In-situ observation of Asian pollution transported into the Arctic lowermost stratosphere

On a research ﬂight on 10 July 2008, the German research aircraft Falcon sampled an air mass with unusually high carbon monoxide (CO), peroxyacetyl nitrate (PAN) and water vapour (H 2 O) mixing ratios in the Arctic lowermost stratosphere. The air mass was encountered twice at an altitude of 11.3 km, ∼ 800 m above the dynamical tropopause. 5 In-situ measurements of ozone, NO, and NO y indicate that this layer was a mixed air mass containing both air from the troposphere and stratosphere. Backward trajectory and Lagrangian particle dispersion model analysis suggest that the Falcon sampled the top of a polluted air mass originating from the coastal regions of East Asia. The anthropogenic pollution plume experienced strong up-lift in a warm conveyor belt (WCB) 10 located over the Russian east-coast. Subsequently the Asian air mass was transported across the North Pole into the sampling area, elevating the local tropopause by up to ∼ 3 km. Mixing with surrounding Arctic stratospheric air most likely took place during the horizontal transport when the tropospheric streamer was stretched into long and narrow ﬁlaments. The mechanism illustrated in this study possibly presents an important 15 pathway to transport pollution into the polar tropopause region.

may take place in the vicinity of fronts (Esler et al., 2003) or in the WCB outflow region (Cooper et al., 2002;Mari et al., 2004). Mixing with stratospheric air may also occur in the mid-and upper troposphere (Parrish et al., 2000;Brioude et al., 2007;Stohl et al., 2007). In the extra-tropical UTLS (upper troposphere/lowermost stratosphere) region, large-scale stirring and inter-mingling of air masses leads to the formation of 10 a mixing layer, and can be quantified using tracer relationships (Fischer et al., 2000;Zahn et al., 2000;Hoor et al., 2002Hoor et al., , 2004Pan et al., 2004Pan et al., , 2007Kunz et al., 2009). Exchange across the extratropical tropopause plays an important role for the trace gas composition in the lowermost stratosphere, affecting both local chemistry and the radiative budget. It is therefore notable that Lagrangian transport climatologies show 15 a potential for WCB trajectories to reach the tropopause region (Stohl, 2001;Wernli and Bourquoi, 2002). A few percent of these trajectories even enter the lowermost stratosphere, not only within active WCBs, but also a few days after the WCBs decay (Eckhardt et al., 2004). Since some of the WCBs reach polar latitudes, it is conceivable that anthropogenic pollution can be transported into the polar lowermost stratosphere 20 within a few days after emission. This transport pathway has not yet been verified by observations, however, measurements show in general relatively high CO mixing ratios of 100 nmol mol −1 and more in the polar lowermost stratosphere (Cooper et al., 2005;Tilmes et al., 2010).
In the present paper we demonstrate that Asian pollution can reach the Arctic 25 tropopause region within a few days after emission. To the best of our knowledge, our study presents the first in-situ observation of a distinct Asian pollution plume recently transported into the Arctic lowermost stratosphere. The anthropogenic emissions were up-lifted within a warm conveyer belt over eastern Russia, transported across the Arctic and probed by the Falcon above northern Greenland, where it was already mixed with stratospheric air. The aim of this paper is to present in-situ trace gas observations, to analyze the origin of the encountered air mass, and to discuss its lifting process and transport history. The present paper is organized as follows: Sect. 2 describes the methods used in 5 the analysis. A detailed description of the case study is given in Sect. 3, beginning with a presentation of the meteorological situation at the time of our measurements (Sect. 3.1) and a discussion of the in-situ observations (Sect. 3.2). Subsequently, the source region of the encountered air mass is identified by using a Lagrangian particle dispersion model and a backward trajectory model (Sect. 3.3). The analysis of the 10 meteorological situation in the source region (Sect. 3.4) as well as during transport to Greenland (Sect. 3.5) is followed by a discussion of tracer correlations (Sect. 3.6.1) and mixing processes that have occurred during the cross-polar transport (Sect. 3.6.2). Section 4 provides the summary and conclusions.
2 Data and model descriptions 15

Chemical measurements
The data presented in this paper were measured on 10  performed, covering latitudes from 57 • N to 81.5 • N, longitudes from 66 • W to 16 • E and altitudes up to 11.8 km. The Falcon was equipped with instruments to measure trace gases, aerosol microphysical properties as well as meteorological parameters. The present study focuses on the trace gas measurements. CO was detected by vacuum-UV fluorescence (Gerbig et al., 1999; accuracy ±5 %, detection limit 2 nmol mol −1 ), 5 ozone by UV absorption (TE49C; ±5 %, 1 nmol mol −1 ) and CO 2 via an infra-red absorption technique (Li-COR 7000; ±5 %, 5 nmol mol −1 ). NO and the sum of oxidized nitrogen compounds, NO y , were measured with a chemiluminescence detector (Ziereis et al., 2000; ±10/15 %, 10/15 pmol mol −1 ). NO y species (NO y = NO, NO 2 , NO 3 , N 2 O 5 , HNO 3 , PAN. . . ) hereby were first converted to NO at the surface of a gold converter by 10 adding CO as reducing agent, and then detected as NO. Peroxyacetyl nitrate (PAN) measurements were performed with a Chemical Ionisation -Ion Trap Mass Spectrometer (Roiger et al., 2011; ±10 %, 25 pmol mol −1 ), and the H 2 O data shown in the present paper were obtained by a Lyman-alpha hygrometer (Zöger et al., 1999; ±6 %, 0.1 µmol mol −1 ).

FLEXPART simulations
For the analysis of all GRACE flights, the FLEXPART model was used in order to attribute anthropogenic and biomass burning pollution sources to the in-situ observations. FLEXPART is a Lagrangian particle dispersion model (Stohl et al., 2005, and (Stohl et al., 2007). All pollution tracers underlie passive transport without involving removal processes. After a life-time of 20 days, the tracers are removed under the assumption that they become incorporated into the so-called atmospheric background . 15 The present study also uses FLEXPART backward simulations. These were started along the flight track every time the Falcon moved by 0.15 degrees in either longitude or latitude, or whenever the altitude changed by 10 hPa. For each release 60 000 particles were followed backward in time for 20 days. The model output of such backward simulations consists of fields of emission sensitivities which can be multiplied with emis-20 sion fields to yield maps of source contributions for passive tracers . When spatially integrated, tracer mixing ratios are obtained, allowing to construct tracer time series along the flight path. We consider CO, NO 2 and SO 2 separately for both anthropogenic and biomass burning pollution from Asia, North America and Europe. Further information is given by the fraction of particles remaining in the stratosphere Introduction

LAGRANTO trajectories
The Lagrangian Analysis Tool (LAGRANTO; Wernli and Davies, 1997) was used to investigate the origin of the observed air masses over Greenland. The calculations are based on three-dimensional wind fields of 6-hourly ECMWF operational analyses at a 1 • horizontal grid resolution. Several variables characterizing the physical state of the air parcels including potential temperature, specific humidity and potential vorticity were traced along the trajectory paths.
In this case we used two different set-ups of LAGRANTO to calculate forward and backward trajectories. 8-day backward trajectories were started close to the observation time on 10 July,18:00 h UTC at a constant pressure level (see Sect. 3.3.2).
Additionally, a method to diagnose warm conveyor belt trajectories was used (Wernli and Davies, 1997), by selecting only boundary layer air parcels that ascend poleward to the upper troposphere and lowermost stratosphere over a short time period (see Sect. 3.4). 15 The Antarctic Meteorological Research Center (AMRC) provides Arctic composite satellite images every three hours, creating a total of eight images per day (Lazzara and Knuth, 2011). Geostationary and polar-orbiting satellites are used to generate the composite and, depending on the geographical location, can include observations from the following: Polar Orbiting Environmental Satellite (POES), Geo- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | polar orbiting observations. In general terms, the composite satellite images depict the temperature of the Earth as seen by the imagers on board the satellites, which measure the intensity of the radiation emitted by the Earth at approximately 11.0 µm, the infra-red window channel. The images are colour enhanced such that yellow and red colours highlight the cold upper-level cloud tops, green colours underscore the tops of the warmer mid-level and low-level clouds, and blue colours are the much warmer surface of the Earth. In the present paper, the composites are combined with overlays of the FLEXPART Asian anthropogenic CO tracer fields. The passive CO tracers with a life-time of 20 days were obtained from the FLEXPART forward model runs (see Sect. 2.2) and are represented in the satellite images by white contour lines.

Meteorological situation at upper levels
The weather situation close to the time of observation (10 July 2008, 18:00 UTC) is presented in Fig. 1a using 300 hPa geopotential height (black isolines) and pressure altitude p 2PVU of the dynamical tropopause, defined as the 2 potential vorticity units 15 (PVU) surface (colour-coded). The red line gives the Falcon flight path. The geopotential height shows a large-scale trough over north-eastern Canada with a discrete low over the Davis Strait. The entire trough region is characterized by a low dynamical tropopause (equivalent to high values of p 2PVU , up to ∼400 hPa within the closed contour line of the low). On its leading edge, the low has southerly flow at 300 hPa 20 (see geopotential isolines) and the dynamical tropopause is situated at ∼300 hPa. Above northern Greenland, north of a nearly east-west oriented narrow filament with a lower tropopause (p 2PVU > 400 hPa), the dynamical tropopause is at higher levels (p 2PVU ∼ 230 hPa). The connected air mass is located in the diffluent exit region of a jet. At this time the jet was located right over the pole, as shown by the widening spac- 25 ing of geopotential isolines from the North Pole towards the most northern part of the 10

In-situ observations
On the flight on 10 July 2008, CO values up to 138 nmol mol −1 were observed at 11.3 km altitude, whereas during GRACE typical CO mixing ratios at this altitude were ∼40 to 50 nmol mol −1 . Figure 1b shows the Falcon flight-path, for illustration purposes it is colour-coded by the measured ∆CO above the altitude-dependent CO background 5 (see colour-scale on the right). The CO background here is defined as the 30th percentile of all POLARCAT-GRACE measurements, calculated separately for altitude bins of 1 km. ∆CO values of <10 nmol mol −1 are shown in dark blue. The marker size represents the flight altitude (larger markers indicate higher altitudes). After take-off in Kangerlussuaq, the Falcon climbed up to 11.3 km altitude while fly- 10 ing north (up to point P1), and then ∼60 km to the east (towards point P2). On the way back to the south, the Falcon descended at ∼75.5 • N to 7.9 km before landing again in Kangerlussuaq. The layer with enhanced CO was traversed twice, both on the outbound and the return leg, right before point P1 and after point P2. The maximum horizontal separation is 60 km (P1 to P2). Comparison with Fig. 1a shows that the layer 15 with enhanced CO agrees well with the region of elevated tropopause, as indicated by the lower atmospheric pressure at the 2 PVU surface. Figure 2 presents the time-series for several trace gases for a part of the flight between 15:30 and 17:00 UTC (see time markers in Fig. 1b). It shows CO and O 3 (a), NO and NO y (b) together with PAN and CO 2 (c). Also given are the potential temperature 20 Θ and the water vapour mixing ratio (d). The time resolution is 1 s, except for PAN, which is reported every 2 s. Panel (e) shows the flight altitude (colour-coded by the latitude of the Falcon), and potential vorticity as interpolated from the ECMWF analysis fields. The two way points P1 and P2 are marked as vertical lines (see Fig. 1b). The red shaded area in Fig. 2  ("LMS"), which indicates that the aircraft possibly did not traverse the entire mixing region. The time-series of the trace gases inside the "MR" is nearly symmetric due to its twofold sampling. The CO enhancement in the "MR" was accompanied by an increase of the tropospheric pollutant PAN (up to 330 pmol mol −1 compared to ∼40 pmol mol −1 in the "LMS"), and of water vapour (∼80 µmol mol −1 compared to ∼20 µmol mol −1 in the 10 "LMS"). The trace gases with generally higher values in the stratosphere than in the unpolluted regions of the troposphere (i.e. O 3 , NO and NO y ) showed smaller mixing ratios within the "MR" of down to 100 nmol mol −1 , 30 pmol mol −1 and 900 pmol mol −1 , respectively. The anticorrelation of the different trace gases indicates that the Falcon sampled an air mass of both tropospheric and stratospheric origin. The interpretation of the 15 chemically inert greenhouse gas CO 2 (atmospheric lifetime of ∼100 years) however is more complicated, due to underlying temporal and spatial variations (see Sect. 3.6). In some parts of the "MR" some smaller scale filaments are observed (indicated by black arrows). Thermodynamic parameters also vary within the "MR". Potential temperature Θ and 20 potential vorticity both decrease, pointing to a less stable stratification in the "MR". Potential vorticity during this part of the flight had a minimum value of ∼5 PVU, and the minimum vertical distance between flight altitude and the 2 PVU tropopause was ∼800 m (see Sect. 3.5).
ACPD 11,2011 Asian pollution transported into the Arctic lowermost stratosphere

FLEXPART backward simulations
The results of the FLEXPART backward runs initialised along the flight track are shown in Fig. 3b. To allow an easier comparison, the in-situ O 3 and CO time-series are repeated in Fig. 3a. Generally, the structure of the "MR" is very well captured by the 5 FLEXPART model. The blue line in the middle panel gives the fraction of air with a stratospheric origin which decreases inside the "MR" to a minimum of ∼50 %. The red pattern in Fig. 3b represents the Asian excess-CO, indicating that the tropospheric part of the encountered air mass was polluted by anthropogenic emissions of Asian origin. According to the FLEXPART simulations, contributions from industrial pollution sources from other continents or from biomass burning are negligible. The CO source contribution plot for 16:11 UTC (CO maximum) is given in Fig. 3c. This plot represents the product between the anthropogenic emission flux (taken from the emission inventories) and the footprint emission sensitivity. The latter is proportional to the residence time of the particles over a unit area in the lowest 100 m (Stohl 15 et al., 2005). The CO source contribution plot shows that the pollution mainly originates from the North China Plain, one of the most densely populated regions in China, with some minor contributions from the North East China Plain and Korea. According to the model, the pollution was emitted approximately 6-10 days before our measurements (not shown).

LAGRANTO backward trajectories
In order to analyze the transport history of the sampled "MR" and "LMS" air masses, we calculated an ensemble of 8 day LAGRANTO backward trajectories. The results are illustrated in Fig. 4 Figure 4b to d shows the temporal evolution of meteorological parameters along the trajectories (e.g. pressure, humidity, potential temperature, potential vorticity). We can mainly distinguish between two different air mass origins in this area. Trajectories in the southern part of the box were transported from northern North Amer-5 ica towards Greenland, remaining above 300 hPa. These trajectories characterize the pathway for the sampled "LMS" air masses. The cyclonic curvature indicates transport within the large scale trough described in Sect. 3.1 (see also Fig. 1a). Trajectories in the northern part of the box were traced back across the North Pole to Siberia/Russia. Some parcels moved directly over the pole (yellow), while others were influenced by the upper level high that caused an anti-cyclonic rotation (blue and green).
The time series of pressure along the trajectories shows that independent of their origin, most of the parcels were transported at high altitudes above 300 hPa (see Fig. 4b).
Their PV values were predominantly larger than 2 PVU which points to stratospheric transport (not shown). Only a small number of the trajectories experienced a strong 15 vertical displacement. Remarkably, three of these parcels (thick grey lines in Fig. 4) originate from the boundary layer in the industrialized coastal regions of China and Korea, where they had the possibility to take up anthropogenic pollution. This is in good agreement with the results of the FLEXPART analysis (see Sect. 3.3.1). The subsequent fast and strong ascent from ground level to the upper troposphere is ac-20 companied by a strong loss of humidity and an increase of the potential temperature as a result of latent heat release. The strong up-lift of the trajectories took place over eastern Russia between 5 July, 12:00 UTC (−126 h) and 6 July, 18:00 UTC (−96 h), as indicated by the grey dots in Fig. 4. According to FLEXPART, the pollution was emitted between ∼1-4 July, i.e. ∼6-10 days before our measurements. The polluted 25 air mass was transported for a few days at low levels towards the north-east before it was exported from the boundary layer. The characteristics of strong and poleward ascent in combination with latent heat release suggests up-lift by a WCB, which is further investigated in the following section. 11,2011 Asian pollution transported into the Arctic lowermost stratosphere  Figure 5 shows the position of the WCB parcels during the evolution of a cyclone with an embedded WCB. As the up-lift of the relevant air mass took place between 5 and 6 July (see Fig. 4), we only show WCB trajectories that ascended between 4 July, 00:00 UTC to 7 July, 18:00 UTC. On 5 July, 00:00 UTC (Fig. 5a), the WCB parcels were located over the coastal regions of China and Korea as well as above the Yellow Sea, predominantly at altitudes 15 below 900 hPa, and transported by south-westerly winds. The distribution matches partly the FLEXPART CO source contribution (Fig. 3c), with a shift towards the northeast due to the preceding low-level transport. Twelve hours later, the air mass was advected further north-eastward and a portion of the eastern parcels began rising when an upper level trough approached and caused cyclogenesis (Fig. 5b). At 6 July, 20 00:00 UTC a surface low developed and the parcels were situated close to the cyclone center and in the associated warm sector (Fig. 5c). North-east of the low, the parcels lined up along the warm front and reached the upper troposphere with pressure values up to 300 hPa. The parcels south of the surface low were located closer to the ground (Fig. 5d). Another 12 h later (6 July, 12:00 UTC) the surface low propagated 25 further north eastward (Fig. 5e). Except for some parcels located at mid-levels (500-800 hPa), the bulk reached the upper troposphere and was embedded in the jet stream located above the east coast of Asia. The majority of the WCB parcels were located ACPD 11,2011 Asian pollution transported into the Arctic lowermost stratosphere in the diffluent exit region of the jet stream. Most parcels moved southward, however, those on the cyclonic side of the jet were embedded in the north-eastward flow on the leading edge of a trough approaching from the north (see Fig. 5f). These parcels subsequently moved over the pole following the pathways illustrated in Fig. 4. Of the 4296 selected WCB trajectories, a subset of 191 trajectories (4.5 %) was 5 transported into the Arctic. 92 % of these Arctic WCB trajectories intersected the dynamical tropopause (PVU > 2), compared to only ∼18 % of the remaining WCB trajectories. Most parcels reached the 2 PVU level right after the WCB ascent or when they were again lifted on the leading edge of the upper level trough. The latter ascent is also reflected in the pressure time series of Fig. 4 (see yellow and blue trajectories between 10 −84 and −72 h). The trajectories most likely to reach the Arctic were those initialised from the cyclone centre (not shown).

Cross-polar transport and vertical structure of the polluted Asian air mass
The transport of the Asian pollution plume across the pole is illustrated in Fig. 6. The left panel shows satellite image composites for 4 points in time, covering the time period 15 after the up-lift within the WCB (6 July 2008) to the day of our measurements above northern Greenland (10 July 2008 of the cold front (Fig. 6a). Soundings from four weather stations (black asterisks in Fig. 6a) indicate that the minimum cloud top temperatures of ∼240-223 K within the WCB correspond to a maximum altitude of ∼11 km. CO is used as a passive tracer with a life-time of 20 days in the FLEXPART model runs, therefore other CO pollution "remnants" north of Kamchatka/east of North Siberia are also visible in the upper tropo-5 sphere, possibly from earlier lifting events. However, the polluted air mass lifted within the studied mid-latitude cyclone reached high potential temperatures of 340 K, shown in a comparison with Fig. 6b. As indicated by the red line, a tropospheric streamer had already begun to intrude into the Arctic, ahead of an upper level trough. In the following hours and days, a low pressure system formed, deepened and moved further to the north (indicated with "L" in Fig. 6a, c, e). The polluted Asian air mass therefore was advected towards the North Pole (Fig. 6c, 7 July, 18:00 UTC). Whereas most of the Asian CO tracer was advected eastwards, some of the Asian CO was embedded in the streamer (Fig. 6d). A comparison of Fig. 6c and Fig. 6d shows that the region of enhanced FLEXPART tracer is accompanied partly by upper level clouds. During 15 its journey, the polluted streamer was elongated and deformed ( Fig. 6e and f, 9 July, 06:00 UTC). Finally, a portion of the polluted streamer reached northern Greenland, where it was sampled by the Falcon on 10 July ( Fig. 6g and h, 10 July,18:00 UTC). Figure 6h illustrates the good agreement with our in-situ data, showing that the polluted streamer was located almost perpendicular to our flight track, and was not completely 20 traversed (see also Sect. 3.6).
During the cross-polar transport, the tropospheric streamer elevated the polar tropopause substantially (see also Fig. 1a). Figure 7 illustrates the temporal evolution of the vertical temperature structure during the advection of the streamer. Temperature profiles from soundings before, during and after the passage of the polluted Arctic air mass (blue dotted lines), whereas the tropopause temperature is up to ∼7 K lower. The region above ∼13 km is less affected by the dynamics of the tropospheric streamer. During the passage of the Asian air mass the tropopause is sharper than before and after. The maximum thermal tropopause is ∼11.5 km (11 July, 00:00 UTC), which is slightly above our flight altitude (11.3 km). Note however the considerable tem-5 poral and spatial difference between the soundings and our in-situ measurements. An analysis of ECMWF temperature profiles indicates that e.g. at the time of the observation, the thermal tropopause was ∼500 m lower in the sampled part of the streamer than above Eureka. Figure 8 shows a vertical cross section of the FLEXPART Asian excess-CO tracer 10 around the time of the flight (see black line in Fig. 6h). Also given are isentropes (black lines), the 2 PVU dynamical tropopause (thick black line), as obtained from the ECMWF analysis data, as well as isotaches (blue lines). The red crosses indicate the position of the WCB forward trajectories. Only WCB trajectory points located within 200 km of the Falcon flight path are shown (within 16:30 h ± 3 h). Their abundance close to or above 15 the 2 PVU tropopause reflects the strong selection criteria of the WCB forward analysis (∆600 hPa within 24 h).
According to the FLEXPART model simulation, the Falcon probed the topmost part of an Asian pollution plume with a vertical extension of several km. The Asian excess-CO shows a maximum of ∼90 to 100 nmol mol −1 at about 9 km just below the dy-20 namic tropopause, at the edges of the tropospheric streamer. A part of the pollution reached the lowermost stratosphere, as indicated both by the 2 PVU tropopause and the closer spacing between the isentropes. According to ECMWF analysis, the plume was sampled ∼800 m above the dynamic tropopause and ∼400 m above the thermal tropopause, respectively. The Falcon nearly reached the centre of the streamer which Siberian forest fires, but only at altitudes between ∼4 to 9 km and hence, below our flight level and below the 2 PVU tropopause (not shown).

Tracer-tracer correlations
As shown in Sects. 3.2 and 3.3, both in-situ measurements and the FLEXPART anal-5 ysis indicate that we probed an air mass of both stratospheric and tropospheric origin.
In this section we discuss the chemical properties of the sampled air mass using correlations between different trace gases. Tracer-tracer relationships are a common tool to investigate mixing, chemical and/or transport processes in the tropopause region (Hoor et al., 2002(Hoor et al., , 2004Pan et al., 2004Pan et al., , 2007Kunz et al., 2009).

10
The red crosses in Fig. 9a-d represent the 10 July data observed inside the "MR". For comparison, the 10 July "LMS" data are also shown (black circles). The grey open circles represent all other GRACE data, as obtained from all 16 local flights. For the following discussion it is worth noting that although the GRACE measurement area was far from pollution sources, the sampled data do not represent summertime CO 15 background values. In the free troposphere values were biased towards higher CO mixing ratios (as well as PAN and NO y ). The Arctic mid-troposphere frequently was impacted by long-range transport of aged but CO-enriched pollution plumes, originating mainly in the boreal forest regions in Canada and Siberia (Singh et al., 2010;Schmale et al., 2011). Therefore, the median CO value between 4 to 9 km during GRACE was   (Solomon et al., 1985). In the troposphere, O 3 values are highly variable due to in-situ photochemical production/destruction, dry deposition and import of stratospheric air. The main CO source in the background troposphere and the stratosphere is the oxidation of CH 4 . Other tropospheric CO sources are incomplete combustion processes from human activities or biomass burning, which makes CO an excellent pollution tracer. Its major sink is the reaction with OH radicals, which results in generally lower CO background values in the summer troposphere compared to winter, and for the same reason in the tropics compared to the middle and polar latitudes (Novelli et al., 1992(Novelli et al., , 2003. 10 The "MR" data in the O 3 -CO scatter plot in Fig. 9a can be represented by a regression line, a so-called mixing line, which is formed by irreversible mixing of air parcels of both stratospheric and tropospheric origin. The slope of such a mixing line in a tracer-tracer space is dependent on the trace gas mixing ratios of the initial air masses (Plumb and Ko, 1992;Hoor et al., 2002). The tropospheric character is found to be 15 highest in the centre of the "MR" (see also Fig. 1b). Here, the CO mixing ratio of ∼138 nmol mol −1 suggests predominantly tropospheric origin whereas the ozone mixing ratio of ∼100 nmol mol −1 is already at the upper limit of typical tropospheric values.
Although photochemical in-situ production in strongly polluted air masses may cause such high ozone values, this might indicate that also this part of the streamer had al-20 ready mixed with ozone-rich stratospheric air masses. At the edges of the "MR", the entrainment of O 3 -rich, stratospheric air is more prominent. A comparison of the "MR" data with the remaining GRACE data shows that the "MR" CO values in the tropopause region and above (i.e. O 3 > 100 nmol mol −1 ) are biased towards higher CO mixing ratios. This suggests that a CO-enriched air mass was mixed 25 into the lowermost stratosphere, which confirms the FLEXPART model results. If mixing and not chemical transformation is the dominant process, the extrapolation of the regression slope to an assigned tropospheric value of O 3 gives an indication of the CO concentration of the "tropospheric end-member" (Hintsa et al., 1999;Hermann et al., Hoor et al., 2002). If we assume an upper tropospheric ozone mixing ratio of 70 (±30) nmol mol −1 , we can calculate an initial CO mixing ratio of 143 (±10) nmol mol −1 , as indicated by the dark red marker in Fig. 9a. It has to be pointed out that the tropospheric end-member represents only the tropospheric air mass just before it mixed with the stratospheric air. The WCB inflow regions however can be quite large (Wernli, 5 1997;Eckhardt et al., 2004), which means that the anthropogenic emissions certainly have been diluted with less polluted air already in the troposphere. Figure 9b shows the relation between O 3 and H 2 O. Water vapour mixing ratios show an even stronger gradient across the extratropical tropopause than O 3 and CO. The entry of H 2 O into the stratosphere occurs primarily in the tropics, however in the extratrop-10 ics a moist mixing layer in the lowermost stratosphere is an indicator of troposphere-tostratosphere transport occurring at higher latitudes (e.g. Krebsbach et al., 2006). The moisture in the extratropical tropopause region is largely a function of the temperature at the cold point tropopause, but depends also on the relative strength of the different transport paths. Water vapour has a long chemical life-time in the tropopause region, 15 but it is not conserved if condensation followed by precipitation occurs.
Apparently the "MR" data are divided into two regimes A and B, which are sepa- lowermost stratosphere, though an upscaling cannot be done from our single event study. This might have implications for the temperature and chemistry of the UTLS, as water vapour plays an important role in the radiation budget and is a primary source of HO x radicals (e.g. Fueglistaler et al., 2001;Kunz et al., 2009). The data of group B correspond to the centre of the "MR", and clearly deviate from the regression line.
Obviously water vapour was removed in this part of the air mass due to phase transition followed by sedimentation. This assumption is supported by the occurrence of the high-level clouds in some parts of the streamer, visible in the satellite images (e.g. Fig. 6c, e). A CO 2 -CO scatter plot is presented in Fig. 9c. Apart from an increasing trend dur-10 ing the last decades due to industrialization, CO 2 has seasonal, diurnal, and spatial variations. It is, for example, depleted during summertime in the lower troposphere by plant assimilation. This seasonal cycle, typically ±1 µmol mol −1 in the Southern Hemisphere up to ∼15 µmol mol −1 in the northern boreal forest zone (Wigley and Schimel, 2000), is propagated into the lowermost stratosphere (e.g. Sawa et al., 2008;Boenisch et al., 2009). During the GRACE campaign, the observed CO 2 mixing ratios were on average ∼2 µmol mol −1 lower in the troposphere than in the lowermost stratosphere: Median CO 2 for O 3 < 100 nmol mol −1 was 381.4 ± 0.4 µmol mol −1 compared to 383.3 ± 0.2 µmol mol −1 for O 3 values >100 nmol mol −1 .
The "MR" data show a linear dependence between CO 2 and CO. Positive correla-20 tions between CO 2 and CO are typically observed for combustion sources, and the CO 2 /CO emission ratio may provide information about the combustion efficiency. The CO 2 /CO slope in this case is mainly driven by the mixing processes between the tropospheric and the stratospheric air masses, but it points clearly towards higher CO 2 values with increasing CO. The CO 2 /CO ratio in this tropospheric pollution therefore 25 was rather high compared to the ratios observed in all other CO plumes, which originated mainly from boreal biomass burning. This might be due to a combination of two effects. The fuel carbon conversion of biomass burning is relatively low, especially compared to industrial pollution, e.g. to power plants with high combustion efficiencies. This leads to initial CO 2 /CO slopes of up to a magnitude lower than those from anthropogenic plumes (Andreae and Merlet, 2001;Suntharalingam et al., 2004). Additionally, the CO 2 "background" might differ for these two different source regions. In densely populated areas like the North China Plain, respiratory CO 2 of urban residents as well as agricultural soils and livestock is part of the anthropogenic plume and increases the 5 CO 2 /CO ratio (Wang et al., 2010). On the contrary, CO 2 plant uptake in boreal regions during summer leads to a decrease of this ratio. Figure 9d shows a NO y -O 3 scatter plot. In the stratosphere, NO y is primarily composed of HNO 3 and NO x , produced mainly in the tropics by the photolysis of N 2 O followed by oxidation (Murphy et al., 1993). The NO y composition in the troposphere 10 might differ considerably depending on sources/sinks, as well as region and altitude. As obvious in the GRACE data, correlations between NO y and O 3 are typically steeper in the troposphere than in the stratosphere (Murphy et al., 1993). The NO y values of the "MR" data do not stand out from the remaining GRACE measurements, suggesting that although the air mass was influenced by anthropogenic pollution and hence, by 15 emission of NO x , the main part of the emitted NO y did not reach high altitudes. This is in agreement with earlier studies which show that the export of NO y from the boundary layer is not very efficient Koike et al., 2003;Nowak et al., 2004), mainly because of the removal of water-soluble HNO 3 due to precipitation processes. Furthermore, the polluted air mass resided a few days in the boundary layer before it 20 ascended within the WCB (see Sect. 3.3.2), which means that also dry deposition at the surface might have played a role.

ACPD
Beyond HNO 3 , PAN is the most dominant NO y species found in aged pollution (Koike et al., 2003;Nowak et al., 2004;Singh et al., 2010). PAN is a product of hydrocarbon-NO x chemistry and has long life-times in the mid-and upper troposphere (Talukdar 25 et al., 1995). In the region of weakest stratospheric influence (i.e. at highest CO), a maximum PAN/NO y ratio of ∼0.3 is observed. This ratio is rather low, PAN/NO y ratios of up to ∼0.85 have been measured in other CO plumes during GRACE, and ratios of e.g. ∼0.65 in aged Asian pollution (Nowak et al., 2004 deposition and survives the up-lift in a mid-latitude cyclone (e.g. Miyazaki et al., 2003), but it has a life-time of only a few hours at the warm temperatures prevalent in the boundary layer (Talukdar et al., 1995). The low PAN content therefore is very likely the result of thermal decomposition during the low-level transport in the first days after emission. PAN may recycle back especially during night-time. However, in the humid 5 boundary layer the released NO 2 is efficiently oxidized to HNO 3 (e.g. Bradshaw et al., 2000), especially during summer at these latitudes. As discussed earlier, most of the HNO 3 in the WCB was presumably removed by wet deposition. Compared to the "MR" slopes, the regression lines of the "LMS" data are much steeper for all correlations. The CO values of up to 50 nmol mol −1 are well above the stratospheric equilibrium value of 12 to 15 nmol mol −1 (Flocke et al., 1999) which suggests also for these air masses recent influence from the troposphere (e.g. Hoor et al., 2002Hoor et al., , 2004Tilmes et al., 2010), but to a lesser extent. Compared to the tropospheric air mass mixed into the "MR", the tropospheric air in the "LMS" case contained much less H 2 O, PAN and CO, but slightly higher NO y and much higher CO 2 .

Mixing of Asian pollution with Arctic stratospheric air
As discussed before, the polluted Asian air mass had already mixed with stratospheric air when it was sampled by the Falcon, although some interleaved filaments were still observed (see Sect. 3.2). An interesting question is of course, when and why did these air masses mix? As mentioned in Sect. 3.4, most air parcels reached the 2 PVU 20 level during the up-lift within the WCB, or later during the ascent at the leading edge of the upper level trough. We can not rule out that during this up-lift some mixing took place between the ascending Asian pollution and the lowermost stratospheric air located above, especially at the top of the lifted air mass. However, we suggest that the main part of the mixing happened very likely during the period of cross-polar transport at elevated levels. As mentioned in Sect. 3.5, the polar tropopause was vertically displaced during the passage of the tropospheric streamer.  Fig. 6g) show an up-lift of the thermal tropopause of 2.5 to 3 km. This in turn implies that the polluted air mass was above ∼9 to 10 km embedded by stratospheric air from the Arctic, while it was advected across the pole. As discussed in Sect. 3.5, the tropospheric streamer was elongated and stretched into long and narrow filaments during its journey. Therefore the most likely scenario is that horizontal shear induced 5 mass exchange, and as a result, turbulent mixing across the air mass boundaries took place. As suggested by the "MR" in-situ data, O 3 -rich and CO-poor stratospheric air masses were entrained at the boundaries of the polluted air mass. However, we have to keep in mind that our data only represent a "snap-shot" in the temporal evolution of 3-dimensional mixing events, documenting the mixing stage only at one location and 10 one point in time. The role of radiation (cooling/heating) is not discussed here, but may be relevant due to the existence of clouds, or at least humidity differences. Another point worth mentioning is that not only does the chemical composition change through entrainment of stratospheric air, but also thermo-dynamic properties. The polar-crossing WCB trajectories show for example a slight but continuous increase 15 in potential temperature and PV, as do forward trajectories initialized from the locations of our observation. This possibly reflects the entrainment of warmer stratospheric air, at least to some extent, and this gain in Θ and PV suggests in turn that the pollution might have been completely mixed into the lowermost stratosphere. 20 During the research flight on 10 July 2008 an air mass containing unusually high CO, PAN, and H 2 O mixing ratios of 138 nmol mol −1 , 330 pmol mol −1 and 81 µmol mol −1 was sampled in the Arctic tropopause region at 11.3 km, ∼800 m above the dynamical tropopause (2 PVU). In-situ tracer correlations and FLEXPART backward simulations show that we probed the topmost part of an air mass with recent tropospheric origin, 25 containing anthropogenic pollution from East Asia mixed with O 3 -rich stratospheric air. As supported by a detailed trajectory analysis, the Asian pollution was up-lifted within a WCB connected to a low pressure system over Northern Russia. The analysis also showed that 92 % of the pole-crossing WCB trajectories reached the dynamical tropopause, mainly within the WCB ascent or shortly later. A part of the Asian pollution was embedded in a tropospheric streamer, which was advected across the pole. During its journey, the top of the polluted air mass was surrounded by Arctic stratospheric 5 air. Mixing most likely took place along the sides of the plume, when the tropospheric streamer was stretched into long and narrow filaments. Correlation analyses show that the polluted tropospheric air mass was enriched mainly in CO, CO 2 and H 2 O before it was mixed into the stratosphere, whereas PAN and NO y have been efficiently removed before and during the up-lift within the WCB. While the wet removal of HNO 3 is a regular occurrence in the transport pathway discussed herein, PAN generally has the potential to reach the tropopause region within WCB ascents, if the up-lift is not preceded by multi-day-transport in the warm boundary layer. The transport of pollution across the Pole was associated with an Arctic low-pressure system. Such cyclone events in the proximity of the North Pole are regular phenomena of the polar circulation, especially 15 during summer (Serreze and Barret, 2007;Orsolini and Sorteberg, 2009), though the described cross-polar transport period was unusually strong .

Summary and conclusions
Our study presents the first in-situ measurement of WCB-lifted Asian pollution being stirred and finally irreversibly mixed into the polar lowermost stratosphere. In the context of our observations the question arises: is this transport process a common sce-20 nario that brings pollution to the polar lowermost stratosphere, as already suggested by climatological studies (Stohl, 2001;Wernli and Bourqui, 2002)? Our measurements are possibly linked to other observations: Dessler (2009) showed that even at high latitudes of 75-80 • N, cloud tops are found ∼20 % of the time above the summertime tropopause. Since convection is not very frequent in these regions, WCBs may pos-25 sibly be one reason for these observations. Furthermore, the lowermost stratosphere north of the jet stream is found to have high CO mixing ratios of 100 nmol mol −1 and more (Cooper et al., 2005;Tilmes et al., 2010), which points to regular entrainment of CO-rich polluted air masses. The influence on local stratospheric chemistry is not Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Bönisch, H., Engel, A., Curtius, J., Birner, Th., and Hoor, P.: Quantifying transport into the lowermost stratosphere using simultaneous in-situ measurements of SF 6 and CO 2 , Atmos. Chem. Phys., 9, 5905-5919, doi:10.5194/acp-9-5905-2009Phys., 9, 5905-5919, doi:10.5194/acp-9-5905- , 2009. Bradshaw, J., Davis, D., Grodzinsky, G., Smyth, S., Newell, R., Sandholm, S., and Liu, S.: Observed distributions of nitrogen oxides in the remote free troposphere from the NASA