Low-frequency intraseasonal variations of the wintertime very high latitude mesopause regions

Authors


Abstract

[1] We have performed a spectral analysis of three fall and winter airglow data sets each for South Pole Station and for Eureka, Canada (80°N). These analyses show significant spectral features in the mesopause region at very high latitudes that are similar in period to low-frequency intraseasonal (IS) variations seen in analysis of length of day, atmospheric angular momentum variations, and outgoing long wave radiation, as well as in height or wind fields up to 10 hPa (∼32 km). These are fairly broadband spectral features centered approximately at 17, 23, and 45 days. The 17-day oscillation we find is more prominent than that found in the data sets reflecting variations mainly in the troposphere and stratosphere. This oscillation is most likely associated with the second symmetric s = 1 free Rossby mode. The data from South Pole Station should not reflect this mode, as such, because the spectral components that produce airglow variations should vanish at the pole unless they are zonally symmetric; however, free oscillations having zonal wave number zero with such long periods should not exist. We suggest that, at least at South Pole Station, the 17-day oscillation is driven by interactions between the global Rossby mode and stationary wave number one planetary waves. In addition to the low-frequency oscillations that have been reported in the IS literature, we see a wave in the 12–14 day period range. At South Pole Station that wave may also reflect forcing by a global free mode.

1. Introduction

[2] Considerable evidence has accumulated that periodic disturbances of the airglow reflect periodicities in the underlying wind and temperature field in the upper mesosphere and lower thermosphere. For example, airglow fluctuations exhibit periodicities that reflect the perturbing effects of tides and planetary waves on airglow emissions [Myrabø, 1984; Takahashi et al., 1984; Walterscheid et al., 1986; Hernandez et al., 1992a, 1992b; Sivjee and Walterscheid, 1994; Sivjee et al., 1994; Wiens et al., 1995; Espy and Witt, 1996; Hecht et al., 1998; Haque and Espy, 2000; Schubert et al., 1999; Walterscheid et al., 1999]. Collateral evidence of the presence of tidal and planetary wave oscillations at very high latitudes comes from airglow and radar observations [Hernandez et al., 1992a, 1992b; Sivjee et al., 1994; Sivjee and Walterscheid, 1994; Espy and Witt, 1996; Hecht et al., 1998; Portnyagin et al., 1997; Forbes et al., 1995a; Walterscheid and Sivjee, 1996]. These oscillations cover periods of a few hours to ∼2 weeks.

[3] Here, we extend the range of periodicities previously covered at very high latitudes (latitudes poleward of ∼ ±80°) to include low-frequency oscillations intraseasonal periods. Following Ghil and Mo [1991a], we define low frequency as periods greater than 10 days, this being roughly the limit of predictability for atmospheric disturbances. The oscillations are reflected in the absolute brightness of the near infrared (1.0μ–1.7μ) airglow OH-Meinel (OH-M) band emissions. The observations are airglow brightness from South Pole Station and Eureka, Canada (80°N). The data comprise three winter/fall data sets each for South Pole Station and for Eureka. The data are taken at 3 azimuths at South Pole Station and Eureka and also from the zenith direction at Eureka. The data for azimuth pairs at South Pole Station are cross-spectrally analyzed for wavelength bands that exhibit coherent oscillations and for phase differences between azimuths. Phase differences indicate longitudinal structure at South Pole Station.

[4] The existence of significant low-frequency intraseasonal (IS) oscillations at middle latitudes is well established [Walterscheid, 1980, and references therein; Dickey et al., 1991; Magaña, 1993; Ghil and Mo, 1991a, 1991b; Mechoso and Hartman, 1982; Branstator, 1987; Elsner, 1992; Schubert and Park, 1991; Berbery and Nogués-Paegle, 1993; Schubert et al., 1993; Strong et al., 1993; Cheong and Kimura, 2001]. The periods of significant northern hemisphere IS oscillations are around 17, 23 and 45 days, with amplitudes generally ordered as period (i.e., the 45-day wave is strongest). These periodicities are seen in analyses of length of day (LOD), atmospheric angular momentum (AAM) variations, outgoing long wave radiation (OLR) and analyses of the geopotential heights of pressure surfaces and wind fields. We refer to these waves as the 17-, 23- and 45-day oscillations, even though they are observed to be fairly broadband (for example the 45-day oscillation occupies the 40 to 50-day band).

[5] The 23-day oscillation is a prominent and persistent feature of the high latitude northern hemisphere disturbance field, having its largest amplitude near 70°N. Its amplitude increases upward at least as high as 10 hPa (∼32 km). It may, on occasion, be the strongest IS oscillation [Branstator, 1987].

[6] Similar periodicities are found in the southern hemisphere, and, as in the northern hemisphere, the 23-day oscillation can dominate [Ghil and Mo, 1991b; Lau et al., 1994]. The 17-day oscillation is again the weakest [Ghil and Mo, 1991b]. An analysis of LOD and AAM gives a somewhat different mixture of waves, with a strong maximum near 31-days, in addition to moderate peaks in the 45-day band and weak peaks in the 17- and 23-day bands [Dickey et al., 1991].

[7] The IS waves are also seen in model simulations [Ghil and Mo, 1991a; Li and Nathan, 1997; Strong et al., 1993; Schubert et al., 1993; Simmons et al., 1983]. The 45-day oscillation is similar in period to the Madden-Julian oscillation in the tropics, but the two do not seem to be necessarily connected [Magaña, 1993; Dickey et al., 1991; Lau et al., 1994; Schubert et al., 1993]. The middle latitude oscillation occurs in GCM results that do not exhibit a Madden-Julian oscillation [Ghil and Mo, 1991a]. Berbery and Nogués-Paegle [1993], on the other hand, found a connection between tropical and extra-tropical IS that appears to be related to oscillations in the 30–60 day band.

[8] The 17-day oscillations may be the second symmetric s = 1 Rossby mode [Longuet-Higgins, 1968; Salby, 1981; Walterscheid, 1980, and references therein; Forbes et al., 1995b]. This mode has been identified in data analysis as early as 1954 [Kubota and Iida, 1954] and has been seen over a wide range of latitude and altitudes, including the very high latitude mesopause region [Sivjee et al., 1994]. However, there is some uncertainty in the identification of the 17-day wave as a normal mode, and it may be in the nature of barotropic instability or a normal mode forced by instability [Walterscheid, 1980; Ghil and Mo, 1991a]. The wave at high latitudes is weakly correlated with the wave at lower middle latitudes, and the two may have different sources [Mechoso and Hartman, 1982]. The 23- and 45-day oscillations may be barotropic modes of instability, or originate from instabilities associated with flow over topography [Legras and Ghil, 1985; Simmons et al., 1983; Ghil and Mo, 1991a, 1991b; Strong et al., 1993]. The amplitude of IS disturbances can show considerable intra-annual and interannual variability [Branstator, 1987; Lau et al., 1994]. Sources of variability may be the differences in the background atmosphere and complex interactions between IS disturbances [Simmons et al., 1983; Ghil and Mo, 1991a; Lau et al., 1994].

[9] The 23- and 45-day traveling disturbances are not pure zonal wave number harmonics. The leading disturbances in the northern hemisphere are dominated by wave numbers 1–3 [Branstator, 1987; Ghil and Mo, 1991a]. At high latitudes, the 45-day disturbance at 700 hPa is dominated by westward traveling s = 1, and the 23-day disturbance by westward traveling wave numbers 1–3 at 500 hPa and 1 and 2 at 50 hPa [Branstator, 1987; Ghil and Mo, 1991a]. In the high-latitude southern hemisphere, the traveling components at 500 hPa exhibit higher wave number content than for the northern hemisphere, with wave numbers 3–5 dominating [Ghil and Mo, 1991b; Lau et al., 1994]. The disturbances are characterized by temporal complexity, perhaps reflecting a less dominant role for topography [Ghil and Mo, 1991b].

[10] Numerical simulations and empirical orthogonal eigenfunction (EOF) analysis show the low-frequency IS oscillations should extend to the pole, although with greatly reduced amplitude [Simmons et al., 1983; Ghil and Mo, 1991a, 1991b]. This indicates a zonally symmetric component. To the extent that the IS oscillations are reflected in the airglow at South Pole Station, the existence of s = 0 components of the IS oscillations are indicated. In spectral terms, this most likely implies a coupling between the dominant zonally asymmetric components and the zonally symmetric wave number component induced nonlinearly by zonal asymmetries in the background. The low-frequency nature of the oscillations of interest means that they are more sensitive to the background state, including nonzonal variations, than tides or short-period normal modes. In generalized terms, the eigenfunctions associated with a low-frequency eigenfrequency need not be a pure zonal harmonics and it may be somewhat artificial to view the various wave numbers as distinct but coupled oscillations [Salby, 1984].

[11] In the remaining sections we discuss the airglow measurements and data, present the analysis approach, present the results of the analysis and discuss the results for the low-frequency IS oscillations evident in the airglow at very high latitudes.

2. Data

[12] Near infrared (NIR) Michelson Interferometers (MI), with 2″ optics, about 3° field of view (throughput = 0.06 cm2-sterad) and fitted with a 3-stage thermoelectrically (TE) cooled 1 mm diameter InGaAs detector (NEP < 10–14 Ws1/2), operate at the South Pole Station (SPS) in Antarctica and at Eureka (80°N). Each MI scans from 1.0 μ to 2.0 μ, at 4 cm−1 resolution, every 30 seconds s. Absolute brightness calibration of the MIs is accomplished with the aid of a blackbody source operating at 1273 K and illuminating a Lambertian screen. A periscope directs sky light sequentially from three grid-azimuths (0°, 120° and 240°) at 25° elevation into the MI; the periscope dwells at each of the three positions for fifteen minutes, completing the cycle every 45 min. For airglow OH-M emissions peaked around 87 km, the 25° elevation at the SPS corresponds to viewing the mesopause airglow emitting region that is located at 88.6°S. Thus a 15-minute averaged OH-M NIR spectra at 86.6°S from the three grid-azimuths are recorded every 45 min.

[13] Water vapor absorption of the NIR airglow OH-M emissions is minimal at SPS and Eureka due to very low air temperature, and consequently very dry polar lower atmosphere. The effects of low absorption slightly impact the OH-M (2,0) band, but the P1 and P2 rotational lines of the OH-M (3,1) and (4,2) bands are free of any measurable absorption effects.

[14] The rotational population of the OH (X, v′) states is thermalized in the mesopause region [Sivjee and Hamwey, 1987]. Hence the brightness of each rotational emission line (from each rotational level J) can be expressed in terms of the Boltzman distribution, the radiative transition probability (A) and statistical weight (2J + 1) as follows:

equation image

where F(J) is the rotational energy level in wave numbers (cm−1) and T is the ambient kinetic temperature in the mesopause region around 87 km. Taking the natural log of equation (1) leads to

equation image

where α = hc/k = 1.4387755 cm-K . Data plotted by Walterscheid and Sivjee [2001] (Figure 1) show a linear relation between ln{B(J)/(2J + 1)A(J)} and F(J), demonstrating the thermalization of the OH-M (3,1) P1(2,3,4) and P2(2,3,4) rotational line emissions and the absence of any significant telluric absorption effects on the first six rotational lines of the P branch.

Figure 1.

The airglow brightness in Rayleighs versus day of year for the South Pole Station for 1995. The data have been averaged over 5 days to eliminate the large-amplitude oscillations with periods of 5-days and less that dominate the airglow fluctuations at very high latitudes (solid curve). Also plotted is a 3-wave least squares fit to the data with periods of 22, 33 and 45 days. (dashed curve).

[15] The total brightness of the OH-M (3,1) band, for each sampling period, is derived by two different methods. A least squares synthetic profile fit to the observed OH-M (3,1) band brightness profile, coupled with the absolute brightness calibration, yields the total band brightness. In addition, the brightness of the first six rotational lines of the P branch, coupled with T determined from equation (2), leads to the total band brightness (B) as follows:

equation image

The two methods yield consistent brightness values of the OH-M (3,1) band. The fractional error in brightness is less than 0.5%.

[16] We report on data taken in three fall/winter seasons at Eureka (1995,1995/1996 and 1997) and South Pole Station (1995, 1996, and 1997). Figure 1 shows the airglow brightness time series for the South Pole Station for 1995 for an azimuth of 240°. The abscissa is the day of the year. The data have been averaged over 5 days to eliminate the large-amplitude oscillations with period of 5-days and less that dominate the airglow fluctuations at very high latitudes [Sivjee and Walterscheid, 1994; Sivjee et al., 1994; Walterscheid et al., 1986; Walterscheid and Sivjee, 1996, 2001]. The experimental error for the averaged data is negligible (≪0.5%). Also plotted is a 3-IS wave least squares fit to the data with IS periods of 22, 33 and 45 days. The three monochromatic IS waves do a good job of representing the low-frequency content of the brightness variation, despite the fact that the periodogram shows that the IS oscillations are rather broadband. Higher-frequency non-IS content in the range ∼10–15 days is required to provide a fit to the higher frequency fluctuations seen in the time series.

3. Data Analysis

[17] To search for wave-caused variations in airglow emission brightness, we use standard spectral techniques [Hecht et al., 1987; Sivjee et al., 1994]. Specifically we perform spectral and cross-spectral analyses on the time series for each of the azimuthal directions. Cross-spectra and derived quantities (coherency and phase) are obtained for pairs of azimuths: 0–120°, 120–240° and 240–360° for South Pole Station. At South Pole Station the various azimuths correspond to meridians.

[18] We analyzed the longest period in each year that did not contain cloudy periods (gaps) exceeding 10 days. The longest gap in the data thus selected was 6.2 days (Eureka, 1996–1997). The longest gap for other periods did not exceed 3.7 days. Lomb-Scargle periodogram analyses were performed on the unevenly spaced data in lieu of conventional power spectral density analyses [Press et al., 1992]. However, to facilitate the cross-spectral analyses, the data were interpolated onto an evenly spaced grid using a least squares procedure that preserved the frequency content. The original brightness and temperature data were fit by a Fourier series up to the Nyquist frequency defined for Lomb-Scargle periodograms for unequally spaced data. The fit was then projected onto an equally spaced grid having a temporal spacing of 5 days. To check the effect of this procedure on the spectral content of the data set, we calculated the power spectrum from an FFT performed on the equally spaced data and compared the results with the Lomb-Scargle periodogram based on the original time series. Over the range of periods of interest we found no significant spectral features in the FFT that were not in the Lomb-Scargle periodogram and visa versa.

[19] The data are quadratically detrended over each time series to remove the seasonal variation. (An exception is the data for the winter of 1997/98 at Eureka. This time series was only about 40 days long and detrending would remove periods of interest.) A Hanning windowing function is used to suppress artifacts of the finite duration of the times series, and a Bartlett windowing function is applied to reduce the variance of the spectral estimate [Jenkins and Watts, 1968; Sivjee et al., 1994]. The bandwidth of the Bartlett window is 0.1.25 × 10−4 day−1 for 40-day data segments [Jenkins and Watts, 1968, Table 6.6]. For longer data segments, this bandwidth is multiplied by 40/DL where DL is the length of the data set in days. For the longer data sets, this resolution is more than adequate for resolving the spectral peaks of interest, but marginal for shortest data set (42 days). The raw data were divided by the mean brightness for each period to facilitate comparisons between data sets.

[20] In the cross-spectral analysis we analyze brightness fluctuations at three azimuths located 120° apart. This means that s = 3n (n an integer) waves will always be confounded with s = 0 waves. Significant spectral features were identified on the basis of high coherency and also proximity to known periodicities. High coherency is defined as values large enough such that the 95% confidence bounds do not include zero coherency. Confidence bounds were calculated for squared coherency and the phase spectrum following Jenkins and Watts [1968, pp. 379–381]. As a means of building up the statistical reliability of the spectral estimates of CSD, coherency and phase we co-added cross-spectra from the different azimuths. The results of the spectral analysis are presented in the following section.

4. Results

[21] In this section we present the results of a spectral and cross-spectral analysis for airglow brightness observed from South Pole Station and Eureka, Canada.

4.1. South Pole Station

[22] Figures 24 show the Lomb-Scargle periodograms for airglow brightness for South Pole Station for the years 1995–1997. They are the average of the periodograms for each of the 3 azimuths. We have used oversampling factors (OSF) of 1 and 4 [Press et al., 1992]. The higher oversampling provides a more structured spectra than an OSF of 1. The spectra with higher over-sampling factors do not necessary add significant information. However, significant information may be added when the spectra is band limited, and all finite duration samples are to some extent band limited. This will be especially true for the limited 42-day time series for Eureka in 1997. We will emphasize the OSF = 4 periodograms, but when appropriate, we refer to periodograms with OSF = 1 to help clarify the significance of some details of the spectra.

Figure 2.

Azimuth-averaged periodogram of airglow brightness for South Pole Station for a period in austral fall and winter (days-of-year 104–238) in 1995 (a); days-of-year 104–238 in 1996 (b); and days-of-year 105–242 in 1997 (c). The dashed horizontal lines are the 1% probability levels for the null hypothesis calculated following Press et al. [1992].

Figure 3.

Average amplitude and phase of the cross-spectrum for azimuth pairs as described in the text for days-of-year 104–238 in 1995. The top panel gives the squared coherency versus period and the bottom panel gives the phase of the cross-spectrum. The upper dashed curve in each panel gives the upper 95% confidence limit and the lower dashed curve gives the lower 95% confidence limit [Jenkins and Watts, 1968].

Figure 4.

Same as Figure 4 except days-of-year 104–238 in 1996.

[23] All spectral features noted below are highly significant. We have compared the spectral amplitudes to the power levels associated with the 1% probability level for the null hypothesis calculated according to an approximation used by Press et al. [1992]. The computed confidence level is plotted as a horizontal dashed line. Hernandez [1999] suggests a different computation for the confidence interval based on the work of Fischer [1929]. We have also calculated a confidence bound (not shown) following Hernandez [1999] and found that the bound is higher (more stringent) by about 5 units. All of the putative IS oscillations seen in our data are significant by both measures of confidence.

[24] Figure 2 shows periodograms for three periods in austral fall and winter: a 134-day period in 1995 for days of the year (DOY) from 104 to 238, (a); a 111-day period in 1996 for DOY 130 to 241 (b); and a 137-day period in 1997 for DOY from 105 to 242 (c). In 1995 there are two large distinct spectral features present: one centered near 21 days, the other near 33 days. Lesser peaks near 12–14 days and 50 days are also present. There is also a narrow peak near 17 days, but this peak does not appear for OSF = 1. The spectrum is dominated by three large spectral features: a large double-humped spectral feature in the range 15–17 days; a broad feature (that also may be double humped) from 30–60 days having its peak power near 45 days; and a narrow peak located near 12 days. The double humped nature of the 17- and 41-day features is not present at an OSF of 1, and, while there is a suggestion of a feature at 17 days, it is not distinct. The spectrum exhibits considerable structure. It shows a strong dominant peak near 36 days, smaller but still prominent peaks near 14 and 60 days, and smaller peaks near 20 and 25. For OSF = 1, these peaks merge into a peak near 15 days, a shallow peak near 20 days and a large broad peak near 35 days.

[25] Figures 35 show the squared amplitude (squared coherency) and phase of the normalized cross spectrum. The dashed curves represent the upper and lower 95% confidence bounds. Figure 3 is for the same period in 1995 as Figure 2a. The coherency is high throughout. Also, the phase is not significantly different from zero. Most likely this is because of changes in the large-scale background that are coherent over the limited region covered by the observations (the area poleward of 88.6°S). However, there is clearly increased coherency at 12 and 20 days, especially the latter. The fact that the phase in the 12 and 20-day bands does not deviate from zero is consistent with a dominant zonally symmetric component. It is also consistent with a s = 3 component, but this seems unlikely so close to the pole. Figure 4 is similar except for the period in 1996. Again coherency is high at all periods shown. There are two broad and not too distinct maxima covering the bands from about 15 to 20 days and 25 to 50 days, though the limits of the latter band are poorly defined. Again the phase does not depart significantly from zero. Figure 5 shows the cross-spectral results for 1997. Here the coherency appears to reflect the IS disturbances to a much greater extent. There are dominant statistically significant peaks near 14 and 35 days. The phase is consistent with the dominance of zonally symmetric disturbances.

Figure 5.

Same as Figure 4 except days-of-year 105–242 in 1997.

4.2. Eureka

[26] Figure 6a shows the periodogram for the 70-day period for DOY 290–360 in 1995. It shows two dominant peaks near 17 and 25 days. The elevated power at the lowest frequencies is most likely due to imperfect removal of the seasonal variation. There is also a suggestion of a peak near 12 days. Figure 6b shows the periodogram for the 152-day period for DOY 284–436 (DOY 284–366 of 1996 and DOY 1–70 of 1997). There is a large dominant peak near 25 days, strong peaks near 18 and 35 days. There is also a lesser peak near 12 days. Small, but possibly still significant peaks are located near 15 and 17 days. The increased power at the lowest frequencies is again probably the result of imperfect detrending. At an OSF of 1 the 18-day peak is much less distinct, merging with minor peaks at 14 and 16-days on its flank to form a broad region of increasing power to the left of a dominant 25-day peak. Figure 6c is similar except it covers the limited 42-day period in 1997 for DOY 303–345. The results show a dominant peak near 17 days and a smaller peak near 12 days. At an OSF of 1 the peaks are not resolved.

Figure 6.

Azimuth-averaged periodogram of airglow brightness for Eureka for a period in boreal fall and winter (days-of-year 290–360) in 1995 (a); days-of-year 284–366 in 1996 and 1–70 in 1997 (b); and days-of-year 303–345 in 1997. The dashed horizontal line is the 1% probability level for the null hypothesis calculated following Press et al. [1992].

5. Discussion

[27] The low-frequency IS spectral features seen in the mesopause region at very high latitudes (South Pole Station and Eureka) reflect, at least in period, the IS waves seen in height and wind fields in the lower atmosphere and in LOD and AAM variations. There are distinct features corresponding, at least in frequency, to the 17-, 21 and 45-day oscillations. The double-humped features may arise from structure in the periodogram due to unsteadiness in the IS disturbances. A distinct ∼35 day oscillation is seen in the airglow over South Pole Station. This may be associated with the 31-day feature seen in AAM variations in the southern hemisphere extratropics [Dickey et al., 1991].

[28] Our results differ from previous results in that they are for very high latitudes and for altitudes near the mesopause. In spectral terms our results differ mainly in the prominence of the 17-day oscillation, especially at Eureka, and the existence of an oscillation with a period near 12 days. The 17-day mode has been seen in earlier analysis of airglow data at high latitudes, including Eureka, where next to the 5-day wave it is the dominant oscillation at sub-tidal frequencies [Sivjee et al., 1994]. The 17-day oscillation is most likely the second symmetric Rossby mode [Sivjee et al., 1994]. This identification was made on the basis of theoretical modeling, the prior identification of this mode in meteorological data fields and the ubiquitous nature of the oscillation [Madden, 1979, and references therein; Walterscheid, 1980, and references therein; Salby, 1981; Mechoso and Hartman, 1982; Cheong and Kimura, 2001]. However, as mentioned, one cannot rule out a barotropic mode of instability [Ghil and Mo, 1991a].

[29] The airglow brightness at South Pole Station in 1996 shows a dominant spectral feature near 17 days. There is little evidence of a 17-day wave in 1995, and in 1997 the peaks show a preponderance of power short-periodward of the 23-day band in the range 13–14 days. This is evident also in the coherency (Figure 5).

[30] If the 17-day feature observed from South Pole Station is related to the global second symmetric Rossby wave its prominence so close to the pole must be evidence that it is a zonally symmetric component forced by the global mode. This is because the s = 1 global oscillation in the quantities that force airglow variations (temperature, major gas density, vertical winds, and divergence) must vanish at the pole [Sivjee et al., 1994]. In addition, the cross-spectral analysis shows no evidence of significant zonal variations. Slow variations such as the 17-day wave are subject to considerable modification by the background winds, and it seems reasonable that interactions of the global 17-day mode could interact with, for example, stationary s = 1 planetary waves to produce a zonally symmetric wave with the same period. Similar arguments apply to the slower IS waves.

[31] The occurrence of a wave with a period close to a previously reported 10-hour s = 1 normal mode period was observed in the airglow at the South Pole Station [Walterscheid and Sivjee, 2001]. This wave has zonally symmetric structure and was attributed to an s = 0 normal mode having a similar period [Walterscheid and Sivjee, 2001]. Though less likely than in the case of the much slower 17-day wave, it is also possible that the 10-hour wave in the neighborhood of the south pole is an s = 0 wave forced by interactions with the background stationary waves.

[32] The longer period waves seem likely to be associated with the IS variations seen in the LOD, AAM and analysis fields at 700 and 500 hPa and higher. This interpretation is supported by the very large 23-day wave seen in the Eureka data for 1997. The 23-day wave studied by Branstator [1987] was the dominant IS in some periods and was largest in the high latitude North American sector where Eureka is located. Also the amplitude of the disturbance was greatest at 10 hPa (∼32 km), the highest altitude studied, indicating that the wave is not confined to the lower stratosphere [Simmons et al., 1983]. If the waves with yet longer period belong to the same class of oscillations, then perhaps upward penetration of these waves to the uppermost regions of the middle atmosphere is plausible. This possibility must await modeling that includes the upper mesopause and lower thermosphere for more theoretical support.

[33] Finally, the only feature that we have observed that does not match previously reported IS waves is the feature with a period in the range 12 to 14 days seen in the south pole data. There are no zonally symmetric normal modes with such long periods [Longuet-Higgins, 1968]. However, there are a variety of free Rossby modes with similar periods, and perhaps the oscillation observed from South Pole Station is forced by one of these modes by means of the coupling mentioned above. If so it may be related to the second anti-symmetric s = 1 Rossby mode 10-day wave [Cheong and Kimura, 2001].

Acknowledgments

[34] Work at The Aerospace Corporation was supported by NASA Grants NAGW-2887 and NAG5-4528 and by Embry Riddle Aeronautical University subrecipient contract on NSF Grant ATM97-14648. Programming support by Ms. X. Tran is gratefully acknowledged. Work at the Embry Riddle Aeronautical University was sponsored by the National Science Foundation through grants ATM-9714648, OPP-9909339, OPP-9910950, ATM-9528953, ATM-9804674 as well as by NASA grant NAGS-10066. Engineering support for the Polar research facilities and data analyses services provided by John Pesce are gratefully acknowledged.

[35] Janet G. Luhmann thanks Roger W. Smith and another referee for their assistance in evaluating this paper.

Ancillary