Energetic particle precipitation during solar active periods induces enhancements of NOx (N, NO, NO2) in the lower thermosphere/mesosphere which can be transported to the stratosphere within the polar vortex. The quantitative contribution of these NOx intrusions to ozone chemistry in the stratosphere is still under discussion. Here we present simulations with a three-dimensional model of the middle atmosphere where NOx enhancements in the lower mesosphere have been taken from the observations of the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) instrument on the European Space Agency satellite ENVISAT. Covering the period from July 2002 to March 2004 these observations include the strong solar proton event in fall 2003 and intrusions connected to auroral events during the Arctic and Antarctic winters. The comparison of the disturbed run with the undisturbed model run allows a quantitative assessment of the long-term influence of NOx intrusions on stratospheric chemistry in general and the ozone concentration in particular. From the model simulation, we estimate for the period from July 2002 to March 2004 that in total, an additional 5.4 Gigamol NOy has been brought into the middle atmosphere. This represents in early 2004 about 5% of the global NOy mass in the middle atmosphere. A 2 year decay time for such enhancements is estimated from the model. Persistent reduction of ozone concentration in the stratosphere caused by the NOx intrusions can be followed in the simulation for several months. This reduction is restricted to high latitudes and amounts to several Dobson units in the total ozone column.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The influence of energetic particle precipitation (EPP) induced NOx on the chemistry of the middle atmosphere was first suggested by Crutzen et al. . The role of the polar night mesosphere in coupling the thermosphere to the stratosphere was pointed out by Solomon et al. . Observational evidence for regular long-range NOx descent came from several satellite experiments [Callis et al., 1996; Randall et al., 1998, 2001; Rinsland et al., 1996; Russel et al., 1984], for example from the Halogen Occultation Experiment (HALOE) instrument on UARS, and from the ATMOS and POAM experiments. These new observational results triggered several model studies [e.g., Siskind et al., 2000] (see Jackman and McPeters  for an overview of related work). Recent 3-D model studies of the effects of EPPs have been performed with chemistry climate models (CCMs) applying artificial NOx enhancements [Langematz et al., 2005; Rozanov et al., 2005] or modules calculating NOx and HOx production from prescribed ionization rates [e.g., Jackman et al., 2008]. Baumgaertner et al.  adapted results of Randall et al.  to estimate effects of NOx produced by low-energy electrons in the middle atmosphere. Whereas these model simulations qualitatively reproduce various effects connected with EPPs, the combination of chemistry and transport is especially important for the simulation of the EPP related effects. Model simulations using actual meteorological conditions and comparisons with corresponding observations are therefore needed as a validation of such experiments.
 Since the middle of the year 2002, the satellite ENVISAT of the European Space Agency (ESA) observes the middle atmosphere from the upper troposphere region up to the mesosphere with its Michelson Interferometer for Passive Atmospheric Sounding (MIPAS). MIPAS is a Fourier transform spectrometer for the measurement of high-resolution gaseous emission spectra at the Earth's limb [Fischer and Oelhaf, 1996]. It allows retrieval of global distributions of NO2 on a nearly daily basis up to the lower mesosphere. These observations represent one of the most complete data sets for studying the influence of energetic particle–induced changes in the middle atmosphere as they cover the middle atmosphere during polar night. Only here can NOx descend to the stratosphere where it might influence ozone chemistry substantially. Vogel et al.  studied the Arctic winter 2003–2004 in detail with their 3-D Chemical Lagrangian Model of the Stratosphere (CLaMS) with respect to the influence of NOx on the ozone chemistry during this winter using MIPAS data as a upper boundary condition. Here we present model simulations using the 3-D model of the middle atmosphere KASIMA (KArlsruhe SImulation Model of the middle Atmosphere) with and without a NOx disturbed middle atmosphere, where the NOx enhancements have been taken from the MIPAS data set of its phase I (full resolution) observation period from July 2002 until March 2004, extending the study of Vogel et al. . The comparison of the disturbed and undisturbed runs allows a quantitative assessment of the influence of NOx intrusions on stratospheric chemistry in general and on ozone in particular for this period.
2. Model Experiments
2.1. Model Description
 Model experiments have been performed with the KASIMA model. The KASIMA 3-D model is a mechanistic model of the middle atmosphere which can be coupled to specific meteorological situations by using analyzed lower boundary conditions and nudging terms for vorticity, divergence and temperature. The model is based on the solution of the primitive meteorological equations in spectral formulation and uses the pressure altitude z, with z = −H ln(p/p0), as the vertical coordinate. Here p is the pressure, H(=7 km) is the atmospheric-scale height and p0 = 1013 hPa. (Note that throughout the paper, altitudes are given as pressure altitude as defined above.) A detailed description of the model is given by Kouker et al. [1999a]. The model has been previously used for investigations of stratospheric transport and chemistry [Kouker, 1993; Reddmann et al., 1999; Ruhnke et al., 1999; Kouker et al., 1999b; Reddmann et al., 2001].
 Here we use the version as described by Reddmann et al. , but with ECMWF operational analyses. Temperature, horizontal divergence and relative vorticity of the analyses are used up to 18 km, up to 48 km the model is relaxed to the analyses with a time constant of 4 h. Between 48 km and upper boundary at 120 km the prognostic part of the model is used. From the lower boundary at 7 km up to a pressure height of 25 km, the vertical resolution is 750 m. From 25 km up to the upper boundary the vertical spacing between the levels gradually increases to 3.8 km. The triangular truncation T21 used corresponds to a horizontal resolution of about 5.6° × 5.6°. A numerical time step of 12 min was used in the experiments. Heating rates were calculated using a climatologic ozone field. The model simulates long-term transport in the stratosphere reasonably well. This has been shown by comparisons with the inert tracer SF6 and derived mean age of air [Engel et al., 2006; Stiller et al., 2008].
 The full stratospheric chemistry as described by Ruhnke et al.  is used in an updated version with reaction constants from JPL2002 [Sander et al., 2002]. The photolysis rates are calculated with a vectorized version of the Fast-J2 photolysis code [Bian and Prather, 2002], supplemented by Lyman-α photolysis of O2, CH4, CO2, and H2O using the Lyman-α actinic flux parameterization of Reddmann and Uhl . The chemistry is calculated up to 90 km, above which only transport is applied. The chemical fields have been initialized from the Mainz 2D model (J. U. Grooß, personal communication, 2006).
2.2. NOx Intrusions in the Model
 The operational ESA level 2 MIPAS/ENVISAT data product yields a rather continuous and complete data set for ozone, NO2, CH4, N2O, HNO3, and H2O for the time period July 2002 to March 2004 where MIPAS operated in full resolution mode. As MIPAS observes the thermal emission of the atmosphere, observations are available also for polar night. This is an important advantage of the MIPAS/ENVISAT data set compared to other observations from solar occultation instruments. Here we use reprocessed NO2 data of the version 4.61/4.62. Generally the maximum altitude where NO2 data are available is around 50 km; for solar/geomagnetically active periods with higher mesospheric NO2 concentrations the data reach into the mesosphere. The operational retrieval algorithm is described by Ridolfi et al. , and results have been presented by Carli et al. . NOx has a photochemical lifetime of a few days in the mesosphere, and during night most of NOx is in the form of NO2. Nighttime observations of NO2 are taken therefore as a representative for the concentration of NOx (see section 5 for caveats).
Figure 1 shows time-height cross sections of the mean NO2 volume mixing ratio of the northern and southern polar cap (geographical latitudes poleward of 60°) in ppb, from the ESA MIPAS/ENVISAT nighttime observations as described above. Several strong NOx enhancements are present in the MIPAS/ENVISAT observations, which can be attributed to auroral activity in the respective polar winters and the solar proton event (SPE) in fall 2003. During periods given in Table 1 where enhanced mesospheric NO2 is present, model NOx values are taken from the MIPAS/ENVISAT observations for latitudes and altitudes given in Table 1 using nighttime NO2 as a proxy for NOx, overwriting background values from the model. In order to reduce the uncertainty, for these periods, 5 day averages are calculated. The sampled data are interpolated to a regular grid of 20° × 5° in latitude and longitude having a vertical resolution of 2.5 km using an inverse distance weighting method. The method of interpolation has also been used for the comparison between the model and the observations. The ESA data set has been preferred over other data sets based on MIPAS/ENVISAT observations as it provides a rather complete coverage for the MIPAS/ENVISAT phase I observation period, but it lacks the accuracy of the IMK-IAA data set as it does not include non-LTE effects and does not consider horizontal gradients of mixing ratios in the retrieval [see Wetzel et al., 2007].
Table 1. Periods With Enhanced NOx in the Upper Stratosphere/Lower Mesosphere Used in the Disturbed Model Simulation Run B for Initialization
ϕ > 60°N, z > 55 km
ϕ > 60°S, z > 55 km
solar proton event (SPE)
ϕ > 50°N and S, z > 45 km
ϕ > 60°N, z > 55 km
ϕ > 60°N, z > 55 km
 As described above and given in Table 1, we use the NOx derived from the observation to overwrite the NOx fields in the model only during strong enhancements in the mesosphere and only there. In this sense we treat the NOx intrusions as an additional (mostly mesospheric) NOx source. Changes of chemistry in other parts of the middle atmosphere are consequences of NOx transported downward from the mesosphere or from changes of the photolysis rates through its effect on ozone. A run without the additional NOx source is taken as reference to investigate chemical effects in the stratosphere caused by the NOx intrusions. In the following, we use the term run A for the reference simulation without additional NOx, and for the disturbed simulation run B. As we use only observations in the mesosphere, the known overestimation of the MIPAS-ESA data set in the stratosphere during high mesospheric NO2 concentrations [Wetzel et al., 2007] has no consequences for the model simulation. From the few comparisons made between the ESA and the more sophisticated IMK-IAA retrieval of the MIPAS/ENVISAT data, we expect the data set to underestimate the NO2 concentration in the mesosphere.
3.1. Model Dynamics and Comparison With Observations
 In order to characterize the model dynamics, Figure 2 shows the zonal mean temperature at 80° in both hemispheres from July 2002 to March 2004 for the MIPAS/ENVISAT data set and the model. Generally the agreement in the stratosphere is as expected for a model nudged to ECMWF temperatures up to 1 hPa. Main differences are related to the upper stratosphere/lower mesosphere where the stratopause is warmer than observed and the lower summer mesosphere is colder. For the Northern Polar region the winter 2002–2003 is well reproduced whereas in the winter 2003–2004 the stratopause reforms too fast at 50 km. For this winter the prolonged midwinter warming period [Manney et al., 2005] is evident from Figure 2 with a minor warming end of December and a major warming in early January. In the Southern Polar region the outstanding feature is the vortex split in September 2002 with temperature exceeding 240 K at about 30 km. During Antarctic winter 2003 the stratopause is about 10 K warmer in the model compared to the observations. As a further illustration of the dynamics in the model, Figure 3 shows the zonal mean wind with the warming periods of the different winters.
3.2. NOx in the Model and Comparison With Observations
 In order to demonstrate how the model simulates NO2 under nondisturbed conditions, Figure 4 shows a correlation between observed and simulated NO2 values during spring equinox 2003 in the low and mid latitude stratosphere of both hemispheres. The model mostly reproduces the observed values but has its NO2 maximum slightly higher than observed and therefore underestimates NO2 at the observed maximum and overestimates it above. Figure 5 gives NOx in the model including the additional NOx (run B) for the MIPAS/ENVISAT period in the same way as shown in Figure 1 as an overview. First we see a reasonable agreement between model and observation in the regions where the model is initialized by the data. Outside the initialization domain, the downward transport of NOx into the stratosphere seems to be generally well represented by the model in regard to descent rate and mixing ratio. During Arctic winter 2003–2004 an extended stratospheric warming period caused the disruption of the polar vortex and subsequent destruction of NOx which can be seen in both observations and model. Obvious discrepancies can be found at 30–40 km altitude during Antarctic midwinter 2003 and Arctic late winter 2003–2004 where the observations show enhanced NOx values which have been, as mentioned above, identified as retrieval artifacts.
Figures 6 and 7show detailed plots of the periods of Antarctic winter 2003 and Arctic winter 2003–2004 when strong NOx intrusions have been observed. We focus on the period after the initialization when the transport of mesospheric NOx into the stratosphere has been essentially completed and compare the observations with run B. In the Antarctic winter 2003 the observations show for 1 September maximum values of about 15 ppb NO2 at 39 km. The simulation of NOx yields about 14 ppb at an altitude of 41 km. Figure 7 shows the two NOx intrusion events in the Arctic winter 2003–2004: the direct in situ production of NOx with its maximum at about the stratopause directly following the SPE in fall 2003 and the very intense NOx disturbance in late winter 2003–2004 which could be observed by MIPAS only until the end of March 2004. For the SPE event, we compare observations and model results before the minor stratospheric warming in December 2003. At the end of November the observations show maximum NO2 values of about 30 ppb at about 46 km; the model yields 20 ppb NOx at about 43 km. At the end of March the observations yield maximum values of 50 ppb NO2 at 46 km, whereas the model yields about 45 ppb at about 42 km. Analyzing the contours, we find a corresponding descent in November 2003 of about 7 km in the observations and about 9 km in the model, and in February 2004 a descent of 7 km in the observations and 10 km in the model. To summarize, the model underestimates the entry of NOx only slightly for the Antarctic winter 2003, for the SPE 2003 by about 30% and for late northern winter 2003–2004 again only a slight underestimation of NOx transported from the mesosphere is observed. The descent in the Antarctic winter seems to be reproduced by the model very well, whereas for the Arctic winter the model overestimates the descent by 2–3 km per month.
 In the middle stratosphere, the reactive nitrogen compounds NO and NO2 are converted to reservoir gases, of which HNO3 is the most abundant. Orsolini et al.  found much higher HNO3 concentrations observed by MIPAS/ENVISAT when compared to their model for the winter 2003–2004; Stiller et al.  analyzed the Antarctic winter 2003, when a distinct secondary maximum of HNO3 was found in MIPAS/ENVISAT data at about 40 km. These observations are in line with findings from earlier satellite missions [Austin et al., 1986; Kawa et al., 1995]. The latter explained this discrepancy as a result of reactions of N2O5 with water cluster ions. This reaction, originally proposed by Böhringer et al. , had been combined with heterogeneous reactions on sulfate aerosols by de Zafra et al.  in order to understand HNO3 satellite and ground based observations in polar winters.
 First comparisons of the HNO3 observations with the KASIMA model also showed a pronounced underestimation of HNO3 in the late polar winter when high NOx concentrations indicate strong NOx intrusions. We therefore included the parameterization of de Zafra et al.  in our chemistry module. Note that we do not intend to simulate the period during the SPE in fall 2003 when we expect higher ionization rates. Rather we are interested in the conversion of downward transported NOx into HNO3 which takes place weeks after NOx enters the stratosphere and away from periods with enhanced ionization rates caused by solar proton events.
Figure 8 shows a time-height cross section of HNO3 of the observations and the disturbed simulation (run B) for the Antarctic winter 2003. In Figure 9 the same is shown for the Arctic winter 2003–2004. In the Antarctic winter the HNO3 buildup starts in mid-June and reaches its maximum in mid-July with values of about 7 ppb. In the model, HNO3 also increases in mid-June but reaches its maximum at the end of July or beginning of August with maximum values of about 6 ppb. For the Arctic winter 2003 the increase of HNO3 following the SPE is qualitatively reproduced with an onset at mid-November. Maximum values occur at lower altitude at 40 km of only about 3.5 ppb already at beginning of December compared to observed maximum values of about 7 ppb in the first half of December 2003 at 43 km. The most striking difference is the total absence of observed HNO3 enhancements in late winter 2003–2004 where the model again clearly shows enhancements of several ppbs starting in March.
Figures 10 and 11 show the comparisons between observations and model for ozone. The model underestimates ozone by about 1 ppm from the beginning of the Antarctic winter, decreases further compared to the observation till middle of August and recovers to about observed values at the end of September. The reason of the deficit of modeled ozone seems to be missing transport of ozone rich air from mid to high latitudes (>70°) at altitudes of about 42 km in the fall season of the Southern Hemisphere as can be seen by comparisons between the observations and the model (not shown). In the model this ozone rich air doesn't reach the high latitudes and cannot be transported downward during the winter. For the Arctic winter 2003–2004 the agreement is generally better. The observed ozone maximum altitude is higher in the observations by about 4 km after the stratospheric warming in early January 2004. Note that at 50 km the ozone observations start to be sporadic and limited in latitude.
4. Chemical Effects of the NOx Intrusions
 Comparing the observations and the model results including the additional NOx as presented above we find in general an approximate agreement in the downward transport and the mixing ratios of nighttime NO2 and NOx, respectively. Against the background of the undisturbed stratosphere, the NOx enhancements in the observations and in the model can be clearly traced as seasonal events restricted to polar latitudes, leaving the rest of the stratosphere mostly unaffected. This allows to estimate the chemical effects of NOx intrusions by comparing the disturbed and the undisturbed model run, and to analyze the individual events which show different characteristic between intrusions starting in the upper mesosphere and the SPE with enhancements down to the stratosphere. Note that the chemical fields are not used for the calculation of heating rates. This makes a direct evaluation of the chemical effects possible, but it is not possible to evaluate effects caused by chemical radiative coupling of the disturbed ozone field to the dynamics. In the real atmosphere these coupling effects take place and are at least partly included in the dynamical fields used in the nudged model. To that extent the chemical effects are influenced by the coupling effects as the reference dynamical state does not correspond to a real undisturbed one. Nevertheless for the purpose of quantifying the strength of the NOx intrusions on chemistry this is only an effect of minor importance.
4.1. Additional NOy
 The sum of all nitrogen-containing species except N2 and the source gases like N2O is called NOy [Brasseur and Solomon, 2005]. The contribution of solar- and aurora-induced NOx to the total NOy budget is shown in Figure 12. There the zonal mean time cross section of the NOy column difference and its relative change as percentage of the total column is shown for the time period 2002–2005. In the cold polar winter stratosphere NOy is not completely in gaseous form. As a consequence, the relative change peaks during the Antarctic winter. Substantial and enduring changes of total NOy are seen essentially for the Northern Hemisphere following the winter 2003–2004. Here in summer 2004 an approximate 5% increase of NOy is seen for latitudes higher than about 30°N. Even for the following year 2005 a noticeable increase of NOy is left.
Figure 13 shows the global excess NOy in mol for the period July 2002 to the December 2005. The auroral winter intrusions in the Arctic winter 2002–2003 and Antarctic winter 2003, the SP event in fall 2003 and the strong intrusion in the Arctic winter 2003–2004 are clearly discernible as single events adding to the total NOy background. The characteristic time scale for the decay of the additional NOy in the model lies in the range of about 2 years.
 We estimate the additional NOy in Arctic winter 2002–2003 at 0.4 Gigamol (GM), in Antarctic winter 2003 at about 1.4 GM, the NOy from the SPE at about 1.5 GM and in January 2004 at about 2 GM. As in the Arctic winter both intrusions overlap, they cannot clearly be separated. The value estimated for the Antarctic winter 2003 is about 75% of the value [Funke et al., 2005] deduced from their observation based study with MIPAS/ENVISAT data. Randall et al.  give a range of 1 to 2 GM estimated from HALOE observations for this winter, depending on the use of average or maximum observed NOx values. They deduce the total additional NOy from EPP produced NOx at 45 km alone, assuming essentially that NOx has not already been converted into reservoir gases. We remark, that according to our simulation, this is not obvious for the Antarctic midwinter period with the observed buildup of HNO3. Globally, the additional NOy in the model which we derive amounts at its maximum to about 5% of the total 70 GM NOy in the middle atmosphere.
 The effects of the additional NOy in the atmosphere on the formation of particles in the cold polar vortex (for example by changing the saturation vapor pressure) is small, but still noticeable. The mass of NOy in NAT particles increases in the Antarctic winter 2004 and the Arctic winter 2004–2005 by about 5%.
 Our simulation does not include mesospheric production of N2O [see Semeniuk et al., 2008] which probably lowers the NOy entry somewhat as N2O at altitudes above 50 km is easily photolyzed.
4.2. Effects on Ozone
 The relevance of energetic particle precipitation is determined by its chemical coupling to NOx and NOy interacting themselves on radiatively active gases such as ozone and H2O. This causes changes of the stratospheric circulation and could possibly influence climate on longer time scales. Figure 14 shows changes of ozone caused by the NOx intrusions in the Antarctic winter 2003 and for the Arctic winter 2003–2004. In the Antarctic winter the maximum additional ozone loss closely follows the descent of NOy and has its maximum with about 1.5 ppm at the southern spring equinox. Below the zone of ozone loss a small healing effect caused by the lower ozone column can be observed. The SPE of fall 2003 has essentially no effect in the Southern Hemisphere in this simulation. In the Arctic winter we can see the two distinct events, where after the SPE a maximum ozone loss is already observed during November 2003 with about 0.9 ppm. The maximum ozone loss in this winter occurs as in the Southern Hemisphere at the spring equinox and is related to NOx of thermospheric origin. Ozone healing can be discerned during the period after the SPE whereas from January 2004 the ozone healing from the ozone loss at about 40 km interferes with the ozone loss caused by the SPE NOx and causes small positive ozone change only at the beginning of April.
Figure 15 shows the effects on ozone for the whole simulation period again for 75°S and 75°N. Additional ozone loss lasts for the following year in the Southern Hemisphere at the level of about 0.2 ppm. At the end of 2004 a small ozone change can be found above the band related to the Antarctic winter 2003. Closer inspection of time-latitude cross sections of NOy at different altitudes show that during summer 2004 NOy is transported from the Northern to the Southern Hemisphere at the stratopause level (not shown) exceeding 1 ppb, causing the additional ozone loss. During the Arctic winter 2002–2003 the minor amounts of additional NOx cause only small ozone changes. The maximum changes occur during spring 2004, exceed 30% and follow closely the downward transport of the additional NOx. The changes caused by the SPE in fall 2003 can be traced up to spring 2004 below 30 km but are lost to lower latitudes. The enormous amounts of additional NOx from January cause ozone changes of several percent even in the following spring at altitudes of about 30 km.
 The NOx-induced additional ozone loss causes also changes in total ozone. Figure 16 shows the change of total ozone for the whole simulation period. In the Southern Hemisphere ozone changes are restricted to the Antarctic region. In the Northern Hemisphere a decrease of total ozone of several Dobson units reach the midlatitudes. At high latitudes changes of about 5 DU can be found even in spring 2005.
 Our model setup allows to separate the effects of the several NOx intrusions during the ENVISAT observing period 2002–2004 on ozone. An obvious result is the strong ozone reduction in September 2003 in Antarctic winter and in March 2004 in Arctic winter, at least compared to the effects following the solar proton event. Most of the NOx produced in the SPE is lost during and after the stratospheric minor warming in late December minimizing its effect on ozone. Even the calculated maximum NOx mixing ratios in the SPE are lower than the NOx mixing ratios observed in late winter when compared at the maximum altitude of NOx enhancement after the SPE. Obviously, NOx, first observed in the upper mesosphere by MIPAS, and transported downward in the course of the winter makes a bigger contribution to NOx produced by EPP than the NOx produced directly in the stratosphere during energetic particle events. This is in line with recent analyses of observations by Randall et al. .
 The method of NOx reinitialization we used in our simulation (overwriting NOx in the mesosphere applying observed mixing ratios of nighttime NO2) gives a realistic estimation of the enhanced NOy in the stratosphere when comparing NO2 and HNO3 at the end of the respective winter. Nevertheless, this method has obvious deficits: the concentration of NOx in the mesosphere derived from NO2 is probably underestimated as above 60 km NO is also present during night in the atmosphere at least near the evening terminator. In addition, as at first NO has to react with O3 to build NO2 during strong descent and for low ozone concentrations NO2 and NO can still be far off chemical equilibrium. Indeed, B. Funke (personal communication, 2008) found NO/NOx nighttime ratios which are between about 10% at beginning January and up to 100% in February 2004. In part this could explain why NOx in the model shows its maximum at lower altitudes than observed during this winter. Vogel et al. , with their model, observe a general underestimation of NOx values in their model and therefore perform an additional simulation where they use estimated maximum NOx values. As they use NOx values from our KASIMA simulation to set their upper boundary condition the vertical transport in their model is probably responsible for the different result. Comparing with their reference run the KASIMA simulation also keep elevated NOx values for a longer period.
 The simple parameterization we apply for the buildup of HNO3 via protonized water clusters according to de Zafra et al.  yields surprisingly good results when compared with the observations, especially for the Antarctic winter 2003. Here even quantitatively the simulated maxima values agree with the observations but it seems that HNO3 production starts earlier and lasts longer. We note that in this winter the simulation reproduces the descent in the winter well, bringing additional NOy in time to altitudes where the mechanism via water clusters can work. Specifically in the late Arctic winter 2003/2004, the simulated descent is too strong and the enhanced levels of NOy reach lower altitudes too quickly in the simulation where HNO3 buildup can work. This would explain the HNO3 enhancement in the simulation missing in the observations. The observed HNO3 increase in early November cannot be simulated with this method and is probably related to a different ion chemistry [Verronen et al., 2008].
 The estimated additional total NOy in Antarctic winter 2003 is lower than the value Funke et al.  derive from observations but higher than the value Randall et al.  estimate as the average value. At least for this winter, KASIMA seems to simulate a realistic mesospheric contribution of NOy. The KASIMA model yields a generally good representation of transport within the stratosphere as demonstrated by simulated mean age of air values which agree very well with observed ones [Stiller et al., 2008]. Taking the decay of the additional NOy as shown in Figure 13 as a realistic estimation, and assuming a constant additional NOx source caused by NOx intrusions as observed during the period 2002–2004, NOy would globally increase by about 8 GM corresponding to about 10% of the total NOy mass. As we deduce the lifetime from a rather short period this can be only a rough estimation.
 Comparing the effects on ozone for the Arctic winter 2003–2004, we simulate higher changes of ozone than Vogel et al.  (their Figure 10 in their CLaMS reference run in March). Our values are similar to their maximum NOx run (their Figure 11). Additional loss of total ozone seems longer present in the CLaMS run but that could be in part just a question of the presentation at equivalent latitudes in that paper. In summary, both simulation give about the same amount of loss of ozone for the Arctic winter 2003–2004. Baumgaertner et al.  also simulate the effect of NOx intrusions with the EMAC model and compare with MIPAS/ENVISAT observations in Antarctic winter 2003. They find good agreement down to the 1500 K potential temperature level but the their model overestimates transport further downward. As a consequence, their estimated maximum additional ozone loss of 30 DU is probably overestimated. Langematz et al.  and Rozanov et al.  performed idealized 3-D coupled chemistry model experiments to study the effect of EPP. Both find important additional ozone loss connected with NOx intrusions, but it is difficult to compare their results quantitatively with the results presented in this study. Decreases of the annual mean ozone concentration on the order of 5% at low latitudes as reported by Rozanov et al.  cannot be confirmed by our simulations.
 When comparing the different winters and events and how they affect ozone, clearly the additional NOx brought into the atmosphere in the Arctic winter 2003–2004 has by far the strongest effect on ozone, stronger than the higher amount of additional NOx would suggest. Additional NOy in the winter 2002–2003 is about 20% of the amount in late winter 2003–2004, but ozone changes here are less than 2 DU. This illustrates that the time when EPP occurs in the mesosphere, the strength of the flux of EPPs, the specific dynamical situation of the winter, and especially the descent within the polar vortex determine how strongly the additional NOx will change ozone.
 Periods of enhanced mesospheric NOx during the MIPAS/ENVISAT observation period and their consequences on ozone have been successfully simulated with the KASIMA model. The NOx intrusions can be clearly separated in the model simulation and allow to study the chemical effects of the different events. From the comparison with the MIPAS data, we conclude that the method of overwriting NOx values at about 55 km by observed NO2 values yields realistic enhancements of NOy in the stratosphere which amount to about 5.3 Gigamol for the simulated period. The derived changes of O3 caused by the additional NOx are restricted to high latitudes. We estimate an effective lifetime of the additional NOy of about 2 years in the middle atmosphere. Periods of enhanced NOy as observed between 2002 and 2004 would increase the total NOy mass by about 10%.
 This research was funded by the project MANOXUVA within the DFG priority project 1176 CAWSES. MIPAS ENVISAT data have been used within the project VAMP, AO-Id 127. The authors thank Bernd Funke for helpful discussions. The helpful reviews were highly appreciated by the authors.