Balloon-borne measurements of temperature, water vapor, ozone and aerosol backscatter at the southern slopes of the Himalayas during StratoClim 2016-2017

. The Asian summer monsoon anticyclone (ASMA) is a major meteorological system of the upper troposphere-lower stratosphere (UTLS) during boreal summer. It is known to be enriched in tropospheric trace gases and aerosols, due to rapid lifting from the boundary layer by deep convection and subsequent horizontal confinement. Given its dynamical structure, the ASMA offers a very efficient pathway for the transport of pollutants to the global stratosphere. Detailed understanding of the ASMA structure and processes requires accurate in-situ measurements. Here we present balloon-borne measurements of tem- perature, water vapor, ozone and aerosol backscatter conducted within the StratoClim project from two stations at the southern slopes of the Himalayas. In total we performed 63 balloon soundings during two main monsoon-season campaigns, in August 2016 in Nainital, India (NT16 AUG ) and July-August 2017 in Dhulikhel, Nepal (DK17), and one brief post-monsoon campaign in Nainital in November 2016 (NT16 NOV ). These measurements provide unprecedented insights into the ASMA thermal structure and its relations to the vertical distributions of water vapor, ozone and aerosols. To study the structure of the UTLS during the monsoon season, we adopt the thermal definition of tropical tropopause layer (TTL), and define the region of altitudes between the lapse rate minimum (LRM) and the cold-point tropopause (CPT) as the Asian Tropopause Transition Layer (ATTL). Further, based on air mass trajectories, we define the Top of Confinement (TOC) level of ASMA, which divides the lower stratosphere (LS) into a Confined LS (CLS), below the TOC and above the CPT, and a Free LS (FLS), above the TOC. Using these thermodynamically-significant boundaries, our analysis reveals that the composition of the UTLS is affected by deep convection up to altitudes 1.5-2 km above the CPT due to the horizontal confinement effect of ASMA. This is shown by enhanced water vapor mixing ratios in the Confined LS compared to background stratospheric values in the Free LS, observed in both NT16 AUG (+0.5 ppmv) and DK17 (+0.75 ppmv), and by enhanced aerosol backscatter of the Asian tropopause aerosol layer (ATAL) extending into the Confined LS, as observed in NT16 AUG . The CPT was 600 m higher in altitude and 5 K colder in DK17 compared to NT16 AUG and strong ozone depletion was found in the ATTL and CLS in DK17, suggesting stronger convective activity during DK17 compared to NT16 AUG . An isolated water vapor maximum in the Confined LS, about 1 km 5 above the CPT, was also found in DK17, which we argue is due to overshooting convection hydrating the CLS. These evidence show that the vertical distributions and variability of water vapor, ozone and aerosols in the Asian UTLS are controlled by the top height of the anticyclonic confinement in ASMA, rather than by CPT height as in the conventional understanding of TTL, and suggest that the ASMA contributes to moistening the global stratosphere and to increase its aerosol burden.

enhanced water vapor mixing ratios in the Confined LS compared to background stratospheric values in the Free LS, observed in both NT16AUG (+0.5 ppmv) and DK17 (+0.75 ppmv), and by enhanced aerosol backscatter of the Asian tropopause aerosol layer (ATAL) extending into the Confined LS, as observed in NT16AUG. The CPT was 600 m higher in altitude and 5 K colder in DK17 compared to NT16AUG and strong ozone depletion was found in the ATTL and CLS in DK17, suggesting stronger convective activity during DK17 compared to NT16AUG. An isolated water vapor maximum in the Confined LS, about 1 km 5 above the CPT, was also found in DK17, which we argue is due to overshooting convection hydrating the CLS. These evidence show that the vertical distributions and variability of water vapor, ozone and aerosols in the Asian UTLS are controlled by the top height of the anticyclonic confinement in ASMA, rather than by CPT height as in the conventional understanding of TTL, and suggest that the ASMA contributes to moistening the global stratosphere and to increase its aerosol burden.

Introduction 10
Large-scale deep convection associated with the Asian summer monsoon (ASM) during boreal summer induces a strong and persistent anticyclonic vortex in the upper troposphere -lower stratosphere (UTLS), known as ASM anticyclone (ASMA) (e.g. Park et al., 2007) or, previously, as Tibetan high (e.g. Krishnamurti and Bhalme, 1976). The ASMA is confined by the subtropical westerly jet stream to the north (40-45°N) and the equatorial easterly jet stream to the south (10-15°N), and spans roughly one third of the northern hemisphere's longitudes (20-140°E), between the eastern Mediterranean Sea and the western 15 Pacific Ocean. Its geographic center is above the Tibetan plateau and the altitude of maximum strength of the anticyclonic circulation is around the local tropopause (17-18 km), which is the highest worldwide during the ASM season (Dethof et al., nature of the transport. Nevertheless, these studies are consistent with satellite observations (Fu et al., 2006), regional weather forecasting (Heath and Fuelberg, 2014) and global atmospheric circulation models (Fadnavis et al., 2013;Pan et al., 2016) in indicating the region of the southern slopes of the Himalayas (i.e. approximately latitudes 25-35°N south of the Tibetan plateau) as a hot-spot for the transport of boundary layer pollutants to the ASMA. Considering the recent rapid increase of pollutant emissions from India (e.g. Krotkov et al., 2016) and south-east Asia, due to their fast economic growth, it is crucial for global 5 chemistry climate models to properly represent the ASMA dynamics, thermodynamic structure and processes.
Up to date, most of the observational evidence regarding the chemical composition of the Asian UTLS is derived from satellite measurements, providing information with good regional and temporal coverage, but with limited vertical resolution. High vertical resolution is important to understand the physical boundaries that control the vertical distribution of chemical species, and microphysical processes like the nucleation of cirrus clouds and aerosols. This requires accurate in-situ measurements in 10 a notoriously hardly accessible region such as the UTLS, and balloon-borne measurements are particularly suited for this aim.
In this article we present and discuss the data collected by the StratoClim balloon campaigns, carried out from two sites at the southern slopes of the Himalayas through the years 2016 and 2017. In total 63 balloon soundings were performed in Nainital (India) and Dhulikhel (Nepal) during three distinct periods of time, including two main monsoon-season campaigns (August 2016 in Nainital and July-August 2017 in Dhulikhel), and one brief post-monsoon campaign (November 2016 in Nainital). 15 State-of-the-art instruments were used to measure vertical profiles of temperature, water vapor, ozone and aerosol backscatter, from the surface to the middle stratosphere. Here we first provide an overview of all measurements, showing mean profiles and their standard deviation ranges of natural variability for the different campaign periods, and then focus on analyzing the thermodynamic structure of the UTLS during the ASM season and how it relates with the distributions and variability of water vapor, ozone and aerosols. One aim of this work is also to pave the way for ongoing more targeted modeling and intercompar -20 ison studies within StratoClim and other activities.
attempts of conducting field campaigns in the ASM region were often hindered by operational and logistic limitations, which also affected our DK17 campaign resulting in a reduced measurement schedule (most notably, the number of backscatter measurements was limited: see Table 1). Nevertheless, important scientific data were collected. The DK17 campaign took place simultaneously with the StratoClim aircraft campaign, based in Kathmandu Airport (Nepal), which performed 8 scientific flights with the high-altitude Geophysica-M55 research aircraft. 5 All soundings employed meteorological latex balloons (Totex, Japan) filled with hydrogen gas in order to ascend at a rate of about 5 m/s. Maximum burst altitude of these balloons is around 35 km, and more than 70% of our soundings reached at least 30 km (see Table S1 in supplementary material). A standard meteorological radiosonde was used to host additional instruments through its XDATA interface (Oelsner and Tietz, 2017), and to transmit the data of all instruments to the ground station at 1 Hz frequency. In particular, we used RS41-SGP (Vaisala, Finland) radiosondes (Vaisala, 2017), and the DigiCora MW41 10 sounding system (Vaisala, Finland) as ground station (Vaisala, 2014). Additional instruments employed were: Electrochemical Concentration Cell (ECC, manufacturer: En-Sci, USA) (Komhyr, 1969) for ozone (O3) mixing ratio, Cryogenic Frost-point Hygrometer (CFH, En-Sci, USA) (Vömel et al., 2007; for water vapor (H2O) mixing ratio, and Compact Optical Backscatter Aerosol Detector (COBALD, MyLab, Switzerland) for aerosol backscatter.
For the pressure (p) and temperature (T) measurements that we analyze in this work, uncertainties of RS41-SGP (hereafter: 15 RS41) given by the manufacturer are 0.6/1 hPa (at pressures lower/higher than 100 hPa) and 0.3/0.4 K (at altitudes lower/higher than 16 km), respectively. The performances of ECC sondes have been assessed by many studies (e.g. Smit et al., 2007), and uncertainties are estimated as 5-10% in terms of O3 partial pressure. CFH is a frost-point hygrometer based on the chilledmirror principle with uncertainty on H2O mixing ratio lower than 10% up to 28 km altitude (Vömel et al., 2007). ECC and CFH are regularly deployed in the ASM region since 2009(Bian et al., 2012. COBALD is a detector for aerosol backscatter 20 measurements at optical wavelengths of 455 nm (blue visible) and 940 nm (infrared) developed at ETH Zürich, downscaling the original backscatter sonde by Rosen and Kjome (1991) in weight and size. COBALD is able to detect cirrus clouds (e.g. Brabec et al., 2012) as well as aerosol layers, such as ATAL (Ve rnier et al., 2015). In addition, one RS92-SGP radiosonde (Vaisala, Finland) was added to almost all payloads for an intercomparison with the performances of RS41 (not discussed in this paper). Finally, we note that for logistical constraints, the first two soundings in NT16AUG employed iMet-1-RSB radio-25 sondes (InterMet, USA) (InterMet, 2006), offering the XDATA interface (Wendell and Jordan, 2016), instead of RS41, and SkySonde version 1.9 (Jordan and Hall, 2016) as data acquisition software. The number of deployments of each instrument during the different campaign periods are summarized in Table 1. A list of all the 63 soundings with date and time of launch, payload description, burst altitude and other notes is given in Table S1 in supplementary material.
In this study, we use the pressure measured by RS41 as the main vertical coordinate for all instruments. All variables are 30 binned in pressure intervals of 1 hPa for p > 300 hPa, and 0.5 hPa for p < 300 hPa, yielding an improved signal-to-noise ratio and a dataset with consistent vertical levels. This binning corresponds to a vertical resolution of approximately 25 m in the UTLS. A quality check is performed for all instruments based on interpretation of their house-keeping data, and data points showing anomalous behavior are rejected (note that the number of malfunctioning events per instruments are noted in brackets Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-222 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 11 April 2018 c Author(s) 2018. CC BY 4.0 License.
in Table 1). In this context, CFH deserves a special mention, since a behavior we named "CFH contamination" was observed in a significant number of cases. This consists in the drift towards too high frostpoint temperatures, hence too high water vapor mixing ratios, in the stratosphere (see Figure S1 in supplementary material) which we attribute to the deposition of supercooled water droplets onto the inner walls of the instrument's inlet tube while passing through mixed-phase clouds. This hypothesis is currently subject of a dedicated modeling study. To avoid such instrumental artifact, for the analysis in this paper we do not 5 accept H2O mixing ratio measurements higher than 10 ppmv in the stratosphere, which are unrealistic, as well as all measurements at pressures below 20 hPa (see Section 3.3). Ice saturation (Sice), i.e. relative humidity with respect to ice, is calculated using the frost-point temperature measured by CFH, air temperature by RS41, and the parameterization for saturation vapor pressure over ice by Murphy and Koop (2005). The COBALD data are expressed as backscatter ratio (BSR), i.e. the ratio of the total-to-molecular backscatter coefficient. This is calculated by dividing the total measured signal to its molecular contri-10 bution, which is computed from the atmospheric extinction according to Bucholtz (1995), and using air density derived from the measurements of temperature and pressure (see Cirisian et al., 2014). Uncertainty on COBALD BSR as inferred by this technique is estimated around 5% (Vernier et al., 2015). Finally, note that for conversion of pressure to geometric altitude (z), mean profiles of p vs. z measured by RS41 are shown in Figure S2 in supplementary material. Figure 1 shows mean profiles and standard deviations of temperature, H2O mixing ratio and O3 mixing ratio calculated from all measurements performed during the three campaign periods, namely NT16AUG (blue), NT16NOV (green) and DK17 (red).

General overview of the data 15
Note that for improved recognition, we keep this color coding also in subsequent figures. Panels a-c show the entire measured profiles, while Panels d-f show zooms into the UTLS region. In this section we briefly discuss their main features. Aerosol backscatter measurements will be discussed in Section 6. 20

Troposphere
In the troposphere, seasonal variability of water vapor mixing ratio is the most evident feature, with differences up to a factor of 40 in the lower troposphere between the dry (November) and the ASM season (July-August). Massive latent heat release by condensation during the wet period, in contrast to dry conditions in winter, is reflected in different lower tropospheric lapse rates for the two seasons, with about 5.5 K/km (close to the moist adiabat) in July-August, and 8 K/km (close to the dry adiabat) 25 in November. Lower tropospheric O3 in NT16AUG compared to NT16NOV is likely due to enhanced washout of ozone precursor gases during the wet season. Higher O3 in the lower troposphere in NT16AUG vs. DK17 might be due to photochemical smog transport from the New Delhi urban area and the highly populated Indo-Gangetic plane (e.g. Kumar et al., 2010). Additionally, in the upper troposphere, the same feature is likely also related to stronger convective activity occurring during DK17 compared to NT16AUG (discussed in Section 5.2).

Tropopause region
The structure of the tropopause region is very different during the three measurement periods. In contrast to the sharp coldpoint tropopause (CPT) of the ASM season, typical of the tropics, the tropopause structure in November has more similarity to that in the mid-latitudes, with an almost isothermal layer above the lapse-rate tropopause (LRT), defined according to the World Meteorological Organization (WMO, 1957). As a consequence, in NT16AUG and DK17 the LRT coincides with the 5 CPT, whereas in NT16NOV the average LRT was found at 108 hPa and the CPT at 69.5 hPa, i.e. about 2.5 km apart. Interestingly, comparing the two ASM seasons datasets also reveals significant differences. The average CPT is 10 hPa higher (88 vs. 97.5 hPa, corresponding to about 600 m in altitude) and 5 K colder (-81.7 vs . -76.8 °C) in DK17 compared to NT16AUG. Water vapor in the UTLS is minimum in NT16NOV, with mixing ratios around 2.5 ppmv above the LRT. During the ASM, UTLS H2O is higher, but different vertical distributions are observed. In NT16AUG, H2O mixing ratio decreases monotonically with 10 altitude, with mean value of 6.8 ppmv at the CPT. In DK17, H2O mixing ratio shows a minimum at the CPT (3.5 ppmv), and a local maximum in the LS (6 ppmv). Lower O3 in the tropopaue region during the ASM than in November is likely related to the monsoonal deep convection, transporting ozone-poor air from the boundary layer to the UT. Mean altitude, pressure, potential temperature and temperature of the LRT and CPT for the three campaign periods are summarized in Table 2.

Stratosphere 15
Lower stratospheric temperatures  differ by about 2-4 K between November and July-August, which is consistent with the climatological annual cycle of stratospheric temperature (Randel et al., 2003). Stratospheric H2O mixing ratios are in the range of 4-6 ppmv up to 20 hPa, with a slight increase with altitude due to oxidation of methane. Above approximately 20 hPa (27 km), all CFH measurements show an unrealistic increase in H2O mixing ratio, which is a measurement artifact. At such high altitudes and the corresponding dry conditions and low air densities, outgassing from the balloon skin and the payload 20 train can play a significant role in contaminating the humidity measurements (Kräuchi et al., 2016). Hence, water vapor data in this range will not be considered in this analysis. Differences in stratospheric ozone between NT16AUG and DK17 are likely due to interannual variability.

Meteorological analysis
For relating the measurements to the large-scale atmospheric flow, here we analyze meteorological data from the European 25 Center for Medium-range Weather Forecast (ECMWF) for the three campaign periods, focusing on the UTLS region.

Seasonal variability
In particular, Figure 2 (left column) shows latitude-pressure cross sections of potential vorticity (PV), potential temperature 5 (θ), zonal wind speed and ice water content, for NT (longitude 80° N, Panels a, c) and DK (longitude 85°E, Panel e) from ECMWF analysis data on selected days. During the ASM season (Panels a, e), the subtropical westerly jet stream limits the ASMA on its northern flank around 40-45°N, and marks the border between a "tropical regime" to the south, including the ASMA, and the mid-latitudes to the north. The dynamical tropopause (i.e. the PV = 2 PVU surface) is found at about 100 hPa inside the ASMA, while north of the jet it decreases steeply with altitude. After the end of the monsoon season, the subtropical 10 westerly jet migrates southward to 30-35°N and intensifies in strength. Hence, in November (Panel c), NT is located at the boundary between the tropical and mid-latitude regimes, i.e. below the jet and the associated tropopause break, resulting in the quasi-isothermal lapse rate between the LRT and CPT discussed in Section 3.2. Additionally, the large standard deviation of temperature in the tropopause region in NT16NOV is likely related to varying meridional position of the jet during the measurement period (see Figure 1, Panel d). 15  (Wernli and Davies, 1997), based on ERA-Interim re-analysis wind fields, and initialized at 100 hPa at the time of each sounding. During the ASM season (Panels b, f) the UTLS flow at the southern slopes of the Himalayas is easterly and follows the southern branch of the ASMA, transporting air which has been circulating around the anticyclone already for several days. In contrast, in November (Panel d), the UTLS winds are westerly and follow the subtropical jet stream, carrying 20 air from above the Saharan desert and air masses deflected from the equatorial easterlies. This is consistent with wind speed and direction measurements by RS41 shown in Figure S3 in supplementary material.

Interannual and regional variability
To assess whether the differences between the observations in NT16AUG and DK17 are caused mainly by geographic difference, and associated different mesoscale weather features, or by interannual variability between the ASM 2016 and 2017 seasons, 25 in Figure 3 we examine time series of UTLS temperature and H2O mixing ratio from ECMWF analysis data for both stations and both campaign periods.
In August 2016 (Panels a-d), the UTLS was relatively warm at both locations, with CPT temperatures rarely below -80°C and DK slightly colder than NT (0.4 K on average at 100 hPa), and H2O mixing ratio did never decrease below 4.5 ppmv at both sites. The same day-to-day variability features are visible at both locations with a time shift of about 6-12 h, which is consistent 30 with DK being systematically upstream of NT along the southern branch of the ASMA, and a wind speed of around 20 m/s in the UTLS. In July-August 2017 (Panels e-h), T and H2O values and features are similar to 2016 until 3 August. Then, a period characterized by extremely cold and dry tropopause starts in both NT and DK, peaking between 7-10 August with CPT temperatures colder than -83°C and H2O mixing ratios lower than 3 ppmv. The minima are slightly more pronounced in DK than in NT but are clearly correlated in time, suggesting that these features are related to a large-scale cooling and drying pattern occurring in the ASMA. Interestingly, we also note that a layer of high H2O rises to high altitudes (70-85 hPa) after 3 August 5 (Panel f), forming the local maximum above the CPT which we also find in our DK17 measurements (Figure 1, Panel e).
To quantify this resemblance, Figure 4 shows mean measured profiles of temperature (Panels a-b) and H2O mixing ratio (Panels d-e) in comparison with ECMWF operational data (analysis and forecast) for the NT16AUG and DK17 campaign periods.
Good agreement in the location and magnitude of the CPT and of the main water vapor features, despite a slight moist bias of the ECMWF data, suggests that the ECMWF model captures the main dynamical and microphysical features of the measure-10 ment periods. The H2O maximum above the CPT observed in DK17 and reproduced by the ECMWF model is remarkable and not in accordance with more typical climatological conditions observed during NT16AUG. As will be discussed in Section 5, several features in our measurements suggest that the anomalous characteristics of the UTLS observed in DK17 are due to stronger convective activity compared to NT16AUG.
In summary, here we argue that the differences between the NT16AUG and DK17 datasets are not due to local meteorological 15 effects, which appear to have negligible impact on the UTLS temperature and water vapor at the two measurement sites.
Rather, these differences are to be attributed to interannual variability, and in particular to anomalously strong convective activity occurring in the ASM after 3 August 2017 and persisting on a large scale.

Horizontal confinement in ASMA
In conclusion of the meteorological overview, we evaluate the vertical extent of the horizontal confinement effect of ASMA 20 during the NT16AUG and DK17 campaign periods by means of air mass trajectories. For this, we calculate 2-weeks backward trajectories with LAGRANTO using ERA-Interim re-analysis wind fields, initialized between 40 and 150 hPa at the time of each balloon sounding in NT16AUG and DK17 (i.e. same as shown in Figure 2, Panels b, f for 100 hPa) and 6 h before and 6 h after each sounding. For each pressure level, we then calculate the "confined fraction" of trajectories, defined as the fraction of trajectories which were already located inside the anticyclone 2 weeks before the measurements. The ASMA area is ap-25 proximated as the box of 10-50°N latitude, 20-140°E longitude (shown by grey dashed lines in Figure 2, Panels b, f). The resulting confined fractions for NT16AUG and DK17 are plotted as a function of pressure in Figure 5. In both campaign periods, the confined fraction is high (above 60%) up to 70-80 hPa, while above it sharply decreases to zero. Therefore, we define the top of confinement (TOC) level of ASMA as the level of confined fraction = 50%, corresponding to 73 hPa in NT16AUG and 63.5 hPa in DK17. This level separates altitudes which are affected by the anticyclonic confinement of ASMA (below the 30 TOC), from the confinement-free lower stratosphere above the ASMA (above TOC). Mean altitude and potential temperature of the TOC levels derived from the balloon measurements are given in Table 2. In Sections 5 and 6, we will show that the TOC plays an important role in controlling the vertical distributions and variability of water vapor, ozone and aerosol backscatter in the Asian UTLS.

UTLS structure during the ASM season
In this section we focus on analyzing the UTLS structure of the NT16AUG and DK17 measurements. The NT16NOV measurements will be discussed again in Section 6. 5

Thermodynamic structure
In the Brewer-Dobson circulation, tropical CPT temperature is thought to control the amount of water vapor entering the stratosphere (e.g. Mote et al., 1996). However, the thermodynamic transition between the troposphere and the stratosphere does not occur instantaneously, but over a layer of several kilometers in thickness, termed tropical tropopause layer (TTL), and the properties of the TTL rather than CPT temperature alone control water vapor transport through the tropopause (High-10 wood and Hoskins, 1998;Holton and Gettelman, 2001;Fueglistaler et al., 2009) lapse rate switches from convectively-dominated, in the troposphere (small dθ/dz, low stability), to radiatively-controlled in the stratosphere (high dθ/dz, high stability) (Gettelman and de F. Forster, 2002). In addition, the LRM coincides with the mean convective outflow level (Gettelman and de F. Forster, 2002;Vömel et al. 2002;Paulik and Birner, 2012). Using this definition of TTL, Pan et al. (2014) showed that water vapor has low variability above the CPT, and ozone below the LRM.
Based on the similarity between the Asian tropopause region during the ASM season and that of the tropics (see Sections 3.2, 20 4.1), here we adopt this definition of TTL to study the thermal structure of the UTLS in o ur NT16AUG and DK17 datasets.
However, being our measurement sites not tropical in a strict geographic sense (NT and DK are at latitudes 27-30°N), we refer to the TTL in this region and season as the Asian tropopause transition layer (ATTL, or Asian TTL). Figure 6 shows mean profiles and standard deviations of T, θ and dθ/dz as a function of pressure for the two ASM season datasets. As already noted in Section 3.2, the average CPT is higher in altitude by about 10 hPa and 5 K colder in DK17 than 25 in NT16AUG (Panel a). Because of this large temperature difference, potential temperature levels in DK17 are shifted to lower pressures than in NT16AUG (Panel b), and the resulting difference between the two CPTs in isentropic coordinates is small (1.5 K), with both located slightly above the 380 K θ-level. The average LRM is also found at higher altitude by roughly 10 hPa in DK17 compared to NT16AUG (Panel c), corresponding to a 2.5 K difference near the 360 K isentropic level. Higher LRM and higher and colder CPT in DK17 suggest that, on average, convection reached higher altitudes in the ASM 2017 compared to 30 the year 2016, at least prior and during the time of our measurements. The exact ATTL boundaries in terms of pressure and average potential temperature are 180-97.5 hPa and 360-382 K for NT16AUG, and 169.5-88 hPa, 362.5-383.5 K for DK17 (see Table 2: note that the potential temperature values are the average θ in the pressure levels where the LRM and CPT occur, and the data of individual soundings are not binned with respect to potential temperature). We also note that the measured profiles of dθ/dz and mean LRM levels are in good agreement with ECMWF operational data (Figure 4, Panels c, f). 5 Figure 7 highlights the levels and layers used throughout this work (see also the summary in Figure 12 at the end of this paper).
Based on the definitions of LRM, CPT and TOC (introduced in Section 4.3), hereafter we will refer to: UT as the region of altitudes below the LRM, Asian TTL (ATTL) as the region between LRM and CPT, Confined LS (CLS) as the region between CPT and TOC, and Free LS (FLS) as the region of altitudes above the TOC. In the following of this section, we use these thermodynamically-significant levels and layers to study the vertical distribution and variability of water vapor, ozone and ice 10 saturation.

Water vapor and ozone
To analyze the vertical distributions and variability of tracers in relation to the thermodynamic structure of the UTLS, we define for each balloon sounding the altitude relative to the CPT as a new vertical coordinate. Figure 7 shows mean profiles and standard deviations of temperature, H2O mixing ratio and O3 mixing ratio in this new coordinate system. Besides the CPT 15 (black dashed line), the average LRM and TOC levels are shown by dashed and dotted lines, respectively (blue for NT16AUG, red for DK17). Water vapor mixing ratio in DK17 shows a minimum at the CPT and a local isolated maximum in the Confined LS (Panel b), centered about 1 km above the local CPT (i.e. not the average CPT, but evaluated for each profile individually).
The H2O minimum at the CPT is conceivably due to unusually high frequency of occurrence of ice clouds near the CPT in DK17 (see Ice saturation discussion in Section 5.3), which deplete water vapor from the gas phase in favour of the condensed 20 phase, hence resulting in a strongly dehydrated CPT. The H2O maximum in the CLS is likely due to overshooting convective updrafts, injecting ice crystals directly above the CPT, which then evaporate at higher levels and hence hydrate the CLS. A minimum in O3 mixing ratio slightly above the LRM is found in DK17 (Panel c), which is characteristic of very strong convection, quickly transporting ozone-poor air from the boundary layer to the TTL (Gettelman and de F. Forster, 2002;Vömel et al. 2002;Paulik and Birner, 2012). The absence of this feature in NT16AUG corroborates the statement that convection was 25 stronger during DK17 compared to the ASM 2016 season. H2O mixing ratios in the CLS compared to the FLS in DK17 are obviously related to the previously discussed isolated maximum, yet enhanced frequency of occurrence of high H2O mixing ratios is also observed in NT16AUG despite no local maximum 5 was found in this dataset. We argue that this enhancement is due to the anticyclonic confinement effect of ASMA, which keeps the moist convective outflow horizontally confined and thereby increases the frequency of occurrence of high H2O mixing ratios above the CPT in ASMA. To quantify this enhancement, we consider the mean values of H2O mixing ratio in the CLS and FLS, which are respectively 4.93 ppmv, 4.42 ppmv in NT16AUG, and 5.05 ppmv, 4.29 ppmv in DK17. This means that, on average, 0.49 ppmv (NT16AUG) and 0.76 ppmv (DK17) more H2O are found in the Confined LS inside the ASMA, compared 10 to background stratospheric water vapor in the Free LS above the ASMA, i.e. that H2O mixing ratio is enhanced by 11-18% in the lowermost stratosphere due to the confinement effect of ASMA.
The PDFs of O3 mixing ratio in the CLS, ATTL and UT (Panels f, i, l) show higher frequency of occurrence of low ozone values in DK17 compared to NT16AUG, again suggesting stronger convective activity in the ASM 2017 than in 2016. Ozone depletion extending into the Confined LS confirms that the influence of deep convection on the composition of the UTLS 15 persists up to altitudes well above the CPT, due to the horizontal confinement in ASMA. The fact that this feature vanishes in the Free LS (Panel c), suggests that our trajectory-based definition of TOC is a good measure of the topmost extent of the anticyclonic confinement.
These evidence show that, in the ASM system, the top height of the horizontal confinement in ASMA controls the distribution and variability of water vapor and ozone in the lowermost stratosphere, rather than CPT height as it is in the conventional 20 understanding of the TTL.  (Figure 7, Panel a), much higher and more persistent ice saturations were measured in DK17 than in NT16AUG 25 throughout the entire ATTL. In both datasets, ice saturation is higher in the ATTL compared to the UT, and in DK17 it shows a pronounced maximum with average supersaturated conditions (i.e. more than 50% of the measurements reach Sice > 1) in the 1.5 km directly below the CPT. In contrast, the supersaturated fraction is 10-15% in NT16AUG over the same range of altitude.

Ice saturation
The ATTL ice saturations and supersaturated fractions of NT16AUG are comparable with previous measurements in Lhasa and Kunming, China between 2009-2012(Bian et al., 2012 while the measurements in DK17 range significantly higher, confirm-30 ing that the ASM 2017 season was indeed extraordinary in terms of strong convection, and also suggesting particularly high frequency of occurrence of cirrus clouds in the ASMA 2017. Interestingly, we also note that ice supersaturations in DK17 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-222 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 11 April 2018 c Author(s) 2018. CC BY 4.0 License. extend frequently into the CLS, with about 30% supersaturated fraction in the first 500 m above the CPT (Figure 9, Panels b, c). This implies that, in overshooting convective updrafts, ice crystals can regularly penetrate the CPT as condensed phase (e.g. see Figure 10, Panel f) and evaporate at higher levels in the Confined LS, similarly but likely more frequently than was found in the tropics (Corti et al., 2008). We consider this to be the reason for the isolated H2O maximum above the CPT observed in DK17 (Figure 7, Panel b), which is, to our knowledge, unparalleled in the tropics. 5

Aerosol and cloud backscatter
Finally, here we analyze the aerosol and cloud backscatter measurements by COBALD, which have not been discussed so far.
COBALD was originally designed and used to investigate the properties of ice cloud, including cirrus (e.g. Brabec et al., 2012;Cirisian et al., 2014) and polar stratospheric clouds (e.g. Engel et al., 2014), but as already mentioned, recent measurements from Lhasa, China were also used for in-situ detection of the ATAL aerosols and validation of the Cloud-Aerosol Lidar and 10 Infrared Pathfinder Satellite Observation (CALIPSO) satellite retrieval (Vernier et al., 2015). In this section we address both aspects.
Since the BSR of aerosol droplets is 1-2 orders of magnitude smaller than that of cirrus clouds, the characterization of ATAL requires cloud-filtering techniques to eliminate in-cloud measurements, and therefore a large dataset for a statistically-significant evaluation (e.g. 18 soundings are used in Vernier et al., 2015). We performed 17 COBALD soundings in NT16AUG, but 15 due to logistical constraints only 3 could be realized during the DK17 campaign, which furthermore mostly sampled cloudy conditions in the ATTL. For this reason, a clear-sky aerosol BSR profile cannot be established from the DK17 dataset. On the other hand, the 3 COBALD soundings available from NT16NOV are almost fully clear-sky measurements and therefore allow to calculate a clear-sky aerosol BSR profile, providing a useful reference state of background aerosols (i.e. without ASMA confinement) for comparison with NT16AUG. In the following, we first provide an overview of the main characteristics of the 20 observed cirrus clouds, and then detail the cloud-filtering technique and ATAL detection during the year 2016.  (BSR455 -1). CI has the property of being independent of number density, hence it is a useful indicator of particle size (e.g. Cirisian et al., 2014) as long as particles are sufficiently small, so that Mie scattering oscillations are avoided. Based on the size-dependence of CI, considerations on the typical size range of ice crystals and aerosol droplets, and the evaluation of ice saturation measurements by CFH, a threshold of CI = 7 was empirically developed to discriminate in-cloud (CI > 7) from clear-sky measurements (CI < 7) (Vernier et al., 2015). This helps discerning the BSR features in Figure  30 Atmos. Chem. Phys. Discuss., https://doi.org /10.5194/acp-2018-222 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 11 April 2018 c Author(s) 2018. CC BY 4.0 License. the sharp feature at 145 hPa with CI ≈ 10 is likely an ice cloud (note the concomitant ice supersaturation above the thin cloud layer, suggesting sedimentation), while the broad enhancement in BSR between 95-140 hPa without CI enhancement is the signal of ATAL. The main common characteristics of the cirrus clouds in Figure 10 is their very small spatial and optical thickness, with BSR940 < 20, while much larger values (BSR940 >> 100) are expected in homogeneously-nucleated cirrus clouds, 5 as often observed in the midlatitudes (e.g. Brabec et al., 2012;Cirisian et al., 2014), and as also evidenced by the thick outflow cirrus below 120 hPa in DK002 (Panel f). Low BSR indicates low ice crystal number densities, suggesting that these clouds are most likely formed by heterogeneous nucleation on solid ice nuclei, rather than by homogeneous freezing of sulfate aerosol liquid droplets. This hypothesis is currently being investigated by a dedicated microphysical modeling study. Similarly thin cirrus clouds were observed in about half of the COBALD soundings in NT16AUG (9 out of 17) therefore they occur very 10 frequently in the ASMA, and they were often found embedded within the ATAL, as shown in Figure 10 (Panels a, e). Figure 11 shows all clear-sky (i.e. aerosol only) BSR455 data points and mean profiles from the NT16AUG and NT16NOV datasets.

ATAL in the ASM 2016
Similarly to Vernier et al. (2015), the cloud-filtering criterion we applied consists of three thresholds from two independent measurements, namely: BSR940 < 2.5 and CI < 7 from COBALD, and Sice < 0.7 from CFH. Only data points which simultane-15 ously fulfill all the three criteria above are classified as clear-sky and shown in Figure 11. The cloud-filtering method is illustrated by a scatter plot of BSR940 vs . CI shown in Figure S4 in supplementary material.
In NT16AUG, clear-sky BSR455 enhancement starting approximately at the LRM (180 hPa) and extending up to the TOC (73 hPa) is the signature of ATAL (Figure 10), showing intensity and vertical extent comparable to those derived by CALIPSO satellite retrievals and previous COBALD soundings (Vernier et al., 2015). The aerosol enhancement covers both the Asian 20 TTL and the Confined LS with maximum BSR455 at the CPT, suggesting that the LRM (i.e. the mean convective outflow level) is also the onset of the horizontal confinement in ASMA, and the CPT its level of maximum strength. Above the CPT, the BSR455 enhancement gradually faints with altitude until the TOC, as the horizontal confinement effect by ASMA vanishes (see Figure 5). Above the TOC, the ATAL signal merges with the Junge layer of stratospheric aerosols, which extends into the Free LS up to about 10 hPa. As expected, the aerosol enhancement is absent in NT16NOV, showing that ATAL does not outlive the 25 break-up of the anticyclonic vortex at the end of the ASM season.

Discussion and conclusions
We have analyzed 63 balloon measurements of temperature, water vapor, ozone and aerosol backscatter, collected in the region season (June-August), characterized by typical climatological conditions. The DK17 campaign took place over two weeks (29 July-12 August) with average frequency of two balloon soundings per day, coincident with a period of remarkably anomalous characteristics of the UTLS. The measurements in NT16NOV span an even shorter time period (5 days), but provide a useful reference state for seasonal variability considerations.
The structure of the tropopause region at the southern slopes of the Himalayas exhibits strong seasonal variability, with typi-5 cally tropical characteristics (sharp CPT) during the ASM season due to deep convection, and "mid-latitude" features (separated LRT and CPT) during the dry season. Therefore, for the analysis of our monsoon-season measurements, we adopt the concept of tropical tropopause layer (TTL) and define the region of altitudes between the LRM and the CPT as the Asian Tropopause Transition Layer (ATTL, or Asian TTL). Further, based on air mass trajectories, we define the top of confinement (TOC) level of ASMA, which separates the LS into a Confined LS (CLS), below the TOC and above the CPT, and a Free LS 10 (FLS), above the TOC. These thermodynamically significant levels and layers, qualitatively illustrated in Figure 12, provide physically-meaningful boundaries for the analysis of the vertical distribution and variability of water vapor, ozone and aerosols in the UTLS. , possibly suggesting an orographic influence. It is also interesting to notice that, in both datasets, the TOC coincides 20 with a change in slope of dθ/dz. Due to stronger convection, both the LRM and CPT are found at lower pressures in DK17 compared to NT16AUG (by about 10 hPa), but due to concomitant colder temperatures in DK17, the shift in potential temperature space is small (around 2 K). Ozone mixing ratio in DK17 shows a slight minimum above the LRM, which indicates strong convection (Vömel et al., 2002;Paulik and Birner, 2012) and was not observed in NT16AUG. In both datasets, water vapor is higher in the Confined LS compared to the Free LS (see also Figure 8), and in DK17 the H2O mean profile shows a local 25 maximum about 1 km above the CPT. This feature contradicts the common conceptual understanding of TTL, which claims no water vapor variability above the CPT (Pan et al., 2014), and we suggest this is due to overshooting convection injecting condensed-phase ice directly above the CPT, which subsequently evaporate and hydrate the CLS. Ice saturation is minimum at the LRM and increases in the ATTL, similarly to as in the tropics (Vömel et al., 2002). Due to extremely cold temperatures, average Sice in DK17 is remarkably higher than in NT16AUG, as well as than in previous measurements from the Tibetan plateau 30 (Bian et al., 2012). Aerosol backscatter enhancement of the ATAL was detected in NT16AUG with similar vertical extent and BSR intensity as in previous measurements from Lhasa, China (Vernier et al., 2015). On average, ATAL extends from the LRM to the TOC with maximum backscatter at the CPT, suggesting that the mean convective outflow level (i.e. the LRM) is also the onset of the horizontal confinement in ASMA, CPT its level of maximum strength, and TOC its topmost extent. Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-222 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 11 April 2018 c Author(s) 2018. CC BY 4.0 License.
In conclusion, our analysis shows that the composition of the tropopause transition layer and the lowermost stratosphere within the ASM anticyclone are influenced by deep convection up to 1.5-2 km above the cold-point tropopause, due to the horizontal confinement effect of ASMA. Hence, the vertical distributions and variability of water vapor, ozone and aerosols in this region are controlled by the topmost extent of the anticyclonic confinement, rather than the altitude of the cold-point tropopause, as it is in the conventional understanding of the "generic" TTL. This is shown by enhanced water vapor and aerosol backscatter 5 found in the confined lower stratosphere, in contrast to typical background stratospheric values observed above the top of confinement. These evidence strongly suggest that the ASM anticyclone contributes to moistening the global stratosphere and to increase its aerosol burden. For our two monsoon-season campaigns in 2016-2017, the water vapor enhancement is quantified respectively as 11-18% (0.5-0.75 ppmv), in terms of average H2O mixing ratio difference between the confined lower stratosphere (i.e. above the cold-point and inside the ASM anticyclone) and the background stratospheric H2O measured in the 10 free lower stratosphere above the anticyclone.