Temperatures and horizontal winds in the Antarctic summer mesosphere



[1] A series of 26 meteorological rockets (“falling spheres,” FS) were launched in January and February 1998 from the Antarctic research station Rothera (68°S, 68°W). These flights gave densities and temperatures below ∼93 km and horizontal winds below ∼75 km, respectively. The lowest altitude is approximately 35 km. The instrumental technique is identical to the one applied in similar studies in the Northern Hemisphere (NH). In this paper, we summarize the experimental results and compare them with climatologies in the NH summer and with empirical models. We concentrate on the mesosphere. In January, temperatures in the upper mesosphere are very low (<135 K) and are very similar to the NH. In February, temperatures increase substantially, certainly more than in the corresponding time period in the NH. The zonal winds show a similar behavior: SH/NH values are very similar in January/July but differ in February/August. This indicates that (at least in 1998) the seasonal transition from summer to winter occurs earlier in the SH compared to the NH. Mass densities are generally similar in both hemispheres. The difference is less than 2–6% and shows a seasonal variation similar to temperatures and zonal winds. Our experimental results at Rothera differ significantly from empirical reference atmospheres such as the COSPAR International Reference Atmosphere (CIRA). For example, in the upper mesosphere our mass densities are up to 35–40% smaller compared to CIRA. Such differences can be important, for example, when modeling the sedimentation of ice particles leading to noctilucent clouds (NLC) and polar mesosphere summer echoes (PMSE). Furthermore, zonal winds in the mesosphere in January/July in the SH/NH are very similar in our measurements but different in CIRA. Recent lidar and radar measurements of NLC at Rothera and PMSE at Davis (68.6°S), respectively, show very similar mean altitudes compared to the NH supporting the similarity of the thermal structures in both hemispheres in January/July.

1. Introduction

[2] The summer mesopause region at high latitudes is one of the least understood regions in the entire terrestrial atmosphere since main contributions to the energy and momentum budget are not fully resolved. Since the first temperature measurements in the late 1950s it is obvious that radiative processes cannot explain the curious seasonal variation of temperatures with extremely low values at the summer mesopause (in fact the lowest temperatures in the entire terrestrial atmosphere) and comparatively large values in winter, some 50–70 degrees warmer compared to summer [Stroud et al., 1959; Theon et al., 1967]. It was first suggested by Murgatroyd and Singleton [1961] that dynamical ageostrophic processes drive the upper atmosphere away from radiative equilibrium and finally cause upwelling and downwelling in summer and winter, respectively, with corresponding cooling and heating. Subsequent theoretical studies suggested that gravity waves provide the momentum deposition required to cause ageostrophic motions [Leovy, 1964; Lindzen, 1981]. Gravity waves are generated in the lower atmosphere, propagate upward and break in the mesosphere thereby producing turbulence. The main uncertainty in our understanding of the dynamical control of the mesopause region is related to the unknown seasonal and spatial variation of gravity wave sources and their filtering and breaking in the upper atmosphere. Since the topography in the troposphere plays an important role in the creation of gravity waves, it is expected that the mesosphere in the Southern Hemisphere is less controlled by dynamical processes compared to the Northern Hemisphere since there is less orographic terrain in the Southern Hemisphere. Therefore temperatures in the austral mesosphere are expected to be closer to radiative equilibrium and therefore higher in the summer season compared to northern polar latitudes.

[3] The question whether or not the thermal structure in the Southern Hemisphere (SH) during summer differs from the corresponding Northern Hemisphere (NH) summer triggered an initiative to measure temperatures in the summer mesosphere at austral polar latitudes by means of the falling sphere technique. This method has frequently been used at northern polar latitudes. Climatologies of summer season temperatures in the Northern Hemisphere have been published for 69°N (Andøya Rocket Range, northern Norway) and for 78°N (Longyearbyen, Spitsbergen) [Lübken, 1999; Lübken et al., 2004]. The former will hereafter be referred to as “FJL99.” In 1998 a field campaign called PMSE Observation and Rocket Temperature measurements over Antarctica (PORTA) was set up to launch falling spheres from Rothera (68°S). First temperature results have been published by Lübken et al. [1999]. In this paper, we present more details on temperatures, densities, and horizontal winds. In section 2 the falling sphere technique is introduced and some exemplary results are shown. Climatologies of temperatures, densities, and winds are presented in section 3. We compare our measurements with NH high latitudes and with layered phenomena in section 4. A summary is presented in section 5.

2. Experimental Method and Rocket Launches

[4] For the so-called falling sphere (FS) technique a small rocket transports a sphere, made of metalized mylar, to an altitude of typically 110 km. After release the sphere inflates to 1-m diameter and passively falls through the atmosphere whereby it decelerates. A high-precision radar tracks the descent trajectory which is then used in the equations of motion to determine atmospheric densities (from deceleration) and horizontal winds (from horizontal deflection). Temperatures are obtained by integrating the density profile assuming hydrostatic equilibrium. The temperature at the top of the FS profile (“start temperature” equation image) has to be taken from independent measurements or from a model. We chose to take equation image from FJL99 which implies that per definition the temperatures in the SH are practically identical to the NH at the top of the profiles (∼93–95 km). It is important to note that this coupling disappears quickly with decreasing altitude. As can be seen in Figure 1 an arbitrary variation of equation image by ±15 K at 93 km reduces to an uncertainty of ±2 K and <1 K at 85 km and 80 km, respectively. Even at mesopause altitudes (∼88 km) where uncertainties increase to approximately ±5–10 K, a SH/NH comparison is meaningful taking into account the even larger errors of alternate techniques.

Figure 1.

Temperature profiles of FS flights labeled TPS04 (blue) and TPS23 (red), respectively. The “start temperature” equation image has been varied by ±15 K (light blue and red lines; see text for more details). The black lines give the corresponding NH temperature profiles published by Lübken [1999].

[5] The height-dependent sphere reaction time constant causes a smoothing of the density, temperature, and wind profiles. The smallest scales detectable in the temperature profile are typically 8, 3, and 0.8 km at 85, 60, and 40 km, respectively [Schmidlin, 1991]. The uncertainty of temperatures is typically 7, 3, and 1.5 K at 90, 80, and 70 km. We note that the FS technique shows excellent overall agreement with entirely different rocket borne temperature measurements with much better altitude resolution [Rapp et al., 2002]. In particular the mean mesopause structure is nicely reproduced.

[6] For winds the maximum altitude where reliable data can be derived is ∼75 km, i.e., significantly lower compared to temperatures [Meyer, 1985]. Typical uncertainties are ±3 and ±1 m/s at 70 and 50 km, respectively.

[7] From the British Antarctic Survey research station Rothera (67°34′S, 68°07′W) a total of 26 falling spheres were launched in the time period from 4 January to 27 February 1998. The spheres were successfully tracked in 24 out of 26 flights and altitude profiles of atmospheric densities, temperatures, and horizontal winds were derived. A complete list of the rocket launches is presented in Table 1. Not all atmospheric parameters could be deduced in all flights in the entire altitude region. Restrictions are listed in Table 1. From a few flights we have temperatures but no absolute densities. This is because certain instrumental deficiencies in the FS technique can lead to errors in absolute densities, whereas density gradients (and thereby temperatures) are not affected.

Table 1. Rocket Launches During the PORTA Campaign at Rothera (68°S)
LabelDateTime, UT
  • a

    No data available because of technical problems.

  • b

    Temperatures not used for climatology.

  • c

    Densities not used for climatology.

  • d

    Data below ∼80 km only.

TPS01a4 Jan. 19981910:00
TPS025 Jan. 19982200:00
TPS03a8 Jan. 19981810:00
TPS048 Jan. 19982115:00
TPS0510 Jan. 19981500:00
TPS06b14 Jan. 19981740:00
TPS0716 Jan. 19981635:00
TPS08c19 Jan. 19981700:00
TPS0921 Jan. 19981305:00
TPS10d23 Jan. 19981712:00
TPS11c26 Jan. 19981703:00
TPS12c28 Jan. 19981705:00
TPS1330 Jan. 19981700:00
TPS1401 Feb. 19981815:00
TPS1504 Feb. 19981830:00
TPS1606 Feb. 19981718:00
TPS1707 Feb. 19981610:00
TPS1809 Feb. 19981810:00
TPS1911 Feb. 19981800:00
TPS2013 Feb. 19981700:00
TPS2117 Feb. 19981334:00
TPS2217 Feb. 19981647:00
TPS2319 Feb. 19981700:00
TPS2423 Feb. 19982015:00
TPS2525 Feb. 19981605:00
TPS2627 Feb. 19982100:00

[8] As an example the temperature profile of flight TPS04 from 8 January 1998 is shown in Figure 1 together with the corresponding high-latitude profile from FJL99 and x = 7.25 (early July) in the nomenclature introduced in FJL99. The similarity between the profile measured at Rothera and the mean profile from the Northern Hemisphere is striking (note that the SH/NH stations are located at almost the same latitude). In the entire mesosphere the difference is typically less than 3 K. Other flights in January show similar agreement but with slightly larger variability around the mean profile. The profile from flight TPS04 in Figure 1 shows very low temperatures at the mesopause with a minimum value of 133 K at an altitude of 87 km.

[9] In Figure 1 we also show a temperature profile measured later in the campaign, namely on 19 February 1998 and the corresponding FJL99 profile (x = 8.50, i.e., mid-August). It is obvious that in the upper mesosphere the SH temperatures are significantly larger compared to the NH. Though not present in all individual profiles this is a persistent feature: Temperatures in the SH upper mesosphere in February are significantly larger by more than 10 K compared to NH August (see next section).

3. Climatology of Temperatures, Densities, and Winds

[10] The individual temperature profiles were smoothed and interpolated to obtain a seasonal variation of the thermal structure in the time period from beginning of January until end of February. The data processing is similar to the procedure described in FJL99, i.e., the temperature data at a given altitude where smoothed by fitting a polynomial function. Typical root-mean-square (RMS) deviations from the fit are 10, 7, and 3–5 K at 90, 85, and below 80 km, respectively. The final temperature climatology is shown in Figure 2 and listed in Table 2 for reference. As can be seen from this figure, temperatures at the mesopause are as low as ∼135 K in January, i.e., as cold as at the northern polar mesopause (a detailed comparison is presented in section 4). In February, however, temperatures in the upper mesosphere increase rapidly and reach 160 K at the end of the month.

Figure 2.

Temperature climatology from beginning of January until end of February at 68°S derived from falling sphere measurements. Individual profiles have been averaged and smoothed (see text for more details). Temperatures are listed in Table 2.

Table 2. Temperature Climatology Deduced From FS Measurements at Rothera (68°S)a
z, kmMonth of the Year
  • a

    Temperatures in Kelvin. Month of the year: 1.5 ≡ 16 January, 2.0 ≡ 1 February, etc.


[11] The mass densities ρ(z) were also smoothed and interpolated, again similar to the procedure described in FJL99. The results are listed in Table 3. Since the smoothing procedure was done independently for temperatures and densities, respectively, we cannot expect the corresponding altitude profiles to be in hydrostatic equilibrium. It turns out, however, that the deviation from a hydrostatic profile is typically less than ±1–2 K below 85 km, which is smaller than the uncertainties generated by the limited number of digits given in Table 3. Instead of showing absolute mass densities we present the relative deviations from CIRA-1986 in Figure 3. As can be seen from this figure the differences are negative everywhere (i.e., FS densities are smaller compared to CIRA) and can be as large as −40% in the upper mesosphere.

Figure 3.

Relative density difference (FS-CIRA)/CIRA in percent. Negative numbers indicate that FS densities are smaller compared to CIRA-1986.

Table 3. Seasonal Variation of Mass Densities at Rothera (68°S)a
z, kmMonth of the Year
  • a

    Read 2.03 as 10−2.03 kg/m3. Month of the year: 1.5 ≡ 16 January, 2.0 ≡ 1 February, etc.


[12] We also derived a climatology of zonal winds in a procedure similar to temperatures and densities, i.e., smoothing the individual data points at a given altitude. In Figure 4 the mean zonal winds are shown. In this figure we have marked the altitude range 70–75 km to indicate relatively large uncertainties, large response time constants of the sphere, and large natural variability (the RMS deviation from the fit is typically 4–10 m/s below 70 km, and 10–15 above 70 km). The zonal wind climatology is listed in Table 4.

Figure 4.

Climatology of zonal winds at Rothera (68°S) derived from falling sphere measurements. Individual profiles have been smoothed. Negative values indicate winds from the east (i.e., westerlies). The altitude range above 70 km has been marked by a grey bar to indicate rather large uncertainties (see text for more details).

Table 4. Zonal Wind Climatology at Rothera (68°S)a
z, kmMonth of the Year
  • a

    Winds in meters per second. Month of the year: 1.5 ≡ 16 January, 2.0 ≡ 1 February, etc. Positive values indicate winds toward the east.


[13] We have also analyzed meridional winds. The results using January flights only are shown in Figure 5. Meridional winds are generally smaller in magnitude compared to zonal winds. We have not compiled a climatology and will not discuss meridional winds further, because in the NH our meridional winds show a substantial tidal variation (different from zonal winds, see next section) which complicates a comparison of our SH data with NH values and with models. We note that the mean winds in Figure 5 are directed toward the pole, i.e., against the mean direction expected from the residual circulation. We should keep in mind, that meridional winds achieve substantial magnitude only at and above the mesopause. Taking into account the winds from the NH we think that this is due to a tidal effect (see next section). Taking into account these limitations and using data from a comparable time period and time of the day only (around noon), we find that the meridional winds in the SH are very similar to the NH. Further details of winds in the NH and comparison to SH are presented by Müllemann and Lübken [2004].

Figure 5.

Meridional winds at Rothera (68°S) from all flights in January. The mean is indicated by a thick black line. Negative values indicate winds toward the South Pole.

4. Discussion

[14] Before we start the discussion on hemispherical differences in the thermal and dynamical structure we need to make sure that no systematic biases hamper this comparison. We note that the experimental technique applied at Rothera and Andøya is identical (and also identical at Spitsbergen). For example, the same supplier was used for the rocket hardware, the same radar was used to track the spheres, and the same data reduction procedure was used to analyze the flights. Therefore the experimental advantages and constraints (e.g., height resolution) are identical in our SH and NH data sets.

[15] Could there be a tidal bias in the mean values and/or in the SH/NH comparison? Nearly all flights in Rothera took place around local noon (≈1600–1900 UT), i.e., at a fixed local time, in order to avoid tidal variations and to facilitate comparison with NH launches. In the NH nearly all flights were performed around noon or midnight. As is discussed in more detail by Lübken [1999], there is most likely no bias in the mean temperatures and in the SH/NH comparison since all measurements have been performed in the same phase of the prevailing (but still small) semidiurnal tide [Forbes, 1995]. Concerning wind measurements in the NH we found a significant tide in the meridional winds. We attribute this observation to the diurnal tide which, according to theory, prevails below 80 km altitude (for a detailed discussion on tides in the NH winds, see Müllemann and Lübken [2004]). No tidal signature was present in the zonal winds in the NH which is probably because our measurements took place at the zero transitions of the diurnal tide. In summary, we do not expect a potential tidal bias in our SH/NH comparison of temperatures and zonal winds nor a bias in the mean values. For meridional winds, however, we expect an influence of tides. In fact, we attribute the poleward wind direction observed at Rothera to a tidal effect (winds in the NH are also poleward around noon and are equatorward around local midnight).

[16] As has been shown by Becker et al. [2004] interhemispherical coupling can influence the dynamical and thermal structure in the mesosphere if planetary wave activity is large in the opposite hemisphere. This mechanism has presumably lead to larger mesopause temperatures in the NH in 2002 when the circumpolar vortex in the SH was distorted because of planetary waves. We note that such a SH disturbance has not been observed before, i.e., our NH temperature climatology is not influenced. Also, the planetary wave activity in the NH was normal during our SH measurements in 1998. We conclude that our climatologies for NH and SH are most likely representative.

4.1. Comparison With NH Climatology

[17] A comparison of the Antarctic mean temperatures with the climatology of FJL99 is shown in Figure 6. The difference is generally very small (within ±3 K) in the upper stratosphere and mesosphere (below ∼75 km) in the entire time period. Furthermore, the difference is very small (less than 3 K) at 80–88 km in January, whereas the difference increases rapidly toward the end of February (up to 12 K warmer at Rothera compared to Andøya). This indicates that the transition from summer to winter occurs earlier in the SH compared to the NH. At the top of the altitude range (∼93 km) the difference is again very small which is a consequence of the data reduction procedure described in section 2.

Figure 6.

Difference of mean temperatures at Rothera (68°S) compared to corresponding summer values at Andøya (69°N). Positive values indicate larger temperatures in the SH.

[18] During the entire PORTA time period mean mass densities in the stratosphere and mesosphere are somewhat smaller in the SH compared to the NH (typically, −2% to −6%). Differences increase to +10% (larger in the SH) above ∼90 km which is in the uppermost altitude bins of the FS technique. The SH/NH mass density differences show a seasonal variation from −2% to −6% which mimics the variation of temperatures and zonal winds (see below). This is not a self-evident result since temperatures are deduced from the mass density gradients, not from the absolute mass densities.

[19] A comparison of zonal winds is shown in Figure 7 (the zonal wind climatology at 69°N is published by Müllemann and Lübken [2004]). Mean differences are rather small in January until mid-February (less than ±3 m/s) but increase to more than 20 m/s toward the end of February. This implies that the SH/NH similarity of the thermal structures in January/July discussed above is also present in the dynamical structure. Furthermore, the seasonal variation of the difference in zonal winds mimics the SH/NH difference in the thermal structure and the mass densities presented above. With other words, the zonal winds also showed that in 1998 the summer season in the SH developed differently from a typical Northern Hemisphere behavior. It is interesting to note that a shift by ∼2 weeks in the seasonal variation of winds has also been observed in MF radar winds at somewhat higher altitudes [Dowdy et al., 2001].

Figure 7.

Zonal mean winds in meters per second at Rothera (68°S) minus corresponding summer values at Andøya (69°N). Positive values indicate larger winds at Rothera compared to Andøya.

4.2. Comparison With Other Measurements and With Empirical and Theoretical Models

[20] In the SH there are basically no other measurements of temperatures available which are of comparable altitude range and seasonal coverage. Chu et al. [2004] have recently published first results from their Fe Boltzmann lidar located at Rothera. Their temperatures are substantially larger compared to our FS climatology and are in fact larger than frost point temperatures (Tf) indicating that it would be too warm for ice particles to exist under these conditions (to compute Tf we have made use of water vapor concentrations from the model of Körner and Sonnemann [2001]; however, Tf is much more sensitive to temperatures than to water vapor). It is important to note that the Fe lidar data were used only in the absence of noctilucent clouds (NLC) because the ice particles introduce large uncertainties in the lidar temperature derivation. This selection has probably introduced a bias of lidar temperatures toward too large temperatures. Taking into account the fraction of NLC presence during the lidar measurements (6 hours in a total of 40 hours) it is still difficult to attribute the FS/lidar temperature difference to this bias. We should keep in mind, however, that ice particles can be present in the summer mesopause region much more frequently than suggested by lidar NLC data. This is known, for example, from a comparison of polar mesosphere summer echoes (PMSE) and noctilucent clouds at northern latitudes [see, e.g., Lübken et al., 2004]. Roughly speaking, the lidar sees the “large” ice particles only, whereas smaller particles can still produce PMSE. Therefore very different occurrence frequencies for NLC and PMSE can emerge. For example at 69°N mean occurrence rates of 30–40% and 80% are observed for NLC and PMSE, respectively [Fiedler et al., 2003; Bremer et al., 2003]. We therefore speculate that ice particles have affected the Fe lidar temperatures also at times when the lidar did not observe NLC.

[21] Our temperatures are much lower by up to 15–20 K compared to the COSPAR International Reference Atmosphere, CIRA-1986 [Fleming et al., 1990], and are also significantly lower compared to the empirical model based on mass spectrometer and incoherent scatter data, MSIS [Hedin, 1991; Lübken et al., 1999]. In Figure 8 we show the difference of our FS temperature climatology to CIRA-1986. In the mesosphere FS temperatures are smaller compared to CIRA by up to 12 K. The differences are smaller around the stratopause. The temperature difference to CIRA-1986 does not vary much with season, whereas the difference to MSIS increases from the beginning of January until the end of February. Regarding SH mass densities (Figure 3) we find small differences to CIRA-1986 (approximately ±5%) in the upper stratosphere and lower mesosphere, whereas relatively large deviations of up to -45% are found in the upper mesosphere, in particular in midsummer. Such large differences can play an important role, e.g., when modeling ice particle sedimentation or when calculating trace gas mixing ratios or collision frequencies.

Figure 8.

Temperate difference of the Rothera climatology compared to CIRA-1986. Negative values indicate lower FS temperatures compared to CIRA.

[22] In Figure 9 we compare all individual zonal wind profiles from January with the CIRA-1986 profiles from January (SH) and July (NH), respectively. This figure shows that zonal winds in the SH are more negative compared to CIRA-1986 even if we consider natural variability. In fact, the mean zonal winds in the SH are very similar to the corresponding July CIRA winds in the NH (we have already noted above that the SH/NH difference in FS winds is small in January/July). This means that the hemispherical difference of zonal winds present in the CIRA climatology is not present in our FS data. Theoretical models on hemispherical differences which use CIRA winds can therefore come to misleading conclusions since they introduce an asymmetry in the dynamical structure which is presumably not real [Siskind et al., 2003].

Figure 9.

Individual zonal wind profiles at Rothera in January 1998. For comparison, the CIRA-1986 mean profiles for 70°S (January, blue) and 70°N (July, red) are shown. The thick black line indicates the mean of the FS profiles. Negative values indicate winds from the east (i.e., westerlies).

4.3. Comparison With Layered Phenomena

[23] There are two phenomena in the polar summer mesopause region which are related to low temperatures, namely polar mesosphere summer echoes (PMSE) and noctilucent clouds (NLC) (the latter are also sometimes called polar mesosphere clouds, PMC). More precisely, these layered phenomena are related to the existence of water ice particles which require temperatures smaller then the frost point temperature Tf. Noctilucent clouds are due to ice particles which occur around 83 km during the summer season at midlatitudes and polar latitudes [Gadsden and Schröder, 1989; Lübken et al., 1996; Fiedler et al., 2003]. These clouds can be detected by lidar or, if the observer is in the dark, by naked eye because they scatter light emitted from a laser or from the sun, respectively. PMSE are very strong radar echoes received from the summer mesopause region at polar and, to a lesser extent, also at midlatitudes. These echoes exist because charged ice particles reduce the mobility of free electrons [Cho and Röttger, 1997; Rapp and Lübken, 2003]. When comparing SH/NH differences in layered phenomena, we should keep in mind that the frost point temperature depends on the water vapor abundance, i.e., whether or not temperatures are small enough for ice particles to exist depends on the water vapor concentration, although only weakly compared to temperatures. Very little is known about the H2O concentration in the upper polar mesosphere during summer, even less on a potential SH/NH difference.

[24] Some of the motivation to make the FS measurements at Rothera came from the observation that PMSE were absent in the Antarctic [Balsley et al., 1993]. We note, however, that the radar observations in the Antarctic were performed at comparatively low latitudes (62°S) and even more equatorward comparing geomagnetic latitudes. Furthermore, a shipborne VHF radar passed by Rothera in February 1998, i.e., during our PORTA campaign, and indeed detected PMSE in the upper mesosphere where simultaneous FS flights showed temperatures smaller than frost point temperatures (R. Woodman, personal communication, 2004). We should also keep in mind that PMSE require not only ice particles but also neutral air turbulence and free electrons. Details of the role of turbulence and charged aerosols on the creation of PMSE have become clear only recently [Rapp and Lübken, 2003]. Whereas there are presumably sufficient free electrons available at both hemispheres (because of ionization by solar Lyα radiation), very little is known about a potential SH/NH difference in turbulent activity. Despite all these uncertainties about PMSE it is striking to note that the first radar measurements at Davis (68.6°S) in 2003/2004 have detected PMSE which in fact show a very similar altitude distribution compared to the NH [Morris et al., 2004; Bremer et al., 2003]. This again confirms the SH/NH similarity of the thermal structure.

[25] Chu et al. [2004] have recently published first results from their lidar detection of NLC at Rothera. They observed NLC from 29 December until 26 January 2003 and derived a mean centroid altitude of 83.7 km. This is amazingly close to the mean value at 69°N (83.3 km, see Fiedler et al. [2003]), in particular if we consider (1) geophysical variability of the centroid altitude (roughly ±1 km), (2) the very different sampling statistics, and (3) the different instrument characteristics.

[26] The NLC layer and in particular the peak altitude is closely related to the temperature profile, more precisely to Tf or equivalently to the degree of saturation S (S = 1 where T = Tf). Assuming a water vapor profile from a model and the temperatures in Table 2 we find the condition S = 1 at an altitude of ∼82 km in January, i.e., approximately 1–1.5 km below the mean lidar NLC peak altitudes mentioned above. This difference is expected considering the standard scenario of NLC development: ice particles start to nucleate at the mesopause, sediment to lower altitudes while growing, and finally evaporate once they reach the altitude level where S = 1. Since evaporation is faster compared to growth the NLC layer is expected to be asymmetric around the altitude where S = 1 and is shifted toward higher altitudes by roughly 1 km (see Figure 7 of Rapp et al. [2002]). This is exactly the height difference between the NLC peak altitudes from lidars (NH and SH) and the altitude levels where S = 1 (NH and SH). We conclude that the NLC observations confirm the FS temperature climatology below the mesopause (certainly, the temperature profile published by Chu et al. [2004] is not compatible with the mean NLC layer height). Furthermore, the amazing SH/NH similarity of NLC properties confirms the similarity of the thermal structures.

[27] Satellite borne measurements of PMC have been used to study SH/NH differences. From the MSX satellite nearly identical mean PMC altitudes at 67.5 degree latitude in the SH and NH are observed: 82.3 km and 82.6 km, respectively (see Figure 6 of Carbary et al. [2001]). We note that these altitudes are approximately 1 km lower compared to the lidar values mentioned above which is certainly outside the uncertainty of the lidar data and could be due to a different definition of cloud altitude and/or a systematic instrumental bias in the satellite data. Former PMC altitudes published from instruments on the SME satellite are much higher in the NH (85 km) but have recently been corrected to smaller altitudes (G. Thomas, private communication, 2004). Taking into account the error bars of satellite PMC the altitudes from SME are now very similar to lidar NLC and are very similar in both hemispheres. A systematic comparison of NLC and PMC altitudes from lidar and satellites is beyond the scope of this paper. We conclude that the satellite observations of PMC altitudes support the similarities between the hemispheres.

5. Summary

[28] A series of 26 rocket borne instruments (“falling spheres”) were launched from the Antarctic research station Rothera (68°S) in January/February 1998. These experiments gave mass densities, temperatures, and horizontal winds in the altitude range from approximately 35 km to 93 km (densities and temperatures) and 75 km (winds), respectively. These measurements showed very low temperatures (∼130 K) in the mesopause region in January, in fact very similar to the Northern Hemisphere (NH) July climatology. Throughout February, temperatures in the SH increased up to ∼160 K at 85 km, significantly warmer than in the corresponding period in the NH. Mass densities are slightly smaller in the SH compared to NH (typical difference is −4% to −6%, except above 90 km where the difference increases up to +10%). The SH/NH mass density difference shows a seasonal variation similar to temperatures and zonal winds. Temperatures and densities are significantly different to empirical reference atmospheres such as CIRA-1986 and MSIS. For example, FS temperatures at and below the mesopause are smaller by up to 12 K compared to CIRA-1986, and the mass densities in the same altitude range are also smaller by up to 40%. Such large differences can play an important role when modeling ice particle sedimentation or when calculating mixing ratios or collision frequencies.

[29] Horizontal winds are negative (westward) in January in the upper stratosphere and mesosphere and turn to positive (eastward) values in mid-February. This wind reversal occurs rather early in the season compared to the NH. We therefore conclude, that in the SH in 1998 the thermal and dynamical structure as well as the mass densities showed a transition from summer to winter conditions which occurred earlier in the SH compared to NH summer. In January, horizontal winds from FS are very similar to the NH July values. This closeness differs from CIRA-1986, where SH/NH horizontal winds deviate by more than 20 m/s. Model studies of hemispherical differences which introduce a SH/NH asymmetry by using CIRA winds may therefore come to misleading conclusions. It is important to note that the horizontal winds from FS are derived independently from densities and temperatures. Winds are deduced from the horizontal drift of the sphere, whereas mass densities and temperatures are derived from the (vertical) deceleration. We conclude, that the general morphology of the dynamical structure in the SH shows a striking similarity to the thermal structure and thereby affirms the similarities and differences to the NH and to empirical models.

[30] Our understanding of NLC/PMC and PMSE and their dependence on ice particles (i.e., on low temperatures) has improved substantially in the last years. Recent lidar measurements of NLC altitudes at Rothera show a striking similarity to NH data. The mean altitudes are in agreement with expectations from models when using our SH/NH temperature climatology. In summary, the morphology of NLC support the similarity between the Southern and Northern Hemisphere thermal structures during summer. As far as satellite detection of PMC is concerned, most measurements show very similar altitudes in both hemispheres, but the mean altitudes for different instruments sometimes differ. The reason for this difference is not yet understood. We note that small differences in the thermal structure can lead to large differences in PMC signals, mainly because optical methods used to detect ice clouds are highly sensitive to the particle size (∼r6), whereas PMSE are much less sensitive (∼r2).

[31] Regarding PMSE, we now have a better understanding of the role of charged ice particles and neutral air turbulence creating these strong radar echoes. The apparent lack of PMSE at 62°S cannot easily been taken as an indication of a general SH/NH temperature difference. In fact, the first detection of PMSE at Davis (68.6°, i.e., nearly identical latitude than Rothera) in the season 2003/2004 shows a very similar altitude distribution compared to the NH. This again confirms the similarity in SH/NH temperatures and demonstrates that our observations in 1998/1999 are probably representative for other seasons. Future observations of PMSE and PMC will give more insight into SH/NH similarities and differences.

[32] A complete understanding of the similarity of the SH/NH thermal and dynamical structure in January/July and the difference in February/August requires detailed modeling. The excitation, filtering, and destruction of gravity waves will probably play a key role in understanding these similarities and differences.


[33] The PORTA rocket campaign was conducted with the excellent support of the crew of the Mobile Raketenbasis (DLR, Germany) and the staff of the British Antarctic Survey. This project was supported by the Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie, Bonn, under grants 50 OE 99 01 (ROMA) and 50 OE 9603 4 (TRAMP).