Figure 3 shows the ionic concentration of ground samples (GRND) as well as those of airborne samples (ZF2 and ZF3) collected aboard zero-pressure flights on 15 and 21 August 2017. Additionally, the figure also shows the ionic concentration of the flight held during the winter 2018 campaign denoted as (ZFW). In GRND samples, Na+ and Ca2+ cations are seen on S-1 and S-2 with corresponding anions (NO3-, SO42-, and NO2-) coexisting at the same stage. High NH4+ is observed only on S-3, with a concentration of 212 ng m-3 at STP. K+ was also seen on S-3, with a concentration of 26 ng m-3 at STP (fine mode), and could have originated from biomass burning. City pollution from Hyderabad is likely the source of those aerosols observed on the GRND filters. Flights ZF2 and ZF3 show significant amounts of NO3- and NO2- (87–343 ng m-3 at STP) with trace amounts of mineral dust proxies (Ca2+). Biomass burning (K+) was observed in the results of flight ZF2 only. The presence of non-sea-salt Ca2+ in aerosols is often used as a proxy for mineral dust (Schüpbach et al., 2013), whereas non-sea-salt K+ in aerosols is often used as a proxy for biomass burning (Li et al., 2003). Although their concentrations were too low (close to the detection limit), their presence indicates the possibility of trace contributions from mineral dust and biomass burning.

Other species were below 5 (for cations) to 10 (for anions) ng m-3 at STP, the detection limit of the IC instrument for our analytical setup. Charge balance was not achieved due to a higher negative charge, mainly from NO3- and NO2-, than positive charge, mainly from NH4+, Ca2+, and K+ (Fig. 3), implying the existence of NO3- and NO2- in forms other than salt. For instance, nitric acid trihydrate (NAT, HNO3 ⋅ 3H2O) could be another aerosol cluster in which NO3- may be present in the tropical UTLS (Voigt et al., 2000). We did not find a significant amount of ammonium in our ZF flight samples during the summer. Overall, the concentration of nitrate (80–100 ng m-3 at STP) found on both flights seems to be lower than the levels observed during StratoClim (Höpfner et al., 2019). In the only flight during the winter of 2018, Na+ and K+ were almost inexistent. In comparison, the mineral dust proxy (Ca2+) was present in trace amounts on all four impactor stages and could be associated with SO42-, which was also found on all four stages (Fig. 3b).

Balloon-borne and aircraft sampling techniques have been used since the early 1970s to study the composition of aerosols in the UTLS region (Lazarus et al., 1970). Höpfner et al. (2016) reported the presence of enhanced amounts of NO3 between 12 and 15 km within the Asian summer monsoon region. While sulfate tends to be stable enough to be collected and further analyzed without major chemical transformation, other nitrate-containing particles can be more unstable. NO3- salts apart from NH4NO3 are not significantly volatile after sampling (Newman, 1993). The dissociation of NH4NO3 into gas-phase HNO3 and NH3 increases sharply with increasing temperature and relative humidity (Stelson and Steinfield, 1982; Lightstone et al., 2000), leading to a significant loss of particulate nitrate (PN). The slight retention of HNO3 (gas) on the PTFE (polytetrafluoroethylene) filter could represent a significant source of particulate nitrate on filters at low concentrations and has been used in the past to estimate stratospheric HNO3 (Lazarus et al., 1970). Additional information available during ZF2 will be discussed to assess the presence of ice clouds.

We will now focus the discussion on ZF2, which included a COBALD backscatter sonde and was launched to be co-located in space and time (within 20 km and 1 h, respectively) with satellite observations from the CALIOP lidar aboard the CALIPSO satellite. Figure 4a and b show scattering ratio (SR) and color index (CI) profiles from COBALD (470 and 940 nm) as well as CALIOP SR and volume depolarization profiles at 532 nm. Both balloon and satellite observations show a layer between 13.5 and 16 km with high depolarization (CALIOP) and a high color ratio (COBALD), likely made of aspherical particles. The derived particulate depolarization ratio from CALIOP Level 2v4.1 within the layer was 0.47+/-0.06 (Fig. S3 in the Supplement) and was associated with an optical depth of 0.03+/-0.02, indicating the presence of a subvisible cirrus cloud. Flight ZF2 floated near 14.5–17 km for more than 2 h (Fig. 1c). The time series (Fig. 4c) indicate that the measurements took place within two different air masses: first within an ice cloud, as discussed above, and then in a cloud-free region. The pump connected to the impactor was switched on below 150 hPa and run for ∼16 min within the cloud and for ∼1 h 30 min in a cloud-free region.

For the only flight during the winter of 2018, Na+ and K+ were almost inexistent. In comparison, the mineral dust proxy (Ca2+) was present on all four impactor stages: 30 ng m-3 on stage 1 (particle sizes >2 µm), 46 ng m-3 on stage 2 (particle sizes between 0.5 and 2 µm), 11 ng m-3 on stage 3 (particle sizes between 0.15 and 0.5 µm), and 29 ng m-3 on stage 4 (particle sizes between 0.05 and 0.15 µm). Interestingly, SO42- was also found on all four stages (Fig. 3b): 14 ng m-3 on stage 1, 21 ng m-3 on stage 2, 15 ng m-3 on stage 3, and 12 ng m-3 on stage 4. Satellite analysis of aerosol extinction at 1020 nm from the Stratospheric Aerosol and Gas Experiment III (SAGE III) was conducted to understand the origin of those particles. We found high aerosol extinction values in the Northern Hemisphere from August 2017 to February 2018, consistent with the presence of smoke from the 2017 Canadian fire (Fig. S2).

Fires that intensify towering thunderstorms have the tendency to inject aerosols into the lower stratosphere that were otherwise thought to have originated from volcanic plumes (Fromm et al., 2010). The 2017 Canadian wildfire event led to the formation of multiple pyrocumulonimbus (PyroCb) episodes, resulting in a vast aerosol cloud. Within a few weeks, a portion of this initial plume was transported by the polar jet streams across the Atlantic Ocean in the Northern Hemisphere (Peterson et al., 2018), causing a strong perturbation of the stratospheric aerosol loads (Stocker et al., 2021). The quantity of smoke injected was enormous – so much so that it was observed for more than 8 months (Yu et al., 2019. The presence of the resultant aerosol layer was pointed out by high ultraviolet aerosol index values and confirmed with CALIOP lidar observations in the UTLS (Torres et al., 2020). The aerosol mass increase and subsequent adiabatic aerosol self-lofting as a result of absorption of solar radiation were also observed by the Earth Polychromatic Imaging Camera (EPIC) sensor aboard the Deep Space Climate Observatory (DSCOVR) satellite. Kloss et al. (2019) used SAGE III aerosol extinction values to show that the fire plume was transported within the AMA circulation in August 2017. Our analysis suggests that the smoke plume was still present at 18 km above Hyderabad between January and February 2018, indicating that aerosols sampled during the winter flight were influenced by this smoke plume.

Deep convection, emanating from Southeast Asia, and maritime convection over surrounding seas serve as a conduit for the transport of boundary layer (BL) pollutants (CO, HCN, CH4) to the UTLS (Belikov et al., 2013; Park et al., 2007, 2006; Randel et al., 2010). Wind-driven physical processes lead to the accumulation of pollutants due to the limited exchanges of air between the interior and exterior of the Asian monsoon anticyclone (Fairlie et al., 2014, 2020; Ploeger et al., 2015).

To study the impact of convection on our measurements, we calculate back trajectories from ZF2 and ZF3 using the Langley Trajectory Model (LaTM; Fairlie et al., 2014) driven by winds from the NASA Global Modelling and Assimilation Office (GMAO) Goddard Earth Observing System, Version 5, Forward Processing (GEOS-5 FP) product (Lucchesi, 2018). We locate the intersection with anvils and deep convective clouds observed through cloud-top brightness temperature from the Himawari-8 satellite (Vernier et al., 2018). Figure 5 shows the position of those 5 d back trajectories (colored lines) and deep convective influences (black dots). Air sampled during ZF2 on 15 August 2017 traveled along two branches influenced by convection over southern/eastern China and western China, respectively. Air masses sampled by ZF3 originated from convection over Laos, Myanmar, the Bay of Bengal, and possibly local convection over the Indian eastern shore close to the measurement location.

The role of clouds in nitrite formation is further discussed in this section. To date, only a few nitrite measurements have been reported, mainly because of its low concentrations and the fact that nitrite ions are easily oxidized (Lammel and Cape, 1996). The first quantitative information on nitrite in cloud water was detected at Mt Tsukuba, Japan: values of 400–1050 µg L-1 with pH levels of 5.7–6.5 were reported. In contrast, acidic cloud water samples (pH of 3.4–4.3) collected at significantly higher altitudes showed low nitrite values (15–104 µg L-1) (Okita, 1968). Nitrite and nitrous acid are short-lived intermediates of reactive oxidized nitrogen in the atmosphere. There is limited information on nitrite measurements in the atmosphere (Lammel and Cape, 1996). Nitrite was also measured in fog water samples in a polluted region in Germany (Lammel and Metzig, 1998). Moreover, Bachmann et al. (1989) directly measured nitrite in rain and fog water samples using ion chromatography; respective values of 1.8 and 16 µmol L-1 were found. Photolysis of particulate nitrate, hydrolysis of NO2, and uptake of HNO2 by particles are the sources of particulate nitrite in the atmosphere (Chen et al., 2019).

HNO2 is an important precursor for nitrite formation, but there are challenges involved with making reliable HNO2 measurements at desired concentrations, leading to a lack of information about HNO2 in the troposphere. Firstly, due to its short lifetime (because of photolysis) and subsequent very low concentration from sources, HNO2 is a challenging species to observe. Secondly, HNO2, being sticky in nature, may be lost to the walls of sampling tubes or absorbed on filters. Thus, as nitrite is present in very low concentrations and is easily oxidized.

Intensive agricultural activities have led to the Indo-Gangetic Plain having the highest ammonia (NH3) loading worldwide (Wang et al., 2020), as revealed by satellite observations (Van Damme et al., 2018; Warner et al., 2016) and ground-based measurements (Carmichael et al., 2003). Dentener and Crutzen (1994) reported ammonia (NH3) emissions related to DMS (dimethylsulfide). Höpfner et al. (2016) reported the first detection of ammonia in the ATAL. Nitrite and nitrate are formed by the oxidation of NH3 through the process of nitrification [NH3 + O2 → NO2- + 3H- + 2e-]. In addition, the existence of NH3 in the presence of nitrate leads to the formation of ammonium nitrate which could neutralize aerosol particles and favor the persistence of nitrite, as revealed by a few existing measurements in the polluted region (Lammel and Metzig, 1998). The StratoClim campaign also revealed the presence of ammonium nitrate in the UTLS which would confirm that neutralization of nitrate is effective at high altitudes and may explain the persistence of nitrite found with our balloon measurements.

We conducted GEOS-Chem model simulations to put our observations in the context of the large-scale transport and distribution of atmospheric composition. GEOS-Chem is a state-of-the-art global 3-D chemical transport model that includes fully coupled ozone–NOx–volatile organic compound–aerosol chemistry for both the troposphere and stratosphere (Bey et al., 2001; Park et al., 2004; Eastham et al., 2014). Here, we use the model version 11-01 (http://wiki.seas.harvard.edu/geos-chem/index.php/GEOS-Chem_v11-01, last access: 29 July 2022). A previous version of the model was used to study the origins of aerosols in the ATAL by Gu et al. (2016) and Fairlie et al. (2020). The model simulates black carbon (Park et al., 2003), primary and secondary organic aerosols (POAs and SOAs, respectively; Pye et al., 2010), sulfate–nitrate–ammonium–aerosol thermodynamics coupled to ozone–NOx–hydrocarbon–aerosol chemistry (Park et al., 2004), mineral dust (Fairlie et al., 2007; Ridley et al., 2014), and sea salt (Jaeglé et al., 2011), treated as an external mixture. SOA simulation uses the volatility-based scheme (VBS) of Pye et al. (2010). Sulfate–nitrate–ammonium thermodynamics is computed using the ISORROPIA II thermodynamic equilibrium model of Fountoukis and Nenes (2007). ISORROPIA has been evaluated with in situ measurements during several campaigns in the United States (Guo et al., 2016, 2020) and, more recently, in Korea (Ibikunle et al., 2020). Aerosol wet deposition includes rainout and washout due to large-scale precipitation as well as scavenging in convective updrafts (Liu et al., 2001). Scavenging of aerosols by snow and mixed precipitation is described by Wang et al. (2011) and Q. Wang et al. (2014). Dry deposition of dust and sea-salt aerosols uses the size-dependent scheme of Zhang et al. (2011). Dry deposition for other aerosols follows the resistance-in-series scheme of Wesely (1989). Anthropogenic emissions use the Emissions Database for Global Atmospheric Research (EDGAR; Olivier and Berdowski, 2001) with regional options, including the MIX inventory over East Asia (Li et al., 2014) and the EPA/NEI 2011 inventory over North America (Travis et al., 2016). Biofuel emissions are from Yevich and Logan (2003). Carbonaceous aerosol emissions are provided by Bond et al. (2007). Biogenic emissions are calculated by the Model of Emissions of Gases and Aerosols from Nature (MEGAN; Guenther et al., 2012). Biomass burning emissions use the Quick Fire Emissions Dataset (QFED; Darmenov and da Silva, 2015). Lightning NOx emissions (LNOx) are as described by Murray et al. (2012) and match the Lightning Imaging Sensor and the Optical Transient Detector (LIS/OTD) climatological observations of lightning flashes.  Volcanic SO2 emissions are provided by the AeroCom project (data available from http://wiki.seas.harvard.edu/geos-chem/index.php/Volcanic_SO2_emission, last access: 29 July 2022). The model simulations are driven by the Modern-Era Retrospective analysis for Research and Applications (MERRA-2) reanalysis from the NASA GMAO (Gelaro et al., 2017). For computational efficiency, MERRA-2 fields have been mapped from the native grid to a 2.5∘ × 2∘ (longitude × latitude) horizontal resolution for input to GEOS-Chem. Further, we used the simulations with and without lightning NOx emissions to understand the contribution of lightning to the formation of nitrate aerosol.

In situ chemical analysis are compared with GEOS-Chem simulations. Figure 6 shows the maps of CO, nitrate, sulfate, ammonium, black carbon (BC), and dust aerosol concentrations averaged over 100–150 hPa at 22:00 UTC for 15 and 21 August during the ZF2 and ZF3 flights, respectively (white circle on the map). On 15 August, CO, BC, nitrate, ammonium, and dust aerosol concentrations are enhanced over West China, Nepal, and northeastern India, with the center of the anticyclone positioned over West China. On the contrary, on 21 August during ZF3, the position of the anticyclone was shifted to the east, and the flight apparently sampled air at the edge of the anticyclone. The simulation shows a 20 % and 50 % increase in CO and BC on 21 August, respectively, compared with 15 August. Additionally, the ammonium concentration was decreased by 50 % and dust was reduced by 60 % on 21 August compared with the first flight on 15 August. However, the SO42- concentration is seen to be stable at ∼80 ng m-3 for both 15 and 21 August above Hyderabad. The simulated NO3 concentrations near the location of ZF2 and ZF3 are spatially inhomogeneous, with variations between 30 and 2700 ng m3 across South India. Figure 7 shows the time series of model 3-hourly CO, sulfate, and nitrate concentrations averaged over 100–150 hPa within the model grid point where Hyderabad is located during August 2017. The CO concentration shows a decrease of 14 %, while an increase of 21 % in the SO4 concentration is seen in the ZF3 flight that took place on 21 August. The measured nitrate concentration during ZF2 and ZF3 (around ∼100 ng m-3) is within the range of values simulated within 24 h of the observations. The results of the GEOS-Chem model simulation indicate that lightning NOx could significantly contribute (up to ∼75 % on 10 August) to the formation of nitrates during certain time periods. The lifetime of NOx is approximately 3 h in the region of the outflow of thunderstorms due to the production of methyl proxy nitrate and alkyl, and multifunctional nitrate species and their lifetime is believed to increase downwind of the outflow (Nault et al., 2017). Also shown in Fig. 7 are nitrate concentrations attributed to lightning, as determined by the difference between simulations with and without lightning NOx emissions. ZF2 and ZF3 occurred during a period where the levels of nitrate were ≈50 ng m-3 at STP on 15 August and ≈30 ng m3 at STP on 21 August. The influence of lightning NOx emissions was minimal during this time. Nevertheless, CO levels are slightly higher during ZF2 (80 ppbv) than during ZF3 (60 ppbv), indicating that measurements made during ZF2 may have been more influenced by pollution. The latter is also reflected by the higher BC levels during ZF2 in the model.

We extracted CO, nitrate, and sulfate concentrations from the GEOS-Chem simulation along the calculated trajectories initialized from the ZF2 and ZF3 measurement locations in Fig. 8. The lines are colored according to the balloon GPS altitudes that are used to initialize the trajectory model. Figure 5 uses GEOS-FP winds (meteorology) to convey that GEOS-Chem could simulate convective activities reaching levels between 14 and 15 km. This is confirmed by cloud-top heights (black circle) derived from Himawari-8 crossed by trajectories originating from the troposphere for both ZF2 and ZF3. ZF2 was influenced by convective activities over Western China, whereas ZF3 sampled air masses originating from convection in Southeastern Asia (Myanmar, Laos). CO levels with initial altitudes near 14–15 km (green color) for ZF2 are shown to decrease from 120 to 80 ppbv along the back trajectories, confirming the influence of Chinese pollution and its progressive dilution. At the same initial altitudes, the CO levels along ZF3 back trajectories are significantly lower (near 50–80 ppbv), possibly indicating minimal impacts from polluted sources. The levels of NO3- show significant variability along the trajectories for both cases but are more pronounced in ZF3, with levels above 400 ng m3, again emphasizing the likely importance of LNOx in the production of nitrate aerosols.

Sulfate concentrations are much higher (100–200 ng m3 at STP) for air parcels initialized near 16–17 km for ZF2 and ZF3, likely indicating stratospheric sources, whereas air parcels near 14–15 km show levels below 100 ng m3 at STP. We note that sulfate along the trajectories influenced by Chinese pollution during ZF2 increases by 60 % approximately 50 h before our measurements, which could indicate the formation of sulfate aerosol from SO2. It has previously been reported that sulfate has a lifetime of a few days in the troposphere (Hidy and Blanchard, 2016). The rather short lifetime of sulfate is due to absorption in precipitation or its solubility (Hidy, 1973). The global mean residence time of tropospheric sulfate against dry and wet deposition is approximately a few days (e.g., Park et al., 2004).

The GEOS-Chem model showed higher sulfate levels than the results from IC due to relatively weak scavenging of SO2 and/or SO42-.

The aircraft field campaigns of the StratoClim project were held in July and August 2017 at the Tribhuvan International Airport (KTM; 27.70∘ N, 85.36∘ E; Katmandu, Nepal). In situ aerosol measurements within the AMA were carried out using the ERICA (European Research Council Instrument for Chemical composition of Aerosols) aerosol mass spectrometer. Flights KTM 01 and KTM 02 held on 27 and 29 July showed a low level of sulfate at 360 K (potential temperature), corresponding to an altitude of 15 km. The sulfate concentration was almost equal to zero on 10 August during flight KTM 08 at 16 km and 370 K (Stephan Borrmann, 4th ACAM workshop, 28 June 2019). The very low levels of sulfate sometimes observed in the StratoClim campaign near 360–380 K are consistent with our IC analysis results with respect to the sulfate ionic concentration during flight ZF2 held on 15 August at the same altitude and potential temperature.

The chemical composition of the ATAL has been investigated using offline IC analysis of aerosol-impacted samples collected aboard zero-pressure balloon flights as part of the BATAL campaigns. The measurements of the 2017 summer campaign indicate the dominant presence of nitrate and nitrite aerosols with concentrations between 88 and 374 ng m-3 at STP. Our first flight (ZF2) on 15 August 2017 occurred within the AMA and, thus, sampled air masses therein. In situ measurements revealed the presence of NO3-, and NO2- aerosols (60–200 ng m-3 at STP) with sizes ranging from 0.05 to 2 µm. The second flight (ZF3) on 21 August 2017, however, occurred at the edge of the anticyclone, and subsequent in situ measurements revealed the presence of larger-particle-size NO3- and NO2- aerosols at higher concentrations (87.3–343 ng m-3 at STP). Throughout the flights during the 2017 summer campaign, sulfate aerosol remained below the detection limit of the system (10 ng m-3 at STP) and were much lower than the results from the GEOS-Chem model simulation (80–120 ng m-3 at STP). The higher model sulfate levels compared with the values from IC are believed to be due to relatively weak scavenging of SO2 and/or SO42- in the model. Unlike the summer, Ca2+ and SO42- were found on all four stages (sizes ranging between 0.2 and 0.05 µm) along with traces of NH4+, which could not be quantified in the winter campaign. The winter flight sampled residuals from the 2017 Canadian wildfires affected stratospheric aerosol loadings for several months.

We study the influence of convection on those measurements using back-trajectory calculations co-located with geostationary satellite observations. We show that ZF2 and ZF3 were influenced by convection over Western China, the Bay of Bengal, and Myanmar, Thailand, and Laos. The model was able to reproduce the convective transport from the mid-troposphere (9–12 km) to the upper troposphere (14–15 km). There was no indication of the transport of these air parcels from the boundary layer. Although Himawari-8 observations showed the convective transport reproduced in MERRA-2, the mixture between horizontal and vertical transport was not visible in trajectory calculations. Tropical convection could explain the rapid ascent of the air parcels to higher altitudes because other mechanisms, namely radiative heating, would delay the transport of air parcels from the middle to the upper troposphere. While the model seems to represent convection in the upper troposphere (14–15 km), with the rapid ascension of air parcels, the model's ability to simulate convective influence at higher altitudes seems to be limited.

We used the GEOS-Chem model simulations with and without lightning NOx emissions to understand the contribution of lightning to nitrate aerosol. Flights ZF2 (15 August) and ZF3 (21 August) occurred during a period when the levels of nitrate were relatively small (<100 ng m-3 at STP) with minimal influence of lightning NOx, in contrast with other periods largely affected by nitrate produced by LNOx. As shown by trajectory calculations in Fig. 5, flights ZF2 and ZF3 sampled air masses localized at the border of the Asian anticyclone. Fairlie et al. (2020) showed that the eastern part of the ATAL anticyclone depicts a peak of ammonium contribution from Chinese emissions. The western core of the ATAL, on the other hand, is seen to be enriched with 80 % of anthropogenic sources from India, with the southern and eastern flanks of the anticyclone showing peaks of Chinese contribution wherein nitrate concentrations were found to be the highest.

As the ASM (Asian summer monsoon) varies with respect to spatial dimensions and methodology, inconsistencies in the seasonal and interannual contribution to the ATAL are expected. Mineral dust is considered to be the most abundant type in the troposphere, with its main emission source being from arid and semiarid regions (Huneeus et al., 2011). CaCO3 is considered to be one of the most important components of mineral dust, and about 1.3 Tg of CaCO3 is loaded in the troposphere, corresponding to approximately 8 % of the total loading of mineral dust (Scanza et al., 2015). During atmospheric transport, heterogeneous reactions occur with trace gases, forming more soluble species and resulting in the increased CCN (cloud condensation nuclei) activity of mineral dust particles. Flight ZF2 sampled air masses within a cloud showing the presence of Ca2+ and NO3- on the same stage (particle size <0.15 µm). This implies the formation of Ca(NO3)2 in the presence of HNO3.

Indeed, the atmosphere is an amalgamated den in which gaseous species, particulates, and liquid droplets coexist at the same time. Through our balloon campaigns during the ASM with simultaneous offline measurements of inorganic species and, subsequently, comparing these results with model simulations, we were able to understand, if not fully answer, the many unanswered questions regarding the existence and behavioral patterns of these ionic species of interest. We will continue to research this area with improved techniques and additional experimentation.