Pronounced upper stratospheric and mesospheric NOx enhancements were measured by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) in the Southern Hemisphere (SH) polar vortex from May to August 2003, reaching average abundances of 60 ppbv at 50–60 km in July. Peak mixing ratios of around 200 ppbv were measured in the polar night, representing the highest values ever recorded in the SH. The observed NOx enhancements are attributed to production by electron precipitation in the upper mesosphere and lower thermosphere and subsequent descent with the meridional circulation. Using measured CH4 and CO distributions as dynamic tracers, the downward transport of NOx-rich air masses into the lower and middle stratosphere has been investigated. Upper atmospheric air with average NOx abundances of 15 ppbv reached the 800–1000 K potential temperature region (around 30 km) by the end of July, where it remained until the final warming in late October. The NOx descent was confined to the polar vortex, although significant mixing of tropical and NOx-rich vortex air masses began already in August above 40 km. The amount of upper atmospheric NOy measured inside of the SH vortex in late spring was 1.1 Gigamoles (GM) which is in good agreement with previous estimates from HALOE data. The global coverage of MIPAS data further allows to quantify the upper atmospheric NOx dispersed into the stratosphere during August-September, estimated in 1.3 GM. The net deposition of NOx into the stratosphere during the 2003 Antarctic winter (2.4 GM) makes up 9% of the N2O oxidation source in the SH, twice as much as estimated in previous studies. NOx and tracer distributions observed on several days during the NH winter 2002/2003 have been analyzed for comparison. We found that high planetary wave activity, resulting in the major midwinter warming had led to a rather inefficient NOx downward transport with negligible deposition of NOx into the lower and middle stratosphere.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Nitrogen oxides, NOx = NO + NO2, are the major drivers of catalytic ozone loss in the middle stratosphere. In this altitude region, NOx forms a pronounced layer peaking around 35–40 km with volume mixing ratios (VMRs) of typically 10–17 ppbv. The major chemical source for stratospheric NOx is oxidation of N2O injected from the troposphere in the tropics. A second NOx layer exists in the thermosphere and is produced by dissociation of molecular nitrogen by solar photons and energetic particles, in particular auroral and precipitating electrons from the outer trapping region of the magnetosphere, and subsequent reaction of N(4S,2D) with oxygen to form NO. Depending on solar and geomagnetic activity, maximum thermospheric NO VMRs can vary from 100 up to 2000 ppmv, which, in terms of number densities, becomes comparable to stratospheric NOx abundances. Both stratospheric and thermospheric NOx layers are normally separated by a mesospheric minimum caused by reactions
which reduce the chemical lifetime of NO to only a few days.
 In the absence of sunlight during polar winter, however, large amounts of upper atmospheric NOx can be transported down to the mesosphere and stratosphere by the meridional circulation without being photochemically destroyed. This mechanism was already proposed by Solomon et al.  and Frederick and Orsini . Once transported down to the stratosphere, NOx is photochemically stable and may thus contribute to the stratospheric NOy budget and hence to the global O3 variability [Siskind, 2000]. Enhanced upper stratospheric and mesospheric NO2 reaching levels of about 175 ppbv at 70 km at polar night conditions was measured by the Limb Infrared Monitor of the Stratosphere (LIMS) instrument [Russell et al., 1988] during the 1978/1979 winter in the Northern Hemisphere (NH), representing the first experimental evidence for downward transport of thermospheric NOx. By correlating CH4, NO, and NO2 measured by the Halogen Occultation Experiment (HALOE) experiment on the Upper Atmospheric Research Satellite (UARS), Siskind et al.  clearly demonstrated the existence of NOx-rich upper atmospheric air in the Southern Hemisphere (SH) polar winter stratosphere during 1992–1996. Similar observations were made by the Atmospheric Trace Molecule Spectroscopy (ATMOS) experiment [Rinsland et al., 1999] and the Polar Ozone and Aerosol Measurement (POAM II) Instrument [Randall et al., 1998] for different NH and SH winters. All measurements suggest that the NOx descent is well confined within the polar vortex. In various model studies and comparisons to NOx data from HALOE and the Stratospheric Aerosol and Gas experiment (SAGE II), Callis et al. [1998a, 1998b, 2001, 2002] investigated the role of electron precipitation from the outer trapping region of the magnetosphere as observed aboard TIROS spacecraft on the polar winter descent of NOx. They identified this production mechanism as the dominant NOx source in the polar upper atmosphere and found a significant modulation of stratospheric NOy and O3 by electron precipitation within the solar cycle. Randall et al.  and Siskind et al.  have shown that interannual variations of the NOx enhancements in the polar winter stratosphere are closely linked to variations of the geomagnetic Ap index, suggesting that NOx downward transport is predominantly controlled by the upper atmospheric source rather than by dynamical conditions. On the other hand, a pronounced hemispherical asymmetry was found in NOx HALOE observations, with larger polar winter enhancements occurring in the SH [Siskind, 2000]. This hemispheric asymmetry could be reproduced by 2-D chemical transport model (CTM) calculations when including a larger gravity wave drag in the SH and enhanced planetary wave forcing in the NH which demonstrates the important role of dynamics [Siskind et al., 1997]. However, the CTM model failed to reproduce mesospheric NO enhancements measured by HALOE at latitudes as far equatorward as 30°–40° and it overestimated the net deposition of NOx in the stratosphere. Both limitations were attributed to an underestimation of horizontal mixing in the lower mesosphere. Thus, despite the numerous measurements confirming the downward transport of NOx in polar winter, the overall understanding of its mechanism is limited by the lack of NOx data poleward of 50° along the whole winter [Siskind, 2000].
 A local source of mesospheric and upper stratospheric NOx is highly energetic solar protons (>1 MeV) ejected during strong solar storms and penetrating into the Earth's middle atmosphere in the polar regions. During the July 2000 SPE, NOx enhancements of 50–100 ppbv were detected by HALOE at 65–70°N in the 55–70 km altitude range [Jackman et al., 2001]. In October and November 2003, an unusually powerful solar proton event (SPE) led to local NOx enhancements reaching 180 ppbv in the North polar upper stratosphere as measured by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) experiment [López-Puertas et al., 2005]. In principle, SPE-induced NOx production takes place in both hemispheres with a similar magnitude. However, the impact on stratospheric ozone chemistry is more severe in the winter hemisphere because of the longer photochemical lifetime of NOx and favorable dynamical conditions. López-Puertas et al.  have reported significant NOx-induced O3 depletion after the October-November 2003 SPE in the North polar stratosphere while in the Austral polar regions changes in O3 and NOx disappeared within a few days.
 In situ production of lower mesospheric and upper stratospheric NOx is also possible by precipitating electrons from the outer trapping region with energies larger than 100 keV. Callis and Lambeth  reported NO2 enhancements of up to 140 ppbv around 60 km during the 1991 NH winter and the 1992 SH winter which had been attributed to such electron events.
 The Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) [Fischer and Oelhaf, 1996; European Space Agency, 2000] measures both NOx constituents, NO and NO2, as well as the dynamical tracers CH4 and CO with global coverage and independent of illumination conditions. It is thus perfectly suited for studying polar winter NOx enhancements in the mesosphere and upper stratosphere and its downward transport. Here, we analyze MIPAS measurements regarding NOx, CH4, and CO covering the winter poles between September 2002 and October 2003. In the Antarctic winter 2003, strong NOx enhancements were observed which were larger than ever reported in the SH. In this study, we focus on this winter in order to assess the deposition of upper atmospheric NOx into the stratosphere.
 In section 2, MIPAS observations and the data analysis method are described. In section 3, we discuss the NOx and tracer distributions observed during the Antarctic winter 2003. Possible NOx production mechanisms are assessed in section 4. The modulation of the NOx descent by atmospheric dynamics is discussed in section 5. A quantitative estimation of the stratospheric deposition of upper atmospheric NOx is given in section 6. Finally, we compare the observed NOx descent in the Austral winter 2003 with the Arctic winter 2002/2003 in section 7.
2. MIPAS Observations and Data Analysis
 MIPAS is a limb emission Fourier transform spectrometer operating in the midinfrared spectral region. It has been designed for measurement of atmospheric trace species from space [European Space Agency, 2000; Fischer and Oelhaf, 1996]. It is part of the instrumentation of the Environmental Satellite (ENVISAT) which was launched into its Sun-synchronous polar orbit of 98.55° inclination at about 800 km altitude on 1 March 2002. MIPAS operated from July 2002 to March 2004 at full spectral resolution of 0.05 cm−1 in terms of full width at half maximum (apodized with the strong Norton and Beer  function). MIPAS observes the atmosphere during day and night with global coverage from pole to pole and thus provides trace gas distributions also during polar night. Within its standard observation mode, MIPAS covered the altitude range from 6 to 68 km with tangent altitudes from 6 to 42 km every 3 km, and further tangent altitudes at 47, 52, 60, and 68 km. MIPAS passes the equator in southerly direction at 10.00 am local time 14.3 times a day. During each orbit up to 72 limb scans are recorded. The Level-1b processing of the data, including processing from raw data to calibrated phase-corrected and geolocated radiance spectra, is performed by the European Space Agency (ESA) [Nett et al., 1999].
 Data presented and discussed in this paper are vertical profiles of abundances of NO2, NO, CH4 and CO, which were retrieved with the dedicated scientific IMK-IAA data processor [von Clarmann et al., 2003a, 2003b] from spectra (versions 4.55 to 4.59) recorded from November 2002 to October 2003. The temporal coverage of observations analyzed in this study are summarized in Table 1. Retrieval strategies considering nonlocal thermodynamic equilibrium (non-LTE) effects, error budget and altitude resolution for the species under investigation are reported by Funke et al.  for NO and NO2, Funke et al. [2001, 2003] for CO, and Glatthor et al.  for CH4. The estimated precision in terms of the quadratic sum of all random errors is better than 1 ppbv for NO, at an altitude resolution of 4–7 km. The accuracy, derived by quadratically adding the errors due to uncertainties in spectroscopic data, temperature, non-LTE related parameters, and horizontal gradients to the measurement noise error, varies between 0.6 and 1.8 ppbv. The precision, accuracy, and altitude resolution of the NO2 retrieval is estimated at 0.2–0.3 ppbv, 0.3–1.5 ppbv, and 3.5–6.5 km, respectively. For CO, precision, accuracy, and altitude resolution are 0.01–0.1 ppmv, 0.03–0.2 ppmv, and 4–15 km, respectively, and for CH4 0.015–0.14 ppmv, 0.11–0.24 ppmv, and 3–5 km, respectively.
Table 1. Temporal Coverage of Available IMK/IAA Data
Number of Orbits
Number of Observations
16 Nov. 2002
24 Nov. 2002
1 Jan. 2003
11 Jan. 2003
26 Jan. 2003
18 Feb. 2003
30 Mar. 2003
11 Apr. 2003
19 Apr. 2003
29 Apr. 2003
11 May 2003
17 May 2003
23 May 2003
6 June 2003
9 June 2003
20 June 2003
1 July 2003
11 July 2003
21 July 2003
1 Aug. 2003
11 Aug. 2003
21 Aug. 2003
31 Aug. 2003
22 Sep. 2003
30 Sep. 2003
10 Oct. 2003
 Since these episode-based scientific MIPAS-IMK-IAA data are available only for selected periods, we have also included the operational ESA NO2 and CH4 data (reprocessed data version 4.61/4.62) in our analysis. ESA data are retrieved with the operational retrieval algorithm as described by Ridolfi et al.  and Carli et al. . These operational MIPAS data include neither NO nor CO; further, effects due to non-LTE are ignored in the retrieval of NO2 there.
 In order to analyze the retrieved trace gas profiles in a dynamical context, potential vorticity data from ECMWF analysis has been used. The operational ECMWF medium-range forecasting system is based on a global atmospheric model with comprehensive parameterizations of physical processes, tightly coupled with an ocean wave model. Variables are represented at 60 levels in the vertical, ranging from the ground 0.1 hPa (around 65 km). Incremental four-dimensional variational data assimilation (4D-Var) is used to produce initial conditions, on the basis of in situ and remotely sensed observations of temperature and winds.
 Whenever the boundary of the polar vortex was relevant to our analysis, we have applied the vortex boundary criterion suggested by Nash et al.  modified such that, the second indicator besides potential vorticity, was the first derivative of a horizontal tracer distribution instead of the wind field. Above the 2000 K potential temperature level we have used the CO field, while CH4 was used below.
3. Temporal Evolution of NOx and Tracer Distributions
 A comprehensive picture of the observed NOx descent during Antarctic winter 2003 can be gained from the MIPAS NO2 and CH4 ESA operational data with quasi-continuous temporal coverage. Because of the rapid conversion of NO into NO2 in the absence of sunlight, nighttime NO2 is a reasonable proxy for total NOx. However, it should be noted that because of the neglect of non-LTE effects in ESA processing, their NO2 VMRs might be underestimated by up to 30% above 50 km [Funke et al., 2005]. Figure 1 (top) shows nighttime NO2 abundances at potential temperatures between 625 K and 3000 K (approximately 25–60 km) averaged on a daily basis within 60–90°S equivalent latitudes representing roughly the polar vortex in the March–November 2003 period. The NO2 enhancement appeared at 3000 K in early May. It reached its maximum of 60 ppbv around 1 July. NO2 then decreased continuously until it disappeared at the beginning of September. At lower altitudes, the enhancements appeared later and with damped intensity. Finally, a tongue of enhanced NO2 reaches the 700–800 K level by the end of September. The observed NO2 enhancements are correlated in time and altitude with decreased CH4 abundances (Figure 1, bottom) confirming that the enhanced NO2 was descending from the upper atmosphere. However, the CH4 decrease above 2500 K started already in April while NO2 began to build up in May. This delay was caused by efficient photochemical loss of NOx in April when the polar night area was still small. Besides the region of subsiding air with low CH4 concentrations, a tongue of enhanced CH4 can be seen, localized around 3000 K in March and descending to 1200 K in June. These CH4 enhancements were generated by an accelerated Brewer-Dobson circulation with strong poleward transport of midlatitudinal air masses rich in CH4 preceding the downward motion. In August, CH4 abundances started to raise again above 1500 K, indicating that isentropic transport to the polar region augmented in the upper stratosphere and mesosphere, probably at the same time when subsidence began to weaken. The increase of CH4 in August went along with a pronounced decrease of NO2. The steep increase of CH4 between 700 and 1200 K in mid-October indicates the final warming, which led to the polar vortex rupture. The South polar region was filled with midlatitude air masses rich in CH4 and NO2. Thus the NO2 increase after 15 October around 1000 K is not related to upper atmospheric NOx. The instantaneous NO2 enhancement on 29 October at the right border of Figure 1 (top) was produced by the major SPE discussed in detail by López-Puertas et al. .
 NOx and tracer distributions have been analyzed in more detail using the IMK/IAA data of NO, NO2, CO, and CH4 for the available days during Antarctic winter 2003. Since the efficiency of NOx downward transport depends on the extension and permeability of the polar vortex, the vortex boundaries have been determined from the ECMWF potential vorticity together with measured CH4 and CO distributions, as described in section 2. Given that the temperature and wind measurements in the upper stratosphere assimilated in the ECMWF model are sparse, ECMWF wind and potential vorticity data represent an unverified forecast. Therefore consistency of these data has been checked with the measured tracer distributions. As an example, measured CO and CH4 abundances and potential vorticity along equivalent latitude are shown in Figure 2 at the 2750 K isentrope for days 23 May, 9 June, and 21 August 2003. Pronounced horizontal gradients of all quantities can be seen at the vortex boundary close to 35°S in May and June, indicating a strong and isolated vortex. In the tracer distributions, there is no hint at transport out of the vortex. This changed in August, when horizontal gradients in the tracer abundances and potential vorticity were weak. There is evidence of mixing across the vortex boundary, then. This is in agreement with the temporal evolution of CH4 ESA data (Figure 1), showing a CH4 increase in August above 1500 K in the polar region. Also, the temporal evolution of the ECMWF zonal mean zonal winds at 2000 K potential temperature in the SH (Figure 3) suggests a strong and extended vortex in midwinter and a weakening of the vortex in mid-August. The reversal of the mean zonal winds from easterlies to westerlies indicate the final breakup of the vortex in October.
Figure 4 (top) shows the temporal evolution of the vortex boundary position in equivalent latitudes. Until 15 May, the vortex edge was localized around 50–60°S. From June to the end of July, when the NOx enhancements were most pronounced, the vortex above 1500 K was extended up to 30°S equivalent latitudes. With the weakening of the vortex in August, its latitudinal extension shrank again until its final breakup above 1500 K at the beginning of September. The apparent expansion of the polar vortex above 1700 K in September and October is produced by advection of the vortex tracers to the tropics after the final breakup.
 The temporal evolution of the total NOx abundance averaged over the vortex area is presented in Figure 4 (middle). The pattern of the NOx descent is very similar to that observed from the ESA NO2 nighttime data. Maximum NOx vortex averages of up to 60 ppbv were found at 3000 K in July. The vortex-averaged CO abundances (Figure 4, bottom) increased steadily until mid-May because of the descent of mesospheric air, while decreasing slowly afterward. During June and July, this decrease went along with the expansion of the vortex to the tropics. The stronger decrease in August above 1500 K was related to the weakening of the vortex with enhanced mixing across the boundary. Descent rates of 400 m per day are derived from the CO isolines above 1500 K in May which is qualitatively consistent with typical wintertime descent rates in this altitude range [Garcia and Solomon, 1985]. Lower descent rates are derived from CH4 (250 m per day). However, CH4 data could be affected by strong mixing below the subsidence zone leading to underestimated descent rates.
 The distribution of NOx and CO inside the vortex at the 2500 K isentropic surface is shown in Figure 5 for days 9 June and 1 July along with ECMWF potential vorticity. High values of both NOx and CO were confined to the vortex with sharp horizontal gradients at the edge. As expected, highest NOx abundances were found in the polar night region with VMRs up to 8 times higher than in the illuminated part of the vortex. An opposite behavior is seen in the CO distributions which were more abundant in the outer part of the vortex, suggesting enhanced descent in the subpolar regions. Strong vertical transport close to the inner edge of the vortex boundary is also visible in Figure 2 (top), always showing the highest CO abundances close to the vortex boundary. Since the Eliassen-Palm flux divergence often peaks at midlatitudes in the SH [Garcia et al., 1992], enhanced descent in the subpolar regions is not unusual. Potential vorticity had an isotropic distribution in the vortex boundary region and outside of the vortex. Inside the vortex, and most pronounced in the outer vortex region, however, potential vorticity was rather randomly distributed indicating strong turbulent mixing. This stirring process could have been responsible for the pronounced horizontal gradient in the NOx VMR at the polar night terminator being slightly smoothed out on 1 July compared to 9 June. In fact, filaments of enhanced NOx released from the polar night region showed up at around 150°E on both days. At altitudes above 50 km where NO photolysis is efficient, turbulent mixing inside the vortex could have contributed significantly to an accelerated NOx loss by transporting NOx to illuminated regions where it could be more easily destroyed.
Figure 6 shows the measured NOx distribution over equivalent latitude as 10° running mean for 8 days between 11 April and 31 August along with the vortex boundary. Single NOx measurements and potential vorticity along equivalent latitude for potential temperature levels of 1000 K, 1750 K, and 2500 K are shown in Figure 7. As early as 11 April, a slight increase of 5 ppbv above 2500 K can be seen in the mean NOx abundances close to the vortex core (Figure 6). Single measurements reached here values of 14 ppbv (Figure 7). Mean NOx abundances increased to 15–20 ppbv by the end of April.
 On 17 May, apart of the enhancements at the vortex center, a second increase of the mean NOx abundances shows up close to the vortex edge above 2000 K (Figure 6). This provides further evidence for a strong descent at the inner side of the vortex boundary. Highest NOx abundances with mean values exceeding 150 ppbv were found on 9 June. Even at 1500 K, the mean NOx VMR was around 100 ppbv. Peak VMRs of single measurements exceeded 200 ppbv (Figure 7). The abundance of NOx decreased strongly toward the outer regions of the vortex as already seen in Figure 5. On 1 July, NOx mean VMRs were smaller again, although more homogeneously distributed throughout the vortex. Since on this day the highest NOx concentrations were still found in the polar night region (see Figure 5), the homogeneous distribution along equivalent latitudes demonstrates that potential vorticity was rather randomly distributed inside the vortex, i.e., vortex core and polar night region did not coincide. Enhanced turbulent mixing inside the vortex, smoothing out differences of the NOx load inside and outside the polar night region, is evident in the distributions of single NOx measurements on 1 July compared to 9 June above 1750 K (Figure 7). A high degree of confinement of the NOx enhancements to the vortex area is clearly visible.
 On 21 July, and even more pronounced on 11 August, mean NOx abundances were decreasing in the outer region of the vortex above 1500 K (see Figure 6). This could have been caused by either photochemical loss or by dilution due to NOx-poor tropical air brought into the vortex. The first possibility seems to be unlikely since NOx depletion was not more pronounced at higher altitudes, as one would expect in the case of photochemical loss. Furthermore, as shown in section 6, photochemical NOx loss rates inside the vortex were in the order of 1% per day, much less than required to explain the observed decrease. The observed NOx depletion is thus more likely caused by mixing across the vortex boundary. In the 1000–1500 K region, a tongue of enhanced NOx moving out of the vortex might hint at a NOx outflow on these two days. A flow of NOx-rich upper atmospheric air out of the vortex at 1750 K and 2500 K can also be observed on 21 August in the bottommost plot of Figure 7.
 On 31 August, the pronounced horizontal gradients in the NOx distribution at the vortex edge had disappeared above 1500 K indicating that the vortex was seriously weakened and did not act as a transport barrier anymore. Mean NOx abundances above 1500 K decreased to approximately 10 ppbv which is close to the mixing ratio expected under undisturbed conditions. Below 1500 K, however, NOx enhancements persisted and were confined to the vortex.
4. NOx Sources
 In this section, possible sources for the NOx enhancements in the Antarctic winter 2003 are discussed. It is of particular importance to clarify if in situ production due to energetic particle precipitation had contributed to the observed enhancements or if the NOx was mainly produced in the upper mesosphere/lower thermosphere and subsequently transported downward. Three minor SPEs happened during the Austral winter 2003 on 28 May, 31 May, and 18 June. However, GOES 11 proton fluxes of 10 MeV energy (http://www.sec.noaa.gov/Data/goes.html) able to produce NOx at altitudes around 65 km, were in all occasions 200 times less intense than in the October–November 2003 SPE where an instantaneous NOx increase of up to 180 ppbv was measured in the polar upper stratosphere [López-Puertas et al., 2005]. 30 MeV proton fluxes producing NOx down to 50 km were even smaller by a factor of 700. Thus it is rather unlikely that in situ production due to solar protons had significantly contributed to the NOx enhancements measured in the 2003 SH polar winter. The relevance of NOx production in the lower thermosphere by solar protons of lower energy (i.e., 1 MeV) can be assessed by comparing ion pair formation rates around 90 km due to these protons with the 9-years average (1979–1987) of ion pair formation rates due to energetic electron precipitation reported by Callis et al. [1998b]. The daily mean ion pair formation rate at 90 km due to 1 MeV protons on 29 May, when the most intense SPE happened, can be estimated from the GOES 11 proton fluxes, assuming that 0.1 ion pairs are produced per cubic centimeter and second by a proton flux of 1 proton cm−2 s−1 sr−1 [Jackman and McPeters, 2004]. The resulting rate of 280 cm−3 s−1 is about 10 times less than the average ion pair formation rate due to energetic electrons and decreased rapidly after the SPE. We thus conclude that NOx formation due to solar protons in the lower thermosphere contributed only to a minor extent to the upper atmospheric NOx production in the Austral winter 2003.
 A further local source of NOx in the stratosphere and lower mesosphere is highly energetic (>100 keV) precipitating electrons from the outer trapping region of the magnetosphere. Daily fluxes of precipitating electrons with energies >100 keV, and >300 keV averaged over L > 2.5 and 60–90°S as measured by the MEPED instrument on NOAA 16 (http://www.poes.ngdc.noaa.gov/data/avg) are shown in Figure 8 (top). Precipitating electrons of energies 100 keV and 300 keV are able to produce NOx at altitudes around 75 km and 60 km, respectively. If in situ production of NOx due to these electrons was active, an immediate or slightly retarded temporal correlation of measured NOx at the upper observational limit (around 70 km) and the electron fluxes should be visible. Electron fluxes show pronounced peaks around 1 May, 29 May, 1 August, and 29 October. Electron flux increases coincident with SPEs (i.e., 29 May and 29 October) shall be interpreted with caution since MEPED electron measurements are compromised by the presence of protons, although SPEs are thought to be associated with elevated electron fluxes inside the polar caps. Figure 8 (bottom) shows measured nighttime NO2 from ESA data at 2750 K potential temperature averaged over 60–90°S equivalent latitudes. No temporal correlation with any of the electron fluxes is seen. The apparent double-peak structure in the measured NO2 in early June and July can be explained by dynamics: the first peak is related to the coincident maximum of downward transport visible in the minimum of CH4 abundances, while the second maximum is produced by the maximum NOx load of subsiding air around solstice with the minimum photochemical NOx loss rates. This becomes even clearer when looking at the measured NOx/CO ratio from IMK/IAA data which is a good measure of the NOx load of subsiding upper atmospheric air and thus not affected by dynamical modulation: The ratio is increasing monotonically until 1 July and decreased afterward without any visible modulation related to highly energetic electron sources. Instead, the NOx/CO ratio was temporally correlated with the area covered by polar night, controlling the NOx photochemical loss during the downward transport, with a delay of approximately 2 weeks. We thus conclude that no significant in situ production of NOx below 75 km due to electron precipitation took place in the Antarctic winter 2003. The NOx enhancements seem to be more likely produced at higher altitudes by medium energy electrons (<30 keV) either precipitating from the outer trapping region or from auroral events. No significant temporal correlation of medium energy electron fluxes and measured NOx around 60–70 km is expected since diffusive vertical and horizontal mixing in the mesospheric surf zone [Dunkerton and Delisi, 1985] may smooth out the temporal variability of the NOx source while being transported downward from the lower thermosphere.
 In order to assess the magnitude of the lower thermospheric electron source during the Austral winter 2003 in an interannual context, geomagnetic Ap indices have been averaged over the polar winter season (May–August) for the years 1997–2003. Ap indices have been widely used as a qualitative indicator of electron precipitation in the lower thermosphere [Siskind et al., 2000; Randall et al., 1998]. The seasonal average Ap of 2003 is 23, which is considerably higher than for the years 1991–2002 (mean value of 12, ranging between 7 and 17); with the exception of 1991 with a seasonal Ap of 30. Enhanced electron precipitation is expected in the declining phase of the solar cycle [Rangarajan and Barreto, 2000; Callis et al., 2001], which is the case in 1991 and 2003. In this sense, NOx production in the lower thermosphere in 2003 was rather high within a time frame of a decade, but might have been typical in the context of solar cycle progression.
5. Dynamical Processes
 Apart from the strength of its upper atmospheric source, the NOx flux into the stratosphere is modulated by time-dependent dynamical factors. Gravity wave breaking leads to accelerated vertical velocities and eddy diffusion, acting on the stratospheric NOx deposition in a double sense: first, the downward flux is increased, and, second, the photochemical loss of NOx during the descent is reduced because of a shorter exposure time to sunlight. Siskind  reported estimated vertical advection velocities at 70 km ranging from 2 km per day at the pole to 850 m per day at 50°S. From other CTM calculations [Garcia and Solomon, 1985, 1994; Garcia et al., 1992], vertical velocities of 0.8–1.3 km per day are estimated in the 50–90°S region in July at 70 km in agreement with Siskind , while at 50 km, model vertical velocities are only around 300–500 m per day. We have derived vortex average downward velocities of 400 m per day at 45–55 km from the temporal evolution of measured CO (see section 3) which is in agreement with CTM predictions. Above the stratopause, however, descent rates in the SH polar winter might be overestimated by models, as recent LIDAR temperature measurements over the South pole suggest [Pan and Gardner, 2003].
 Horizontal transport processes affect the NOx flux into the stratosphere by redistributing NOx from the polar night area to illuminated regions where it is photochemically destroyed. The breaking of planetary waves accelerates horizontal mixing, increases the extension and permeability of the polar vortex, and determines the timing of the spring warming and the breakdown of the vortex. By inclusion of planetary wave forcing in model calculations of the NOx descent during a typical Antarctic winter, Siskind  achieved good agreement with HALOE observations above 60 km. In the lower mesosphere (50–60 km), however, an additional mixing term was required to reproduce the HALOE observations of high NOx at midlatitudes. As discussed in section 3, MIPAS data give evidence for isentropic mixing inside the vortex at this altitude range. The latitudinal distribution of MIPAS NOx at 2750 K (approximately 0.27 hPa) in July is in very good agreement with both HALOE observations in June–August averaged over the years 1992–1996 and the model calculations of Siskind  with inclusion of the additional mixing term (compare Figure 9 of this work and Figure 8 of Siskind ). The MIPAS data overcome the observational limitations of HALOE and demonstrate further agreement of model and observations in the 90–50°S region.
 The MIPAS NOx and tracer distributions also suggest that the polar vortex above 1500 K is extended to latitudes equatorward as far as 30–40°S. The average SH midwinter vortex boundary at 2000 K was determined from U.K. Met Office analyses from 1991 to 2001 to be located at 40–45°S [Harvey et al., 2002]. Thus the upper stratospheric polar vortex in the 2003 SH winter was more extended than usual, but probably not exceptional in this context.
 In the model calculations of Siskind , the vortex boundary was assumed to act as an efficient transport barrier. This assumption seems to be justified by the pronounced horizontal gradients in the observed NOx and tracer distributions at the vortex boundary at least until the end of July 2003. However, as discussed in section 3, this assumption no longer holds from August to October 2003. Above 1500 K in this period, the vortex started to weaken and mixing across the vortex boundary took place, resulting in decreased in-vortex NOx abundances. This is different in the model calculations which show enhanced NOx (>30 ppbv) at these altitudes until end of September. The early spring warming in the upper stratosphere and mesosphere in August as observed in 2003 is not unusual: Monthly averaged zonal winds from the CIRA-86 climatology at 50 km show a pronounced decrease from 90–100 m s−1 at their maximum at 40°S in June/July to around 40–55 m s−1 in August/September. This agrees well with the temporal evolution of ECMWF zonal winds at 2000 K during June–September 2003 (Figure 3).
6. Stratospheric NOx Deposition
 The quantification of the amount of NOx brought down into the stratosphere before being photolyzed is extremely important in order to assess the impact of upper atmospheric NOx on stratospheric ozone chemistry. This amount was calculated by Siskind  for a typical Antarctic winter as the difference of amounts modeled with and without including a thermospheric source. Under consideration of planetary wave forcing and additional mixing in the lower mesosphere, this amount was determined to be 6 Gigamoles (GM). However, estimations of the stratospheric NOx deposition from HALOE data [Siskind et al., 2000] gave maximum amounts of only 1 GM during Antarctic winters with elevated electron precipitation. This amount was estimated from the stratospheric subcolumn measured by HALOE between approximately 600 and 1000 K around beginning of October. Within this altitude range, highest peak mixing ratios of around 11 ppbv at 800 K were measured in 1991 and 1994. MIPAS NOx shows peak mixing ratios of 13–15 ppbv in early October 2003 at slightly higher potential temperature levels (1000 K). These small differences suggest less subsidence in 2003 compared to the HALOE observations of previous years. However, subcolumn amounts are quite similar when integrating the MIPAS data up to 1250 K. The severe disagreement between the model of Siskind  and experiments can only be explained by either an overestimation of the upper atmospheric source or by underestimation of the dynamical or photochemical losses. As modeled and observed NOx distributions agree rather well until July, the disagreement seems to be caused by an underestimation of transport out of the upper stratospheric vortex in late winter by the model. Dynamical losses due to enhanced mixing across the weakened vortex boundary in the upper stratosphere reduces the NOx flux into the region below 30 km, where it acts on the polar springtime ozone chemistry. However, NOx transported out of the vortex at 30–50 km represents still an additional source for upper stratospheric NOx on a global scale which has not been assessed so far.
 The study of the temporal evolution of the total amount of NOx accumulated inside the vortex in the stratosphere and lower mesosphere allows to estimate (1) the overall amount of NOx transported downward from the upper atmosphere during the polar winter, (2) the fraction remaining in the vortex in spring, and (3) the fractions lost by mixing out of the vortex at different altitude regions. MIPAS NOx data offers an excellent opportunity to assess this temporal evolution since global data coverage is given all over the Austral winter 2003. By integrating measured NOx densities inside the vortex on isentropic surfaces and subsequently integrating over altitudes between potential temperatures of 625 K to 1250 K, 1250 K to 2000 K, and 2000 K to 3000 K, we have calculated total amounts of NOx in three different regions. The lowest one, 625–1250 K, represents the part of the vortex which remains stable until mid-October. The lower boundary 625 K was chosen because CH4 abundances at this level remain constant from July to October indicating that no further subsidence took place below. This altitude region is approximately the same as that chosen by Siskind et al.  in order to estimate the upper atmospheric NOx content of the SH polar vortex in spring. The 1250–2000 K region represents the part of the vortex which is weakened already in August but photochemical loss is not significant. Finally, the highest region accounts for the remaining altitudes up to the highest levels which has been measured.
 In order to account for photochemical losses, daily mean photochemical loss rates Ld at different potential temperature levels have been derived by time integration of the NO loss rates due to photodissociation at the location of each observation within 24 hours and subsequently averaging these losses over the vortex area. Measurements were weighted with the number density at the considered altitudes and with cos(θ) to account for the variation of the representative surface of each measurement with latitude, θ. Solar zenith angle–dependent NO photolysis rates have been calculated using the parameterization of Minschwaner and Siskind . It has been assumed that each photolysis event destroys two NO molecules because of subsequent recombination of N and NO (R2). The calculated daily mean loss rates are illustrated in Figure 10. At 3000 K, Łd varies between 0.007 and 0.01 day−1 before August with a minimum in June. Later in the year, Ld increases up to 0.1 day−1 in October. At 1700 K, Ld stays below 0.002 day−1, except for October. Below 1500 K, values of Ld are negligible small. The amount of photochemically destroyed NOx accumulated until a given day within the altitude regions defined above was then calculated by integrating the product of Ld and the total NOx amount from 11 April until the day under consideration. Ld and NOx amounts between the days with data availability of IMK/IAA products have been linearly interpolated.
 To estimate the amount of upper atmospheric NOx deposited in the middle stratosphere, the conversion of NOx to NOy reservoir gases has to be taken into account. We thus add to the NOx amounts measured at 625–1250 K and 1250–2000 K the amounts of HNO3 and 2N2O5 inferred from MIPAS data as described in a companion paper [Stiller et al., 2005]. Only the minor species ClONO2, BrONO2, HNO4, and NO3, which do not significantly contribute in this altitude region, were not included in the NOy budget. In order to remove from the measured NOy amounts the contribution which has no upper atmospheric origin, we subtract the values measured on 11 April for the 1250–2000 K and the 2000–3000 K regions. For the 625–1250 K region, the amount of NOy measured on 1 July has been subtracted, assuming that on this day the upper atmospheric NOx had not reached yet this altitude region while background NOx and NOy had already subsided. For NOx, this assumption is justified from the temporal evolution shown in Figure 1. For the other NOy species, Figure 4 of Stiller et al.  indicates that the choice of 1 July results in the best trade off between excluding background NOy and including NOy produced by conversion of upper atmospheric NOx into its reservoirs.
Figure 11 shows the temporal evolution of the amounts of upper atmospheric NOx and NOy observed inside the vortex in the different altitude regions. The amounts of NOx lost by NO photolysis and converted into the reservoir gases are presented separately. The increasing peak amount with decreasing altitude reflects the fact that vertical velocities increased with altitude which led to an accumulation of NOx molecules at lower altitudes. The resulting contributions on the day when NOy amounts maximize, and on 10 October are summarized in Table 2. The total amount of NOy of upper atmospheric origin measured in the whole altitude region (625–3000 K) reached its maximum around 10 August with a value of 2.7 GM. Adding to this the amount of NOx photolyzed until 10 August, we get a proxy of the total amount of NOx brought below 3000 K which is 3.1 GM; assuming that (1) loss by mixing out of the vortex is negligible before beginning of August and (2) further injection of NOx from the upper atmosphere after 10 August can be neglected. The first assumption is supported by the observed NOx and tracer distributions discussed in section 3. The second assumption is justified since, first, photochemical loss rates at the 3000 K (and above) were extremely increased (see Figure 10) which led to destruction of almost all NOx before the air reached the 3000 K level (see Figure 4), and second, vertical velocities should be significantly decreased after the weakening of the vortex.
Table 2. Measured Amounts of Upper Atmospheric NOx and Reservoir NOy, as Well as Accumulated Photochemical NOx Losses in the Three Altitude Regions (See Text for Definition) Inside the Polar Vortex, on Day D When Maximum NOy Amounts Are Found, and on 10 October (in GM)
Potential Temperature Region
Day D of Maximum NOy
Reservoir NOy (D)
Chemical Loss (D)
NOx (10 Oct.)
Reservoir NOy (10 Oct.)
Chemical Loss (10 Oct.)
 The accumulated NOx losses in the 625–3000 K region during the winter can be calculated by subtracting from the total amount brought in from above (3.1 GM) the NOy amounts measured there at the end of the winter (1.1 GM). Of this total loss of 2 GM, accumulated photochemical loss accounted for 0.7 GM. The remaining amount (1.3 GM) represents dynamical loss due to mixing out of the vortex. It is further interesting to estimate the contributions to the dynamical loss at the different altitude regions. In the 2000–3000 K region, the decay of measured NOx after 10 August is compensated by the estimated photochemical losses, indicating that all NOx molecules were destroyed before they could have contributed to the dynamical loss. The amount of NOy molecules being transported out of the 625–1250 K region is given by the difference between the measured amounts at this region on 10 August and 10 October, i.e., 0.65 GM, assuming that after mid-August the downward flux from above was weak. The remaining amount (another 0.65 GM) represents then the dynamical loss of the 1250–2000 K region.
 Finally, the net deposition of upper atmospheric NOy below 3000 K (i.e., the amount which is not photochemically destroyed) can then be estimated at 2.4 GM, an amount which still is less than half of the model prediction (6 GM) by Siskind . The annual NOy production in the SH by oxidation of N2O has been estimated at 26–29 GM [see Siskind et al., 2000, and references therein]. Our analysis of MIPAS data suggests that upper atmospheric NOx has contributed to the SH stratospheric NOx source in 2003 with 9% of the N2O oxidation source. 4% of this amount stayed in the polar vortex until the final warming in October, and 5% was mixed out of the vortex before.
7. Comparison to Arctic Winter 2002/2003
 The asymmetry between the NOx downward transport in the NH and SH has already been mentioned in section 1. In this section we discuss NOx and tracer observations obtained from MIPAS data in the Arctic winter 2002/2003 and compare them to the following Austral winter. From the dynamical aspect, the Arctic winter 2002/2003 was affected by a major midwinter warming event during January, a phenomenon which is rather common in the NH. The temporal evolution of nighttime NO2 and CH4 abundances from ESA off-line data averaged over the 60–90°N equivalent latitude region is shown in Figure 12. NO2 enhancements above 2000 K are visible from 10 November until 20 February with a pronounced interruption in January due to the major warming. The maximum enhancements, however, reached only 15 ppbv, a factor of 4 less than in the Antarctic winter 2003. During the first injection, NO2 rich upper atmospheric air had reached the 1500 K level before it disappeared by the end of December. The high NO2 abundances below 1500 K showing up around 1 January were due to mixing of midlatitude NOx-rich air into the polar region rather than transported down from above, as the simultaneous increase of CH4 at these altitudes indicate.
 It is interesting to note that upper atmospheric air, which had already reached the 1250 K level at this time, descended rapidly to levels around 800 K, as elevated CO abundances measured below 1000 K demonstrate (not shown here). This rapid descend is a typical feature of warming events [Manney et al., 2005]. However, this descend brought no NOx into the lower stratosphere, since the NOx amount of the descended air masses was still rather low. The evolution and origins of middle and lower stratospheric NOx during the Arctic winter 2002/2003 is discussed in more detail by P. Konopka et al. (Ozone loss driven by nitrogen oxides and triggered by stratospheric warmings may outweigh the effect of halogens, submitted to Geophysical Research Letters, 2005).
 The second injection of upper atmospheric NO2 during February reached down to the 2000 K level with VMRs of around 8 ppbv. It is unlikely that this NO2 rich air was transported further down, since the vortex between 1000 K and 2000 K remained weak after the warming as high CH4 abundances in this altitude range indicate.
 It is illuminating to examine the spatial NOx and CO distributions from IMK/IAA data on the isentropic surface of 2500 K for days before and at the beginning of the warming event (Figure 13). As early as 24 November, the vortex seems to have been disturbed by planetary wave activity resulting in an influx of tropical air in the 120–150°E region and a shift of the vortex core to the 45°W sector. On 1 January the vortex had nearly split and the bulk of CO was then seen around 65°N/90°W outside the polar night region. As a consequence, the NOx downward flux took mainly place in illuminated regions, significantly reducing the amount reaching the 2500 K level.
 Averaged daily photochemical loss rates of NO exceeded 0.02 day−1 in the 2000–3000 K region on 1 January, 3 times larger than found in the Antarctic winter 2003 in the corresponding period. Since the photochemical loss during the transport from the source region down to the stratosphere scales exponentially with the loss rates, the impact of the higher values in the Arctic winter 2002/2003 compared to the following Austral winter on the stratospheric NOx net deposition is severe.
 Apart from the dynamical factors, the small NOx downward flux in this Arctic winter is also related to a relatively weak electron precipitation activity as indicated by a seasonal mean Ap index of 15, much lower than in the Austral winter 2003 of 23.
 In summary, we conclude that high planetary wave activity resulting in the major midwinter warming led to a rather inefficient NOx downward transport with negligible deposition of NOx in the lower and middle stratosphere. Since major warming events are a typical feature of the Northern Hemisphere, this Arctic winter is certainly not an unusual case.
8. Summary and Conclusions
 Pronounced upper stratospheric and mesospheric NOx enhancements have been measured by the MIPAS instrument in the SH polar vortex from May to August 2003, reaching maximum average abundances of 60 ppbv inside the vortex in the 50–60 km region on 1 July. Peak VMRs of around 200 ppbv have been measured in the polar night region between 2500 and 3000 K, representing the highest values ever recorded in the SH and exceeding those measured by the LIMS instrument in the NH winter 1978/1979. However, the lack of NOx measurements covering the polar winter regions at latitudes >50° in previous SH winters makes it difficult to evaluate the measured enhancements in 2003 in the interannual context.
 No temporal correlation to local production mechanisms below 75 km by SPEs or highly energetic electrons has been found. Hence, we attribute the observed NOx enhancements to production by a lower thermospheric electron source and subsequent descent with the meridional circulation. An indication for enhanced electron precipitation in 2003 is given by a seasonal average Ap index of 23, a value which is larger than the ten years average (1992–2002) by a factor of 2. Within the last 15 years, only in 1991 a higher seasonal Ap index has been measured.
 In the course of the Antarctic winter 2003, the NOx enhancements inside the polar vortex descended from the mesosphere into the middle and lower stratosphere. No outflow of NOx from the vortex was observed until August, confirming the assumption of Siskind  that the NOx descent is confined to the polar vortex. By the end of July, upper atmospheric air with average NOx abundances of 15 ppbv reached the 800–1000 K region (i.e., 30–35 km) where it remained until the final warming in late October. However, above 1500 K, the polar vortex began to weaken already in August and significant mixing of tropical and vortex air masses took place. As a consequence, NOx above 1500 K was considerably depleted.
 In July, good agreement of the MIPAS NOx observations from 2003 with model calculations for a typical Antarctic winter and HALOE observations averaged over 1992–1996 [Siskind, 2000] was found. In contrast to HALOE observations, our observations allow for comparison to modeled data in the whole latitude range covering the polar regions. The agreement of model results and observational data in July, in particular with respect to the latitudinal extension of the descending NOx as far equatorward as 30–40°, confirms the hypothesis of Siskind et al.  that isentropic mixing inside of the vortex in the lower mesosphere takes place. Evidence for enhanced horizontal mixing is also found in the spatial distributions of NOx and CO inside the vortex on the isentropic surface of 2500 K in June and July.
 The total amount of NOx injected into stratosphere without being photochemically destroyed has been estimated to be 2.4 GM. 1.1 GM of this amount remained in the polar vortex until the final warming in mid-October, while 1.3 GM were transported out of the vortex by isentropic mixing before. The amount of NOx remaining inside of the SH vortex in late spring had been previously estimated from HALOE data to 0.8–1.3 GM [Siskind et al., 2000] which is in good agreement with our results. The portion of NOx transported out of the vortex represents a new source of NOx to the stratosphere which thus doubles the previous estimates of the stratospheric NOx deposition derived from HALOE data. Upper atmospheric NOx contributes then to the total NOy source in the stratosphere with 9% of the dominant N2O oxidation source.
 No evidence for an unusual dynamical situation was found in the 2003 SH winter from the derived descent rates and from the temporal evolution and extension of the polar vortex. Taking into account that the dynamical conditions in the SH are generally less variable than in the NH, the NOx descent in the Austral winter 2003 might be representative for the SH from the dynamical point of view. If so, the polar winter descent of upper atmospheric NOx can have a significant impact on the stratospheric NOy budget on a longer timescale and might explain the underestimation of NOx at its VMR peak height by chemical transport models [TOPOZ III, 2005]. Since electron precipitation from the outer trapping region and auroral events as upper atmospheric NOx production mechanisms are more frequent in the declining phase of the solar cycle [Rangarajan and Barreto, 2000], a solar cycle modulation of the upper atmospheric NOx source with a phase lag of 2–3 years is expected. Because of the dynamical coupling of the upper atmosphere and the stratosphere, this modulation will affect stratospheric NOy and O3. Model calculations performed by Callis et al.  indicate that the amplitude of the solar cycle modulation by electron precipitation of the global ozone column from 25 to 45 km could be in the order of 2.5%, nearly twice as much as the amplitude due to UV effects.
 NOx and tracer distributions observed during several days in the NH winter 2002/2003 have also been analyzed. We found that high planetary wave activity resulting in the major midwinter warming led to a rather inefficient NOx downward transport with negligible deposition of NOx in the lower and middle stratosphere. Since major warming events in midwinter are a typical feature of the Northern Hemisphere, an insignificant NOx deposition due to unfavorable dynamical conditions, as observed in the 2002/2003 winter, is expected to happen frequently. On the other hand, dynamical conditions in the NH winters are much more variable, resulting thus in a higher variability of the NOx deposition. The MIPAS observations of NOx in the NH winter 2003/2004 [López-Puertas et al., 2005] with record values of around 350 ppbv detected in the strong polar vortex during February and March, demonstrate how drastically dynamical conditions and NOx deposition can vary within two successive years in the NH. Although it might not be excluded that the October/November 2003 SPEs have contributed to the extraordinary NOx enhancements detected in the upper stratosphere in early 2004, the bulk of descended NOx was attributed to electron precipitation similar as in the Antarctic winter 2003 [López-Puertas et al., 2005]. The net deposition of NOx in the NH stratosphere in early 2004 is a topic of future work.
 The authors acknowledge ESA for providing MIPAS spectra and L2 data, as well as NILU and ECMWF for meteorological data. The IAA team has been supported by Spanish Ministerio de Educación y Ciencia under projects REN2001-3249/CLI and ESP2004–01556 and EC FEDER funds. The IMK team was supported by SACADA (BMBF 07ATF53) and by the EU-Project TOPOZ-III (EVK2-CT-2001-00102).