Global modelling studies of composition and decadal trends of the Asian Tropopause Aerosol Layer

. The Asian Summer Monsoon (ASM) traps convectively-lifted boundary layer pollutants inside its upper-tropospheric lower-stratospheric Asian monsoon anticyclone (AMA). It is associated with a seasonal and spatially-confined enhanced aerosol layer, called the Asian 15 Tropopause Aerosol Layer (ATAL). The knowledge of the ATAL properties in terms of aerosol budget, chemical composition, as well as its variability and temporal trend is still largely uncertain, due to the dynamical variability of the AMA, the dearth of in situ observations in this region, the complex transport pathways of pollutants and its atmospheric chemical processes. In this work, we use the Community Earth System Model (CESM 1.2 version) based on the 20 coupling of the Community Atmosphere Model (CAM5) and the MAM7 (Modal Aerosol Model) aerosol module to simulate the composition of the ATAL and its decadal trends. Our simulations cover a long-term period of 16 years from 2000 to 2015. We identify a “double-peak” aerosols vertical profile for the ATAL. We attribute the upper peak (around 100 hPa, predominant during early ATAL in June) to dry aerosols, possibly from nucleation processes and the lower peak 25 (around 250 hPa, predominant for a well-developed and late ATAL, in July and August) to cloud-borne aerosols associated with convective clouds. We find that mineral dust is the dominant aerosol by mass in the ATAL showing a large interannual variability, but no long-term trend, due to its natural variation. The results between 120-80 hPa (dry aerosol peak) suggest that for aerosols other than dust the ATAL is composed of around 40 % of sulfate, 30% of secondary 30 and 15% of primary organic aerosols, 14% of ammonium aerosols and less than 3% of black carbon. The analysis of the anthropogenic and biomass burning aerosols shows a positive trend for all aerosols simulated by CESM-MAM7. (Lau et al., 2018, Basha et al., 2019, Yuan et al., 2019) and this affects the ATAL formation, location and composition. Wei et al., (2019) have also found that the AMA exhibits intraseasonal variability between the Iranian Plateau and the Tibetan Plateau with a quasy-biweekly oscillation. This study provides further insight on the chemical composition of the ATAL and assesses its decadal variability composition and aerosols trends for the first time. To asses this, we have carried out long-term modeling of the ATAL using the Community Earth System Model (CESM 1.2) which embeds the Community Atmosphere Model (CAM5) coupled with the MAM7 (Modal Aerosol Model) aerosol module. Our simulations cover an overall extended period of 16 years, 115 from January 15th 2000 to December 15th 2015. Yuan et al., 2019 derived decadal trends for carbonaceous aerosols and dust in the ATAL using only meteorological reanalysis data, while in the present study a detailed chemistry and microphysic modelling is used to estimate trends for a larger aerosols composition.

. The upper atmospheric circulation is dominated by the related Asian

40
Monsoon Anticyclone (AMA), which is known to contain enhanced concentration of tropospheric trace gases and aerosols, due to rapid lifting from the boundary layer by deep convection and subsequent horizontal confinement. The AMA is confined by the subtropical westerly jet stream in the north (~40-45°N) and the equatorial easterly jet stream in the south (~10-15°N), and spans from 20-140°E in the northern hemisphere. The altitude of maximum strength of the 45 anticyclonic circulation is around the local tropopause (17-18 km) (e.g., Dethof et al., 1999;Bian et al., 2012;Ploeger et al., 2015;Garny and Randel, 2016;Pan et al., 2016, Brunamonti et al., 2018. On a daily basis, the specific location, spatial extent and strength of the AMA depend on the internal dynamical variability of the ASM (Randel and Park, 2006;Garny and Randel, 2013;Vogel et al., 2015;Pan et al., 2016). As suggested, the AMA can effectively trap 50 boundary layer pollutants and is associated with the formation of the Asian Tropopause Aerosol Layer (ATAL) (Vernier et al., 2011, Vernier et al., 2013. The ATAL refers to an enhanced aerosol layer near the tropopause over the Asian monsoon region extending from~13 and 18 km altitudes. Its horizontal extension is determined by the AMA geometry, roughly in the broad region bounded by approximately 5-105°E, 15-45°N (e.g. Vernier et al., 2015, Lau et al., 2018, 55 Bian et al., 2020. Combined satellite observations from SAGE (Stratospheric Aerosol and Gas Experiment) II and CALIOP (Cloud-Aerosol LIDAR with Orthogonal Polarization) have highlighted the presence of the ATAL since 1998, while it was not observed prior to that year (Vernier et al., 2015). Model studies have suggested that the ATAL might have been present previously but was masked by the overwhelming UTLS aerosols produced by the Mount Pinatubo eruption 60 (Neely et al., 2014).
The sources, chemical composition and spatial and temporal variability of the ATAL are not yet well understood. Recent observations from the StratoClim (Stratospheric and upper tropospheric processes for better climate predictions) aircraft campaign in 2017 and a few recent balloon measurements from the BATAL (Balloon measurement campaigns of the Asian 65 Tropopause Aerosol Layer) 2015 campaign (Vernier et al., 2018), suggest that aerosol particles in the ATAL may contain large amounts of sulfate, as well as organics, nitrates (including ammonium nitrate), black carbon and dust (Vernier et al., 2015(Vernier et al., , 2018Yu et al., 2017;Höpfner et al., 2019). Different indications on the ATAL composition have been brought by a number of modeling studies. Yu et al. (2015), using the CESM1 (Community Earth System Model) global

70
Earth system model coupled with the CARMA (Community Aerosol and Radiation Model for Atmospheres) aerosol model, have suggested that the ATAL might be principally composed of secondary organic and sulfate aerosols, as well as of primary organic aerosols. Using GEOS-Chem (Goddard Earth Observing System with Chemistry) chemical transport model, Fairlie et al. (2020) have found significant ammont of sulfate, ammonium, organic aerosol and nitrate in the 75 ATAL, with a predominat contribution of niitrate, as was identified previously by Gu et al. (2016) using an earlier version of the model. Fadnavis et al. (2017)  aerosol eXtension (GMXe) aerosol module have found that mineral dust and water-soluble compounds, like nitrate and sulfate, are the principal aerosols typology over the Tibetan Plateau, within the AMA. Therefore, existing modeling studies have proved to be able to simulate the enhancement concentration of aerosols in the AMA region, even if a very large 85 uncertainty in the composition of the ATAL remains.
A rising temporal trend of the ATAL optical signature in the AMA region has been observed (Vernier et al., 2015). The recent rising trends of sulfur dioxide and volatile organic compounds emissions in India are a candidate for explaining the appearance of the ATAL and its evolution.
This region have been proposed as the main source region of the ATAL (Vernier et al., 2015).

90
Continental convective regions have also been shown to be the main contributors to the air trapped within the AMA with North India and South of the Tibetan Plateau as specific source areas (Tissier and Legras, 2016;Legras et al., 2019).
On the other hand, important contributions of natural sources, like dust, are expected. Xu et al. (2015), using CALIOP and MISR (Multi-angle Imaging SpectroRadiometer) satellite data, have 95 found that dust is one of the predominant aerosol over the Tibetan Plateau, and the same finding is reported by Ma et al. (2019), using the ECHAM model. Lau et al. (2018), driven by MERRA-2 data, have found abundant amounts of dust in the mid-and upper-troposphere over India and China during May to June due to the westerly transport from the Middle East desert, during July-August these large quantities of dusts transported from the deserts are trapped and 100 accumulated within the AMA and contributing to the ATAL formation.
It's also important to note that the ATAL formation and possible spatial and temporal variability is closely related to the dynamical variability of the AMA. Basha et al. (2019) have suggested that the spatial extent and strength of the AMA is greater during July and August compared to June and September, and that the decadal variability is bigger at the edges of the anticyclone.

105
As a consequence of the variability of atmospheric dynamics, some years show a stronger monsoon activity than others (Lau et al., 2018, Basha et al., 2019, Yuan et al., 2019 and this affects the ATAL formation, location and composition. Wei et al., (2019) have also found that the AMA exhibits intraseasonal variability between the Iranian Plateau and the Tibetan Plateau with a quasy-biweekly oscillation.

110
This study provides further insight on the chemical composition of the ATAL and assesses its decadal variability composition and aerosols trends for the first time. To asses this, we have carried out long-term modeling of the ATAL using the Community Earth System Model (CESM which is calculated using a ternary parameterization (H2SO4-NH3-H2O) and boundary nucleation (Merikanto et al., 2007). The inter-and intra-modal coagulation is calculated for Aitken, Accumulation and Primary Carbon modes.
In MAM7 the aerosol particles (AP) can exist in the "interstitial" state (AP that are suspended in clear or cloudy air) and "cloud-borne" state (AP attached to or contained within different 150 hydrometeors, such as cloud droplets and/or ice crystals). The cloud-borne aerosols only include AP that are within stratiform (anvil) clouds, while the interstitial aerosols include both clear-sky AP and AP contained within convective clouds. This means that the AP in convective cloud droplets are lumped with the interstitial AP in the model and the interstitial aerosol mixing ratios include the truly interstitial (i.e. "clear-sky/dry") AP and the "convective" cloud-
As has been detailed in Wang et al. (2013), in CAM5-MAM7 cloud-borne aerosols in stratiform clouds are treated in a prognostic way in CAM5: their mixing ratios are saved between model time steps and evolve as a result of source, sink, and transport processes. Their activation is http://www.globalchange.umd.edu/ceds/ceds-cmip6-data/). According to CEDS, the anthropogenic emissions are first scaled to EDGAR database for most emission species, then to national/regional inventories, e.g. REAS 2.1 (Regional Emission inventory in ASia version 2.1, Kurokawa et al., 2013) in Asia, for SO2, NOx, NMVOCs, CO and CH4. For each inventory, scaling We run our simulations for 16 years, from January 15th 2000 to December 15th 2015, using the CESM1.2 (CAM5) initial atmosphere state file at that date.

2.2-Correlative satellite data
Our simulations have been compared to satellite data from the Microwave Limb Sounder (MLS)  (Livesey et al., 2020). The data precision is about 16 ppbv and the data accuracy is estimated at ±26 ppbv and ±30%.

210
The ACE-FTS instrument is an in infrared solar occultation spectrometer, providing profiles of the Earth since February 2004 from the Canadian satellite SCISAT-1 (Bernath et al., 2005). It operates in the wavelength range from 2.2 to 13.3 µm (750-4400 cm-1) with a spectral resolution of 0.02 cm-1. The data set provides 30 measurements per day for over 30 chemical species from 5 km (or cloud top) up to 150 km. The horizontal weighting function of a 215 measurement has typically a width of~300 km. The vertical resolution is < 4 km.

3-Model evaluation with satellite observations: CO distribution
We compare CO measurements from MLS and ACE-FTS with our simulations. While a direct comparison of aerosol extinction observations from various satellite instruments with CESM-MAM7 is not easy, e.g. due to the interference of clouds, using a trace gas (like CO) is a more of Africa the CO biomass burning emissions are lower (see Van de Werf et al. 2017). This could explain the low bias in CO mixing ratios for our comparisons with satellite measurements.
While reproducing average (monthly) features is a probing test for our simulations, catching shorter-term processes and variability is even more challenging towards the description of a complex phenomenon as the ATAL. Thus, we have also tested the model's ability to reproduce 260 observed specific features in time frames of the order of a few days. Figure 1c shows an possible 'eddy shedding' case on July 4th to 6th 2005. During this short time period, a multicentric AMA is observed by MLS, with rather multiple maxima in eastern Asia, instead of a classical individual maximum above the Himalayan region. Our CESM-MAM7 simulations reasonably reproduce this pattern. They show a distributed pattern with maxima above eastern

265
Asia, but also above western Asia (Fig. 1d), which is very consistent with MLS observations of   (~40%), while an aerosol enhancement due to dust above Africa is also obtained.

320
As discussed in previous studies, the spatial extent, strength and position of the AMA is highly variable due to the dynamical seasonal variability of the ASM (Randel and Park, 2006;Garny and Randel, 2013;Lau et al., 2018;Basha el al., 2019). Due to this dynamical variability the tracer concentrations are strongly controlled by the oscillations and shedding of the AMA, that therefore affect the ATAL extent and composition. For the subsequent analyses, we have 325 defined a criterion to isolate the ATAL horizontal extent based on the geopotential height (GPH) values. For values of the GPH greater than 16.7 km, at 100 hPa, we tag the area as the AMA region. Then, we use this criterion to choose a box with the highest probability to find air masses delimited by the anticyclone. This criterion of empirically selected GPH values to represent anticyclone boundaries has been discussed and used by several authors e.g.

-Vertical distribution of the ATAL
In Fig. 3 we show CESM-MAM7 vertical aerosol mass mixing ratio profiles for the accumulation mode, averaged from June to August within the blue box of Fig. 2 and 2014. Our focus on the accumulation mode is justified by the fact that it is the principal mode that contributes to the ATAL (see Fig. S1 in Supplement), with mostly anthropogenic origin. In this first analysis, we have excluded dust. Dust is still the most important ATAL component, in our simulations, in terms of mass, but its burden and variability is mostly subject to natural factors and their variability.

365
Our simulations show a characteristic "double-peak" vertical configuration, with a bi-modal vertical distribution and two relative maxima, one at higher altitudes (~80-120 hPa) and the other at lower altitudes (~200-300 hPa). During early phases of the ASM (e.g., June, Fig. 3a) the maximum of aerosol concentrations is generally located between 200 and 80 hPa; later on (e.g., July and August, Figs. 3c,e) an aerosol enhancement at lower altitudes (around 250 hPa),

370
superimposed with a maximum at around 100 hPa, is found. This "double-peak" vertical structure could be explained looking at the interplay of interstitial and in-clouds aerosols in CESM-MAM7. As was detailed in the Sect 2.1, the interstitial aerosols include both clear-sky/dry aerosols and aerosols contained within convective clouds. Our simulations show that that during the mature phase of the AMA (July and August), at the same time of increased 375 convection, the AP in convective clouds (maximum of convective outflow at~250 hPa) also increase. This causes a maximum of aerosols at lower altitudes. Figure S2 shows the cloud ice fraction for 2014 averaged for the blue box. In June the fraction of clouds is much smaller than in July and August.
We have also tried to separate the overall in-clouds and the purely dry aerosols (these latter 380 likely coming from nucleation/condensation processes). In order to analyze the contribution of dry aerosols to the ATAL we have carried out an analysis to reduce/eliminate the contribution from convective cloud-borne aerosols. For this purpose, we have filtered out the profiles, in our blue box, for which the extinction coefficient is larger than an arbitrary threshold (1.0 10-3 km-1 in our case). Figure S3 shows the evaluation of different filters for the extinction coefficient 385 applied for our box domain for August 2014 (same behavior is observed for July, not shown).
We have applied a filter of 8.0 10-4, 9.0 10-4, 1.0 10-3 and 2.0x10-3 km-1, respectively and have evaluated the maximum value obtained at around 100 hPa. By varying these threshold values, we arrive to the point of isolating the upper peak, which is satisfactory for 1.0 10-3 km- which are larger in size due to liquid phase formation, freezing and/or hygroscopic growth (depending on the primary or secondary nature of aerosols). We then identify as clear sky/dry AP the ones associated with this upper peak (120-80 hPa). The comparison with AP vertical 395 profiles from Fig. 3a,c,e allows us to point out that in CESM-MAM7 both type of aerosols contribute to the ATAL, i.e. clear-sky/dry aerosols and convective cloud-borne aerosols.
It is worth noticing that, for these two selected years (2000 and 2014), the aerosol profiles can differ from one aerosol type to another, but are quite similar for a given month/year, and a double peak/single peak structure can be observed for one aerosol type or another. This  (Figs. 4a,c) and the isolated accumulation mode (Fig. 4b,d). To account for the whole double-peak phenomenology 415 and to isolate the single dry AP peak (see discussion in Sec. 4.2), the concentrations are averaged between 200-80 hPa (Fig. 4a,b) and 120-80 hPa (Fig. 4c,d). These two vertical ranges allow the differentiation of the ATAL composition based on in-clouds processes or, from another point of view, to describe how the composition changes depending on the altitude. No filter has been applied to show the contribution of all aerosols.

420
The aerosol type that dominates the ATAL, for both altitude ranges, is dust, followed by sulfates and organic particles (secondary and primary). The comparison between Fig 4a and 4c shows that at higher altitudes the amount of sulfates increases slightly and, more markedly, dust amount decreases. Figure 4e shows the percent contribution of aerosols types to the ATAL, between 120-80 hPa. It is evident that although less dust reaches higher altitudes, this aerosol

435
This discrepancy is probably due to the fact that: 1) they report averaged values from June to September and, 2) there are marked differences in the model online calculation for dust emission in ECHAM5 (e.g. wind speed, hydrological parameters and soil properties), with respect to our model. In the work of Yu et al. (2015), based on the CESM1/CARMA model, an enhancement of concentration of aerosols above Africa, mainly attributed to dust, has been 440 found. However, they do not report the amount of dust in the ATAL region. Even if there still is a large disagreement about the exact amount of dust present in the ATAL, it is clear that this natural component contributes significantly to the ATAL seasonal build-up due to its transport from the nearby desert regions, like Taklamakan and Thar deserts, and the northern slope areas of Tibetan Plateau (Lau et al., 2018, Ma et al., 2019. The difference in the amount of 445 dust reported by the different authors may be related to how the topography is represented in the model, the resolution of the model and the parameterization of the convection process (Brühl et al., 2018), as well as the different schemes used for the generation of dust.
With the intention to analyze the composition of the ATAL in terms of anthropogenic and biomass burning emissions we discuss more in details the contribution of the non-dust aerosols,

450
for which the accumulation mode at two different altitude ranges is shown in Fig. 4b, d.
Hence, anthropogenic and biomass burning aerosols that reach the ATAL are principally small and young. The same behavior is observed in the 200-80 hPa range (Figure not shown here).

455
Sulfate aerosols from moderate-to-strong volcanic eruptions, with injection in the UTLS, can also interact with the dynamical features of the AMA (e.g. Sellitto at al., 2017)  the AMA. The subsequently formed volcanic sulfates from SO2 conversion to particles rise the background inside and outside the AMA and therefore contributed to the ATAL burden, during these two years. For these years influenced by moderate volcanic eruptions the concentration of sulfate increases drastically and reaches or even exceeds the dust concentration (see Fig. 4).
Excluding dust and focusing on the mostly anthropogenic accumulation mode, Fig. 4f suggest 465 that the fraction of the ATAL of anthropogenic origin is composed of about 40% Sulfate, 30% SOA, 15% POM, 14% ammonium and less than 3% BC. Compared to the results reported by Yu et al. (2015), our results show about the same percentage of sulfate in the ATAL but less organics, i.e~45% aggregating SOA and POM for our study compared with 60 % of organic as reported by Yu et al. (2015).

475
In the following, we evaluate the decadal trends of the different aerosol types in the ATAL. In particular, we have estimated the trends for the dust in the fine soil dust mode (Fig. 5a) and all other aerosol types in the accumulation mode (Figs. 5b,c). The concentrations for each year are averaged between 120-80 hPa pressure levels and over the domain chosen above (blue  than values reported without applying the filter. This reflects the fact that at 120-80 hPa the dry aerosols contribute to a larger fraction of the ATAL than convective cloud-borne aerosols. The analysis of differences without and with the application of the extinction filter (i.e, (dry + convective) -(dry) aerosols) reveals that the increase for convective cloud-borne aerosols between 120 and 80 hPa in our box domain is~22% for sulfate,~10% for SOA,~28% for POM,
We have also carried out the same analysis for the larger altitude interval of the ATAL, i. e. between 200 and 80 hPa (Fig. 5d and e). More convective cloud-borne aerosols are likely in this case. Thus, the differences for the cases without versus with the extinction filter (calculated from Tab. 2) are larger than the previous case (~36% for sulfates,~44% for POM,~32% for 525 NH4, 47% for BC and~21% for SOA).

5-Conclusions
In this paper, we have presented the results for our long-term simulation, i.e. 16 years (January

605
In terms of vertical distributions, our simulations reveal a marked double-peak structure. The CESM-MAM7 simulations have allowed as to analyze separately 'clear sky/dry' aerosols and 'cloud-borne' aerosols, including those from convective clouds. Dry aerosols contribute to one higher peak (peaking around 80-120 hPa) and convective cloud-borne aerosol to one lower peak (peaking around 200-250 hPa). We show that the contribution of the convective cloud- AB also would like to thank the NCAR/CESM online discussion board for many helpful technical discussions that helped throughout this study, specially thanks to Louisa Emmons and Simone
Furthermore, the authors thank the ACE-FTS and MLS-teams.