency in the annual cycle does not appear clearly at km, that is, the spring enhancement of E p is not recognized obviously, and the magnitude of the annual variation is also small. [44] So as to compare the range of the annual variation of E p, we estimated the mean E p in summer (December February) and spring (September and October). The annual mean E p, average of E p in summer and spring are summarized in Table 4. At km the mean E p is the largest (3.06 ), and it is smaller at km and km ( ). E p in spring is large (4.64 ) at km and is small (2.12 ) at km. E p in summer is large (1.80 ) at km and is small ( ) at km and km. [45] Figure 9 shows the variation of the monthly zonal mean E p in four latitude regions: S, S, S and S. We can generally recognize annual variations in all of the four latitude regions. However, some differences can be recognized between S and S. At S, E p gradually increases in July September and reaches the maximum in October before decreasing rapidly in November. (Note that the variation in 2002 is unusual, i.e., the E p peak appeared in September 2002, probably because of the effects of the major sudden warming [Baldwin et al., 2003; Ratnam et al., 2004].) However, the sharp spring peak of E p cannot clearly be recognized at S, which suggest that the spring enhancement is a peculiar phenomenon that occurs at latitudes higher than 60 S. [46] The E p peak in spring simultaneously occurred in September or October in the three latitude bands in S. However, the year-to-year variations of the E p amplitudes are not coherent between the three latitude bands. Table 4. Zonal Mean E p at S in the Antarctic Region a Altitude Range Annual Mean E p, E p in Spring, E p in Summer, Ratio km km km a Annual mean E p, mean E p in spring (September and October) and E p in summer (DJF). 9of15

10 Figure 9. Monthly variation of zonal mean E p at km in four latitudinal regions: S, S, S and S. [47] The mean E p in summer (DJF) and the maximum value in spring (September or October) are summarized in Table 5. The annual mean E p is larger at S (2.23 ) and smaller (1.63 ) at S. E p in spring is the largest at S ( ) compared to that at S and S ( ). E p in summer is larger at S (1.54 ) than at S ( ) Comparison of E p With Mean Horizontal Wind, Planetary Wave Activity, and Surface Wind [48] Similar to the analysis in the Northern Hemisphere, time variations of zonal mean E p are compared with V, A p, A p 0, F z, DF and V s in the Antarctic region. The zonal mean E p is averaged over km at S. V is averaged zonally in the height range from 250 to 7 hpa at S. Table 5. Zonal Mean E p at km in the Antarctic a Latitude Range Annual Mean E p, E p in Spring, E p in Summer, Ratio S S S S a Annual mean E p, mean E p in spring (September October) and mean E p in summer (DJF). Figure 10. Time variation of zonal mean E p zonally averaged at km and S (first panel); V averaged zonally at hpa in altitude and S in latitude (second panel); dv/dt, the time derivative of V at hpa and S (third panel); A p at 10 hpa and S (fourth panel); A p 0 at 10 hpa and S (fifth panel); F z averaged zonally at S and 10 hpa (sixth panel); absolute value of DF averaged zonally at S and 10 hpa (seventh panel); and mean V s averaged zonally and at S (eighth panel). Asterisks show the mean value of each variable in winter (July September). Note that we also include dv/dt, the time derivative of V 0 with respect to month. A p and A p is also calculated by using 8. F z and DF are averaged at S at the height of 10 hpa. V s is averaged zonally at S. Figure 10 shows the time variations of E p together with the variation of the other parameters. Asterisks in Figure 10 show the mean values between July and September. Considering that we have defined December February as the winter months in the Northern Hemisphere, we could select June August as winter in the Southern Hemisphere by shifting six months. However, the largest V values appear in August; therefore we here define July, August and September as the winter-like season in Antarctica. Again, we analyzed a 10 of 15

11 Figure 11. Distribution of E p at (top) km, (middle) 8 at hpa, and (bottom) V at hpa. Distribution in (left) September in 2003 and (right) October cross correlation function between E p and other reference variables. [49] E p shows a clear annual cycle which is, however, different from that in the Northern Hemisphere. That is, E p gradually increases between July and September and reaches the maximum in spring (September October) before decreasing rapidly. V shows an annual cycle with the peak mostly in August but the E p peak occurs about one or two months later mostly in October, which is clarified by the lag of cross correlation function between E p and V. This time 0 lag also appears for A p, A p and F z. The E p enhancement occurs during the rapid decrease of the polar night jet in spring, which can be clearly recognized by a good correlation between E p and dv/dt in Figure 10. [50] Figure 11 shows the distribution of E p at km, 8 and V at hpa in September 2003 and October 2003 in the Antarctic region. In September the map of both V and 8 indicate that a strong polar vortex existed along a circle at 60 S and E. The horizontal distribution of E p shows a very similar structure with V, although the longitude range is limited to E. Comparison of V between September and October indicates that the polar vortex rapidly decayed in one month. In October, the E p distribution is again similar to the V and 8 structure, and the E p amplitudes became much larger than those in September. This result suggests that gravity wave generation is enhanced during the decay phase of the polar vortex, and the wave horizontal distribution is closely related to the vortex structure. [51] It is noteworthy that the large DF in Figure 10 is seen simultaneously with the E p enhancement, Like in the Northern Hemisphere, we can assume that the planetary wave activity generates gravity waves via geostrophic adjustment related to planetary wave transience and/or breaking. As a matter fact, when V, A p, A 0 p and F z are large in winter months, E p are enhanced moderately. However, over Antarctica, the largest DF sometimes occurred one month later than the peak of F z, although they mostly coincided in the Arctic region. Therefore DF is responsible for additional excitation of gravity waves in early spring. [52] We show the mean values between July and September for some of the selected parameters (E p, V, F z and DF) and the values corresponding to the E p peak in September or October in Tables 6 and 7, respectively. Note that these values are normalized relative to the values in The maximum value of E p in spring does not vary largely ( ) in in Figure 10. DF shows a large value in 2003, but except for that the variation of the DF is small. Our study suggests that DF is the key parameter to explain the E p enhancement in spring, as their narrow peaks coincides, and there exists a reasonable quantitative consistency. Then, we need to explain the broader enhancement of E p in winter. [53] We have reported in the previous section a good correlation between E p and F z during winter months in the Northern Hemisphere, where they not only coincide but also show a quantitative consistency. Therefore we investigate here a similar relation in the Southern Hemisphere. In Table 6 we show the mean value of E p and F z in July August Table 6. Ratio of Mean Value of the Parameters (E p, V, F z and DF) in Winter in the Antarctic Normalized by the Mean Value in 2002 Parameter E p V F z DF of 15

12 Table 7. Ratio of the Mean Value of the Parameters (E p, V, F z and DF) in Spring in the Antarctic Normalized by the Mean Value in 2002 Parameter E p V F z DF September. In the E p value does not vary largely, showing the range of the variability of about 10% but it became somewhat smaller in In 2002, F z was considerably smaller than that in other years. This comparison suggests a general consistency for the relation between E p and F z in wither months. (Note, however, that in 2002 the large E p that appeared in September were associated with a sudden warming event [Ratnam et al., 2004].) The cross correlation function between E p and F z shows a time lag of one month, because the annual cycle of E p is largely affected by the spring peak. [54] As is well known, in the Southern Hemisphere, the polar night jet during midwinter is significantly stronger than that in the Northern Hemisphere. Resultant strong westerlies would suppress planetary wave activity in the stratosphere. In late winter, westerlies become moderate allowing more active planetary wave conditions, which leads to the excitation of gravity waves through geostrophic adjustment. However, the largest DF in the Southern Hemisphere occurs one month later than the peak of F z, although they coincide in the Northern Hemisphere. This may be understood by taking account of planetary wave activity in early spring (September October). The largest DF occurs mostly in October when the polar vortex usually breaks down. The coincidence of the time derivative of V with the peak of E p also suggests that the E p enhancement is related to the decay of the polar vortex. In that case, planetary wave breaking might easily occur because of the presence of critical surfaces in the stratosphere, to make DF large despite rather weak planetary wave activity. Consequently, the gravity wave excitation in winter months seems to be related to the planetary wave activity in a similar way to the Arctic case. [55] Similar to our investigation for the Arctic results, we compare the horizontal distribution of E p with topography in order to examine the orographic effects. Figure 12 shows the 5 year mean ( ) distribution of E p in August at km and km and the topographical distribution in the Antarctic region. As Baumgaertner and McDonald [2007] already reported, a localized enhancement of E p is recognized around the Antarctic Peninsula at km. Note also that significant E p values are seen around the Andean mountain in South America. Therefore orographic effects are obviously one of the important mechanisms for generating gravity waves in the polar regions. However, larger E p values are mostly recognized over the ocean. When E p is averaged over the wide latitude and longitude ranges in the Antarctica, the statistical behavior of the wave energy may not be fully explained by the orographic effects. It is most likely explained by considering geostrophic adjustment, which is discussed in more detail in the next subsection Comparison of the E p Characteristics Between Arctic and Antarctic Regions [56] We here discuss characteristics of E p from the Antarctic region, investigating the similarity and differences with the results from the Arctic region. Figure 13 summarizes the height variations of the annual mean E p, and the E p values in winter (DJF) and summer (JJA) in the Northern Hemisphere (50 80 N), which are taken from Table 1. The corresponding E p values in the Southern Hemisphere (50 80 S) in Table 4 are also plotted in Figure 13. Note that the E p values in spring are used instead of winter results. The height structure of the annual mean E p is in general quite similar between the Arctic and Antarctic regions, but the E p values were about 10 25% larger in the Antarctic. The summer results show a similar tendency. The E p values in winter and spring in the Arctic and the Antarctic, respectively, agree well at km and km, although the E p was largely enhanced at km in the Antarctic. The integrated amount of gravity wave energy is larger over Figure 12. Five year averaged distribution of E p at (left) km and (middle) at km and (right) topographical distribution in the Antarctic region. 12 of 15

13 Figure 13. Height variations of the annual mean E p (diamond), E p in winter (DJF) (triangle) and summer (JJA) (square), respectively, in the Arctic region (solid symbol with a solid line). Annual mean, spring (SO) and summer (DJF) values of E p are similarly plotted for the Antarctic region (open symbol with a dashed line). the Antarctic but the peak E p values seem similar between the Arctic and Antarctic regions. [57] Figure 14 shows the latitude dependence of the zonal mean E p at km in the Arctic and Antarctic regions which are taken from Tables 2 and 5, respectively. In both hemispheres, the annual mean E p exhibits a similar latitude variation such that the values are nearly the same between 50 to 80, and decrease at by a factor of about 3/4. The E p values are persistently larger in the Southern Hemisphere by about 15%. [58] Similarities in the height and latitude variations between the Northern and Southern Hemisphere shown in Figures 13 and 14 suggest that a common mechanism seems to be dominant in generating gravity waves, although the wave generation is more active in the Southern Hemisphere. The latitude distribution of E p does not seem to reflect topography which is considerably different between the hemispheres. [59] Figure 15 shows time variations of E p averaged at km and in each hemisphere (reproduced from Figures 4 and 10) in comparison with DF at 10 hpa averaged over In the Northern Hemisphere the stratospheric major warming is quite common [e.g., Manney et al., 2005]. By referring to the NCEP analysis the warming events were reported to occur in December 2001, February 2002, January 2003, December 2003, January 2004 and January 2005 (M. V. Ratnam, personal communication, 2007), which coincided well with the large positive and negative enhancement in DF in Figure 15. Most of the large E p values are associated with these DF peaks. The correlation between E p and DF is more conspicuous in the Southern Hemisphere, that is, all of the E p enhancement in October in occurred when DF became sharply enhanced. It is noteworthy that in 2002 a major warming occurred in September [e.g., Baldwin et al., 2003], and both E p and DF in Figure 15 simultaneously became large. [60] Regarding the horizontal distribution of gravity waves, we do not find a meaningful correlation with the topography. Planetary waves seem to be active in the vicinity of the polar night jet, so, provided the planetary wave activity excites gravity waves, E p enhancements can be localized along the polar night jet. Our study has confirmed the similar distribution of E p and the polar night jet. [61] Orographic excitation of gravity waves is obviously an important mechanism, but, in a statistical sense it may not be a major generation source of gravity waves at high latitudes (higher than 50 ). It seems most likely that geostrophic adjustment is responsible for producing gravity waves in both the Arctic and Antarctic regions. We should note, however, our analysis does not correctly estimate the energy for gravity waves with short horizontal wavelengths, Therefore we should not exclude the possibility that orographic effects could be effective in exciting different parts of the gravity wave spectra. 5. Concluding Remarks [62] We studied the climatological behavior of the atmospheric gravity waves in the stratosphere in both the Arctic and Antarctic regions by using GPS RO data. We used level-3 version 005 GPS RO data of the CHAMP satellite from May 2001 to December From the GPS RO temperature profiles, we calculated the background mean temperature, T 0, the Brunt-Väisälä frequency squared, N 2 and the temperature fluctuations, T 0, with vertical wavelengths shorter than 7 km. Then we determined the wave potential energy E p at km, km and km. E p values were averaged every one month in a cell in longitude and latitude. The GPS RO measurements are sensitive to gravity wave components with relatively long horizontal wavelengths because of the limb sounding configuration. We also refer to ERA-40 and objective Figure 14. Latitude variations of E p in the Arctic and Antarctic regions. Symbols and lines are the same as in Figure of 15

14 Figure 15. Time variations of (top) E p and (bottom) DF in the (left) Arctic and (right) Antarctic regions, respectively. analysis at ECMWF and JMA to estimate the temporal and spatial variations of the mean wind, surface wind, planetary wave activity, E-P flux and its divergence. [63] In the Arctic region (50 90 N), E p shows a clear annual variation with a maximum in winter (DJF), which is consistent with the annual variation of F z. The large F z value indicates that planetary waves are active, then the planetary waves could distort the polar vortex. Unbalanced flow due to the distortion of the polar vortex can excite gravity waves through geostrophic adjustment. [64] Planetary waves are active near the polar night jet, which excite gravity waves, therefore the E p enhancement can be localized along the polar night jet. Our study has confirmed the similarity of the distribution of E p and the polar night jet. We also found that E p became large in response to the enhancement of DF during a major warming event in winter, which suggests that active planetary wave is also effective in generating gravity waves through planetary wave transience and/or breaking. To summarize, in the Arctic region geostrophic adjustment relating to planetary wave transience and/or breaking are important in understanding the enhancement of E p. However, we cannot clearly deny the possibility that the wave-mean flow interaction is responsible for the annual and interannual variation of E p. [65] In the Antarctic region (50 90 S), E p gradually increases from July to September and reaches maximum in early spring (September October) before decreasing rapidly. This seasonal variation is different from the Arctic result. The time derivative of V in terms of month coincides with the peak of E p and moreover the horizontal distribution of E p shows a very similar structure with V and 8. These results suggest that the E p enhancement is related to the decay of the polar vortex. Because the enhancement of DF in September October correlates very well with the E p peaks, planetary wave breaking seems to be related to the generation of gravity waves. In the winter months, we can recognize a good correlation between E p and F z. They not only coincide but also have a quantitative consistency. As in the Northern Hemisphere, we can assume that planetary wave activity generates gravity waves via geostrophic adjustment related to planetary wave transience and/or breaking during the winter months. [66] The statistical results for the horizontal distribution of E p are compared with the topography in the Arctic and Antarctic regions. Moderate enhancement of E p is recognized over a few specific areas, such as Scandinavia and the Antarctic peninsula. Therefore orographic effects seem to be one of the mechanisms for generating gravity waves in the polar regions. However, the horizontal E p pattern has a higher correlation with the structure of the polar vortex. Therefore the climatological behavior of the gravity wave energy is most likely to be explained by considering geostrophic adjustment, while orographic effects are less important. [67] Acknowledgments. We would like to thank Y. Kawatani, S. P. Alexander and M. V. Ratnam for their valuable discussions. We are also grateful to T. Horinouchi and J. Furumoto for their helpful comments. This study is partially supported by MOE s (Ministry of the Environment) Global Environmental Research Fund of Japan (A-10) and by a Grant-in- Aid for Scientific Research (B) from Japan Society of Promotion of Science. We are very grateful for the suggestions by the two anonymous reviewers. References Allen, S. J., and R. A. Vincent (1995), Gravity wave activity in the lower atmosphere: Seasonal and latitudinal variations, J. Geophys. Res, 100, Andrews, D. G., J. R. Holton, and C. B. Leovy (1987), Middle Atmosphere Dynamics, 489 pp., Elsevier, New York. Baldwin, M., T. Hirooka, A. Oneill, and S. Yoden (2003), Major stratospheric warming in the Southern Hemisphere in 2002: Dynamical aspects of the ozone hole split, SPARC Newsl., 20. Baumgaertner, A. J. G., and A. J. McDonald (2007), A gravity wave climatology for Antarctica compiled from Challenging Minisatellite Payload/Global Positioning System (CHAMP/GPS) radio occultations, J. Geophys. Res., 112, D05103, doi: /2006jd Eckermann, S. D., and P. Preusse (1999), Global measurements of stratospheric mountain waves from space, Science, 286, Fritts, D. C., and M. J. Alexander (2003), Gravity wave dynamics and effects in the middle atmosphere, Rev. Geophys., 41(1), 1003, doi: /2001rg Hirota, I. (1984), Climatology of gravity waves in the middle atmosphere, J. Atmos. Terr. Phys., 46, Hirota, I., and T. Niki (1985), A statistical study of inertia-gravity waves in the middle atmosphere, J. Meteorol. Soc. Jpn., 63, Kitamura, Y., and I. Hirota (1989), Small-scale disturbances in the lower stratosphere revealed by daily rawin sonde observations, J. Meteorol. Soc. Jpn., 67, of 15