Transport of Antarctic stratospheric strongly dehydrated air into the troposphere observed during the HALO-ESMVal campaign 2012

Dehydration in the Antarctic winter stratosphere is a well-known phenomenon that is occasionally observed by balloon-borne and satellite measurements. However, in-situ measurements of dehydration in the Antarctic vortex are very rare. Here, we present detailed observations with the in-situ and GLORIA remote sensing instrument pay- 5 load aboard the new German aircraft HALO. Strongly dehydrated air masses down to 1.6 ppmv of water vapor were observed as far north as 47 ◦ S and between 12 and 13 km in altitude, which has never been observed by satellites. The dehydration can be traced back to individual ice formation events, where ice crystals sedimented out and water vapor was irreversibly removed. Within these dehydrated stratospheric air 10 masses, ﬁlaments of moister air reaching down to the tropopause are detected with the high resolution limb sounder, GLORIA. Furthermore, dehydrated air masses are observed with GLORIA in the Antarctic troposphere down to 7 km. With the help of a backward trajectory analysis, a tropospheric origin of the moist ﬁlaments in the vortex can be identiﬁed, while the dry air masses in the troposphere have stratospheric 15 origins. The transport pathways of Antarctic stratosphere/troposphere exchange are investigated and the irrelevant role of the Antarctic thermal tropopause as a transport barrier is conﬁrmed. Further, it is shown that the exchange process can be attributed to several successive Rossby wave events in combination with an isentropic interchange of air masses across the weak tropopause and subsequent subsidence due to radiative cooling. Once transported to the troposphere, air masses with stratospheric origin are able to reach near-surface levels within 1–2 months.


Introduction
Antarctic stratospheric dehydration occurs regularly every winter and spring in the very isolated southern hemispheric polar vortex (e.g. Vömel et al., 1995;Kelly et al., 1989; and stratosphere. Here, we want to note that HALO was equipped with in-situ instruments as well as the high resolution limb sounder, GLORIA (Gimballed Limb Observer for Radiance Imaging of the Atmosphere). This combination of high-precision in-situ measurements together with sophisticated remote sensing observations in the UT/LS is an outstanding attribute of both campaigns. Another objective of the ESMVal flights 10 was to get a full meridional cross section of atmospheric measurements for global chemistry model evaluation. Hence, the Antarctica flight was performed in an effort to get as far south as possible and reach the stratospheric vortex. During this flight, dehydrated air masses were measured in-situ quite far north up to 45 • S between 12 and 13 km. In contrast, Aura-MLS and POAM 3 satellite measurements (Nedoluha et al.,15 2002; Jimenez et al., 2006;Schoeberl and Dessler, 2011) do not show dehydrated air masses as far north (not beyond 57 • S) and as low in the stratosphere and upper troposphere as was observed during this ESMVal flight. In general, the process of dehydration is well understood. Relatively less is known about the fate of the dehydrated air masses. The air within the Antarctic polar vor-20 tex is highly isolated with a weak exchange of trace gases across the vortex edge driven by stratospheric planetary Rossby waves propagating from the troposphere and Rossby wave breaking (RWB) events. However, these mainly disturb the bottom of the polar vortex. Normally, the vortex edge in the Southern Hemisphere is less strongly disturbed than in the Arctic due to less wave activity (Schoeberl et al., 1992). In the els. The first process consists of katabatic winds over the Antarctic Plateau caused by the high topography that create a general downwelling above the Antarctic continent as reported by Roscoe (2004). The second process is driven by mid-latitude cyclones on the poleward side of the jet stream that support RWB events and a corresponding stratospheric intrusion as reported by Ndarana et al. (2012). Rossby wave induced 10 stratospheric intrusions, such as tropopause folds, occur more often further north and in the midlatitudes than directly above the Antarctic continent (James et al., 2003). Once an airmass is in the troposphere, the mean cooling rates cause a reduction in potential temperature and the airmass will descend from the tropopause to near-surface heights within 10 days (descent rate of 5 mm s −1 ) as reported by van de Berg et al. 15 (2007). However, the frequency, the seasonality, and the process behind tropopause folds and stratospheric intrusions in the Antarctic region is still under debate (Stohl and Sodemann, 2010;Ndarana et al., 2012;Mihalikova and Kirkwood, 2013). The study presented here is structured as follows: in Sect. 2, a brief overview of the different instruments and data used is given. The in-situ and remote sensing mea- 20 surements across the Antarctic polar vortex are described in Sect. 3, where it is also shown that dehydration occurs directly in the transition region between the upper troposphere and lower stratosphere (UT/LS). Furthermore, Sect. 4 includes a case study of observed stratosphere/troposphere exchange of dehydrated air masses combined with an extensive trajectory analysis showing how deep below the thermal tropopause 25 dehydrated air masses can be found.

Instrumentation and meteorological data
The new German research aircraft, HALO, deployed during TACTS and ESMVal has a long flight endurance of up to 12 h. This enables air masses in the Antarctic vortex to be sampled without landing in Antartica. Altogether, a set of nine in-situ instruments for measuring trace gases and one remote sensing instrument were installed aboard 5 HALO. For the study presented here, we use the water vapor data of FISH and HAI as well as measurements from TRIHOP for methane and FAIRO for ozone. The remote sensing instrument, GLORIA, provides cross-sections of trace gases that are fairly parallel to the flight track of the aircraft. GLORIA's high vertical resolution makes it particularly suited for the investigation of small-scale structures.

10
In addition to the aircraft measurements, satellite observations from CALIPSO and meteorological data from ECMWF are used for further interpretation of the observed situation. The instruments and meteorological data are described in the following subsections. 15 The airborne Lyman-α photofragment fluorescence hygrometer FISH (Fast In-situ Stratospheric hygrometer) is a well-established closed-path instrument for measuring water vapor in the range of 1 to 1000 ppmv (Zöger et al., 1999). The FISH is especially built to measure the low water vapor mixing ratios prevailing in the upper troposphere and lower stratosphere. It is regularly calibrated on the ground to a reference frost point 20 hygrometer (MBW DP30) and had an accuracy of 6 % ± 0.4 ppmv during TACTS and ESMVal campaigns (Meyer et al., 2015). The air supply for the measuring cell is provided by a forward facing inlet, which also samples ice crystals if present (H 2 O cond

HAI -total water measurements (in-situ H 2 O)
The novel Hygrometer for Atmospheric Investigation (HAI) (Buchholz, 2014) is a fast, airborne, in-situ hygrometer capable of multi-phase water detection that employs direct Tunable Diode Laser Absorption Spectroscopy (dTDLAS) in combination with a special data retrieval method, which allows absolute water vapor measurements without any 5 sensor calibration (Buchholz et al., 2013). This data retrieval was recently successfully validated via a direct side by side comparison with the German metrological water vapor scale (Buchholz et al., 2014). HAI is a 2 × 2, multi-channel TDLAS spectrometer which realizes, in a unique concept, a simultaneous gas phase and total water measurement via a parallel water detection in an open-path cell, outside the aircraft fuselage, and in extractive (i.e. gas sampling) closed-path cells inside the aircraft. To enlarge the dynamic range to 1-40 000 ppmv, both cell types are simultaneously probed by two laser wavelengths, at 1.4 µm for the lower to mid troposphere and at 2.6 µm for the upper troposphere and the stratosphere. The ranges of the 1.4 and 2.6 µm spectrometer channels overlap and therefore provide a redundant, independent detection 15 between 10 and about 5000 ppmv. The metrologically calculated spectrometer uncertainties are 4.3 % ± 3 ppmv at 1.4 µm and 5.9 %±0.4 ppmv at 2.6 µm. In this study, data from the two closed-path spectrometers are used, i.e. total water measurements similar to FISH. 20 The TRIHOP instrument is a three channel Quantum Cascade Laser Infrared Absorption spectrometer capable of the subsequent measurement of CO, CO 2 , N 2 O and CH 4 . During TACTS/ESMVal the instrument was calibrated in-situ against secondary standards of compressed ambient air which are traceable against NOAA standards. Integration time for each species was 1.5 s at a duty cycle of 8 s, which finally limits Introduction ber 2012, TRIHOP CH4-data achieved a precision (2σ) of 9.5 ppbv and accuracy of 13.5 ppbv, respectively, without any corrections applied.

FAIRO -ozone measurement (in-situ O 3 )
FAIRO is a new accurate ozone instrument developed for use on board the HALO aircraft. It combines two techniques, the UV photometry (light absorption of O 3 at λ = 250-5 260 nm) with high accuracy and chemiluminescence detection with high measurement frequency. A UV-LED is used as a light source for the UV photometer, which can be controlled well (in contrast to Hg lamps) for constant light emission. The 1-sigma precision is 0.08 ppbv at a measurement frequency of 4 s and a cuvette pressure of 1 bar and the total uncertainty is 2 %. The chemiluminescence detector shows a measurement frequency of 12.5 Hz and a high precision of 0.05 ppbv (at 10 ppbv absolute, a measurement frequency of 5 Hz, and a pressure of 1 bar) (Zahn et al., 2012).

GLORIA (aircraft remote sensing)
The Gimballed Limb Observer for Radiance Imaging of the Atmosphere (GLORIA) combines a two-dimensional focal plane detector array with a Fourier-transform spec- 15 trometer to capture about 6000 infrared limb spectra simultaneously. This enables remote sensing observations with high vertical and horizontal resolution to resolve small scale structures (see Riese et al., 2014). The spectral sampling can be switched between 0.625 and 0.0625 cm −1 at the cost of an increased acquisition time in case of higher spectral resolution . The gimbal mount and inertial alti-20 tude control system allows GLORIA to maintain a steady pointing on a moving aircraft; it can also be used to point the instrument at a range of azimuth angles with respect to the aircraft, covering about 70 • . This allows for tomographic measurement patterns and 3-D reconstruction of fine-scale filamentary structures (Ungermann et al., 2011; atmosphere). Thus, while the location of quantities retrieved at flight level lies in the vicinity of the aircraft, the locations of quantities at lower altitudes are several tens to hundreds of kilometers away. Concerning the water vapor product used in this paper,  showed that GLORIA water vapor at flight level agrees fairly well with the in-situ FISH measurements, within error bars. The deviations to FISH 5 during the ESMVal/TACTS flights are mostly less than 0.4 ppmv.

CALIPSO (satellite remote sensing)
The Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellite is one of five satellites in the NASA A-train constellation. CALIPSO completes 14.55 orbits per day with an inclination of 98.2 • and thus delivers good cov-10 erage above the polar regions. Besides one wide field camera and an imaging infrared radiometer, the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) is aboard CALIPSO. The lidar operates with two wavelengths (532, 1064 nm) with additional polarization-sensitivity, providing high-resolution vertical backscatter profiles of aerosols and clouds. In this study we use the CALIPSO Lidar Level 2 Polar Strato- 15 spheric Cloud (PSC) data product, which is described in Pitts et al. (2009) and Pitts et al. (2011). This data product provides a PSC composition scheme on a daily basis for all nighttime orbits with a resolution of 5 km horizontally by 180 m vertically.

Meteorological situation/flight pattern
The ESMVal flight on 13 September 2012 was performed from Cape Town (South Africa) heading towards Antarctica until 65 • S at an altitude of 12-13 km (see Fig. 1).
After a dive from 12.5 km down to 3 km to sample the transition from the polar strato-5 sphere to the troposphere, the return flight took place at 12.5 km and rose up to 15 km shortly before descending back to Cape Town. A hexagonal flight pattern was included in the flight path back to Cape Town to enable the tomographic study of trace gases with GLORIA. The ECMWF wind field (black lines) as well as the strongly increasing potential vorticity (PV, see Fig. 2) show that the vortex edge (yellow dots) was far north, up to 47 • S. The vortex edge is determined according to the definition of Nash et al. (1996). The color code in Fig. 1a represents the in-situ measured water vapor mixing ratios from FISH along the flight path. Directly south of the vortex edge the water vapor mixing ratios decrease rapidly to below 2.5 ppmv (purple dots along the flight path). The latitudinal cross section of equivalent latitude (calculated from the ECMWF PV 15 field) along the flight path is shown in Fig. 1b. The thermal tropopause (light blue dots) stays almost constant at around 10 km and rises up to 12 km further poleward. The PV and equivalent latitude are nearly conserved with time in the stratosphere but less strongly so in the troposphere due to small scale mixing. The equivalent latitude shows high values within the vortex and thus marks clearly the vortex air masses. However, 20 high equivalent latitudes of 70-80 • S show up far below the thermal tropopause down to 5 km altitude in the latitude range from 60-45 • S. This indicates that these air masses originate from higher latitudes and were transported from the stratosphere through the thermal tropopause. The transport process can be shown and confirmed with the trajectory analysis in Sect. 4.2. Interestingly, the 310 K isentrope crosses the thermal 25 tropopause once (65 • S) and the 320 K isentrope crosses it three times (58, 55 and 50 • S) on the poleward side of the jet core. Therefore, air masses can be potentially transported from the stratosphere into the troposphere and vice versa. crosses the air masses with high equivalent latitude in the stratosphere before and after the dive. The dive down to 3.5 km altitude was performed on the poleward side of the high equivalent latitude tropospheric air masses.

Observations
The time series of the in-situ measurements are shown in Fig. 2, where the mea- increase. This is the time when HALO penetrated the Antarctic vortex (marked in blue in Fig. 2). After ascending to a higher flight level (08:35 UTC), water vapor decreases to slightly above 2 ppmv and stays there until the southernmost point was reached and the dive started at 11:00 UTC. This low water vapor in the time between 08:15 and 11:00 UTC coincides well with high equivalent latitude (see Fig. 1b). Subsequently, the 20 strong increase in methane and decrease in ozone and PV indicate the penetration of the aircraft into the Antarctic troposphere at 11:00 UTC. After returning to the previous flight level (11:45 UTC), the vortex signatures with low values of water vapor and methane as well as high values of ozone and PV show up again. However, at this point (∼ 12:30 UTC) the water vapor mixing ratio reaches the lowest value of 1.6 ± 0.5 ppmv ACPD 15,2015 Transport of Antarctic dehydrated air into the troposphere C. Rolf et al. The saturation mixing ratio with respect to ice within the vortex is three to four times larger than the measured water vapor mixing ratio. Thus, the sampled air masses were clearly sub-saturated with relative humidities of 25-33 % and it is unlikely that ice particles remained in the probed air masses. In addition, HAI and FISH show no peaks in the total water vapor time series (based on 1 Hz data), which would indicate the pres-5 ence of ice particles on the flight level. Especially, the water vapor mixing ratios around 2 ppmv during the stratospheric flight legs (06:30 to 11:00 and 11:45 to 15:45 UTC) are very smooth and show no significant variations. Thus, both water vapor instruments, FISH and HAI, measured only gas-phase water vapor, and no cloud particles were present in the vortex air. Both FISH and HAI observed these low water vapor mixing 10 ratios independently.
In order to show the fairly good agreement of both water vapor instruments FISH and HAI based on 1 Hz data at the observed low water vapor values, we choose two flight legs for the comparison (leg one: 09:45-10:45 UTC and leg two: 13:00-14:00 UTC). During flight leg one, both hygrometers have an absolute difference of 0.21 ppmv and 15 a mean relative difference of −14.9 % (±10.5 %, 1σ). During flight leg two, a difference of 0.26 ppmv and a mean relative difference of −5.9 % (±7.3 %, 1σ) is found. These rather small differences are consistent with the uncertainties of both hygrometers, which are ±0.45 and ±0.5 ppmv at the 2 ppmv level for FISH and HAI, respectively. The HAI data measured before 09:10 UTC (marked with gray in Fig. 2) are influ-20 enced by an untypical memory effect of the measurement cell during ascent, caused by a valve which was opened too late, but only when flying in the upper troposphere. As a result, the "wet" measurement cells were not sufficiently dried off with the high mass flows during take-off and ascent. This also explains the asymmetry in the deviations to FISH before and after the dive. In addition, the agreement of GLORIA water vapor ob-25 servations at flight level with the two in-situ instruments, FISH and HAI, is remarkably good, considering that GLORIA is a remote sensing instrument.
The GLORIA instrument measures quasi-vertical profiles along the flight path by viewing the atmosphere on the right side of HALO. Here, we focus on the H 2 O re- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | trieval product, which is shown in Fig. 3. We consider only the times from 08:00 to 14:00 UTC to focus on the air masses where GLORIA observed vortex air (the whole dataset is from 06:30 to 15:30 UTC). The dive is noticeable as the white area between 11:00 and 12:00 UTC, where GLORIA did not measure in order to prevent condensation of tropospheric water vapor on the cold instrument. The dry vortex air masses are 5 clearly visible between 08:00 and 13:00 UTC at flight level (black solid line in Fig. 3), but also below, down to altitudes of 7 km. As mentioned in Sect. 2.5, the quantities retrieved from GLORIA are approximately placed on a parabola following the tangent points through the atmosphere (e.g. 250 km horizontal distance to flight path at 9 km altitude). The dry region just before the dive was measured by GLORIA in the westerly 10 direction to the flight path on the way towards Antarctica, while the air masses after the dive are measured in the easterly direction on the way back to Cape Town. Thus, the dehydrated air masses below the thermal tropopause seem to cover a large region, having a dimension of at least 500 km horizontally at 9 km altitude. The thermal tropopause is also derived from the GLORIA retrieved temperature pro- 15 files and is marked with black dots. Interestingly, low water vapor mixing ratios around and below 2 ppmv are observed beneath the thermal tropopause. Especially in the time range from 09:00 to 10:30 UTC and from 12:00 to 13:30 UTC, very dry air masses with water vapor mixing ratios of 2-3 ppmv can be found far below the thermal tropopause, down to 7 km. This indicates that dehydrated stratospheric air masses could have 20 been transported through the thermal tropopause into the troposphere. The dynamic tropopause, which is between the −4 and −2 PVU isoline, is somewhat lower than the thermal tropopause at ∼ 7 km in the time range from 12:00 to 13:30 UTC and indicates a proceeding stratospheric intrusion. In Sect. 4.2, we analyze this transport process with the help of airmass backward trajectories. In addition, no water vapor values less 25 than 6 ppmv are found below the −4 PVU isoline in ECMWF data (not shown here). This shows, at least for this situation, that the transport process of dry stratospheric air masses into the troposphere is not captured by the ECMWF meteorological analysis (see Sect. 4). Embedded in the dry vortex air masses, small filaments (marked in Fig. 3 with red boxes) of enhanced water vapor are visible. These filaments may indicate rehydration layers, where sedimented ice particles from upper layers have sublimated and resulted in layers of enhanced water vapor as described in Khaykin et al. (e.g. 2013) for the Arctic. However, trajectory reverse domain filling (RDF) (Sect. 4.2) showed that these 5 air masses could not be attributed to rehydration. Also the in-situ measurements of FISH/HAI showed no evidence of a rehydration layer during the dive from ∼ 12.5 to 3 km and back. We have analyzed these somewhat moister air masses in more detail, as discussed in the next section.

ACPD
4 Trajectory based case study of 13 September 2012 10 Detailed investigation of the observed Antarctic dehydration is performed using airmass trajectories calculated with the tool CLaMS-traj which is part of the Chemical Lagrangian Model for the Stratosphere (CLaMS) (McKenna et al., 2002;Konopka et al., 2007). The trajectories are driven by the ERA-Interim reanalysis horizontal winds and diabatic heating rates (Dee et al., 2011;Ploeger et al., 2010). Besides temperature, 15 pressure, and humidity along the trajectories, the potential vorticity (PV) and thermal tropopause height (WMO lapse rate criterion) are added from the ECMWF meteorological fields. The temperatures from ECMWF along the flight path agree very well with the measured temperatures from the HALO aircraft (not shown here). The mean deviation on each flight level is mostly around 0.4 K, and the temperatures in ECMWF are 20 slightly lower than the measured temperatures.

History of dehydrated air masses along flight path
In this section, we investigate the history of the in-situ measured dehydrated air masses. For that purpose, the trajectories are calculated every 10 s along the flight path, each reaching 50 days backwards in time. For further analysis, we selected all Introduction trajectories between 12:20-12:29 UTC (in total 54), where the lowest water vapor mixing ratios were measured. However, the other trajectories with measured low water vapor mixing ratios between 2 and 3 ppmv show very similar history. Figure 4 shows the median saturation mixing ratio with respect to ice (orange line) and the frequency distribution (gray scale color-code) of every 1 h trajectory time step of all trajectories 5 calculated by ClaMS along the 50 days back in time. The trajectories show the same history in saturation mixing ratio from 13 September until 14 August, where the gray colors start to spread out. This indicates that the trajectories have the same path with the same temperature history for at least 30 days. After this time, the trajectories split up into two branches, indicating a bifurcation point and associated mixing on 14 August. 10 The trajectories within the branch with lower saturation mixing ratios of 2 to 3 ppmv below the median line stay together on the same path until the beginning of August. marked with a blue bar (see Fig. 4). Figure 4 illustrates the coincidence of low saturation mixing ratios along the trajectories with observed ice formation (CALIPSO) in the considered air masses. The trajectories pass several cold regions associated with low saturation mixing ratio where ice formation is initiated. If the ice cloud and supersaturation exist long enough (typ-25 ically around 0.5-1 day, Nedoluha et al., 2002) to allow the ice particles to grow to sizes of several microns (diameter of 10 to 20 µm), sedimentation of these ice particles begins and causes irreversible dehydration. The sedimented ice particles with these sizes fall with velocities of around 2.5 km d −1 (see Müller and Peter, 1992) until warmer ACPD 15,2015 Transport of Antarctic dehydrated air into the troposphere C. Rolf et al. temperatures in lower regions causes them to evaporate. With each pass of a cold region, more and more water vapor is removed from the airmass. As stated by Nedoluha et al. (2002), air masses must undergo several periods with dehydration before reaching the 1 to 2 ppmv mixing ratio level, which is typically observed by satellites within the vortex in the Antarctic spring. This periodic behavior is confirmed by Fig. 4, where trajectories undergo four events with low saturation mixing ratios and simultaneous ice cloud observations by CALIPSO. The first and the second period (1 and 9 August) reveal rather high saturation mixing ratios of around 3 ppmv. In every later period, colder temperatures and lower saturation mixing ratios are necessary to initiate additional ice formation and dehydration. So the third and the fourth period on 16 and 29 August 10 where CALIPSO observed ice show lower saturation mixing ratios of 1.5 and 1.2 ppmv, respectively. The last two events will have dried out the air masses to the final state of about 1.6 ppmv, which is observed by the FISH and HAI instruments. After this strong dehydration, the temperatures became warmer in the late Antarctic spring and summer and no additional dehydration took place according to the simulations. 15 The recurring low saturation mixing ratios and corresponding low temperatures are primarily caused by gravity waves, which are induced by the high topography of the Antarctic continent (Stohl and Sodemann, 2010). Especially, the first three ice formation events occurred just as the trajectories passed the Antarctic Peninsula (northernmost part), which is a strong source of gravity waves (e.g. Ern et al., 2011;Hoffmann et al., 20 2013). Only the last freezing event with the lowest supersaturation occurs above the Antarctic Plateau (central part) in the north-east of the continent.
After freezing, the ice particles can even reach the troposphere due to sedimentation, especially if dehydration takes place at low altitudes (11.5-12.5 km) as observed during the Antarctic ESMVal flight with HALO. The lack of rehydration signatures in the 25 GLORIA measurements (see Fig. 3) also indicates the permanent removal of water vapor from the stratosphere, as will be further discussed in Sect. 4.2. The air masses from which the ice particles originate and the air masses where they sublimate behave very differently in dynamical perspective. Due to the length of time since the last ACPD 15,2015 Transport of Antarctic dehydrated air into the troposphere C. Rolf et al. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | dehydration on 28 August (16 days before measurement) it is very unlikely that rehydrated air masses would still be found below the observed dehydrated air masses in the troposphere. In addition, the in-situ measurements taken during the dive confirm the absence of any observed rehydration features.

5
The trajectories for analyzing the transport of dehydrated air masses observed by GLO-RIA are calculated 150 days backward in time from each tangent point (in total 30 000 single trajectories). In addition, the trajectories are also calculated 30 days forward to show where the air masses are further transported. With this set of trajectories, the GLORIA measurements can be interpreted and the associated transport pathways can 10 be investigated.

Vortex air masses
From the in-situ ozone and methane measurements, it is visible that the HALO aircraft has penetrated the Antarctic vortex (see Sect. 3.2). The identification of vortex air masses in the GLORIA trace gas cross-sections is performed using the tangent point 15 trajectories. Figure 5 shows the amount of time that the trajectory for each GLORIA tangent point stayed within the vortex before the measurement took place. To calculate this so-called vortex indicator, the vortex border needs to be determined at each trajectory time step with the Nash criterion (Nash et al., 1996). Then, the time that each individual trajectory remains within the vortex border is counted backwards from 20 the time of measurement. This procedure is limited to the preceding 50 days. The blue color in Fig. 5 shows the trajectories that stay in the vortex for at least the last 50 days. In contrast, air masses that were never in the vortex are indicated by a deep red color. This includes tropospheric air, but also stratospheric air masses north of 48 • S. Interestingly, a sharp boundary between deep red and deep blue colors appears between 25 09:00 and 11:00 UTC and also between 11:45 and 12:20 UTC, which exactly follows the measured thermal tropopause (black dots). Here, the fundamental role of the thermal tropopause as a vertical transport barrier for stratosphere/troposphere exchange emerges.
In contrast, the air masses that spent between 5 and 40 days in the vortex (marked by light blue and reddish colors) indicate mixing of outside air into the vortex. Even at 5 the vortex edge and in the core of the vortex itself, small filaments are apparent. As indicated in Sect. 3.2, some filaments with elevated amounts of water vapor were also observed with GLORIA. Two of these filaments observed around 13:00 UTC between 10 to 12 km altitude can be found with this Nash PV criterion showing vortex residence times of 5 days (filament 6) and 35 days (filament 7), respectively. This shows that these 10 filaments are likely freshly mixed in from outside the vortex and thus could potentially contain higher water vapor content compared to the dehydrated vortex air and therefore weaken the hypothesis of rehydration signatures.
To show that all observed filaments (red boxes in Fig. 3) are caused by mixing, reverse domain filling (RDF) is applied to all trajectories, as already stated in Sect. 3.2 15 (similar as in Beuermann et al., 2002). For the RDF calculation, only advection and no mixing along the trajectories is assumed. In each time step of the RDF method, the ECMWF water vapor field is interpolated onto the locations of the trajectories and then projected onto the observed time-altitude grid of the GLORIA observations. Figure 6 shows ECMWF water vapor mixing ratios five days prior to the measure-20 ment time as projected by the trajectories. The red boxes highlight the position of the water vapor filaments observed by GLORIA (see Fig. 3). With some small deviations in time and altitude, almost all of the observed filaments in the vortex can be reproduced by the RDF method. Note, that the RDF method cannot produce the effect of rehydration. This indicates that the water vapor filaments observed by GLORIA were 25 not generated by rehydration, but are likely the signatures of moister air masses that were mixed into the vortex during the preceding 5 days. Continuing the RDF method further back in time, the filament 7 (≈ 12 km) at 13:00 UTC emerges 35 days before observation and indicates the origin to be another mixing event. The time of in-mixing of the filament 7 was already visible in our analysis in Fig. 5. Thus, it seems that these filaments originate from several mixing events, not from a single mixing event only. Furthermore, very dry air masses observed in the time between 09:00 to 10:30 and 12:00 to 13:30 UTC down to 7 km (see Sect. 3.2) can also be partly reproduced with the RDF method, indicating that transport is the main reason for the occurrence of dry air below 5 the thermal tropopause. The origin of these air masses and the transport mechanism are analyzed and discussed in the next subsection.

Transport of dehydrated air masses across the tropopause
The thermal tropopause in the Antarctic region is formed under unique climate conditions (Evtushevsky et al., 2011). The upper troposphere and lower stratosphere are 10 generally very cold with a very weak vertical temperature gradient during winter and spring. Therefore, the thermal tropopause in the Antarctic region is rather poorly defined (Stohl and Sodemann, 2010). This is also implied by the tropopause heights derived from GLORIA, which have a broad scatter (± 0.5 km) at some places and indicate the weak vertical temperature gradient. In addition, the PV gradient at the 15 transition between troposphere and stratosphere within the vortex (08:30-13:30 UTC) is small. The PV isolines (−4 and −2) in the vortex have a larger spacing than at the edge (jet region), where the vertical distance between the isolines decreases (08:15 and 13:30 UTC in Fig. 5). As a consequence, the tropopause and the PV gradient cannot serve as a strong transport barrier like they do in the mid-latitudes. Especially in for the last 70 days within the troposphere (dark red colors) can clearly be assumed to be tropospheric nature. The stratospheric and vortex air masses did not cross the tropopause (dark blue colors). The air masses below the measured tropopause are obviously more patchy and contain more trajectories originating from the troposphere (reddish colors). How-5 ever, some freshly mixed in stratospheric air masses are also discernible below the thermal tropopause (light blue colors), down to 7 km. The green framed areas in Fig. 7 mark all the trajectories having tropopause crossing times between 0 and 70 days and where GLORIA observations reveal water vapor content of less than 3 ppmv. Indeed, there is a very good correlation between the light blue air masses, which crossed the 10 tropopause between 2 and 6 days before, and the green framed areas marking the dehydrated air masses. In addition to the low water vapor content in the green framed areas, higher values of ozone were also observed by GLORIA, which reinforces the stratospheric origin of these air masses (not shown here). Figure 8a and b shows the median potential temperature (theta) and altitude distri- 15 butions of all trajectories within the green framed areas from Fig. 7 for the period of 5 months preceding (blue lines) and the first month following (red lines) the observation. The potential temperature is decreasing from about 380 to 310 K within five months, which reveals the well-known general descent of air masses in the polar vortex due to radiative cooling. After the trajectories have left the stratosphere and penetrated 20 the troposphere, the general subsidence in the Antarctic troposphere caused by the katabatic surface winds leads to adiabatic heating of these air masses. Additionally, RWB events facilitate a wave-driven, secondary circulation that transports warmer air masses from around the vortex edge to the colder inner vortex. Despite the dryness of these air masses, the slight enhancement of ozone and the residual water vapor 25 enable radiative cooling and a reduction in potential temperature i.e. descent of air masses continue down to 5 km altitude (300 K) in the time after the observation. Once transported into the troposphere, air masses will be transported down to near-surface level as suggested already by Roscoe (2004) and Stohl and Sodemann (2010). The point of in-mixing of stratospheric air masses into the troposphere and vice versa cannot be associated with a single event. In fact, successive Rossby wave activity at the vortex edge facilitates the transport process. If the isentropes cross the thermal tropopause (see Fig. 1b), and if the thermal tropopause is very weak at the lower border of the stratosphere, the tropopause is not able to prevent nearly vertical 5 stratospheric/tropospheric exchange. This is a completely different situation compared to the subtropical jet, where the thermal tropopause and the PV gradient are typically strong enough to prevent such an exchange.

ACPD
Here, strong Rossby wave activity is made visible by comparing the median latitude and equivalent latitude just a few days before the observation (6 to 17 September) as 10 seen in Fig. 8c and d. The latitude is oscillating from near 60 down to 80 • S, up to 55 • S, and then back to 65 • S, while the equivalent latitude in this time frame stays constant at around 70 • S. This implies no PV change and only an isentropic latitudinal displacement due to planetary Rossby waves. Thus, Rossby wave breaking events drive the mixing process together with isentropic transport through the thermal tropopause. 15 Just in the middle of the time frame, on the point furthest south at latitudes around 80 • S (see Fig. 8c) on the 9 September, the back trajectories split up into two different branches, indicating a bifurcation point (not shown here). This implies a mixing of air masses with different histories and also that this Rossby wave event is not the sole reason for the transport into the troposphere. Indeed, if one looks into the airmass 20 history, additional Rossby wave events occurred in the days and weeks before. So it is not possible to match one specific event to the transport of all air masses which were observed below the tropopause with low water vapor content (green framed areas in Fig. 7). Also, the aforementioned filaments with enhanced water vapor in the vortex correspond to in-mixing due to the wave activity just 5 days before observations. 25 The vortex becomes more and more unstable during the Antarctic spring and one may assume that this in-mixing event is already a first sign of the vortex break-up. Haigh and Roscoe (2009)  the observed in-mixing of vortex air masses occurs long (at least two months) before the final warming and break-up of the vortex. It is therefore more or less independent of the final warming event and seems to be a usual process in the Antarctic winter and spring. Finally, the reason for the large downward transport is a mixture of strong Rossby 5 wave activity, the presence of a very weak thermal tropopause, which fails to act as a strong enough transport barrier, and the additional radiative cooling with corresponding subsidence of the air masses.

Conclusions
Detailed observations of dehydration and stratospheric/tropospheric exchange in the 10 Antarctic UT/LS are very rare. In this study, high resolution in-situ measurements of strongly dehydrated air in the lower vortex with the hygrometers FISH and HAI are presented. The final dehydration can be traced back to several ice formation events, where water is sedimented out of the observed air masses. Dehydrated air masses are also observed with the high resolution remote sensing instrument GLORIA down to the 15 thermal tropopause, as well as below this point, down to the free troposphere to 7 km in altitude. Besides dehydrated air masses, small filaments of enhanced water vapor are observed by GLORIA in the vortex air. These filaments can be clearly assigned to in-mixing of moister tropospheric air masses into the dry vortex at several points in time, indicating that this transport process occurs frequently. 20 With an extensive trajectory case study, dry air masses below the thermal tropopause are determined to have stratospheric origin. It is shown that these dry air masses are transported further down to near surface levels within one month. Both the in-mixing of moist tropospheric air into the stratosphere as well as the large-scale transport of dehydrated air masses into the troposphere confirm the ineffectiveness 25 of the thermal tropopause as a transport barrier during Antarctic spring and winter (e.g. Roscoe, 2004;Evtushevsky et al., 2011). This transport is not caused by a single Introduction Rossby wave event, but rather the result of frequent in-mixing due to strong wave activity in combination with isentropes crossing the weak tropopause. Once the air masses have been transported into the troposphere, radiative cooling causes their subsidence to near surface levels around 5 km. The implication of frequent in-mixing of dry stratospheric air masses into the troposphere could on the one hand significantly reduce 5 cloudiness and precipitation, and on the other hand increase ozone in the troposphere above Antarctica. In conclusion, frequent tropospheric/stratospheric intrusions can influence or be one cause of the unique Antarctic climate. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | in the lower Antarctic stratosphere during late winter and early spring, 1987, J. Geophys. Res., 94, 11317-11357, doi:10.1029/JD094iD09p11317, 1989  The black contours illustrate the horizontal westerly wind from ECMWF data. Yellow dots represent the vortex edge derived from the Nash criterion (Nash et al., 1996) based on ECMWF data. (a) Horizontal map of water vapor mixing ratios measured by FISH (5 min averaged data) is color coded on the flight path (gray color indicate data gaps due to the dive). (b) Meridional cross-section of equivalent latitude along the flight path (orange line) calculated from ECMWF data. Blue dots represent the ECMWF thermal tropopause. 15,2015 Transport of Antarctic dehydrated air into the troposphere C. Rolf et al.    Fig. 7 for the period of 5 months before (blue lines) to 1 month after the measurement (red lines): (a) median of potential temperature (theta), (b) median altitude above mean sea level, (c) median latitude from 1 month before to 1 month after observation, (d) median equivalent latitude from 1 month before to 1 month after observation. The gray shaded areas marks the SD of the 1400 trajectories.