Model simulations of stratospheric ozone loss caused by enhanced mesospheric NOx during Arctic Winter 2003/2004

Satellite observations show that the enormous solar proton events (SPEs) in November 2003 had signiﬁcant e ﬀ ects on the composition of the stratosphere and mesosphere in the polar regions. After the October–November 2003 SPEs and in early 2004 signiﬁcant enhancements of NO x ( = NO + NO 2 ) in the upper stratosphere 5 and lower mesosphere in the Northern Hemisphere were observed by several satellite instruments. Here we present global full chemistry calculations performed with the CLaMS model to study the impact of mesospheric NO x intrusions on Arctic polar ozone loss processes in the stratosphere. Several model simulations are preformed with dif-ferent upper boundary conditions for NO x at 2000 K potential temperature ( ≈ 50 km 10 altitude). In our study we focus on the impact of the non-local production of NO x which means the downward transport of enhanced NO x from the mesosphere in the stratosphere. The local production of NO x in the stratosphere is neglected. Our ﬁndings show that intrusions of mesospheric air into the stratosphere, transporting high bur-dens of NO x , a ﬀ ect the composition of the Arctic polar region down to about 400 K 15 ( ≈ 17 – 18 km). We compare our simulated NO x and O 3 mixing ratios with satellite observations by ACE-FTS and MIPAS processed at IMK/IAA and derive an upper limit for the ozone loss caused by enhanced mesospheric NO x . Our ﬁndings show that in the Arctic polar vortex (Equivalent Lat. > 70 ◦ N) the accumulated column ozone loss between 350 – 2000 K potential temperature ( ≈ 14 – 50 km altitude) caused by the SPEs 20 in October–November 2003 in the stratosphere is up to 3.3 DU with an upper limit of 5.5 DU until end of November. Further we


Introduction
During periods of solar disturbances solar events can affect the concentration of constituents in the middle atmosphere. Protons, electrons, and alpha particles released from the sun are channelled along the Earth's magnetic field and cause ionization, ex-5 citation, dissociation, and dissociative ionization of the background constituents when they reach the Earth's atmosphere. Solar disturbances can also lead to solar proton events (SPEs) which are characterized by emission of protons with higher energies. Some of these highly energetic protons can penetrate to the stratosphere, but generally only in the polar regions. In general, the major effects of solar energetic particle 10 precipitation (EPP) and SPEs have been found to significantly increase the production of odd nitrogen NO y (N,NO,NO 2 ,NO 3 , N 2 O 5 , HNO 3 , HO 2 NO 2 , ClONO 2 , and BrONO 2 ) and odd hydrogen HO x (H, OH, HO 2 ) (e. g. Crutzen et al., 1975;Jackman et al., 1980;Solomon et al., 1981;López-Puertas et al., 2005a;Verronen et al., 2006). Both NO y and HO x play a key role in the ozone balance of the middle atmosphere 15 because they destroy odd oxygen by catalytic cycles. During polar night conditions, NO y is long lived and can be transported downwards. Therefore, extra NO y produced in the upper atmosphere could be important to the ozone budget of the lower and middle stratosphere. In contrast enhanced HO x produced by SPEs in the mesosphere or upper stratosphere is short-lived and therefore only yields short-time ozone decrease. 20 The fourth largest period of SPEs measured in the past 40 years happened in October-November 2003 (known as the "Halloween sunstrom"), which resulted in ionization effects in the atmosphere down to 30 km altitude near the geomagnetic poles (e. g. Jackman et al., 2005b). An especially vigorous period of high fluxes of energetic protons occurred from 28 -31 October 2003 and a second maximum was between Introduction the Arctic stratosphere and mesosphere. After the SPEs, ozone depletion signatures associated with significant enhancements of NO x (=NO + NO 2 ) (e. g. Seppälä et al., 2004;López-Puertas et al., 2005a;Rohen et al., 2005), of several NO y components like HNO 3 or N 2 O 5 (e. g. López-Puertas et al., 2005b;Orsolini et al., 2005), and of HO x (e. g. Degenstein et al., 2005;von Clarmann et al., 2005) in the upper stratosphere and 5 lower mesosphere were observed. Further, upper stratospheric and mesospheric enhancements of NO x were observed at high northern latitudes in early 2004 (e. g. López-Puertas et al., 2005aRandall et al., 2005) due to downward transport of upper atmospheric NO x produced throughout the winter by auroral and precipitating electrons . Re-10 sults by Randall et al. (2006) confirm that the impact of EPP on the atmosphere are modulated by meteorological conditions; their results suggest that a stronger vortex leads to increasing NO x values in the stratosphere caused by descent of NO x -rich air masses from the mesosphere and thermosphere within a well-isolated vortex.
In fact, the dynamic situation of the Arctic winter 2003/2004 was very anomalous. 15 A major warming beginning in late December 2003 led to nearly 2 months of vortex disruption . The upper stratospheric vortex broke up in late December, but began to recover by early January. In February and March it was the strongest since regular observations began in 1979. The lower stratospheric vortex broke up in late January, split into two fragments on 1 February with the fragments coa-20 lesced again by 17 February. Thus the period when the vortex was strong in the upper stratosphere but very small and weak in the lower to middle stratosphere was quite uncommon . The temperatures in the lower stratosphere were unusually high and temperatures in middle and upper stratosphere were unusually low during and after February 2004. These dynamic disturbances supported the downward 25 transport of NO x leading to the NO x enhancements in the upper stratosphere in early 2004.
In addition to this downward transport of NO x in the lower stratosphere, dynamic disturbances of the vortex also yielded enhanced meridional transport of NO x -rich as 4914 well as O 3 -rich air masses from the subtropics into the polar regions in the lower stratosphere. This effect was discussed by Konopka et al. (2007a) for the winter 2002/2003. They found that the enhanced meridional transport of NO x also causes ozone loss that can outweigh the halogen-induced ozone loss processes occurring in winter and early spring in the Arctic lower and middle stratosphere. 5 In this paper, model simulations with the Chemical Lagrangian Model of the Stratosphere (CLaMS) (McKenna et al., 2002a,b;Konopka et al., 2007b) will be presented. We analyze for the Arctic winter 2003/2004 the long-term impact of enhanced NO x in the mesosphere caused by the extremely strong SPEs in October-November 2003 and the mesospheric intrusion in early 2004 on stratospheric ozone budget. Here 10 we focus on the impact of enhanced NO x on ozone loss caused by downward transport from the mesosphere into the stratosphere, whereas enhancements of different NO y species below 55 km downward to 30 km altitude observed immediately after the SPEs (e. g. López-Puertas et al., 2005b) due to local production of NO x are not considered. 15 To study the impact of downward transport of enhanced NO x in the Arctic winter 2003/2004 on the stratospheric ozone budget, full chemistry studies are performed with the chemical transport model CLaMS (e. g., Konopka et al., 2003;. The simulations cover the altitude range from 350 -2000 K potential temperature (≈14-50 km altitude). The horizontal and vertical transport is driven by ECMWF 20 winds and heating/cooling rates derived from a radiation calculation. The mixing procedure uses the mixing parameter as described in Konopka et al. (2004) with a horizontal resolution of 200 km and a vertical resolution increasing from 3 km around 350 K potential temperature to approximately 13 km around 2000 K potential temperature according to the model set up (Konopka et al., 2007b) (IMK) and at the Instituto de Astrofísica de Andalucía (IAA), Granada, Spain. CFC-11 is provided by L. Hoffmann (Hoffmann et al., 2007). Some species measured by MIPAS are not available over the full model altitude range, there-5 fore these data are interpolated to typical tropospheric and mesospheric values at the lower and upper boundary, respectively. Other species are taken from the HALOE climatology (Grooß and Russell, 2005) (HCl and HF) or from results of a simulation with the Mainz 2-D model (Gidel et al., 1983;Grooß, 1996). For all species the lower boundary conditions at 350 K are taken through the course of the simulation from these initial values.

The model study
In the model, the flux of enhanced NO x from the mesosphere is implemented in form of the upper boundary conditions at 2000 K potential temperature (≈50 km altitude) which are updated every 24 h. The NO y constituents NO x , N 2 O 5 , HNO 3 and the tracers CH 4 , CO, H 2 O, N 2 O, and O 3 are taken from results of a long-term simu- 15 lation performed with the KASIMA model (Karlsruhe Simulation Model of the middle Atmosphere) (Kouker et al., 1999). In this KASIMA simulation (T. Reddmann, personal communication, 2007), enhanced NO x concentrations in the mesosphere during disturbed periods have been derived above 55 km from MIPAS observations (provided by the European Space Agency ESA). 20 In addition to this CLaMS reference model run (referred to as "ref" run) a model simulation without an additional NO y -entry (this means no NO x , N 2 O 5 , and HNO 3 ) at the upper boundary is performed. Because most NO y at the upper boundary is NO x in the following we refer this simulation without an additional NO x -source at the upper boundary as "no NO x " run. In addition, to estimate the possible maximum impact 25 of NO x on stratospheric ozone loss, we derive a maximum NO x -entry at the upper boundary condition from several satellite measurements of NO and NO 2 by the MIPAS instrument on board the ENVISAT satellite (provided by IMK/IAA) , by the HALOE instrument on board the UARS satellite (Russell et al., 1993), and by 4916 the ACE-FTS instrument on board the SCISAT satellite (Bernath et al., 2005). For this we modify the upper boundary conditions derived from KASIMA. For each day at the upper boundary the NO x mixing ratios for equivalent latitudes greater or equal 60 • N are replaced by the maximum NO x value observed by any satellite instruments within 60 • and 90 • N at 2000 K potential temperature at that day, respectively, by the maximum 5 NO or NO 2 value if no NO x measurement exists in this 24 h period. For days where no satellite observations are available we take the maximum NO x derived by KASIMA at that day. This model simulation is referred to as "max NO x " run.

Model results
To facilitate an analysis of the downward transport in CLaMS a mesospheric tracer 10 (MS) is implemented in the model which is initialized with 100% at the upper boundary and with zero below. During the model simulation this tracer is held constant at the upper and lower boundary with 100% and 0%, respectively. This tracer quantifies the percentage of air originating from the upper boundary at 2000 K potential temperature. Figure 1a shows that the influence of the upper boundary and mesosphere, respec-15 tively, extends downward to about 600-700 K potential temperature from February until June 2004, i. e. slightly above the region where halogens typically effectively destroy ozone in the Arctic polar winter. Strong downward transport from the mesosphere down to 900 K (MS=80%) occur at the beginning of January caused by the recovery of the upper vortex in early January. Further, strong transport down to 1000 K (MS=80%) is 20 found at the beginning of March where the vortex was very strong (cf. Sect. 1, . Figure 1b shows that the upper and middle stratospheric temperatures become atypically low during and after February as also reported by . The lower stratospheric temperatures remain extremely high after the major warming. The final warming and the vortex breakup, respectively, occurred not until end of April Introduction October-November 2003 and propagate downwards from 2000 K to about 800 K potential temperature. Further, Fig. 1c shows that model signatures of increased NO x in February and March due to downward transport of upper atmospheric NO x produced throughout the winter by auroral and precipitating electrons  propagate down to approximately 800 K -900 K until March 2004. Figure 1d shows NO x 5 without mesospheric NO x sources as simulated in the "no NO x " run. Here a typical seasonal variation of NO x can be seen that is not disturbed by an additional downward transport. The accumulated chemical ozone change (∆O 3 ) shown in Fig. 2c is derived from the difference between the simulated (see Fig. 2a) and passively transported (see Fig. 2b) 10 ozone. It allows to quantify the chemical contribution to ozone change during the winter. From late January on, meridional transport of ozone-rich air from low-latitudes into the polar regions yields high ozone mixing ratios up to 5-6 ppmv between 500-800 K (see Fig. 2b) caused by the vortex disruption in late January and the very weak lower stratospheric vortex during February and March (cf. Sect. 1). Below about 800 K, 15 accumulated ozone loss dominates the ozone evolution during the winter and early spring (see Fig. 2c). Here, about 0.5 ppmv accumulated ozone loss is simulated from Above about 800 K, accumulated ozone production dominates the ozone evolution during the winter and early spring (see Fig. 2c). The accumulated ozone production is 20 strongest in late February and early March and yields values up to 4 ppmv between 1300-1700 K. Here the accumulated ozone production is increasing because the sunlight is returned after the polar night. Our findings show that above 800 K the impact of highly enhanced NO x values in the stratosphere due to downward transport from the mesosphere on ozone is small since these NO x values do not yield accumulated 25 ozone loss like in altitudes below 800 K. Thus, above 800 K, there is ozone production in the descending mesospheric air only counteracted to a certain extent by ozone loss caused by the very high NO x intrusion from the mesosphere. In the Arctic polar vortex in winter 2002/2003, only small enhancements of NO x were found in the stratosphere, but enhanced ozone was found in an intrusion of mesospheric air descending into the lower stratosphere . The fact that air masses originate from the mesosphere are characterized by enhanced O 3 values is consistent with our simulations showing ozone production in the descending mesospheric air for the Arctic winter 2003Arctic winter /2004 Besides NO x -induced ozone loss also halogen-induced and HO x -induced ozone loss have an important impact on polar ozone loss processes. However, halogen-induced ozone loss plays a minor role during the winter 2003/2004, because the lower stratospheric temperatures are unusually high caused by the vortex disruption in the lower stratosphere at the end of January. An analysis of ECMWF temperature data shows 10 that PSCs could only occur during December 2003 and January 2004 between approximately 400-800 K potential temperatures (see Fig. 3). Limited to this time period the model simulates a small chlorine activation of up to 700 pptv ClO x (see Fig. 4a). Note that chlorine activation is only confined to the lower part of the region where most ozone depletion occurred in this time period (see Fig. 4b). This is consistent with re-15 sults by Vogel et al. (2003) and Konopka et al. (2007a), who found that NO x -induced ozone loss occurs slightly above the region where halogens destroy ozone in the Arctic polar winter. Consequently, the ozone loss between 500-800 K is not only driven by ClO x chemistry. Above 800 K, the ClO x altitude profiles have a maximum around 1600 K potential temperature (≈43 km) during sunlight conditions. During polar night 20 conditions the ClO mixing ratios are low in these altitudes (see Fig. 4a). In general, by the absence of PSCs in the middle stratosphere the destruction of ozone is mainly driven by NO x , whereas HO x -induced ozone loss plays an important role in the lowermost stratosphere and in the upper stratosphere and mesosphere, respectively.  shows the mean deviations of the simulated NO x mixing ratios from the satellite observations. The mean deviations are calculated for bins with an altitude range of 200 K potential temperature and for equivalent latitudes poleward of 70 • N. Figure 5 shows that the reference run (top row) yields lower NO x mixing ratios than observed by satellites for the upper levels. This effect is still small about −15% in November (−5 ppbv at 10 1800 K-2000 K) and is increasing to −101% (−17 ppbv at 1600 K-1800 K) and −151% (−19 ppbv at 1800 K-2000 K), respectively, in December. In January, no satellite data are available poleward of 70 • N equivalent latitude for these altitudes. In February a deviation of up to −129% (−77 ppbv at 1800 K-2000 K) is found to ACE-FTS measurements. In March the deviation is even up to −398% (−121 ppbv at 1600 K-1800 K) 15 and −379% (−136 ppbv at 1800 K-2000 K) measured by MIPAS and up to −146 % (−76 ppbv at 1600 K-1800 K) and −202% (−129 ppbv at 1800 K-2000 K) measured by ACE-FTS. The underestimation of upper stratospheric NO x in the reference run during periods of enhanced transport of NO x -rich air masses from the mesosphere into the stratosphere indicates that the NO x mixing ratios prescribed at the upper boundary 20 are too low. As shown above, this effect is very small during the SPEs in October-November, but larger a few weeks after the SPEs in December and enormous during the strong mesospheric intrusion in night. In consequence, the average NO contribution to NO x at 55 km depends on the availability of ozone and on the extension of the polar night area. Therefore NO 2 nighttime measurements are clearly a lower boundary for NO x . Further two retrieval codes are used. For the KASIMA longterm simulations the ESA operational processor (Raspollini et al., 2006) is used because more NO 2 data are available which is 10 essential to perform the simulations. The observations shown in this work are NO x data provided by the IMK/IAA retrieval code (Funke et al., 2005) which allows in contrast to the ESA operational processor accurate inference of NO 2 volume mixing ratios under consideration of all important non-LTE processes. The findings of a validation study of MIPAS-ENVISAT NO 2 operational data by Wetzel et al. (2007) are that large To study the impact of the underestimation of NO x mixing ratios at the upper boundary, we perform sensitivity tests with modified upper boundary conditions. At the upper boundary all NO x mixing ratios for equivalent latitudes greater or equal 60 • N are replaced by the maximum NO x value derived by the IMK/IAA-MIPAS retrieval code 25 (V3O NO 9 and V3O NO 2 9), observed by HALOE (V19) or by ACE-FTS (V2.2) as described in Sect. 2. In general for this "max NO x " run the simulated NO x mixing ratios at the upper levels are somewhat higher than the satellite observations (see Fig. 5, bottom row). In February the deviation is up to 51% (154 ppbv at 1600 K-1800 K) and 77% (163 ppbv at 1800 K-2000 K) and in March only up to −15% (7 ppbv at 1600 K-1800 K) and 5% (33 ppbv at 1800 K-2000 K) measured by MIPAS and up to 38% (107 ppbv at 1600 K-1800 K) and 33% (116 ppbv at 1800 K-2000 K) measured by ACE-FTS. This is because at the upper boundary the NO x mixing ratios for equivalent latitudes greater or equal 60 • N are mainly replaced by the maximum NO x value observed by satellites at 5 that day (see Sect. 2). Thus for the "max NO x " run the CLaMS NO x values are clearly on average higher than the satellite data as expected. Further, significant differences between the mean deviation calculated with IMK/IAA-MIPAS and ACE-FTS observations occur. Since NO x decays from the vortex center towards the vortex boundary in the reference run while it remains constant in the "max NO x " run, differences in the spatial NO x distribution between the reference and the "max NO  Figure 7 illustrates as discussed before that ACE-FTS and IMK/IAA-MIPAS NO x at 2000 K are much higher than the CLaMS reference run. On 4922 18 March 2004, the maximum NO x mixing ratio measured by MIPAS and ACE-FTS is 328 ppbv and 216 ppbv, respectively, at 2000 K. That is a factor of 2-3 higher than the NO x values simulated by CLaMS reference run for that day. Nevertheless Fig. 7 shows that simulated and observed ozone mixing ratios are in very good agreement for that day. We will discuss this in the next section in more detail. Finally, Fig. 7 shows that 5 an equivalent latitude of 60 • N describes very well the areas of enhanced NO x mixing ratios because the enhancements depends on the downward transport of NO x which in turn depends on the vortex strength.
To summarize, we compared the results of two model simulations with the observations: the reference run yielding rather lower NO x mixing ratios than observed and the 10 "max NO x " run yielding in general clearly higher NO x mixing ratios and a larger area of enhanced NO x mixing ratios compared to satellite observations. Thus these two simulations represent a lower and a clear upper limit of the real distribution and absolute values of NO x within the Arctic polar vortex. 15 To compare our simulated O 3 mixing ratios with satellite observations conducted by IMK/IAA MIPAS (Version: V3O O 3 7 until V3O O 3 9) and ACE-FTS (V2.2 update) (Dupuy, E. et al., 2007) we calculate the mean deviations of the simulated ozone mixing ratios from the satellite observations similar as in Sect. 4.1 for NO x mixing ratios. A comparison between MIPAS ozone data retrieved with the IMK/IAA procesor with 20 several ground-based, balloon-borne, and satellite instruments for stratospheric ozone yields mean differences of generally ±10% (Steck et al., 2007). A currently reported validation of ozone from ACE-FTS (V2.0 update) with MIPAS ESA ozone found that the O 3 data are within ±10% between 10-42 km altitude, but increase above this range, with ACE-FTS reporting larger volume mixing ratios than MIPAS by up to 40% around 25 53 km altitude (Cortesi, U. et al, 2007). ACE-FTS), respectively. The "max NO x " run yields in general smaller O 3 mixing ratios as in the reference run, because the NO x -induced ozone loss is higher in the "max NO x " run. Thus in February and March 2004, where the largest discrepancies between simulated and observed NO x occur (cf. Sect. 4.1) the simulated O 3 agrees best with the reference run. For these months simulated O 3 values in the "max NO x " are clearly lower than the satellite observations. Thus the "max NO x " run is a clear lower limit for O 3 compared to satellite observations. Furthermore, Fig. 6 shows that enhanced NO x at the upper boundary has an impact of the O 3 mixing ratios over the whole altitude profile (cf. below, Sect. 5). Finally it is shown in Fig. 6 that the ozone mixing ratios at the upper boundary (that is in the KASIMA model) are a bit too low compared to satellite 15 observations. In general, differences between observations conducted by IMK/IAA-MIPAS or ACE-FTS and the CLaMS simulations are increasing with altitude, but the form of both altitude profiles is very similar. This is consistent with the fact that above approximately 40 km altitude ACE-FTS reporting larger volume mixing ratios than ESA-MIPAS by up 20 to 40% (Cortesi, U. et al, 2007). The length of the error bars in Fig. 6 indicates the variability in the difference between simulated and observed O 3 values.

O 3 mixing ratios
Summarizing, the comparison between simulated and observed ozone mixing ratios confirm that the "max NO x " run is a upper limit case which overestimates the strength of the NO x intrusions at the upper boundary and underestimates the ozone mixing ratios i. e. the O 3 destruction in this model run is too high compared to IMK/IAA-MIPAS and ACE-FTS measurements. Thus, the reference run yields the best agreement with O 3 satellite observations. 4924

Ozone loss
To answer the question what is the quantitative influence of the mesospheric NO x sources on stratospheric ozone loss, CLaMS simulations with and without a mesospheric NO x source at the upper model boundary are compared. For the reference model run the respective differences of NO x (∆ abs NO x ) and O 3 (∆ abs O 3 ) shown in and ∆ abs O 3 correlates fairly well. The strongest impact of enhanced NO x on O 3 mixing ratios is found in January between 850 K and 1100 K potential temperature. Here as shown before (cf. Sect. 3) a strong downward transport caused by the recovery of the upper vortex occur that cause up to 1 ppmv (28%) lower ozone mixing ratios than in the "no NO x " run. Further during the period of the very strong upper vortex in February 15 and March 2004 (cf. Sect. 1) our model results show up to 0.6 -0.7 ppmv (17%) lower O 3 mixing ratios at the end of March 2004 between 1300 K and 1700 K potential temperature. Furthermore, from the beginning of March on a small impact of enhanced NO x on ozone is found down to 400 K potential temperature (17-18 km altitude).
The enhanced NO x mixing ratios have also an impact on HO x . The spatial distribu-20 tion of the minima and maxima between ∆ abs O 3 and ∆ abs HO x correlates fairly well (not shown here). Up to approximately 150 pptv (112%) lower HO x mixing rations as in the "no NO x " run are found in January between 850 K and 1100 K potential temperature. That means the HO x -induced ozone loss is lower in the reference run than in the "no NO x " run. 25 The impact of mesospheric NO x sources on the column ozone is shown in Fig. 9a  Interactive Discussion simulation without mesospheric sources ("no NO x "), and in yellow for the run with a maximum NO x source ("max NO x "). The corresponding accumulated column ozone loss in the stratosphere calculated from the simulated O 3 values minus the passively transported ozone (as shown in Fig. 2) is shown in Fig. 9a (middle panel). Until the end of April a column ozone loss of up to approximately 80 DU is simulated for the refer-5 ence run and up to approximately 90 DU for the "max NO x " run. Thereafter the column ozone loss is strongly decreasing because the final warming occurred at the end of April 2004 resulting in transport of midlatitude air masses to the pole that have encountered less chemical ozone depletion. In Fig. 9a (bottom panel) the ozone loss attributed to enhanced mesospheric NO x intrusion are shown. This value is calculated from the 10 difference between the "no NO x " run and the reference run (red) and the "max NO x " run (yellow), respectively. At the end of March the ozone loss attributable to mesospheric NO x intrusion in early 2004 is up to 10 DU (equivalent to ≈3% of the ozone column) for the reference run and up to 18 DU (equivalent to ≈6% of the ozone column) for the "max NO x run (see Fig. 9a, bottom panel). That means that nearly twice the ozone loss 15 occurs in the "max NO x " run as for the reference run. Further Fig. 9a (bottom panel) shows ∆O 3 of approximately 19 DU at the end of May. This is after the final warming occurred at the end of April. This feature shows that enhanced mesospheric NO x values have still an impact on ozone chemistry also during the change from polar winter to summer conditions. Thus we conclude that the column ozone loss in the strato-20 sphere at the end of March assigned to mesospheric NO x sources is about 10 DU but lower than 18 DU. The column ozone loss caused by the SPEs in October-November 2003 in the stratosphere is up to 3.3 DU with an upper limit of 5.5 DU in mid-November.
These absolute values are small compared to the ozone loss in spring 2004 caused by the mesospheric NO x intrusions, but their fraction at the total accumulated ozone loss 25 occurred until mid-November is up to 36% and 48%, respectively. The presented column ozone loss in the stratosphere is derived for equivalent latitudes greater 70 • N. The impact of mesospheric enhanced NO x is strongest for high equivalent latitudes because the downward transport depends on the vortex strength.
4926 Therefore choosing a wider latitude range to analyze the average column ozone loss yields lower column ozone loss.
Further we also calculate the ozone loss over the winter and the ozone loss caused by mesospheric NO x sources for the lower stratosphere over an altitude range from 350 K until 700 K (≈14-27 km, see Fig. 9b). Here a total ozone loss until end of April 5 up to 68 DU for the "ref" run and up to 71 DU for the "max NOx" run are found. A maximum column ozone loss at the end of April assigned to mesospheric NO x sources is up to 3.5 DU for the reference run and 6.5 DU for the "max NO x " run, respectively. Unfortunately no analyses to the polar stratospheric ozone loss for the Arctic winter 2003/2004 exist so far because the halogen-induced ozone loss plays a minor role and yield very 10 low ozone loss over the winter as mentioned before. But a three-dimensional model study of the Arctic ozone loss in that winter results in an O 3 change of approximately 50 DU between 345-670 K (poleward of 65 • equivalent latitude) until end of March and beginning of April (Feng et al., 2005). Our simulations yield 45-50 DU ("ref" run) and 47-52 DU ("max NO x " run) until end of March and beginning of April. However the 15 simulation by Feng et al. (2005) are started at the beginning of December 2003 and our simulations are started at beginning of October 2003 and second our analysis is for equivalent latitudes poleward of 70 • in contrast to 65 • in the Feng et al. (2005) study. Nevertheless, we can conclude that our simulations are the same range as the results by Feng et al. (2005). Thus our finding shows that the impact of enhanced mesospheric 20 NO x values on ozone loss processes in the lower stratosphere (350-700 K) is about 5% but lower than 9% for the Arctic winter 2003Arctic winter /2004 For cold Arctic winters an accumulated ozone loss in the lower stratosphere was derived in the range of about 100 DU (e. g. Rex et al., 2004;Tilmes et al., 2006). Thus in cold Arctic winters with strong halogen-induced ozone loss the impact on the column Interactive Discussion the non-local production of NO x caused by SPEs of October-November 2003 have a negligibly small impact on ozone loss. In our simulations the enhancements of different NO y species below 55 km downward to 30 km altitude observed immediately after the SPEs (e. g. López-Puertas et al., 2005b) caused by particle precipitation reaching the stratosphere are not considered. 5 The impact of local NO x production on polar ozone loss is considered in a study by Jackman et al. (2005a), where the influences of all the solar proton fluxes between 2000 and 2003 with a 2-D CTM is simulated. They found an impact on total ozone of 0.5-3% (≈2-10 DU) (WMO, 2007), whereas we derive a column ozone loss about 2.5 DU, but lower than 4.5 DU caused by the October-November 2003 SPEs. Tak- 10 ing into account that we neglect the local NO x production below 55 km in October-November 2003 in our model simulations and only consider the downward transport of enhanced NO x it is not surprising that our results are at the lower limit of the predictions by Jackman et al. (2005a). 15 The impact of downward transport of enhanced upper atmospheric NO x caused by solar proton events in October-November 2003 and by mesospheric NO x intrusions in early 2004 produced throughout the winter by auroral and precipitating electrons on the stratospheric ozone budget was studied performing model simulations with the chemical transport model CLaMS. Upper boundary conditions were taken from results 20 of a longterm simulation conducted with KASIMA, where increased NO x concentration in the mesosphere have been derived from MIPAS measurements (reference run). For the Arctic polar region (Equivalent Lat. >70

Summary and conclusions
• N) we found that enhanced NO x caused by SPEs in October-November 2003 is transported downward into the middle stratosphere to about 800 K potential temperature (≈30 km) until end of December 2003. 25 The mesospheric NO x intrusion due to downward transport of upper atmospheric NO x produced by auroral and precipitating electrons  affects NO x mix-4928 temperature during the period of the strong polar vortex in the middle stratosphere. We point out that the enhancements of different NO y species below 55 km downward to 30 km altitude due to local production of NO x observed immediately after the SPEs (e. g. López-Puertas et al., 2005b) caused by particle precipitation down in the stratosphere are not considered in our simulations as well as the local production of HO x . 10 Thus in our simulations only the non-local production of NO x , namely the transport of enhanced NO x from the mesosphere (upper boundary condition) to the stratosphere is considered and the local production of NO x and HO x is neglected.
An intercomparison of simulated NO x and O 3 mixing ratios with satellite observations by ACE-FTS and MIPAS shows that the NO x mixing ratios at the upper boundary 15 derived from KASIMA simulations are in general too low. Therefore a model simulation with higher NO x mixing ratios at the upper boundary derived from satellite measurements was performed. For this model run ("max NO x ") the simulated NO x mixing ratios and the total area of enhanced NO x mixing ratios are in general higher compared to satellite observations and provide so an upper limit for the impact of mesospheric NO x 20 on stratospheric ozone chemistry. The comparison between simulated and observed ozone mixing ratios confirm the "max NO x " run is an upper limit which overestimates the NO x entry at the upper boundary and underestimates O 3 caused due to stronger O 3 destruction. Moreover in the "ref" run the simulated O 3 is in very good agreement with satellite measurements. D. Y.: Retrieval of stratospheric NO x from 5.3 and 6.2 µm nonlocal thermodynamic equilibrium emissions measured by Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) on Envisat, J. Geophys. Res., 110, D09302, doi:10.1029/2004JD005225, 2005 and Belotti, C.: Comment on "Origin of the January-April 2004 increase in stratospheric NO 2 observed in northern polar latitudes" by Jean-Baptist Renard et al., Geophys. Res. Lett., 34, l07813, doi:10.1029/2006GL027518, 2007 Gidel Ozone loss in DU for these model runs. Bottom panel: ∆O 3 caused by the addition NO x -source at the upper boundary. Shown is the ozone column of the "no NO x " run minus the ozone column of the "ref" (red) and the "max NO x " run (yellow). (b) same as (a), but for an altitude range from 350 K until 700 K (≈14-27 km) potential temperature.