Transport of trace gases via eddy shedding from the Asian summer monsoon anticyclone and associated impacts on ozone heating rates

. The highly vibrant Asian summer monsoon (ASM) anticyclone plays an important role in efﬁcient transport of Asian tropospheric air masses to the extratropical upper troposphere and lower stratosphere (UTLS). In this paper, we demonstrate long-range transport of Asian trace gases via eddy-shedding events using MIPAS (Michelson Interferometer for Passive Atmospheric Sounding) satellite observations, ERA-Interim reanalysis data and the ECHAM5–HAMMOZ global chemistry-climate model. Model simulations and observations consistently show that Asian boundary layer trace gases are lifted to UTLS altitudes in the monsoon anticyclone and are further transported horizontally eastward and westward by eddies detached from the anticyclone. We present an event of eddy shedding during 1–8 July 2003 and dis-cuss a 1995–2016 climatology of eddy-shedding events. Our analysis indicates that eddies detached from the anticyclone contribute to the transport of Asian trace gases away from the Asian region to the western Paciﬁc (20–30 ◦ N, 120– 150 ◦ E) and western Africa (20–30 ◦ N, 0–30 ◦ E). Over the last two decades, the estimated frequency of occurrence of eddy-shedding events is ∼ 68 % towards western Africa and ∼ 25 % towards the western Paciﬁc. Model sensitivity experiments considering a 10 % reduction in Asian emissions of non-methane volatile organic compounds (NMVOCs) and nitrogen oxides (NO x ) were performed with ECHAM5–HAMMOZ to understand the impact of Asian emissions on the UTLS. The model simulations show that transport of Asian emissions due to eddy shedding signiﬁcantly affects the chemical composition of the upper troposphere ( ∼ 100–400 hPa) and lower stratosphere ( ∼ 100–80 hPa) over western Africa and the western Paciﬁc. The 10 % reduction of NMVOCs and NO x Asian emissions leads to decreases in peroxyacetyl nitrate (PAN) (2 %– 10 % near 200–80 hPa), ozone (1 %–4.5 % near ∼ 150 hPa) and ozone heating rates (0.001–0.004 K day − 1 near 300– 150 hPa) in the upper troposphere over western Africa and the western Paciﬁc.

eral tropospheric Asian trace gases over the last decade, e.g. ozone at ∼ 1 % yr −1 -3 % yr −1 (Verstraeten et al., 2015), CO at 3 % yr −1 (Strode and Pawson, 2013) and NO x at ∼ 3.8 % yr −1 -7.3 % yr −1 (Schneider and van der A, 2012; Ghude et al., 2013). Biomass-burning is another major contributor to the observed growth in these trace gases (van der Werf et al., 2006). Peroxyacetyl nitrate (PAN), a powerful pollutant formed in biomass-burning plumes (Wayne, 2000), is a secondary pollutant produced through the oxidation of hydrocarbons released from anthropogenic and biogenic sources. It is a reservoir of reactive nitrogen and plays a fundamental role in the global ozone budget (Tereszchuk et al., 2013;Payne et al., 2017). PAN can also be formed in the upper troposphere through the production of NO x from lightning (Zhao et al., 2009). Simulations of the Model of Ozone and Related chemical Tracers (MOZART) show an increase of 20 %-30 % of PAN concentrations in the upper troposphere and lower stratosphere (UTLS) over the Asian summer monsoon (ASM) region produced from lightning (Tie et al., 2002). While in the lower troposphere, PAN has a short lifetime (a few hours), in the UTLS it has a longer lifetime (3-5 months) and can therefore act as a reservoir and carrier of NO x (Tereszchuk et al., 2013). Recent satellite observations show an increasing trend in PAN (∼ 0.1 ± 0.05 to 2.7 ± 0.8 ppt yr −1 ) in the UTLS over Asia (Fadnavis et al., 2014).
Monsoon convection plays an important role in the lofting of boundary layer Asian air masses to the UTLS (e.g. Randel et al., 2010;Fadnavis et al., 2015;Santee et al., 2017). The uplifted air masses become confined into the anticyclone enclosed by jets (westerly and easterly jets to the north and south, respectively), which act as a strong transport barrier and restrict isentropic mixing into the extratropical lower stratosphere or the equatorial tropics (Ploeger et al., 2015. Confinements of high concentrations of trace gases, including ozone precursors (e.g. hydrogen cyanide (HCN), CO, hydrochloric acid (HCl), NO x and PAN), and low ozone in the anticyclone are evident in satellite and aircraft observations (Randel et al., 2010;Vogel et al., 2014;Fadnavis et al., 2015;Ungermann et al., 2016;Santee et al., 2017). The observed ozone minimum is still an open question in spite of high amounts of its precursors in the anticyclone. A fraction of these trace gases enters the lower stratosphere and affects the UTLS chemical composition (Randel et al., 2010;Fadnavis et al., 2015;Roy et al., 2017;Garny and Randel, 2016), with associated radiative forcing impacts . Cross-tropopause transport associated with the Asian monsoon is evident in a number of species, including aerosols, hydrogen cyanide (HCN) and PAN (Randel et al., 2010;Fadnavis et al., 2014Fadnavis et al., , 2015Bourassa et al., 2012).
The ASM anticyclone is highly dynamic in nature (e.g. Hsu and Plumb, 2000;Popovic and Plumb, 2001;. On the subseasonal scale, it shows variation in strength and location (Garny and Randel, 2016). It frequently sheds eddies, and on occasion, it splits into two anticyclones, namely the Tibetan and Iranian anticyclones Nützel et al., 2016). An eddy detached from the anticyclone carries Asian air masses (trace gases) away from the ASM region. There are scattered studies indicating eddy shedding to the west (Popovic and Plumb, 2001) and east Vogel et al., 2014) of the anticyclone. An eddy-shedding event causes irreversible mixing in the surrounding air changing the chemical composition and radiative balance of that region (Garny and Randel, 2016). Here, we analyse transport of Asian trace gases via eddies in detail, subsequent mixing into the extratropics and radiative impact of eddy-shedding events on decadal scales. In this study, we answer the following questions: (1) how frequent were eddy-shedding events during the last two decades?
(2) Which regions are the most affected? (3) Does the transport of Asian trace gases arising from eddy shedding affect UTLS ozone concentrations and heating rates at remote locations?
To address these questions, we first consider an eddyshedding event demonstrating eastward and westward shedding from the ASM anticyclone during 1-8 July 2003. This year was chosen since the monsoon season was quite normal (i.e. no evidence of El Niño or Indian Ocean dipole phenomenon influencing the monsoon circulation). We then present a climatology of eddy shedding events and lead-lag relations of eddies with the anticyclone. We also evaluate the impact of increasing Asian emissions of NO x and NMVOCs on ozone and PAN during the eddy-shedding event, using model sensitivity simulations. Finally, we estimate the associated changes in ozone heating rates in the UTLS due to Asian trace gases being transported via eddy-shedding events.
2 Model set-up and satellite observations

Satellite observations
The Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) on-board the European ENVIronmental SATellite (ENVISAT) (MIPAS-E) was launched in March 2002 into a polar orbit of 800 km altitude. Its orbital period is about 100 min. MIPAS-E provided continual limb emission measurements in the mid-infrared over the range 685-2410 cm −1 (14.6-4.15 µm) (Fischer et al., 2008). MI-PAS monitored many atmospheric trace constituents including CO, PAN and O 3 . The details of the general retrieval method and set-up, error estimates and use of averaging kernel and visibility flag are documented by von . Here, we analyse the MIPAS observations of CO, PAN and O 3 during 1-8 July 2003.
To account for the comparatively low and altitudedependent vertical resolution of MIPAS, the model data were convolved with the MIPAS averaging kernel to be directly comparable to MIPAS measurements of CO, PAN and ozone. MIPAS vertical resolution for CO, O 3 and PAN in the UTLS is 5, 3.5 and 5 km, respectively. The data are contoured and gridded. For each grid point, the surrounding MIPAS data points are averaged while applying a distance weighting. The maximum distance for which MIPAS data points are considered is ±7 • in latitude and ±15 • in longitude (covering a box of 14 • in latitude and 30 • in longitude), and a minimum number of two data points per interpolation grid point are required. The data quality specifications as given in the metadata of the MIPAS data files were employed, namely only data with a visibility flag equal to 1 (all gases) and a diagonal value of the averaging kernel greater than 0.03 were used for ozone and PAN, while the diagonal value of the averaging kernel had to exceed 0.008 for CO to be used (personal communication with with Bernd Funke, 2018; see also Glatthor et al., 2007;Funke et al., 2009).

Model set-up
We employ the ECHAM5-HAMMOZ (Roeckner et al., 2003) aerosol-chemistry-climate model to understand redistribution of Asian trace gases via eddy shedding from the anticyclone. ECHAM5-HAMMOZ comprises the general circulation model ECHAM5 (Roeckner et al., 2003), the tropospheric chemistry module MOZ (Horowitz et al., 2003) and the aerosol module Hamburg Aerosol Model (HAM) (Stier et al., 2005). The chemistry of ozone, VOCs, NO x , and other gas-phase species is based on the MOZART-2 chemical scheme (Horowitz et al., 2003). It includes O x -NO xhydrocarbons with 63 tracers and 168 reactions. The details of the parameterizations and emissions used in the model as well as a validation of the results are described by Fadnavis et al. (2013Fadnavis et al. ( , 2014Fadnavis et al. ( , 2015 and Pozzoli et al. (2011).
The model simulations were performed with a T42 spectral resolution corresponding to about 2.8 • × 2.8 • in the horizontal dimension and 31 vertical hybrid σ -p levels from the surface up to 10 hPa. Here, we note that our base year for aerosol and trace gas emissions is 2000. We performed two simulations: (i) a control experiment (CTRL) and (ii) a sensitivity experiment (Asia-10), where emissions of both NO x and NMVOCs were simultaneously reduced by 10 % over Asia (10 • S-50 • N, 60-130 • E) similarly to earlier publications (Naik et al., 2005;Fadnavis et al., 2015). This fixed 10% reduction was chosen due to the spatial-temporal variability of NMVOCs over Asia and the inherent difficulty in obtaining a common trend value (Li et al., 2014). The impacts of this NMVOCs and NO x emission perturbation are investigated by analysing the associated anomalies (Asia-10 -CTRL) in ozone, PAN and ozone heating rates.
Both simulations were performed for the year 2003 driven by European Centre for Medium-Range Weather Forecasts operational analyses (Integrated Forecast System, IFS, cycle-32r2) meteorological fields (available every 6 h) (Uppala et al., 2005). All simulations include lightning NO x and the subsequent PAN production. Since the lightning parameteri-zation is the same in the CTRL and sensitivity simulations, its impact may be negligible. However, there may be an indirect impact of changed emissions on lightning and thus on NO x or PAN production. The model simulations used here are the same as those used by Fadnavis et al. (2015). The climatology of ozone mass mixing ratio, winds and potential vorticity (PV) are obtained from ERA-Interim reanalysis data for the period 1995-2016. The anomalies are obtained from differences between daily mean values of July 2003 and daily climatology. Power spectral analysis and lag-lead correlations have been carried out on PV data for the period 1995-2016 to show climatological features.
Instantaneous ozone heating rates are calculated using the Edwards and Slingo (1996) radiative transfer model. We used the offline version of the model, with six shortwave and nine long-wave bands, and a delta-Eddington 2-stream scattering solver at all wavelengths in a set-up similar to other recent studies (Rap et al., 2015, Roy et al., 2017.

A typical case study of eddy shedding from the monsoon anticyclone
The dynamics of the monsoon anticyclone is best portrayed at the 370 K potential temperature surface and the monsoon anticyclone is obvious as an area of low PV values (PV < 2 PVU, 1 PVU = 10 −6 K m 2 kg −1 s −1 ) (indicating tropospheric air mass) at this surface (Garny and Randel, 2016). Eddies are identified as air with low PV emanating from the monsoon anticyclone (Popovic and Plumb, 2001;Vogel et al., 2014). Past studies have shown that, during the monsoon season (June to September), the bulk of the low PV air at the isentropic level of 370 K is confined between about 20-35 • N and 20-120 • E, indicating the spatial extent of the anticyclone (Popovic and Plumb, 2001;Vogel et al., 2014;Garny and Randel, 2016). A pocket of low PV air mass detached from the boundary of the anticyclone (outside the anticyclone, 20-35 • N and 20-120 • E) is considered as an eddy. Figure    Plumb, 2000; Popovic and Plumb, 2001;Fadnavis and Chattopadhyay, 2017). The RWB is manifested as a rapid and large-scale irreversible overturning of PV contours on the 350 K isentropic surface. It is accompanied with a cyclonic circulation at 200 hPa (Strong and Magnusdottir, 2008;Fadnavis and Chattopadhyay, 2017). Figure 2a-h show the distribution of PV at the 350 K surface and the circulation at 200 hPa during 1-8 July 2003. It can be seen that, during 1-8 July 2003, three RWB events occurred: one near 30 • E (referred as RWB-1), one near 70 • E (referred as RWB-2) and another one near 120 • E (referred to as RWB-3). Since RWB-3 was outside the region of the ASM anticyclone (over the western Pacific ∼ 150-170 • E), it did not play a role in the eddy-shedding event of 1-8 July. If we track the locations of these RWB events (indicated by the black and red arrows), one can see that, with the progression of time, the RWB feature moved eastward. The eastward migration of RWB is linked to its movement along the subtropical westerly jet (Fadnavis and Chattopadhyay, 2017). Initially, during 1-5 July, RWB-1 was strong (PV > 2 PVU), while RWB-2 (PV < 2 PVU) was weak. During this period the southwardand westward-moving RWB-1 leads to eddy shedding over western Africa. Later, during 4-8 July, RWB-2 strengthened while RWB-1 weakened and disappeared. The southwardand eastward-moving RWB-2 was responsible for the eddyshedding event near the western Pacific (see Fig. 2d-h).

Climatology of eddy shedding from the monsoon anticyclone
A power spectrum analysis (PSA) has been performed on the PV data (averaged for 300-100 hPa) during 1995-2016 for western Africa (20-30 • N, 0-30 • E) and the western Pacific (20-30 • N, 140-150 • E). The PSA uses the temporalto-frequency fast Fourier transform in order to identify dominant signal frequencies. It provides information on signal power (square of variance) associated with the frequency components of the signal, with the dominant signal periodicity being the inverse of the dominant signal frequency. Figure 3a-b show the distribution of power spectral variance over the western Africa and western Pacific regions.
The variances corresponding to the periodicities of 3-5 days, 12-15 and 18-21 days are significant at 95 % confidence level for both the regions, indicating that the eddy-shedding activity is dominated in the range of synoptic frequency (∼ 10 days). Popovic and Plumb (2001) also indicated a typical duration of an eddy-shedding event of ∼ 4-8 days. We compute the frequency of eddy-shedding days (PV < 1 PVU) occurring over western Africa and the western Pacific. The ERA-Interim data for the last two decades show that eddy shedding is quite frequent over western Africa (∼ 68 %) and the western Pacific (∼ 25 %). The lag-lead correlation of PV (averaged for 200-100 hPa) for the central region of the anticyclone (85-90 • E, 28-30 • N) with PV averaged over the western Pacific shows a maximum positive lead correlation at 3-4 days (Fig. 3c). Similarly, PV over western Africa shows a maximum positive lead correlation for 5-6 days with the PV averaged over the monsoon anticyclone (Fig. 3d). This indicates that the transport of the eddies from the anticyclone (source region) has a typical duration of 3 to 4 days over the western Pacific and 5 to 6 days over western Africa. This transport time is the timescale over which the trace gases are moved to remote locations from the ASM anticyclone.

Horizontal transport of ozone, CO and PAN via eddies
Biomass burning over south-east Asia and east Asia produces large amounts of CO, NO x , VOCs, PAN, ozone and aerosols (e.g. Streets et al., 2003;Fadnavis et al., 2014). The monsoon convection over the Bay of Bengal, southern slopes of Himalaya and South China Sea (see Fig. S1 in the Supplement) lifts up these species into the anticyclone, where they may be dispersed in the UTLS by the vibrant anticyclone and its associated eddies. Figure 4a- (Randel et al., 2010;Roy et al., 2017). The spatial pattern of low amounts of ozone in the anticyclone and the associated eddies is evident in all of the data sets during 1-8 Similarly to ozone the maximum in the CO distribution is not collocated with eddies. Further, slight differences between model simulations and MIPAS observations are found. These differences may be due to coarse resolution, uncertainties in emissions, chemistry represented and transport processes in the model. There are differences in amounts of ozone, CO and PAN from model simulation, satellite observations and ozone from ERA-Interim. These differences may be due to a number of reasons, e.g. different grid sizes on MIPAS, ERA-Interim and model data, binning of MIPAS data for 2 days to accom- In (a)-(b) the dotted green line indicates the spectrum and blue and red lines indicate 5 % and 95 % confidence levels for lag-1 autocorrelation. Any spectral peak above the red line is statistically significant at the 95 % confidence level. modate better spatial coverage, uncertainties in the model emission inventory and retrieval errors in the satellite data. A maximum in PAN near the location of eddies differs in MI-PAS and model. Comparison of Fig. 1 and Figs. 4-6 shows that the minimum in ozone and maximum in CO and PAN is not collocated at eddies. The location varies slightly in species and data sets (in MIPAS, ERA and model). This may be due to differences in data sets and production and loss processes of each species.

Vertical distribution of CO, PAN and ozone
Further, we analyse the vertical distribution of CO and PAN as an indication of Asian biomass-burning emissions. Figure 7 shows longitude-pressure cross sections (averaged for 20-40 • N) of CO and PAN from the CTRL simulation, with wind vectors depicting circulation patterns. It illustrates that during 1-5 July 2003 a plume of CO and PAN was uplifted from the Asian region (80-120 • E), moving further upward into the UTLS. The location of the plume (Fig. 7) coincides with a strong convection region -see Fig. S1 -showing combined cloud droplet (CDNC) and ice crystal (ICNC) number concentrations from the CTRL simulation. Figures 7 and S1 together indicate that surface emissions are lifted up by the monsoon convection. In the upper troposphere (∼ 120 hPa), westward horizontal transport of CO/PAN towards western Africa is obvious as a result of eddy shedding during the respective days. In particular, during 2-4 July high amounts of CO and PAN are observed near 0-30 • E at 100 hPa ( Fig. 7ab and e-f). On 2 July, there is some PAN transport over the western Pacific. During 4-8 July 2003, eddy shedding occurs to the east of the anticyclone over the western Pacific (120-150 • E) (see Fig. 1e-f). Eastward horizontal transport of CO/PAN in the regions of eddy shedding is evident in Fig. 7c-d and g-h. The Asian trace gases then disperse downward into the troposphere (to ∼ 500 hPa over the western Pacific and to ∼ 200 hPa over western Africa) and are partially lifted into the lower stratosphere.
The vertical distribution of ozone shows low amounts of ozone extending from the convective regions of the Bay of Bengal and the South China Sea (∼ 15-25 • N) to the upper troposphere ( Fig. S2a-d), with amounts of ozone of ∼ 100-200 ppb near the tropopause (see also Fig. 4i-l). The lower amounts of ozone over the Asian troposphere may be due to  clean marine air masses during the monsoon season (Zhao et al., 2009). The feature of low ozone air mass ascent is less evident than the CO and PAN vertical ascent due to a number of factors which influence ozone production and loss processes at different altitudes in the troposphere and lower stratosphere, such as stratospheric intrusions, lightning, etc. (see discussions in Sect. 3.4).

Influence of Asian emissions on extratropical UTLS
In this section, we investigate the influence of Asian anthropogenic emissions of NMVOCs and NO x on the distribution of PAN and ozone in the tropical/extratropical UTLS from sensitivity experiments.   ated eddies (1-5 July over western Africa and 4-8 July over the western Pacific). These anomalies portray the effect of Asian boundary layer emissions (NMVOCs and NO x ) on the upper level anticyclone and the associated eddies. A number of studies (Randel et al., 2010;Fadnavis et al., 2013Fadnavis et al., , 2015Vogel et al., 2014) have shown lifting of Asian emissions to the UTLS by the monsoon convection and its confinement in the anticyclone. A 10 % decrease in Asian NMVOCs and NO x emissions decreases amounts of PAN by ∼ 5-23 % in the ASM anticyclone and the associated eddies over western Africa and the western Pacific. Further, we analyse the vertical distribution of anomalies of PAN and ozone. Figure 8e-h show longitude-pressure sections of anomalies of PAN. It shows negative anomalies (in response to reduced Asian emissions) along the transport pathways (Fig. S1), i.e. from the boundary layer of the Asian region (80-120 • E) into the upper troposphere and westward/eastward transport from the anticyclone owing to eddy shedding. These anomalies extend above the tropopause, indicating cross-tropopause transport. Upward transport across the tropopause in monsoon season has been demonstrated to occur in recent tracer studies Vogel et al., 2018). PAN is rather long-lived in the cold tropopause region and should therefore behave similarly to inert trace gases in the model simulation (Fadnavis et al., 2014(Fadnavis et al., , 2015. Our simulations show that a 10 % reduction in Asian emissions of both NMVOCs and NO x , results in a decrease in the amount of PAN by ∼ 2-10 % over north-western Africa during 1-5 July and over the western Pacific during 4-8 July 2003. The vertical distribution of ozone anomalies (Fig. 8i-l) show negative values (−1 to −4.5 %) in the troposphere extending from the surface up to ∼ 180 hPa along the transport pathways (∼ 90 • E) and in the region from where crosstropopause transport occurs. Near the tropopause (except in the region of cross-tropopause transport; indicted by boxes in Fig. 8i-l) ozone anomalies are positive, varying between 1 and 8 % (Fig. 8i-l). In contrast to PAN, ozone will be chemically active during the slow ascent over the monsoon area for several months (Vogel et al., 2018). Ozone loss rates are likely to be affected in the Asia-10 simulations. For example reduced NO x will lead to a lower efficiency of ozone loss providing a reason for higher ozone in the Asia-10 runs. Further, less NMVOCs in Asia-10 simulations might lead to lower OH concentrations in the lowermost stratosphere above the monsoon region. The major ozone loss cycle in the lowermost stratosphere in the tropics is driven by HO x radicals with the rate limiting step being the reaction of OH with ozone. The anomalies of OH concentrations are negative near the tropopause indicating lower ozone loss rates (Fig. S3). The changes in dynamics (e.g stratospheric intrusions and lightning) due to emission sensitivity may also partially contribute to positive anomalies of ozone near the tropopause. Ozone distributions from CTRL simulations show stratospheric intrusion in the northern part of the anticyclone ∼ 30 • N (Fig. S2) which is enhanced (positive anomalies) in the Asia-10 simulations (Fig. S4a-d). The spatial distribution of ozone anomalies (Fig. S4e-h) indicate that the response to emission reductions generates negative anomalies of ozone in the southern part of anticyclone (15-25 • N; 60-120 • E) (may be due to cross tropopause of monsoon air), while ozone anomalies are positive in the northern part of the anticyclone (which may be associated with stratospheric intrusions). The ozone variability near the tropopause is generally driven by the strong mixing of tropospheric and stratospheric air masses.
In Fig. 8i-l, negative values of ozone anomalies extending from the surface to ∼ 180 hPa may likely be related to the vertical extent of transport and associated outflow. A plume of high values of CO (∼ 95 ppb) and PAN (∼ 260 ppt) (Fig. 7), together with relatively low amounts of ozone (70-80 ppb) (Fig. S2) reaching to ∼ 180 hPa and leads to a strong gradient near the tropopause. This also indicates that the outflow of uplifted trace gases in the upper troposphere reaches to ∼ 250-180 hPa. The moderate concentrations of CO and PAN between 180 and 70 hPa may also be due to the slow ascent into the lower stratosphere of these Asian pollutants (Park et al., 2008).
During the monsoon season, marine air masses containing low amounts of ozone prevail over the Asian land mass. The monsoon air mass gathers Asian boundary layer ozone precursors (and other trace gases) and are uplifted to the UTLS by the monsoon circulation. It should be noted that a decrease in emissions of NO x and NMVOCs in the Asia-10 simulations produces lower amounts of ozone in the troposphere than in the CTRL simulation. Therefore, in the regions of eddy shedding, negative anomalies near 200-300 hPa indicate transport of monsoon air (via eddies) towards western Africa during 1-5 July and to the western Pacific during 4-8 July.

Influence of Asian emission of trace gases on ozone heating rates
Ozone is a dominant contributor to radiative heating in the tropical lower stratosphere, impacting the local heating budget and non-local forcing of the troposphere below (Gilford and Solomon, 2017). We estimate changes in ozone heating rates caused by a 10 % decrease in Asian NMVOCs and NO x emissions. Figure 9a-d, show anomalies of ozone heating rates on 1-8 July (plotted on alternate days), indicating a reduction in ozone heating rates in response to a decrease in Asian NMVOCs and NO x emissions, coincident with the region of convective transport (see also Fig. S5). In the upper troposphere (300-180 hPa), the negative anomalies in ozone heating rates vary between −0.001 and −0.0045 K day −1 . Interestingly, reduced Asian emissions (NMVOCs and NO x ), lead to a reduction in ozone, which leads to a reduction in ozone heating rates (−0.001 to −0.003 K day −1 ) in the region of eddy shedding over western Africa (1-5 July) and the western Pacific (4-8 July). The ozone poor Asian air mass trapped within eddies has reduced the heating over western Africa and the western Pacific. Influence of Asian NO x emissions on ozone heating rates (mean for June-September ∼ 0.0001-0.0012 K day −1 for 38 % increase over India) in the upper troposphere (300-200 hPa) have been reported in the past (Roy et al., 2017). Near the tropopause ozone heating rates are positive (0.001-0.005 K day −1 ) except in the region of cross-tropopause transport (marked in Fig. 8i-l). The positive anomalies of ozone heating rates are associated with positive anomalies of ozone near the tropopause. The ECMWF data set for 44 years  shows an interannual amplitude of the ozone heating rate ±0.00025 K day −1 near the tropopause over 30 • S-30 • N (Wang et al., 2008).

Summary and discussion
In this study, we showed evidence of eddy shedding from the ASM anticyclone to both its eastern and western edges, during 1-8 July 2003 based on MIPAS satellite observations and ERA-Interim reanalysis data as well as the associated transport patterns of trace gases from the ASM region to remote regions. The transport diagnostic based on ERA-Interim data shows that eddy shedding events are associated with RWB in the subtropical westerly jet. The RWB feature moves eastward in the subtropical westerly jet. Initially, during 1-5 July 2003, RWB occurs in the western part of the anticyclone and then sheds over western Africa (20-30 • N, 0-30 • E). Later, during 5-8 July 2003, RWB moves to the eastern part of the anticyclone and sheds an eddy over the western Pacific (20-30 • N; 120-150 • E). Analysis of ERA-Interim PV data for the last two decades  shows that the occurrence frequency of eddy shedding from the ASM anticyclone over western Africa is ∼ 68 % and ∼ 25 % over the western Pacific. In the UTLS (300-100 hPa), eddies (PV < 2 PVU) over western Africa and the western Pacific shows the highest correlations with the PV in the anticyclone after accounting for 3-4 days or 5-6 days of lag. This indicates that the anticyclone sheds eddies with a transport duration of typically 3 to 4 days to western Africa and 5-6 days to the western Pacific. We employed the chemistry climate model ECHAM5-HAMMOZ to investigate the transport of Asian boundary layer trace gases (CO, ozone and PAN) into the monsoon anticyclone and the associated eddies. The model simulations show that Asian trace gases transported into the monsoon anticyclone are further carried away horizontally towards western Africa and the western Pacific by eddies which detach from the anticyclone. These eddies protrude down to ∼ 200 hPa over western Africa and ∼ 500 hPa over the western Pacific. They redistribute Asian trace gases downward into the troposphere over these regions. Moreover, part of this air mass is also transported upward into the lower stratosphere. A higher frequency of eddy shedding over western Africa (68 %) during the last two decades  indicates a greater influence of Asian trace gases on the UTLS over western Africa than the western Pacific over this period.
We evaluated the impact of Asian NO x and NMVOCs emissions on ozone and PAN in the regions of the ASM anticyclone and the associated eddies. The model sensitivity simulations for a 10 % reduction in Asian emissions of NMVOCs and NO x indicate significant reduction (∼ 2 − −10 %) in the concentration of PAN in the UTLS (300-80 hPa) over western Africa and the western Pacific. The vertical distribution of anomalies of PAN shows negative val-ues along the transport pathways, i.e. rising from the Asian region (80-120 • E) into the upper troposphere, and in both westward and eastward transport towards the region of eddy shedding. Tropospheric ozone (1000-180 hPa) shows a decrease of up to −4.5 % in response to a 10 % decrease in Asian emissions of NMVOCs and NO x , while positive ozone anomalies (up to 8 %) are seen near the tropopause. In general, negative ozone anomalies in response to 10% reduction of NO x and NMVOCs in the region of convective transport are seen in Fig. 8i-l. However, positive anomalies of ozone are observed near the tropopause (except in the region of cross-tropopause transport), which may be due to reduction in the efficiency of ozone loss induced by lower concentrations of NO x and OH in the Asia-10 simulations and changes in dynamics due to emission changes, e.g. stratospheric intrusions and lightning. The mixing of tropospheric and stratospheric air masses near the tropopause generates ozone variability. However, such an analysis is beyond the scope of the paper.
Our analysis indicates that transport of Asian trace gases from the anticyclone to western Africa and the western Pacific via eddies causes a change in the chemical composition of the UTLS and may therefore impact the radiative balance of the UTLS. We also estimate that a 10 % reduction in Asian NMVOCs and NO x emissions leads to a decrease in ozone heating rates of 0.001 to 0.004 K day −1 in the region of transport into the troposphere and an increase of 0.001 to 0.005 K day −1 near the tropopause and lower stratosphere (180-50 hPa) over Asia (20-150 • E; 20-40 • N). Previous studies showed that ozone changes in the lower stratosphere have the largest impact on the ozone radiative forcing . Interestingly, in the upper-troposphere (200-300 hPa) negative anomalies of ozone heating rates (∼ 0.001-0.003 K day −1 ) are seen in the region of eddy shedding over western Africa and the western Pacific. Thus, transport of Asian air masses via eddies eventually alters the heating rates in the UTLS in the regions of eddy shedding and may thus affect radiative forcing and local temperature. However, such questions are beyond the scope of this study. It should be noted that the distributions of MIPAS concentration fields look different from those of ERA-Interim and ECHAM5-HAMMOZ. These differences may be due to a number of reasons, e.g. different grid sizes of MIPAS, ERA-Interim and model data, binning of MIPAS data for 2 days to accommodate better spatial coverage, uncertainties in the model emission inventory, and retrieval errors in the satellite data. The ozone heating rates estimated from the model simulations will vary accordingly. Notwithstanding, we suggest further scrutiny of long-range transport of Asian trace gases via eddies shedding from the anticyclone and its impact on ozone heating rates in the respective regions.