In order to improve our understanding of the effects of energetic particle precipitation on the middle atmosphere and in particular upon the nitrogen family and ozone, we have modeled the chemical and dynamical middle atmosphere response to the introduction of a chemical pathway that produces HNO3 by conversion of N2O5 upon hydrated water clusters H+·(H2O)n. We have used an ensemble of simulations with the National Center for Atmospheric Research (NCAR) Whole-Atmosphere Community Climate Model (WACCM) chemistry-climate model. The chemical pathway alters the internal partitioning of the NOy family during winter months in both hemispheres, and ultimately triggers statistically significant changes in the climatological distributions of constituents including: i) a cold season production and loss of HNO3 and N2O5, respectively, and ii) a cold season decrease and increase in NOx/NOy-ratio and O3, respectively, in the polar regions of both hemispheres. We see an improved seasonal evolution of modeled HNO3 compared to satellite observations from Microwave Limb Sounder (MLS), albeit not enough HNO3 is produced at high altitudes. Through O3changes, both temperature and dynamics are affected, allowing for complex chemical-dynamical feedbacks beyond the cold season when the pathway is active. Hence, we also find a NOxpolar increase in spring-to-summer in the southern hemisphere, and in spring in the northern hemisphere. The springtime NOxincrease arises from anomalously strong poleward transport associated with a weaker polar vortex. We argue that the weakening of zonal-mean polar winds down to the lower stratosphere, which is statistically significant at the 0.90 level in spring months in the southern hemisphere, is caused by strengthened planetary waves induced by the middle and sub-polar latitude zonal asymmetries in O3and short-wave heating.
 Nitric acid (HNO3) is considered to be a very important chemical species in the stratosphere. As a key reservoir for the reactive nitrogen (NOx), HNO3intervenes in the nitrogen-dominated ozone (O3) depleting cycle, and it plays an important role in the halogen-dominated catalytic cycle leading to polar O3 depletion. For these reasons, the distribution of HNO3in the stratosphere has been studied closely with ground-based and satellite instruments. In the last two decades, a large amount of observations from several satellite-borne instruments, such as the Limb Infrared Monitor of the Stratosphere Experiment (LIMS) on the NIMBUS 7, the Sub-Millimeter Radiometer (SMR) aboard Odin, the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) aboard ENVISAT, the Microwave Limb Sounder (MLS) aboard UARS and Aura, and the High Resolution Dynamics Limb Sounder (HIRDLS) also aboard Aura, has allowed the distribution of HNO3 in the stratosphere to be characterized [Gille et al., 1984; Santee et al., 2004; Stiller et al., 2005; Kinnison et al., 2008; Orsolini et al., 2009; Urban et al., 2009].
 The primary chemical source of HNO3is the three-body gas-phase reaction between the hydroxyl radical (OH) and nitrogen dioxide (NO2), while its major sinks are photolysis and oxidation by OH. While HNO3 peaks in the winter lower stratosphere at high latitudes, enhanced abundances of HNO3are commonly observed in the polar mid and upper stratosphere, as revealed by ground-based [de Zafra et al., 1997; McDonald et al., 2000] and satellite observations [Austin et al., 1986; Kawa et al., 1995; Orsolini et al., 2005; Stiller et al., 2005; López-Puertas et al., 2005; Orsolini et al., 2009]. Figure 1 shows the monthly evolution of HNO3 (averaged poleward of 70° in each hemisphere) observed by Aura/MLS (data version 3.3; see Appendix A) from August 2004 to February 2011. Descending tongues of HNO3-enriched air are apparent above 10 hPa during winter in both hemispheres above the main HNO3 layer between approximately 100 and 10 hPa, which leads to a secondary maximum in the HNO3 profile between 10 and 1 hPa.
 The intensity of these recurring HNO3 enhancements between 10 and 1 hPa varies from year to year and appears stronger in the southern hemisphere (SH), where enhanced mixing ratios can reach approximately 6 ppbv. Up to the stratopause, the vertical extent of HNO3is comparable in both hemispheres. These MLS observations are in overall agreement with other instruments, as polar-averaged daily values ranging from 5 to 10 ppbv were reported between 25 and 40 km by SMR [Orsolini et al., 2009] and MIPAS [Orsolini et al., 2005; López-Puertas et al., 2005]. Note that the early winter disappearance of the HNO3 below about 30 hPa in the SH corresponds to the beginning of the polar stratospheric cloud (PSC) formation period, when uptake from the gas phase chemistry exists. In the northern hemisphere (NH) some discontinuities linked to stratospheric sudden warmings occur when the vortex undergoes weakening and displacement.
 Besides in situ stratospheric production, NOx enhancements can arise from the descending flux of NOx-rich air through the stratopause in the winter polar regions, the so-called EPP/NOx “indirect” effect. NOx is produced in the mesosphere and thermosphere during the pervasive low energy auroral electron precipitation, or during more sporadic and higher energy electron precipitation events. Recurrent HNO3 enhancements are associated with the EPP/NOx “indirect” effect; however, the production of NOxas well as the efficiency of the mesosphere-to-stratosphere transport exhibits some variability, especially in the more dynamically active NH. Heterogeneous conversion of dinitrogen pentoxide (N2O5) to HNO3 upon hydrated ion clusters [Böhringer et al., 1983; Kawa et al., 1995] has been suggested as a probable explanation for the slower enhancements following the EPP/NOx “indirect” effect. In this case, the anomalously large NOx concentration leads to excess N2O5 and then to the increased production of HNO3 by conversion of N2O5 on hydrated water cluster (H+·(H2O)n):
 The high-altitude EPP-related source of HNO3is not well accounted for in stratospheric chemistry-transport models (CTM), or in middle atmosphere chemistry-climate models (CCM) as these models do not include the relevant ion chemistry. An exception to the former case is the recent study byReddmann et al. , who incorporated the ion cluster reaction (R1)into a chemistry-transport model. Focusing on the stratospheric NOx and HNO3 descent following the SPEs of October 2003, the HNO3 abundance in their model improved significantly with the inclusion of the afore mentioned reaction. Going beyond the study with prescribed stratospheric dynamics by Reddmann et al. , in this paper we use a free-running CCM, which can produce coupled responses in both chemistry and dynamics. To this end, we use the Whole-Atmosphere Community Climate Model (WACCM) to investigate the role ofreaction (R1) in the formation of HNO3 enhancements through the NOxindirect effect, and more generally its impact on the year-round evolution of stratospheric chemistry. We use an ensemble approach to study the primary chemical effects of the reaction as well as secondary effects due to combined, coupled changes in chemistry and dynamics. The model and the experimental setup are presented inSection 2, whereas the main results are presented in Section 3. Section 4 briefly addresses some of the biases in WACCM, and a discussion and a summary are provided in Section 5.
2. Modeling Methodology
2.1. Whole-Atmosphere Community Climate Model (WACCM)
 We use version 3.1.9 of WACCM developed at the National Center for Atmospheric Research (NCAR) [Garcia et al., 2007]. WACCM covers the vertical region from the surface up to 6.0 × 10−6 hPa (equivalent to approximately 130 km geometric altitude) with 67 hybrid sigma-pressure levels. We have run WACCM with a horizontal resolution equivalent to 1.9° × 2.5° (latitude × longitude) and a time step of 1800 s for the physical parameterizations; this time step is further subdivided as necessary to ensure computational stability. WACCM incorporates a detailed neutral advection and chemistry model for the middle atmosphere, including heating due to chemical reactions; a model of ion chemistry in the mesosphere/lower thermosphere (MLT); ion drag and auroral processes; and parameterizations of shortwave heating at extreme ultraviolet wavelengths and infrared transfer under nonlocal thermodynamic equilibrium conditions. The WACCM chemistry module is derived from the three-dimensional chemical transport Model for OZone And Related chemical Tracers version 3 (MOZART-3) described inBrasseur et al.  and (http://gctm.acd.ucar.edu/gctm/mozart). It solves numerically 51 neutral species, including all the significant members of the Ox, NOx, HOx, ClOx and BrOx chemical families and respective reservoir species. Tropospheric source species such as N2O, CH4, H2O, chlorofluorocarbons (CFCs) and other halogenated compounds are included. Heterogeneous processes on sulfate aerosols and polar stratospheric clouds (liquid binary sulfate, supercooled ternary solutions, nitric acid trihydrate, and water ice), as well as aerosol sedimentation, are represented. The photolysis rates in WACCM for all absorbing species are calculated interactively. Further details regarding photolytic reaction rates and the chemistry module in general are given in Kinnison et al. .
2.2. Model Setup
 To investigate the impact of HNO3 formation by heterogeneous chemistry on hydrated ion clusters, we ran WACCM for 4 successive years. A simulation with an additional effective chemical reaction mimicking (R1) (“perturbed simulation” detailed below) is compared with a control simulation without this reaction. Prescribed climatological sea surface temperatures were used, but the solar forcing and surface emissions correspond to the years 1998 to 2002. Since our aim is to model HNO3 enhancements due to the EPP/NOx“indirect” effect, individual SPEs are not introduced as forcing in our simulations. The diagnostics presented in this paper were carried out after a spin-up period of 6 months.
 Since WACCM is a free-running CCM, the inter-annual dynamical variability can be large. To minimize this internal variability, we used an ensemble approach in which we ran 4 pairs of 4-year control and perturbed simulations. This provides a total of 16 modeled years for each of the perturbed and control simulations. These 4 control and perturbed pairs (i.e., ensemble members) are initiated from different initial conditions, lagged by a day. Except insection 3.1, we only present ensemble-mean results with the statistical significance estimated by the standard Student t-test.
 Monthly mean distributions of minor species are interpreted with the aid of the Transformed Eulerian Mean (TEM) circulation. The total mass-weighted meridional circulation is calculated using monthly averaged dynamical fields. To represent the forcing of the TEM circulation by the resolved waves, we calculated the divergence of the Eliassen-Palm flux (EPFD), also using monthly averaged fields.
2.3. Additional HNO3 Chemistry Introduced in WACCM
 The rate of the chemical reaction (R1) converting N2O5 to HNO3 is controlled by the amount of N2O5 and H+·(H2O)n available. Following de Zafra and Smyshlyaev , we treated the reaction as a pseudo first-order reaction with a rate constantkp= k2(H+·(H2O)n) where k2= 3.5 × 10−10 cm3 s−1, yielding
 There is some uncertainty in the value of the k2 constant. Based on Böhringer et al. , de Zafra and Smyshlyaev  give a range of 8.0 × 10−10 < k2 < 4.0 × 10−12. We used k2= 3.5 × 10−10 cm3 s−1 near the upper limit, as suggested by de Zafra and Smyshlyaev . The reaction, in which one N2O5 molecule is lost and two HNO3 molecules are produced, is processed in the implicit chemical solver for both species. The reaction has a side reaction consuming water vapor to regenerate the hydrated water cluster. However, this secondary effect was neglected here since it contributes as a marginal sink for stratospheric water vapor.
 We introduced a prescribed meridional distribution of hydrated water cluster H+·(H2O)n shown in Figure 2, corresponding approximately to the distribution shown in Figure 5a from Beig et al. . The prescribed time-independent distribution depends upon latitude and height, and is symmetrical around the equator. The densities peak near 4000 molec/cm3at polar latitudes and are tapered to zero at 30°N or S. Under normal conditions, galactic cosmic rays are the primary source of ions in the stratosphere. While varying with geomagnetic latitude and moderately influenced by the 11-year solar cycle and large magnetic storms, the ion distribution exhibits relatively little short-term variation [de Zafra and Smyshlyaev, 2001]. As such, the use of a time-independent distribution of ion clusters serves as a good first order approximation.
3. Simulation Results
3.1. Primary Chemical Response in N2O5 and HNO3
Figure 3 shows the time evolution of differences in HNO3 and N2O5(averaged poleward of 70° in both hemispheres) between the perturbed and the control simulations over the pressure range 100–0.5 hPa. While we will display 4-member ensemble means in the rest of the paper, we show inFigure 3the 4-yearlong simulation for a particular member to demonstrate the recurring conversion of N2O5 to HNO3in the mid-stratosphere in any given simulation. The corresponding distributions from the control simulation are also shown as dashed contour lines. A recurrent annual HNO3 increase in the perturbed simulation appears when and where N2O5 is present. N2O5 is only abundant in the polar regions during the cold and dark season, i.e., during October to February in the NH, and April to August in the SH, when its thermal decomposition and photochemical conversion are slow. The magnitude of HNO3 enhancements in the perturbed simulation are up to 6 ppbv near 20 hPa. In both hemispheres, the recurrent increase in HNO3 in the perturbed simulation, around 30–10 hPa, is found above the maximum volume mixing ratio (vmr) in the control simulation. While the increase coincides in time with the winter maximum in the NH, its occurrence is after the seasonal maximum in the SH. In the lower stratosphere, differences in HNO3vary both in magnitude and sign interanually and are more pronounced in the SH. This characteristic was also evident in the other ensemble members. Such year-to-year differences are only marginally seen in N2O5.
Figure 4 shows the climatological annual cycle of HNO3 in both hemispheres, averaged poleward of 70°, for Aura/MLS (Figures 4a and 4b) based on monthly averages from the years 2004–2011, and for WACCM as an ensemble-mean based on monthly averages from 16 simulated years (Figures 4c and 4d: perturbed simulation, and Figures 4e and 4f: control simulation). The inclusion of the ion cluster reaction significantly improves the HNO3 distribution throughout the year, especially in the SH (Figures 4b, 4d and 4f). The wintertime high-altitude abundance above 10 hPa, albeit improved, remains clearly lower than MLS, as the HNO3 enhancements in WACCM do not extend as high as those measured by MLS or by SMR [e.g., Verronen et al., 2011; Orsolini et al., 2009]. In addition, the implementation of reaction (R1) has better aligned the model seasonal cycle with the observations above 10 hPa. In the NH (Figures 4a, 4c and 4e), the control simulation showed a warm season maximum, but the perturbed simulation now reveals a winter maximum, as observed by MLS. Below about 30 hPa however, HNO3 is already overestimated in the control simulation compared to MLS, and the introduction of the pathway further increases this high bias. Although not included in the model, the MLS HNO3 data also include the effects of solar proton events, e.g., in January 2005 (NH), September 2005 (SH), and December 2006 (NH).
 The large occasional enhancements of HNO3 in the lower stratosphere in the perturbed simulations, as in the first year in Figure 3b, can be explained by a spring break-up of the austral vortex. Such a break-up would lead to a poleward flux of HNO3 from midlatitudes, where in fact HNO3 is more abundant than at high latitudes [e.g., Urban et al., 2009], due to permanent removal of HNO3by sedimentation of polar stratospheric cloud particles in winter. When the break-up occurs in late spring or summer, the meridional gradient of HNO3 has waned, and it leaves a weak signature in HNO3. We will return in Section 4to the issues of potential WACCM biases and the impact of our assumptions regarding the background ion cluster distribution. In that section, we will relate the lack of high-altitude HNO3 formation to a lack of available N2O5 in WACCM, and the high HNO3 bias in the lower stratosphere to an excessive winter descent.
3.2. Response in NOx, N2O and O3
 The distributions of O3 and NOx are governed by chemistry as well as transport processes. The competition between these two processes differs in various regions and seasons, and we will use the ratio of NOx/NOy to separate these processes. The difference in this ratio between the perturbed and the control simulations will deviate from zero (positively or negatively) if chemistry alters the NOx abundance. Figure 5shows the climatological annual cycle of the ensemble-mean difference between perturbed and control simulations, for NOx, NOx/NOy-ratio and O3in the polar regions. The figure includes contours of statistical significance at the 0.9 and 0.75 levels. The corresponding ensemble-mean annual cycle from the control simulation is overlaid as dashed contours.
 We begin by describing the NOx and NOx/NOy-ratio (Figures 5a–5d). Between approximately 5 and 50 hPa, the introduced re-partitioning within the nitrogen family induces a clear deficit in the NOx/NOy-ratio during the cold season when N2O5 is available and R1 is active. This is due to the fact that HNO3is a longer-lived reservoir for NOx, and the regeneration of NOxhas slowed down. The polar-averaged deficit in NOx/NOy-ratio begins in the autumn in both hemispheres and is strongest during early and late winter - characteristics that are more salient in the SH.
 We infer that the NOx positive anomalies in regions where the NOx/NOy-ratio difference is close to zero (e.g., in the NH spring above 25 hPa, and in the SH winter, spring and summer above 50 hPa) are caused by transport and not by the direct influence of nitrogen chemistry. To further elucidate the origin of these anomalies, we examine inFigures 6a and 6bthe climatological annual cycle of the ensemble-mean differences between the two simulations at 10 hPa as polar stereographic maps (40°–90° N inFigure 6a, and 40°–90° S in Figure 6b), for N2O, NOx, O3, and NOx/NOy-ratio. WhileFigure 6 shows every other month, the auxiliary materialFigures S1a and S1b show the entire seasonal cycle. Ensemble-mean 10-hPa geopotential height differences are overlaid on the N2O panels, while ensemble-mean geopotential height for the control and the perturbed simulations are shown on the O3 and NOx/NOy panels, respectively. Figures 6a and 6b nicely show the annual cycle of the anomalies in NOx (or in NOx/NOy), with a pronounced high-latitude deficit in the cold season. Polar, positive N2O anomalies are evident from late winter to summer in the SH, and in spring in the NH, correlated with anomalies in geopotential height. Such positive correlations between N2O and geopotential height reflect the enhanced poleward transport from the N2O-rich lower latitudes in a weakened polar vortex. As N2O is photochemically converted into NO in the sunlit spring and summer, a concomitant increase in NOx occurs. In the SH, the positive anomalies in NOx and N2O are both stronger in magnitude and more prolonged, lasting well into the summer season (e.g., Figures 6a and 6b, first two rows).
 The perturbed run with its stronger poleward advection in spring and summer has in fact brought the WACCM annual climatological cycle of N2O closer to MLS observations. This is demonstrated in Figure 7, which is the analogue of Figure 4 for N2O. In the control simulation, in the SH in particular, the late spring and summer polar abundances were too low and peaked too late compared to MLS. The perturbed run has higher abundances and a steeper seasonal increase, albeit still smaller than in the observations. This is likely due to a tendency of this WACCM version to produce an austral polar vortex that is too long-lived, remaining climatologically strong until January (Figure 6b, third or fourth row). This would inhibit the poleward transport of N2O rich air toward high latitudes. The earlier vortex weakening would also tend to alleviate the low N2O bias in the control run. Because MLS data only covers the years 2004–2011, the observed higher abundances in the NH winter above 30 hPa could reflect the occurrences of several major stratospheric sudden warmings in years 2004, 2006, 2009 and 2010 [e.g., Orsolini et al., 2010; Manney et al., 2008], which would bring N2O-rich air from lower latitudes poleward. The SSWs would not occur at the same frequency in these WACCM runs.
 In summary, in the perturbed case, a pattern of decreased NOx abundance during winter is evident. The weaker spring vortex allows for a stronger poleward transport of N2O, which translates into positive NOx anomalies, stronger and longer lasting in the SH. The recurrent spring and summer NOxexcess in the mid-stratosphere is perhaps the most striking and unanticipated chemical anomaly arising from dynamical-chemical feedbacks, as it occurs when the cold-season ion clusterreaction (R1) is not active anymore.
 These altered NOx levels have important consequences for the O3 abundance. While the O3 response is more complex, Figure 6reveals strong positive anomalies in winter at mid and high latitudes, turning into weaker negative anomalies in the spring-summer (SH) or spring (NH). In both hemispheres, when sunlight returns to the polar regions in the spring, NOx and O3are also clearly anti-correlated between the 30–5 hPa pressure range, where the NOx catalytic cycle dominates (Figures 5a, 5b, 5e, and 5f). Below about 30 hPa, the O3 increase is not accompanied by a negative anomaly in NOx, but the change in the partitioning in NOy implicated by the anomaly in the NOx/NOyratio is indicative of a mitigating effect upon the halogen-induced O3 depletion in this region.
3.3. Response in Zonal-Mean Winds and TEM Circulation
 The results above indicate that reaction (R1) leads to a weakened polar vortex during springtime. However, the origin of the zonal wind anomalies requires a careful analysis. This is further demonstrated in Figure 8, which shows the annual climatological cycle of zonal-mean zonal wind at 10 hPa averaged over the latitude band 60°–70°. The westerlies are weaker in the perturbed simulation during spring in the NH, and from winter to early summer in the SH. The zonal wind anomaly exceeds one standard deviation of the variability in the control simulation during July and November in the SH. These zonal wind decreases are not confined to the 10 hPa level, but rather extends throughout the entire stratosphere and lower mesosphere. This is demonstrated inFigure 9, which shows the monthly mean, ensemble-mean difference in zonally averaged zonal winds during the months where the zonal anomalies inFigure 8are large, i.e., April–May in the NH and November–December in the SH. Overlaid are the corresponding zonal winds from the control simulation. In both hemispheres, these monthly mean negative wind anomalies are statistically significant, especially in late spring, at the 0.75 to 0.90 level depending on latitude or pressure. We now demonstrate that the weaker polar vortex in the perturbed run coexists with a stronger planetary wave forcing diagnosed by the EPFD, and with a stronger TEM circulation. To this end, the ensemble-mean TEM circulation differences between the two simulations are diagnosed alongside the meridional distributions of the long-lived tracer N2O. In Figures 10a, 10c, 10e and 10g, we present the climatological seasonal latitude-height distributions of the relative differences between the two simulations for N2O (in percent and color shading). Note that small absolute differences in N2O can appear as large relative differences. The corresponding N2O distributions from the control simulation are overlaid as black contours. The corresponding TEM circulation is also overlaid as black vectors. The joint examination of the N2O and TEM cross-sections further illustrates the polar enhancement in N2O in the spring, as seen in the maps at 10 hPa (Figures 6a and 6b) and discussed in Section 3.2. In the SH, the enhancement started in winter and persisted during the summer season, before waning in the lower polar stratosphere, around 50 hPa, during the autumn season.
Figures 10b, 10d, 10f and 10h show the difference in TEM circulation between the simulations, as vectors, as well as the difference in the EPFD in color shading. The largest differences in TEM are found in winter above 10 hPa, at midlatitudes in the NH and at mid or high latitudes in the SH. These differences correspond to a strengthening of the TEM circulation, which appears consistent with the negative (westward) anomaly in the EPFD. The strengthened circulation can only be attributed to forcing by planetary waves.
3.4. Link Between the O3 and Dynamical Response
 The depth of the wind anomalies, extending throughout the stratosphere, the altitude of the largest incipient O3 perturbations in autumn and early winter, near 10 hPa (Figure 5e) and the pronounced EPFD anomaly, all point toward amplification of planetary waves as being the lead factor.
 In winter, the NOxcatalytic cycle can be active up to sub-polar latitudes, giving rise to zonally asymmetric O3 anomalies in response to opposite anomalies in NOx (Figure 6). Recent work with chemistry-climate models has demonstrated that using radiatively active O3 tends to alter the generation and propagation of planetary waves, with a strong influence on the circulation throughout the stratosphere, compared to simulations where only zonally averaged O3 is used radiatively [Kirchner and Peters, 2003; Gabriel et al., 2007; Gillett et al., 2009; Waugh et al., 2009; McCormack et al., 2011]. While we compare here two fully coupled simulations, the perturbed run has a higher degree of zonal asymmetry in O3. This is demonstrated in Figures 11a and 11b, which show the ensemble-mean climatological annual cycle of the standard deviation of ozone in the longitudinal direction at midlatitudes (30°S–60°S or 30°N–60°N), for the control simulation (dashed contours) and for the difference (perturbed minus control, color shading). In winter, the difference in the standard deviation is positive throughout the middle atmosphere in the SH, and above 10 hPa in the NH. This enhanced asymmetry in O3gives rise to short-wave heating anomalies, as demonstrated inFigures 11c and 11dat 10 hPa for the months of February (NH) and August (SH). The ozone ensemble-mean differences (color shading) are identical to those shown inFigure 6, but are shown here to be strongly correlated, as expected, to shortwave radiative heating rate anomalies, locally as high as 10%.
 It is not possible to fully disentangle the causality of the co-variability of planetary waves and O3resulting from multiple radiative-dynamical feedbacks between temperature and O3. Neither was it possible even in the simpler case of comparing runs with O3 fully coupled to the radiation to runs with only zonally symmetric O3 radiatively active in papers cited above. Nevertheless, our diagnostics support the argument that the enhanced zonal asymmetries in O3 induced by the initial NOxperturbations in winter middle and sub-polar latitudes would tend to amplify the planetary waves forcing leading to a strengthened TEM circulation, and a vortex that is significantly weakened earlier in the winter.
4. Further Discussion of Potential Biases in WACCM
 Sensitivity simulations with several different distributions of H+·(H2O)n have been completed (not shown), but failed to reproduce the polar HNO3 enhancements reaching the stratopause, as observed in MLS (Figures 1 and 4) and SMR [Orsolini et al., 2009] data. This failure appears to be due to a low bias in the climatological distribution of N2O5 in the upper stratosphere in WACCM, and does not necessarily suggest that the observed HNO3enhancements originate from a different chemical processes. A comparison of the polar-averaged profile of N2O5 (>70° and DJF in the NH, >70° and MJJ in the SH) from both perturbed and control simulations together with MIPAS satellite observations is presented in Figure 12 (see Appendix A for more details about MIPAS data). It reveals a low bias in WACCM in the upper stratosphere, above the levels where the HNO3 perturbations in Figure 3 have decayed, i.e., 3 hPa in the NH, and 6 hPa in the SH. Hence, the potential for HNO3 production in the upper stratosphere is insufficient due to a too low abundance of N2O5 in WACCM. In the upper stratosphere the use of an averaging kernel might help to reduce the differences between model and MIPAS observations. However, such a detailed comparison is beyond the scope of this study. Since the reaction that decomposes N2O5 is strongly dependent upon temperature, a warm model bias would cause the loss to be too large and hence a negative bias in N2O5. A temperature comparison with MIPAS (not shown) indeed reveals a warm bias in the upper stratosphere in WACCM. We also surmise that the source of N2O5, i.e., ultimately NOx, is too low due to weak production from EPP or galactic cosmic rays. Another possibility is that the amount of NOx descending into the stratosphere might be underestimated in WACCM because the elevated levels of NOx created in the lower thermosphere do not reach the top of the descending branch of the TEM circulation [Smith et al., 2011]. From the middle mesosphere down to the mid stratosphere, the TEM circulation is reasonably well captured in WACCM as demonstrated in the transport of carbon monoxide [Kvissel et al., 2012]. Nevertheless, a further exploration of the causes of this upper stratospheric N2O5 deficit in WACCM is beyond the scope of this study. Below these levels, the high bias in the control run has been improved by the introduced N2O5 conversion through reaction (R1), and abundances of about 2 ppbv are closer to the MIPAS observations. In the middle stratosphere, the perturbed run still shows an excess of N2O5 that could be alleviated by a higher conversion rate to HNO3 or a higher background abundance of hydrated ion clusters. We have chosen an idealized distribution of ion clusters, which is still poorly constrained by observations. The reaction rate k2 and its potential dependence on cluster size could also be constrained by better experimental data.
 Another bias in these WACCM version 3.1.9 simulations concerns the anomalous persistence of the austral winter vortex into spring and early summer (e.g., Figures 8 or 9), which causes O3 zonal anomalies in Figure 11 to be off phase with those in observations (Gillett et al., 2009), and the downward TEM circulation in the lower stratosphere to excessively compress and lower the N2O contours (e.g., Figure 7).
5. Summary and Discussion
 In this paper, we have studied the WACCM response to the introduction of a chemical pathway that produces HNO3 by conversion of N2O5 upon hydrated water clusters H+·(H2O)n. The current study is the first of its kind, where this chemical reaction is introduced in a free-running CCM. As such, it is exploratory, and a time-independent distribution of hydrated water clusters has been prescribed in the model simulations. We aim to improve our understanding of the EPP-NOx “indirect effect”; therefore, individual solar proton events are not included in the present study.
 We have used an ensemble approach, carrying out four pairs of control and perturbed simulations each lasting four years, and addressed the statistical significance of the results. The main focus has been on annually re-occurring chemical signatures in the polar regions. Despite the two model simulations being similar in all other aspects, it is not surprising that there are differences in the distributions of trace gases that can be attributed to changes in transport. The introduced chemical pathway for HNO3 alters the internal partitioning of the NOy family during winter months in both hemispheres, and the perturbed HNO3 ultimately impacts the distribution of O3. Through O3changes, both temperature and dynamics are affected allowing for complex chemical-dynamical feedbacks lasting beyond the cold season when thereaction (R1) is active.
 Statistically significant changes in the climatological distributions of constituents from the nitrogen family or O3 compared to the control run include: i) a cold season production and loss of HNO3 and N2O5, respectively, ii) a cold season decrease and increase in NOx/NOy ratio and O3, respectively, in the polar regions of both hemispheres, and iii) a NOx and N2O polar increase in spring-to-summer in the SH, and in spring in the NH. Associated with the later (iii) is an O3 decrease, which is only statistically significant in the SH.
 Through the inclusion of the ion cluster reaction, we see an improved seasonal evolution of modeled HNO3compared to satellite observations from MLS, esp. regarding the seasonal cycle in the polar mid-stratosphere. However, a prior overestimation at lower levels in the NH is worsened by implementingreaction (R1). The N2O distribution is also improved in the perturbed run compared to the MLS observations, esp. in the spring-to-summer in the SH. Our comparison of WACCM versus MLS observations also include O3, shown in auxiliary material Figure S2, but discrepancies in the lower stratosphere in winter and spring are difficult to ascribe to this effect only and not to other model processes, such as the anomalous vortex persistence or heterogeneous chemistry on PSCs.
 The seasonal evolution of zonal-mean zonal winds and temperature is not identical in the two simulations, which we argue is caused by strengthened planetary waves induced by the midlatitude zonal asymmetries in O3and short-wave heating. The strengthening in the perturbed runs ultimately results from the influence of the NOx cold season deficit which leads to a higher degree of zonal asymmetry in O3at mid and sub-polar latitudes. In the SH, the weakening of the zonal-mean winds indeed begins in winter but is statistically most significant in late spring and summer, and in the NH, it is significant in the spring.
 The well-studied EPP-NOx “indirect” effect allows O3 depletion in the spring through the catalytic NOx cycle [Jackman et al., 2005, 2008; Randall et al., 2006]. Our study uncovers a new way for the EPP-NOx “indirect” effect to be further amplified in a model. The springtime polar NOx increase arises from a weaker polar vortex that allows more transport of N2O polewards, instead of in situ production of NOx. On the other hand, the WACCM biases, e.g., in N2O5, makes reaction (R1) ineffective above approximately 10 hPa, as shown in Figure 4. If WACCM would allow higher levels of NOx to reach the stratosphere from above [Smith et al., 2011], the EPP-NOx “indirect” effect would yet be further amplified by the mechanism we propose here. As mentioned earlier, heterogeneous conversion of N2O5 into HNO3 onto volcanic aerosols can also occur in the lower stratosphere [Bekki et al., 1997], and the question arises whether analogous dynamical couplings would be at play when aerosol loading is large enough. The issues of N2O5 and dynamical biases in WACCM 3.1.9 make a compelling case for investigating the dynamical feedbacks of the reaction (R1)in other chemistry-climate models. Nevertheless, it is quite remarkable that the inclusion of this chemical pathway has changed the magnitude and the seasonal march of the stratospheric jet. Such change highlights the importance of NOx in modulating ozone abundances, and of EPP processes for the entire middle atmosphere.
 In this study, we use the version 3.3 of Microwave Limb Sounder (MLS) L2 data. All the data were screened according to the instruction given in the MLS Data Quality and Description Document [Livesey et al., 2011]. The MLS instrument is mounted on NASA Aura satellite, launched in 2004 into a sun-synchronous near-polar orbit. It measures the thermal emission of more than 15 trace species in the microwave range, in both day and night conditions, covering latitudes up to 82° on each orbit. These measurements can be inverted into altitude profiles with a range between 215 and 0.01 hPa (≈10–80 km). The quality of the data varies considerably depending upon the height of the retrieval and the component retrieved. For HNO3, the standard recommended altitude range is between 215 and 1.5 hPa (≈10–50 km) [Livesey et al., 2011, chapter 3.11]. The vertical resolution in the middle/upper stratosphere is 4–5 km up to about 1 hPa, where it starts to degrade rapidly down to 12 km around 0.1 hPa. Retrieved mixing ratios at 1 hPa are also scientifically useful during episodes when production of HNO3increases and the signal-to-noise ratio improves. These episodes can for instance happen during strong solar proton events [Verronen et al., 2008, 2011], or during winters when an enhanced downward circulation advects NOx-rich air from the mesosphere into the stratosphere.
 We have also used 3 years of N2O5observations by MIPAS over the years 2007–2009. The MIPAS IMK-IAA data set from the Institute for Meteorology and Climate Research has been used (Atmospheric Trace Gases and Remote Sensing website:http://www.imk-asf.kit.edu). We have used the visibility flag and the averaging kernel (values less than 0.03 is removed) to remove bad data and data gaps.
 O.K.K., Y.O.R. and F.S. were supported by the Norwegian Research Council NORKLIMA program (Project Arctic_Lis, project number #178629). O.K.K. has had a three-week stay at NCAR supported by The Norwegian Research School in Climate Dynamics (ResClim). Work at the Jet Propulsion Laboratory, California Institute of Technology, was done under contract with NASA. The authors would like to acknowledge Varavut Limpasuvan and Amund Søvde for useful comments.