Mixing and ageing in the polar lower stratosphere in winter 2015/2016

. We present data from winter 2015/2016, which were measured during the POLSTRACC (The Polar Stratosphere in a Changing Climate) aircraft campaign between December 2015 and March 2016. The focus of this work is on the role of transport and mixing between aged and potentially chemically processed air masses from the stratosphere with mid and low latitude air mass fractions with small transit times originating at the tropical lower stratosphere. By combining measurements of CO, N 2 O and SF 6 we estimate the evolution of the relative contributions of transport and mixing to the UTLS composition 5 over the course of the winter. We ﬁnd an increasing inﬂuence of aged stratospheric air partly from the vortex as indicated by decreasing N 2 O and SF 6 values over the course of winter. Surprisingly we also found a mean increase of CO by (3.00 ± 1.64) ppb V from January to March relative to N 2 O in the lower stratosphere. We show that this increase of CO is consistent with an increased mixing of tropospheric air as part of the fast transport mechanism in the lower stratosphere surf zone. The analysed air masses were partly 10 affected by air masses which originated at the tropical tropopause and were quasi-horizontally mixed into higher latitudes. This increase of the tropospheric air fraction partly compensates for ageing of the UTLS due to the diabatic descent of air masses from the vortex by horizontally mixed, tropospheric inﬂuenced air masses. This is consistent with simulated age spectra from the Chemical Lagrangian Model of the Stratosphere (CLaMS), which show a respective fractional increase of tropospheric air with short transit times lower than six months and a simultaneous increase of aged air from deep stratospheric and vortex 15 regions with transit times larger than two years. We thus that from deep and which led to a simultaneous increase of the fraction of in the Arctic lowermost stratosphere over the course of the overall ageing. These aged air masses are isentropically mixed with younger air masses out of the TTL region. The observations are in-line with the climatology of mixing from 2005-2015 on the basis of Era-interim 20 by the CLaMS model.

To estimate the radiative effects of these species with large gradients at the tropopause, the details of mixing are essential Shine, 1997, 2002;Riese et al., 2012). Uncertainties arising only from uncertainties in mixing may lead to significant uncertainties of the radiative forcing, which are on the order of 0.5 W m −2 .
In our study we focus on the transition of the tracer composition in the subvortex region up to Θ = 410K during winter 2015/2016. We will quantify the effects of quasi-isentropic mixing from the tropics and diabatic downwelling and its effect on 5 the chemical composition as well as the evolution of the age spectrum and the mean age in this region.

Meteorological conditions during winter 2015/16
The early Arctic winter 2015/16 was the coldest winter in the lower stratosphere (LS) since 1948. These extreme cold conditions could establish due to a strong and cold Arctic polar vortex which developed in November 2015 due to very low planetary wave activity in the stratosphere (Matthias et al., 2016). From late December 2015 to early February 2016 the temperatures at 10 Θ = 490 K decreased below 189 K. Therefore strong dehydration and denitrification was seen in low H 2 O and HNO 3 volume mixing ratios, which finally led to a strong chlorine activation in early winter. Using MLS data the chemical influence of the vortex could be observed to isentropes below Θ = 400 K (Manney and Lawrence, 2016).
The major final warming (MFW) occurred on 5 March 2016 which led to a split of the vortex one week later. This early final warming was unusual, as only five other MFWs since 1958 appeared before middle of March. Due to this early warming air 15 masses in the polar lower stratosphere were mixed with non-vortex air and prevented chemical ozone depletion reaching record low values during winter 2015/16 (Manney and Lawrence, 2016, and references therein).
The winter 2015/16 was characterised by an unprecedented anomaly of the quasi biannual oscillation (QBO) with a westward jet formed within the eastward phase in the lower stratosphere (Newman et al., 2016;Osprey et al., 2016). Since the QBO affects the zonal wind direction in the tropical lower stratosphere (Niwano et al., 2003) its strength and phase is crucial for 20 stratospheric transport processes and westerly phases are related to strong and cold polar vortices (Holton and Tan, 1980). Further the winter 2015/16 was also affected by a strong warm phase of the El-Niño Southern Oscillation (ENSO) (McPhaden et al., 2015). A direct influence on the polar vortex is still under debate, but according to Matthias et al. (2016)

this strong
El-Niño is suggested to account for a weakening of the polar vortex.
3 Project overview and measurements 25 This work will address the evolution of composition, age structure and the influence of transport and mixing of air masses in the lower stratosphere. The composition of air masses inside the LS, which is affected by diabatic descent of upper stratospheric air masses, irreversibly mixed with younger air from the TTL is analysed by combining measurements of in-situ data with model calculations of the Chemical Lagrangian Model of the Stratosphere (CLaMS) (McKenna et al., 2002;Grooß et al., 2014;Ploeger et al., 2015). 30

The POLSTRACC campaign 2015/16
The data presented in this study were obtained during the POLSTRACC (Polar Stratosphere in a Changing Climate) mission, which was part of the combined PGS (POLSTRACC/GW-LCYCLE/SALSA) framework. The main objectives of the POL-STRACC mission were the investigation of structure, composition and dynamics of the Arctic LMS and processes involving chemical ozone depletion and polar stratospheric clouds in the Arctic winter UTLS. In total 17 scientific flights were performed from Oberpfaffenhofen, Germany (48.05 • N, 11.16 • E) and Kiruna,Sweden (67.49 • N,20.19 • E) covering the region from 25 • N to 87 • N and 24 • E to 80 • W (Fig. 2). Typical flight altitudes ranged from 10 km asl 1 to 14.5 km asl corresponding to potential temperatures in the stratosphere from Θ = 320 K up to Θ = 410 K. The total flight time was about 157 hours, of which 19 hours were spent in December 2015, 62 hours were spent from January to February and 76 from February to 10 March, respectively. For this study we focus on Arctic measurements starting from Kiruna, which took place during two campaign phases, representing flights from 12. January 2016 to 02. February 2016 (phase 1) and from 26. February 2016 to the 18. March 2016 (phase 2). For the aim of this work we use approximately 50 hours of measurements of those flights which were conducted to probe air masses predominantly underneath the polar vortex above PV = 7 PVU.
The research aircraft HALO is a modified business jet type Gulfstream G-550. It has a maximum range of 12500 km with a 15 maximum altitude of 15.5 km and can carry about up to 3 tons of scientific payload. The payload was a combination of different remote sensing (e.g. WALES lidar (Wirth et al., 2009;Fix et al., 2016), Väisälä RD 49 dropsondes and GLORIA limb sounder (Friedl-Vallon et al., 2014;Kaufmann et al., 2015)) and in-situ instruments of trace gases with different lifetimes, sources and sinks.

In-situ trace gas measurements
In this study we analyse measurements of N 2 O, CO, which were measured with the TRIHOP instrument  and SF 6 by the GhOST-MS instrument (Sala et al., 2014). For our analysis the data are synchronised to a common time resolution of 10 seconds or 0.1 Hz respectively, corresponding to a horizontal resolution of 2.5 km at typical HALO flight speeds.
GhOST data is available with a resolution of 60 seconds at an integration time of one second which leads to a horizontal 25 resolution of 15 km.

The TRIHOP instrument
The TRIHOP instrument (Schiller et al., 2008) is an infra red absorption laser spectrometer with three quantum cascade lasers (QCL) operating between wavenumbers 1269 cm −1 and 2184 cm −1 . It was set up to measure CO, N 2 O and CH 4 during the 30 POLSTRACC campaign. To quantify mixing ratios in the order of ppb V the instrument uses a multi pass White-cell which 1 above sea level (2014) and Müller et al. (2016). The thermal tropopause is denoted by the thick black solid line. The measurement region is depicted as blue box subdivided into the extratropical tropopause layer (ExTL), the lowermost stratosphere (LMS) and the lower stratosphere (LS). LMS and LS are separated by the 380 K isentrope (green dashed line). Transport pathways of air masses are denoted by coloured, thick arrows from the tropical tropopause layer (TTL) (red) and the polar Arctic upper stratosphere (blue) with respective mean age Γ and trace gas volume mixing ratio χ. Quasi horizontal mixing is represented by wavy double side arrows, indicating no net mass transport of air masses. Dotted lines are isentropes in K, the solid dark blue line indicates the 7 PVU contour, which is used to separate the regime of the ExTL from the LMS and LS (for details see text). Thin orange contour lines depict the zonal view of the jet stream. is pressure regulated at a pressure of 30 mbar to minimize pressure broadening of the absorption lines. The measurements were performed with an integration time of 1.5 seconds per species. The three species are subsequently measured during a full cycle which finally leads to a time resolution of 7 seconds due to additional latency times when the channels are switched. The instrument is regularly calibrated in-flight against compressed standards of ambient air which are in turn calibrated prior and after the campaign against primary standards, connected to the World Meteorological Organisation Global Atmosphere Watch 5 Central Calibration Laboratory (WMO GAW CCL) scale (X2007) for greenhouse gases. During POLSTRACC it was possible to achieve a (2σ) precision of CO, N 2 O and CH 4 of 1.15, 1.84 and 9.46 ppb V respectively.

GhOST-MS in-situ measurements
The GHOST-MS instrument is a two channel gas chromatograph for airborne measurements of trace gases. One channel uses a mass spectrometer (Agilent MSD 5975) for the detection of atmospheric trace gases at a time resolution of four minutes.

10
This channel uses negative ion chemical ionization as described in Sala et al. (2014) to measure brominated hydrocarbons. The other channel measures SF 6 and CFC-12 using an ECD (Electron capture detector) with a time resolution of one minute. For the POLSTRACC campaign the precision for SF 6 was 0.6% and the precision for CFC-12 was 0.2%.
Mean age of air is inferred from SF 6 measurements (Engel et al., 2009). Due to its much higher atmospheric mixing ratio, the precision of CFC-12 measurements is better than that of SF 6 measurements. Prior to calculating mean age, the SF 6 time 15 series has therefore been smoothed using the CFC-12, by applying a local (ten minutes of data before and after the time of measurement) fit between CFC-12 and SF 6 . This procedure removes parts of the instrumental scatter but retains the local information and does not introduce any offset to the mean age values. Mean age derived in this way has an overall precision   (Ploeger et al., 2015).
Recently, a method to calculate the age of air spectrum has been implemented in CLaMS (Ploeger and Birner, 2016), which will be used in the following analysis. The age spectrum is the transit time distribution of air masses for transport from a control surface (usually take as the tropical tropopause or the Earth's surface) to a given location in the stratosphere (e.g. Hall and Plumb, 1994;Waugh, 2002) and can be related to the Greens function of the transport equation. The calculation method in CLaMS is based on inert tracer pulses, with different tracers released every other month at the surface in the tropics. This method allows calculating time dependent age spectra for the non-stationary atmospheric flow at any location and time in the model domain (see Ploeger and Birner (2016) for further details). 5 For the aim of this work also a simulation with full stratospheric chemistry was performed by CLaMS with the setup as described by Grooß et al. (2014). This setup is typically used for periods up to six months. The upper boundary is set to Θ =900 K potential temperature, where tracers like O 3 , N 2 O and CO are constrained by MLS satellite observations. Due to its Lagrangian formation, a box-trajectory model setup is also possible in which the identical chemistry scheme is used along single air mass trajectories. This setup is also used here to diagnose chemical pathways and chemical conversion rates. This boxmodel setup 10 is also used here to estimate CO production and loss rates.

Results
As shown in Hoor et al. (2010) rapid and frequent mixing with tropospheric air mainly affects the region of PV < 7 PVU.
To exclude mixing with air masses of recent tropospheric origin or from the exTL (extratropical tropopause layer) we only selected data above this level of potential vorticity. Therefore the composition of analysed data is mainly affected by isentropi- 15 cally, irreversibly mixed air mass signatures originating out of the tropics and diabatically descended air masses from the upper stratosphere in the polar region. In this analysis we further excluded flights, which were dedicated to the observation of gravity waves.  Hoor et al., 2004;Hegglin et al., 2006). Equivalent latitude is directly linked to the potential vorticity, which is conserved under adiabatic processes (Holton, 2004). Therefore, these coordinates are suitable to account for reversible adiabatic tracer transport. 25 An air parcel in the stratosphere is a mixture of fractions of air with different histories, transport pathways and individual transit times. The several transport pathways constitute to an age spectrum or transit time distribution, respectively. The age spectrum can be obtained by calculation of the Green's function of the tracer continuity equation for a conserved and passive species (Hall and Plumb, 1994).

Age of air
The mean age is defined as the first moment of the transit time distribution. To determine the mean age from measurements,  (Hall and Waugh, 1997). Since SF 6 is a long-lived inert trace gas with a well known increase of its mean surface mixing ratio, it is a commonly used species for calculations of mean age (Boenisch et al., 2009). The sink of SF 6 is in the mesosphere, where it is destroyed by shortwave UV radiation. The lifetime of SF 6 is assumed to be 3200 years, but recent studies indicate a significantly shorter lifetime of about 850 years (Ray et al., 2017). This implies that mean age derived from SF 6 may be too old. Especially for polar vortex air, this has been modelled and observed to cause a high bias of up to one 5 year (Ray et al., 2017) or even more in mesospheric air (Engel et al., 2006b). While this may cause a significant offset in mean age for polar vortex air, it is estimated that relative changes in mean age as discussed in this paper can be reliably derived from SF 6 observations. Figure 3 a) and b) show the distribution of mean age calculated from SF 6 measurements for phase 1 (January) and phase 2 (February / March), respectively. As is evident on panel (a) during phase 1 the LS is dominated by air masses of mean ages 10 between 0.5 years to less than three years at maximum. The oldest air masses with mean ages older than two years were encountered at largest distances from the tropopause and potential temperatures ranging from Θ = 360 K to Θ = 380 K. In contrast, during phase 2 (panel (b)) in general much older air masses up to five years were found at potential temperatures of Θ = 410 K. These higher potential temperatures at flight altitude are the result of the diabatic descent over the course of winter and indicate an increasing influence of air masses originating deeper in the stratosphere or from the Arctic polar vortex. To 15 directly compare the temporal evolution of the age of air in the lower stratosphere panel (c) shows the difference of age of air between both phases. It can be seen that the bulk of air inside the LS is getting older between Θ = 330 K and Θ = 380 K. The mean increase is 0.29 years, indicating diabatic downwelling due to the evolution of the polar vortex and thus an increased mean age in late winter. 20 Nitrous oxide (N 2 O) is a very stable molecule with a lifetime of 123 years (Ko et al., 2013). Its sources are at the surface due to natural and anthropogenic emissions with a very small seasonal variability (Dils et al., 2006). As a result of the well-mixed troposphere and the absence of tropospheric sinks (Ciais et al., 2013) N 2 O has a distinct background value in the troposphere, so mixing ratios below this value can be identified as stratospheric influenced . The tropospheric background value of N 2 O between November 2015 and March 2016 in the northern hemisphere was 329.3 ppb V , measured by the 25 NOAA Global Monitoring Division (NOAA). Its annual increase was found to be 0.78 ppb V in the last years (Hartmann et al., 2013) with an variability of 3-5 ppb V (Kort et al., 2011). The main sink reactions of N 2 O are due to photolysis in the UV-band (190 nm ≤ λ ≤ 220 nm) and the reaction with O( 1 D) which only occurs within the upper stratosphere (Ko et al., 2013). Since there are no sources of N 2 O in the stratosphere its profile above the tropopause changes and shows a weak negative vertical gradient. During winter and spring a stronger negative vertical gradient establishes because of the enhanced diabatic down-   Panel (a) shows data for phase 1, (b) for phase 2 and panel (c) shows their absolute difference (phase 2-phase 1). The colour code represents the mean age. Blue colours in panel (c) indicate an increase of mean age in the subvortex region from January to March. Only bins with more than ten data points are shown.

Nitrous oxide
and it is evident that there is a general decrease of N 2 O observed in the whole LS, consistent with the measurements of mean age, indicating an enhancement of the diabatic downwelling over the course of winter.

Carbon monoxide
Carbon monoxide (CO) is released to the atmosphere mainly through incomplete combustion processes and methane oxidation 5 as the only significant in-situ source. It has therefore a large variability in the troposphere which is also affected by anthropogenic emissions. Due to the high variability of surface emissions CO has variable mixing ratios in the range of 70 ppb V to 200 ppb V (Prinn et al., 2000) in the northern hemispheric troposphere with lifetimes on the order of weeks. In the lower stratosphere the main source of CO is the methane oxidation with the OH radical. The main sink reaction is the oxidation with the OH radical where CO gets oxidized into CO 2 , which leads to a longer lifetime in the order of several months under dark 10 vortex conditions: We found an equilibrium value of 10-15 ppb V in winter 2015/2016, depending on the integrated temperature history of the respective air mass in agreement with previous studies (Müller et al., 2016;Herman et al., 1999).
A potential additional source of CO in the Arctic winter stratosphere is the reaction of CH 4 with reactive chlorine (Cl) which is nearly insignificant for the lower stratosphere (Flocke et al., 1999). Transport from the mesosphere, where CO is produced from the photolysis of CO 2 , also provides a potential source of CO via strong diabatic descent during winter under persistent polar vortices (Engel et al., 2006a). These potential influences are discussed in chapter 6. Figure 5 shows the distribution of CO. During phase 2 (panel (b)) the lowest mixing ratios of 15 ppb V were found at potential temperatures between Θ = 380 K and Θ = 410 K and equivalent latitudes > 60 • N. As can be seen by the vertical branch 5 of the CO-N 2 O correlation (Fig. 7), this value is the stratospheric equilibrium during late winter. Phase 1 (panel (a)) values ranged between 60 ppb V and 17 ppb V , hence the stratospheric background value was not measured in January 2016. A strong tropospheric influence is evident below Θ = 340 K with CO values up to 57 ppb V at phase 1 and 47 ppb V at phase 2. Hence the overall distribution of carbon monoxide in the UTLS seems to be consistent to N 2 O and mean age obtained from SF 6 measurements, despite its much shorter lifetime compared to the other species.

10
However, when comparing the differences of the respective phases (panel c), we see a different behaviour compared to N 2 O and SF 6 . We encountered an increase of carbon monoxide mixing ratios over the course of winter, which is at first glance in contradiction to the distributions of mean age and N 2 O. While the distributions of long-lived tracers SF 6 and N 2 O indicate an ageing of air masses, the increase of short-lived CO indicates a source of CO either from the troposphere or the stratosphere.
Note that the increase is observed above Θ = 360 K, whereas below this level a decrease occurs. We will analyse the potential 15 sources of CO in the following and rise the hypothesis that CO increases due to an enhancement of mixing of tropospheric air from the tropical lower stratosphere over the course of winter without a direct strengthening of tropospheric source emissions and mixing ratios from below.

Analysis
We found a decrease of the long lived species SF 6 and N 2 O with their lowest values in the furthest regions from the troposphere in late winter, which fits well in the general picture of the Brewer-Dobson circulation and the enhanced downwelling in late winter/spring. The contradicting, simultaneous increase of the short lived CO over the course of winter could indicate a strengthening of tropospheric transport by enhanced mixing with fraction of air with low transit times into the lower 5 stratosphere.

Identification of mixing on the basis of tracer-tracer correlations
In the following we will discuss this hypothesis and also other potential sources for the additional CO mixing ratios. We will show that, despite of different potential source regions as the mesosphere or chemical in-situ production, this increase is originating from an enhanced isentropic mixing out of the TTL, interacting with the diabatic descent in the polar stratosphere.

10
To identify mixing processes across the tropopause CO-O 3 correlations have been widely used (Fischer et al., 2000;Zahn et al., 2000;Hoor et al., 2002;Pan et al., 2004;Müller et al., 2016). Since ozone is affected by chemical processes particularly in the vortex region we use N 2 O as stratospheric tracer instead of ozone. Carbon monoxide, here used as tropospheric tracer also has sources in the mesosphere and via chlorine chemistry in the stratosphere. In the LS the influence of chlorine is small, compared to the reaction with the hydroxyl radical, therefore we investigated the influence of chlorine chemistry regarding 15 methane which leads to the formation of CO. This influence will be discussed in detail later.
To analyse the effect of transport and mixing on the evolution of the UTLS composition we used the N 2 O-CO relation as shown  (Fischer et al., 2000). In the presence of actual 5 rapid mixing a straight mixing line between two end members of the correlation establishes (Hoor et al., 2002;Müller et al., 2016). As stratospheric CO will relax towards its stratospheric equilibrium value while N 2 O is chemical inert, the initial linear correlation will become curved with time in case of inefficient mixing when the chemical lifetime is shorter than the time scale of mixing. Depending on the strength of mixing relative to the chemical CO sink the curvature will change and is less pronounced as the mixing gets more efficient. It is important to note that the change of CO relative to a given N 2 O value 10 can only be explained by a change of the ratio between mixing and chemical time scales. Mixing alone acts on both tracers N 2 O and CO. Therefore a change of the shape of the curve is a direct result of the increased mixing relative to the chemical timescale, which is less efficient when mixing becomes stronger. Panel (d) of Fig. 6 shows additionally the correlation under mesospheric influenced conditions. In this case the correlation would rise to higher CO mixing ratios and lower N 2 O mixing ratios, since N 2 O gets destroyed and CO produced in the mesosphere. This is a remarkable result since we expect that due to the ageing of air inside the lower stratosphere in winter, the CO mixing ratio decreases with time. It is important to note that the correlation along the mixing line which connects tropospheric values with the stratosphere shows higher CO relative to N 2 O. Furthermore phase 1 shows higher CO values relative to N 2 O com-25 pared to phase 2 for N 2 O values larger than 313 ppb V . Therefore we can conclude that regarding to the CO-N 2 O correlation the direct tropospheric impact was greater in phase 1 than in phase 2, indicating enhanced mixing with tropospheric influenced air originating in the TTL region during phase 2.
A potential mesospheric impact is highly unlikely due to the fact that during phase 2 the N 2 O-CO correlation tends towards the equilibrium value in the region of lower N 2 O values. This influence will be discussed later in detail. 30 As shown before the analysed measurement region, which is covered in both phases, lies between potential temperatures of Θ = 340 K and Θ = 380 K. Additionally, the measurement data is filtered for potential vorticity values larger than PV = 7 PVU.    (c)). Further on, rapid eddy mixing of air from the TTL leads to an increase of tropospheric tracer signatures in the Arctic region (Rosenlof et al., 1997).
To quantify the increasing influence from tropospheric air masses in the lower stratosphere, we applied a simple mass balance approach to quantify the composition of the lower stratosphere. Therefore, we assume an air parcel in the lower stratosphere may consist of either upper stratospheric or tropospheric origin (Fig. 1). This mass balance system is solved to get the amount 5 of tropospheric fraction f trop of the measured air.
For a mixing ratio χ on a specific isentrope θ we assume and f trop + f strat = 1 ( 3) which leads to the tropospheric fraction f trop based on CO measurements with χ CO,m the measured CO mixing ratio, χ CO,strat the stratospheric CO background which was set to 15.7 ppb V as mean of the vertical branch of the CO-N 2 O correlation and χ CO,trop the tropospheric CO entry value in the TTL.
Earlier studies have shown that CO mixing ratios above the tropical tropopause are at levels between 50 and 60 ppb V (Herman 15 et al., 1999;Marcy et al., 2007).
The difference of the calculated tropospheric fraction f trop between phase 2 and phase 1 is shown in Fig. 8 as a function of N 2 O, which acts as a quasi vertical coordinate. The CO increase over the course of winter corresponds to an increase by f trop of 6.8(3.7)% between 313 ppb V and 273 ppb V N 2 O. Note that additionally the tropospheric fraction decreases towards more tropospheric N 2 O values from phase 1 to phase 2. This is a clear evidence that an increase of the CO mixing ratio at the 20 tropopause is not the cause for the observed lower stratospheric CO increase. This would be consistent with an increase of the fraction of young air of tropospheric origin and more efficient mixing as indicated in Fig. 6.
Panel (b) shows the distribution against equivalent latitude. Note that the observed increase is most prominent above Θ = 360 K.
This is a clear indication that mixing at Θ < 360 K is suppressed due to the strong subtropical jet, which acts as a barrier for mixing (Haynes and Shuckburgh, 2000) and would be consistent with enhanced mixing out of the TTL region.  track and therefore can directly compare our measurements with the spectrum.
To test if the model is able to reproduce the observations of tracers we compared CO and N 2 O from CLaMS with the measurements (Fig. 9). Model output is available along the flight track with a time resolution of ten seconds. Figure  This remarkable agreement between model and observations further motivates the usage of CLaMS for age analysis of our measurements.
As mentioned before, CLaMS is able to calculate the full transit time distribution of analysed air masses along the flight track. Figure 10 shows the averaged age spectra of the CLaMS model for the respective phase (panel (a)) and the difference of them (panel (b)). Vertical dotted lines represent the mean age of the respective phase (blue and red) calculated by the CLaMS model, the dashed vertical lines separates young air masses with a mean age lower than 0.5 years and old air masses with mean age larger than 2 years. Since we have the full transit time distribution of each data point, we can compare this relation between the different parts of the age spectrum. An increase of the tropospheric fraction would be linked to an increase of the part of 15 the age spectrum with low transit times as indicated by the observed increase of CO relative to N 2 O.
From Fig. 10 panel (b) it is evident that there is an increase of air masses older than two years up to 0.3%. In the range of air masses younger than six months, there is also an increase of the age spectrum between phase 2 and phase 1 evident which is, with maximum values up to 0.9%, larger than the increase of the old air masses. The increase of the young fraction is in agreement with the observed increase of CO. It is therefore an indication for an increased mixing with air from the TTL at the end of winter.
Since the mean age is calculated as first moment of the distribution its value is most sensitive to changes by the old tail of the distribution (Hall and Waugh, 1997). Therefore the mean age rises by 0.27 years from 1.71 years to 1.98 years as a result of 5 the increase of the age spectrum distribution for air masses older than two years. This matches the mean age increase of SF 6 and indicates, combined with the decrease of N 2 O, the overall ageing in the lower stratosphere over the course of winter. Since the integral over the Greens function is normalised to one, the increases of air masses older than two years and younger than six months must result in a relative decrease in between. Therefore air masses with mean ages between 0.5 years and 2 years are more enhanced in phase 1 than in phase 2, which is evident by the change of the transit time distribution up to −1.9%.

10
To further investigate the relationship of young versus aged air we used the spectral information for each individual data point in the following way. We calculated the accumulated fraction of air masses with transit times lower than six months and older than two years, respectively, for each data point. Figure 11 shows the binned fraction of air masses with transit times lower than six months versus the modelled mean age.
The comparison of the correlation for different times (phase 1 and phase 2) shows that for a given mean age a significant 15 increase of the young tropospheric contribution is evident. Thus, according to the model and in agreement with the observed increase of CO, the late winter LS is more affected by tropospheric young air.
During winter 2015/16 CO mixing ratios in the LS increased from January to March while long-lived trace gases denote an , : Figure 11. Mean age versus air fractions with transit times < 6 months from the age spectra simulated by CLaMS for phase 1 (blue) and phase 2 (red). Each data point is binned in steps of 5 ppbV N2O. The variability in each bin is given by the vertical and horizontal lines,

Discussion
Since there are different sources for CO at different locations in the atmosphere an increase of carbon monoxide mixing ratios 5 can be due to (i) an increase of isentropic mixing out of the TTL, (ii) an increase of the tropospheric source strength, (iii) a potential influence of the mesosphere and (iv) a change of chemical reaction cycles due to higher amounts of reactive chlorine in the stratosphere. As already discussed the increase of enhanced tropospheric source emissions (ii) is highly unlikely (see Fig. 7). Since our analysis points to an increase of isentropic mixing out of the TTL (i), the possible influence of points (iii) and (iv) have to be further discussed.

10
Carbon monoxide is produced in the mesosphere due to the photo-dissociation of carbon dioxide through high energetic sunlight. Therefore the composition of mesospheric air masses is clearly distinct from air mass composition of the stratosphere. This is reflected in the MLS observations that determine the CLaMS upper boundary at Θ = 900 K potential temperature. The simulation indicates the expected downward transport but sees no mesospheric influenced air masses in the LS.
It is evident that CO mixing ratios increase at Θ = 900 K from December 2015 to end of February 2016 which descended 10 from Θ = 900 K down to Θ = 600 K at the end of March 2016. Therefore these enhanced CO mixing ratios do not affect our measurement region below about Θ = 410 K. For potential temperatures below Θ = 420 K a slight increase in the CO mixing ratio with time is simulated, which does not originate in the stratosphere or mesosphere in agreement with our observations (see Fig. 7). Regarding to the composition of CO (Fig. 5) it is evident that the measured CO mixing ratios decrease with altitude and the lowest values are found at the highest regions and equivalent latitudes. This indicates that the increase of measured 15 CO mixing ratios has no mesospheric origin, because the enhanced CO mixing ratios are only transported down to Θ = 600 K potential temperature in the CLaMS model.
Furthermore, an additional influence of descended mesospheric air into the lower stratosphere would not only impact CO but also N 2 O. Due to the chemical differences between the stratospheric and the mesospheric composition, mixing of mesospheric air, enriched in CO and depleted in N 2 O, would lead to mixing lines very strongly differing from the observed relationship (see 20 Fig. 6). Importantly, the CLaMS N 2 O-CO correlation ( Fig. 9) almost perfectly mirrors the observations, further indicating no mesospheric influence on the simulated correlation. Additionally the age spectrum calculations of the CLaMS model provide mass fractions of air masses regarding their stratospheric residence time. As is evident from Fig. 11 there is a significant increase of air masses younger than six months at typical mean ages for lower stratospheric air masses and mesospheric influence on the basis of our analysis is highly unlikely.

25
In general, another important source of carbon monoxide in the atmosphere is the reaction of methane with reactive chlorine, which is not significant in the lower stratosphere (Flocke et al., 1999). Eventhough the influence of the methane reaction with Cl on CO is low, air masses enriched in reactive chlorine can be transported downwards, providing potentially reactants for the chemical production of CO. It may not be the case in this specific winter because of unprecedented low temperatures and resulting higher chlorine activation. Therefore this aspect was also investigated in more detail. For this aim we simulated  Figure 13 shows the statistical evaluation of the net CO change due to chemistry over the period as function of potential temperature on 15 March. The blue line represents the statistical mean and the dashed lines the 1-σ standard deviation. As is evident the mean overall change is even negative over the entire profile, which is due to the oxidation of the produced CO by the reaction with OH. Therefore we conclude that the observed increase of CO in phase 2 is not due to the additional chemical source reaction.

5
To investigate if transport and increased mixing of air mass fractions with transit times smaller than six months in winter 2015/16 was special compared to other years we analysed the climatology of these fractions from 2004 to 2016 and compared it to the calculated fractions in winter 2016, both from the CLaMS model (Fig. 14). The colour code represents the fractions of air masses with transit times smaller than six months, the contour lines represent the mean age and the thick black line indicates the WMO tropopause. Note, that mixing of these air masses is significantly stronger depicted in the southern hemispheric polar region. This must not directly be compared to the northern hemisphere, since this time span represents summer months in the southern hemisphere, where mixing is larger compared to the winter.
From the climatology it is evident that the largest fraction of air masses with transit times smaller than six months exceeding 73% is found between 30 • S and 30 • N up to Θ = 430 K. In January this strong signal has a sharp gradient at Θ = 450 K. These air fractions are transported from January to March to the poles so that between 70 • N and 90 • N the fraction of air masses 5 with transit times smaller than six months increases by 5 % at Θ = 380 K. From January 2016 to March 2016 this transport is even stronger than in the climatology, as is evident from the convex structure of the distribution gradient to the north pole.
From Fig. 14 it is also evident, that the mean age in March compared to January at Θ = 400 K shows a simultaneous higher value in both, the climatology as well as the winter 2015/2016, whereas the structure of the mean age contours show a more horizontal meridional gradient in the winter 2016 compared to the climatology.  times larger than two years and fractions of air with transit times smaller than six months. Since the mean age itself is most sensitive to changes on the old tail of the age spectrum, the ageing of air masses in the LS over the course of winter can be explained by the increase of old air masses, characterised by low N 2 O and SF 6 measurements. Increased mixing of young air masses adds to this and leads to an increased fraction of the younger part of the age spectrum, consistent with the observed increase of CO. It is evident that this enhancement is due to stronger mixing processes out of the TTL region, where fresh 15 tropospheric air is mixed into the polar lower stratosphere. Other potential sources of CO like mesospheric air and chemical reaction of CH 4 with chlorine are unlikely to have caused the observerd increase of CO.
Therefore we conclude that the Arctic lower stratosphere in March was strongly affected by mixing with young tropospheric air, which partly compensates for the overall ageing. These aged air masses are isentropically mixed with younger air masses