Persistence of moist plumes from overshooting convection in the Asian monsoon anticyclone

. The Asian Monsoon Anticyclone (AMA) represents the wettest region in the lower stratosphere (LS) and is the key contributor to the global annual maximum in LS water vapour. While the AMA wet pool is linked with persistent convection in the region and horizontal confinement of the anticyclone, there remain ambiguities regarding the role of tropopause-overshooting convection in maintaining the regional LS water vapour maximum. This study tackles this issue using a unique set of observations from onboard the high-altitude M55-Geophysica aircraft deployed in Nepal in Summer 2017 within the EU StratoClim project. 25 We use a combination of airborne measurements (water vapour, ice water, water isotopes, cloud backscatter) together with ensemble trajectory modeling coupled with satellite observations to characterize the processes controlling water vapour and clouds in the confined lower stratosphere (CLS) of AMA. Our analysis puts in evidence the dual role of overshooting convection, which may lead to hydration or dehydration depending on the synoptic-scale tropopause temperatures in AMA. We show that all of the observed CLS water vapour enhancements are traceable to convective events within AMA and furthermore bear an isotopic 30 signature of the overshooting process. A surprising result is that the plumes of moist air with mixing ratios nearly twice the background level can persist for weeks whilst recirculating within the anticyclone, without being subject to irreversible dehydration through ice settling. Our findings highlight the importance of convection and recirculation within AMA for the transport of water into the stratosphere. combine local airborne measurements with global satellite observations to characterize the mechanisms of convective impact on water vapour and clouds through both mass and energy transport above the cold point tropopause. We observational evidence of convectively-induced lower stratosphere hydration and dehydration of both irreversible and reversible 70 types. The link between the local variations and phase transitions of water with deep convection across the Asian anticyclone is investigated using ensemble trajectory modeling constrained by satellite detections of convective


Persistence of moist plumes from overshooting convection in the Asian monsoon anticyclone
x is the flight number) performed every second day during the period 27 July -10 August in both the morning and afternoon hours.
Three of the flights were performed within the Nepali borders, whereas in the other flights the airplane flew out to southwest, south and southeast from Nepal reaching the Bay of Bengal (see Fig. S1 of the Supplement). The Geophysica aircraft hosted a large number of in situ and remote sensors for measuring gaseous and particulate UTLS composition. A full description of the campaign is provided by Stroh et al. (2021, in prep. same issue). In this study, we use in situ measurements of water vapour, total water and 85 water isotopologues respectively by FLASH, FISH and ChiWIS instruments as well as particle backscatter measurements by the onboard MAS scatterometer and MAL lidar.

FLASH-A (Fluorescent Lyman-Alpha Stratospheric Hygrometer for Aircraft) is an airborne instrument of the FLASH
hygrometer family designed specifically for the M55-Geophysica aircraft . The instrument was redesigned in 90 2009 (Khaykin et al., 2013) for the RECONCILE campaign (von Hobe et al., 2013) and then substantially improved for the StratoClim experiment. FLASH-A is mounted inside a gondola under the right wing of Geophysica and has a rear-facing inlet, enabling water vapour measurements. With the aspiration rate of 470 cm 3 /s, the air samples in a 90 cm 3 measurement chamber are fully exchanged every 0.19 s. The chamber is maintained at constant temperature (24 °C) and pressure (36 hPa). Before a flight, the instrument is ventilated for several hours using dry air (< 1 ppmv) whereas the inlet tube, heated to 30 °C, is kept sealed before 95 the aircraft climbs to 250 hPa level to avoid chamber contamination by moist tropospheric air.
Unlike the previous airborne versions of FLASH-A with transverse optical setup, the StratoClim FLASH-A rendition has a coaxial optics similar to that of the FLASH-B balloon-borne instrument (Yushkov et al., 1998). The water vapour mixing ratio is detected by sensing the fluorescence light yielded by photodissociation of water molecules after their exposure to Lyman-alpha radiation. A near Lyman-α line (123.6 nm), is produced by a krypton lamp whereas the hydroxyl fluorescence at 300 -325 nm 100 wavelength range is detected by a photomultiplier operating in photon-counting mode. The accuracy of water vapour measurements in 1 -100 ppmv range is estimated at 8%, whereas the precision of 1 Hz data in the stratosphere is 0.6 ppmv with a detection limit of 0.1 ppmv for 5 s integration time. FLASH-A was calibrated against a reference MBW-373L frost-point hygrometer before and after the aircraft deployment as well as during the campaign using FISH calibration facility. During StratoClim campaign, FLASH-A operated in all the eight scientific flights as well as during the transfer flight to Kathmandu.

105
ChiWIS (Chicago Water Isotope Spectrometer) is an airborne implementation of the ChiWIS-lab instrument (Sarkozy et al., 2020) designed for atmospheric chamber measurements of water vapour and water isotopologues under UTLS conditions. The new version of the instrument is a tunable diode laser (TDL), off-axis integrated cavity output spectrometer (Clouser et al., 2021, in prep. same issue). The spectrometer scans absorption lines of both H2O and HDO near 2.647 µm wavelength in a single current sweep. With a 90 cm-long multi-pass cell, the effective path length amounts to more than 7 km. During the airborne campaign, the 110 instrument has demonstrated measurement precision for ten second integration times of 18 ppbv and 80 pptv in H2O and HDO, respectively. The measurements were reported at 0.2 -0.5 Hz frequency depending on the ambient mixing ratio and the desired signal-to-noise ratio. Periods of the flights where the internal cell pressure of ChiWIS was below 30 hPa are not reported because

115
FISH (Fast In situ Stratospheric Hygrometer) is a closed-path Lyman-α fluorescence hygrometer with a forward-facing inlet, which enables measurement of total water (sum of gas phase water and sublimated ice crystals). The measurement accuracy is 6 % -8 % whereas the precision of 1 Hz data is estimated at 0.3 ppmv (Zöger et al., 1999;Meyer et al., 2015). Inside the cirrus clouds, the ice water content (IWC) is calculated by subtracting the FLASH-A gas-phase water from the total water measured by FISH, as described by Afchine et al. (2018). The minimum detectable IWC is 3×10 −2 ppmv (∼3×10 −3 mg m −3

130
The temperature was measured by TDC (ThermoDynamic Complex), a modified Rosemount 5-hole probe that provides an accuracy of 0.5 K and precision of 0.1 K for temperature measurements at 1 Hz frequency (Shur et al., 2007). We used TDC measurements of temperature and pressure to convert FLASH-A water vapour mixing ratio into relative humidity over ice (RHi) as well as to compute the saturation mixing ratio using the saturation vapour pressure equation by Murphy and Koop (2005). The accuracy of TDC measurements is discussed by Singer et al. and Stroh et al. (2021, in prep, same issue) 135 NIXE-CAPS (New Ice eXpEriment: Cloud and Aerosol Particle Spectrometer) is mounted under the right wing of Geophysica and measures the cloud particle number size distribution in the size range of 3-930 µm diameter at a time resolution of 1 Hz (Meyer, 2012). The IWC derived from particle size distribution are found to be in good agreement with those derived from FISH total water measurements . The lower detection limit of the instrument is 0.05 ppmv (≈0.005 mg m −3 ). NIXE-CAPS provided measurements in all the flights.

140
In situ measurements of cloud/aerosol backscatter and with a time constant of 10 s were provided by the forward-looking backscatter probe MAS (Multiwavelength Aerosol Scattersonde) described by Buontempo et al. (2006). To distinguish between clear-sky and in-cloud measurements, here we use a threshold of 1.2 units of backscatter ratio at 532 nm together with a 2.5% threshold in the volume depolarization (corresponding to particle depolarization of 8-10%). MAS instruments operated in all the flights except F1, for which we used NIXE-CAPS data  to detect the clouds. Singer et al. (2021, in prep. same 145 issue) showed good agreement between the cloud detections by both of these instruments.
Remote measurements of cloud backscatter below and above the aircraft were conducted by Miniature Aerosol Lidar (MAL) (Mitev et al., 2002). Backscatter ratios at 532 nm are derived after applying a noise filter, range correction and correction for incomplete overlap in the near range, allowing observations as close as 40 meters from the aircraft.  Livesey et al. (2017), who report for the lower-middle stratosphere a vertical resolution of 2.8 -3.2 km and an accuracy of 4 -9%. The data screening criteria specified by Livesey et al. (2017) have been 155 applied to the data. To interpolate the water vapour profiles onto a common potential temperature grid, we use the MLS temperature product provided at the same pressure levels.
Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) is a primary instrument onboard the CALIPSO satellite, operational since 2006 (Winker et al., 2009) and providing backscatter coefficients at 532 and 1064 nm with a vertical resolution of 30 m in the UTLS. Here we use CALIOP 532 nm level 1B version 4.0 product for diagnosing the cloud vertical cross-sections 160 and for quantifying the cloud top altitude. To enhance the sampling of clouds, we also use level 1 backscatter product at 1064 nm provided by NASA's Cloud-Aerosol Transport System (CATS) lidar operating onboard International Space Station (Yorks et al., 2016). The CATS 1064 nm backscatter is converted to 532 nm using CALIOP color ratio.
The GPS Radio Occultation (RO) technique provides vertical profiles of atmospheric variables with high vertical resolution (~0.5 km around the tropopause), global geographical and full diurnal coverage, and high accuracy (<1 K) (Steiner et al., 1999).

165
We use RO "dry" temperature profiles from COSMIC (Anthes et al., 2008); GRACE (Beyerle et al., 2005) and Metop A/B missions (Luntama et al., 2008) for analyzing the temperature and minimum saturation mixing ratio within AMA during July and August 2017.

Definitions
The vertical boundaries of the tropical tropopause transition layer (TTL) can be defined using two different approaches 170 reviewed by Pan et al. (2014). The mass-flux approach (Fueglistaler et al., 2009) defines the lower boundary as the tropicallyaveraged level of all-sky zero net radiative heating (14 km, 355 K) and the upper boundary as 18.5 km (425 K), where the local mass flux becomes comparable to that of the Brewer-Dobson circulation (Fu et al., 2007). Another approach is based on the TTL thermal structure, where the lower and upper boundaries are defined respectively as the level of minimum stability and the cold point tropopause (CPT) (Gettelman and Forster, 2002). In this study, we adopt the thermal definition of the TTL as in this case the 175 boundaries can be derived from the local instantaneous measurements provided by the Geophysica. Pan et al. (2014) found that the thermally-defined TTL boundaries are consistent with those derived from the ozone-water vapour relationship, which renders the thermal approach to TTL definition most suitable for this study.
Since the location of the Geophysica deployment is not tropical in the geographical sense, we refer to the TTL in this region as the Asian tropopause transition layer (ATTL) with an upper boundary at the CPT derived from ERA5 temperature profiles 180 collocated with the flight tracks and using airborne temperature profiles. Following Brunamonti et al. (2018), we refer to the upper layer of the Asian anticyclone as confined lower stratosphere (CLS) with a lower boundary at the CPT level and an upper boundary corresponding to the top level of confinement, which they estimate as 63.5 hPa (~440 K) for the 2017 AMA season.
A convective overshoot (also termed "ice geyser" by Khaykin et al. (2009)) is defined as detrainment of ice crystals above the local CPT. Depending on the relative humidity at the level of detrainment, this process can lead either to CLS moistening by rapid 185 ice sublimation, or to irreversible dehydration via uptake of vapour by the injected ice crystals, their growth and sedimentation.
The clouds that have formed in the CLS as a result of local cooling are termed in situ cirrus. A secondary cloud refers to an in situ cirrus cloud that has nucleated from an air mass enhanced in water vapour as a result of convective overshoot.  classes that are representative of deep convection. In this study, we consider the convective hits above 100 hPa only, corresponding to the cloud tops potentially overshooting the tropopause. They constitute 24.3% of the total number of convective hits identified by this analysis. The convective origin of the sampled parcels is statistically diagnosed in terms of the fraction of convective hits per 1 s sample as well as the convective age of parcel, i.e. the time since convective hit. A detailed description of the trajectory modeling setup and geostationary satellite products is provided by Bucci et al. (2020). The time evolution of temperature in terms of H2O saturation mixing ratio (Fig. 1c) and water vapour (Fig. 1d) within the flight domain shows an interesting development of the UTLS conditions before and during the campaign period. In mid-July, a humid layer in the ATTL starts to build up and propagates above CPT up to about 410 K by early August. The first four Geophysica flights were conducted during this moisture build-up period, characterized by relatively warm CPT temperatures (Fig. 1c), which 230 we term "warm/wet" period. It should be noted that the 2017 AMA season was marked by a strong positive anomaly in water vapour ranging 1 -2 ppmv but without a significant tropopause temperature anomaly ( Supplementary Fig. S2). The positive water vapour anomaly is not specific to AMA region and reflects the global wet anomaly in the tropics and subtropics, as revealed by MLS observations (not shown).
In early August, after the warm/wet period, the ATTL experienced a rapid cooling and the last four flights sampled a colder 235 and drier ATTL. This "cold/dry" period is marked by stronger convective activity in the region reflected by low OLR (Fig. 1c) and higher carbon monoxide ( Fig. 1d), indicative of an enhanced flux into the stratosphere. This period is also marked by a widespread occurrence of ice clouds above the CPT (377 -390 K) and as high as 415 K level according to high-resolution cloud profiling by CALIOP and CATS satellite lidars (Fig. 1d). We note that the cold convective period had a transient effect on the CLS water vapour, which mostly recovers the late July values, after the cease of convective activity and tropopause warming in the flight 240 domain. A similar inference was reported by Brunamonti et al. (2018) on the basis of balloon soundings in Nepal during the airborne campaign.

Airborne perspective
The airborne measurements of water vapour and temperature shown in Fig. 2 reflect the satellite-derived development of the 245 UTLS conditions. The ensemble of water vapour profiles obtained using the FLASH hygrometer during the eight StratoClim flights is shown in Fig. 2a. The H2O vertical profiles at and above the CPT level show a striking variability over the two-week campaign period with mixing ratios ranging from 2.8 to 10.2 ppmv. On average, the warm/wet period yielded an L-shaped mean H2O profile (solid curve), which is characteristic of the Boreal subtropical conditions, although with notable enhancements at and above the CPT. In contrast, the cold convective period revealed the vertical distribution more typical for the tropical tropopause conditions,

250
with the hygropause at the CPT level. The airborne measurements during both synoptic periods show an accumulation of sharp moist layers above the CPT and up to 410 K level which are diagnosed in the following section. These layers constitute the CLS wet pool seen by MLS, although their sharp vertical structures of sub-kilometer scale can hardly be resolved by the satellite.
The large variability of water vapour is consistent with the tropopause temperature variability, showing minimum saturation mixing ratio between 2.5 and 6.5 ppm and highly variable CPT vertical structure (Fig. 2b), presumably modulated by gravity 255 waves. The CPT potential temperature varied between 370 -391 K, which is fully consistent with the GPS-RO data.
The highly-variable thermal conditions led to a remarkable dispersion of RHi around the CPT. The clear-sky measurements reveal both a subsaturated and strongly supersaturated environment with RHi spanning 40 -175 % (Fig. 3a). The credibility of RHi data is ensured by an excellent agreement across the three airborne hygrometers and temperature sensors (TDC and UCSE) (Singer et al., in prep. same issue). In the presence of ice crystals (Fig. 3b), the RHi is generally well above 100% although the 260 subsaturated cloud occurrences were also observed in both dry and wet parts of the water vapour spectrum. Those are mainly caused by short excursions of temperature above the frost point, which does not necessarily lead to permanent evaporation and depends on the Lagrangian temperature history. The occurrence of ice crystals was recorded at levels 15 K (~1 km) above the local CPT. The highest-level clouds were detected by the upward-looking MAL lidar at 412 K (18.5 km), which is consistent with NIXE-CAPS detection of cloud particles up to 415 K (Krämer et al., 2020, their Fig. 11) as well as with the maximum cloud altitudes 265 inferred from satellite lidars (415 K). The presence of ice in supersaturated air is more specific to the cold dry parcels (see also Krämer et al., 2020, their Figure 10d), which suggests a local dehydration during the cold/dry period.
A different perspective on the environmental conditions of cloud occurrence around the CPT is provided in Fig. 3c, showing the distribution of IWC as a function of RHi. The binned ensemble is restricted to the samples, for which both MAS and NIXE-CAPS data indicate the presence of ice particles. The ice crystals found in the subsaturated air above the local CPT are likely to be 270 in the process of sublimation and therefore have a potential for a permanent CLS hydration. Conversely, the crystals in the supersaturated environment will retain their aggregate state and the largest ones (characterized by higher IWC) will sediment down below the tropopause thereby causing permanent dehydration of the CLS. We note that the ice particles in the subsaturated environment account for 14% of the particles detected above the local CPT.

275
The influence of overshooting convection on the observed water vapour variability was investigated using TRACZILLA

Isotopic composition of convective plumes
The relation of moist layers with overshooting convection can be reliably diagnosed using the isotopic ratio of water 290 (HDO/H2O), which is enhanced for water vapour molecules sublimated from ice. Figure 4b clearly shows that the wetter parcels in the lower stratosphere are isotopically enhanced, and the wettest of them bear the strongest isotopic signature. This unambiguously points out that the hydrated layers have been produced by overshooting ice geysers. Remarkably, the wet and isotopically enhanced pixels in Fig. 4b are found as high as 420 K level, that is 30 -50 K above the cold point. Given the diabatic heating rate of 1.1 K/day in AMA and the average recirculation time of 16 days within the anticyclone ,

295
these hydrated parcels could, in principle, have recirculated twice before being sampled by the aircraft. N. The convective age for these parcels varies between 2.6 and 9.9 days. Sensibly, the shortest age corresponds to the lower-height moist layer at 390 K level (F2), whereas the longest age is found for the wet and heavy parcels detected as high as at 410 K in F4.
The moist features in the flight F2 (A2 and B2, see Sect. 5), found at 390 and 399 K levels, are sourced to different convective 310 events that occurred 2.6 and 4.7 days before being sampled by Geophysica.
While the warm/wet period flights were largely influenced by convection in the Northeastern part of AMA, the wet and heavy parcels sampled during flights F5 -F7 are sourced to various different locations. A large convective system over North-Eastern India in the vicinity of the flight on the same day is responsible for the hydration feature in F6 at 380 K level. The convective source of the wet air sampled by F5 is found above Western India, although we note that no isotopic data are available for this 315 flight, whereas the number of parcels with mixing ratio exceeding one standard deviation is small for this flight. In flight F7, the enhanced water vapour features above 400 K (A7 and B7, see Sect. 5) originate from two different sources: the lower-level feature (A7) is traced back to a group of relatively small systems along the Eastern Chinese coast that occurred 3.9 days before the measurement, whereas the upper one (B7) originates from a large cluster of small-scale convective systems in the center of Asian anticyclone above the northern foothills of Himalayas. We note that this particular region is marked by enhanced water vapour 320 amount according to MLS averages over the campaign period (cf. Fig. 1b). The B7 parcels have thus followed the anticyclonic circulation path for nearly a full loop before arriving to the flight domain, which took 12.7 days.
The potential for a vapour-rich parcel travelling within AMA CLS to permanently hydrate the stratosphere is determined by the Lagrangian temperature history. We did not analyze the RHi variation along the trajectories, however we quantified the minimum temperatures encountered across AMA using high-resolution GPS-RO profiling. As follows from Fig. 1a, the subtropical 325 part of AMA has never cooled below the H2O saturation mixing ratios of around 8 ppmv in July-August 2017, enabling the vapourrich patches to travel along the northern flank of the anticyclone without freezing. Remarkably, the majority of convective systems identified as the most probable sources of wet and heavy parcels (shown in Fig. 5 and marked by black pixels in Fig. 1a) have occurred within the warm tropopause environment in the northern subtropical part of AMA.
It is noteworthy that the probed wet and heavy parcels (shown along the flight tracks in Fig. 5) are all located in the 330 northernmost part of the flight domain, i.e. nearer the center of AMA. This is consistent with the spatial distribution of AMA CLS water vapour inferred from MLS (Fig. 1b), showing the maxima above the Tibetan plateau and Sichuan region, that is around the center of the anticyclone. With that, the air circulating near the outer edge of the anticyclone is bound to pass the colder TTL above (cf. Fig. 1) has led to cooling and dehydration at around the CPT level. The efficiency of the convectively-induced dehydration,

335
counteracting with the convective moistening in the warmer TTL regions of AMA is considered on a case by case basis in the next section.

Long-range transport and evolution of moist convective plumes
The hydrated layers in the CLS characteristic of elevated convective hits fraction and/or isotopic enhancement (wet and heavy) were detected at altitudes between 16.9 -19.0 km (380 -415 K) in all the flights except F8 with the magnitude of mixing 340 ratio enhancement between 0.9 -5 ppmv (see Supplementary Fig. S3). The largest enhancement (5 ppmv) was observed in F2 at 399 K (B2 feature), whereas the highest altitude of hydrated layer centered at 18.9 km (411 K) was sampled in F7 (B7 feature).
The flights F2 and F7 represent respectively warm/wet and cold/dry regimes (see Sect. 3), however in both of these flights the observed moist layers originated from distant convective events. In this section, we provide further insight into the results of F2 and F7 and describe the evolution of the respective moist convective plumes using airborne and satellite measurements.

Warm and wet regime: Flight 2
During the warm/wet regime, the mean CPT-level water vapour mixing ratio was 7.2 ppmv, whereas the minimum saturation mixing ratio ranged from 5.5 to 6.9 ppmv. During F2, the aircraft was cruising side to side along the Himalayan foothills within Nepali borders gaining altitude in 500 m steps before climbing to 21 km (Fig. 6c). The water vapour vertical profile in Fig.   6a,b reveals two layers above the CPT (marked A2 and B2) with water mixing ratio peaking at 10.2 ppmv, twice the campaign- The upper layer (B2) topping at 399 K (~18 km) is characterized by very large fraction of convective hits reaching 0.9 ( Fig. 6a) with an average age of 4.7 days (cf. Fig. 5). The convective origin of B2 is unambiguously confirmed by a strong 355 enhancement in the HDO/H2O ratio of -340 per mill. This is substantially higher than the isotopic ratio found for the equivalenthumidity air below the CPT (about -480 per mill at 373 K level). The enhanced isotopic ratio in this layer clearly indicates that the water vapour enhancement was produced by sublimation of ice. It is remarkable that after nearly 5 days, the convective plume responsible for B2 feature has retained such an amount of moisture.
The underlying wet layer (A2) at ~390 K (~17.5 km) is traced back to a different convective event aging 2.4 days (cf. Fig.   360 5). However, given that the magnitude of enhancement is nearly the same as that of its upper-level twin, it is conceivable that both A2 and B2 represent the outflow of the same convective event in Northeastern China, and the lower-level A2 feature is a result of gravitational settling of ice crystals shortly after injection.

Secondary cloud formation
For an air parcel at 82 hPa bearing 10 ppmv of water vapour (as reported for B2), the saturation is achieved at -78.5 °C.

370
Interestingly, the bottom of this cloud is found at the same potential temperature level as the hydrated layer B2 and only about 350 km away from it. Nevertheless, these features appear to have different convective sources. Figure 7a shows the back trajectories released from this cloud intersecting a large convective system above Northeast China on 21 July, 8 days before F2. In an attempt to investigate the evolution of humidity of this air mass, we searched for the MLS swaths collocated in space and time (within 500 km and 1 hour) with the tracked parcels. A perfect match was found on 24 July: the MLS swath lies precisely across 375 the cluster of the tracked parcels as shown in Fig. 7b. The nearest MLS profile reports 8 ppmv at the parcel level, which is 2 -3 ppmv wetter than the neighboring measurements along the same orbit. This suggests that the moist plume remained compact up to 3 days after the convective event.
The Lagrangian temperature history of this air mass (Fig. 7c) suggests that since the convective encounter, the parcels remained subsaturated most of the time and, in particular, during the collocated measurement by MLS. The RHi was estimated 380 from the ERA5 temperature and pressure along the back trajectories, whereas the mixing ratio was assumed to be constant 12 ppmv. The episodes of moderate supersaturation with RHi reaching 140% were encountered between about 144 to 170 h before the sampling and it is conceivable that cirrus could have formed around that time and some water was lost to sedimentation.
However, the episodes of strong supersaturation with RHi reaching the homogeneous freezing threshold were encountered only during the last day before the measurement, when the parcels were entering the colder CLS above the Southern slopes. The RHi 385 along the back trajectories during the last day was reaching 160%, which would enable ice nucleation and repartitioning of the excessive vapour into a secondary cloud.

Cold and dry regime: Flight 7
The cold/dry regime was marked by a synoptic-scale cooling throughout the 370 -400 K layer extending across the CPT.
The largest vertical extent of the cold layer was observed in F7, where the saturation mixing ratio dropped below 4 ppmv throughout 390 370 -397 K layer (cf. Fig. 2b). The northbound flight leg of F7 (see flight track in Supplementary Fig. S1) included several porpoises across the CPT level (varying between 375 -383 K) as shown in Fig. 8a. The time series of potential temperature is marked with ice particle occurrence detected by MAS, which shows the presence of subvisible cirrus clouds with scattering ratio below 6 extending up to 400 K level.
The water vapour time series in Fig. 8a reveals a remarkably large horizontal variation of mixing ratio in this layer, spanning 395 3.0 to 6.2 ppmv on a horizontal scale of hundred kilometers. Almost the entire CPT-porpoising segment of F7 shown in Fig. 8a (20000 -22200 s) is supersaturated with RHi reaching 155%, whereas the water vapour variation follows the saturation mixing ratio with a high degree of correlation (r = 0.97). The occurrence of ice particles detected by MAS is reflected by enhancements in IWC shown as blue shading in Fig. 8a,b. The magnitude of IWC enhancements (up to 3.3 ppmv) is comparable to the magnitude of water vapour reduction, which suggests that these ice crystals have formed in situ as a result of synoptic-scale CPT cooling.

400
Indeed, as shown in Fig. 8b, the total water does not exceed the background level represented by the mean water vapour from the previous flights. This does not however rule out that some of these crystals were produced by overshooting as suggested by Lee et al. (2019) for this particular flight.
Above the layer of thin cirrus reaching 400 K level, the water vapour profiles in F7 reveal two enhancements, marked in between 405 -415 K (18.5 -19 km). Both A7 and B7 moist features are characteristic of significant isotopic enhancement (Fig.   8c), whereas the B7 is also marked by an enhanced fraction of convective hits (cf. Fig 4a and Supplementary Fig. S3). From the convective sources' analysis in Sect. 4.2, we know that the hydrated feature B7 has a convective age of 12.7 days during which the moist convective plume has made a nearly complete circle within AMA. During this time, the mixing within the moist layer is expected to smoothen its vertical structure, however the B7 enhancement reveals a rather sharp vertical structure. Such a sharp 410 structure is normally associated with recently sublimated ice crystals from a nearby overshoot (e.g. Khaykin et al., 2009;2016).
The absence of recent (< 5 days) convective events (see Supplementary animation) upwind of B7 has led us to investigate the satellite cloud measurements and temperature history along the corresponding backward trajectories. On 6 August, neither CATS nor CALIOP transects (marked as T1 and T2 in Fig. 9a) show the presence of clouds above 18 420 km, which is consistent with the parcel's temperature history in Fig. 9c showing sub-saturated conditions before T1. After passing the T2 point, the parcel has experienced a strong cooling episode, boosting the maximum RHi above the homogeneous freezing threshold. The next collocated lidar overpass (T3) took place on 7 August when the parcel's temperatures have just relaxed down to saturation levels. The lidar curtains labeled T3 and T4 (Fig. 9b) show evidence of partial evaporation of the cirrus cloud crosssampled by CATS at 17 UT (T3) and by CALIOP three hours later (T4). The location of the tracked parcels (marked by a white 425 rectangle in Fig. 9b and Supplementary Fig. S5b), matches precisely the evaporating fraction of the cloud. Thus, the final sublimation of this secondary cloud has occurred about 13 hours before B7 sampling, which can explain its sharp vertical structure.

Secondary cloud sublimation
At the time of B7 sampling, the parcel's RHi -computed from ERA5 temperatures and FLASH peak value of 7.3 ppmv in the hydrated layer -amounts to nearly 100%, which is consistent with the airborne temperature measurement. Downwind of the flight track there are two transects (T5 and T6), not necessarily collocated in time with the westward progression of B7 parcel, but

430
showing an absence of ice particles at the respective level ( Fig. 9 and Supplementary Fig. S5).
The above led us to conclude that while the B7 water vapour enhancement was produced by a 12.7-days old convective plume that circumnavigated AMA, its vertical structure was modified by strong yet transient cooling episodes that acted to temporally repartition the vapour into ice on a scale of several hours. As inferred from NIXE-CAPS particle size distribution measurements in F8, a freshly nucleated in situ cirrus at around the CPT level is dominated by very small ice crystals with effective 435 diameter of 4 -10 μm ( Supplementary Fig. S6). According to Muller and Peter (1992) such crystals would sediment at a rate of 0.6 -2 cm/s. Assuming the onset of ice crystals nucleation at B7 -20 h (corresponding to the onset of the strong cooling episode) and their evaporation at T4 point, the cloud particles should have sedimented by less than 200 -700 meters during their lifetime.
With this case we point out that the homogeneously-nucleated crystals smaller than 10 μm occurring in the CLS as a result of convectively-induced radiative cooling and/or gravity waves-induced temperature perturbations do not last long enough to 440 sediment out from the stratosphere and therefore have limited potential to dehydrate the CLS. The occurrence of water vapour enhancements in the lower stratosphere associated with overshooting convection has been reported in several studies based on in situ measurements in the deep tropics over Western Africa (Khaykin et al., 2009;Schiller et al., 2009), Northern Australia (Kley et al., 1993;Corti et al., 2008), South America (Khaykin et al., 2013), Central America 445 (Sargent et al., 2014), Western Pacific (Jensen et al., 2020) as well as at midlatitudes over North American monsoon (Hanisco et al., 2007;Weinstock et al., 2007;Smith et al., 2017) and Asian monsoon (Vernier et al., 2018;Brunamonti et al., 2018;Krämer et al., 2020). We note that the reported cases represent a small fraction of in situ measurements acquired; there is typically no more than one case of water vapour enhancement above the tropopause detected during a field campaign.
Compared to other field campaigns, the StratoClim aircraft deployment in Nepal provided a lot of evidence for moist layers 450 above the tropopause. Their convective overshooting origin is unambiguously supported by both the enhanced isotopic ratios in the moist plumes and by their traceability to convective events. Notably, the occurrence of lower stratospheric moist plumes above the monsoon regions is also supported by satellite observations (Fu et al., 2006;Schwartz et al., 2013;Werner et al., 2020), whereas the enhanced water isotopic ratios observed over these regions (Hanisco et al., 2007;Randel et al., 2012) support the role of overshooting convection in maintaining the water vapour maximum in the summer monsoon anticyclones. This process adds to 455 the radiatively-driven slow ascent of wet air through the warm tropopause in the northern part of AMA. Another possible pathway of water into the CLS in addition to the slow ascent and overshooting may be the isentropic transport across the CPT from the Tibetan plateau (characterized by highest CPT) to the southern slopes of Himalayas.
Using MLS observations and OLR data, Randel et al. (2015) concluded that stronger convection in the Asian monsoon region leads to colder and drier lower stratosphere whereas the opposite is true for the weaker convection. They also pointed out 460 the importance of subseasonal variations of deep convection driving the water vapour amount near the tropopause. Interestingly, the composited maps of OLR anomalies for wet and dry regimes (Fig. 5 in Randel et al. (2015)) reveal an east-west dipole and in both cases this dipole is centered exactly on the StratoClim flight domain. Furthermore, the evolution of the UTLS conditions in the flight domain, switching from warm/wet to cold/dry regime over the course of the campaign, allowed for sampling the oppositesign effects of deep convection on the water vapour above the tropopause.

465
Our analysis confirms the conclusion of Randel et al. (2015) regarding the strong relation between convectively-modulated temperature and water vapour around the tropopause. Indeed, the second (cold/dry) period of the campaign with organized convection in the region led to synoptic-scale CPT cooling and a drastic drop of water vapour by ~30% around the tropopause. We note though that the dehydration layer did not extend above 395 K. A similar conclusion was made by Brunamonti et al. (2018) on the basis of balloon soundings of water vapour and ozone in Nepal as part of StratoClim campaign in 2017. They argued that 470 overshooting convection is responsible for an isolated maximum of H2O in the CLS observed in July 2017, whereas the water vapour minimum at the CPT level is caused by synoptic-scale cold anomaly above the southern slopes that maximized around 9 August.
Our trajectory analysis suggests that the convective origin is characteristic of the wettest and the driest parcels (Fig. 4a), which points out the dual role of overshooting convection on the AMA water vapour. With that, we note that the probability of 475 dehydration decreases with the age of convective outflow, ascending within AMA at an average 1.1 K/day rate in potential temperature . This way, a hydrated air mass, circulating within the confined anticyclone, progressively moves up and away from the tropopause and becomes less likely to encounter permanent dehydration. A similar inference was made by Ueyama et al., (2018)

495
such phenomena were never before observed in the deep tropics. The recirculation of water vapour-enhanced air masses was reported in the Antarctic and Arctic vortices (Voemel et al., 1995;Khaykin et al., 2013) where the hydration of lower stratosphere occurs through sedimentation of ice PSCs.
Overall, our results suggest a complexity of processes controlling water abundance and its aggregate state in the lower stratosphere of AMA. The strong isotopic enhancements specific to the moist layers in the CLS and their traceability to convective 500 events consistently suggest that overshooting convection is an important contributor to the seasonal maximum of water vapour in the AMA lower stratosphere. At the same time, the large-scale organized convection in the southern part of AMA is shown to cause synoptic-scale dehydration around the tropopause through radiative cooling. Another mechanism of dehydration is the overshooting of ice crystals into the supersaturated environment above the tropopause, which leads to their rapid growth and sedimentation. The evidence of such a process was obtained in a particular flight (F8) and will be a subject of a separate study. data. MK, AA, CR and NS provided airborne total water and ice particles size data. FC provided airborne particle backscatter in situ data. VM and RM provided airborne lidar data. VV provided airborne temperature data. EM, CS, BC, MK, CR, BL and FS provided useful comments and participated in the redaction of the paper.

520
Data availability The airborne data will be available from the HALO database at https://halo-db.   detected by MAS as a function of relative humidity over ice (RHi) and potential temperature relative to the local CPT level. The presence of ice crystals in subsaturated air (RHi<100%) above the cold point tropopause potentially leads to permanent hydration of the CLS, whereas in supersaturated air the ice crystals are expected to sediment out of the CLS, thereby leading to its dehydration.