The Earth System Model Validation (ESMVal) field experiment with the High Altitude and Long Range Research Aircraft (HALO, http://www.halo.dlr.de) was conducted during 10–24 September 2012 in close cooperation with the HALO TACTS mission (Jurkat et al., 2014; Vogel et al., 2015; Müller et al., 2016). During the 65 HALO flight hours of the ESMVal campaign, trace gas distributions were sampled from the ground to a maximum altitude of 15.3 km along the following route: Oberpfaffenhofen (Germany), Sal (Cape Verde), Cape Town (South Africa), boundary of Antarctica, Cape Town, Male (Maldives), Larnaca (Cyprus), Oberpfaffenhofen, Spitzbergen (Norway) and again to Oberpfaffenhofen. The goal was to gather in situ observations for the evaluation of Earth system models and to improve process understanding. Specific areas of interest included regions impacted by deep convection, lightning and biomass burning in West- and South Africa, anthropogenic pollution in Europe and the Mediterranean, the northern and southern polar regions, and the northern African and Asian monsoons.

The Asian summer monsoon (also known as Indian or south-west monsoon), sensu stricto, is a prevailing sea breeze, lasting from June to September (Gettelman et al., 2004; Lawrence and Lelieveld, 2010). Different mechanisms may contribute to the formation of a conduit of rising air, centred over the southern Tibetan Plateau (Bergman et al., 2013) during Northern Hemisphere summer. A high-pressure area forms in the convective detrainment altitudes, sustaining a coherent anticyclone, centred at 200 to 100 hPa (Dunkerton, 1995; Randel and Park, 2006; Garny and Randel, 2016).

Polluted boundary layer air is entrained from throughout the region. It is effectively uplifted first in the narrow conduit to detrainment altitudes of about 200 hPa, later by large-scale upward motion at the eastern side of the anticyclone and then confined by the Asian summer monsoon anticyclone (ASMA) (Lelieveld et al., 2001; Li et al., 2005; Randel and Park, 2006; Park et al., 2007, 2008, 2009; Chen et al., 2012; Bergman et al., 2013; Vogel et al., 2015; Ploeger et al., 2015). As a consequence, trace gas mixing ratios within the anticyclone are mainly shifted towards lower-tropospheric values, e.g. relatively increased carbon monoxide (CO) (Li et al., 2005; Park et al., 2008) and decreased ozone (O3) (Randel and Park, 2006; Park et al., 2007, 2008; Kunze et al., 2010). The in situ measurements considered in our study also show enhanced CO mixing ratios in the ASMA, but instead of decreased O3 we found significantly increased O3 mixing ratios – relative to the UT air encountered south of the anticyclone.

While several studies looked into the boundary layer sources for ASMA air, entrainment of stratospheric or TL air has received much less attention. We are not aware of a study focusing on it, although the possibility of stratospheric entrainment at the eastern flank of the ASMA has already been recognised (Plumb, 2005; Randel and Park, 2006; Ren et al., 2014). Park et al. (2007) found a relatively high frequency of TL air at 100 hPa (MLS satellite data, July–August 2005) at the eastern ASMA flank, and interpreted this as an indicator of frequent stratosphere–troposphere exchange. This is consistent with Konopka et al. (2010), who found that the ASMA enhances horizontal transport of O3-rich air from the extratropics into the stratospheric part of the tropical tropopause layer. Other studies (Cristofanelli et al., 2010; Barret et al., 2016) indicate the near absence of stratospheric intrusions during the monsoon season. At least the entrainment of O3-rich TL air is supported by the HALO ESMVal in situ observations considered here. Enhanced O3 was also found in CARIBIC (http://www.caribic-atmospheric.com) in situ measurements in the monsoon region. Trace gas signatures from the northern part of the ASMA were interpreted as photochemically older than those from the more central region, sampled at the southernmost parts of the CARIBIC flights (Baker et al., 2011; Rauthe-Schöch et al., 2016). However, the origin of enhanced O3 in the old air is not entirely clear, and Baker et al. (2011) also mentioned the possibility of stratospheric influences. Whether O3 in air originating in the ASMA is generally enhanced or depleted was pointed out as one of the major open questions related to the Asian monsoon already by Lawrence and Lelieveld (2010).

Here we present a unique set of upper troposphere (UT) in situ measurements in the ASMA, obtained during the HALO ESMVal campaign. The focus is on the measurements, their representation in a global chemistry climate simulation and the origin of air masses. We pinpoint the processes that led to the observed trace gas signatures using a more detailed analysis of an exemplary flight segment, and conclude showing how the present measurements could be reconciled with previous, seemingly contradictory studies. In a follow-up study the processes that determine the ASMA composition are further analysed by putting the HALO ESMVal measurements into a regional, seasonal and multi-annual perspective. We refer to that study (Gottschaldt et al., 2017) as accompanying paper in the following. The accompanying paper is mainly based on EMAC simulations, which also show that our in situ data reflect rather common processes in the ASMA. Both studies shall help to explain the highly variable composition of the ASMA and its outflow, addressing the following key aspects of the ASMA that were recently identified as poorly understood (Randel et al., 2016): dynamical and chemical coupling with convection, composition/reactive chemistry in the monsoon region, mixing of higher-latitude lower-stratospheric air into the tropical TL by the ASMA.

The paper is structured as follows: Sect. 2 provides a description of the instruments and techniques used for the in situ measurements of selected tracers during the ESMVal campaign, the Eulerian global chemistry climate simulation hindcasting the synoptic situation of the measurements, and the trajectory model. The transport pathways of air masses that contributed to the observed chemical composition and periods of interest from the measured time series are identified in Sect. 3. Section 4 shows that the EMAC global simulation may be used for the interpretation of the in situ measurements, because the main features are reproduced well. Section 5 is dedicated to the discussion of selected tracer–tracer relations in the in situ data. The eastern flank of the ASMA is found to be crucial for the generation of the observed trace gas signatures, which is discussed in Sect. 6 for that part of the flight providing the most direct observations of it. In Sect. 7 we reconcile the HALO ESMVal observations of increased O3 with previous studies that found decreased O3 in the ASMA, and then conclude with a summary in Sect. 8.

An air mass with enhanced mixing ratios of O3, CO, NO, NOy and HCl (Fig. 1) was sampled during a flight on 18 September 2012 from the Maldives to Cyprus over the Arabian Sea. Here, HALO flew at an altitude of 160 to 170 hPa, just before reaching the Oman coast (Fig. 2). We attribute the sudden increase of the above-mentioned trace gases at 07:46 UTC to entering the ASMA from the south. The sampling of this air mass was interrupted by a dive to probe the lower boundary of the ASMA, but after that HALO continued to fly in UT air related to the ASMA.

We divide the ESMVal-flight from Male to Larnaca pragmatically into seven parts (Table 1), called periods of interest (POI) in the following. Thereby we refer to the central region as the “interior”, and to the boundary region, i.e. the outer streamlines of the ASMA circulation, as the “fringe”. The terms “interior” and “fringe” characterise actual positions of streamlines within the anticyclone, independent of the trace gas signatures they carry. Streamlines represent an instantaneous snapshot of transport barriers, because there is no large-scale transport perpendicular to streamlines. Figure 3 illustrates the evolution of the ASMA circulation at flight altitude from 9 days before to 1 day after our measurements. Besides streamlines, Fig. 3 shows geopotential height, because increased geopotential is a proxy for characterising the extent of the ASMA on pressure levels (Barret et al., 2016). Alternatively, PV has been proposed to delineate the ASMA in isentropic coordinates (Ploeger et al., 2015) and the corresponding equivalent to Fig. 3 is provided in the Supplement S1. However, the above ASMA boundary definitions (Ploeger et al., 2015; Barret et al., 2016) and another one based on maximal wind speed, used by Pan et al. (2016), emphasise concepts of a closed ASMA volume or transport barriers on monthly or seasonal timescales. In contrast, our analyses rather aim to explain observed UT tracer distributions resulting from daily-scale dynamics. The latter is best captured by Lagrangian trajectories, which unlike the above Eulerian approaches inherently reflect the time dependence of the flow. Our choice of POIs and the delineation of the ASMA are thus based on back-trajectories from the flight path and the observed trace gas signatures.

Clearly ASMA-related air masses (POI3/5/6) are marked in shades of yellow in the left column of Fig. 1, and indicated by three black ovals in Fig. 2a. For the sake of brevity we focus on POI3 in the following, while more information about the other flight segments is available in Appendix A.

The flight segment POI3 is characterised by almost parallel back-trajectories along the southern ASMA fringe (Fig. 4c). The outer trajectories show air masses circling around the ASMA within 10 days (Fig. 4a) at 160–170 hPa (Fig. 4b), while the inner trajectories were first uplifted at the southern/south-western flanks of the Himalayas, then the Tibetan conduit to merge with the UT ASMA circulation at its eastern flank (Fig. 4d). The back-trajectories of POI3 mainly encompass South Asia and the Arabian Peninsula. We define the start of POI3 by a sudden increase of O3, CO, HCl and nitrogen oxide mixing ratios, compared to the clean air encountered before (POI2). The transition from POI2 to POI3 is not clearly reflected in the back-trajectories, because the northernmost back-trajectories of POI2 resemble the adjoining, southernmost back–trajectories of POI3 (see Appendix A, Fig. A1). We consider this to be an artefact of very dynamic wind fields that are not well represented in the trajectory calculations (see also discussion of POI2 in Appendix A). The problem is most likely related to the combination of the rapidly splitting ASMA (Fig. 3) and the strongly divergent flow at the eastern ASMA boundary (Fig. A2), where the southbound ASMA circulation separates from the eastward subtropical jet. POI3 ended when HALO started to descend for a dive over Oman. Almost immediately below the flight altitude of POI3 the back-trajectories no longer clearly indicate direct transport of air from the eastern ASMA flank.

During POI3 the fringe was transected from the outer to the inner streamlines. Deep convection at the eastern ASMA flank contributed considerably to POI5 (Fig. A4d), and very little to POI6 (Fig. A5d). POI3 passed the eastern ASMA flank on 15 September (Fig. 4a), POI5 on 12 September (Fig. A4a), POI6 on 13 September 2012 (Fig. A5a). Note that the eastern ASMA flank moved eastward during that period, and the area enclosed by the back-trajectories did shrink (Figs. 3, 4a, A4a, A5a). A schematic of the synoptic situation for POI3/5/6 is given in Fig. 5. All three POIs are part of a filament that spent at least 10 days in the UT of the ASMA region, and was entrained by updraughts at the eastern ASMA flank. The curled-in structure of the filament indicates that the ASMA split into a Tibetan and an Iranian part around 18 September 2012 (Fig. 3). It is not clear whether the steep O3 gradient, chosen as the beginning of POI3, corresponds to where outside flow later separates from the ASMA circulation. Thus we do not attempt to estimate whether or how much outside air becomes part of the ASMA circulation by entrainment at the southern edge. This uncertainty is not important for the present study, but might need to be addressed before quantitatively estimating trace gas budgets within the ASMA.

Six-hourly satellite images show no signs of fresh convection in the vicinity of the HALO track on the days before the flight, in contrast to the eastern ASMA edge and the Himalayas (see Supplement Fig. S8). Shorter-lived, localised convective events were identified in 15 min satellite images (not shown) over the Hajar mountains (Oman) and east of the Strait of Hormuz on 17 and 18 September 2012, afternoon. We set a more detailed discussion of this aspect aside here, since dispersion calculations (not shown) indicate that the UT ASMA measurements during the ESMVal campaign (POI3/5/6) were not affected by those convective plumes. Furthermore, videos from the cockpit camera show that HALO did not transect the convective region over the Hajar mountains on 18 September 2012.

POI3 is less affected by stirring during transport from the eastern ASMA flank to the measurement location than POI5/6 and thus provides a more direct view of the remote and so far unsampled eastern flank. In addition to stirring, diffusion may also act to conceal features of trace gas distributions during transport. However, assuming a diffusion coefficient of 15 m2 s-1 (Schumann et al., 1995), purely diffusive mixing is negligible here. It acts on a scale of about 1 km per day, and air parcels needed less than a week from the eastern flank to their respective measurement locations (Figs. 4a, A4a, A5a). For comparison, two measurement points in 10 s time resolution at typical HALO speed are about 2.5 km apart.

Here we discuss to what extent the aforementioned (Sect. 2.2) simulation with the EMAC model can reproduce the O3, HCl, CO, NO and NOy measurements in order to use that simulation for further interpretation of the measurements.

The EMAC simulation has a horizontal grid resolution of about 300 km in the ASMA region, and the time step length is 12 min. Processes acting on smaller, unresolved scales need to be parameterised. This is compared to in situ data with a time resolution of 10 s, corresponding to a spacing of about 2.5 km. Due to the different resolutions, a perfect match between the simulation and observations cannot be expected. Additional differences might be caused by non-perfect representations of emissions, physical and chemical processes.

EMAC in the set-up used here is known to simulate a high O3 bias in the tropics, more specifically of 5–25 % at 100–250 hPa, compared to ozonesonde data, and 30–50 % in the tropospheric column compared to satellite data (Jöckel et al., 2016). However, the relative enhancement observed during the POIs is reproduced by the simulation (Fig. 1a).

The HCl mixing ratios encountered during the flight from Male to Larnaca were at the detection limit of the AIMS instrument; therefore they had enhanced noise. They were interrupted by missing value periods due to calibrations and background measurements. In order to carve out variations on a timescale relevant for this study, we smoothed the HCl in situ data as follows: each original value of the time series is substituted by the average of two means – the mean of all values 150 s before and the mean 150 s after the original value. Missing values are ignored for calculating the mean, but the average is missing if at least one missing value is present. Each operation is based on equal weights. This procedure gives values in periods of sparse data greater weight, but was found to preserve the shape of the time series better than a conventional running mean filter. Note that the time series technically is still in 10 s resolution after the smoothing, but with regard to contents, time resolution has been traded for a better signal-to-noise ratio. In situ measurements in the subtropical UTLS over North America (Marcy et al., 2004) indicate that UT background mixing ratios of HCl may be of the order of 5 pmol mol-1. Such low values are found in our data (Fig. 1c) during clean-air-dominated POI2 and in the middle of POI5, where back-trajectories point to lower-tropospheric air (Fig. A4c, d). Relative HCl enhancements in other sections of the flight are also clearly visible in Fig. 1c, indicating in-mixing of stratospheric or TL air. The simulation reproduces the magnitude of measured HCl and also the rough time evolution during the POIs. We consider the agreement as reasonable, given the uncertainties of the measurements, as well as the possibility of spurious washing out and slightly misrepresented gradients of trace gas mixing ratios (Figs. 1d, 2b) in the simulation. The relative minimum in free tropospheric HCl is best seen in the curtain (Fig. 1d), together with a filament of increased HCl extending from the tropopause to the flight track around 08:00 UTC.

Considering the use of monthly instead of daily resolved biomass burning emission data, there is a surprisingly good agreement between measured and simulated CO. The air masses encountered during the measurements might have experienced sufficient mixing since last boundary layer contact to lose memory of any high-frequency emission variations, making monthly emissions in the simulation a viable approximation here. There is a negative bias of simulated CO of about 10 nmol mol-1 during POI5/6 (Fig. 1e). Figure 2d shows that POI5/6 coincide with a region of strong CO gradients. This may result in some inaccuracies in the simulated values along the flight path, even if the synoptic situation is captured well by the simulation. Uncertainties in the chemical mechanism also have the potential to cause a low bias of CO and a high bias of O3 (Gottschaldt et al., 2013; Righi et al., 2015). In any case, caution is needed when interpreting those measurements based on the simulation. We focus on the best represented flight section (POI3) whenever possible.

The relative changes of observed NOy (Fig. 1i, j) and NO (Fig. 1g, h) mixing ratios are captured by the simulation, in particular the enhancements of those trace gases in the ASMA. However, observed short timescale variations during POI5/6 are smoothed out in the simulated data because of the coarser temporal resolution of the output (direct) and the representation of processes is limited by the grid resolution (indirect). The representation of nitrogen oxides in the simulation also depends on the quality of the corresponding emission inventories, and is further complicated by the shorter photochemical lifetime compared to CO and O3. NO – and to a lesser degree also NOy – mixing ratios have steep vertical gradients at the flight altitude during POI5/6 (Fig. 1h, j). We do not expect a global simulation to perfectly reproduce the time and location of such features, and the corresponding inaccuracies are most likely to print through in the vicinity of steep gradients. Also, parameterisations of subgrid-scale processes are mainly designed to reproduce climatological characteristics and individual convective events in the simulation may not be triggered at the same times and locations as in reality. There are regions of over- and underestimated NO and NOy respectively, and we don't expect any systematic bias in the representation of nitrogen oxides in the simulation. In particular, we note that the magnitude of measured NO mixing ratios is reproduced by the simulation, despite most UT NOx in the ASMA being produced by lightning (see accompanying paper for details) and estimates of lightning NOx emissions include large uncertainties (Schumann and Huntrieser, 2007).

To summarise, the limited resolution of the simulation is at the core of most of the deviations between observed and simulated trace gas mixing ratios. This means that, in return, a large-scale feature like the ASMA is likely to be represented well by the specified dynamics simulation set-up, which is also well suited to reproduce the corresponding trace gas distributions. Overall, we are confident that the simulation reproduces the atmospheric situation well enough to be utilised for interpreting the in situ data of the POIs. The overall agreement between observed and simulated data is best for POI3.

Enhanced tropospheric tracers (CO) fit the climatological picture of the ASMA, but at the same time enhanced O3 and HCl is notable and indicate enhanced in situ production, or contributions of stratospherically affected air, e.g. from the TL. In this section we determine the origins of the measured trace gas signatures with the help of tracer relations. The following analysis focuses on CO vs. O3, and is supplemented by other relations (HCl vs. O3, NO vs. O3, NOx vs. O3, NOx vs. NOy). POI3, POI5 and POI6 are part of one filament, and all are characterised by mixing of UT air with uplifted lower-tropospheric air at the eastern ASMA flank (Fig. 5). In the following, we exemplify tracer–tracer relations in the filament by a discussion of POI3. We focus on that period, because it is best represented in the simulation and the dynamics are less complicated than for POI5 and POI6. The latter means that POI3 provides the most direct view of the eastern flank and the relevant processes of that key region are least concealed by stirring. Furthermore, the air encountered during POI5/6 has more remote source regions (Figs. A4a, A5a) and was subject to longer transport since passing the eastern ASMA flank. Thus, it is easier to disentangle the relevant processes for POI3.

Based on the general picture provided by previous studies, we had expected to find decreased O3 in the ASMA compared to the surrounding UT, but found increased mixing ratios instead. Thus either our presumption of generally decreased O3 in the ASMA is wrong, or HALO encountered an unusual situation. In the following we revisit studies that advocated the picture of decreased O3 in the ASMA (Randel and Park, 2006; Park et al., 2007, 2008; Kunze et al., 2010, 2016), while the frequency of occurrence of the processes needed to explain the in situ data is further discussed in the accompanying paper.

Randel and Park (2006) and Kunze et al. (2010) base their analyses of O3 on isentropic vertical coordinates, mainly the 360 K potential temperature level. Kunze et al. (2016) find a quite persistent pattern of decreased O3 concentrations during strong monsoon seasons on the isentropic levels 360 to 380 K. Isentropes tend to form a trough in the ASMA, when viewed in pressure coordinates, due to diabatic heating over the Tibetan Plateau (Ren et al., 2014). Thus it is more likely to find lower-tropospheric trace gas signatures in the ASMA interior on potential temperature surfaces than on pressure levels. EMAC simulated O3 in the ASMA is indeed decreased on isentropic levels, but at the same time increased on various UT pressure levels in the same altitude range (see Supplement S4).

However, Park et al. (2007) report an O3 minimum in the ASMA at the pressure level of 100 hPa, for July and August 2005. That is based on MLS retrievals, which were recently found to have some low O3 bias at around 100 hPa (Yan et al., 2016). Our simulation, nevertheless, reproduces an O3 minimum at 100 hPa in the ASMA for July and August in 2005 and 2012 (see Supplement Figs. S2 and S3). No O3 minimum is simulated for September (2005, 2012), and neither for 150 nor 200 hPa in any month during the monsoon season.

Park et al. (2008) report an O3 minimum inside the ASMA based on retrievals from another space borne sensor (ACE-FTS). They flag profiles as being “inside” the ASMA, if CO ≥ 60 nmol mol-1 at 16.5 km. Co-located retrievals then show decreased mixing ratios of O3 and stratospheric tracers (e.g. HCl) for the inside bin in contrast to the outside bin. This approach rather identifies an anti-correlation of CO and O3 in the UT than generally decreased O3 inside the ASMA. Such an anti-correlation is also indicated in the EMAC simulated snapshot shown in Fig. 2.

Summarising this section, our simulation is able to reproduce decreased O3 in the ASMA for those special circumstances it has been reported by other studies. At the same time, we found no indication in our simulation to expect decreased O3 in the ASMA at about 150 hPa in September. The O3 measurements taken during HALO ESMVal in the fringe (POI3) and more inside the ASMA (POI5, POI6) are consistent with that simulated September mean O3 distribution, although only two analysed seasons strictly may provide not more than a strong indication of climatological trace gas signatures in the ASMA. Rather than attributing the trace gas signatures measured by HALO to an unusual situation, we assume that presumptions dominated by midsummer data might not be valid towards the end of the monsoon season. Possible reasons for the differences between July/August versus September might include the longer time available for photochemical build-up of O3 in the ASMA, decreasing resupply of O3-poor air towards the end of the monsoon season, more in-mixing of O3-rich TL air into the decaying ASMA (dynamical instabilities), and changes to the altitudes of maximum O3 production.

Our study contributes to the so far sparse ASMA in situ measurements, allowing us to address some of the aspects of this important UT phenomenon that were recently identified as poorly understood (Randel et al., 2016): dynamical and chemical coupling with convection, composition/reactive chemistry in the monsoon region, and mixing of higher latitude lower-stratospheric air into the tropical TL by the ASMA.

Data from the HALO ESMVal campaign that were gathered during a flight from Male (Maldives) to Larnaca (Cyprus) on 18 September 2012 are presented and analysed. That region is particularly unexplored by in situ measurements. HYSPLIT backward trajectories show that HALO was in an UT filament most of the time, which had been part of the UT ASMA circulation for at least 10 days, thereby circulating around the anticyclone. Uplifted air was entrained into the UT filament at the eastern ASMA flank, which was then transported by almost unperturbed, parallel streamlines in the southern ASMA fringe (Fig. 4). Back-trajectories indicate that HALO crossed the filament three times (Fig. 5) in the zone where an originally larger ASMA was just splitting into an Iranian and a Tibetan part (Figs. 3, A2). At least a part of the filament from the eastern ASMA flank is diverted into the new Iranian anticyclone. Based on the in situ measurement data, the first transect of the fringe filament provides the hitherto most direct view of the upstream eastern ASMA flank, where several processes act that have the potential to strongly modify UT trace gas mixing ratios.

A global simulation with the EMAC model is in reasonable agreement with observed trace gas mixing ratios along the HALO flight track. The specified dynamics set-up (nudging) certainly enforces a better agreement between simulation and observations, compared to what could be expected from a free-running simulation. The synoptic scale of the ASMA acts to alleviate discrepancies that are related to the limited spatial and temporal resolution of the simulation, but a perfect agreement cannot be expected. Overall we find that this simulation is well suited to be used for further interpretation of the measurements. An ASMA splitting event indicated by back-trajectories is also reproduced by the EMAC simulation (Fig. 3) and further analysed in the accompanying paper.

Based on the general picture provided by previous studies, depleted O3 mixing ratios were expected in the ASMA compared to the surrounding UT. However, enhanced O3 was found in the ASMA filament encountered by HALO. In order to identify the processes that generated this O3 signature, additional tracers are considered for further analyses: CO as marker for lower-tropospheric air, HCl for stratospheric or TL origins, NO and NOy as important players in O3 photochemistry. All above-mentioned tracers were measured in situ, and their mixing ratios steeply increase across the boundary of the ASMA filament compared to the adjoining clean air in the south.

Tracer–tracer relations of the in situ data are consistent with a mix of UT and lower-tropospheric air in the ASMA fringe. Two effects likely have contributed to the observed signatures of increased O3: photochemical O3 production and entrainment of stratospheric or TL air. The EMAC simulation indicates that net photochemical O3 production is maximal, where uplifted air with O3 precursors originating in boundary layer pollution (e.g. CO) mixes with UT air that is enriched in (lightning) NO, another precursor. Besides increased O3, mixing ratios of the stratospheric tracer HCl are also relatively enhanced in air that had been part of the UT ASMA for longer. This trace gas signature cannot be explained by photochemical ageing of uplifted, lower-tropospheric air alone. The EMAC simulation indeed shows that a TL filament with more stratospheric trace gas signatures than the surrounding UT air is entrained into the ASMA fringe at a tropopause trough at the eastern flank of the anticyclone (Figs. 2 and 7). It is dragged away from the TP and deeper into the troposphere, circling around the ASMA interior. That particular event hardly contributed to the simulated data on the flight track, but timing and location are such that – given the uncertainties of the simulation – in reality the corresponding event might still have contributed to the observed air composition. If this is not the case, then earlier such entrainment events contributed to the ASMA trace gas signatures – in both the simulation and the measurements.

Dynamical instabilities, like the ASMA splitting event encountered by HALO, provide a means to overcome the radial transport barriers presented by closed streamlines and to effectively stir the previously entrained TL air into the ASMA interior.

Our current study focuses on the detailed analysis of a single transect of in situ data through one part of the ASMA, close to the end of the monsoon season. The relevance of the involved processes – entrainment of TL air into the ASMA fringe, photochemistry and stirring – for the trace gas distributions in the ASMA is further explored in the accompanying paper. We nevertheless found that the EMAC simulation is able to reproduce decreased O3 mixing ratios in the ASMA at 100 hPa for July and August as reported by previous studies, but it also reproduces increased O3 as observed during the HALO ESMVal campaign. Decreased O3 was found in the simulation neither for lower altitudes nor September monthly mean values, and the apparent contradiction to previous studies vanishes in this more differentiated view. The incidence of O3-rich air in the ASMA – as seen in the simulated monthly mean data – indicates that the ESMVal in situ measurements could even represent a common composition of the ASMA at about 150 hPa.