Vortex‐Wide Detection of Large Aspherical NAT Particles in the Arctic Winter 2011/12 Stratosphere

Micron‐sized HNO3‐containing particles in polar stratospheric clouds are known to denitrify the polar winter stratosphere and support chemical ozone loss. We show that populations of nitric acid trihydrate (NAT) particles with volume‐equivalent median radii of 3–7 μm can be detected vortex‐wide by means of infrared limb sounding. Key for detection are the applied optical characteristics of highly aspherical particles consisting of the β‐NAT phase. Spectroscopic signatures and ambient conditions of detected populations show that these particles play a key role in denitrification of the Arctic winter stratosphere. Complementary gas‐phase HNO3 observations indicate collocated highly efficient HNO3 sequestration within days and are consistent with measured spectral signals of populations of large NAT particles. High amounts of condensed gas‐phase equivalent HNO3 exceeding 10 ppbv and long persistence of detected populations, despite expected gravitational settling, imply that our understanding of the particles is incomplete.


Introduction
Large HNO 3 -containing particles in polar stratospheric clouds (PSCs) sediment and denitrify the polar winter stratosphere and thus support chemical ozone loss (Fahey et al., 1990;Peter, 1997;Poole & McCormick, 1988;Salawitch et al., 1989;Solomon, 1999;Toon et al., 1986;Waibel et al., 1999). The particles are thought to be composed predominantly of solid nitric acid trihydrate (NAT) and have radii between 5 and 10 μm (Fahey et al., 2001;Hanson & Mauersberger, 1988). While extensive denitrification is common in the cold Antarctic winter stratosphere, the Arctic winter stratosphere is mostly warmer, and extensive denitrification is linked to individual cold winters (Fahey et al., 1990;Toon et al., 1990). In context of climate and atmospheric composition change, a high year-to-year variability and a trend to colder stratospheric temperatures were diagnosed in the Arctic (Rex et al., 2006;Rieder & Polvani, 2013;WMO (World Meteorological Organization), 2018). Extensive denitrification and ozone loss were clearly observed in recent extreme Arctic winters (Khosrawi et al., 2011;Khosrawi et al., 2017;Kuttippurath et al., 2012;Manney et al., 2011;Sinnhuber et al., 2011). Accurate representations of Arctic winter denitrification and chemical ozone loss are essential for future projections of stratospheric ozone (WMO (World Meteorological Organization), 2018). Among other factors, significant uncertainties remain due to limited knowledge of the characteristics of large NAT particles (Grooß et al., 2014;Tritscher et al., 2019;Zhu et al., 2015).
NAT is stable at temperatures below~195 K (Hanson & Mauersberger, 1988;Tizek et al., 2004;Worsnop et al., 1993) and was confirmed by in situ observations in the Arctic stratosphere (Voigt et al., 2000). Further constituents of PSCs are supercooled ternary solutions (STS) and ice (Peter, 1997). Large HNO 3 -containing particles denitrifying the Arctic winter stratosphere were detected in situ and attributed to NAT (Brooks et al., 2003;Fahey et al., 2001;Northway et al., 2002). Lidar remote sensing observations allow detection of NAT-containing PSCs operationally (Achtert & Tesche, 2014;Pitts et al., 2018) but are hardly sensitive to particle sizes. A characteristic spectral peak in space-borne infrared limb observations was clearly assigned to the v 2 out-of-plane deformation band of NO 3in β-NAT, and a detection method sensitive to small NAT particles in high volume densities, which occur for example in lee-wave events, was established (Höpfner et al., 2006;Spang & Remedios, 2003). Laboratory experiments suggest that β-NAT particles are highly aspherical (Grothe et al., 2006). In a case study involving airborne in situ and infrared limb observations, large highly aspherical β-NAT particles with equivalent radii exceeding 10 μm were detected (Molleker et al., 2014;Woiwode et al., 2016). Compared to previous observations, a high gas-phase equivalent HNO 3 mixing ratio of 9 ppbv was inferred from the airborne limb observations, while the in situ observations supported even higher values. State-of-the art model parameterizations were not able to reproduce the observed combinations of large particle sizes and high HNO 3 content inferred from these case studies (Grooß et al., 2014;Zhu et al., 2015).
In the past, irreversible denitrification of the Arctic winter stratosphere was successfully simulated using a microphysical model of large NAT particles . Similar and advanced parameterizations Hoyle et al., 2013;Zhu et al., 2015) were applied to simulate denitrification in subsequent and, particularly, colder Arctic and Antarctic winters (Grooß et al., 2005;Grooß et al., 2014;Zhu et al., 2015;Khosrawi et al., 2017;Tritscher et al., 2019). However, nucleation, growth, and sedimentation characteristics of large NAT particles remain largely uncertain. Ice is likely to play an important role in NAT nucleation (Weiss et al., 2016). While state-of-the-art parameterizations of large NAT particles are successful in reproducing major patterns of denitrification and the state at the end of the winter, large discrepancies remain in direct comparisons with rare Arctic particle field observations (Grooß et al., 2014;Molleker et al., 2014;Woiwode et al., 2016;Zhu et al., 2015). Depending on the different models and scenarios, simulated particle sizes are significantly smaller and/or number densities significantly lower than observed. Condensed HNO 3 inferred from the observations is underestimated by up to one order of magnitude. Moreover, relying on the common approximation of spherical particles, field observations of extremely large NAT particles would imply a much faster sedimentation and more efficient irreversible denitrification than simulated. As a consequence, uncertainties remain in simulations and projections of polar stratospheric reactive nitrogen (NO y ) and ultimately chemical ozone loss. Clearly, vortexwide observational capabilities sensitive to large NAT particles that allow model parameterizations to be tested are lacking.

New Method to Detect Large NAT Particles
Here, we introduce a new method for vortex-wide and specific detection of aspherical β-NAT particle populations with median radii of 3-7 μm. The method is based on radiative transfer calculations of infrared limb observations (Stiller et al., 2002) accounting for the light scattering by micron-sized particles (Höpfner et al., 2006). We apply our method to space-borne observations by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) (Fischer et al., 2008). Particle size distributions are implemented by means of lognormal distributions and are characterized by median radius r, mode width σ, and number density N. Simulations of spectra of spherical β-NAT particles, STS droplets, and spherical ice particles are based on Mie calculations (Höpfner et al., 2006). Simulations of highly aspherical β-NAT particles are based on T-Matrix calculations (Bi et al., 2013) and are used to define the criteria of our method. As in Woiwode et al. (2016), we adopt elongated spheroids with an aspect ratio of 0.1. Note that highly oblate spheroids with an aspect ratio of 10 produced similar results in the previous study. Refractive indices of β-NAT, STS, and ice reported in the literature are used (Biermann, 1998;Biermann et al., 2000;Zasetsky et al., 2005;Höpfner et al., 2006). Figure 1a shows a typical MIPAS spectrum which exhibits the characteristic spectral signature of large aspherical β-NAT particles: a "shoulder-like" signature around 820 cm −1 and an overall high and flat spectral baseline. Geolocation and time of the observation approximately coincide with airborne observations reported in the literature (compare (Molleker et al., 2014;Woiwode et al., 2016)). Also shown are two simulations of aspherical β-NAT particles (including and excluding climatological trace gas profiles), spherical β-NAT particles, STS droplets, and ice particles. The same size distribution constrained by the previous airborne case study is used (gray in Figure 1b; see Woiwode et al., 2016 and based on Molleker et al., 2014). Other than small discrepancies (discussed below), the spectral baselines of both simulated spectra of aspherical β-NAT particles and the observation agree well, particularly in the narrow spectral windows  Fahey et al., 2001), and used for sensitivity simulations (red, pink, black and blue solid/dashed lines) in (c,d). (c) Sensitivity simulations of spectra of aspherical β-NAT particles using size distributions with different r (indicated), excluding and including effects by high tropospheric clouds (black and red solid lines, respectively). (d) Range of detection, determined from simulations including effects by a high tropospheric cloud and using different combinations of r and N (red, black, and blue, corresponding with Panel b). (e) MIPAS observations of a cloud-free case (black), a developed β-NAT "peak" (r ≤ 3 μm, high volume density, red) and a developed "shoulder" of aspherical β-NAT particles (r ≥ 3 μm, high amounts of condensed HNO 3 , blue). Microwindows used to detect aspherical NAT populations with r ≥ 3 μm are marked in cyan in (a,c,d, and e). Gray dotted horizontal lines in (c,d) indicate the overall threshold radiance level of our method.
"A", "B", and "C" (cyan) which are used in our method. The simulation of spherical β-NAT particles does not adequately reproduce the signature around 820 cm −1 and shows an underlying negative gradient (i.e., "tilt" of the spectral baseline with increasing wave numbers to lower radiances). The STS simulation shows hardly any modulation around 820 cm −1 and also has an overall negative gradient, and the ice simulation exhibits a steep negative gradient and lacks any specific signatures. Thus, the signature of large aspherical β-NAT particles can be clearly distinguished from STS and ice.
The simulations of aspherical β-NAT particles also show some noticeable discrepancies from the observations. Discrepancies between 850 and 925 cm −1 are explained by the fact that the climatological gas-phase HNO 3 profile used in the simulation does not represent the actual state of denitrification. Therefore, the gas-phase HNO 3 spectral lines are overestimated. In contrast, stratospheric CO 2 mixing ratios are wellknown, and corresponding spectral lines above 925 cm −1 are reproduced well. Furthermore, the observation shows a negative baseline gradient below 850 cm −1 . This discrepancy between observation and simulation is explained by altitude-dependent variations of the β-NAT size distribution, weak contributions by potential ice particles present at higher altitudes (i.e., inducing an underlying negative gradient), and uncertainties in the simulated dense ozone lines based on a climatological ozone profile. However, these discrepancies hardly affect our method which is based on the microwindows A, B, and C.
The following integrated radiances and their differences are used to detect the flat "hockey stick" signature of large aspherical β-NAT particles from 817.5 to 961 cm −1 . (i) An integrated radiance of ≥100 nW cm −2 sr −1 in the spectral window "C" (960.0 to 961.0 cm −1 ). (ii) A difference between the integrated radiances in the spectral windows "B" (833.0 to 834.0 cm −1 ) and "C" of ≤100 nW cm −2 sr −1 . (iii) A difference between the integrated radiances in "A" (817.5 to 818.5 cm −1 ) and "B" of ≥50 nW cm −2 sr −1 . More information on PSC detection methods used here is provided in the supplement.

Sensitivity
To test the sensitivity of our method, modified simulations are shown in Figures 1c and 1d. Our baseline case involves a size distribution with r = 5 μm, σ = 1.35, and N = 1.5·10 −3 cm −3 (pink solid line in Figure 1b). The spectrum of large β-NAT particles is modulated by the radius and greybody-like radiation originating from the ground or tropospheric clouds scattered by the β-NAT particles. Figure 1c shows simulations of aspherical β-NAT particles including the baseline case with modified radii from 2 to 6 μm. Note that except for the size distribution adapted from the previous study for the same day (grey), unbinned size distributions are shown in Figure 1b for easier reading. The size distributions used in Figure 1c correspond to gas-phase equivalent HNO 3 mixing ratios of 0.6 (r = 2 μm), 2 (r = 3 μm), 5 (r = 4 μm), 10 (r = 5 μm), and 16 ppbv (r = 6 μm). These scenarios cover cases with low fractions of HNO 3 being condensed to cases with almost all HNO 3 being condensed (compare MLS ambient HNO 3 mixing ratios in Figures 2a and 2b). Note that radii of volume-equivalent spheres are given here and not diameters (compare Carslaw et al., 2002;Fahey et al., 2001;Molleker et al., 2014). Black solid lines in Figure 1c show simulations assuming no tropospheric clouds. Red solid lines show the same simulations including an optically dense tropospheric cloud top at 10 km. The simulations show that a cold tropospheric cloud top induces a weaker overall signal and modulates the spectrum. However, the criteria of our method are fulfilled in the cases with median radii equal to or larger than 3 μm. Only the r = 3 μm scenario including a cold tropospheric cloud top shown here falls slightly below the detection threshold.
The spectrum sensitivity to different particle number densities is investigated in Figure 1d for r = 3, 5, and 7 μm. The solid red, black, and blue lines correspond to cases which are just detectable under the attenuating conditions of high tropospheric clouds, which are often found in Arctic winters. Here, the NAT "shoulder" at 820 cm −1 is developed sufficiently (see criterion (iii)). The scenarios correspond to 2 (r = 3 μm), 5 (r = 5 μm), and 13 ppbv (r = 7 μm) of gas-phase equivalent HNO 3 . The loss in sensitivity with increasing median radius is a consequence of the increasing size parameter (see Bi et al., 2013 and references therein) and flattening of the signature at 820 cm −1 . The dashed lines are cases with 15 ppbv of gas-phase equivalent HNO 3 , approximating situations with complete condensation of all available HNO 3 . Associated number densities are also indicated in Figure 1d, and the corresponding size distributions are shown in Figure 1b using the same color scheme. Note that the signature around 820 cm −1 shows a partially peak-like shape for r = 3 μm (cf. Höpfner et al., 2006). So in summary, within the assumptions made in our sensitivity scenarios, our method based on integrated radiances and their differences is capable of detecting NAT particle populations with equivalent median radii of 3-7 μm and condensed gas-phase equivalent HNO 3 exceeding 2-13 ppbv.
In Figure 1b, the applied size distributions are compared to the size distribution reported for the Arctic winter 1999/2000 field observations (yellow, Fahey et al., 2001). All size distributions with r ≥ 3 μm accessible to our method considerably exceed this size distribution in terms of number densities and condensed HNO 3 . Furthermore, size distributions simulated in different model studies are well exceeded for the same parameters (compare, Carslaw et al., 2002;Grooß et al., 2014;Zhu et al., 2015).
Note, our method may also respond to NAT populations with r ≤ 3 μm in high volume densities, if the integrated radiance in the spectral window "C" exceeds 100 nW cm −2 sr −1 and the slope of the spectral baseline is sufficiently low (i.e., "B"-"C" ≤ 100 nW cm −2 sr −1 ). Therefore, our method should be used in combination with the established method for the detection of small NAT particles in high volume densities reported in the literature (Höpfner et al., 2006) to distinguish between large NAT particles and small NAT particles in high volume densities. MIPAS spectra of a cloud-free case and cases with developed signatures of small NAT particles in high volume densities and large NAT particles with r ≥ 3 μm consuming high amounts of HNO 3 are shown in Figure 1e.

Complementary MIPAS and MLS Observations
In the following section, we apply our method sensitive to populations of aspherical β-NAT particles with median radii of 3 to 7 μm and gas-phase equivalent HNO 3 exceeding 2 to 13 ppbv to MIPAS observations during the onset of denitrification in the 2011/12 Arctic winter. The formation of such populations is associated with significant sequestration of HNO 3 from the gas phase. The space-borne microwave limb sounder (MLS) is capable of measuring gas-phase HNO 3 in the presence of PSCs (Livesey et al., 2018;Manney et al., 2015;Waters et al., 2006). We use MLS observations to test whether gas-phase HNO 3 depletion detected by MLS is consistent with populations of large NAT particles detected by MIPAS. Furthermore, we assess the extent of gas-phase HNO 3 sequestration using the MLS data and thereby constrain the maximum possible amount of HNO 3 condensed in the detected populations. Finally, we estimate condensed HNO 3 from MIPAS observations and compare with that inferred from the MLS gas-phase HNO 3 observations.
The onset of widespread PSC formation and HNO 3 sequestration is detected by MIPAS and MLS in early December 2011. The left and middle columns of Figure 2 show MLS gas-phase HNO 3 observations at the pressure (altitude) levels of 31 hPa (~22 km) and 46 hPa (~20 km). These are the altitudes where PSCs, HNO 3 sequestration, and denitrification are typically found in the Arctic (Pitts et al., 2018;Waibel et al., 1999). In the right column, we show MIPAS observations at 20 km, where NAT particles are expected to grow to large sizes during gravitational settling Fahey et al., 2001). PSCs detected by the sensitive but unspecific cloud index method (see Spang et al., 2004), populations of small NAT particles (r ≤ 3 μm) in high volume densities (see Höpfner et al., 2006), and populations of large aspherical NAT particles (r ≥ 3 μm) detected by our new method are highlighted. ECMWF ERA-Interim (European Centre for Medium-Range Weather Forecasts, https://www.ecmwf.int/) temperature contours at the same altitude level provide the thermodynamic context. Temperature isolines are shown at 200 K, which is an upper limit allowing the persistence of NAT particles under enhanced HNO 3 partial pressures, and at 195 K, which is the typical value of the existence temperature of NAT (T NAT ) in the stratosphere (Hanson & Mauersberger, 1988). The isoline at 190 K marks lower temperatures which are still above the ice frost point (T ice~1 88 K (Murphy & Koop, 2005)).

Vortex-Wide Detection in Arctic winter 2011/12
On 4 December 2011, the gas-phase HNO 3 distribution north of 60°N is largely unperturbed. At 22 and 20 km, HNO 3 mixing ratios are mostly higher than 12 and 10 ppbv, respectively (Figures 2a and 2b). Only a weak local minimum is found at 20 km above the north-west coast of Greenland. An unspecified PSC population is detected by MIPAS over the pole at temperatures around T NAT (Figure 2c). Two days later, significant local gas-phase HNO 3 depletion of more than 5 ppbv (2-3 ppbv) relative to ambient HNO 3 mixing ratios is identified above north-west Greenland and Baffin Bay at 22 km (20 km), indicating efficient gas-phase HNO 3 sequestration by PSCs (Figures 2d and 2e). At the same time, the PSC population detected by MIPAS has spread out in the same region, and a population of large NAT particles is detected coincident with local gas-phase HNO 3 depletion detected by MLS (Figure 2f).
On 8 and 10 December 2011, regions depleted in gas-phase HNO 3 at 22 km (Figures 2g and 2j) and 20 km (Figures 2h and 2k) spread out rapidly on synoptic scales across Greenland to Scandinavia. Local minimum gas-phase HNO 3 mixing ratios fall below 5 ppbv (2 ppbv) on 8 December 2011 (10 December 2011), which are in contrast to much higher mixing ratios of 10 to 15 ppbv in surrounding air masses. At the same time, the population of large NAT particles detected by MIPAS (Figures 2i and 2l) spreads out further in exactly the same region. Ambient temperatures remain slightly below T NAT and mostly well above T ice . On 10 December 2011, indications of small NAT particles in high volume densities are found locally above the Arctic sea and close to the pole (Figures 2l). On 14 December 2011, the regions depleted in gas-phase HNO 3 at 22 and 20 km have spread out further and cover wide regions from Greenland to northern Siberia (Figures 2m and 2n). Gas-phase HNO 3 mixing ratios approach zero, indicating almost complete condensation of HNO 3 . The detected populations of large NAT particles follow the regions depleted in gas-phase HNO 3 consistently, whereas temperatures remain below T NAT and above T ice (Figure 2o).
Strong short-term modulations of populations of large NAT particles are documented in Figure 3. By 24 December 2011, after extensive gas-phase HNO 3 depletion and populations of the particles persisting over weeks (not shown), the region depleted in gas-phase HNO 3 covers a wide area in the sector from Siberia to the pole (Figures 3a and 3b). While synoptic minimum temperatures at 20 km remain below T NAT and above T ice , indications of NAT particles with volume-equivalent median radii ≥3 μm remain only locally above Siberia (Figure 3c). Wide areas have been irreversibly denitrified. HNO 3 removed from this level by gravitational settling of large NAT particles might be still present in the condensed phase at lower altitudes.

Geophysical Research Letters
To test this hypothesis, we show also MIPAS observations at the level of 17 km (Figure 3d). In fact, a widespread population of large NAT particles that coincides well with the region depleted in gas-phase HNO 3 above is found at 17 km. Two days later, the denitrified area stretches across the pole to Greenland and Canada (Figures 3e and 3f). Hardly any indications of PSCs remain due to warmer temperatures (Figures 3g and 3h). However, only another 2 days later, the denitrified air masses are located mainly at a region around Greenland and the Arctic sea (Figures 3i and  3j), and a PSC including populations of large NAT particles has been reestablished in the same region (Figures 3k and 3l). Note that such populations are located preferentially in regions where enhanced supersaturation of gas-phase HNO 3 versus the NAT phase is found locally, which is consistent with the thermodynamic characteristics of NAT (Hanson & Mauersberger, 1988).
Examples of MIPAS spectra of NAT populations detected by our method are shown in Figure 4 together with reference simulations (see Figure 1 c). For the reference simulations, the corresponding gas-phase equivalent HNO 3 mixing ratio at 20 km is indicated for comparisons with the MLS data (see Figure 2, middle column). On 4 December 2011, before the beginning of significant HNO 3 sequestration, the shown MIPAS spectrum corresponds to less than 3 ppbv of condensed gasphase equivalent HNO 3 . The spectrum of 6 December 2011 corresponds to more than 3 ppbv and the spectrum of 10 December 2011 to up to 6 ppbv of condensed gas-phase equivalent HNO 3 . The spectrum of 14 December 2011 corresponds to a rather high amount of condensed HNO 3 of about 12 ppbv. All gasphase equivalent HNO 3 mixing ratios estimated from the MIPAS spectra are in keeping with HNO 3 depletion seen in the complementary MLS data.

Conclusions
The MLS and MIPAS observations in the Arctic winter 2011/12 clearly show fast and extensive sequestration of gas-phase HNO 3 within only a few days and coincident formation of widespread populations of large aspherical NAT particles. The observations are consistent with established spectral (Biermann, 1998;Höpfner et al., 2006) and thermodynamic (Hanson & Mauersberger, 1988;Tizek et al., 2004;Worsnop et al., 1993) characteristics of β-NAT and show that large aspherical β-NAT particles play a key role in denitrification of the Arctic winter stratosphere. Extensive gas-phase HNO 3 sequestration measured by MLS and condensed HNO 3 inferred from the MIPAS spectra of 10 ppbv or more exceed model simulations for different Arctic winters by up to one order of magnitude Fahey et al., 2001;Grooß et al., 2014;Zhu et al., 2015). NAT particles with radii around 5 μm are expected to sediment with fall speeds of about one kilometer per day (Fahey et al., 2001), which would result in the removal of particles of the sizes observed here after only a few days. In contrast, we find populations of large NAT particles at 20 km persisting over more than 1 week. The long persistence and slow sedimentation of the detected populations could be explained by rather low fall speeds of highly aspherical particles (Westbrook, 2008). Further focused field observations would be extremely helpful to constrain the microphysical properties of large NAT particles. Model parameterizations should be revisited in order to reproduce observations of large NAT particles to improve future projections of polar stratospheric ozone.

Data Availability Statement
The MIPAS observations used here are accessible via https://earth.esa.int/web/guest/pi-community/applyfor-data/fast-registration. The MLS data used here are available at https://disc.gsfc.nasa.gov/datasets/ ML2HNO3_V004/summary. The ECMWF ERA-Interim data used here are accessible at https://www. ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era-interim. . Observed spectra of NAT populations and estimates of condensed HNO 3 . The MIPAS spectra with a tangent altitude of 20 km are associated with large NAT particle detections. Sensitivity simulations for r = 3, 4, and 5 μm (orange, magenta, and red) including and excluding effects by high tropospheric clouds (solid and dashed lines, respectively) are superimposed. Microwindows used to detect aspherical NAT populations with r ≥ 3 μm are marked in cyan.