References

N 2 O abundances ranging from 0.5 to 6 ppbv were observed in the polar upper stratosphere/lower mesosphere by the MIPAS instrument on the Envisat satellite during the Arctic and Antarctic winters in the period July 2002 to March 2004. A detailed study of the observed N 2 O-CH 4 correlations shows that such enhancements cannot be explained by dynamics without invoking an upper atmospheric chemical source of N 2 O. The N 2 O enhancements observed at 58 km occurred in the presence of NO x intrusions from the upper atmosphere which were related to energetic particle precipitation. Further, the inter-annual variability of mesospheric N 2 O correlates well with observed precipitating electron fluxes. The analysis of possible chemical production mechanisms shows that the major part of the observed N 2 O enhancements is most likely generated under dark conditions by the reaction of NO 2 with atomic nitrogen at altitudes around 70–75 km in the presence of energetic particle precipitation (EPP). A possible additional source of N 2 O in the middle and upper polar atmosphere is the reaction of N 2 (A 3 Σ u + ), generated by precipitating electrons, with O 2 , which would lead to N 2 O production peaking at altitudes around 90–100 km. N 2 O produced by the latter mechanism could then descend to the mesosphere and upper stratosphere during polar winter. The estimated fraction of EPP-generated N 2 O to the total stratospheric N 2 O inside the polar vortex above 20 km (30 km) never exceeds 1% (10%) during the 2002–2004 winters. Compared to the global amount of stratospheric N 2 O, the EPP-generated contribution is negligible.


Introduction
Nitrous oxide is the main precursor of odd nitrogen in the middle atmosphere.Its major sources, both natural and manmade, are located at the surface from where it is transported into the stratosphere.Photolysis by solar UV is its major sink and the reaction with O( 1 D) leads to the formation of NO.
Due to its long chemical lifetime and the apparent absence of sources in the middle atmosphere, N 2 O is an excellent tracer for stratospheric transport processes.In particular, correlations of observed N 2 O abundances with other tracers such as CH 4 or CFCs have been widely used in transport studies (Michelsen et al., 1998a;Plumb et al., 2000;Ray et al., 2002).
Recently, however, Funke et al. (2008) reported polar stratospheric and mesospheric N 2 O enhancements in the Northern Hemisphere (NH) in the aftermath of the large solar proton event (SPE) which took place in October-November 2003.These N 2 O enhancements were attributed to chemical production of N 2 O by NO 2 + N( 4 S) → N 2 O + O. (R1) Both species (NO 2 and N( 4 S)) were largely enhanced as a consequence of ionization caused by solar protons.Semeniuk et al. (2008) reported polar mesospheric N 2 O enhancements in the NH observed by the Fourier Transform Spectrometer on SCISAT-1 during February-April 2004 which were also attributed to the reaction of NO 2 and atomic nitrogen.Since no major SPE occurred in this period, these authors assumed that enhanced N abundances, required for the formation of N 2 O, were generated by ionization caused by auroral electron precipitation.
An alternative production mechanism of middle and upper atmospheric N 2 O was proposed by Zipf and Prasad (1980): i.e., metastable N 2 (A 3 + u ) is produced by electron impact during auroral substorms and reacts with O 2 to form N 2 O: Published by Copernicus Publications on behalf of the European Geosciences Union.In accordance with laboratory measurements (Zipf, 1980), these authors considered an efficiency of 0.6 to form N 2 O by the reaction of N 2 (A 3 + u ) with molecular oxygen, which would then result in production of enormous amounts of N 2 O around 90-100 km during geomagnetic perturbations.However, the observation of such a high efficiency for producing N 2 O by Reaction (R2), initially reported by Zipf (1980), has not been confirmed by other groups, and the branching ratio for this channel is probably less than 0.02 (de Sousa et al., 1985;Iannuzzi et al., 1982).
Here, we report upper stratospheric and mesospheric N 2 O enhancements observed by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) instrument on Envisat in polar winters during 2002-2004.In Sect.2, we present the observed spatial distributions of polar winter N 2 O and their temporal evolution.The chemical origin of these N 2 O enhancements is demonstrated in Sect. 3 by means of an analysis of N 2 O-CH 4 tracer-tracer correlations, and possible production mechanisms are discussed in Sect. 4.

MIPAS observations and data analysis
MIPAS is a limb emission Fourier transform spectrometer designed for the measurement of trace species from space (Fischer and Oelhaf, 1996;European Space Agency, 2000;Fischer et al., 2008).It is part of the instrumentation of the Environmental Satellite (ENVISAT) which was launched into its sun-synchronous polar orbit of 98.55 • N inclination at about 800 km altitude on 1 March 2002.ENVISAT passes the equator in southerly direction at 10.00 a.m.local time 14.3 times a day.MIPAS operated from July 2002 to March 2004 at full spectral resolution of 0.035 cm −1 (unapodized) in terms of full width at half maximum and has resumed operation with reduced resolution, after an instrument failure, since August 2004.MIPAS observes the atmosphere during day and night with global coverage from pole to pole.Within its standard observation mode at full spectral resolution, MI-PAS covers the altitude range from nominally 68 km down to 6 km with tangent altitudes at 68, 60, 52, 47, and then at 3 km steps from 42 to 6 km.Occasionally, MIPAS also operates in several upper atmospheric measurement modes scanning up to 170 km.The field of view of MIPAS is 30 km in the horizontal and approximately 3 km in the vertical.During each orbit up to 72 limb scans are recorded.The Level-1b processing of the data (version 4.61/62 was used here), 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(Nett et al., , 2002)).
2.1 Analysis of IMK/IAA-generated N 2 O data Data presented and discussed in this section are vertical profiles of abundances of N 2 O and CH 4 retrieved with the scientific IMK-IAA data processor (von Clarmann et al., 2003a) developed and operated by the Institute of Meteorology and Climate Research (IMK) in Karlsruhe together with the Instituto de Astrofísica de Andalucía (IAA) in Granada.This data processor is based on a constrained non-linear least squares algorithm with Levenberg-Marquardt damping and line by line radiative transfer calculations with the Karlsruhe Optimized and Precise Radiative Transfer Algorithm (KOPRA) (Stiller et al., 2002).The first step in the L2 processing is the determination of the spectral shift, followed by the retrieval of temperature and elevation pointing (von Clarmann et al., 2003b), where pressure is implicitly determined by means of hydrostatic equilibrium.The retrieval of volume mixing ratio (vmr) profiles of species is carried out in the following order: O 3 , H 2 O, HNO 3 , and then CH 4 and N 2 O simultaneously.The results of the species firstly retrieved are used in the retrievals of the subsequent species.The N 2 O vmr is retrieved from the MIPAS spectra around 1284.9 cm −1 , where the v 1 band of N 2 O is located (Glatthor et al., 2005).The retrievals are performed from selected spectral regions (micro-windows) which vary with tangent altitudes in order to optimize computation time and minimize systematic errors (Echle et al., 2000).Thus, height dependent combinations of micro-windows were selected with a trade-off between computation time and total retrieval error.The retrieval noise error in the N 2 O vmr is typically 3% at 10-44 km and 22% at 50 km.The total error varies between 10 and 20% at 10-35 km and is about 30% between 35-50 km (Glatthor et al., 2005).The resulting vertical resolution was about 4 km in the altitude range 15-40 km and decreased to more than 10 km below and above this region.More details on the N 2 O retrieval strategy can be found in Glatthor et al. (2005).
Two data sets with different retrieval versions are used in this study: The first one includes N 2 O and CH 4 data versions V3O N2O 8.0 and V3O CH4 8.0 (in the following referred as V3O 8.0) and covers 45 days between 13 September 2002 and 21 October 2003.The second one (V3O N2O 12.0 and V3O CH4 12.0, in the following referred as V3O 12.0) was derived with an updated retrieval version and includes 54 days between 9 September 2003 and 25 March 2004.The retrieval updates applied to the latter data set include a weaker regularization of N 2 O at altitudes above 45 km in order to achieve a more realistic shape of the retrieved profiles in the presence of extraordinary mesospheric N 2 O enhancements such as those found in the aftermath of the 2003 "Halloween" SPE and the Arctic winter 2004.It should be noted, however, that above 60 km, retrieved profiles of enhanced N 2 O tend to be low-biased in both data sets due to the regularization.In consequence, the peak altitudes of the derived mesospheric N 2 O enhancements could appear lower than in reality.
In order to analyze the retrieved trace gas profiles in a dynamical context, potential vorticity data from the European Centre for Medium-Range Weather Forecasts (ECMWF) analysis has been used for the representation of N 2 O and CH 4 data in equivalent latitudes as described by Nash et al. (1996).Figure 1 shows potential temperature-equivalent latitude daily mean cross sections of N 2 O and CH 4 vmr on representative days in the NH winter 2002/2003, SH winter 2003, and NH winter 2003/2004. Mesospheric   opposite sign, indicating that the polar vortex boundary was located at 60 • N equivalent latitudes.Indeed, an unusual strong vortex together with a very fast and efficient descent of mesospheric air has been observed in this particular Arctic winter (Manney et al., 2005).N 2 O enhancements observed during the other two polar winters do not show such pronounced gradients and extend towards low latitudes, reaching background values around 30 • .Thus, it seems that in the 2002-2003 winters stronger mixing of polar and tropical air occurred through a weaker polar vortex boundary.
Figure 2 shows the temporal evolution of mean N 2 O and CH 4 abundances (Version V3O 8.0) within 70-90 • S equivalent latitudes in the period September 2002 to November 2003.In September 2002, characterized by a major warming which led to a split of the stratospheric polar vortex (e.g., Newman and Nash, 2005), mesospheric N 2 O abundances around 0.5 ppbv were observed.Given the unusual dynamical situation which allowed for intrusions of tropical air masses into the polar region, it is not clear whether these small N 2 O enhancements are of chemical or dynamical origin.The apparent descent visible in the temporal evolution of N 2 O and CH 4 abundances during the following polar summer (i.e.December to April) is originated by photochemical losses of both species involving photolysis and reactions with OH and O( 1 D).Later on, a tongue of enhanced CH 4 can be seen, localized around 60 km in March and descending to 30 km in July/August.These CH 4 enhancements were generated by an accelerated Brewer-Dobson circulation with strong poleward transport of tropical air masses rich in CH 4 preceding the downward motion during polar winter.The temporal evolution of N 2 O shows a similar behavior, though transport-generated N 2 O enhancements are less pronounced at altitudes above 50 km compared to the corresponding CH 4 distributions.During June-August 2003, three N 2 O peaks reaching values higher than 1 ppbv show up at altitudes around 55-65 km while CH 4 vmrs were generally low due to polar winter descent.As already seen in the N 2 O zonal mean distributions on 8 June, these N 2 O peaks are most likely related to chemical production.From August until the final breakup of the polar vortex in October, CH 4 abundances started to increase above 40 km due to stronger mixing across a weakened vortex boundary (Funke et al., 2005).These intrusions of tropical air into the polar regions makes it more difficult to distinguish whether mesospheric N 2 O enhancements in August and September are related to chemistry or to transport.
The version 4.61/4.62).This further allows to corroborate the evidence for chemically-produced mesospheric N 2 O enhancements found in the IMK/IAA data with data generated with an independent retrieval algorithm and, hence to exclude retrieval artifacts as a possible explanation.ESA data are retrieved with the operational retrieval algorithm as described by Raspollini et al. (2006)  ation value ε of all truncated negative values to be zonally averaged can be determined by x 0 is the "true" zonal mean value and σ is the standard deviation assumed to be identical to ally averaged noise error.After determination of ε by resolving numerically the integral in the corrected zonal mean can be calculated by substitution of all truncated profile points by ε bsequent averaging.Since the "true" zonal mean value is unknown at the beginning, we start e uncorrected value and iterate until x 0 changes by less than 2%.This convergence criterion ually fulfilled after a few iterations.since negative vmrs are not supported by this retrieval algorithm.Whenever negative vmrs occur at a given altitude during an iteration of the retrieval, the corresponding vmr profile points are set arbitrarily to a value of 10 −10 ppmv.The rationale of this procedure is to avoid numerical retrieval instabilities.This truncation of negative values, however, results in a positive bias of ESA vmr mean values, if the "true" mean value is close to the noise error or below.In order to correct for this bias, the following procedure was applied: Assuming zonal homogeneity at a given altitude and a Gaussian distribution of N 2 O measurements affected by random errors, the expectation value ε of all truncated negative values to be zonally averaged can be determined by where x 0 is the "true" zonal mean value and σ is the standard deviation assumed to be identical to the zonally averaged noise error.After determination of ε by resolving numeri-cally the integral in Eq. ( 1), the corrected zonal mean can be calculated by substitution of all truncated profile points by ε and subsequent averaging.Since the "true" zonal mean value is unknown at the beginning, we start with the uncorrected value and iterate until x 0 changes by less than 2%.This convergence criterion was usually fulfilled after a few iterations.
Figure 5 shows the temporal evolution of ESA N 2 O vmr averaged within 70-90 • geographical latitudes at 58 km before and after the statistical correction, together with IMK/IAA-retrieved N 2 O averages and mean single measurement noise errors.Excellent agreement between the corrected averaged ESA and IMK/IAA data is found, while the uncorrected ESA data shows a positive bias up to 1.2 ppbv in dependence of the noise error.We thus conclude that the statistical correction applied to the zonally averaged ESA N 2 O works satisfactorily.It can further be excluded that the N 2 O enhancements discussed in the previous section are generated by a retrieval artifact in any of the data sets, given that they appear with good quantitative agreement in two independently retrieved data sets.

Analysis of N 2 O-CH 4 correlations
In the tropics and middle latitudes, N 2 O and CH 4 show compact correlations, only varying little between seasons and years (Michelsen et al., 1998a,b).N 2 O vmrs generally increase monotonically with increasing CH 4 with a non-linear relationship.In polar spring, this relationship tends to be more linear due to mixing of subsided vortex air and midlatitude air masses.Without mesospheric sources, N 2 O should increase with CH 4 in any case, while the presence of such source should invert this dependence for low CH 4 vmrs.Therefore, N 2 O-CH 4 correlations represent an excellent tool for the detection of mesospheric N 2 O sources.
We have analyzed N 2 O-CH 4 correlations determined from ESA data on a monthly basis, using all available observations between 45 km and 60 km within 60-90 • S and 60-90 • N. We first performed a histogramming of the observations within bins of CH 4 =10 ppbv and N 2 O=0.1 ppbv.Median values of the obtained N 2 O probability density function (PDF) at a given CH 4 level were then determined in a second step.
Figure 6  A further anomaly is found during November 2003 in the NH, when N 2 O was significantly increased after the "Halloween" SPE.In contrast to the perturbed polar winter correlations discussed above, however, enhanced N 2 O is found at nearly all CH 4 levels.This is expected, since solar protons led to in situ production of N 2 O at altitudes above 45 km.The fact that polar winter episodes of enhanced N 2 O are characterized by inverted N 2 O-CH 4 correlations at CH 4 levels smaller than 0.5 ppbv hints thus at descent of air masses enriched in N 2 O from higher altitudes rather than in situ production.
Increased median values of N 2 O PDFs at CH 4 levels higher than 0.1 ppmv are not only found in the aftermath of the "Halloween" SPE.The finding that chemical production of N 2 O occurred in the mesosphere during all polar winters in the period 2002-2004, though with variable magnitude, makes it rather unlikely that mesospheric N 2 O enhancements are sporadic singular phenomena.It is further evident that polar N 2 O-CH 4 correlations are significantly perturbed by this mesospheric N 2 O source.

Discussion
In order to understand which are the chemical mechanisms responsible for the observed polar winter N 2 O enhancements and which are the atmospheric parameters affecting their magnitudes and inter-annual variations, we have assessed the temporal evolutions of polar mesospheric N 2 O, CH 4 , and NO 2 at 58 km from the ESA data set, jointly with precipitating electron fluxes observed by the Medium Energy Proton and Electron Detector (MEPED) on NOAA 16 (poes.ngdc.noaa.gov/data/avg)and calculated ion pair production rates due to solar protons (Jackman et al., 2008).
The temporal evolution of these quantities are shown in Figs.7 and 8 for the SH and NH, respectively.In order to isolate the N 2 O amount generated by the mesospheric source from the contribution of N 2 O of tropospheric origin, we have estimated the latter by applying a typical N 2 O-CH 4 correlation to the CH 4 observations.This correlation has been determined from an average of the monthly N 2 O-CH 4 correlations presented above, excluding the polar winter periods (June-September in the SH and November-February in the NH) perturbed by mesospheric N 2 O generation.A dominant contribution of tropospheric N 2 O transported up to 58 km was   A further source of enhanced N or N 2 (A 3 + u ) could be energetic electron precipitation (EEP).Precipitating electron fluxes of different energies observed by the MEPED instrument are shown in Figs.7 and 8 (lower panels).These fluxes have been averaged over a 30-day period in order to reflect that N 2 O losses are very small during polar winter and EEPinduced N 2 O production would therefore tend to accumulate.Fluxes of precipitating electrons of energies 30 keV, 100 keV and 300 keV included in Figs.7 and 8 cause ionization peaks at altitudes around 85, 75, and 60 km, respectively (Callis et al., 1998).Therefore, they can be used as indicator for EEP-induced atmospheric ionization at these altitudes.Since only relative temporal variations of these fluxes are of interest when comparing them to the evolution of polar winter N 2 O enhancements, fluxes of different energies have been scaled arbitrarily for the sake of a better representation.Electron flux increases coincident with SPEs (i.e., 29 May and 29 October 2003) should be interpreted with caution since MEPED electron measurements are compromised by the presence of protons (Evans and Greer, 2000), although SPEs are thought to be associated with elevated electron fluxes inside the polar caps (e.g., Baker et al., 2004).
It becomes evident from Figs. 7 and 8 that strongest N 2 O enhancements at 58 km occur in polar winters with highest electron precipitation, i.e. the Antarctic winter 2003 and the Arctic winter 2004.The correlated inter-annual variations of mesospheric N 2 O and electron fluxes hint at an implication of EEP in the production of N 2 O. On the other hand, no temporal correlation of the N 2 O evolution with the short-term fluctuations in any of the electron fluxes is visible.This suggests that a dominant local contribution of an EEP-related source at altitudes as low as 58 km is rather unlikely.
N 2 O enhancements of mesospheric origin seem to occur always in presence of elevated NO 2 , although the observed NO 2 vmrs vary drastically from winter to winter (see Figs. 7 and 8, upper panels).This could hint at an implication of NO 2 in the chemical production of N 2 O. On the other hand, mesospheric NO 2 enhancements during polar winters are known to be generated by descent of upper atmospheric NO x produced by energetic particle precipitation (Callis et al., 1998;Siskind, 2000;Funke et al., 2005;Randall et al., 2007).Then, the simultaneous occurrence of both species could simply reflect that their sources are located above the observed altitude (58 km) and their descended contributions are modulated by the meridional circulation in the same manner.A common modulation of N 2 O and NO 2 by variable descent velocities and horizontal mix-ing is further supported by the good temporal correlation of small-scale structures in the evolution of both species which, in turn, is anti-correlated with the evolution of CH 4 vmrs.It is also evident from Figs. 7 and 8 that polar winter enhancements of N 2 O are by far not proportional to the available NO 2 amounts.The average N 2 O/NO 2 ratios observed during the periods with strongest N 2 O enhancements, i.e. in June-August 2003 in the SH and January-March 2004 in the NH, were considerably smaller than those observed in periods with weak N 2 O enhancements.Further, mesospheric N 2 O enhancements tended to start earlier and lasted longer than the NO 2 enhancements.
In order to qualitatively assess the relative magnitudes and peak altitudes of the EEP-related N 2 O sources, we have estimated N 2 O production rates due to Reactions (R1) and (R2) for different atmospheric conditions.
N( 4 S) is generated in presence of EEP by a chain of ionization processes from N 2 .It is commonly assumed that each ion pair produced by electron impact leads to the formation of 0.55 N( 4 S) atoms (e.g., Jackman et al., 2005).During night, this mechanism represents the only source of N( 4 S), while at daytime, NO photolysis leads to additional N production.The dominating losses of atomic nitrogen are and Reaction (R1).Assuming atomic nitrogen to be in steady state and an efficiency of Reaction (R1) to form N 2 O of 0.5 (Funke et al., 2008), the N 2 O production rate P 1 due to Reaction (R1) is where p is the ion pair production rate, J NO the photolysis rate of NO, and k 1 , k 3 , and k 4 are the rate coefficients for the atomic nitrogen loss Reactions (R1), (R3), and (R4), respectively.N 2 (A 3 + u ) is also produced by the impact of precipitating electrons.Quenching by atomic oxygen, i.e.
and radiative emissions in the Vegard-Kaplan bands with an Einstein coefficient of 0.52 s −1 are the prominent losses in the thermosphere, while at lower altitudes Reaction (R2) becomes dominant.As reported by Zipf and Prasad (1980), we assume that each ion pair produces 0.35 metastable N 2 (A 3 + u ) molecules.In accordance with the more recent laboratory work of de Sousa et al. (1985), we apply an efficiency of 0.02 to form N 2 in Reaction (R2) as an upper limit.The N 2 O production rate P 2 due to Reaction (R2) is then olume mixing ratios of NO2 and NO during polar night (solid and dotted lines, respectively) and NO2 ay (dashed line) used in the nominal (black) and in the "high NOx" scenario (red).tmospheric air occurred in the NH polar vortex.In agreement with observations of ACE-FTS nd et al., 2006) and MIPAS (Funke et al., 2007), we have assumed in this scenario 100 ppbv , increasing to 4 ppmv at 75 km (20 ppmv at 100 km).The NO x partitioning was taken hole Atmosphere Community Climate Model (WACCM) calculations (Garcia et al., 2007) ary, 90 • N (polar night) and 60 • N at noon (i.e., 80 • SZA).In these simulations, the polar O x partitioning in the mesosphere shows only small inter-and intra-annual variations.Nomperatures were taken from the Mass Spectrometer and Incoherent Scatter Radar empirical (NRLMSISE-00) (Picone et al., 2002) for January, 90 • N. We also included an additional ture scenario for "strong descent" conditions which reflects the thermal structure over the ole during end of January 2004, when temperatures 30 K higher (40 K lower) than MSIS served at 70 km (50 km) by the Sounding of the Atmosphere using Broadband Emission Ray (SABER) instrument (Hauchecorne et al., 2007).Profiles of temperature and vmrs of NO 2 for the different scenarios are shown in Fig. 9. Atomic oxygen densities required for the ion of P 2 have been taken from WACCM calculations for January, 90 • N. Rate constants for ns (R1-4) were taken from Sander et al. (2006), while the quenching rate k 5 of Reaction (R5) en from Zipf and Prasad (1980).
e 10 shows the estimated N 2 O production rates P 1 and P 2 due to Reactions (R1) and (R2) vely, in the presence of EEP.Apart of the nominal scenario, P 1 is also shown for the "high cenario combined with the nominal and the "strong descent" temperature scenarios.Since ends on illumination, both, polar night and sun overhead conditions have been considered.oduction due to Reaction (R2) is independent of temperature and the NO x abundance.Thus, considered the nominal scenario only for the estimation of P 2 .17 and "strong descent" (dotted line) scenario.Right: Volume mixing ratios of NO 2 and NO during polar night (solid and dotted lines, respectively) and NO 2 during day (dashed line) used in the nominal (black) and in the "high NO x " scenario (red).
where k 2 and k 5 are the rate coefficients for the Reactions (R2) and (R5), respectively.
As a representative profile of atmospheric ionization due to EEP, we took the average ion pair formation rates calculated by Callis et al. (1998) from MEPED data taken in the period 1979-1987 (see Fig. 2a of their work).These rates might underestimate the EEP-induced ionization for conditions of elevated geomagnetic activity as encountered during 2003.On the other hand, due to the averaging over nearly an entire solar cycle, they are useful for the estimation of an average N 2 O production to be expected during any polar winter.Calculations of the N 2 O production rates have been performed for different scenarios of polar winter NO x descent.In the nominal scenario, NO x abundances are 20 ppbv at 50 km and increase to 800 ppbv at 75 km (10 ppmv at 100 km).The "high NO x " scenario reflects the NO x abundances observed in February 2004 when the unusual descent of upper atmospheric air occurred in the NH polar vortex.In agreement with observations of ACE-FTS (Rinsland et al., 2006) and MIPAS (Funke et al., 2007), we have assumed in this scenario 100 ppbv at 50 km, increasing to 4 ppmv at 75 km (20 ppmv at 100 km).The NO x partitioning was taken from Whole Atmosphere Community Climate Model (WACCM) calculations (Garcia et al., 2007) for January, 90 • N (polar night) and 60 • N at noon (i.e., 80 • SZA).In these simulations, the polar night NO x partitioning in the mesosphere shows only small inter-and intra-annual variations.Nominal temperatures were taken from the Mass Spectrometer and Incoherent Scatter Radar empirical model (NRLMSISE-00) (Picone et al., 2002) for January, 90 • N. We also included an additional temperature scenario for "strong descent" conditions which reflects the thermal structure over the North pole during end of January 2004, when temper- The peak nighttime production rate P 1 due to Reaction (R1), located around 73 km, is order of 15 cm −3 s −1 (see Fig. 10).The daytime contribution of this reaction has its max around 55 km with a N 2 O production rate considerably smaller than the corresponding nig contribution.This is expected, since daytime NO 2 abundances are very small due to NO 2 p ysis.Daytime N concentrations are driven by NO photolysis rather than by electron impact, 380 induces a pronounced dependence of P 1 on the NO x availability during day (compare dash solid red line in Fig. 10).This dependence on NO x is, however, negligible at the P 1 peak h around 73 km, during night.The reason is that for the atmospheric conditions there, atomic ni losses are dominated by Reactions (R1) and (R3), both involving NO x .In consequence, th production is mainly driven by the ionization rate and the NO/NO 2 partitioning, there.The 385 however, is expected to introduce only a small variability of P 1 .At altitudes below the peak h the nighttime production rate P 1 depends more strongly on the NO x vmr, since atomic ni losses due to Reaction (R4) are getting more important towards lower altitudes.An 800% in of NOx in the "high NO x " scenario compared to the nominal scenario results in a 50% incre the N 2 O production at 60 km.

390
Temperature also affects the N 2 O production by Reaction (R1).This is related to the temperature dependence of the atomic nitrogen losses by Reaction (R4), leading to a loss a peratures around 255 K faster by one order of magnitude compared to temperatures around When considering the "strong descent" temperature scenario, reflecting the conditions duri Arctic mid-winter 2004, the P 1 profiles are significantly increased below 63 km and sligh atures 30 K higher (40 K lower) than MSIS were observed at 70 km (50 km) by the Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) instrument (Hauchecorne et al., 2007).Profiles of temperature and vmrs of NO and NO 2 for the different scenarios are shown in Fig. 9. Atomic oxygen densities required for the calculation of P 2 have been taken from WACCM calculations for January, 90 • N. Rate constants for Reactions (R1-4) were taken from Sander et al. (2006), while the quenching rate k 5 of Reaction (R5) was taken from Zipf and Prasad (1980).
Figure 10 shows the estimated N 2 O production rates P 1 and P 2 due to Reactions (R1) and (R2) respectively, in the presence of EEP.Apart of the nominal scenario, P 1 is also shown for the "high NO x " scenario combined with the nominal and the "strong descent" temperature scenarios.Since P 1 depends on illumination, both, polar night and sun overhead conditions have been considered.N 2 O production due to Reaction (R2) is independent of temperature and the NO x abundance.Thus, we have considered the nominal scenario only for the estimation of P 2 .
The peak nighttime production rate P 1 due to Reaction (R1), located around 73 km, is on the order of 15 cm −3 s −1 (see Fig. 10).The daytime contribution of this reaction has its maximum around 55 km with a N 2 O production rate considerably smaller than the corresponding nighttime contribution.This is expected, since daytime NO 2 abundances are very small due to NO 2 photolysis.Daytime N concentrations are driven by NO photolysis rather than by electron impact, which induces a pronounced dependence of P 1 on the NO x availability during day (compare dashed and solid red line in Fig. 10).This dependence on NO x is, how-ever, negligible at the P 1 peak height, around 73 km, during night.The reason is that for the atmospheric conditions there, atomic nitrogen losses are dominated by Reactions (R1) and (R3), both involving NO x .In consequence, the N 2 O production is mainly driven by the ionization rate and the NO/NO 2 partitioning, there.The latter, however, is expected to introduce only a small variability of P 1 .At altitudes below the peak height, the nighttime production rate P 1 depends more strongly on the NO x vmr, since atomic nitrogen losses due to Reaction (R4) are getting more important towards lower altitudes.An 800% increase of NO x in the "high NO x " scenario compared to the nominal scenario results in a 50% increase in the N 2 O production at 60 km.
Temperature also affects the N 2 O production by Reaction (R1).This is related to the strong temperature dependence of the atomic nitrogen losses by Reaction (R4), leading to a loss at temperatures around 255 K faster by one order of magnitude compared to temperatures around 220 K.When considering the "strong descent" temperature scenario, reflecting the conditions during the Arctic mid-winter 2004, the P 1 profiles are significantly increased below 63 km and slightly decreased above, compared to the nominal temperature scenario (compare dotted and dashed lines in Fig. 10).
For the interpretation of the observed N 2 O enhancements in terms of possible production by Reaction (R1), it is necessary to assess the dependence of P 1 during night on both temperature and NO x availability in more detail.Therefore, we have performed calculations of P 1 for various temperatures between 220 and 270 K and NOx vmrs up to 4 ppmv (not shown).It turned out that for typical polar winter temperatures of 220 K at 70 km, P 1 is rather proportional to NO x up to approximately 20 ppbv, while for NO x abundances larger than 100 ppbv the impact of the NOx abundance on P 1 is weak.NO x abundances of more than 100 ppbv are generally found above 70 km during typical polar winters.For perturbed temperatures of 255 K, the linear dependence on NO x is extended up to 150 ppbv.The region of weak NO x dependence starts at NO x vmrs around 500 ppbv.Since such high temperatures are generally found in winters with strong subsidence and hence high NO x availability, P 1 is expected to be rather independent on NO x at its peak height also in presence of high temperatures.
The maximum of the production rate P 2 is located at 95-100 km with a magnitude of 25 cm −3 s −1 and decreasing to values smaller than 1 cm −3 s −1 below 75 km.P 2 is mainly driven by the EEP ionization rate, though above 90 km, increasing atomic oxygen densities make P 2 dependent on the O/O 2 partitioning.At the P 2 peak height, around half of the EEP-generated N 2 (A 3 + u ) molecules are quenched by atomic oxygen before they can react with O 2 to form N 2 O.
Despite this dependence on the O/O 2 partitioning, N 2 O production by Reaction (R2) shows similarities with the EEP-induced NO production in terms of both source region and variability.The magnitude of the NO production, however, is higher than our estimates of the N 2 O production by Reaction (R2) by approximately a factor of 200, assuming that 0.7 NO molecules are produced by each ion pair (Jackman et al., 2005).In consequence, we would expect that during polar winter, when photochemical losses are small, the N 2 O produced by Reaction (R2) is about 0.5% of the NO x abundance in air masses descended from the upper atmosphere.
During mid winter, the average polar NO 2 abundances at 58 km, as shown in Figs.7 and 8, represent a good proxy of total NO x .During the 2002 Antarctic and 2002/2003 Arctic winters, observed NO 2 abundances at 58 km were generally lower than 15 ppbv.Hence, a negligible contribution by Reaction (R2) to the observed N 2 O enhancements of less than 0.07 ppbv would be expected at this altitude.However, during the Antarctic winter 2003, when NO 2 abundances of more than 100 ppbv were measured, this contribution could have made up to 0.5 ppbv, which is about 40% of the N 2 O enhancements observed in this period.During January-March 2004 in the Arctic, the contribution due to Reaction (R2) could have even been as high as 1.5 ppbv which makes up 25% of the observed N 2 O enhancements.We recall, however, that these estimates represent an upper limit for a possible contribution of Reaction (R2).
In the lower thermosphere, N 2 O abundances on the order of several 100 ppbv would then be expected due to Reaction (R2).Unfortunately, there are no thermospheric N 2 O measurements available which would support such high N 2 O amounts.We thus have looked at MIPAS spectra taken in the upper atmospheric observation mode on 11 June 2003, which includes tangent heights up to 170 km.No evidence for N 2 O emissions at tangent heights above 90 km was found in any of the N 2 O bands included in the MIPAS spectra.However, due to the low signal-to-noise ratio at these altitudes, only N 2 O abundances higher than approximately 200 ppbv could have been detected by MIPAS at these tangent heights.
In this sense, a possible contribution of Reaction (R2) can neither be proven nor excluded.In any case, even when assuming a maximum efficiency of this reaction to form N 2 O, its contribution could only explain a small fraction of the observed N 2 O enhancements.
Assuming a continuous EEP-induced N 2 O production with typical rates as provided by our estimates of P 1 and a transport-limited N 2 O lifetime of a few weeks, Reaction (R1) would lead to N 2 O abundances on the order of several ppbv at its source region around 70 km.Since our estimations reflect conditions of average geomagnetic activity, even higher mesospheric N 2 O abundances would be found in periods of elevated geomagnetic activity, e.g. during the Antarctic winter 2003 and Arctic winter 2004.Although N 2 O abundances are expected to decrease towards lower altitudes due to dilution during the descent, it seems plausible that N 2 O abundances such as observed at 58 km during the 2002-2004 polar winters could have been caused by Reaction (R1).It has already been demonstrated by CMAM model calculations (Semeniuk et al., 2008) R1) is mainly driven by the EEP-induced ionization rate around 73 km.Hence, except for dynamical modulations, the inter-annual variations of the observed polar winter N 2 O enhancements should be correlated to the flux variations of precipitating electrons with energies greater than 100 keV.Indeed, we found the highest N 2 O amounts in the 2003 SH and 2004 NH winters with most elevated fluxes of >100 keV electrons.However, it is striking that the N 2 O enhancements observed during the 2003 SH winter at 58 km were about 5 times smaller than those of the 2004 NH winter, although the observed >100 keV electron fluxes were comparable.This apparent asymmetry could be explained by the unusually strong polar vortex during the 2004 NH winter leading to a more efficient descent of N 2 O produced around 73 km than in the 2003 SH winter.Further, favored by the low temperatures and high NO x availability, significant N 2 O production could have occurred at altitudes below 65 km during the 2004 NH winter.
We have further observed that the N 2 O/NO 2 ratio tended to be considerably higher at the beginning and the end of the polar winters.As discussed above, the non-linear dependence of the nighttime fraction of P 1 on the NO x availability could be responsible for variations of the N 2 O/NO 2 ratio: At the beginning and the end of the winter, when only small amounts of NO x are available, N 2 O increases linearly with NO 2 , while during mid winter, sufficient NO x is available to make the N 2 O production dependent on the EEP-induced ionization and the NO x partitioning, only.
In summary, we have shown that the magnitude and temporal evolution of the observed N 2 O enhancements during the 2002-2004 polar winters are in concordance with a dominant nighttime N 2 O source around 73 km due to Reaction (R1) which is driven by energetic electron precipitation.The variability of N 2 O at 58 km over various winters seems to be driven by both dynamical conditions and variations of the source strength.Further, we cannot exclude that Reaction (R2) contributed additionally by up to 25-40% to the observed enhancements.

Conclusions
We have presented observations of enhanced N 2 O abundances, ranging from 0.5 to 6 ppbv in the polar upper stratosphere/lower mesosphere, which have been taken by the MI-PAS instrument on the Envisat satellite during the Arctic and Antarctic winters in the period July 2002-March 2004.These N 2 O enhancements have been found in two data sets resulting from independent processing, the first, generated at IMK/IAA, and the second, generated by the operational data processing performed by ESA.The good agreement of both data sets makes is unlikely that the observed N 2 O enhancements are related to retrieval artifacts.
Simultaneous N 2 O and CH 4 observations show a pronounced anti-correlation during polar winters at CH 4 levels lower than 0.1 ppmv.This behavior gives clear evidence that the N 2 O enhancements are of chemical rather than dynamical origin.As a consequence, polar winter N 2 O-CH 4 correlations should be used with caution in tracer-tracer studies due to the perturbations by this mesospheric N 2 O source.
The finding that chemical production of N 2 O occurred in the mesosphere during all polar winters in the period 2002-2004, though with variable magnitude, makes it rather unlikely that mesospheric N 2 O enhancements are infrequent isolated phenomena.
The polar winter N 2 O enhancements observed at 58 km occurred in the presence of NO x intrusions from the upper atmosphere which were related to energetic particle precipitation.Further, the inter-annual variations of polar winter averages of mesospheric N 2 O correlate well with those of precipitating electrons fluxes as measured by the MEPED instrument, which hints at an EEP-related N 2 O source.On the other hand, we found a pronounced anti-correlation of the temporal evolutions of N 2 O and CH 4 at 58 km, which hints at a dynamical modulation of descending N 2 O from a source region at higher altitudes.
The analysis of possible chemical production mechanisms shows that the major part of the observed N 2 O enhancements is most likely generated under dark conditions by the reaction of NO 2 with atomic nitrogen in the presence of energetic particle precipitation.N 2 O production due to this mechanism has its maximum around 73 km.The polar winter N 2 O abundances observed at 58 km seem to be modulated by both variations of the source strength and dynamical factors driving the efficiency of the descent from the source region.An additional source of N 2 O in the middle and upper polar atmosphere could represent the reaction of N 2 (A 3 + u ), generated by precipitating electrons, with O 2 , which would lead to N 2 O production peaking at altitudes around 90-100 km.N 2 O produced by the latter mechanism could have then descended to the upper stratosphere and mesosphere during the 2002-2004 polar winters, where it could have contributed to the observed N 2 O enhancements by up to 25-40%.
EEP-generated mesospheric N 2 O represents a continuous, though variable, source of stratospheric odd nitrogen during polar winters.This source, however, is of minor importance when compared to EEP-induced NO production.
The total fraction of stratospheric N 2 O produced in the upper atmosphere is difficult to assess directly from MI-PAS observations since the EEP-generated N2O which has descended into the middle or lower stratosphere during the winter can hardly be separated from the much higher background N 2 O abundances.However, it can be estimated indirectly from the deposition of EEP-generated NO y into the stratosphere (Funke et al., 2005), assuming a constant ratio of N 2 O and NO y of upper atmospheric origin below 50 km.This is justified since at these altitudes none of N 2 O and NO y is produced in situ and none of them show significant photochemical losses during polar winter.The estimated fraction of EEP-generated N 2 O to the total stratospheric N 2 O inside the polar vortex above 20 km (30 km) never exceed 1% (10%) during the 2002-2004 winters.Compared to the global amount of stratospheric N 2 O, the EEP-generated contribution is negligible.

Fig. 1 .Fig. 1 .
Fig. 1.Potential temperature-equivalent latitude daily mean cross sections of N2O (left) and CH4 (right) vmr retrieved by IMK/IAA for the NH for 8 December 2002 (top), SH for 8 June 2003 (middle), and NH for 9 February 2004 (bottom).Data versions used are V3O 8.0 (top and middle) and V3O 12.0 (bottom).Mean geometric heights are indicated by dotted white lines.Note the different scale in the bottom left panel.

Fig. 2 .
Fig. 2. Temporal evolution of N 2 O (top) and CH 4 (bottom) VMR retrieved by IMK/IAA (Versions V3O 8.0) averaged over equivalent latitudes 70-90 • S from September 2002 until October 2003.An area-weighting factor (cosine of equivalent latitude) has been applied.Mean potential temperatures are indicated by dotted white lines.
shows the obtained temporal evolution of the N 2 O-CH 4 correlation for 60-90 • S and 60-90 • N. Deviations from the typical correlation, characterized by monotonically increasing N 2 O with CH 4 , are clearly visible in all four polar winters at CH 4 levels below 0.1 ppmv, providing an additional proof for mesospheric N 2 O sources.Maximum median values of the N 2 O PDFs of 2.2 and 4 ppbv are found at the lowest CH 4 levels during June 2003 in the SH and January 2004 in the NH, respectively.During these months, a pronounced anti-correlation of N 2 O and CH 4 below 0.1 ppmv was observed.During the Antarctic winter 2002 and Arctic winter 2002/2003, deviations from the typical correlation were less pronounced.In these winters, maximum median values of the N 2 O PDFs of 1 ppbv are found at CH 4 levels of 0.03-0.05ppmv.
Tongues of high N 2 O are visible during July 2002, June 2003, and September 2003 in the SH, as well as February 2003 and January 2004 in the NH.Such enhancements hint at mixing of descended N 2 O-rich and CH 4poor mesospheric air masses with ambient air masses from lower latitudes.

Fig. 6 .Fig. 7 .Fig. 7 .Fig. 8 .
Fig. 6.Monthly mean median values of N 2 O PDFs at given CH 4 levels calculated from ESA 4.61/4.62data products for 60-90 • S (top) and 60-90 • N (bottom).White regions reflect an availability of less then 30 observations at the corresponding CH 4 level.Note the different color scales for SH and NH!
temperatures used in the nominal (solid line) and "strong descent" (dotted line) scenario.

Fig. 9 .
Fig.9.Left: Kinetic temperatures used in the (solid line) and "strong descent" (dotted line) scenario.Right: Volume mixing ratios of NO 2 and NO during polar night (solid and dotted lines, respectively) and NO 2 during day (dashed line) used in the nominal (black) and in the "high NO x " scenario (red).

Fig. 10 .
Fig. 10.Estimated EEP-related N2O productions by Reaction (R1) at polar night (solid black line) a (SZA=80 • ) conditions (solid red line), as well as by Reaction (R2) (solid blue line) for the nominal sc Dashed and dotted lines represent N2O productions for the "high NOx" scenario combined with the n and the "strong descent" temperature scenarios, respectively (only Reaction R1).See text for further de

395Fig. 10 .
Fig.10.Estimated EEP-related N 2 O productions by Reaction (R1) at polar night (solid black line) and day (SZA=80 • ) conditions (solid red line), as well as by Reaction (R2) (solid blue line) for the nominal scenario.Dashed and dotted lines represent N 2 O productions for the "high NO x " scenario combined with the nominal and the "strong descent" temperature scenarios, respectively (only Reaction R1).See text for further details.
at the MIPAS tangent heights.Retrieved N 2 O abundances on the highest tangent height level are discarded by the operational algorithm, such that highest available N 2 O observations are taken around 60 km with variations of ±2 km related to the orbital characteristics.ESA 4.61/4.62data include all MIPAS observations taken with full spectral resolution between June 2002 and March 2004, representing thus a quasi-continuous data set.Here, we analyze operational N2O data interpolated to an altitude of 58 km, which turned out to be the highest altitude covered by this data set during the whole period.However, care has to be taken when statistically analyzing ESA data products (i.e.zonal mean values) which are close to the detection limit www.atmos-chem-phys.net/8/5787/2008/Atmos.Chem.Phys., 8, 5787-5800, 2008 5792 B. Funke et al.: Mesospheric N 2 O enhancements observed by MIPAS • S (top) and 70-90 • N (bottom) averages of ESA N2O vmr (reprocessed data version 4.61/4.62)d without statistical correction (black and red solid lines, respectively), single measurement noise errors solid line), and N2O vmr retrieved at IMK/IAA (Versions V3O 8.0 and V3O 12.0 shown by red and monds, respectively) at 58 km.Spikes in the N2O averages on 26 September 2002 (SH), 29 April 2003 nd 29 July 2003 (NH) are generated by spurious data.
Funke et al.: Mesospheric N 2O enhancements observed by MIPAS the mesospheric N 2 O enhancements during the Arctic winter 2004 was generated by this reaction during night.It is thus very likely, that its nighttime contribution was also responsible for the major part of the N 2 O enhancements observed in the other polar winters.Its daytime contribution, which does not require the presence of EEP, is expected to be con- , that a dominant contribution to www.atmos-chem-phys.net/8/5787/2008/Atmos.Chem.Phys., 8, 5787-5800, 2008 5798 B.