Nitriﬁcation of the lowermost stratosphere during the exceptionally cold Arctic winter 2015–2016

. The Arctic winter 2015–2016 was characterized by exceptionally low stratospheric temperatures, favouring the formation of polar stratospheric clouds (PSCs) from mid-December until the end of February down to low stratospheric altitudes. Observations by GLORIA (Gimballed Limb Observer for Radiance Imaging of the Atmosphere) on HALO (High Altitude and LOng range research aircraft) during the PGS (POLSTRACC–GW-LCYCLE II–SALSA) campaign from December 2015 to March 2016 allow the in-vestigation of the inﬂuence of denitriﬁcation on the lowermost stratosphere (LMS) with a high spatial resolution. Two-dimensional vertical cross sections of nitric acid (HNO 3 ) along the ﬂight track and tracer–tracer correlations derived from the GLORIA observations document detailed pictures of wide-spread nitriﬁcation of the Arctic LMS during the course of an entire winter. GLORIA observations show large-scale structures and local ﬁne structures with enhanced absolute HNO 3 volume mixing ratios reaching up to 11 ppbv at altitudes of 13 km in January and nitriﬁed ﬁlaments persisting until the middle of March. Narrow coherent structures tilted with altitude of enhanced HNO 3 , observed in mid-January, are interpreted as regions recently nitriﬁed by sublimating HNO 3 -containing particles. Overall, extensive nitriﬁcation of the LMS between 5.0 and 7.0 ppbv at potential temperature levels between 350 and 380 K is estimated. The GLORIA observations are compared with CLaMS (Chemical Lagrangian Model of the Stratosphere) simulations. The fundamental structures observed by GLORIA are well reproduced, but differences in the ﬁne structures are diagnosed. Further,

Abstract. The Arctic winter 2015-2016 was characterized by exceptionally low stratospheric temperatures, favouring the formation of polar stratospheric clouds (PSCs) from mid-December until the end of February down to low stratospheric altitudes. Observations by GLORIA (Gimballed Limb Observer for Radiance Imaging of the Atmosphere) on HALO (High Altitude and LOng range research aircraft) during the PGS (POLSTRACC-GW-LCYCLE II-SALSA) campaign from December 2015 to March 2016 allow the investigation of the influence of denitrification on the lowermost stratosphere (LMS) with a high spatial resolution. Twodimensional vertical cross sections of nitric acid (HNO 3 ) along the flight track and tracer-tracer correlations derived from the GLORIA observations document detailed pictures of wide-spread nitrification of the Arctic LMS during the course of an entire winter. GLORIA observations show largescale structures and local fine structures with enhanced absolute HNO 3 volume mixing ratios reaching up to 11 ppbv at altitudes of 13 km in January and nitrified filaments persisting until the middle of March. Narrow coherent structures tilted with altitude of enhanced HNO 3 , observed in mid-January, are interpreted as regions recently nitrified by sublimating HNO 3 -containing particles. Overall, extensive nitrification of the LMS between 5.0 and 7.0 ppbv at potential temperature levels between 350 and 380 K is estimated. The GLO-RIA observations are compared with CLaMS (Chemical Lagrangian Model of the Stratosphere) simulations. The fundamental structures observed by GLORIA are well reproduced, but differences in the fine structures are diagnosed. Further, CLaMS predominantly underestimates the spatial extent of HNO 3 maxima derived from the GLORIA observations as well as the overall nitrification of the LMS. Sensitivity simulations with CLaMS including (i) enhanced sedimentation rates in case of ice supersaturation (to resemble ice nucleation on nitric acid trihydrate (NAT)), (ii) a global temperature offset, (iii) modified growth rates (to resemble aspherical particles with larger surfaces) and (iv) temperature fluctuations (to resemble the impact of small-scale mountain waves) slightly improved the agreement with the GLORIA observations of individual flights. However, no parameter could be isolated which resulted in a general improvement for all flights. Still, the sensitivity simulations suggest that details of particle microphysics play a significant role for simulated LMS nitrification in January, while air subsidence, transport and mixing become increasingly important for the simulated HNO 3 distributions towards the end of the winter.

Introduction
The processes of denitrification and nitrification are wellknown phenomena occurring in the polar winter stratosphere . They involve the condensation, growth, sedimentation and sublimation of nitric acid (HNO 3 )-containing polar stratospheric cloud (PSC) particles and result in an irreversible vertical redistribution of HNO 3 . Denitrification is known to affect polar winter ozone loss . Denitrification at higher lay-M. Braun et al.: Nitrification of the lowermost stratosphere during cold Arctic winter 2015-2016 ers (i.e. around 16 to 22 km) attenuates fast deactivation of catalytically active chlorine species into the reservoir species chlorine nitrate (ClONO 2 ). However, chlorine deactivation can be fostered at lower layers enriched in HNO 3 (i.e. nitrified) by evaporating nitric acid trihydrate (NAT) particles (Fischer et al., 1997). Observational evidence for particles other than NAT involved in denitrification is sparse (e.g. Tabazadeh and Toon, 1996;Kim et al., 2006). Nitrification of the lowermost stratosphere is of particular interest since the chemical budget of reactive nitrogen (NO y ) and, thereby, its possible effects on ozone are modified in a region important for the radiative budget of the atmosphere (Riese et al., 2012).
While fundamental processes in connection with PSCs are well understood, there are still uncertainties concerning the formation of NAT particles and their characteristics that influence the processes of denitrification and nitrification. Chemistry-transport and global-chemistry models including simplified microphysical properties of NAT are generally successful in simulating denitrification of the polar winter stratosphere (Carslaw, 2002;Grooß et al., 2005Grooß et al., , 2014Khosrawi et al., 2017;Zhu et al., 2017, and references therein). However, parametrizations resulting in agreement with observed size distributions of NAT particles, particularly extremely large NAT particles, and reproducing fine structures of observed denitrification patterns remain an issue (e.g. Molleker et al., 2014;Woiwode et al., 2014).
Hemispheric differences in nitrification are observed due to different conditions in Antarctic and Arctic winter vortices. In the Antarctic, cold and stable vortices result in widespread PSC coverage and denitrification over wide vertical ranges. PSCs are observed less frequently in the Arctic, and the degree of denitrification varies from winter to winter Pitts et al., 2018, and references therein). Increasing greenhouse gas emissions could lead to lower stratospheric temperatures (e.g. Rex et al., 2006), which likely cause stronger denitrification of the Arctic stratosphere.
Numerous studies addressed denitrification in Arctic winters, especially in cold winters, e.g. 1994-1995-2000(Popp et al., 2001(Jin et al., 2006), 2009(Khosrawi et al., 2011Woiwode et al., 2014(Sinnhuber et al., 2011) and 2015(Khosrawi et al., 2017. While most of these studies focused on the denitrification at altitudes higher than roughly 15 km, less attention was paid to the associated nitrification of lower layers. Dibb et al. (2006) reported nitrification at potential temperature levels above 340 K (around 12 km) at the end of January 2005. Further, Hübler et al. (1990) interpreted enhanced mixing ratios of up to 12 ppbv at altitudes between 10 and 12.5 km in the Arctic winter 1988-1989 as resulting from nitrification. Tuck et al. (1997) found indications for nitrification in the Antarctic at levels above 400 K in the Antarctic winter 1994. Particularly, nitrification of the Arctic lowermost stratosphere (LMS) has hardly been investi-gated. This is due to the fact that cold winters with strong denitrification were rare events in the Arctic stratosphere in the past and that the observational capabilities to resolve nitrification of the LMS with sufficient coverage and vertical resolution are sparse. For example, limb sounders, like MLS (Microwave Limb Sounder;Waters et al., 2006) or MIPAS (Michelson Interferometer for Passive Atmospheric Sounding; Fischer et al., 2008) typically have vertical resolutions of around 3-5 km making it difficult to resolve fine-scale structures of NO y redistribution.
The process of vertical HNO 3 redistribution is very sensitive to temperature. NAT particle nucleation may begin as soon as temperatures fall below NAT equilibrium temperature T NAT . However, clear evidence of the precise nucleation conditions of NAT particles is still lacking. NAT particles are nucleated heterogeneously with low number densities on foreign nuclei such as meteoritic dust particles . Below T NAT , these particle grow and sediment downward, and they evaporate as temperatures rise above T NAT . A simulation of this process is challenging as it depends both on the nucleation parametrization and on the precise reproduction of the temperatures around T NAT . This is especially the case during the onset of this process. At a later time, to nucleate new NAT particles in denitrified air, lower temperatures are needed because of the already decreased HNO 3 mixing ratios. This results in a maximum potential denitrification for a given temperature. Since both the nucleation process and the mesoscale temperature modulations (e.g. by gravity waves) are not well known, it is especially difficult to simulate the small-scale structure of HNO 3 during the onset period.
Here, we present observations of nitrification of the LMS in the unusually cold Arctic winter 2015-2016 by the airborne limb-imaging Fourier transform infrared (FTIR) spectrometer GLORIA (Gimballed Limb Observer for Radiance Imaging of the Atmosphere, Friedl-Vallon et al., 2014;Riese et al., 2014). In that winter, an extraordinarily cold and stable polar vortex (Matthias et al., 2016;Manney and Lawrence, 2016) promoted a long-lasting PSC phase from mid-December until the end of February with a large vertical extent Voigt et al., 2018) reaching down into the LMS.
Using the GLORIA observations and simulations by the Chemical Lagrangian Model of the Stratosphere (CLaMS; Grooß et al., 2014, references  Thereby, we attempt to quantify the observed nitrification, which is particularly difficult because the LMS composition is influenced by air masses originating from the Arctic vortex, the extra-vortex stratosphere and the troposphere (Werner et al., 2010;Gettelman et al., 2011;Krause et al., 2018). ClONO 2 also contributed significantly to lowermost stratospheric total NO y during the Arctic winter 2015-2016. This aspect is addressed within a separate study by Johansson et al. (2019). Here we focus only on gas-phase HNO 3 , which is the direct product of nitrification by sublimating NAT particles.
We compare the GLORIA data with simulations by the CLaMS. To test how well different parametrizations within the same model reproduce the GLORIA observations, four sensitivity studies were performed. Those sensitivity simulations investigated the impact of (i) enhanced sedimentation rates in case of ice supersaturation, (ii) a global temperature offset, (iii) modified growth rates and (iv) temperature fluctuations.
2 Aircraft campaign and data

POLSTRACC-GW-LCYCLE II-SALSA
The GLORIA observations analysed in this study were obtained during the combined POLSTRACC (POLar STRAtosphere in a Changing Climate), GW-LCYCLE II (Gravity Wave Life Cycle Experiment) and SALSA (Seasonality of Air mass transport and origin in the Lowermost Stratosphere using the HALO Aircraft) campaigns (PGS). Starting from Oberpfaffenhofen, Germany, or from Kiruna, Sweden, 18 research flights were carried out by the German research aircraft HALO (High Altitude and LOng range research aircraft) between December 2015 and March 2016. The flights probed an entire winter period in the LMS at high northern latitudes. For this study five research flights between December and March were used. The selection of the flight data was based on data availability and scientific requirements. Data availability was limited to flight sections where GLO-RIA was operated in the high spectral resolution "chemistry mode"  used in this study (see Sect. 2.2) and sufficiently cloud-free conditions allowing for the retrieval of HNO 3 . From the scientific point of view, flights with long continuous chemistry mode measurements were chosen to show how patterns in the lowermost stratospheric HNO 3 distribution change during the winter. We furthermore focus on flights in January, where PSCs reached down to the LMS and where the most notable changes are found in the observed HNO 3 distributions. Since we use ozone as stratospheric tracer to quantify nitrification, flights in January are preferable since only little chemical ozone loss was diagnosed at this time of the winter when compared to February and March (see Johansson et al., 2019). Further GLORIA chemistry mode observations can be found in the supplementary information of Johansson et al. (2018a) and at the HALO Database (https://halo-db.pa.op.dlr.de/, last access: 4 November 2019).

GLORIA
GLORIA is an airborne infrared limb imaging spectrometer . During PGS, GLORIA has been operated on board the HALO aircraft and pointed to the right-hand side of the flight path. GLORIA combines a Michelson interferometer with an imaging HgCdTe detector which records 128 vertical and 48 horizontal interferograms simultaneously. All interferograms are transformed into spectra. The spectra from horizontal detector rows are averaged for noise reduction prior to the atmospheric parameter retrieval . In high spectral resolution mode, which is used in this study, the spectrometer covers the range from 780 to 1400 cm −1 with a spectral sampling of 0.0625 cm −1 . For the retrieval, the radiative transfer code KOPRA (Karlsruhe Optimized and Precise Radiative transfer Algorithm; Stiller et al., 2002) and the inversion tool KOPRAFIT (Höpfner et al., 2001) were used. Estimated uncertainties of the GLORIA retrieval results are typically 1-2 K for temperature and 10 %-20 % for trace gases. Typical vertical resolutions of the retrieved profiles are about 400 m at flight altitude and decrease to about 1000 m around the lowest tangent points. A detailed description and validation of the dataset used in this study is given by Johansson et al. (2018a).

CLaMS
The Chemical Lagrangian Model of the Stratosphere (CLaMS) (McKenna, 2002a, b) is a chemistry transport model based on trajectory calculations for an ensemble of air parcels. CLaMS includes modules simulating Lagrangian trajectories, mixing, chemical processes and Lagrangian particle sedimentation. The CLaMS simulations used here were performed with a special setup for the POLSTRACC campaign with a horizontal resolution of about 100 km and a vertical resolution of about 500-900 m in the lower stratosphere above 10 km altitude decreasing to about 2 km below 9 km altitude. Further, this configuration includes a comprehensive stratospheric chemistry as described by Grooß et al. (2014). The simulations were performed for the entire winter and were based on meteorological wind and temperature data from the ECMWF ERA interim reanalysis (Dee et al., 2011) employing a horizontal resolution of 1 degree × 1 degree and a time step of 6 h. To simulate processes connected to NAT particles, particle parcels are implemented (Grooß et al., 2005. Particle size and number concentration are assigned to each particle parcel so that various particle parcels in one air parcel determine the particle size distribution. NAT and ice nucleation is temperature and saturation dependent and is parameterized by the scheme by Hoyle et al. (2013) and Engel et al. (2013), respectively. Particle growth and evaporation are calculated along particle trajectories based on Carslaw (2002) assuming the characteristics of spherical particles . Comparisons with PSC observations  show that the parametrization of nucleation and sedimentation of NAT and ice particles in CLaMS is capable to reproduce the main features of PSC observations. Also, vortex averages of the vertical redistribution of HNO 3 and H 2 O have been reproduced .

GLORIA vertical cross sections of atmospheric parameters
The GLORIA retrieval results in vertical profiles of atmospheric parameters. These vertical profiles are combined to two-dimensional quasi-vertical cross sections along the flight paths and show mesoscale atmospheric structures (Johansson et al., 2018a). Since the observations are performed in limb mode, the distance of the tangent points (i.e where the major information about atmospheric parameters stems from) gradually increases from the observer for the lower limb views. This is reflected by the tangent point distributions discussed in Sect. 4.1 to 4.3. The GLORIA data are filtered for cloudaffected observations, and data points with a vertical resolution worse than 2 km or above flight altitude are neglected for further analysis.

Simulated cross sections from CLaMS
For comparison with GLORIA, the CLaMS data were interpolated to the retrieval grid geolocations, characterized by altitude, latitude, longitude and time of the tangent points. The temporal interpolation with respect to atmospheric dynamics is performed by trajectory calculations. CLaMS output is typically saved daily at 12:00 UTC. Therefore forward trajectories are calculated for points between 00:00 and 12:00 until 12:00 UTC. The corresponding 12:00 UTC volume mixing ratio is then assumed as concentration of the original geolocation based on the assumption that chemical and physical changes in volume mixing ratios during the time of the trajectory calculations are negligible for the chemical species considered here. For points between 12:00 and 00:00 UTC backward trajectories are calculated analogously.

Identification of sub-vortex air
The altitude range of GLORIA observations in this study typically lies within the LMS, ranging from the tropopause to the 380 K isentrope (see Werner et al., 2010, and references therein). It has to be pointed out that robust identification of vortex air in the LMS is not possible due to dynamical disturbances, transport and in-mixing of air masses from different origins. In fact, the sub-vortex region in the LMS has a more filamentary character and is affected by interaction with air masses from the extra-vortex stratospheric overworld, the extra-tropical transition layer (ExTL) and the troposphere (Gettelman et al., 2011). Two filters have been applied to select data points associated with the sub-vortex region.
The first filter applied is the criterion by Nash et al. (1996) at the 370 K isentrope and is based on the potential vorticity (PV) field obtained from the MERRA-2 reanalysis (Gelaro et al., 2017). Grid points with latitude-longitude pairs outside the polar vortex at the potential temperature (θ ) level of 370 K are classified as non-vortex points. Secondly, data are filtered by scaled potential vorticity (sPV) calculated from MERRA-2 reanalysis data with a threshold of 1.2×10 −4 s −1 . sPV is calculated by dividing the potential vorticity (PV) by ∂θ/∂p to obtain similar PV ranges for all isentropic levels that are investigated (Manney et al., 1994;Dunkerton and Delisi, 1986). Therefore this filter takes the altitude information of the grid points into account. Data points in the tracer correlations (see below) are attributed to sub-vortex air, if both criteria are met.

Quantification of nitrification based on tracer-tracer correlations using relative normalized frequency distributions
To quantify nitrification in the LMS tracer-tracer correlations of HNO 3 and O 3 associated with sub-vortex air are analysed.
Here, we use ozone as an approximation of a passive reference tracer, since ozone is well accessible with GLORIA and shows a sufficient vertical gradient in the LMS region. The choice of ozone as a passive tracer is based on the assumption that ozone depletion is small in January as the air is hardly exposed to sunlight. The model study by Khosrawi et al. (2017) supports this assumption. Two aspects can affect the correlation: (1) mixing with extra-vortex air masses not affected by nitrification would lead to an underestimation of HNO 3 introduced into the LMS by nitrification and (2) potential ozone depletion would shift higher HNO 3 mixing ratios to lower ozone values, thus enhancing estimated nitrification.
As correlation scatter plots of measured data for several flights are difficult to assess due to the large number of individual points, estimates of relative normalized frequency distributions (RNFDs), as described by Eckstein et al. (2018), are used in this study. This method calculates a scaled twodimensional histogram on a volume mixing ratio grid. In this study a grid of 0.070 ppmv O 3 × 0.35 ppbv HNO 3 is chosen, which is motivated by the total estimated error of the trace gases (Johansson et al., 2018a). GLORIA data points with a calculated relative error larger than 20 % are neglected in this study. Besides a clearer presentation of the data points, RNFDs filter out single data points with very high HNO 3 volume mixing ratios. Therefore, in the context of a challenging vortex identification, this method offers an additional filter, as single data points that are differing significantly and are erroneously identified as vortex air are filtered out. Here it has to be pointed out that also local non-erroneous points with very high HNO 3 values within the vortex are filtered out applying this method. However, in this study we aim to quantify the overall nitrification of the LMS, while local nitrification is highly inhomogeneous and can reach significantly higher values. The RNFD contour line used for quantification in this study includes points within 2 % of the histograms maximum density. An example of a RNFD for the HNO 3 -O 3 correlation during flight 8 is given in Fig. 1. We also show alternative isolines including 1 % and 4 % of the maximum density of the histogram to visualize weaker and stronger thresholds for statistical outliers. In all cases, the isolines show a similar pattern in general. However, stronger threshold values limit the vertical range of the analysis and filter out valuable significant data points.

GLORIA observations and CLaMS simulations of selected flights from January to March 2016
To investigate how the observed HNO 3 distribution is affected by nitrification, three research flights have been selected. The first flight was carried out on 20 January 2016 during the coldest phase of the winter (Manney and Lawrence, 2016), with PSCs ranging down to flight level. The second flight took place on 31 January 2016 after a strong PSC phase. The last flight was carried out on 18 March 2016 at a late state of the winter -about 2 weeks after the final warming (5-6 March; Manney and Lawrence, 2016). The observed patterns in the HNO 3 distributions are compared with the observed patterns in the ozone distribution. Since ozone and HNO 3 are effected by the same dynamical processes, the different patterns in the observed distributions are likely caused by processes that effect only one species (i.e. nitrification due to sublimation of NO y -containing particles sedimented from higher altitudes). Therefore, the local HNO 3 enhancements seen in comparing adjacent air masses at a given height level and the deviations of their pattern from the pattern seen in the ozone distribution are interpreted qualitatively as a result of nitrification.

Flight 8 on 20 January 2016
Applying the Nash criterion on 20 January 2016, a relatively coherent vortex region is found, with all GLORIA tangent points located inside the vortex region at θ = 370 K (Fig. 2a). Since clouds complicate a robust trace gas retrieval, a number of GLORIA observations were removed by cloud filtering. As a consequence, only limited GLORIA nitric acid data are available in flight sections with sufficiently transparent conditions (Fig. 2b). Further, particulate NO y (i.e. the difference between measured total NO y and gas-phase NO y ) was simultaneously measured in situ by using a chemiluminescence detector in combination with a converter for NO y species (Stratmann et al., 2016). Similar observations have also been made during other aircraft campaigns in the Arctic (Northway et al., 2002).
The measurements shown in Fig. 2c are based on the subisokinetic sampling of particles with a forward-looking inlet (e.g. Fahey et al., 1989;Ziereis et al., 2004). Particles larger than a few tenths of a micrometre are sampled with enhanced efficiency and are detected as gas-phase equivalent NO * y . The efficiency factor depends, among others, on the ratio between aircraft and sampling velocity, pressure, temperature, and particle size (e.g. Fahey et al., 1989;Feigl et al., 1999). The maximum enhancement factor may be achieved for particle sizes larger than about 10 µm and is on the order of several tens, depending on the actual combination of the above mentioned parameters. Here, only equivalent NO * y that was not corrected for enhancement is shown as a qualitative proxy for particulate HNO 3 . As absolute values cannot be obtained, we only use the data as a proxy for condensed HNO 3 -containing particles present at flight altitude. The in situ data clearly confirm the presence of HNO 3 -containing PSC particles at flight altitude and mainly between waypoints A and B in close vicinity to gas-phase HNO 3 maxima detected by GLORIA.
The vertical cross sections of O 3 and HNO 3 volume mixing ratios along the HALO flight track derived from GLO-RIA are depicted in Fig. 2b and d. The ozone distribution shows increasing volume mixing ratios with altitude reach- ing 1.1 ppmv at 13 km. The observed O 3 volume mixing ratios vary only moderately at fixed altitude levels during the whole flight in agreement with the location of the measurements within the vortex and the homogeneity inside the vortex. Compared to the ozone distribution, the HNO 3 volume mixing ratios are varying significantly at fixed altitudes. The HNO 3 distribution shows high HNO 3 volume mixing ratios particularly in the flight segment between the waypoints B and C, reaching up to 8 ppbv at a flight altitude of 13 km compared to 3 ppbv observed in adjacent air around waypoint C. Differences of this maximum structure from the corresponding O 3 distribution are interpreted qualitatively as nitrified air. Further, local maxima are forming coherent structures tilted with altitude and are observed down to 11 km in that flight segment. In addition, small-scale fine structures with enhanced HNO 3 volume mixing ratios appear between 15:30 and 17:00 UTC and reach down to 10 km. The pattern of continuous and slightly tilted vertical bands differ significantly from the ozone distribution and show enhanced values compared to adjacent air masses at a given height level, thus suggesting their formation by redistribution of HNO 3 .
To test this hypothesis, we show normalized GLORIA HNO 3 and O 3 data along selected isentropes in Fig. 2f. Normalization factors are chosen in a way such that the mixing ratios of both gases are close to 1 in air masses which are not affected by nitrification. In particular, HNO 3 volume mixing ratios (in ppbv) were multiplied by a constant factor of 0.5, and O 3 volume mixing ratios (in ppmv) were multiplied by a constant factor of 1.5 to result in the shown unitless normalized mixing ratios. The same factors were applied for the subsequent flights. In unaffected air masses, the normalized mixing ratios of these gases are expected to show the same pattern. In nitrified air masses, locally enhanced HNO 3 and different modulations are expected relative to O 3 . In fact, such local maxima in the HNO 3 mixing ratios can be identified at 340 K around 15:45, 16:25, 18:20 and 18:40 UTC, and are more pronounced at 350 K around 18:35 and after 19:10 UTC. The maxima clearly coincide with local maxima seen in the HNO 3 cross sections. Thus, the simultaneous presence of confined local gas phase HNO 3 maxima in the GLORIA data and HNO 3 -containing particles detected in situ in regions close to the equilibrium temperature of NAT, well above the equilibrium temperature of ice (see GLORIA temperature data shown in Johansson et al., 2018a), suggests that an ongoing nitrification process was probed.
The vertical cross section of HNO 3 volume mixing ratios modelled by CLaMS is shown in Fig. 2e. HNO 3 volume mixing ratios reach maximum values of 8 ppbv at flight altitude in the flight segment between B and C. While maximum HNO 3 volume mixing ratios in this flight are well represented by CLaMS, slight differences in the location of the maximum occur. Overall HNO 3 mixing ratios are clearly underestimated by CLaMS, and CLaMS mostly misses the vertical fine structure.

Flight 12 on 31 January 2016
At the end of January 2016, applying the Nash criterion at θ = 370 K, a more disturbed lower vortex region is observed. As shown in Fig. 3a, a large region between Greenland, central Europe and northern Siberia fulfilled the vortex criterion. However, filaments of lower PV are found from Greenland to southern Scandinavia and around the eastern rim of Scandinavia. Flight 12 was carried out starting and ending in Kiruna on 31 January and intersected several times with filaments outside the vortex.
Measured cross sections of O 3 and HNO 3 volume mixing ratios are depicted in Fig. 3b and c. Except for flight segments between 08:40 and 09:20 UTC, as well as between 11:30 and 11:50 UTC, that are associated with the vortex edge or non-vortex air (indicated by grey shading in Fig. 3be), ozone values increasing with height are observed. Compared to the ozone values only varying slightly along an isentrope, the HNO 3 volume mixing ratios show larger variations at levels of constant potential temperature and suggest local enhancements by nitrification. Again, the analysis of normalized HNO 3 and O 3 along the selected isentropes clearly shows enhanced and more variable HNO 3 mixing ratios relative to O 3 inside air masses attributed to the sub-vortex region (Fig. 3e, for normalization see Sect. 4.1). During this flight, high local maximum values of HNO 3 well above 10 ppbv are found between 13:30 and 14:30 UTC in the GLORIA observations. Fig. 3d. When compared to GLORIA, locally more confined and weaker HNO 3 maxima are modelled after 12:15 UTC reaching down to altitudes of 12 km. Maximum HNO 3 volume mixing ratios are found at flight altitude showing narrow peaks up to 10 ppbv at 12:15 UTC and 8 ppbv at 12:50 UTC, at waypoints C and D. Overall, CLaMS shows a higher spatial variability and underestimates the maximum values during large parts of the flight.

Flight 21 on 18 March 2016
For the flight on 18 March 2016, the PV distribution shows a patchy pattern of regions inside the remains of the vortex according to Nash et al. (1996) around Scandinavia, with the GLORIA observations being located partly inside and outside these regions (Fig. 4a).
The measured O 3 distribution (Fig. 4b) shows increasing values with altitude and reaches values of 1.2 ppmv at flight level. Ozone values along the isentropes vary only slightly. The measured HNO 3 distribution (Fig. 4c) shows a slightly higher variability along the isentropes with local maxima for altitudes higher than 9 km reaching maximum values of up to 6 ppbv at flight altitude embedded in background values of 2 to 3 ppbv. Filamentation and mixing following the earlier vortex break-up (Manney and Lawrence, 2016) resulted in less spatial variability when compared to the previous flights. Since flight 21 was carried out after the vortex break-up and the correlation of HNO 3 and O 3 was altered by in-mixing of extra-vortex air, this flight is included in the model comparisons in Sect. 6 but not in the quantification of nitrification of the LMS in Sect. 5. However, well-defined local maxima qualitatively attributed to result from nitrification by the comparison with the ozone distribution still persisted in this late stage of the winter. The analysis of normalized HNO 3 and O 3 along isentropes shows enhanced and slightly more variable HNO 3 mixing ratios relative to O 3 in the sub-vortex region and its vicinity, thus supporting that these patterns are remnants of nitrification (Fig. 4e, for normalization see Sect. 4.1). CLaMS (Fig. 4d) shows HNO 3 volume mixing ratios for altitudes higher than 9 km corresponding well with GLORIA observations. Maximum values of locally 6 ppbv are modelled around 12:30 UTC at 13 km. Again, CLaMS slightly underestimates overall HNO 3 mixing ratios when compared to GLORIA.

Quantification of nitrification of the LMS from
December 2015 to January 2016 To quantify nitrification in the LMS from December 2015 to January 2016 we applied the method described in Sect. 3.4 using selected flights in this period. Figure 5a depicts the distributions for flights 5, 6, 8 and 12 derived from GLORIA observations. Flight 5 (only limited GLORIA data available, see Appendix A) was carried out on 21 December 2015, at the beginning of the winter, with no significant hints to nitrification. Therefore this flight was chosen as the early winter reference. Due to a limited number of points associated with vortex air and since sub-vortex and non-vortex data points show a compact correlation for flight 5, non-vortex points are also included here to extend the available data. For all other flights, only data points associated with vortex air are used. Flight 6 (see Appendix A) covered a broad range of latitudes in the sub-vortex region and below PSCs . HNO 3 volume mixing ratios for flight 5 range up to 3.2 ppbv with an approximately linear relationship to the observed ozone values. In case of flight 6, enhanced HNO 3 volume mixing ratios compared to flight 5 are observed for all ozone values. Flight 8 shows a similar enhancement throughout the whole range of ozone mixing ratios observed. While the enhancement is similar to flight 6, minimum HNO 3 values for ozone values higher than 0.7 ppmv are higher than those for the flights before. For flight 12, the HNO 3 volume mixing ratios reach higher values than those for the previous flights. Altogether, comparing maximum values with the early winter reference, an ongoing nitrification is observed between December 2015 and January 2016 reaching up to 7 ppbv at ozone values of 1 ppmv and 5 ppbv at ozone values of 0.6 ppmv. Johansson et al. (2019) estimated an ozone depletion by 0.15 ppmv at 380 K for ozone values around 1.15 ppmv by the end of January 2016. Assuming this potential ozone depletion of 15 % (dashed profile in Fig. 5a) in the LMS during the given time frame, the estimated nitrification would reduce to 6 ppbv at 1 ppmv O 3 and 3.5 ppbv at ozone values of 0.6 ppmv. This is a lower-limit estimation, especially considering the contrary effect by mixing of non-vortex air masses.
Correlation-based approaches are also suitable for model comparisons. The exact reproduction of complex fine structures by models cannot be expected because of uncertainties in measurements and the model. Differences in the meteo- rological fields used for modelling, especially the temperature, can result in differing local structures. Since the investigated flights probed a wide range of the sub-vortex region, the obtained correlations can be regarded as representative for the (Arctic) sub vortex and allows the comparison between model and measurement. We point out that observed differences in RNFDs can be caused by an inaccurate representation of processes influencing both HNO 3 and O 3 volume mixing ratios.
The distributions simulated by CLaMS are shown in Fig. 5b. For flight 5, the HNO 3 volume mixing ratios range up to 2.5 ppbv in an approximately linear relationship. Flight 6 shows enhanced HNO 3 volume mixing ratios that are significantly lower than those for the GLORIA observations. CLaMS models a further enhancement for flight 8 with a small patch reaching the maximum values observed by GLORIA. Similar to the GLORIA measurements, flight 12 displays the highest HNO 3 volume mixing ratios of all flights. However, the maximum values observed by CLaMS are 2 ppbv lower. Beneath 0.3 ppmv O 3 hardly any enhancement is observed over the duration of the flights. Overall, CLaMS is able to reproduce the general enhancement of HNO 3 during the winter leading to a nitrification of up to 4 ppbv for ozone values of 0.8 to 1 ppmv, which is by 3 ppbv HNO 3 lower than the GLORIA observations.

Comparison of GLORIA results with CLaMS sensitivity simulations
Four sensitivity simulations have been performed to investigate processes and aspects that have not been represented in the model so far. These sensitivity simulations were performed based on assumptions concerning particle formation and shape. Besides the formation of NAT on ice particles, ice can possibly accumulate on NAT particles (Voigt et al., 2018) resulting in larger particles with an enhanced settling velocity. Therefore in the "ice settling" simulation the computed ice settling velocity (computed as described by Tritscher et al., 2019) was increased by a factor of 1.5 at all locations where the saturation ratio of ice, S Ice , is larger than 1.2. Since NAT formation is temperature dependent, a sensitivity simulation is performed with a global temperature offset of 1 K. Particle growth in CLaMS is based on the assumption of growth rates of spherical particles. However, Woiwode et al. (2016) found indications for highly aspherical particles with an enhanced surface compared to spherical particles of the same volume. Since the HNO 3 uptake depends on the surface, a faster particle growth would occur. A 1.5 times enhanced particle growth was implemented in the "aspherical particle" simulation. Changes in settling velocities due to different shapes have not been taken into account here. Several studies suggest a connection between orographically induced gravity waves and NAT formation (Davies et al., 2005;Carslaw et al., 1998;Höpfner et al., 2006). However smallscale temperature fluctuations are not resolved by ERA interim temperatures. Therefore, artificial fluctuations according to Tritscher et al. (2019) have been added in the "temperature fluctuations" simulation. The comparison is based on the RNFDs depicted for the individual flights in Fig. 6. The model cross sections compared to measurements of flights 6, 8, 12 and 21 can be found in the Appendix B (Figs. B1, B2, B3, B4).
The ice settling simulation (yellow) delivers nearly identical results as the reference simulation for all flights. For the "T-1K" simulation (dark blue), enhanced HNO 3 volume mixing ratios are observed down to lower O 3 volume mixing ratios compared to the reference simulations for flight 6 and 8, but they are still not reaching down to the ozone values noticed by GLORIA. The high HNO 3 volume mixing ratios measured by GLORIA are still underestimated here. While there are only slight changes compared to the reference simulation for flight 12, the T-1K simulation is clearly deteriorating for flight 21. Here, lower ozone volume mixing ratios are observed. For this flight none of the CLaMS simulations are able to reproduce the high ozone volume mixing ratios ob-served by GLORIA, which is possibly caused by weaker subsidence in the model. Further, lower absolute HNO 3 values might occur due to stronger mixing in CLaMS. The aspherical particle case results in enhanced values observed down to lower altitudes than for the reference simulation for the flights 6, 8 and 12. Further, it shows higher maximum HNO 3 values than for the reference for flights 6 and 8. However, the absolute values are still underestimated compared to the GLORIA observations. For the flights 12 and 21, indications for points with lower HNO 3 values compared to the reference are found. The RNFD structure of temperature fluctuation simulation for flight 6 and 21 are nearly identical to the reference simulation. For flight 8, lower HNO 3 volume mixing ratios are observed. In contrast to that, the temperature fluctuation simulation for flight 12 shows the best agreement with the observations. However, even though the lower branch is consistent with GLORIA, an upper branch with values lower than the reference simulation and far lower than GLORIA exists.

Discussion and conclusion
Nitrification of the LMS in the Arctic winter 2015-2016 was analysed based on GLORIA measurements during the PGS campaign. Vertical cross sections of HNO 3 volume mixing ratios for several flights throughout the winter show complex fine-scale structures and enhanced values at altitudes down to 9 km. Flight 8 on 20 January 2016 was carried out under cold conditions with PSC observations at flight altitude. For this flight, coherent structures tilted with altitude of locally enhanced HNO 3 volume mixing ratios are observed that most likely indicate defined regions where settled HNO 3 -containing particles recently sublimated. This is supported by simultaneous in situ observations of HNO 3containing particles. The net effect of proceeding nitrification and dynamical processes in the LMS is observed for flight 12 at the end of January with a pronounced pattern of enhanced HNO 3 volume mixing ratios well exceeding 10 ppbv. Nitri- fied filaments with HNO 3 volume mixing ratios up to 6 ppbv persisted until flight 21 in March 2016. While cross sections provide a qualitative insight on local nitrification patterns for selected flights, the extent of overall nitrification has been quantified based on HNO 3 -O 3 correlations. Nitrification reached an extent of up to 7 ppbv at ozone values of 1 ppmv (θ ≈ 370 K) and up to 5 ppbv at ozone values of 0.6 ppmv (θ ≈ 350 K). A conservative correction, assuming a 15 % ozone loss on the correlations would reduce these numbers to 6 and 3.5 ppbv, respectively.
The comparison of GLORIA observations with the chemistry transport model CLaMS confirms the model's ability to reproduce nitrification of the LMS. Large-scale structures are reproduced by the model that also resolves complex fine structures, although differing from measured patterns. CLaMS predominantly underestimates the enhanced values observed by GLORIA. Enhanced values are found less frequently in the simulation and are limited to narrow regions. Further, modelled HNO 3 enhancements reach less far down on 12 and 20 January 2016 when compared with GLORIA. The CLaMS simulations result in a weaker nitrification of up to 4 ppbv for the period of December to January for ozone mixing ratios between 0.8 to 1 ppmv, which is by ∼ 3 ppbv lower than observed by GLORIA. For flight 21 in March, CLaMS underestimates the observed ozone volume mixing ratios, potentially indicating insufficient subsidence and stronger mixing in the model (Johansson et al., 2019). Sensitivity studies with CLaMS considering (i) ice formation on NAT particles, (ii) a 1 K global temperature offset, (iii) growth rates of aspherical particles or (iv) temperature fluctuations were performed. While the "ice formation" simulation shows only slight differences, the other cases show noticeable differences during individual flights. The temperature fluctuation simulation provides improved agreement for the flight on 31 January 2016 but also worsens the results for the flight on 20 January 2016. The T-1K simulation improves the results for the flights 6 and 8 in January but deteriorates the results for the flight in the late winter on 18 March 2016. This shows the sensitivity of the simulation results on temperature. Potentially, a higher resolution in time and space would result in more realistic temperature fluctuations and could improve the simulations. The aspherical particle case shows slightly more pronounced improvements for the flights in mid-January. Even though the sensitivity simulations partially improve the results, distinct differences between model and measurements remain. Therefore, we conclude that a more comprehensive change in the model representations is required. However, the sensitivity simulations suggest that particle microphysics play a significant role for LMS nitrification in January. Increasing discrepancies from the observations towards the end of the winter are attributed to simulated air subsidence, transport and mixing processes.
Several studies investigated nitrification in previous cold winters, although mainly with a focus on higher altitudes. Hübler et al. (1990) interpreted enhanced NO y values of up to 12 ppbv at altitudes between 10 and 12.5 km in the Arctic winter 1988-1989 as a result of nitrification. For the Arctic winter 2002Grooß et al. (2005 modelled a nitrifica-tion of less than 1 ppbv for potential temperatures lower than 360 K. For the Arctic winter , Dibb et al. (2006 observed nitrification of up to 3 ppbv for potential temperatures between 360 and 340 K. Jin et al. (2006) reported an average nitrification of less than 2 ppbv for potential temperatures lower than 370 K for the same winter. Further, during the Arctic winter 2009-2010  modelled a nitrification of less than 1 ppbv for potential temperatures lower than 360 K, while  found no significant indications for nitrification below 370 K. Since Arctic winters might show a tendency towards colder stratospheric temperatures (Rex et al., 2006), disturbances of the LMS NO y budget by nitrification are likely becoming more frequent. During the Arctic winter 2015-2016 exceptionally low stratospheric temperatures occurred, and the vortex was sufficiently stable to allow formation of PSCs down to lowest stratospheric altitudes. Those conditions were the prerequisites for the strong nitrification observed and presented here. The measurements obtained by GLORIA during the POL-STRACC campaign document in detail a strong impact of nitrification on the LMS during an entire Arctic winter.
Data availability. The discussed GLORIA data set is available at the HALO database at https://halo-db.pa.op.dlr.de/ (last access: 4 November 2019) and at the KITopen repository (https: //doi.org/10.5445/IR/1000086506, Johansson et al., 2018b). NASA MERRA-2 reanalysis data are available at https://disc.gsfc.nasa. gov/ (last access: 4 November 2019). Figure A1a shows the flight track and GLORIA tangent of points of flight 5 on 21 December 2015. The flight accessed air masses associated with the sub-vortex and its vicinity in the region around Scandinavia. Figure A1b and c show the associated vertical cross sections of O 3 and HNO 3 derived from the GLORIA observations.      Author contributions. MB conducted the analysis and interpretation of GLORIA level-2 data and model simulations, and prepared the paper with contributions from all co-authors. JUG performed the CLaMS model simulations. SJ, JU and WW performed the level-1 and level-2 analyses of GLORIA data. MH contributed to the GLO-RIA data analysis and interpretation. FFV and PP coordinated the GLORIA operations during the PGS campaign. HO and BMS coordinated the PGS field campaign. HZ provided the particle HNO 3 data. PB contributed to the interpretation and the paper preparation.

Appendix A
Competing interests. The authors declare that they have no conflict of interest.
Special issue statement. This article is part of the special issue "The Polar Stratosphere in a Changing Climate (POLSTRACC) (ACP/AMT inter-journal SI)". It is not associated with a conference.
Acknowledgements. We thank the PGS coordination team and the DLR-FX for successfully conducting the field campaign. The results are based on the efforts of all members of the GLORIA team, including the technology institutes ZEA-1 and ZEA-2 at Forschungszentrum Jülich. We thank NASA for providing their MERRA-2 meteorological reanalysis data set. Review statement. This paper was edited by Martyn Chipperfield and reviewed by two anonymous referees.