A recent model study shows that in response to an increasing aerosol forcing, the Southern Hemisphere (SH) ocean circulation, including the Agulhas outflow, its retroflection, and the entire subtropical gyre circulation intensify and shift polewards. Are these oceanic responses manifested in the SH atmospheric circulation? We demonstrate that as a result of the poleward shift, maximum sea surface temperature (SST) gradients, midlatitude storms and the westerly jet shift southward, intensifying the trend of the southern annular mode (SAM). Because the response of the atmospheric circulation to the underlying oceanic anomalies is equivalent barotropic, a poleward shift and intensification in zonal wind and vertical velocities are generated from the atmosphere-ocean interface to the middle troposphere. These atmospheric circulation responses, in turn, reinforce the ocean circulation changes. This constitutes an air-sea positive feedback. Our results illustrate an impact of Northern Hemisphere (NH) aerosols on the SH atmospheric circulation trends through a SH ocean circulation response.
 Most aerosols reside in the NH, and as such, studies of the impacts predominantly focus on the atmospheric circulation in the NH and the tropics. These include a strong influence on regional climate, such as summertime floods and droughts in China [Menon et al., 2002], a weakening of the South Asian monsoon [Ramanathan et al., 2005], and an increasing rainfall trend over northwest Australia [Rotstayn et al., 2007]. Like volcanic aerosols [Church et al., 2005; Gleckler et al., 2006], anthropogenic aerosols have a strong cooling signature in the globally averaged oceanic heat content [e.g., Cai et al., 2006]. Further, in response to a stronger NH surface cooling effect, there is an oceanic heat redistribution from the SH oceans to the NH, through a strengthened Global Conveyor [Gordon, 1986] associated with the North Atlantic Deep Water Formation (NADWF). This leads to a reduction in the SH oceanic heat content comparable to that in the NH. As a result of the pan-oceanic adjustment, the modeled SH ocean circulation trends over the past decades are rather different with and without increasing anthropogenic aerosols. Is the difference in the SH ocean circulation trends manifested in the SH atmospheric circulation trends? This is the main issue we will address in this paper. We compare two sets of 20th century simulations in a coupled ocean-atmosphere global climate model. We show that increasing NH aerosols intensify the SAM trend and modify its structure.
2. Model and Experiments
 The atmospheric model used in this study is a low-resolution, flux adjusted (spectral R21) version of the CSIRO atmospheric general circulation model [Rotstayn et al., 2007]. The adjustment is carried out in terms of momentum, heat, and freshwater fluxes. The R21 model has 18 hybrid vertical levels and a horizontal resolution of approximately 5.6° in longitude and 3.2° in latitude. The Cox-Bryan oceanic model has the same horizontal resolution as the atmospheric model, with 21 vertical levels. The atmospheric model has a comprehensive aerosol scheme in the simulations for the period 1871 to 2000. Other forcings included are those due to changes in long-lived greenhouse gases, ozone, volcanic aerosol and solar variations. An eight-run ensemble with all of these forcings (ALL ensemble) and a further eight runs with all forcings except those related to anthropogenic aerosols (AXA ensemble) are performed. The AXA differs from ALL in that the anthropogenic emissions of aerosols are fixed at their 1870 levels. The differences between the two ensembles averages, ALL minus AXA (referred to as “aerosol-induced”), were used to deduce the anthropogenic aerosol effect. We focus on the period since 1941. We also deploy a multi-century control climate to assess the significance of the difference between ALL and AXA, as described by Cai et al. .
3. Aerosol-Induced SH Ocean Circulation Trends
Cai et al.  provide a detailed description of the SH ocean response to anthropogenic aerosols. Here, we emphasize that the origin of the SH ocean response lies in the heat transport from the SH ocean to the NH ocean. There is no aerosol-induced trend in the net heat into or out of the SH oceans (Figure 1a, positive into the ocean), yet a significant decreasing temperature trend of the SH ocean surface (Figure 1b) and ocean interior (Figure 1c) is induced. If these oceanic circulation trends were forced by the atmosphere, there would be a trend in the net heat flux, which is inferred and shown in Figure 1c (orange curve). Thus, the oceanic cooling has to be conducted via a heat transport to the NH oceans across the equator. The stronger cross-hemispheric heat transport in ALL is carried out by an intensifying upper branch of the Conveyor, with an increasing Indonesian Throughflow, a stronger Agulhas outflow and retroflection, and strengthening flows toward the North Atlantic (Figures 2a and 2b). The heat for supporting the stronger NADWF in ALL is derived mainly along the Conveyor pathway in the latitude band 10°S–35°S. By contrast, in AXA, there is an accumulation of heat in the latitude band 40°S–50°S (Figure 2b).
 In association, a stronger poleward heat transport in the southern midlatitude Indo-Pacific is generated in ALL, so as to support a stronger heat transport via the Agulhas outflow, through the South Atlantic, eventually to the North Atlantic [Cai et al., 2006, Figure 2a]. That in ALL a stronger NADWF is associated with a stronger SH subtropical gyre is consistent with recent findings about their linkage [Speich et al., 2002; Blanke et al., 2001].
Figures 2a and 2b illustrate another important feature: the Agulhas outflow, its retroflection and the entire SH subtropical gyre do not simply increase in strength along the pathways in AXA. Instead, they shift poleward, with the largest change in the Indian Ocean. Indeed, in ALL, the cooling north of 35°S and the maximum warming to the south (Figure 2a) are a direct consequence of a poleward shift [Cai et al., 2007]. This is further illustrated in Figures 2c and 2d, where in ALL the streamfunction trends have a maximum at about 45°S, compared with a maximum at 42°S in AXA. The difference has been shown to be statistically significant at the 95% confidence level by Cai et al. [2006, 2007] using the control experiment, taking into account each of the ensemble members.
 That the origin of the stronger poleward shift in ALL lies in the response of the SH oceans to the NH cooling is further supported by zonal mean heat flux trends: in the latitude band where the zonal-mean warming is largest (43°S) the ocean loses heat to the atmosphere (figure not shown). The poleward shift and intensification of the subtropical gyres is so powerful in generating warming that it leads to a heat loss to the atmosphere. In AXA, because the poleward shift is weaker, there is a heat gain where the zonal-mean warming is largest (40°S).
4. Aerosol-Induced Poleward Shift of the SH Atmospheric Circulation Trends
4.1. Trends in ALL and AXA
 Because the SH oceanic circulation response in ALL is not a mere intensification along the circulation paths of those in AXA but involves a poleward shift, the SH atmospheric circulation responds strongly with a similar poleward movement and strengthening. This feature is seen in the wind stress curl (Figures 3a and 3b), which, in turn, has the effect of maintaining the ocean circulation trend. Since wind stress curl is dominated by meridional gradients of zonal wind stress, a poleward shift and intensification in ALL from those in AXA (Figures 3c and 3d) is produced. We shall discuss the detailed process in section 4.2.
 A well-known feature of the mid- to high latitude atmospheric circulation is that its response to underlying oceanic anomalies has an equivalent barotropic vertical structure [Peng et al., 1995; Baines and Cai, 2000; Simmonds, 2003]. This nature of the response is illustrated by a similar poleward shift and intensification in zonal winds at 500 mb (Figures 3e and 3f). The trend in ALL is greater than that in AXA, and the location of maximum trends in ALL is approximately 3° further south, similar to the difference at the surface. The shift is most obvious in the Indian Ocean sector, because it is primarily driven by the shift of the Agulhas outflow and its retroflection. Further, trends in both ensembles are SAM-like. In AXA, the SAM trend is primarily caused by Antarctic ozone depletion [Thompson and Solomon, 2002]. Increasing NH aerosols intensify the SAM trend and shift it polewards.
 That aerosol-induced SH atmospheric circulation trends are mainly driven by aerosol-induced SH ocean circulation changes is confirmed by the zonal maximum meridional temperature gradient located near the surface rather than high in the atmosphere (e.g., stratosphere) as with Antarctic ozone depletion. Further, forcing an atmosphere-only model with a SST trend pattern in the NH from ALL produces no SAM-like trend in the SH, consistent with results from an earlier experiment [Simmonds et al., 1989], in which an atmosphere-only model is forced by SST anomalies similar to aerosol-induced SST changes, with cooling in the western Pacific, and produces no SAM-like response.
4.2. Mechanism for the Difference Between ALL and AXA
 We may understand the associated mechanism in terms of circulation changes in ALL from AXA. As discussed in section 3, the stronger poleward shift and intensification of the SH subtropical circulation is accompanied by a stronger SH midlatitude poleward heat transport. These changes move the location of the maximum SST gradients (Figure 4a) and hence the maximum atmospheric baroclinicity polewards [Inatsu and Hoskins, 2004]. This reduces midlatitude storms as indicated by an increasing mean sea level pressure (MSLP) (not shown) and a sharp decrease in rainfall (Figure 4b) around 43°S, with relatively more events occurring to the south. In association, the westerlies to the north reduce (Figure 4c), taking less heat out of the ocean. By contrast, westerlies strengthen to the south, cooling the ocean as storms activities increase with a decreasing MSLP [Inatsu and Hoskins, 2004; Simmonds and Keay, 2000; Simmonds et al., 2003]. The above process operates while the SH ocean circulation shifts polewards so that the stronger, poleward-shifting westerlies and curls are generated (Figures 3a and 3c) in tandem with the poleward shift of the SH subtropical circulation. The stronger, poleward-shifting curl in ALL in turn reinforces the stronger, poleward-shifting gyre circulation, with ensuing changes in temperature gradients, rainfall, and zonal winds. Thus, there is a positive feedback that is consistent with the classical idea that the ocean forces the atmosphere with SSTs, driving wind changes, which in turn act on the ocean, operating in the midlatitudes [Baines and Cai, 2000].
 Given that the midlatitude westerlies and storms are intimately linked to the Ferrel cell [Hall and Visbeck, 2002], we can understand the above process by examining the response of the Ferrel cell. Midlatitude westerlies represent statistical residuals after zonally averaging almost-compensating northward and southward flows associated with cyclones and anticyclones. At any given time, there are many such weather systems operating. The integrated effect of these weather systems is a poleward heat transport. The process involves the vertical and meridional circulation of the atmosphere: in a zonal-mean plane, to the north of midlatitude storm activities, there is a sinking motion, whereas to the south there is a rising branch, such that near the surface there are flows transporting heat poleward.
 As the westerlies and storms shift polewards, we expect the Ferrel cell to move southward and strengthen. This is indeed the case as shown in Figure 5, which depicts the trends in zonally averaged vertical velocities of the atmosphere. In Figure 5, a positive value indicates a sinking motion, and vice-versa. We see that both the rising and sinking branches are stronger and the centers of these branches are further to the south in ALL (Figure 5a) than in AXA (Figure 5b). Thus, in response to NH aerosols, the SH midlatitude circulation from the ocean to the middle troposphere is intensifying and moving polewards as an entity, with consistent circulation changes. The associated SST anomalies in turn strengthen the feedback process. We refer to this process as the “Ferrel cell-ocean circulation feedback”. The central point is that the evolving SH ocean circulation is an active part in the feedback, which is stronger than previously realised without an active ocean [Sen Gupta and England, 2007].
 We have examined the impacts of increasing anthropogenic aerosols on the SH atmospheric circulation. A previous study has showed that as the SH ocean circulation responds to an aerosol-induced stronger global Conveyor, the Agulhas outflow, its retroflection, and the downstream SH subtropical gyre circulation intensify and move polewards. This ocean circulation response has two effects. Firstly, it generates a stronger SH midlatitude poleward heat transport in the South Atlantic, feeding a stronger northward heat transport to the NADWF region. Secondly, it pushes the location of the maximum SST gradients further to the south. The surface atmosphere circulation responds to the change in SST gradients, and the westerlies and storms all shift polewards, strengthening the SAM trend, and pushing it poleward. The associated wind changes, in turn, reinforce the ocean circulation trends, with a stronger SST trend, in a positive feedback process, referred to as the “Ferrel cell-ocean circulation feedback”. Because the atmospheric responses underlying these oceanic anomalies are barotropic in nature, a poleward shift and intensification of the SAM trend, similar to those at the atmosphere-ocean interface, are generated. An issue arises as to whether the model Conveyor is overly efficient in its response to an aerosol-induced cooling in the NH. This will be investigated in a future study. Nevertheless, our results show that long-term trends in the extratropical ocean circulation can influence the trends of the extratropical atmospheric circulation.
 This work is supported by the Australian Greenhouse Office. We thank Leon Rotstayn and Martin Dix for developing the model and running the experiments, and the reviewers for their helpful comments.