Mesoscale simulations of tropical cyclone Enawo (March 2017) and its impact on TTL water vapor

In early March 2017, tropical cyclone (TC) Enawo formed north of Réunion Island and moved westward toward Madagascar. Enawo evolved from a tropical depression on 2 March to an intense TC on 6 March. 15 This study explores the water vapor transport into the tropical tropopause layer (TTL) throughout TC Enawo’s development. High-resolution (2km) mesoscale simulations using the Meso-NH model were performed to cover TC Enawo’s lifecycle over the ocean for the period 2-7 March 2017. The simulated convective cloud field agrees with geostationary satellite infrared observations. Compared to the Global Precipitation Measurements (GPM) and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellite observations, the 20 simulation seems to reproduce well both location and amplitude of the observed precipitation. Simulated and observed ice water content have similar ranges in the upper troposphere but simulated ice above the tropopause is overestimated by a factor 10. Balloon-borne measurements of water vapor, temperature and horizontal winds are also used to validate the Meso-NH simulations in the upper-troposphere and TTL regions. The simulations reveal that the maximum water vapor transport into the TTL occurred on 4 March, when deep (cold) convective 25 clouds were observed. As a result, the lower stratospheric water vapor is increased by ~50% when compared to pre-storm conditions. An increase of ~2ppmv in water vapor mixing ratio was simulated in the lower stratosphere within a 700-km region surrounding Enawo’s center. Our simulation of TC Enawo suggests that TCs over the Southwest Indian Ocean (0-30°S, 30-90°E) could produce a moistening of 0.4 ppmv. We extended our results to the global tropics (30°S-30°N) using the estimates from published work (Allison et al., 2018; Preston et al., 2019) 30 and by calculating statistics on TC numbers and durations using the International Best Track Archive for Climate Stewardship (IBTrACS) dataset. We estimated a global impact of TC induced tropical lower stratospheric moistening of 0.3 to 0.5 ppmv. Our results suggest that TCs may play an important role in the moistening of the TTL/lower stratosphere via direct injection of ice particles and subsequent sublimation.


35
Water vapor in the low stratosphere regulates up to 10% of the greenhouse effect at the earth's surface (Solomon et al., 2010). The stratospheric water vapor concentration is primarily affected by the freeze-drying at the tropical tropopause (Brewer, 1949) and the production of water vapor by methane oxidation (le Texier et al., 1988). It has https://doi.org/10.5194/acp-2020-870 Preprint. Discussion started: 21 October 2020 c Author(s) 2020. CC BY 4.0 License. Saffir Simpson Scale) in the Gulf of Mexico. They identified dehydration between 14.5 and 17.5 km due to ice sedimentation and hydration between ~17.5 and 21 km due to the sublimation of ice crystals.
In addition, Preston et al. (2019) investigated transport of O3/CO/water vapor by typhoon Mireille (a category 3 80 storm) over the Western North Pacific using the WRF model with chemistry (WRF-Chem). Their high-resolution (3km) simulations showed positive vertical fluxes throughout the troposphere and the tropopause, which increased the CO/water vapor concentration in the UTLS region. Despite having different intensities (category 1 versus category 3 storm), overshooting tops were identified for both systems that transported large quantities of water vapor to the UTLS. Both studies also agree in terms of estimation of net water vapor flux within 50%.

85
The current study focuses on an intense TC that occurred over the SWIO in March 2017. (Tao and Jiang, 2013) For the period 1979-2008, an average of ~90 TCs having maximum sustained winds ≥ 63 km/h occurred each year globally (cf. Table 2.5 WMO 2017). The most active TC basin is the Western North Pacific (WNP, ~30% of the global total), followed by the Eastern North Pacific (ENP, ~19% of the global total), the Southwest Indian Ocean (SWIO, ~15% of the global total) and the North Atlantic (NA, ~13% of the global total). The SWIO has 90 been poorly studied so far despite having tropical-cyclone activity that is comparable to that of the NA. Using 11- year Tropical Rainfall Measuring Mission (TRMM) precipitating hydrometeor satellite observations, Tao and Jiang, (2013) identified overshooting tops in tropical cyclones (above 14 km) and showed that the South Indian Ocean is the second basin after the Northwest Pacific in terms of total number of overshooting tops (cf. Table 2 of Tao and Jiang, 2013).

95
The event analyzed is the intense TC Enawo (category 3 on the US Saffir Simpson Scale), that made landfall in northeastern Madagascar on 7 March 2017, killing more than 80 people and causing extensive damage. In a previous study, Evan et al. (2020) investigated the effect of deep convection on the TTL composition over the SWIO during austral summer (December-March). They focused on the origin of convective signatures in two balloon-borne water vapor profiles observed in the vicinity of tropical storms in January 2016 (Tropical Storm 100 Corentin) and March 2017 (TC Enawo). The balloon-borne measurements were made in coordination with lidar observation at the Maïdo Observatory on Réunion Island (21°S, 55°). Using lagrangian backtrajectories, convective activity in both tropical storms was shown to produce significant hydration in the upper tropos phere (UT). In contrast, no water vapor anomaly was found near or above the tropopause region on 3 March 2017 over Réunion Island as the tropopause region was not downwind of TC Enawo. In addition, the balloon was launched 105 at a distance of ~1000 km from Enawo when the storm was still intensifying. Results from Allison et al. (2018) and Preston et al. (2019) suggest that overshooting convection and subsequent water vapor transport to the lower stratosphere (LS) mostly occurs in the eyewall region of TCs. Here, we extend the Evan et al. (2020) analysis of TC Enawo using a mesoscale model to quantify the impact of TC Enawo on UTLS water vapor.
With a background ranging from 2 ppmv to 6 ppmv (Fueglistaler et al., 2009;Rosenlof et al., 2001), the water 110 vapor content in the TTL is very sensitive to small variations. Most GCMs with coarse horizontal and vertical grid spacings underestimate convective transport and its effect on UTLS water vapor. High-resolution modeling allows a better representation of deep convection within TCs and is therefore useful to understand t he relative roles of vertical transport of and ice microphysics on UTLS water vapor (Allison et al., 2018;Chaboureau et al., 2007;Dauhut et al., 2015;Frey et al., 2015;Mrowiec et al., 2012;Ravindra Babu et al., 2015). The main goal of 115 this study is to further our understanding of the impact of TC on the TTL by simulating the TC Enawo over the SWIO basin, for the period 2-7 March 2017 when the storm was intensifying from tropical storm to intense tropical https://doi.org/10.5194/acp-2020-870 Preprint. Discussion started: 21 October 2020 c Author(s) 2020. CC BY 4.0 License. cyclone. over the SWIO. We take advantage of the Meso-NH model that has been previously developed for the simulations of TCs in the SWIO (Hoarau et al., 2018;Pianezze et al., 2018). This study is also part of the CONCIRTO (CONvection CIRrus over the Tropical indian Ocean) project that aims to further our knowledge on 120 how deep convection and cirrus clouds affect the TTL over the SWIO.
The present paper is organized as follows. Section 2 describes the Meso-NH model setup as well as balloon-borne in situ measurement and satellite observations used to evaluate the model representation of TC Enawo. Section 3 presents an overview of the evolution of TC Enawo in both observations and simulations. Section 4 investigates simulated water and ice transport to the low stratosphere. The global impact of TCs to the tropical lower 125 stratospheric water vapor is assessed in Section 5. The model results are discussed in Section 6. Finally, Section 7 contains the conclusions and a summary of our study.

Satellite observations
METEOSAT 8 is a geostationary satellite located at 41.1°E that monitors for the Indian Ocean since March 2017. The Global Precipitation Measurement Integrated Multi-satellitE Retrievals (GPM-IMERG) product from the National Aeronautics and Space Administration (NASA) is used to evaluate simulated rainfall. It uses an algorithm that merges precipitation radar, microwave precipitation estimates, microwave-calibrated infrared, and rain gauge analyses at a spatial resolution of 0.1• over the latitudinal belt 60°N-60°S (Huffman et al., 2018). The contain extinction and Ice Water Content (IWC) profiles at 60-m vertical and 5-km horizontal resolutions, which are retrieved from the 532-nm extinction coefficient (Avery et al., 2012). Between March 2 and 7, there were 9 CALIPSO overpasses over TC Enawo. However, only the overpass on 5 March at 21:30 UTC was over TC Enawo's eye region (cf. Figure 7).

150
A balloon launch was specifically planned using a Lagrangian model and geostationary infrared images to sample the convective outflow from TC Enawo on 3 March 2017 at 18 UTC. The balloon sonde payload consisted of the Cryogenic Frostpoint Hygrometer (CFH) as well as the Intermet iMet-1-RSB and Meteomodem M10 meteorological radiosondes. A detailed description of the balloon measurements is provided in Evan et al. (2020). https://doi.org/10.5194/acp-2020-870 Preprint. Discussion started: 21 October 2020 c Author(s) 2020. CC BY 4.0 License.
The Cryogenic Frost Point Hygrometer (CFH) is an in-situ instrument that measures the water vapor mixing ratio 155 profile from the surface to the stratosphere (~28km).
The CFH was developed to provide highly accurate water vapor measurements in the TTL and stratosphere where the water vapor mixing ratios are extremely low (~2 ppmv). CFH mixing ratio measurement uncertainty ranges from 5% in the tropical lower troposphere to less than 10% in the stratosphere (Vömel et al., 2007) ; a recent study shows that the uncertainty in the stratosphere can be as low as 2-3% (Vömel et al., 2016). The M10 radiosonde 160 provides measurements of Relative Humidity (RH), temperature, pressure, vertical velocity, wind speed and direction from which zonal/meridional winds are derived.
Different simulations (Table 1), were performed with the Meso-NH model. The simulations have vertical grid spacings less than 100 m in the boundary layer and 300m up to 30km (140 verticals levels), with a damping layer 170 in the uppermost 25 km and a model top at 1hPa. All simulations are initialized and forced at the boundaries with 6-hourly analyses from the operational European Centre for Medium Range Weather Forecasts -Integrated Forecast System (ECMWF-IFS) with a grid spacing of ~9km and 137 vertical levels. Simulations were run from 2 March 2017, 00UTC to 7 March, 00UTC to encompass Enawo's evolution from tropical depression to very intense TC just before landfall over Madagascar. The turbulent scheme is based on Cuxart et al. (2000). We use 175 a single-moment bulk mixed-phase cloud parameterization ICE3 (Lac et al., 2018;Pinty and Jabouille 1998). It solves the microphysics of five hydrometeors which are three precipitable species (snow, rain, and graupel) and two non-precipitating hydrometeors (cloud water, and cloud ice). The longwave radiative scheme used in Meso-NH is the rapid radiation transfer model (RRTM; Mlawer et al., 1997). The shortwave radiative scheme is based on Foucart and Bonnel (1980). The subgrid shallow convection is based on the eddy-diffusivity mass flux and

180
Kain-Fritsch approach of entrainment and detrainment in cumulus clouds (Pergaud et al., 2009). Meso-NH is coupled with the surface model SURFEX (Surface Externalisée, Masson et al., 2013) and uses COARE-3 parameterisation for ocean-atmosphere fluxes (Fairall et al., 2003). The first model level is at 10m. Initial simulations were run to test the model's sensitivity to different parameterizations on a domain A covering 46°E-69°E in longitude, 32°S-8°S in latitude at a horizontal grid-spacing of 10 km (Simulation S1). Tests on 185 domain limits have been performed to identify the sensitivity to the eastern boundary. Then, five simulations were performed (Table 1) to test the horizontal resolution, nested domains and sensitivity to SST. Simulations were run in domain A with a horizontal grid-spacing of 10km and 5km (Simulations S1 and S2). The Kain Fritsch cumulus parameterization was used to simulate deep convection in S1 and S2. A second domain, called domain B, centered over a region that encompasses Enawo's life cycle was designed ( Figure 1) for simulations with a horizontal grid-190 spacing of 2km (Simulations S3, S4 and S5). Deep convection was explicitly resolved at that resolution.
Simulation S3 had nested domains having grid spacings of 10 (domain A) and 2km (domain B). It was found that using one-way grid-nesting did not improve the model's representation of Enawo, so higher resolution simulations https://doi.org/10.5194/acp-2020-870 Preprint. Discussion started: 21 October 2020 c Author(s) 2020. CC BY 4.0 License. at 2km were run without nesting (Simulations S4, S5). The sensitivity to the Sea-Surface-Temperature (SST) was also tested. In simulations S1 to S4, the Sea-Surface-Temperature (SST) data from ECMWF were used and 195 updated every 6 hours. Simulation S5 is coupled with the oceanic model CROCO (Coastal and Regional Ocean COmmunity model, http://www.croco-ocean.org, (Debreu et al., 2012;Penven et al., 2006). Further details on the coupling are provided in Voldoire et al. (2017) and Pianezze et al. (2018). The oceanic model has a horizontal resolution of 2km and 32 vertical levels from the surface to the ocean bottom (about a depth of 5km below Enawo) and shares the same domain of simulation as the Meso-NH model. Between 2 and 4 March, Enawo intensified slowly from TD to TS because of a strong east-southeast vertical wind shear. On 4 March, Enawo stalled over the ocean while intensifying. Very cold cloud tops (<-90°C) were observed 215 at that time. Enawo evolved from TS to TC on 5 March at 06UTC with an eyewall with a poorly defined structure.
Later, the intensification speed was slowed by an Eyewall Replacement Cycle (ERC), observed on satellite microwave images. An ERC occurs when the pressure forces at the center of the TC push inward the eyewall. At some point, the eyewall collapses and a new eyewall forms afterward. Because of a decrease in vertical wind shear, Enawo further intensified to reach the ITC stage on 6 March at 12UTC (Category 2-3 on the Saffir-Simpson 220 scale), with 10-minute maximum sustained winds of 46.3 m s-1 (90kts). The 10-minute maximum sustained winds (VMAX) increased by 23 m.s -1 (45 kts) in 24 hours, larger than the top 5% percentile of rapid intensification over the SWIO basin (Leroux et al., 2018). Enawo reached peak intensity on 7 March at 06 UTC, with ten-minute maximum sustained winds at 56.5 m.s -1 (110 kts) and the central pressure at 932 hPa. TC Enawo reached Madagascar's northeastern coast on March 7 at around 09:30 UTC and was the third strongest tropical cyclone on 225 record to strike the island. It was also the strongest TC of the Southern Hemisphere for the 2016/2017 TC season.
The RSMC La Réunion provides 6-hourly best-track data that include the location of the storm center, minima of mean sea level pressure (MSLP) and 10-min maxima sustained wind. Root mean square errors (RMSE) between the Best Track and the simulations were calculated for track (in km) and MSLP (in hPa) following the approach of Chandrasekar and Balaji (2012) and Allison (2018). The RMSE for the track and MSLP for each simulation 230 are presented in Table 1. Based on the results of MSLP and RMSE, simulation S4 does a better job in representing https://doi.org/10.5194/acp-2020-870 Preprint. Discussion started: 21 October 2020 c Author(s) 2020. CC BY 4.0 License. the intensification phase and the trajectory of Enawo. The comparison of S4 to Enawo's best track, precipitation and balloon-borne measurements are presented in the subsequent sections. Simulation S4 will be used in section 4 to investigate the vertical structure of the TTL and water vapor/ice transport to the lower stratosphere.

Trajectory/intensification
235 Figure 1 compares the simulated trajectory of simulation S4 to the best-track data from March 2 at 06UTC (top right) and March 7 at 00 UTC (bottom left). The trajectory in the simulations is defined using a minimum of Mean Sea Level Pressure (MSPP). The simulated trajectory is to the south of the one provided in the best-track data but with a RMSE of 97±54km. After March 4, the simulated trajectory is even more to the south when compared to the best track. However, the westward propagation of the tropical cyclone is relatively well represented with 240 comparable propagation speed. On March 4, the simulation was able to show that Enawo stalled over the ocean near 13.6°S/56.5°E. In the rest of the study, we use the minimum MSLP to define the storm center. March). The simulation has a RMSE for MSLP less than 6hPa. Since no nudging was applied, it is remarkable 245 that the Meso-NH simulations can represent Enawo's initial development from tropical depression on 2 March, 00UTC to moderate tropical storm on 3 March, 18UTC. This is most likely due to the high-resolution operational ECMWF analyses used for initialization and the fact that the initial storm structure was present in the ECMWF analysis on 2 March 00UTC. When initialized a few days before, the Meso-NH did not generate the initial tropical depression that would become TC Enawo.

250
On 5 March 00UTC, Enawo stalled over the ocean after the ERC. The ERC was not reproduced by the simulation and the simulation started to diverge compared to Enawo's MSLP. An ERC is particularly difficult to simulate and a resolution less than 500 m is necessary to resolve this process (S. Bielli, personal communication). Between 60h and 96h, the intensification is slower in the simulation as indicated by the slower decrease in the simulated MSLP. After 96h (6 March, 00 UTC), the rapid intensification when Enawo evolves from TC to intense TC is 255 shown in the best-track data with a drop of 30 hPa in the MSLP over 12 h. The model is not able to capture this rapid intensification but instead simulates a steady intensification phase. As a consequence, the error in the MSLP increases to 20 hPa at the end of the simulation.

260
For this comparison, simulated precipitation was interpolated to the GPM grid (0.1°x0.1°) and integrated over the period 2 March, 06 UTC to 7 March, 00 UTC. Overall, the simulated total rainfall agrees with observations and Meso-NH captures the north-south asymmetry.
Between March 4 and March 5, heavy rain (in red on Figure 5) is observed in the eyewall. Precipitation for this period tends to be overestimated in the S4 simulation. Precipitation within the secondary rainbands are consistent 265 with those observed, especially south of Enawo's trajectory. The model tends to produce lighter precipitation over a larger region. In addition, simulated precipitation over the eastern coast of Madagascar is not observed in the GPM data. https://doi.org/10.5194/acp-2020-870 Preprint. Discussion started: 21 October 2020 c Author(s) 2020. CC BY 4.0 License.

Comparison to balloon-borne measurements.
We further compare the Meso-NH model results to balloon-borne measurements at the Maïdo Observatory on 3 The detrainment of water vapor at ~13 km is resolved in the simulation but remains lower than the observed one.
In the troposphere between 2 and 14 km in altitude, the simulation has a wet bias of +450 ppmv, resulting from a higher moisture transport with an explicit representation of convection at a 2 km grid-spacing. In the TTL (14-18km in altitude), the mean water vapor difference at the location of the Maïdo Observatory is about -1ppmv in 280 S4.
The simulated temperature profile is colder than the M10 temperature profile in the troposphere between 2 and 14 km with a mean bias of -1 K. A cold bias of -2 K is also observed in TTL (14-18 km). However, both the model and the observation indicate a CPT at ~16 km.
On Figure 4, the simulated and observed profiles of zonal and meridional wind speed are in good agreement. The 285 peak in meridional wind at 13km associated with Enawo's convective outflow in the UT is captured by the model.

315
The CALIOP lidar cannot penetrate deep convective clouds, therefore observed IWC is not shown below ~10 km.
From the distribution of IWC on Figure 8, we can see that the model tends to produce less ice in the convective core below ~15km in altitude, and more ice above. On the contrary, less ice is produced in the upper troposphere in Enawo's outer regions (south of 18.5°S and north of 11°S). In the upper troposphere, IWC values range from 10 -4 to 10 -2 g.m -3 in both the observation and the simulation. Maximum values of IWC are observed in the eyewall: 320 ~0.01 to 0.05 g.m -3 in the observation versus ~0.05 to 0.1 g m -3 in the simulation. Simulated IWC is a factor 10 larger than CALIPSO IWC above the CPT. The location of the convective center is fairly well reproduced by the simulation (13°S,55°E in S4 versus 14°S,54.7°E in the CALIPSO observation). The simulation does show ice above the CPT between 11 and 15°S (IWC ranging from 10 -5 to 10 -4 g.m -3 ). However, CALIOP only shows ice above the CPT in the eye region (according to the best-track data the storm center was located at 13.96°S, 55.02°E 325 on 5 March, 18 UTC).
Average mass (φmass), ice (φice) and water vapor (φvapor) flux density (kg.m -2 .s -1 ) to the low stratosphere (~18 km) can be computed as a function of time and distance from the TC center using the equations: ri corresponds to a model grid point, ρ is the air density (kg.m -3 ), qice is the ice mixing ratio (kg.kg -1 ), qvapor is the 335 water vapor mixing ratio (kg.kg -1 ) and w is the vertical velocity (m.s -1 ). Figure 7 illustrates the calculation of φmass and can be applied to φice and φvapor by adding the qice/ qvapor term. We sum the ρ×w term at individual grid point over ring circle regions around the TC center (red symbol on Figure 7) defined by radiuses (rk) varying from 20 to 700 km by 20 km increment (ie., rk+1-rk = 20 km) and divide by the number Nk of grid points in the ring circle region to define average φmass, φice, φvapor (kg.m -2 .s -1 ). Some studies have proposed to calculate upward flux using 340 only grid points containing upward vertical motion (Chaboureau et al., 2007;Dauhut et al., 2016;Mrowiec et al., 2012). Wei, (1987) has estimated mass transport across the tropopause by looking at the contribution from diabatic processes, the temporal change of the tropopause potential temperature and mass exchange due to the potential temperature gradient along the tropopause (Ravindra Babu et al., 2015;Wei, 1987). At 11:55 UTC, vertical winds in the lower TTL troposphere (14-16.5 km) are relatively weak. The 380 K isentropic surface (black lines on Figure 9) is deformed, possibly due to eddies or convectively generated gravity waves. A dry layer can be observed below 380 K, which is linked with the cold point tropopause. Higher values 365 of water vapor mixing ratio ~20 ppmv above the tropopause could be due to a precedent overshooting event within Enawo.
At 12:05 UTC, a stronger updraft develops near 13.5°S which further deforms the 380 K isentropic surface. Direct ice injection can be seen between 17 km and 19 km (ice mixing ratio of 10 ppmv to 1000 ppmv at 18 km) and dehydration below the tropopause between 13.4°S to 13.6°S becomes more important.

370
At 12:15 UTC, stronger upward motion and tropopause's deformation are still observed. The water vapor mixing ratio above 380K increases from 15 ppmv to more than 45 ppmv.
10 minutes later, at 1225 UTC ice mixing ratio values decrease by a factor of 100 in the l ower stratosphere (17-19km, 13.4°S-13.6°S) while water vapor mixing ratios increase by a factor 4 compared to the initial condition (40 versus 10 ppmv initially). This suggests that ice was transported in the lower stratosphere and sublimated in a 375 subsaturated environment thereby hydrating the LS (the full processes are described at high temporal resolution in Dauhut et al., 2018). The 380 K potential temperature surface between 13.4°S and 13.6°S is smoother and returns to its initial altitude (17km During the overshooting event on 4 March, the average LS water vapor mixing ratio increased from 6.1 to 7.4 ppmv around the TC center. It suggests transport of ice/water vapor from the troposphere. Ice sublimation in a sub-saturated LS leads to moistening. These results are further discussed in the discussion section.

Water vapor in TTL
We estimated the temporal change in TTL water vapor mixing ratio between the beginning and end of the Meso-NH simulations. Water vapor profiles were averaged over a 500 km region surrounding Enawo on 2 March 06 UTC (6 hours after the start of the simulation) and 7 March 00 UTC (end of the simulation). The average water vapor mixing ratio difference between these two dates is shown on Figure 10 for the Meso-NH simulation. A  vapor positive anomaly propagates outward Enawo's center, with water vapor mixing ratio of 5 ppmv transported 500 km away from Enawo's storm after 1.5 day of transport. At the end of the simulation, the water vapor mixing ratio background rose from 3.3 ppmv to 4.8 ppmv, an increase of 45%, 500 km away from Enawo's center.

Global impact of TCs to the tropical lower stratospheric water vapor
We compute a net vertical water vapor flux Fmean (t.hr -1 , where t corresponds to the metric ton) at 18 km which  Leroux et al. (2018). We also estimated an average duration of 74±49 hours (± refers to the standard deviation) per cyclone. Using an average net water vapor flux of 2.7x103 t.hr -1 from Enawo's high-

445
resolution simulation, we estimate that TCs over the SWIO could transport 9.6x105 t of water vapor into the lower stratosphere. Figure 10 shows that Enawo moistened the region between ~18 and 20 km. Dauhut et al. (2015) found moistening associated with a Hector thunderstorm between potential temperature levels 380-420 K (∼17-18.5 km). Using CALIOP data, Avery et al. (2017) reported that convective ice could be observed up to up to 2 km above the tropopause over the Central Eastern Pacific (Figure 2a, 2b  compute their mean duration as the period during which VMAX ≥ 64 kt. We also consider VMAX at peak intensity. The results for individual basins and the global tropics are summarized in Table 3. Number of TCs shown in Table 3 can be compared to Table 2.5 of WMO (2017) and Table 3   Therefore, our estimate of global TC tropical lower stratospheric moistening of 0.3 to 0.5 ppmv agrees with previous studies that have considered deep convection on a regional scale.
The Asian Monsoon in the Northern Hemisphere is clearly the most important source region but our analysis suggests that the most intense TCs may contribute to the tropical lower stratospheric water vapor budget, 510 especially in the Southern Hemisphere where there is no similar monsoon circulation. Because climate models and observations predict more intense TC with warming sea surface temperatures (Elsner et al., 2008;Kossin et al., 2020;Sobel et al., 2016), accurate understanding of the stratospheric impact of TC convection is critical in the context of global warming. This is important not only for the lower stratospheric water vapor budget but also for the transport of other chemical species (e.g. ozone, CO, ozone depleting substances) to the UTLS.

515
There are several limitations to the present study. First, we consider a single TC in the SWIO to assess the net TC especially for convective updrafts in the eyewall region, but would require more computer power.
Finally, we use a single-moment microphysics scheme in our 2-km simulation of TC Enawo. One conclusion from Allison et al. (2016) for TC Ingrid is that double-moment microphysics schemes produce more realistic tropical clouds and precipitation, which are important for the representations of updrafts and transport of ice to the TTL.
While our comparison to the GPM satellite observations suggests that the simulation seems to reproduce well both Indian Ocean) show that these storms moisten the lower stratosphere (83 hPa/~18 km) from 0.3 to 0.5 ppmv in a ~5 to 10° region around the eye.

Summary and Conclusions
We extended the analysis of Evan et al. (2020)      The present study 600-km Cat 3 TC Enawo in the SWIO 2 km grid-spacing, net flux estimated over 48 hours 2.7