Climate model simulations of past and future climate invariably contain prescribed zonal mean stratospheric ozone. While the effects of zonal asymmetry in ozone have been examined in the Northern Hemisphere, much greater zonal asymmetry occurs in the Southern Hemisphere during the break up of the Antarctic ozone hole. We prescribe a realistic three-dimensional distribution of ozone in a high vertical resolution atmospheric model and compare results with a simulation containing zonal mean ozone. Prescribing the three dimensional ozone distribution results in a cooling of the stratosphere and upper troposphere comparable to that caused by ozone depletion itself. Our results suggest that changes in the zonal asymmetry of ozone have had important impacts on Southern Hemisphere climate, and will continue to do so in the future.
 Here we study the effects of zonally asymmetric ozone on the Southern Hemisphere climate using a high vertical resolution version of the Hadley Centre slab model, denoted HadSM3-L64. This consists of a 50-m-deep mixed-layer ocean and a dynamic and thermodynamic sea-ice model coupled to a 64-level atmospheric model extending up to 0.01 hPa [Gillett et al., 2003].
 In the absence of a suitable three-dimensional gridded observational ozone dataset, we used ozone concentrations from ERA-40. ERA-40 provides global gridded ozone data of adequate quality for a sensitivity study such as this and has been used for a similar study of the Northern Hemisphere [Gabriel et al., 2007]. Ozone is assimilated from TOMS and SBUV data when available and the reanalysis model contains a chemistry parameterization and tracer transport equation [Dethof and Hólm, 2004]. Despite some deficiencies, the ERA-40 ozone describes well the large scale structures of stratospheric ozone such as the formation and break up of the ozone hole [Dethof and Hólm, 2004].
 Annually repeating ozone was prescribed in the model from the 12 month period of July 2000 to June 2001. This period was chosen because October 2000 shows particularly large zonal asymmetry in its ozone distribution (Figure 1a). The greatest zonal asymmetry is concentrated in September–November from 100 hPa to 10 hPa. In other months the zonal asymmetry is minimal. The ozone distribution in July 2000 and June 2001 was very similar, therefore wrapping the ozone data at this point is sensible. The ERA-40 data only extends to 1 hPa at which altitude the ozone shows little zonal asymmetry. Ozone concentrations above this level were based on climatological zonal mean values [Randel and Wu, 1999b]. The ozone hole of 2000 was particularly large in its extent [Thompson and Solomon, 2002]. The ERA-40 data indicate its shape was similar to the shape of the mean ozone hole over the 1990s, but with greater zonal wave number 1 amplitude. The direction of displacement of the ozone hole from the pole is not random, but is determined by planetary wave sources in the troposphere [Baldwin and Dameris, 2007]. Our use of ozone concentration from 2000 therefore allows us to place a plausible upper limit on the size of the response to zonal asymmetry in ozone.
Grytsai et al.  found a 30–40% increase in the amplitude of the wave one structure of the September–November total column ozone field (TOMS) at 65°S from 1979 to 2005, and we find a 55% increase in the amplitude of the wave one structure of the September–November ERA-40 ozone at 50 hPa, 65°S between 1979 and 1999 based on a linear fit of the 1979–1999 data. The zonal asymmetry of ozone in the Southern Hemisphere during the austral spring is associated mainly with the Antarctic ozone hole due to the hole being centered off the pole. Therefore, we would expect significantly smaller zonal asymmetry before the advent of polar ozone depletion in the mid-1970s, but lack the observations to verify this.
 The impact of the zonal asymmetry in ozone was assessed by comparing two simulations, one using the 3-dimensional ozone field described above and the other using the zonal mean of this ozone field (referred to henceforth as 3DO3 and zmO3 respectively). Both were allowed to run for 25 model years with all other forcings held constant, but the initial 5 years of each, during which the model was reaching equilibrium, were discarded. The significance of differences in simulated temperatures, geopotential heights and heating rates were assessed using a two-sample t-test.
 The zonal asymmetries in ozone produce a general cooling at 50 hPa over the Antarctic (Figure 1d). However, surprisingly the maximum cooling is located around 150°E 70°S close to the region of maximum ozone. This longitudinal pattern of response cannot be explained by changes in radiative heating, which tend to produce a warming in the region of ozone maximum. A similar pattern of response is seen in the Northern Hemisphere, with a region of minimum ozone at 10 hPa associated with a local warming [Gabriel et al., 2007]. Comparing Figures 1c and 1d, it is clear that the shape of the polar vortex is altered by the zonal asymmetries in ozone, with the vortex centre shifted closer to the pole.
 In the zonal mean in October, the whole region between 300 hPa and 5 hPa over the Antarctic is significantly colder when three-dimensional ozone variations are prescribed (Figure 2), with a maximum cooling of about 10 K at 10 hPa. Above 1 hPa where the zonal asymmetry of ozone is minimal, the atmosphere warms. In November, when the zonal asymmetry in ozone is weaker, the temperature differences are slightly smaller (maximum cooling of 7 K at 50 hPa) and the transition from cooling to warming occurs at a lower altitude (around 15 hPa): there is an apparent downward propagation of the temperature and geopotential height response in the stratosphere (Figure 3), of a similar nature to that seen in response to ozone depletion [Gillett and Thompson, 2003; Keeley et al., 2007]. As expected, the stratospheric cooling is associated with large decreases in geopotential height, indicating a pronounced strengthening of the polar vortex. The maximum temperature and geopotential height differences seen around 50 hPa in October are slightly larger in magnitude than those simulated in response to zonal mean ozone depletion over the past 30 years [Gillett and Thompson, 2003; Keeley et al., 2007], indicating their potential importance in explaining climate trends. However, perhaps surprisingly considering the similar magnitude of the stratospheric changes, a significant zonal mean response is not seen in the troposphere. Southern Hemisphere climate trends due to changes in the full 3-dimensional ozone field are the subject of further research. While no significant response is seen in lower tropospheric temperature or geopotential height in the zonal mean (Figure 2), a cooling of around 3 K is seen in October at the surface centered around 180°E 65°S over the Ross Sea, coincident with the region of maximum cooling in the stratosphere and the region of maximum ozone concentration.
Figure 2a shows that while the Antarctic stratosphere experiences a cooling in response to the specified 3D ozone, the tropical stratosphere warms. This zonal mean temperature response is associated with a weakening of the Brewer-Dobson circulation (diagnosed using the transformed Eulerian mean streamfunction, not shown). In order to further examine the causes of the zonal mean temperature response, we examined changes in radiative and dynamical heating using the Eulerian zonal mean thermodynamic energy equation. This equation in spherical coordinates (based on Andrews et al. [1987, equation (3.3.2e)]) is given by:
where J is the radiative heating and other variables have their standard meteorological meanings [see Keeley et al., 2007]. Heating rates were calculated using monthly means of T, v, ω, θ and the primed products.
 There was more shortwave heating in regions of higher ozone concentration and less in regions of lower ozone concentration, but in the zonal mean the difference in shortwave heating between the 3DO3 and zmO3 simulations is close to zero throughout the stratosphere, as expected, since the shortwave heating scales approximately linearly with the ozone concentration [Forster and Shine, 1997].
 Longwave heating is influenced both by the distribution of ozone and the change in temperature. Examination of Figure 3 indicates that the longwave heating changes are largely driven by the temperature change, with colder regions exhibiting less emission in the longwave, i.e. a positive change in longwave heating. The difference in zonal mean longwave heating for 3DO3 (Figure 3c) is representative of the total radiative heating difference, since the shortwave heating difference is close to zero. The radiative heating difference clearly does not explain the temperature differences (Figure 3a), indicating that these must be driven by dynamical changes.
 The zonal mean dynamic heating analysis shows strongly contrasting effects associated with changes in meridional eddy heat flux (first term of equation (1)) and vertical velocity (third term of equation (1)). However, the zonal mean total dynamic heating difference, averaged over the Antarctic region (Figure 3d), is dominated by changes in the meridional eddy heat flux. The total dynamic heating difference is of opposite sign to, but dominates, the total radiative heating difference (Figure 3). The combination of the two explains the temperature difference reasonably well, although the balance is not exact, probably because heating changes are calculated from monthly means. The October–November total heating difference dipole corresponds to a delay in the final warming of the stratosphere, with the delay occurring earlier higher in the stratosphere.
 These results indicate that the zonal mean temperature differences are caused by differences in the dynamics rather than due directly to radiative differences. This reduction in dynamical heating of the polar vortex in spring associated with the zonal asymmetries in ozone is associated with a weakened Brewer-Dobson circulation and reduced wave driving of the stratosphere. Preliminary studies using a simplified global atmospheric circulation model with a fixed wave-one-structured heating perturbation applied at the poles also show the importance of dynamics on the resulting temperature field (A. Charlton-Perez, personal communication, 2007). The equilibrium longitudinal position of the temperature minimum and relative maximum is dependent on the mean strength of the zonal wind, such that a response in antiphase with the heating perturbation is possible. The flux of wave activity into the stratosphere is very sensitive to the position and shape of the stratospheric polar vortex [Scott and Polvani, 2004; Esler and Scott, 2005], therefore we suggest that small changes in the position and shape of the vortex induced by zonal asymmetries in the ozone distribution lead to a large reduction in the upward flux of wave activity from the troposphere [Nathan and Cordero, 2007], which in turn strengthens the polar vortex.
 Prescribing zonally asymmetric stratospheric ozone rather than a zonal mean results in cooling of the stratosphere and upper troposphere during October and November comparable to that caused by ozone depletion itself over the last 30 years. However, it should be noted that we used ozone observations from a particularly zonally asymmetric year for this sensitivity study. The cooling is largely caused by transient dynamical changes. Given the magnitude of the effects seen in this study and the increasing zonal asymmetry of Southern Hemisphere ozone since the 1970s, we suggest this zonal asymmetry is responsible for a substantial fraction of the recent 30-year trends in Southern Hemisphere temperature and geopotential height. As the ozone hole recovers, the zonal asymmetry of ozone is likely to decrease over the next 50 years, leading to further climate effects. To date almost all simulations of stratospheric temperature change and Southern Hemisphere climate change have employed zonally-averaged ozone concentrations. The only exceptions to this pattern are coupled chemistry climate models which interactively simulate ozone [Eyring et al., 2007], although in these models tropospheric climate changes are usually constrained by prescribed sea surface temperatures. We suggest that in future coupled ocean-atmosphere models should also be run with interactive ozone chemistry in order to better account for the effects of three dimensional ozone variations [see also Baldwin et al., 2007].
 JAC and NPG acknowledge support from the Leverhulme Trust. SPEK was funded for this work by NERC grant (NE/D000440/1). We thank Jeff Cole and Lois Steenman-Clark (Reading University) for assistance with 3-D ancillary files, the Met Office for use of the Unified Model and the ECMWF for use of the ERA-40 data. We thank Andrew Charlton-Perez and Alexey Karpechko for useful advice and discussion.