The Asian summer monsoon circulation provides a pathway for anthropogenic pollution into the stratosphere e.g.,, where it may crucially affect stratospheric chemistry and radiation. A related phenomenon is the build-up of the Asian tropopause aerosol layer (ATAL; ), which has recently been estimated to cause a significant regional radiative forcing of -0.1 W m-2 , cooling the Earth's surface. Hence, transport in the Asian monsoon is likely an important factor for climate change.

Transport by the Asian monsoon includes convection over the Bay of Bengal, northern India and the South China Sea e.g.,. At higher levels monsoon transport is dominated by a strong anticyclonic circulation with confinement and slow uplift of air in the upper troposphere and lower stratosphere (UTLS; e.g., ). Related to this transport are increased mixing ratios of trace gases with tropospheric sources and decreased mixing ratios of trace gases with stratospheric sources e.g.,. The detailed upward transport from the convective outflow to higher levels involves a vertical conduit over the southern Tibetan Plateau . In addition, convective uplift by typhoons has been shown to inject air masses into the outer region of the anticyclonic circulation . The interplay of these processes results in fast upward transport into the lower stratosphere and an enhanced fraction of young air in the monsoon UTLS region . Convection over land causes particularly fast upward transport .

Based on global satellite observations of hydrogen cyanide (HCN), argued that upward transport from the Asian monsoon reaches deep into the tropical stratosphere. Water vapor observations and simulations, on the other hand, show transport from the monsoon anticyclone mainly into the extratropical lower stratosphere e.g.,. As stratospheric water vapor is strongly controlled by cold temperatures around the tropopause these results are not necessarily contrary. However, recently even tracer-independent model diagnostics have yielded inconclusive results. On the one hand, the back trajectory study of shows strongest transport from the anticyclone directly into the tropical stratosphere. On the other hand, climate model simulations by show the tropopause crossing of air masses from the anticyclone largely in the extratropics and subsequent transport into the extratropical lower stratosphere.

Here, we use tracer-independent model diagnostics (i.e., independent of species' chemistry and emissions) in combination with satellite observations of the tropospheric tracers hydrogen cyanide (HCN) and carbon monoxide (CO) to investigate the pathways of pollution from the Asian monsoon anticyclone to the lower stratosphere, and quantify the related amount of air originating in the monsoon anticyclone. In Sect. , first we demonstrate how transport from the anticyclone can be divided into two main pathways directing into (i) the tropical pipe and (ii) the Northern Hemisphere (NH) extratropical lowermost stratosphere, over the course of a year following the monsoon season. Second, we discuss the detailed transport across the tropopause in the monsoon. Finally, in Sect.  we argue that regarding air mass transport into the stratosphere, the Asian monsoon acts as a vertical “chimney” with strong horizontal transport on top (above the tropopause).

We quantify air mass transport from the Asian monsoon anticyclone using simulations with the Lagrangian chemistry transport model CLaMS . CLaMS uses an isentropic vertical coordinate throughout the UTLS, and the model transport is driven with horizontal winds and total diabatic heating rates from European Centre of Medium-Range Weather Forecasts (ECMWF) ERA-Interim reanalysis . The horizontal resolution of the model simulation is about 100 km and the vertical resolution about 400 m around the tropical tropopause seefor details. We included an air mass origin tracer in the model to diagnose the fraction of air at any location in the stratosphere which has left the Asian monsoon anticyclone during the previous monsoon season (see below). In addition we consider carbon monoxide (CO), with the CO lower boundary in CLaMS derived from Atmospheric Infra-red Sounder (AIRS) version 6 satellite measurements following the method described in , with relevant chemistry for the UTLS region included .

It has recently been shown that trace gas confinement within the monsoon anticyclone core can be best described by potential vorticity (PV) contours , and that the anticyclone core can be clearly distinguished from the surrounding atmosphere in a layer around 380 K potential temperature . We therefore apply the method of to determine the PV value related to the anticyclone border from the maximum PV gradient on every day during (boreal) summers 2010–2013 at the 370 and 380 K potential temperature surfaces (see Appendix for further details, and the Supplement for the data).

The anticyclone tracer is initialized with unity inside the PV contour enclosing the anticyclone core in the 370–380 K layer, around 16–17 km altitude, on each day during July–August of the years 2010–2013 and is advected as an inert tracer during the following year. On 1 July of the year thereafter, the tracer is set to zero everywhere and is then reinitialized for the following monsoon season. By definition, the tracer mixing ratio at any location in the stratosphere equals the fraction of air which has left the monsoon anticyclone during the previous monsoon season see. Initializing the air mass origin tracer in the UTLS part of the Asian monsoon avoids our results being affected by small-scale transport processes in the troposphere (e.g., convection), whose representation in global reanalysis data is uncertain e.g.,. This choice of method is suitable to study the transport of air from the anticyclone, irrespective of where it originated at the surface. The impact of different boundary layer source regions on the Asian monsoon UTLS is an important research topic itself e.g.,. The monsoon tropopause is mainly located above 380 K (see Appendix and Fig. ) such that the tracer is to a good degree initialized in the troposphere and can be used to study transport from the tropopause region into the stratosphere (see Sect. ).

The anticyclone air mass tracer is compared to global HCN measurements from the Atmospheric Chemistry Experiment Fourier Transform Spectrometer (ACE-FTS) satellite instrument . These data have been presented and discussed by and shown to be a valid tracer for Asian monsoon pollution. For the results of this paper we use HCN from the updated ACE-FTS level 2 data version 3.5 during the period between 1 July 2010 and 30 June 2014, which is in good agreement with the results shown by . Physically unrealistic outliers in the ACE-FTS data have been filtered out following , discarding data with a quality flag greater than 3. Furthermore, we use CO observations from the Microwave Limb Sounder (MLS) on board the Aura satellite for validating Asian monsoon transport in the model simulation. While the vertical resolution for HCN from ACE-FTS (3–4 km) is almost twice as good as for HCN from MLS (about 6 km), MLS has a much higher sampling rate (about 3500 profiles per day) compared to ACE-FTS (maximum 32 occultations per day). Hence, for the considerations of climatological zonal mean HCN it is advantageous to use HCN from ACE-FTS (see Sect. ), whereas for maps of CO within the monsoon region the higher sampling density of MLS is beneficial. (For further comparison of the two instruments see, e.g., .)

Figure  presents the anticyclone air mass fraction and compares with HCN satellite observations from ACE-FTS. During July–September (Fig. a) the anticyclone air is transported into the lower stratosphere mainly in the subtropics (between 20 and 40∘ N). During fall (October–December, Fig. b), the anticyclone air disperses throughout the NH lower stratosphere, even reaching the tropics and Southern Hemisphere (SH). Strong wintertime tropical upwelling related to the stratospheric Brewer–Dobson circulation lifts the anticyclone air in the tropics during the following winter (Fig. c). Related downwelling in the extratropics flushes the anticyclone air out of the NH lower stratosphere. During spring (Fig. d), the anticyclone air in the tropical pipe rises further while the extratropical lower stratosphere is cleaned. Hence, two main pathways emerge for air from the Asian monsoon anticyclone into the stratosphere. First, a fast transport pathway is directed into the NH extratropical stratosphere (extratropical pathway). Second, a slower pathway is directed into the tropical stratosphere and deep into the stratosphere related to ascent within the tropical pipe (tropical pathway).

Contours of ACE-FTS-measured HCN show that the simulated anticyclone air mass fraction correlates well with satellite-observed pollution for a discussion of these data as a tracer for pollution from the Asian monsoon see. In analogy to the model tracer, observed HCN peaks in the subtropical and extratropical lower stratosphere during and directly following the monsoon season (Fig. a, b). During winter and spring (Fig. c, d), both enhanced HCN mixing ratios and anticyclone air mass fractions rise in the tropical pipe and are flushed out of the NH lower stratosphere. The good correlation between the maxima of HCN and anticyclone air mass fraction in the tropical pipe during April–June (Fig. d) renders an origin of enhanced HCN mixing ratios in the Asian monsoon very likely, as proposed by . The fact that the ascending tropical HCN signal slightly lags the model tracer signal (Fig. d) is consistent with the overestimated tropical upwelling in ERA-Interim e.g.,. During July–September (Fig. a) no agreement between the anticyclone air mass tracer and HCN mixing ratios in the tropical pipe can be expected due to the reset of the anticyclone air mass tracer to zero on 1 July (see Sect. ). Similarly, the poor agreement between the anticyclone air mass tracer and HCN in the NH lower stratosphere during April–June (Fig. d) is to be expected, because the enhanced HCN mixing ratios around the tropopause are related to young air masses while the anticyclone tracer originates in the previous monsoon season almost 1 year ago.

HCN exhibits enhanced concentrations also in the SH subtropics during austral spring to summer (Fig. b, c), consistent with independent satellite observations from the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS; ). Hence, a contribution from the SH to stratospheric HCN cannot be ruled out. Furthermore, the irregular tape-recorder signal in the deseasonalized anomaly of tropical HCN during 2005–2008 has been linked to irregularly occurring biomass burning in Indonesia . Compared to these studies, the focus here is on the annually repeating seasonal signal discussed by . The qualitative agreement between the transport pathways of HCN and air mass from the monsoon indicates that transport from the Asian monsoon anticyclone has the potential to significantly contribute to the annual signal in HCN concentrations in the stratosphere. In the following, we focus on air mass transport from the Asian monsoon, which clearly reaches the tropical pipe (Fig. ) and therefore may cause substantial pollution transport deep into the stratosphere.

The time series of air mass fractions in Fig.  show that the amount of anticyclone air peaks in the NH extratropical lowermost stratosphere in October, reaching around 15 % at 380 K. In the tropics at 460 K above the layer of frequent exchange between tropics and midlatitudes; see the amount of air which originated in the anticyclone peaks in December, reaching around 5 %. This later timing of the peak in the tropics compared to the extratropics is related to the higher potential temperature level (460 vs. 380 K) and slow tropical upwelling. At lower levels (here 400 K, red dashed) the tropical anticyclone air mass fraction peaks earlier, around October. The anticyclone air fraction in the extratropical stratosphere peaks with a value that is more than twice as high compared to the tropical anticyclone air fraction. However, the anticyclone air transported to the tropics remains much longer in the stratosphere and exceeds the extratropical amount after about half a year (at levels higher than 460 K the anticyclone air fraction peaks after January with peak values above the extratropical anticyclone air fraction; see Fig. ). The large standard deviation (from the zonal averaging) around the extratropical zonal mean value (grey shading in Fig. ) indicates strong variability in the extratropical lowermost stratosphere tracer distribution, related to the frequent occurrence of smaller-scale structures in the midlatitude tracer distribution due to various processes (e.g., Rossby-wave breaking). At the lower end of the tropical pipe (460 K), the tracer distribution is more homogeneous as reflected in a smaller standard deviation.

To further understand the details of transport from the monsoon anticyclone into the stratosphere we investigate the direction of tropopause crossing. Recently, a question was raised regarding whether the air confined within the monsoon anticyclone crosses the tropopause vertically or horizontally or, in other words, whether the monsoon acts mainly as a vertical “chimney” or as an isentropic “blower” for cross-tropopause transport . The good agreement of carbon monoxide distributions in the monsoon region at 380 K between the CLaMS simulation and MLS satellite observations shows that the model reliably simulates transport in the monsoon anticyclone (Fig. a, b upper panels). Note that the figure shows the deviation of CO from the zonal mean to emphasize the anomalous character of monsoon transport. In particular, the positive CO anomaly in the monsoon agrees well between model and observations, and even the weak positive anomalies to the northwest and northeast of the monsoon indicate regions of frequent eddy shedding .

In order to clearly separate tropospheric and stratospheric air we transform the data to a tropopause-based vertical coordinate, chosen as the distance to the local tropopause in potential temperature before calculating all averages (for using this method in a different context, see, e.g., ). The distributions in the monsoon region change substantially when viewed in tropopause-based coordinates along a surface at 10 K above the local tropopause (Fig. a, b lower panels). The positive CO anomaly significantly weakens, as an effect of the averaging procedure following the tropopause, indicating that a considerable part of the trace gas anomaly in the monsoon is related to the upward-bulging tropopause in the monsoon region. However, the fact that parts of the anomaly remain indicates upward transport across the tropopause above the monsoon. Also, for the tropopause-based map, CO distributions from CLaMS and MLS observations agree reliably well in the monsoon region. Significant differences between CLaMS and MLS exist only at midlatitudes already observed by and above the west Pacific and Maritime continent.

Figure c shows analogous maps as for CO for the anticyclone air mass fraction. Again, tropopause-based averaging weakens the positive monsoon anomaly. However, a clear maximum remains centered in the monsoon region above the tropopause. This indicates that cross-tropopause transport into the stratosphere in the monsoon occurs to a large degree in the vertical direction. Vertical transport, diagnosed from the ERA-Interim total diabatic heating rate, is consistent with this finding showing maximum upward velocity in the anticyclone (grey contours in Fig. ). The stronger degradation of the monsoon anomaly for CO as compared to the inert air mass tracer is related to the finite (∼4 months) lifetime of CO. As a consequence, CO mixing ratios degrade rapidly at levels around the tropopause, where vertical transport is slow. At levels about 30 K above the local tropopause the positive CO anomaly above the monsoon anticyclone almost vanishes, whereas the inert model tracer still shows clearly enhanced values (not shown).

An unambiguous picture of air mass transport across the tropopause can only be deduced from the inert air mass origin tracer in the model. Figure  shows the anticyclone air mass fraction averaged over the zonal section of the Asian monsoon (40–100∘ E) and over periods of about a week (with all averages carried out in tropopause-based coordinates). Directly after the main monsoon season at the end of August (Fig. a) the largest amount of anticyclone air is located in the subtropics between 20 and 40∘ N around and above the tropopause. One month later, this air has been further transported upwards and resides clearly above the tropopause (Fig. b). Hence, cross-tropopause transport of anticyclone air occurs mainly vertically across the subtropical tropopause, like in a chimney using the terminology of. Above the tropopause, however, in a layer between about 380 and 430 K the air from the anticyclone is strongly affected by horizontal transport processes and is largely mixed into the NH extratropics and into the tropics (Fig. b–d). Strong horizontal transport above about 380 K in NH summer and fall is likely related to enhanced subtropical Rossby-wave breaking during this season see. Fastest uplift in the subtropics is consistent with largest upward velocity in that region (black contours in Fig. a). Note that ERA-Interim cross-isentropic vertical velocities in August show even downwelling equatorwards of about 10∘ N in the 380–410 K layer.

There has been a recent scientific debate on if and how the air masses from the Asian monsoon anticyclone reach the lower stratosphere. concluded from 60-day backward trajectory ensembles that the preferred pathway of air masses is to travel from within the upper-tropospheric anticyclone region to the tropical lower stratosphere, but they did not further investigate where (relative to the tropopause) horizontal mixing from the monsoon region to low and high latitudes occurs. analyzed air mass origin tracers in a climate model. They found that Asian surface air is transported upwards in the monsoon, reaches the extratropical tropopause within a few days, and is first transported quasi-horizontally into the extratropical lower stratosphere before eventually being transported subsequently into the tropics. A very recent study by also shows mainly quasi-horizontal isentropic transport out of the monsoon anticyclone into the lower stratosphere.

Here, we focus on transport from the anticyclone deep into the stratosphere. Using a PV-gradient-based definition of the anticyclone edge, we trace the anticyclone air over an entire year following the monsoon season. Our analysis shows that the air from the anticyclone crosses the subtropical tropopause vertically (here cross-isentropical) and is subsequently transported horizontally (along isentropes) in the stratosphere to both the tropics and to NH extratropics, as illustrated in Fig. . The vertical nature of cross-tropopause transport is consistent with the findings of , but with the addition that above the tropopause a substantial amount of anticyclone air is mixed into the NH extratropics. This strong horizontal transport is, on the other hand, consistent with and , but with the difference that horizontal transport (either isentropic advection or mixing) in our case occurs mainly above the tropopause. It is important to note that we defined vertical and horizontal transport with respect to potential temperature as the vertical coordinate. Therefore, horizontal transport can be directly interpreted as isentropical mixing.

Hence, in summary we refine the findings of , and by describing transport from the Asian monsoon anticyclone into the stratosphere as a “blowing chimney”, using the terminology of . This characterization emphasizes the vertical “chimney-like” nature of cross-tropopause transport (with respect to potential temperature as vertical coordinate), but with the pollutants transported away quasi-horizontally along isentropes above the tropopause (see Fig. ). This quasi-horizontal transport pathway from the monsoon into the UTLS is supported by recent in situ measurements . At lower levels below the tropical tropopause (about 380 K) horizontal transport from the anticyclone core to the NH extratropics is very weak due to strong gradients in PV, in agreement with the findings of .

So far, our conclusions concern air masses from the anticyclone core. To investigate differences in transport from the anticyclone edge, we initialized an anticyclone edge tracer in CLaMS (between PV contours of the anticyclone border PV* and PV*+2 PVU; see Appendix), whose mixing ratio by definition yields the fraction of air originating from the anticyclone edge during the last monsoon season. Figure e and f show the air mass fraction from the anticyclone edge at the end of August and at the end of November. Comparison to the air mass fraction from the anticyclone core shows that directly after the monsoon season (Fig. a, e) air from the anticyclone edge is transported faster in the horizontal direction into the tropics and into NH extratropics. This is a consequence of air masses in the anticyclone edge region being less well confined as compared to air masses in the anticyclone core. After a few months, however, the two distributions of anticyclone edge and core air align (Fig. d, f), showing that, in the long term, air masses ascending in the anticyclone core and air masses injected into the anticyclone edge e.g., by typhoons; see follow the same transport pathways. The higher fraction of air from the anticyclone edge compared to the core is likely a result of the larger area of the edge region. Note that air masses in the anticyclone edge may have originated in the anticyclone core at lower levels, as suggested by the vertical transport conduit pathway proposed by .

The anticyclone air fraction of 5 % in the tropical pipe appears small if compared to the 15 % fraction in the NH extratropical lowermost stratosphere (see Fig. ). However, as tropical air ascends deep into the stratosphere with the rising branch of the Brewer–Dobson circulation while extratropical air is flushed out of the stratosphere within a few months, the impact of this tropical anticyclone air on stratospheric chemistry and climate may be substantial. Our model simulation shows that the tropical anticyclone air correlates well with the annual cycle in satellite observed HCN over the course of a year. Hence, the Asian monsoon likely causes pollution transport deep into the stratosphere and contributes to the stratospheric aerosol loading. Therefore, changes in these two pathways of pollution from the Asian monsoon anticyclone into the stratosphere likely affect chemistry and radiation and may be important for causing feedback effects in a changing climate.