Corresponding author: S. Oberländer, Institut für Meteorologie, Freie Universität Berlin, Carl-Heinrich-Becker Weg 6-10, 12165 Berlin, Germany. (email@example.com)
 Climate models project an increase in the Brewer-Dobson circulation (BDC) with increasing future greenhouse gas (GHG) concentrations. This study identifies the causes of future changes in the BDC from sensitivity simulations with the EMAC chemistry-climate model, by changing the external forcings, like GHG concentrations, sea surface temperatures (SSTs) together with sea ice concentrations, and ozone-depleting substances (ODS), separately. The particular influence of rising tropical SSTs is assessed. Contributions of different waves to changes in the residual circulation are calculated as well as changes in mean age of stratospheric air (AoA) to account for the effect of mixing processes. We find that in boreal winter the tropical upward mass flux increases by about 1%/dec in the upper and 2%/dec in the lower stratosphere until the end of the 21st century. Mean AoA decreases by up to 60 and 30 days/dec, respectively. Changes in transient planetary and synoptic waves account for the strengthening of the BDC in the lower stratosphere, whereas upper stratospheric changes are due to improved propagation properties for gravity waves in future climate. Regarding the external forcings, the radiative impact of rising GHG concentrations is detected to affect upper stratospheric layers only, whereas lower stratospheric signals are almost entirely due to rising SSTs. Changes in tropical SSTs influence not only the shallow but also the deep branch of the BDC as confirmed from both changes in residual circulation and mixing. Declining ODS will slightly counteract the BDC increase in the Southern Hemisphere.
 The transport of air and trace constituents in the stratosphere and mesosphere is controlled by the Brewer-Dobson circulation (BDC), which is characterized by rising motion in the tropics, poleward flow in the stratosphere, and sinking motion at high latitudes. In the mesosphere the BDC becomes a global circulation system with rising motion in the summer hemisphere, meridional flow toward the winter hemisphere, and sinking motion at winter high latitudes. The mass transport by the BDC is composed of two components: (a) the zonal mean residual circulation (RC) with a meridional () and a vertical component (), as defined in the Transformed Eulerian Mean (TEM) framework, representing the net advective mass transport [e.g., Andrews et al., 1987], and (b) two-way quasi-horizontal mixing processes. As shown, e.g., by Rosenlof et al. , Plumb , Birner and Bönisch  and Bönisch et al. , the RC can be considered to consist of a lower or shallow branch controlling the transport of air in the tropical lower stratosphere and an upper or deep branch effective in the mid-latitude upper stratosphere and mesosphere. Both the RC and quasi-horizontal mixing are driven by atmospheric waves on different spatial and temporal scales [Plumb, 2002]. In winter, planetary-scale tropospheric Rossby waves propagate upward into the stratosphere where they dissipate and transfer their momentum to the zonal mean circulation. Thereby, a meridional circulation is induced below their breaking levels as has been formulated theoretically in the downward control principle (DCP) [Haynes et al., 1991]. Tropospheric synoptic waves are responsible for the year-round two-cell structure of the RC in the lower stratosphere [e.g., Plumb, 2002]. Additionally, the RC is affected by orographic gravity wave drag (OGWD) in the stratosphere [e.g., McFarlane, 1987; Okamoto et al., 2011] as well as non-orographic gravity wave drag (NGWD) in the upper mesosphere that is induced by tropical convection or shear instability [e.g., Rind et al., 1988]. In a chemistry-climate model (CCM) intercomparison study performed within the Chemistry-Climate Model Validation (CCMVal) initiative, it was concluded that for the multi-model mean, orographic gravity waves (OGWs) contribute about one third to the total wave drag in the lower stratosphere, in contrast to NGWD that is of minor relevance for the tropical upwelling in the lower stratosphere [Butchart et al., 2010].
 Atmospheric waves are also responsible for the mixing of air in the tropical upper troposphere/lower stratosphere (UTLS) where the subtropical transport barrier is weak [Neu and Plumb, 1999] and at the edge of the stratospheric polar vortex in winter. A widely used metric for the BDC, including both the RC and two-way mixing, is the age of air (AoA). Hall and Plumb  defined the AoA as a statistical distribution of transit times of all elements of an air parcel between the entry point into the stratosphere at the tropical tropopause and a point in the middle atmosphere for which the AoA is determined. The average over the distribution is termed the mean AoA. An increase in the BDC would be associated with a decrease in mean AoA, and vice versa.
 Changes in the BDC are of direct relevance for the assessment of climate change as the BDC determines the exchange of air between the troposphere and the stratosphere. A change in the tropical upwelling branch due to climate change might affect the import of anthropogenic emissions into the stratosphere and the future development of the stratospheric ozone layer, while changes in the downwelling are expected to modify tropospheric composition and air quality [World Meteorological Organization (WMO), 2011]. An assessment of changes in the BDC for the past decades revealed substantial uncertainties. Measurements from balloon-borne trace gas data [Engel et al., 2009] and Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) satellite observations [Stiller et al., 2012] indicate that the BDC may have slightly decelerated and mean AoA become older in the mid-latitude upper stratosphere since 1975. By contrast, Randel et al.  related an observed cooling of the tropical lower stratosphere since the beginning of the 21st century to stronger tropical upwelling due to an enhanced BDC, which is consistent with estimates from a number of CCMs [Butchart et al., 2010, and references therein].
 As for the response of the BDC to future climate change, there is agreement between the models of an increase, particularly in the boreal winter season [e.g., SPARC CCMVal, 2010]. On average, CCMs project an increase in tropical upwelling at the 70 hPa level (∼18 km) of about 2%/dec and a reduction in the mean AoA by about 0.05 yr/dec nearly everywhere in the stratosphere [Butchart et al., 2010]. Nevertheless, there are still gaps in the understanding of the mechanisms that lead to the simulated changes. In particular, the contributions of different wave types to the simulated changes in the BDC vary between models. Some studies [e.g., Butchart and Scaife, 2001; Garcia and Randel, 2008; Li et al., 2010] attributed the acceleration of the BDC to changes in the wave drag mainly from resolved waves. Other model studies emphasize the role of OGWs for tropical upwelling, in particular in the lower stratosphere mid-latitudes during boreal winter [e.g., Li et al., 2008; McLandress and Shepherd, 2009; Okamoto et al., 2011]. The CCMVal model intercomparison identified gravity wave drag as the main driver of the positive trend in tropical upwelling, contributing up to 77% in DJF and 59% in the annual mean. However, the partitioning between OGWD and resolved wave drag varied strongly among the five contributing CCMs, which leads to potential biases due to extreme values in individual models [Butchart et al., 2010].
 A further unresolved issue is to what extent the enhanced wave drag in a future climate arises from enhanced wave generation and dissipation or from changes in the conditions for wave propagation. Deckert and Dameris  and Calvo and Garcia  identified an increase in tropical quasi-stationary Rossby wave generation arising from enhanced deep convection in areas with increased sea surface temperatures (SSTs). At subtropical and mid-latitudes, an increase of wave activity into the lower stratosphere was found in different studies, e.g., Sigmond et al. , Garcia and Randel , Oman et al. , and Garny et al. . McLandress and Shepherd  suggest an increased upward EP-flux from resolved waves in the troposphere rather than a changed propagation direction within the stratosphere, i.e., changes in wave refraction. For OGWs they indicate an upward shift of breaking levels and dissipation. As explained by Shepherd and McLandress , the mechanism for increased stratospheric resolved wave drag is essentially the same as tropospheric warming, and the associated increase in the meridional temperature gradient lead to a strengthening of the upper flank of the subtropical jets, shifting the critical layers for Rossby wave breaking upward, so that more Rossby waves reach the lower stratosphere, where they dissipate and provide the drag to enhance the BDC.
 Most of the above described findings have been derived from transient CCM simulations that were performed as contributions to the CCMVal initiative and cover the period from 1960 to either 2050 or 2100. In these simulations, the roles of individual drivers for the simulated changes in the BDC cannot unambiguously be isolated as the models are driven by combinations of prescribed future changes in the concentrations of greenhouse gases (GHGs) and ozone-depleting substances (ODS). Moreover, the results are based on individual simulations or small ensembles of up to three simulations. Given the large interannual variability of the stratosphere in boreal winter, significant changes in future wave driving derived from one future realization may be hard to detect and differ between models. A useful alternative approach to understand the response of the BDC to the different external forcings is therefore the analysis of sensitivity simulations, in which the external forcings are applied individually and independently. As CCMs without interactively coupled ocean model require prescribed SSTs as lower boundary condition, such sensitivity simulations allow the distinction between the direct radiative effect of rising GHG amounts and the indirect effect of rising GHG amounts on the SSTs. In the former case, this is realized by only prescribing changes in GHG concentrations. In the latter case, only SST changes are specified. Usually, they are taken from future projections of the parent atmosphere-ocean general circulation model (AO-GCM) and their response included to the changes in GHG concentrations. This approach of decoupling enables us to estimate the magnitude of the impact from different forcings on the future development of the BDC. Only a few previous sensitivity studies addressed the role of rising GHG amounts. Sigmond et al.  separated tropospheric and stratospheric CO2 changes and suggested that the increase in the stratospheric residual circulation is due about two thirds to tropospheric and one third to middle-atmospheric CO2 doubling. Both forcings were found to be additive. Fomichev et al.  and Garny et al.  reported an effect of rising SSTs on the low latitude wave driving and the tropical upwelling. The effect of changes in ODS was studied by, e.g., Oman et al.  and McLandress et al. , who found that the future GHG induced changes in mean AoA and tropical upwelling will be damped by ozone recovery.
 The aim of this study is to unravel the relative roles of the different external forcings, which cause future changes in the BDC in individual simulations with the EMAC CCM. Our study is complementary to previous work as we compare the effects of all relevant forcings for future climate (i.e., GHGs, ODS, and SSTs) on the BDC in a set of systematic sensitivity simulations with the same model. For this purpose, multi-year timeslice simulations with constant boundary conditions of the prescribed forcings have been performed and analyzed allowing the identification of statistically significant signals. We apply the DCP [Haynes et al., 1991] to separate the contributions to the wave drag by different wave types for the individual external forcings. The contribution of resolved as well as parameterized waves to the residual mean mass stream function and the tropical upward mass flux are analyzed. Resolved waves are split into temporal (stationary and transient) and spatial (planetary and synoptic) components. We will address the question whether the future BDC is more affected by changes in wave generation or propagation, and we shed light on the importance of parameterized gravity waves for future BDC changes. In addition, we discuss changes in the mean AoA to take the additional influence of mixing processes on the BDC into account.
 This study is structured as follows. In section 2, the EMAC model and the simulations are presented and the applied methods are introduced. Section 3 shows the results for changes in temperature and circulation due to the different external forcings. Causes for changes in the RC are isolated by the analysis of the contributions from different waves to changes in tropical upwelling and the residual mean mass stream function for the different forcings. The section ends with an analysis of changes in mean AoA. The paper is completed by a summary and concluding remarks in section 4.
2 Model and Experiments
 The presented model simulations have been performed with the coupled ECHAM/MESSy Atmospheric Chemistry (EMAC) CCM. EMAC is a numerical chemistry and climate simulation system that includes submodels describing tropospheric and middle atmosphere processes [Jöckel et al., 2006]. It uses the first version of the Modular Earth Submodel System (MESSy1) to link multi-institutional computer codes. The core atmospheric model is the fifth generation European Centre Hamburg general circulation model (ECHAM5) [Roeckner et al., 2003, 2006]. For the present study, we applied EMAC (ECHAM5 version 5.3.01, MESSy version 1.7) in the T42L39MA-resolution, i.e., with a spherical truncation of T42 (corresponding to a quadratic Gaussian grid of approx. 2.8° by 2.8° in latitude and longitude) with 39 hybrid pressure levels up to 0.01 hPa. The applied model setup comprised the standard submodels for hydrological and radiative processes, and homogeneous as well as heterogeneous chemical reactions. For a better representation of the UV/VIS spectral range, the FUBRAD radiation-scheme is used [Nissen et al., 2007]. Small scale OGWs are parameterized using the formulation of Lott and Miller . The scheme by Hines[1997a, 1997b] accounts for non-orographic gravity waves (NGWs) that are generated through convection, fronts, or other small scale processes.
 EMAC performs well compared to other CCMs and observations as pointed out in Eyring et al.  and Austin et al. , based on multi-decadal transient SCN-B2c- and SCN-B2d-simulations, defined within CCMVal [Eyring et al., 2008]. Weber et al. , who analyzed the relationship between the polar spring-to-fall ozone ratio and winter eddy heat flux deduced a good representation of the overall BDC pattern for EMAC.
 We present results from five sensitivity simulations performed with EMAC in timeslice mode. Except for the reference simulation that comprises 75 years, each simulation is integrated over 40 years following a spin-up time of 2 years. The boundary conditions vary monthly without interannual variability. The advantage of simulations in timeslice mode is the large ensemble of data for the same model year. Hence, statistically significant results can be obtained with comparatively small computer resources. The individual simulations differ with respect to external forcings for future changes in the BDC, e.g., tropical and global SSTs as well as GHG and ODS concentrations, respectively. With the assumption of linear additivity of the different forcings [Sigmond et al., 2004; Fomichev et al., 2007], we isolate the causes for future BDC changes.
 All simulations are performed under mean conditions of the 11-year solar cycle, using spectral solar flux input data averaged between the maximum and minimum of solar cycle 22 [Lean et al., 2005]. The quasi-biennial oscillation (QBO) and volcanic eruptions are not included in the model simulations. Hence, easterlies prevail in the tropical lower stratosphere. Emissions of ozone precursors are prescribed following the Reanalysis of the Tropospheric chemical composition data set (RETRO) [Schultz et al., 2007]. Ozone and other atmospheric trace gases are calculated interactively. The main features of the timeslice simulations as well as their purpose are as follows:
 The reference simulation 2000 follows the CCMVal REF-B0 simulation [Eyring et al., 2008] except for SSTs and sea ice concentrations (SICs). Observed GHG and ODS concentrations for the year 2000 are prescribed from IPCC  and WMO , respectively. SSTs and SICs are 10-year monthly means (1995–2004) from a simulation with the coupled AO-GCM ECHAM5 Max Planck Institute Ocean Model (MPIOM) [Jungclaus et al., 2006].
 For the future simulation 2095, GHG concentrations follow the IPCC SRES A1B-scenario for 2095 [IPCC, 2001], while ODS concentrations for 2095 are taken from WMO A1-scenario [WMO, 2007]. CO2 concentrations are nearly doubled from 367 ppm in the reference simulation to 689 ppm in the simulation 2095, while the ODS decrease toward the end of the 21st century. SSTs and SICs are prescribed as 10-year monthly means (2090–2099) from a MPIOM simulation with rising GHG concentrations following the IPCC A1B-scenario. The 2095 simulation provides a reference climate of the end of the 21st century. Compared to the 2000 simulation, it projects future climate change due to the main forcings, GHG increases and the recovery of the ozone layer.
 The sensitivity simulation SENGHGSST uses GHG concentrations as well as SSTs and SICs as prescribed for the 2095 simulation and ODS concentrations as specified for the 2000 simulation. The comparison with the 2000 simulation allows us conclusions on the total effect of GHG increases, including the GHG response of the SSTs.
SENSST is a sensitivity study with GHG and ODS concentrations prescribed as in the 2000 simulation, but SSTs and SICs taken from the 2095 simulation. By comparing SENSST with the 2000 and the SENGHGSST runs, respectively, we are able to distinguish the influence of GHG changes on the SSTs (here “SST effect”) from the influence of changing GHG concentrations on radiation and composition (here “GHG effect”).
 The SENtropSST sensitivity run uses artificial SSTs, modified in the tropics only, to examine their impact on changes in wave generation and propagation. SSTs, as prescribed for the 2095 simulation, have been implemented in the inner tropics between 15°N and 15°S and SSTs as specified for the 2000 simulation for the extratropics between 30° latitude and the poles. Between 15° and 30°N/S linear interpolation is used. GHG and ODS concentrations as well as SICs are prescribed as for the 2000 simulation. The impact of extratropical SST changes on the BDC can be estimated from the comparison of the simulations SENSST and SENtropSST.
 A summary of the simulations is given in Table 1. Table 2 provides an overview on how the model runs are combined to derive the effects of specific forcings. We are aware of the limitations of our experimental setup to assess the contributions of GHGs and ODS to future change. However, the assumption of linearity of the external forcings is to a first order an acceptable approximation to reality [e.g., Sigmond et al., 2004; Fomichev et al., 2007]. Therefore, our set of simulations provides a consistent complementation to existing studies on the BDC changes.
Table 1. Timeslice Simulations With EMAC
trop SST: 2090–2099
Table 2. How to Extract Causes for Climate Change From Sensitivity Runsa
The first column shows the differences between the simulations; columns two to four, the inclusion of the forcings from GHGs, ODS, and SSTs/SICs; and the last column, the extracted signal.
GHG + SST
2.2.1 Residual Circulation and “Downward Control Principle” (DCP)
 As stated above, the BDC includes both the RC and two-way quasi-horizontal mixing processes [e.g., Bönisch et al., 2011]. The meridional () and vertical () components of the RC are calculated from 6-hourly model data according to the TEM equations [Andrews et al., 1987]. The mass transport by the RC can be expressed in terms of the residual mean mass stream function Ψ [Holton, 1990]. For the reason of mass conservation, Ψ is given by
where a is the radius of the earth, ρ0 the air density, and ϕ the latitude [Haynes et al., 1991]. The overbars indicate zonal mean quantities. As the RC is driven by atmospheric waves on different temporal and spatial scales [Plumb, 2002], a separation has been applied for those waves that are resolved by the model, i.e., waves long enough to be captured within the model grid. Atmospheric waves are separated by their time scale into stationary and transient waves. Stationary waves are spatial deviations from the monthly zonal mean, while transient waves are derived from 6-hourly data. A spatial separation has been carried out to distinguish between long planetary waves with wave numbers one to three and smaller scale, synoptic waves for the remaining wave numbers four to 42.
 To study the contribution of different scale waves to Ψ, we make use of the DCP [Haynes et al., 1991]. Under steady conditions, the DCP states that the extratropical vertical mass flow through a given surface is determined by the sum of all zonally averaged forces above this level. Hence, Ψ(ϕ,z) for a given latitude ϕ and level z can be calculated for a steady atmosphere, following the TEM framework [Haynes et al., 1991] by
where the term characterizes the zonal mean drag due to the dissipation of atmospheric waves. The zonal mean angular momentum and its derivation are defined as and , respectively, where the rotation rate of the Earth is Ω. Because of the singularity in sinϕ, the DCP is not defined for tropical latitudes.
 The tropical upward mass flux () is commonly used as an indicator for the strength of the BDC. It is defined as the integrated residual mean mass stream function equatorward of the so-called turnaround latitudes (TLs) of the Northern Hemisphere (NH) and the Southern Hemisphere (SH), with the vertical residual velocity pointing upward. Due to mass conservation, can be calculated as the sum of the extratropical downward fluxes in the NH () and in the SH () [Holton, 1990]: . Ψmaximizes at the TLs, where tropical upwelling changes to extratropical downwelling. As Ψ vanishes at the poles, and depend on Ψ at the TLs only, and can be calculated as follows [Holton, 1990; Okamoto et al., 2011]:
The zonal mean forcing term in equation (2) can be expressed in terms of the drag by model resolved and unresolved (i.e., gravity) waves:
The Eliassen-Palm flux divergence () describes the forcing by model resolved waves. The forcing by unresolved gravity waves is decomposed into the forcing by OGWs () and NGWs (), according to their sources. As the OGW part is not a model output quantity, it is calculated here from the difference to the total forcing.
2.2.2 The Concept of Age of Stratospheric Air
 To study the combined effect of the slow circulation and fast mixing processes, we make use of the concept of AoA. The mean AoA can be obtained from increasing surface emissions of sulfur hexafluoride (SF6), which is a stable gas throughout the troposphere and stratosphere. It decreases only in the mesosphere due to chemical loss reactions (not included in EMAC) and photolytical processes. In our study the stratospheric entry region is set to the 100 hPa level in the inner tropics. The mean AoA is calculated from the difference in concentration between the stratospheric entry and at the respective position in the stratosphere divided by the annual mean slope of the linearly increasing SF6 concentration at the ground.
 This section presents the results on changes in temperature and circulation due to the separated influences of future amounts of GHGs and ODS as well as rising SSTs, and investigates the causes for the change patterns. Following the well-established fact that the BDC is strongest in NH winter season [e.g., Olsen et al., 2007], we concentrate on changes in the December-January-February season, referred to as DJF, hereafter. Future changes for the various forcings are shown per decade with respect to the time period of 95 years between reference and future climate. This allows us to compare with transient simulations.
3.1 How Does Climate Change Affect Temperature and Circulation?
3.1.1 Changes in Temperature and Zonal Wind
 Changes in GHG concentrations or ODS affect the radiation balance in the middle atmosphere, inducing changes in meridional temperature gradients and by the thermal wind relation changes in the zonal flow. Figure 1 shows the response of the zonal mean temperature to future changes in individual and combined forcings in NH winter. The headers of the plots indicate the forcings extracted from the differences between the simulations (cf. Table 2). Statistically significant changes are computed with the Student's t-test. Significant positive (negative) changes at the 95% level are shaded in orange (light blue) here and in the following figures. The total signal, i.e., including changes in all external forcings to 2095 conditions (Figure 1a), shows a tropospheric warming and stratospheric cooling pattern. The tropospheric warming reaches up to 0.6 K/dec in the tropics and is almost entirely due to the GHG induced (mainly tropical) SST effect (Figures 1d, 1f) while the direct radiative effect of rising GHG concentrations explains only a small portion of the total tropospheric signal (Figure 1c). The slight cooling in 100 hPa at 60°N is due to the SST forcing, in particular by tropical SSTs. Higher SSTs cause a warmer troposphere and thereby a lifted tropopause in the future (see supporting information, Figure S1). The intense tropospheric warming at NH high latitudes can be explained by large decreases of sea ice that were found in an additional sensitivity run with reduced SICs only (not shown). The stratospheric cooling of up to 0.8 K/dec by the end of the century can be explained by the radiative effect of rising amounts of GHGs (Figure 1c). Tropical SSTs account for an additional cooling signal in the tropical lower stratosphere. The stratospheric temperature change due to ODS is of opposite sign with a slight warming of up to 0.2 K/dec in SH summer in 100 hPa and at the stratopause. According to the prescribed WMO A1-scenario [WMO, 2007], ODS will decrease in the future, leading to a stratospheric ozone recovery and higher temperatures.
 In the Arctic lower and middle stratosphere, EMAC shows a weak nonsignificant warming in boreal winter (Figure 1a). This signal is in good agreement with results from the CCMVal-2 transient simulations that projected no significant future temperature change in northern Arctic winter [SPARC CCMVal, 2010]. A closer inspection of the temperature response in EMAC to the individual forcings reveals that the weak future polar warming is a result of two compensating effects, i.e., a statistically significant cooling induced by the radiative GHG effect (Figure 1c) and a statistically significant warming from the GHG induced SST effect (Figures 1d, 1f). As shown previously in studies of a doubled CO2 climate [e.g., Fomichev et al., 2007] rising SSTs will affect tropospheric temperatures, the tropopause height, and the temperature in the UTLS. In addition, we find a significant impact of the SST effect on the polar lower and middle stratosphere in northern winter. This result confirms earlier simulations with general circulation models (GCMs) by Rind et al.[1990, 1998] and Sigmond et al. .
 The future tropospheric warming and stratospheric cooling cause an enhanced meridional temperature gradient in the UTLS that leads to an intensification of the subtropical jet streams through thermal wind balance. The future increase in the subtropical jets (Figure 2a) is mainly due to rising tropical SSTs (Figure 2f). In connection with the lifted tropopause (Figure S1), the subtropical jets are able to extend to higher altitudes. In the winter polar stratosphere, EMAC shows a significant weakening of the polar night jet (Figure 2a) up to 0.6 m s−1/dec at polar latitudes in 10 hPa that is induced by rising tropical SSTs (Figure 2f). This deceleration of the polar night jet exceeds the weak and nonsignificant signal from transient CCMVal simulations [SPARC CCMVal, 2010], which may be due to the shorter period in the CCMVal trend (up to 2079) and the considered height range (only up to 20 hPa). However, individual CCMs and CMIP5 models, e.g., in McLandress and Shepherd  and Bunzel and Schmidt , show a weakening of the polar night jet that extends down significantly into the troposphere and is of comparable magnitude to EMAC. Tropical SSTs also affect the upper stratosphere and mesosphere, where they strengthen the low latitude summer easterlies (Figure 2f). The radiative GHG effect (Figure 2c) leads to a significant enhancement of the polar night jet in the middle and upper stratosphere, opposite to the zonal wind response to SST changes (Figure 2d). The future decline in ODS has only a minor effect and causes a slight significant strengthening of the stratospheric summer easterlies (Figure 2e), in response to the temperature changes from ozone recovery.
 This analysis has shown that, except for the slight counteracting GHG effect in the middle stratosphere, future zonal wind changes in the troposphere and middle atmosphere will be dominated by the increase in tropical SSTs. The importance of the SST effect was also emphasized by Sigmond et al. , who found a weakening of the polar night jet as a response to tropospheric CO2 doubling, which is equivalent to a CO2 induced SST change. However, while Sigmond et al.  attribute the intensification and upward extension of the subtropical jet mainly to the middle atmospheric GHG effect, our results show that it is entirely due to the effects of tropical SSTs, as also reported in Garny et al.  and Graff and LaCasce .
3.1.2 Changes in the Residual Mean Mass Stream Function, Tropical Upwelling and the Tropical Upward Mass Flux
 Figure 3 shows the response of the RC to the external forcings of Table 2 in terms of changes in the residual mean mass stream function Ψ. The streamlines of the residual mass transport are defined to be positive (negative) for clockwise (counter-clockwise) circulation, i.e., from the tropics to the North (South) Poles. Hence, positive changes in the NH and negative in the SH represent a strengthening of the mass transport, respectively. For the future, an increase in Ψ is visible in both hemispheres, reaching into the upper stratosphere, except for the polar regions in the SH lower stratosphere (Figure 3a). This intensification of the RC in northern winter is a result of increasing GHG concentrations that impact the lower and middle stratosphere by enhanced SSTs (Figure 3d) and the upper stratosphere by their direct radiative effect (Figure 3c). Extratropical SSTs cause a slight but statistically significant decrease of Ψ below 70 hPa in both hemispheres (not shown). In the SH, the future decrease in ODS counteracts the strengthening of Ψ by the other forcings (Figure 3e), which is consistent with transient CCM simulations of McLandress et al. . They found an increase of both the SH stratospheric wave drag and the RC in DJF for the ozone hole period, followed by a projected decrease for the future ozone recovery. This can be explained by the radiative and dynamical effects of the ozone depletion, leading to a more stable and persistent Antarctic vortex in spring [e.g., Langematz et al., 2003] and a shift of the planetary wave driving toward summer. Moreover, idealized GCM studies by Rind et al.  indicate an increase of the RC due to a reduction in the stability of the Antarctic lower stratosphere, and consequently enhanced upward propagation of planetary wave activity which increases Ψ. In the future, when decreasing ODS will lead to a recovery of Antarctic ozone, the change in the RC due to the ozone depletion will reverse, leading to the projected decrease of Ψ (Figure 3e).
 An intensification of Ψ is connected to a strengthening in tropical upwelling and extratropical downwelling, as illustrated in Figure 4 with the residual mean vertical velocity at 70 hPa (∼18 km) for the simulations in Table 1. The reference simulation (black) shows the weakest upwelling in the tropics, lying outside of one standard deviation of the other simulations. Differences between the sensitivity studies under future conditions are small, as changes in tropical SSTs that are included in all of them account for the largest change in tropical upwelling. Moreover, a future shift in the TLs toward the inner tropics of about 5° in the NH and 2° in the SH, which corresponds to a narrowing of the tropical upwelling region for the future, is detectable in our analysis. Future changes in extratropical downwelling are more difficult to assess due to the large interannual variability in NH winter. In both hemispheres, all future simulations consistently indicate stronger downwelling at mid-latitudes (45°–70°S and 35°–60°N) than the 2000 reference run, dominated by the influence of SST changes. A second region with increased future downwelling emerges at northern polar latitudes (60°–90°N). A comparison with the simulation using high levels of ODS (year 2000) (i.e., no ozone recovery, red) shows that polar ozone increase enhances the downwelling over the Arctic. At the same time, polar downwelling will weaken over Antarctica in summer due to ozone recovery, in agreement with Rind et al.  and McLandress et al. . In DJF, future SST changes have no distinguishable effect on polar downwelling over Antarctica.
 Figure 5 shows the relative changes in (left), (middle) and (right) due to the different forcings. In the future, (black) significantly increases throughout the whole stratosphere, with a maximum of about 2% /dec at 50 hPa. This can primarily be attributed to significant changes in the NH extratropical latitudes. Changes in are weak and affect the lower stratosphere only. At altitudes below 10 hPa, GHG induced changes in tropical SSTs are responsible for the significant strengthening of , whereas in the upper stratosphere, the direct GHG effect accounts for about two thirds of the total change in . The future decline in ODS induces a small negative change in in the lower stratosphere, which is related to a decrease of (Figure 5, right). It is, however, more than compensated by rising GHGs and SSTs. A weak increase in due to ODS has only a minor effect on the tropical upwelling (Figure 5, middle).
 The changes in with a maximum increase in the lower stratosphere, weaker increases in the middle and again stronger flux increases in the upper stratosphere, lie in the range of the multi-model mean calculated from eight CCMs within the CCMVal initiative [e.g., Butchart et al., 2010; SPARC CCMVal, 2010]. In EMAC, tropical upwelling intensifies within a narrower upwelling region, which has also been found by Li et al. . Intensified tropical upwelling and extratropical downwelling are consistent with a strengthened RC, which corresponds to tropical cooling and extratropical warming in the lower stratosphere (Figure 1d). The separation of the changes in tropical upwelling by the applied forcings in our study has revealed that the increase of in the lower and middle stratosphere is primarily caused by rising tropical SSTs. These results are in qualitative agreement with, e.g., Fomichev et al. , Oman et al. , and Garny et al. . However, slight quantitative differences exist that may be due to the different model domains and time periods under consideration. Moreover, we find that the radiative effect of rising GHGs becomes more important in the upper stratosphere. The counteracting reduction in lower stratospheric upwelling by declining future ODS agrees with a correlation analysis done by Oman et al.  and transient single-forcing simulations by McLandress et al. .
3.2 Which Waves Are Responsible for the Changes?
 This section addresses the question, which waves are driving the simulated RC changes in response to the different forcings.
 Figure 6 shows the residual mean mass stream function (Ψ) and the contributions to Ψ by the driving of different wave types for the reference simulation (2000) at 70 hPa. Ψ has a maximum at the TLs, decreases toward high latitudes and vanishes at the poles. Resolved and unresolved wave components add up to Ψ. Resolved waves (“EPF”) are separated by time into stationary (“EPF stat”) and transient (“EPF trans”), and by space into planetary (“EPF plan”) and synoptic (“EPF syn”) components (see section 2.2), each pair adding up to the total resolved wave forcing. Unresolved waves consist of OGWs and NGWs. The most important contributor to Ψ are the resolved waves (blue), controlling a large amount of the circulation at all latitudes. Of those, transient waves (blue, dash-dotted) are dominant at low latitudes equatorward of 40°N and in the SH, whereas stationary waves (blue, dashed) explain Ψ north of 60°N. Except for the SH subtropics, planetary waves (green, dashed) contribute strongest to Ψ. In the NH tropics and subtropics, synoptic waves (green, dash-dotted) become important too. This means that transient and synoptic waves mainly propagate into the tropical lower stratosphere, as also supported by the Eliassen-Palm (EP) flux vectors for the reference simulation (not shown). Between 30°N and 50°N OGWs (red) are an important additional driver of the RC. In the SH subtropics, they are crucial for the formation of the summer hemispheric part of the winter circulation. NGWs (yellow) play a minor role in the lower stratosphere and become more important above 5 hPa (not shown). Compared to the annual mean Ψ is by about 80 kg m−1 s−1 larger in the northern winter and by 30 kg m−1 s−1 smaller in the southern summer hemisphere. The contribution of the different waves to Ψ is nearly the same as for DJF. In the SH subtropics, planetary waves become more relevant in the other seasons, while OGWs have their strongest impact in DJF (not shown).
 Based on the difference between Ψ at the NH and SH TLs, the relative contributions from different waves to between 70 and 1 hPa for the reference simulation are shown in Figure 7. Resolved waves (blue, solid) drive up to 90% of in the lower and around 55% in the upper stratosphere. In the lower stratosphere, contributions from transient synoptic and planetary waves dominate the tropical upwelling. With increasing altitude the driving by synoptic waves decreases. Stationary planetary waves have their strongest effects in the middle stratosphere, while their transient components are the most important drivers in the upper stratosphere. In the middle and upper stratosphere, below about 2 hPa OGWs (red) and NGWs (yellow) become equally important drivers of , contributing about 20% each.
 The following Figures 8 and 9 show future changes in Ψ and in DJF. As a result of all forcings, the residual mean mass flux (black) increases at the three selected latitudes by about 1.4 to 1.8%/dec. The largest absolute future increase in Ψ (black) occurs in the subtropics. Annual mean changes reach up to 2.5%/dec in the subtropics of both hemispheres. We find that in DJF, about one third of the increase is due to enhanced activity of resolved waves (blue, +0.45%/dec), primarily transient planetary waves, while the impact of synoptic waves will decline in the future. Two thirds of the total subtropical increase in Ψ will arise from OGWD (red, +0.95%/dec). NGWs (yellow) are of minor importance. The large fraction of OGWD is a special feature of the NH winter. Annual mean future changes at 30°N are primarily explained by resolved waves with major contributions from transient planetary waves (not shown). At 50°N, stronger drag from both resolved waves (+0.55%/dec) and OGWs (+0.80%/dec) will contribute to the strengthening of Ψ in DJF, as well as in the annual mean (EPFD: 0.80%/dec, OGWD: 0.70%/dec). Like in the subtropics, transient (planetary and synoptic) waves will become more important while the stationary components will decrease. In the SH subtropics, OGWs contribute only in the respective winter season to the future change in Ψ (not shown). At 70°N, the increase in Ψ by about 3.2% is explained by resolved stationary waves. Annual mean changes reach 0.9%, with stationary and transient planetary waves accounting for the largest amount. In SH high latitudes, Ψ decrease in DJF as well as in the annual mean.
 In the next step, we will identify which forcings will cause the total future changes in the different wave types. As already shown in Figure 5, GHG induced tropical SST changes drive below 10 hPa while the GHG effect accounts for mass flux changes at higher levels. Figure 9 shows the wave contributions to the total future changes (Figure 9, left) in between 70 and 1 hPa, as well as the changes from the GHG effect (Figure 9, middle) and the SST effect (Figure 9, right). In the lower stratosphere between 50 and 10 hPa the increase in is caused by an enhanced activity of transient planetary waves and by OGWs (consistent with Figure 8). The increase in both wave types occurs in the SST simulations only, indicating the dominant role of the SST effect on the future wave forcing. In addition, tropical SSTs slightly enhance the contribution of synoptic waves (green, dot-dashed) to the lower stratospheric . In contrast, in the upper stratosphere the increased total is solely due to enhanced GW activity. Here the GHG effect will lead to enhanced OGW activity while the SST effect will induce enhanced NGW activity mainly driven by extratropical SSTs. Both, OGWs and NGWs will contribute equally to the future increase in in the upper stratosphere. The GHG effect causes a slight weakening of the tropical upward mass flux by reduced transient planetary wave activity (Figure 9, middle).
 To investigate in more detail the role of resolved waves on the future RC, Figure 10 shows the total future change in the EP-flux and its divergence (Figure 10a) and the respective changes through the GHG effect (Figure 10b) and the SST effect (Figure 10c) for the NH. The individual contributions to the total changes in Figure 10 by planetary and synoptic as well as stationary and transient wave components are shown in Figure 11 for the SST effect and in Figure 12 for the GHG effect. We find a future increase in wave generation at the surface at mid-latitudes and in the middle troposphere at low latitudes (Figure 10a) particularly from stationary waves (Figure 11c). The changes in SSTs increase the land-sea heating contrast and therefore the generation of stationary waves at the surface which is consistent with Garcia and Randel . In the tropics, quasi-stationary wave generation is connected with changes in latent heat release through deep convection as shown in Deckert and Dameris . The changes in EP-flux divergence and the overlying convergence through transient waves (Figures 11a, 11b, 11d) can be associated with an upward shift of the regions of intensified wave breaking in line with the height change of the tropopause in a future climate with higher (tropical) SSTs (cf. Figure S1). In the NH lowermost stratosphere around 100 hPa, a broad area of anomalous EP-flux divergence, centered at mid-latitudes (Figure 10a) is associated with an increase in vertically propagating stationary planetary waves into the stratosphere at mid-latitudes and southward propagating transient planetary and synoptic waves (Figure 11). These changes are triggered by the SST effect (Figure 10c) and are consistent with the changes in that are driven mainly by transient waves (Figures 9a, 9c).
 As the changes in the refractive properties improve the upward propagation of stationary planetary waves in the lower stratosphere at mid-latitudes south of 50°N (see Figure S2d), more stationary planetary waves reach the middle and upper stratosphere and cause an EP-flux convergence through wave dissipation (Figures 11a, 11c). The GHG effect (Figure 10b) is less relevant in the lower stratosphere, as it does not lead to enhanced wave propagation from the troposphere into the stratosphere. However, in the upper stratosphere and lower mesosphere, it contributes to the future anomalous EP-flux convergence (Figure 12). Associated with the positive zonal wind anomaly in the mid-latitude upper stratosphere and lower mesosphere (cf. Figure 2c), enhanced poleward refraction of stationary and transient planetary waves will take place (Figures 12a, 12c, 12d). This is supported by the analysis of the changes in the refractive index for stationary planetary waves (Figure S2c).
 In summary, our analysis of the wave contributions to Ψ and for the reference simulation shows that resolved planetary waves dominate the RC at all latitudes and throughout the stratosphere. Drag from synoptic waves is strongest in the lower stratosphere, but its contribution to the RC decreases with altitude to ∼10% at 30 hPa. Parameterized OGWD is particularly important in the subtropical lower stratosphere in winter where it contributes about one third to the RC, and up to 20% to in the middle and upper stratosphere. The great importance of the OGWD for the formation of the summer hemispheric low-latitude part of the winter circulation is captured in EMAC in good agreement with Okamoto et al. .
 The future increase in the RC in NH winter is by two thirds due to OGWD and by one third to resolved transient planetary wave drag in the subtropical lower stratosphere. This result of our multi-year timeslice simulations supports the 77% increase in by OGWD that Butchart et al.  derived from five transient CCM projections. At mid-latitudes OGWs and transient planetary and synoptic waves contribute equally to the increased RC while the high latitudes will feature an enhancement of stationary planetary waves. These results are in good agreement with transient simulations shown in McLandress and Shepherd . The increase in synoptic and planetary transient wave drag in the subtropics was explained by Shepherd and McLandress  by an upward shift of the critical layers for Rossby wave breaking into the lower stratosphere associated with the intensification of the subtropical upper tropospheric jet (cf. Figure 2). The same mechanism is responsible for the growing future role of OGWD evident in the subtropical middle stratosphere [e.g., Li et al., 2008; Garcia and Randel, 2008; McLandress and Shepherd, 2009; Okamoto et al., 2011].
 Our analysis of the forcings of the above described changes in the RC and the relevant wave drag components clearly revealed the dominant impact of the GHG induced SST effect in the lower stratosphere. The increase in transient planetary and synoptic wave propagation into the lower stratosphere and toward the equator could be attributed to the effects of rising tropical SSTs. Similar to Garny et al. , we find an SST induced increase in stationary planetary wave propagation into the subtropical lower stratosphere which was explained by Deckert and Dameris  and Calvo and Garcia  as a response to GHG induced SST anomalies. However, this effect occurs in the summer hemisphere only and is hardly evident in the NH tropics in DJF. Our results further show that increases in tropical SSTs also affect the upper stratospheric part of the RC by an increase in the resolved wave drag.
 The GHG effect is of minor importance in the lower and middle stratosphere where no significant change in the upward propagation of resolved waves from the troposphere is found. This result is confirmed by an analysis of the refractive index (Figure S2c): The SST effect favors stationary planetary waves of wave numbers one and two to propagate more likely into the stratosphere at mid-latitudes, while there are no lower stratospheric changes in refractive properties for the GHG effect. Hence, upper stratospheric changes in resolved wave drag induced by increased GHGs are more likely associated with enhanced wave refraction than with wave propagation from the troposphere.
 In addition, both the SST and GHG effects induce changes in in the upper stratosphere by enhanced drag from OGWs and NGWs as the propagation conditions for gravity waves improve through changes in zonal mean zonal winds.
3.3 Inclusion of Mixing Processes
 As stated in the Introduction, the RC is only one part of the BDC. The second contributor is quasi-horizontal mixing that is particularly important in the regions of the weak transport barriers in the UTLS [Neu and Plumb, 1999]. Here we employ the concept of AoA, introduced in section 2.2.2 as a further diagnostic to investigate future changes in the BDC.
 Figure 13 shows the mean AoA for the NH winter season in the 2000 simulation (colored shading) and retrieved from MIPAS-SF6 measurements, averaged for the years 2002 to 2010 (contour lines). Mean AoA increases with latitude and height as air is transported from the entry region at the tropical tropopause to the polar upper stratosphere and lower mesosphere. Oldest mean AoA appears in the winter lower mesosphere, while in the respective summer hemisphere, mean AoA is lowest (not shown). In the UTLS, the flatter distribution of mean AoA in the observations indicates stronger mixing processes than simulated with the EMAC CCM. Above ∼35 km altitude, MIPAS shows considerably higher mean AoA than the model. This is due to descending mesospheric air that has been chemically processed and reaches the stratosphere with lower SF6 concentrations than an inert tracer. It leads to an overestimation of mean AoA at mid-latitudes and high latitudes [Stiller et al., 2012]. This so-called “apparent age” does not exist in the model because the mesospheric sink of SF6 due to electron capture processes is not included. Outside the polar vortex, mean AoA in the upper stratosphere is expected to be less than 8 to 9 years [Stiller et al., 2012]. Altogether, EMAC is able to reproduce the observed mean AoA in the lower and in the most part of the middle stratosphere.
 Figure 14 shows future changes in mean AoA separated by the external forcings (cf. Table 2) in the lower (70 hPa) and in the upper stratosphere (1 hPa). At both heights a global future decrease in mean AoA (Figure 14, black) is found that corresponds to an acceleration of the mass transport from the tropics to high latitudes and is in line with a strengthening of the BDC. In the lower stratosphere (Figure 14, left), small decreases in mean AoA occur in the tropics, entirely due to rising tropical SSTs (green). At high latitudes changes approach one month/dec. Here decreases in mean AoA due to tropical SSTs are partly compensated by the effects of ODS (red) in the NH and GHGs (yellow) in the SH. The decrease in total mean AoA intensifies with altitude (Figure 14, right). While in the lower stratosphere, the reduction in mean AoA is symmetric to the equatorial entry region of air parcels and accumulates toward the poles, strongest changes in the upper stratosphere occur at mid-latitudes to high latitudes in winter. The future reduction in total mean AoA at 1 hPa is determined by the rising SSTs. ODS additionally reduce the mean AoA at high northern latitudes, while the atmospheric GHG effect counteracts. Changes in tropical SSTs are the dominant forcing south of 60°N, while extratropical SSTs (indicated from the difference between global and tropical SSTs) have a stronger effect on the reduction of the mean AoA in the NH polar vortex.
 As is obvious from Figure 4, increases in extratropical downwelling occur at mid-latitudes of both hemispheres (45°–70°S and 35°–60°N), while in the polar region, it increases in northern winter only (60°–90°N). As upwelling is restricted to the tropics for both branches, extratropical changes in mean AoA allow us to determine changes in the downwelling branches. The two regions at northern mid-latitudes and polar latitudes with a future increase in downwelling (cf. Figure 4) indicate a potential change in both the lower and the deep branches of the BDC. Figure 15 shows changes in mean AoA in the stratosphere and lower mesosphere at 35°–60°N (Figure 15, left) and 60°–90°N (Figure 15, right), respectively. In both regions, future mean AoA (black) will strongly decrease over the whole vertical range. This confirms the above finding of a future increase in the BDC which was based on the RC only. At northern mid-latitudes (Figure 15, left), mean AoA shows a decrease that rapidly grows with altitude in the lower stratosphere up to 40 days/dec at 30 hPa, consistent with the increase in transient planetary and synoptic wave activity in the lower stratosphere (Figure 9) that leads to stronger mixing between lower latitudes and mid-latitudes. As this change is equatorially symmetric (not shown) and confined to the lower stratosphere, it is indicative of an intensification of the lower branch of the BDC in the future. Figure 15 shows that tropical SST changes (green) explain the total future signal, while changes in GHGs (yellow), ODS (red), and extratropical SSTs do not contribute significantly to the decrease in mean AoA. This suggests that changes in the shallow branch are driven by the SST effect only. At higher latitudes (Figure 15, right), mean AoA decreases more continuously with altitude from 20 days/dec at 100 hPa to around 55 days/dec in the lower mesosphere. These changes can be associated with the upper branch of the BDC that is driven by waves that propagate upward in the westerlies of the winter hemisphere. Also at high latitudes, the changes in tropical SSTs are the main forcing of the future decrease in mean AoA in the lower and upper stratosphere. In the lower stratosphere, the SST effect will be slightly reduced by polar ozone recovery, while the GHG effect reduces the mean AoA. At the upper levels, both tropical and extratropical SSTs are responsible for a further decrease in mean AoA, as is also obvious from the enhanced dissipation of planetary waves in the polar upper stratosphere in Figure 11a. The GHG effect counteracts the upper stratosphere SST effect, while ODS lower the mean AoA. Hence, changes in the deep branch are caused by the interaction of different external forcings.
 In summary, the results obtained from the AoA analysis are consistent with the changes in the RC discussed above, both suggesting a strengthening of the BDC in the future. This will include both the shallow and the upper stratospheric deep branch of the BDC during the 21st century. The future decadal changes in mean AoA in EMAC are comparable in magnitude with those derived from a linear tracer in a transient CCM simulation for the past (1965–2006) with external forcings considered [Garcia et al., 2011]. The large rates of decrease found in the extratropical lower stratosphere, which indicate an enhanced transport of air from the tropics into higher latitudes, confirm the results of Butchart et al.  who analyzed transient future projections from CCMs.
 Rising tropical SSTs were found to mainly influence future changes in mean AoA over a large altitude range from the lower stratosphere up to the lower mesosphere. Oman et al. , who correlated changes in mean AoA with changes in SSTs and ODS in future CCM simulations, came to similar conclusions on the overall decrease in mean AoA by the SST effect and a slight increase in mean AoA due to future ozone recovery. However, our results show enhanced downwelling in the lower stratosphere through increased planetary wave drag.
 In this study we used multi-decadal timeslice simulations of present (year 2000) and future (year 2095) climate, performed with the EMAC CCM, to study the future evolution of the BDC. By varying the effective external forcings, like the concentrations of GHGs and ODS, the individual effects of these anthropogenic composition changes on the future mass circulation could be isolated. In particular, the direct radiative effect of increased GHG concentrations (GHG effect) could be separated from the indirect influence of rising GHG concentrations on the tropical and extratropical SSTs (SST effect). Changes in temperature and the zonal wind background state, as well as in the residual mean mass stream function, tropical upwelling, and extratropical downwelling for the individual forcings were examined. The driving mechanisms for the simulated changes in the RC were isolated for individual wave types and forcings. Mixing processes were included by calculating changes in mean AoA for the effective forcings. Hence, our study provides complementary insight on future BDC changes compared to the transient CCM simulations performed within the CCMVal initiative which combined all known forcings of future climate to simulate the most representative scenario. Our results can be summarized as follows:
 In general, we find very good agreement of the projected atmospheric change between 2000 and 2095 in our timeslice simulation including all forcings with results from the transient CCMVal simulations [e.g., SPARC CCMVal, 2010]. EMAC predicts a global warming of up to 0.6 K/dec in the tropical upper troposphere and a cooling of the stratosphere and mesosphere up to −1 K/dec in northern winter. As compensation of a significant warming due to the SST effect and a significant cooling from the GHG effect, the lower polar winter stratosphere does exhibit a slight insignificant warming. The SST induced polar warming is associated with more stationary planetary waves that propagate into the upper stratosphere and induce an enhanced EP-flux convergence through intensified wave breaking. This is also confirmed from the changes in the refractive properties. The tropospheric subtropical jet significantly intensifies up into the lower stratosphere, while the polar night jet weakens. We further identified an intensification of the RC from changes in the residual mean mass stream function, tropical upwelling, and the tropical upward mass flux. The increase in in the lower stratosphere is largely explained by stronger forcing from transient planetary waves. However, the role of OGWD will increase in the future, particularly in the subtropical stratosphere in DJF, as also pointed out by McLandress and Shepherd  and Okamoto et al. . In the upper stratosphere and mesosphere, parameterized GWs were found to drive most of the strengthening signal. The changes in future wave drag simulated in EMAC are consistent with the changes in critical layers, as proposed in Shepherd and McLandress .
 The consideration of potential changes in mixing processes confirms the result of a future intensification of the RC and thus indicates an intensification of the BDC as a whole. In our simulations, future changes in mean AoA for DJF are negative, consistent with results from Oman et al.  and Butchart et al.  for the annual mean. EMAC simulates a decrease in mean AoA of about 20–40 days/dec in the lower stratosphere at NH mid-latitudes and high latitudes. In the upper stratosphere, the mean AoA decrease ranges between 1 and 2 months/dec, depending on altitude and latitude.
 We found that the GHG induced rise of tropical SSTs triggers the largest amount of change in the BDC identified for the 2095 future simulation, particularly in the lower and middle stratosphere. SST changes are responsible for the warming of the polar lower winter stratosphere and the weakening of the stratospheric polar night jet, consistent with Sigmond et al. . Due to the lifting of the tropical tropopause and the associated enhanced meridional temperature gradient, SSTs also induce a significant intensification of the subtropical troposphere jet up to the middle stratosphere, as also reported by, e.g., Graff and LaCasce . Future tropical SSTs enhance the RC in the lower stratosphere, as obvious in the intensified residual mean mass stream function. Associated with this is an increase in in the lower and middle stratosphere below 10 hPa, which is caused mainly by enhanced activity of transient planetary waves and OGWs. At mid-latitudes, more transient waves will propagate into the subtropical lower stratosphere. Stationary planetary waves will propagate upward and dissipate in the high latitude stratosphere. In addition, tropical SSTs also affect the extratropical upper stratosphere, where they lead to both enhanced dissipation of planetary waves as well as OGWD and NGWD. Rising tropical SSTs were also found to determine future changes in mean AoA over a large altitude range from the lower stratosphere up to the mesosphere.
 The direct radiative effect of increasing GHG concentrations leads to a global cooling of the stratosphere and mesosphere. The GHG effect accounts for an increase in the RC in the upper stratosphere and lower mesosphere, where significant changes in resolved wave forcing were found. Here enhanced convergence arises from a poleward refraction of planetary stationary and transient waves caused by stronger westerlies in the lower mesosphere at mid-latitudes. In addition, OGWD in the upper stratosphere will increase. Consistently, changes in due to the GHG effect are strongest in the upper stratosphere.
 The future ODS decline leads to a slight significant weakening of the SH summer circulation, in response to the temperature changes resulting from ozone recovery. It has its major impact on the polar downwelling in the lower stratosphere where it counteracts the GHG induced downwelling increase in SH summer and slightly enhances the polar downwelling in NH winter.
 In summary, our study revealed a significant strengthening of the BDC as response to future climate change. This result was confirmed by both an analysis of the RC as well as of mean AoA. The effect of rising GHG concentrations was identified as main forcing of future BDC changes, while the decline in ODS and the associated ozone recovery have only a minor effect. Our analysis of mean AoA revealed that in a future climate we expect an intensification of the lower shallow branch as well as of the upper deep branch of the BDC.
 This work was produced in the DFG Research Unit FOR 1095 “Stratospheric Change and its Role for Climate Prediction” (SHARP) under the grants LA 1025/13-1 and LA 1025/14-1, funded by the Deutsche Forschungsgemeinschaft (DFG). We are grateful for the support from Blanca Ayarzagüena, Anne Kubin, and Markus Kunze and we thank three anonymous reviewers for helpful comments on the paper. For providing the computing time, we thank the North-German Supercomputing Alliance (HLRN). Mean AoA from MIPAS was kindly provided by Gabriele Stiller.