Abstract

The separate effects of ozone depleting substances (ODSs) and greenhouse gases (GHGs) on forcing circulation changes in the Southern Hemisphere extratropical troposphere are investigated using a version of the Canadian Middle Atmosphere Model (CMAM) that is coupled to an ocean. Circulation-related diagnostics include zonal wind, tropopause pressure, Hadley cell width, jet location, annular mode index, precipitation, wave drag, and eddy fluxes of momentum and heat. As expected, the tropospheric response to the ODS forcing occurs primarily in austral summer, with past (1960–99) and future (2000–99) trends of opposite sign, while the GHG forcing produces more seasonally uniform trends with the same sign in the past and future. In summer the ODS forcing dominates past trends in all diagnostics, while the two forcings contribute nearly equally but oppositely to future trends. The ODS forcing produces a past surface temperature response consisting of cooling over eastern Antarctica, and is the dominant driver of past summertime surface temperature changes when the model is constrained by observed sea surface temperatures. For all diagnostics, the response to the ODS and GHG forcings is additive; that is, the linear trend computed from the simulations using the combined forcings equals (within statistical uncertainty) the sum of the linear trends from the simulations using the two separate forcings. Space–time spectra of eddy fluxes and the spatial distribution of transient wave drag are examined to assess the viability of several recently proposed mechanisms for the observed poleward shift in the tropospheric jet.

1. Introduction

The circulation of the Southern Hemisphere (SH) extratropical troposphere has changed noticeably over the past several decades, particularly in summer. These changes include a shift toward a more positive phase of the SH annular mode (SAM) which is associated with a poleward (southward) shift of the midlatitude jet, an increase in tropopause height, a poleward expansion of the Hadley cell, an increase in high-latitude precipitation, a warming of the Antarctic Peninsula, and a cooling of the eastern Antarctic Plateau (e.g., Simmonds and Keay 2000; Thompson et al. 2000; Thompson and Solomon 2002; Seidel and Randel 2006; Fu et al. 2006; Marshall et al. 2006). In conjunction with these changes, there have been large changes in the stratosphere. Over the past three decades it has cooled at a rate of 0.5–1.5 K decade−1, with substantially stronger cooling in the Antarctic lower stratosphere in spring and summer (Randel et al. 2009), which has resulted in a delay in the breakdown of the SH polar vortex by several weeks (Waugh et al. 1999). The stratospheric temperature trends are attributed to increasing concentrations of well-mixed greenhouse gases (GHGs) and decreasing concentrations of ozone due to chemical ozone depletion, with ozone depletion being by far the dominant forcing in the Antarctic lower stratosphere (Shine et al. 2003; Langematz et al. 2003).

That the tropospheric circulation changes are largest in austral summer is a strong indication that stratospheric ozone depletion (i.e., the Antarctic ozone hole) has played an important role at high latitudes, as first argued by Thompson and Solomon (2002) and demonstrated in the modeling study of Gillett and Thompson (2003). More comprehensive simulations using coupled (atmosphere–ocean) climate models (i.e., with prescribed stratospheric ozone) and chemistry–climate models (CCMs; i.e., with interactive stratospheric chemistry) indicate that the impacts of ozone depletion are not only confined to high latitudes but extend to middle and low latitudes, affecting the midlatitude jet, tropopause height, Hadley cell width, and precipitation (Son et al. 2009, 2010). These models, which are also forced by increasing GHG concentrations, furthermore predict that as ozone levels recover in the future the trends will reverse sign or weaken substantially (Perlwitz et al. 2008; Son et al. 2008, 2009, 2010).

A limitation of most of the above studies is that they cannot distinguish between the impacts of GHG-induced climate change (henceforth simply climate change) and ozone depletion since the concentrations of GHGs and ozone depleting substances (ODSs) both vary in time in the simulations. Climate change has an important impact on the SH that cannot be ignored, as demonstrated by coupled climate models forced with increasing GHG concentrations. Those models simulate past trends in the SAM index that have the same sign as (but are weaker than) the observed trends in December–February (DJF; e.g., Fyfe et al. 1999; Kushner et al. 2001; Arblaster and Meehl 2006). Arblaster and Meehl (2006) examined the separate impacts of GHG and ozone changes on the SAM index. However, their model did not have a well-resolved stratosphere and employed observed stratospheric ozone. The latter means that tropospheric changes cannot be unambiguously attributed to ODS changes, since the observed ozone includes the impacts of climate change, as well as natural variability. Son et al. (2009) compared similar kinds of models that participated in the Intergovernmental Panel on Climate Change’s (IPCC) Fourth Assessment Report (henceforth referred to as AR4 models), some of which included ozone forcing and some of which did not, which meant that the effects of ozone depletion and recovery could not be unambiguously identified due to intermodel differences. While Perlwitz et al. (2008) used a CCM, they did not perform an attribution of the past changes, and the use of observed sea surface temperature and sea ice distributions (SSTs for short) in the past and modeled SSTs in the future meant that a direct comparison between past and future was not possible. Consequently, the relative contributions of the ODS and GHG forcings to tropospheric circulation changes have not yet been fully resolved. In addition, the extent to which these contributions are additive (i.e., linear) remains as an open question.

The physical mechanisms responsible for the tropospheric circulation changes, in particular the poleward shift of the midlatitude jet, are also not well understood. Although there is evidence that stratospheric wave drag can directly impact surface winds (Thompson et al. 2006), it is widely believed that tropospheric eddy horizontal momentum fluxes also play an important role (e.g., Kushner and Polvani 2004; Song and Robinson 2004; Chen and Held 2007; Wittman et al. 2007; Simpson et al. 2009). Chen and Held (2007) noted that the phase speed spectrum of upper-tropospheric momentum fluxes in SH summer had shifted to larger angular phase speeds in reanalysis data over recent decades, and suggested that this could induce the observed tropospheric wind changes by displacing the region of subtropical wave drag poleward, in accordance with critical-layer control of wave breaking (Randel and Held 1991). However, Chen and Held (2007) were unable to exclude the possibility that the eddy flux (and phase speed) changes were responding to rather than causing the tropospheric wind changes. Moreover, since the reanalysis data and model simulations they examined included the effects of both ozone depletion and climate change, it is not possible to identify the extent to which the changes in tropospheric eddies are initiated by changes in the stratosphere.

As in the first part of this study, which focused on the SH stratosphere (McLandress et al. 2010; henceforth Part I), we analyze simulations from the Canadian Middle Atmosphere Model (CMAM). CMAM was unique among the 18 CCMs participating in phase 2 of the Stratospheric Processes and their Role in Climate (SPARC) CCM Validation Activity (CCMVal2; Eyring et al. 2008; SPARC CCMVal 2010) in that it was coupled to an ocean model. Not only does this allow for air–sea coupling, which produces self-consistent SSTs, it also provides a seamless transition from past to future. Three sets of simulations extending from 1960 to 2099 are examined: a control experiment in which both the GHG and ODS forcings vary transiently in time, and two sensitivity experiments in which either GHGs or ODSs are held fixed at 1960 levels, while the other forcing varies transiently as in the control experiment. The control experiment has been analyzed extensively as part of CCMVal2 (SPARC CCMVal 2010).

Here, we extend the analysis of Part I down to the troposphere. The circulation-related diagnostics we examine comprise the zonal wind, tropopause pressure, jet location, SAM index, Hadley cell width, eddy fluxes of momentum and heat, and high-latitude precipitation, in addition to the Antarctic surface temperature. Use of our three sets of simulations not only enables us to unequivocally identify the relative impacts of the ODS and GHG forcings, it also enables us to assess the additivity of the response to those forcings. As in Part I, additivity is tested by comparing linear trends from the sensitivity experiments to those from the control experiment, taking into account the statistical uncertainty of those trends. Strictly speaking, by “additive” we mean that we are unable to detect nonadditivity within our specified confidence levels; see further discussion in section 2. It is also important to realize that additivity is really only meaningful when the individual trends due to the two forcings are similar in strength, which is the case in the future, but not in the past when ODS effects dominate. As discussed in Part I, an additive response has important implications. For instance, it permits a simpler understanding of the physical mechanisms involved and means that only one of the two sets of sensitivity experiments need be performed for attribution studies like ours.

We also examine the role of tropospheric eddies in driving the zonal mean jet changes. To put our results into the context of other studies, in particular Chen and Held (2007), we compute phase speed spectra of the momentum flux convergence in the upper troposphere, and get results that are qualitatively similar to theirs. Although we cannot identify the dynamical mechanisms for the jet changes, we are able to comment on the viability of the different hypotheses that have been proposed. Furthermore, by comparing the spectra from the three sets of simulations, we are able to clearly identify the region of the atmosphere where the changes are initiated. In particular, the sensitivity experiment in which the GHGs are held fixed at 1960 levels enables us to isolate the role of the stratosphere on the changes in tropospheric eddies.

2. Description of model and simulations

CMAM is the upward extension of the Canadian Centre for Climate Modelling and Analysis’s third-generation coupled general circulation model (CGCM3). The ocean component of CMAM is described in Part I and references therein. The atmospheric component has 71 vertical levels, with a resolution that varies from several tens of meters in the lower troposphere to ∼2.5 km in the mesosphere. A T31 spectral resolution is used in the horizontal, which corresponds to a grid spacing of ∼6°. Detailed descriptions of the stratospheric chemistry scheme and the atmospheric component of CMAM are provided in de Grandpré et al. (2000) and Scinocca et al. (2008), respectively.

Since the simulations are identical to those used in Part I, only a brief description will be given; the reader is referred to that paper for further details. Three sets of coupled simulations extending from 1950 to 2099 are used, each comprising an ensemble of three members. Each ensemble member is initialized in 1950, after spinup of the ocean from a preindustrial control state. The first decade after the chemistry initialization (1950–60) is required to spin up the chemical tracers and is not included in the analysis. The GHG and ODS forcings follow the CCMVal2 specifications given in Eyring et al. (2008).

The first set is the control experiment, the so-called REF-B2 simulation described in Eyring et al. (2008), which employs time-varying concentrations of GHGs and ODSs, with the GHGs prescribed according to the IPCC Special Report on Emissions Scenarios (SRES) moderate A1B scenario (Houghton et al. 2001) and the ODSs according to the A1 scenario (WMO 2007). The REF-B2 simulation has been analyzed extensively in SPARC CCMVal (2010). The second set is the GHG simulation in which the concentrations of GHGs used in the radiation scheme are allowed to vary in time as in REF-B2, but the ODSs are held fixed at their 1960 values in the chemistry scheme. Note that the concentrations of chlorofluorocarbons-11 and -12 (CFC-11 and CFC-12) passed to the radiation code evolve in the same manner as in the REF-B2 simulation. This is a climate-change-only sensitivity experiment since the fixed low concentration of ODSs prevents changes in halogen-driven ozone loss. The third set is the ODS simulation in which the concentrations of ODSs used in the chemistry scheme are allowed to vary in time as in REF-B2, but the GHGs are held fixed at their 1960 values in the radiation scheme. Note that in this simulation the concentrations of N2O and CH4 seen by the chemistry scheme evolve as in REF-B2. This is an ozone-depletion-only sensitivity experiment since the radiative impacts of GHGs are fixed in time. A more detailed discussion of the setup of the two sensitivity experiments is provided in Eyring et al. (2008), which refers to them as SCN2b and SCN2c. As discussed in Part I, one drawback of the two sensitivity experiments is that they both include the chemical effects of transiently varying N2O and CH4, as well as the radiative effects of transiently varying tropospheric aerosols. This results in a “double counting” when additivity is tested. However, the impacts are small and, as will be seen, not detectable in the tropospheric trends examined here. As discussed in Part I, the troposphere in the coupled version of CMAM warms excessively under increasing GHG concentrations during the historical period. However, since we shall show that the response to the GHG and ODS forcings is additive, the excessive warming has no impact on our conclusions concerning the tropospheric response to ODS forcing, which is the main focus of this paper.

A fourth set of simulations is also discussed. This is the REF-B1 simulation of the recent past (Eyring et al. 2008). Unlike the coupled simulations, REF-B1 uses observed SSTs and includes natural forcings of stratospheric sulfate aerosols due to explosive volcanoes, as well as solar variability. The GHG and ODS forcings are identical to REF-B2. The REF-B1 simulation is used to assess the impacts of using the observed SSTs and the additional radiative forcings on the set of tropospheric diagnostics examined by Son et al. (2009) and on Antarctic surface temperatures.

3. Results

As in Part I a linear trend analysis is applied to the periods 1960–99 (“past”) and 2000–99 (“future”). This choice of periods is motivated by the temporal behavior of the stratospheric chlorine in the REF-B2 simulation, which increases rapidly from 1960 to the late 1990s and decreases slowly thereafter (see Fig. 1 in Part I). Linear trends are computed from the ensemble average time series using a least squares fit. Presented together with the trends are the 95% confidence levels, which are computed using the Student’s t test, assuming independent and randomly distributed residuals. All results presented below are ensemble averages and, unless stated otherwise, monthly zonal average data are used.

As in Part I, we assess the additivity of the responses to the ODS and GHG forcings by comparing linear trends from the sum of the GHG and ODS simulations to linear trends from the REF-B2 simulation, taking into account the statistical uncertainty of those trends. We do this by posing the null hypothesis that the response is additive, and then examine the statistical significance of the trends of the residual (i.e., ODS + GHG – REF-B2). If the residual trends are not statistically significantly different from 0 at the 95% confidence level, we cannot reject the null hypothesis, and we conclude that there is no evidence for nonadditivity. We will refer to this as meaning that the response is “additive to within the statistical uncertainty” or simply “additive.” If the residual trends are statistically significant (i.e., the null hypothesis is rejected), we conclude that there is a nonadditive response.

a. Stratosphere–troposphere winds and temperatures

Figure 1 shows trends in monthly polar-cap temperatures and high-latitude zonal winds for the REF-B2 simulation. The past trends (Fig. 1, top) are characterized by a rapid cooling that peaks at ∼6 K decade−1 in November at 50 hPa and a corresponding zonal wind increase that peaks at ∼6 m s−1 decade−1. The cooling is due to reduced solar heating resulting from reduced concentrations of ozone; the stratospheric wind response is from the thermal wind balance. The zonal wind trends are not confined to the stratosphere, but extend down to the surface where they peak at ∼0.5 m s−1 decade−1 in December. In the future (Fig. 1, bottom; note the smaller contour interval), ozone recovery results in a reversal of the trends in late spring and early summer, with a peak warming of ∼1 K decade−1 and a zonal wind decrease of ∼0.8 m s−1 decade−1. The weaker magnitude of the temperature and zonal wind trends in the future is the result of weaker ODS trends compared with the past. The effects of climate change are clearly seen in the future, with its characteristic warming of the troposphere and cooling of the stratosphere, which can be seen in the seasons in which the GHG forcing dominates over the ODS forcing. The springtime temperature and zonal wind trends shown in Fig. 1 are similar in structure to the CCMVal2 multimodel means presented in Son et al. (2010), though are of somewhat larger magnitude, presumably because the ozone hole in CMAM is somewhat deeper than observed (Austin et al. 2010).

Fig. 1.

Linear trends in (a),(c) temperature and (b),(d) zonal wind for the REF-B2 simulations for (top) 1960–99 and (bottom) 2000–99. Temperature and zonal wind are area averaged from 70° to 90°S and 50° to 70°S, respectively. Shading denotes 95% confidence levels. Units are K decade−1 and m s−1 decade−1. Different contour intervals are used in the top and bottom panels. Log-pressure height (computed using a constant scale height of 7 km) is plotted along the right axes.

Fig. 1.

Linear trends in (a),(c) temperature and (b),(d) zonal wind for the REF-B2 simulations for (top) 1960–99 and (bottom) 2000–99. Temperature and zonal wind are area averaged from 70° to 90°S and 50° to 70°S, respectively. Shading denotes 95% confidence levels. Units are K decade−1 and m s−1 decade−1. Different contour intervals are used in the top and bottom panels. Log-pressure height (computed using a constant scale height of 7 km) is plotted along the right axes.

Before showing results from the sensitivity experiments, it is useful to compare the REF-B2 zonal mean winds to those from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis dataset (Kalnay et al. 1996). Figure 2 shows zonal wind trends (black) and climatological westerlies (red) for REF-B2 and the reanalysis for the period 1979–99 for December–February (DJF). The climatological jet is stronger than in the reanalysis [a bias found in all 17 CCMVal2 models examined in Son et al. (2010)] and is displaced ∼10° equatorward. In the troposphere the magnitude and spatial structure of the trends are in good agreement with the reanalysis, but the CMAM trends maximize about 10° more equatorward. At middle to high latitudes in the stratosphere, the trends are larger than in the reanalysis.

Fig. 2.

Linear trends in zonal mean zonal wind for DJF for 1979–99 for (a) the REF-B2 simulations and (b) NCEP–NCAR reanalysis. Contour interval is 0.5 m s−1 decade−1; shading denotes 95% confidence levels. The red lines show the corresponding climatological westerlies using a contour interval of 10 m s−1, starting at 10 m s−1.

Fig. 2.

Linear trends in zonal mean zonal wind for DJF for 1979–99 for (a) the REF-B2 simulations and (b) NCEP–NCAR reanalysis. Contour interval is 0.5 m s−1 decade−1; shading denotes 95% confidence levels. The red lines show the corresponding climatological westerlies using a contour interval of 10 m s−1, starting at 10 m s−1.

Turning to the sensitivity experiments, we now examine the relative roles of the ODS and GHG forcings in producing the wind and temperature trends seen in REF-B2. Figure 3 shows time series of polar-cap temperatures in the lower stratosphere averaged from October to January, the period of peak cooling seen in Fig. 1a. REF-B2 (black) exhibits a rapid decrease in the past followed by a slow increase in the future. The ODS and GHG simulations (red and blue, respectively) indicate that the strong past trend in REF-B2 is due primarily to ozone depletion, while the weak future trend is due to a partial cancelation of the warming from ozone recovery and the cooling from increasing GHG concentrations. As discussed in Part I, the stratospheric polar-cap temperature response to the ODS and GHG forcings is additive. This can be seen by comparing the two right-hand columns in row 1 of Table 1, which list the trends and their uncertainties in 100-hPa polar-cap temperature for REF-B2 and for the time series generated from the sum of the GHG and ODS simulations. Overlap of the error bars indicates that the trends are additive to within the statistical uncertainty.

Fig. 3.

Time series of 100-hPa temperature for October–January area averaged from 70° to 90°S for the REF-B2 (black), ODS (red), and GHG (blue) simulations. The red and blue curves are shifted with respect to the black; their color-coded axes are given on the right. The straight lines are linear fits computed from 1960 to 1999 and 2000 to 2099.

Fig. 3.

Time series of 100-hPa temperature for October–January area averaged from 70° to 90°S for the REF-B2 (black), ODS (red), and GHG (blue) simulations. The red and blue curves are shifted with respect to the black; their color-coded axes are given on the right. The straight lines are linear fits computed from 1960 to 1999 and 2000 to 2099.

Table 1.

Linear trends and their uncertainties in 100-hPa polar-cap temperature (K decade−1), extratropical tropopause pressure (hPa decade−1), latitude of the tropospheric jet maximum (° decade−1), latitude of Hadley cell boundary (° decade−1), SAM index computed from 850-hPa geopotential (decade−1), high-latitude precipitation (mm decade−1), and Antarctic surface temperature (K decade−1) for the GHG, ODS, and REF-B2 simulations for the past and future time periods. All quantities are for DJF, with the exception of the 100-hPa temperature, which is for October–January. The far-right column is for the sum of the GHG and ODS simulations; additivity is determined by comparing it to REF-B2. Uncertainties correspond to the 95% confidence levels. See text for details.

Linear trends and their uncertainties in 100-hPa polar-cap temperature (K decade−1), extratropical tropopause pressure (hPa decade−1), latitude of the tropospheric jet maximum (° decade−1), latitude of Hadley cell boundary (° decade−1), SAM index computed from 850-hPa geopotential (decade−1), high-latitude precipitation (mm decade−1), and Antarctic surface temperature (K decade−1) for the GHG, ODS, and REF-B2 simulations for the past and future time periods. All quantities are for DJF, with the exception of the 100-hPa temperature, which is for October–January. The far-right column is for the sum of the GHG and ODS simulations; additivity is determined by comparing it to REF-B2. Uncertainties correspond to the 95% confidence levels. See text for details.
Linear trends and their uncertainties in 100-hPa polar-cap temperature (K decade−1), extratropical tropopause pressure (hPa decade−1), latitude of the tropospheric jet maximum (° decade−1), latitude of Hadley cell boundary (° decade−1), SAM index computed from 850-hPa geopotential (decade−1), high-latitude precipitation (mm decade−1), and Antarctic surface temperature (K decade−1) for the GHG, ODS, and REF-B2 simulations for the past and future time periods. All quantities are for DJF, with the exception of the 100-hPa temperature, which is for October–January. The far-right column is for the sum of the GHG and ODS simulations; additivity is determined by comparing it to REF-B2. Uncertainties correspond to the 95% confidence levels. See text for details.

Figure 4 shows zonal mean zonal wind trends for DJF for the past (top) and future (bottom). The close similarity of Figs. 4a and 4b indicates that the past REF-B2 trends are driven primarily by ozone depletion. The past trends in the GHG simulation (Fig. 4c) are substantially weaker, but are of the same sign in the troposphere as those of the ODS simulation. Unlike the response to the ODS forcing, which is confined mainly to middle and high latitudes, the response to the GHG forcing is largest at low to middle latitudes. In the future (note the smaller contour interval), the zonal wind trends in the ODS and GHG simulations are of nearly equal strength, but of opposite sign, thus explaining the near-zero tropospheric trends in REF-B2. The panels in the far-right column of Fig. 4 show the “residual” trends and their uncertainties computed from the sum of the time series from the GHG and ODS simulations minus the time series from the REF-B2 simulation. The residual trends are not statistically significant, indicating that the zonal mean zonal wind response to the ODS and GHG forcings is additive to within the statistical uncertainty.

Fig. 4.

Linear trends in zonal mean zonal wind for DJF for (top) 1960–99 and (bottom) 2000–99 for (a),(e) REF-B2, (b),(f) ODS, and (c),(g) GHG simulations, and (d),(h) the residual (computed from the sum of GHG and ODS minus REF-B2). Contour intervals are 0.4 (top) and 0.2 (bottom) m s−1 decade−1. Shading denotes 95% confidence levels. The absence of shading in (d) and (h) indicates that the residual trends are not statistically significant.

Fig. 4.

Linear trends in zonal mean zonal wind for DJF for (top) 1960–99 and (bottom) 2000–99 for (a),(e) REF-B2, (b),(f) ODS, and (c),(g) GHG simulations, and (d),(h) the residual (computed from the sum of GHG and ODS minus REF-B2). Contour intervals are 0.4 (top) and 0.2 (bottom) m s−1 decade−1. Shading denotes 95% confidence levels. The absence of shading in (d) and (h) indicates that the residual trends are not statistically significant.

b. Tropospheric circulation changes

The tropospheric diagnostics examined here comprise extratropical tropopause pressure, tropospheric jet location, Hadley cell boundary, SAM index, high-latitude precipitation, and Antarctic surface temperature. The rationale for choosing this set of diagnostics is that it permits a quantitative comparison to the results of Son et al. (2009). Results from that comparison will be discussed at the end of this section. With the exception of the SAM index, which is defined in two different ways (see below), all diagnostics are computed exactly as described in Son et al. (2009).

Figure 5 shows time series of the six diagnostics for DJF, plotted as anomalies with respect to 1960 baselines, as in Part I. The baselines are given by the 1960 intercepts of the linear fits computed from 1960 to 1999 for each set of ensemble averages. The average baseline for each diagnostic is plotted in the top-left corner of each panel. Linear fits through the time series for the past and future time periods are given by the thin lines.

Fig. 5.

Anomaly time series of (a) extratropical tropopause pressure area averaged from 45° to 90°S (note the inverse relationship between tropopause pressure and tropopause height), (b) tropospheric jet location (latitude of 850-hPa zonal mean zonal wind maximum), (c) Hadley cell boundary (latitude of zero mass streamfunction at 500 hPa), (d) SAM index (computed from 850-hPa geopotential), (e) precipitation rate area averaged from 55° to 65°S, and (f) Antarctic surface temperature area averaged from 75° to 90°S for the REF-B2 (black), ODS (red), and GHG (blue) simulations, and the sum of GHG and ODS (purple dots), for DJF. Anomalies are computed with respect to 1960 baselines (see text); the average baseline (i.e., average of the REF-B2, ODS, and GHG baselines) is plotted in the top-left corner of each panel. The 1960 baseline has not been removed from the SAM index time series. See Fig. 3 for more details.

Fig. 5.

Anomaly time series of (a) extratropical tropopause pressure area averaged from 45° to 90°S (note the inverse relationship between tropopause pressure and tropopause height), (b) tropospheric jet location (latitude of 850-hPa zonal mean zonal wind maximum), (c) Hadley cell boundary (latitude of zero mass streamfunction at 500 hPa), (d) SAM index (computed from 850-hPa geopotential), (e) precipitation rate area averaged from 55° to 65°S, and (f) Antarctic surface temperature area averaged from 75° to 90°S for the REF-B2 (black), ODS (red), and GHG (blue) simulations, and the sum of GHG and ODS (purple dots), for DJF. Anomalies are computed with respect to 1960 baselines (see text); the average baseline (i.e., average of the REF-B2, ODS, and GHG baselines) is plotted in the top-left corner of each panel. The 1960 baseline has not been removed from the SAM index time series. See Fig. 3 for more details.

Figure 5a shows time series of the extratropical thermal tropopause pressure computed using the method of Reichler et al. (2003), and area averaged from 45° to 90°S. The REF-B2 simulation (black) exhibits a rapid decrease with time of tropopause pressure in the past, which is followed by a more gradual decrease in the future. The ODS simulation (red) exhibits a past decrease and future increase, while the GHG simulation (blue) shows a steady decrease from 1960 to 2100. The purple dots denote the sum of the anomaly time series from the ODS and GHG simulations. The dots are seen to envelope the REF-B2 time series, thus providing visual evidence for additivity. Differences in the year-to-year fluctuations of the dotted and black curves are unimportant since only the long-term behavior of the time series is relevant for additivity. Note also that visual inspection reveals that the variability with respect to the linear trends appears to be larger for the ODS + GHG time series than for the REF-B2 time series, which is consistent with what one would expect for the sum of two normally distributed random variables. The trends and their uncertainties for the sum, which are listed in the far-right column of Table 1 (row 2), indicate that the tropopause pressure response to the ODS and GHG forcings is indeed additive to within the statistical uncertainty.

Time series of the tropospheric jet location and the Hadley cell boundary are shown in Figs. 5b and 5c, respectively. The former is defined as the latitude of the zonal mean zonal wind maximum at 850 hPa; the latter as the latitude where the meridional mass streamfunction at 500 hPa is 0. The mass streamfunction (Ψ) is computed from the zonal mean meridional wind (υ) as follows:

 
formula

where ϕ is latitude, p is pressure, and g is gravity. Both time series are computed on a 0.1° latitude grid using cubic spline interpolation. The jet location and Hadley cell boundary show a poleward (southward) shift in the past in REF-B2 that is largely attributed to ozone depletion, as seen by similar changes in the ODS simulation. The very weak REF-B2 trends in the future are due to the oppositely signed trends in the ODS and GHG simulations. The trends are additive to within the statistical uncertainty, as seen in Table 1 (rows 3 and 4). However, because of the larger uncertainties, the precision to which additivity can be determined is less than that for the tropopause pressure.

Two methods are used to compute the SAM index. The first computes it from the leading empirical orthogonal function (EOF) of the 850-hPa geopotential height anomalies from 18° to 90°S (Thompson and Wallace 2000). The second uses the proxy method of Gong and Wang (1999), which defines the SAM index as the difference in zonal average mean sea level pressure (MSLP) between 40° and 65°S; this method is used to compare our results to those of Son et al. (2009), which will be discussed later. Since fixed latitudes are used, the proxy method may potentially be affected by model biases (e.g., due to biases in the latitudinal structure of the EOF). However, a quantitative comparison of the two indices yields an average correlation coefficient of 0.86 for the three sets of simulations, indicating that the proxy method provides a good estimate of the true SAM index for these simulations.

To compare the SAM indices from the three sets of simulations, a single EOF is computed from anomalies generated by removing a single climatology (computed from 1960 to 2100 for each month in DJF using all nine simulations) from the individual ensemble members. The EOF computed in this manner is virtually identical to the EOFs computed from each of the three separate sets of simulations. Moreover, the three separate EOFs explain roughly the same amount of the variance (∼52% plus or minus a percent). Figure 6 shows the leading EOF computed in this manner, which by construction has units of geopotential height. It exhibits a clear annular structure, with positive (red) and negative (blue) values at middle and high latitudes, respectively. The nondimensional SAM index for each ensemble member is computed by projecting the monthly geopotential anomalies onto this EOF. We have also verified that secular trends in global mean height, which may alias onto the SAM and obscure its interpretation as a north–south movement of mass (Gerber et al. 2010), have no perceptible impact at 850 hPa.

Fig. 6.

SH polar stereographic map of the first EOF of 850-hPa geopotential height for DJF computed from the combined anomalies from the REF-B2, ODS, and GHG simulations. Contour interval is 10 m; zero line is thick. Red and blue shading denote positive and negative values, respectively. Latitude circles at 20°, 40°, 60°, and 80°S are shown. This EOF is used to compute the SAM indices shown in Fig. 5d.

Fig. 6.

SH polar stereographic map of the first EOF of 850-hPa geopotential height for DJF computed from the combined anomalies from the REF-B2, ODS, and GHG simulations. Contour interval is 10 m; zero line is thick. Red and blue shading denote positive and negative values, respectively. Latitude circles at 20°, 40°, 60°, and 80°S are shown. This EOF is used to compute the SAM indices shown in Fig. 5d.

Figure 5d shows the SAM index computed as above, and averaged for DJF. Since the time mean SAM index is very nearly 0, the 1960 baseline is not removed. In the REF-B2 simulation the SAM index exhibits a rapid increase in the past, flattening out to near-zero trend in the future. In response to ozone depletion and recovery, the SAM index in the ODS simulation increases rapidly in the past and decreases slowly in the future, whereas climate change (GHG simulation) results in a slow and steady increase in the SAM index from 1960 to 2100. The latter is in accordance with tropospheric climate model simulations using only increasing GHG concentrations (e.g., Fyfe et al. 1999). The additivity of the trends is demonstrated in Table 1 (row 5). While the ODS+GHG trend in the SAM index is 50% larger than the REF-B2 trend in the past, the difference is not statistically significant due to large uncertainties.

Figures 5e and 5f show the high-latitude precipitation and Antarctic surface temperature time series, area averaged from 55° to 65°S and 75° to 90°S, respectively. REF-B2 exhibits increasing precipitation throughout the simulation, with the stronger past trends due primarily to ozone depletion and the weaker future trends to increasing GHGs. This is consistent with the shift in the tropospheric jet, which moves the storm tracks poleward in the past. No trends in polar-cap average Antarctic surface temperature are seen in the ODS simulation. An examination of Table 1 (rows 6 and 7) indicates that both of these quantities are additive to within the statistical uncertainty. However, in the case of Antarctic surface temperature, the additivity is trivial since the trends in the ODS simulation are nearly zero.

The lack of any ODS-induced Antarctic surface temperature trends might seem surprising in light of the emphasis in Gillett and Thompson (2003) on the summertime surface cooling arising from the ozone hole. However, the latitude range used to compute the trends shown in Fig. 5f, namely 75°–90°S, which is what was used by Son et al. (2009), misses a large fraction of eastern Antarctica where significant cooling has been observed in the recent past (Thompson and Solomon 2002). The top panels in Fig. 7 show the spatial structure of the high-latitude summertime surface temperature trends in the ODS simulation for the past and the first half of the twenty-first century. During the period of ozone decline (Fig. 7a), a cooling is indeed seen over the eastern half of Antarctica, together with several isolated regions of warming over western Antarctica along coastal regions. Opposite trends are seen during the period of ozone recovery (Fig. 7b), attesting to the robustness of the response. The past trends are broadly consistent with those shown in Gillett and Thompson (2003), as well as with those in Sigmond and Fyfe (2010), who studied the surface response to an imposed ozone hole in a dynamics-only coupled model. It should be noted that the surface temperature observations shown in Thompson and Solomon (2002) are concentrated on the eastern half of the continent, with the exception of the Antarctic Peninsula, which shows a warming, so we do not really know what the observed temperature trends are over western Antarctica.

Fig. 7.

SH polar stereographic maps of surface temperature trends for DJF: (a) ODS simulation for 1960–99, (b) ODS simulation for 2000–2050, (c) REF-B1 simulation for 1960–99, and (d) difference between ODS and REF-B1 simulations for 1960–99. Colored contours are staggered about 0 using an interval of 0.05 K decade−1 (i.e., ±0.025, ±0.075, …), with red denoting warming and blue cooling. The area enclosed by the black dotted lines indicates regions where the trends are statistically different from 0 at the 95% confidence level.

Fig. 7.

SH polar stereographic maps of surface temperature trends for DJF: (a) ODS simulation for 1960–99, (b) ODS simulation for 2000–2050, (c) REF-B1 simulation for 1960–99, and (d) difference between ODS and REF-B1 simulations for 1960–99. Colored contours are staggered about 0 using an interval of 0.05 K decade−1 (i.e., ±0.025, ±0.075, …), with red denoting warming and blue cooling. The area enclosed by the black dotted lines indicates regions where the trends are statistically different from 0 at the 95% confidence level.

Figure 7c shows the past summertime surface temperature trends from the REF-B1 simulation, which includes both ODS and GHG forcings, but is constrained by observed SSTs. We choose to show REF-B1 rather than REF-B2, because, as discussed in section 2, this version of the coupled CMAM warms excessively, which obscures the ODS effect on surface temperatures. The REF-B1 simulation exhibits cooling over eastern Antarctica that is similar to the ODS simulation, with the differences in the trends (Fig. 7d) not being statistically significant over most of the continent. From this comparison we can conclude that the ODS forcing has been the dominant driver of temperature changes over the Antarctic continent during the recent past.

The seasonal variations of the trends in a subset of the circulation-related diagnostics (tropopause pressure, jet location, and Hadley cell boundary) are shown in Fig. 8. As seen by the blue symbols, the effects of GHG forcing are largely independent of season. By contrast, the effects of ozone depletion and recovery seen in the ODS simulation (red) are confined mainly to summer, with the exception of tropopause pressure, which also has trends that are statistically different from zero in spring and fall. The yellow symbols denote the trends computed from the sum of the time series of the GHG and ODS simulations. Where the yellow and black error bars overlap indicates additivity.

Fig. 8.

Seasonal variation of linear trends in (a),(b) extratropical tropopause pressure, (c),(d) tropospheric jet location, and (e),(f) Hadley cell boundary for (left) 1960–99 and (right) 2000–99 for the REF-B2 (black), ODS (red), and GHG (blue) simulations, and the sum of GHG and ODS (yellow). Seasons are September–November, SON; December–February, DJF; March–May, MAM; and June–August, JJA. Error bars denote 95% confidence levels. See Fig. 5 for details.

Fig. 8.

Seasonal variation of linear trends in (a),(b) extratropical tropopause pressure, (c),(d) tropospheric jet location, and (e),(f) Hadley cell boundary for (left) 1960–99 and (right) 2000–99 for the REF-B2 (black), ODS (red), and GHG (blue) simulations, and the sum of GHG and ODS (yellow). Seasons are September–November, SON; December–February, DJF; March–May, MAM; and June–August, JJA. Error bars denote 95% confidence levels. See Fig. 5 for details.

We end this section with a comparison to the AR4 model results presented in Son et al. (2009). Figure 9 shows trends in all but one of the tropospheric diagnostics examined in Son et al.; polar-cap average Antarctic surface temperature is omitted since it was shown to have near-zero trends in the ODS simulation. For consistency with the Son et al. results, we use the proxy method for computing the SAM index, where MSLP is interpolated onto the 40° and 65°S latitudes using cubic splines, and we compute the future trends over 2000–49 rather than 2000–99. With the exception of the 100-hPa polar-cap temperature, which is averaged from October to January, all quantities are for DJF. Past trends are shown on the left half of each panel, future trends on the right. Since the GHG simulation and the AR4 models without ozone depletion and recovery (blue circles and squares, respectively) are both climate-change-only experiments, they are a comparable set of simulations. Both agree well in the size of the trends and in the absence of any significant change from past to future, with the exception, perhaps, of tropopause pressure in the future. Note, however, that the CMAM error bars are not comparable to the AR4 error bars (see Fig. 9 caption). Similarly, the REF-B2 simulation and the AR4 models with ozone depletion and recovery (black circles and squares, respectively) are comparable. Again there is good agreement, with both showing oppositely signed trends in the past and future, and with past trends exceeding future trends. Comparison of REF-B2 and the ODS simulation (red) shows that ozone depletion and recovery explain most of the trend changes between the past and future time periods, as discussed earlier. Finally, as a check of the REF-B2 results for the past, trends from the REF-B1 simulation (green) are also shown. The trends in all plotted quantities agree with the REF-B2 results within the error bars.

Fig. 9.

Linear trends in (a) 100-hPa polar-cap temperature, (b) extratropical tropopause pressure, (c) tropospheric jet location, (d) Hadley cell boundary, (e) SAM index proxy (MSLP difference between 40° and 65°S), and (f) precipitation rate for the REF-B2 (black circles), GHG (blue circles), ODS (red circles), and REF-B1 (green circles; past only) simulations, and the mean of a total of 20 AR4 models with and without ozone depletion and recovery (black and blue squares, respectively). Left and right halves of each panel show trends computed from 1960–99 and 2000–49, respectively. Error bars denote the 95% confidence levels for CMAM results and one standard deviation for AR4 model results. Time periods are (a) October–January and (b)–(f) DJF. AR4 model data are from Son et al. (2009). See Fig. 5 for details.

Fig. 9.

Linear trends in (a) 100-hPa polar-cap temperature, (b) extratropical tropopause pressure, (c) tropospheric jet location, (d) Hadley cell boundary, (e) SAM index proxy (MSLP difference between 40° and 65°S), and (f) precipitation rate for the REF-B2 (black circles), GHG (blue circles), ODS (red circles), and REF-B1 (green circles; past only) simulations, and the mean of a total of 20 AR4 models with and without ozone depletion and recovery (black and blue squares, respectively). Left and right halves of each panel show trends computed from 1960–99 and 2000–49, respectively. Error bars denote the 95% confidence levels for CMAM results and one standard deviation for AR4 model results. Time periods are (a) October–January and (b)–(f) DJF. AR4 model data are from Son et al. (2009). See Fig. 5 for details.

Figure 9 is comparable to Fig. 4 in Son et al. (2010), which compares AR4 model trends to the multimodel mean trends of 17 CCMVal2 models (one of which is CMAM), but using only the REF-B1 and REF-B2 simulations. Our REF-B1 and REF-B2 results are consistent with theirs, although our past trends are somewhat larger, in accordance with our somewhat deeper ozone hole. The important point to note here, however, is that, unlike our analysis, which includes the two sensitivity experiments, theirs did not attribute the ODS and GHG contributions to those trends.

c. Role of tropospheric eddies

Chen and Held (2007) performed a space–time spectral analysis of the meridional flux of zonal momentum (or simply momentum flux) from reanalysis data and found a marked increase in the eastward (angular) phase speed of the tropospheric eddies, which was accompanied by a poleward displacement of the region of wave breaking in the subtropics. This pattern of behavior was also reproduced in their climate model simulations. The dynamical mechanism they proposed to explain this behavior is discussed in the introduction. However, their analysis could not distinguish between cause and effect since, as they point out, it is unclear whether the shift toward higher phase speeds is the cause or the result of the shift in the tropospheric westerlies. Interpretation of their results is further complicated by the fact that the effects of ozone depletion and climate change are both present, thus making it unclear whether changes in the stratosphere are initiating the changes in the troposphere.

Here, we perform the same analysis as in Chen and Held (2007) using the three sets of simulations. To compare our results to theirs, we show spectra of the momentum flux convergence (M), which are given by

 
formula

where a is the earth’s radius, the overbar denotes a zonal mean, and u′ and υ′ are, respectively, the deviations of the zonal and meridional wind components about the zonal mean. The reason Chen and Held used momentum flux convergence instead of Eliassen–Palm flux divergence is that they were interested in the zonal momentum budget and the changes in surface winds. Momentum flux convergence spectra are computed from gridded daily average horizontal winds that have been truncated to zonal wavenumber 16. Zonal wavenumber–frequency spectra are calculated using a standard Fourier analysis for each 120-day period (mid-November–mid-March) of each ensemble member. To prevent edge effects from contaminating the frequency spectra, the ends of the time series are tapered (windowed) before computing the spectra. As in Randel and Held (1991), the spectra are smoothed in frequency, transformed to zonal wavenumber–angular phase speed space, and summed over all zonal wavenumbers. [Randel and Held used regular phase speed, but as in Chen and Held (2007), we use angular phase speed as it is conserved under meridional propagation through a zonally symmetric flow on a sphere.] Linear trends and their 95% uncertainties are computed from the ensemble-average spectral time series using least squares.

Figure 10 shows the momentum flux convergence trends and climatologies versus angular phase speed and latitude at 250 hPa. The thick solid and dashed lines denote, respectively, the climatological zonal mean zonal winds (divided by the cosine of latitude) and the corresponding wind changes computed from the linear fits. The climatological spectra (colored contours) exhibit the characteristic pattern of convergence (red) in middle latitudes and divergence (blue) in the subtropics. The former reflects the generation of synoptic waves in the region of the jet core; the latter, their dissipation near critical levels as they propagate equatorward. There is also a close correspondence between the momentum flux convergence trends and the zonal wind changes, with positive trends (thin solid black) being associated with wind increases, and negative trends (thin dashed) with wind decreases. This is indicative of the strong coupling between the synoptic waves and the zonal mean wind. Although the results presented here include planetary-scale waves, nearly identical results are obtained if only synoptic-scale waves (i.e., wavenumbers greater than 3) are used (results not shown).

Fig. 10.

Linear trends in momentum flux convergence (black contours) vs angular phase speed (ca) and latitude at 250 hPa for DJF for (top) 1960–99 and (bottom) 2000–99 for the (a),(d) REF-B2, (b),(e) ODS, and (c),(f) GHG simulations. Contour intervals are (top) 0.006 and (bottom) 0.003 m s−1 day−1 decade−1ca)−1, where Δca = 1 m s−1 is the grid size; thin solid and dashed black lines denote positive and negative trends, respectively. Shading denotes the 95% confidence levels of the trends. The climatological momentum flux convergence is given by the red (positive) and blue (negative) contours, using an interval of 0.1 m s−1 day−1ca)−1. Both the thin black and colored contour lines are staggered about 0 (i.e., ±0.05, ±0.15, …). Spectra are not plotted for |ca| ≤ 2 m s−1 since those slow phase speeds are not resolved. The thick solid lines are the climatological zonal mean zonal winds divided by the cosine of the latitude (U/cosϕ); the thick dashed lines are the changes in U/cosϕ computed from the linear trends (i.e., end of time period minus beginning). Values of U/cosϕ in m s−1 are given along the bottom axis.

Fig. 10.

Linear trends in momentum flux convergence (black contours) vs angular phase speed (ca) and latitude at 250 hPa for DJF for (top) 1960–99 and (bottom) 2000–99 for the (a),(d) REF-B2, (b),(e) ODS, and (c),(f) GHG simulations. Contour intervals are (top) 0.006 and (bottom) 0.003 m s−1 day−1 decade−1ca)−1, where Δca = 1 m s−1 is the grid size; thin solid and dashed black lines denote positive and negative trends, respectively. Shading denotes the 95% confidence levels of the trends. The climatological momentum flux convergence is given by the red (positive) and blue (negative) contours, using an interval of 0.1 m s−1 day−1ca)−1. Both the thin black and colored contour lines are staggered about 0 (i.e., ±0.05, ±0.15, …). Spectra are not plotted for |ca| ≤ 2 m s−1 since those slow phase speeds are not resolved. The thick solid lines are the climatological zonal mean zonal winds divided by the cosine of the latitude (U/cosϕ); the thick dashed lines are the changes in U/cosϕ computed from the linear trends (i.e., end of time period minus beginning). Values of U/cosϕ in m s−1 are given along the bottom axis.

The momentum flux convergence spectra for the past in the REF-B2 simulation (Fig. 10a) exhibit both an increase in time of (angular) phase speed and a poleward shift of the maxima, as indicated by the differences in the positions of the trends (light dashed and solid contours) compared to the relative maxima of the climatology (red and blue contours). This is in good qualitative agreement with the results shown in Fig. 1 of Chen and Held (2007), although the phase speed increase is somewhat smaller here. The close correspondence between the REF-B2 and the ODS simulations (Fig. 10b) suggests that the observed changes in momentum flux convergence found by Chen and Held are driven primarily by ozone depletion. The past trends in the GHG simulation (Fig. 10c) are much smaller and not statistically significant, indicating that climate change is playing only a small role in the past.

The future trends in momentum flux convergence in REF-B2 (Fig. 10d) are substantially weaker than in the past, which is to be expected given the weaker trends in the ODS forcing. There is also a change in structure of the trends, with the past trends peaking at high phase speeds (c ∼ 20 m s−1) and the future trends at low phase speeds (c ∼ 10 m s−1). The structural change can be understood by examining the trends in the sensitivity experiments. In response to ozone recovery, the future trends in the ODS simulation (Fig. 10e) consist of a meridionally oriented dipole centered at c ∼ 15 m s−1 that is opposite in sign to the past trends. The trends in the GHG simulation (Fig. 10f) are the same sign as in the past, but statistically significant on account of the longer time series. They appear to consist of a pair of meridionally oriented dipoles: one at higher phase speeds that is nearly equal and opposite to the future trends in the ODS simulation, and the other at lower phase speeds.

Inspection of Figs. 10e and 10f reveals a near cancelation of the high-phase-speed dipoles, leaving only the low-phase-speed dipole seen in REF-B2, which suggests that the response to the ODS and GHG forcings is additive. To confirm this, Fig. 11, which uses the same contour interval as Fig. 10, shows trends and their 95% uncertainties computed from time series generated from the sum of the time series from the ODS and GHG simulations minus the REF-B2 time series. The fact that the residual trends are not statistically significant (with the exception of several isolated points) indicates that the response is additive to within the statistical uncertainty. The larger residual trends in the past than in the future are a reflection of the smaller sample size. If more ensemble members were available, we would expect that the residual trends would diminish. Additivity in a nonlinear quantity like eddy flux is perhaps somewhat surprising, but not totally unexpected given our earlier finding that the zonal wind trends, which evolve in conjunction with the eddy fluxes, are also additive.

Fig. 11.

Linear trends in momentum flux convergence vs angular phase speed and latitude at 250 hPa for DJF for the residual computed from the sum of GHG and ODS minus REF-B2 for (a) 1960–99 and (b) 2000–99. The contour interval is the same as in Fig. 10. Shading denotes 95% confidence levels. The absence of shading in all but a few isolated regions indicates that the residual trends are not statistically significant.

Fig. 11.

Linear trends in momentum flux convergence vs angular phase speed and latitude at 250 hPa for DJF for the residual computed from the sum of GHG and ODS minus REF-B2 for (a) 1960–99 and (b) 2000–99. The contour interval is the same as in Fig. 10. Shading denotes 95% confidence levels. The absence of shading in all but a few isolated regions indicates that the residual trends are not statistically significant.

Figure 12 shows meridional heat flux spectra at 700 hPa, that is, where baroclinic instability is generating the waves. The climatological values are negative (blue), indicating a southward (i.e., poleward) heat flux and an upward flux of wave activity. The overall features are similar to what was found with momentum flux convergence: the past REF-B2 trends are driven mainly by ozone depletion and show an increase with time of the angular phase speed, as well as a poleward shift in the maximum. Future trends for the ODS simulation are much weaker than in the past and opposite in sign, unlike the future trends for the GHG simulation, which are of the same magnitude and sign as in the past. Inspection of the residual trends for the heat flux (not shown) indicates that this quantity too is additive to within the statistical uncertainty. As before, the results are nearly unchanged if planetary-scale waves are excluded, indicating that synoptic-scale waves are by far the dominant contributors to the heat fluxes in the SH summer troposphere.

Fig. 12.

As in Fig. 10, but for 700-hPa meridional heat flux. Solid (positive) and dashed (negative) black contours denote the trends using intervals of (top) 0.01 and (bottom) 0.005 K m s−1 decade−1c)−1, staggered about 0 as in Fig. 10. Blue contours denote a negative (southward) climatological heat flux, using an interval of 0.2 K m s−1c)−1, starting at 0.1 K m s−1c)−1.

Fig. 12.

As in Fig. 10, but for 700-hPa meridional heat flux. Solid (positive) and dashed (negative) black contours denote the trends using intervals of (top) 0.01 and (bottom) 0.005 K m s−1 decade−1c)−1, staggered about 0 as in Fig. 10. Blue contours denote a negative (southward) climatological heat flux, using an interval of 0.2 K m s−1c)−1, starting at 0.1 K m s−1c)−1.

To obtain a more complete picture of the changes in eddy fluxes, Fig. 13 shows trends in the resolved transient wave drag (black contours) as given by the Eliassen–Palm flux divergence, expressed as force per unit mass, and thus representing the right-hand-side eddy tendency term in the zonal mean zonal wind equation. Color shading in all but the far-right panels denotes the trends in the zonal mean zonal winds. In the upper troposphere a strong positive correlation exists between the wave drag and zonal wind trends, such that positive trends in wave drag (solid black) are associated with positive trends in zonal wind (red), and vice versa. Note that the climatological wave drag in the upper troposphere is negative; so the positive trends represent a decrease in wave drag. This plot reveals that the ozone hole (Fig. 13b) displaces the wave drag from the midlatitude upper troposphere to higher latitudes, in addition to a displacement to subtropical latitudes. This is quite different from what would be expected from the Chen and Held (2007) mechanism, which is confined to subtropical latitudes, and suggests a more drastic latitudinal redistribution of wave drag reaching to higher latitudes. Note that extending the Chen and Held (2007) mechanism to higher latitudes would predict a shift in wave drag toward, rather than away, from the jet core in response to the faster phase speeds seen in Fig. 12.

Fig. 13.

Linear trends in transient wave drag (black contours) and zonal mean zonal wind (color shading) vs latitude and pressure for DJF for (top) 1960–99 and (bottom) 2000–99 for the (a),(e) REF-B2, (b),(f) ODS, and (c),(g) GHG simulations; and (d),(h) the residual (computed from the sum of GHG and ODS minus REF-B2). Wave drag is given by the Eliassen–Palm flux divergence, expressed as force per unit mass. Contour intervals for wave drag are (top) 0.2 and (bottom) 0.1 m s−1 day−1 decade−1. Red and blue shading in (a)–(c) and (e)–(g) denotes positive and negative zonal wind trends, respectively, using intervals of (top) 0.2 and (bottom) 0.1 m s−1 decade−1. Contour lines and color shading intervals are staggered about 0 (i.e., ±0.1, ±0.3, … and ±0.05, ±0.15, …). Gray shading in (d) and (h) denotes the 95% confidence levels of the residual wave drag trends; the absence of shading in all but a few regions indicates that the residual trends are not statistically significant.

Fig. 13.

Linear trends in transient wave drag (black contours) and zonal mean zonal wind (color shading) vs latitude and pressure for DJF for (top) 1960–99 and (bottom) 2000–99 for the (a),(e) REF-B2, (b),(f) ODS, and (c),(g) GHG simulations; and (d),(h) the residual (computed from the sum of GHG and ODS minus REF-B2). Wave drag is given by the Eliassen–Palm flux divergence, expressed as force per unit mass. Contour intervals for wave drag are (top) 0.2 and (bottom) 0.1 m s−1 day−1 decade−1. Red and blue shading in (a)–(c) and (e)–(g) denotes positive and negative zonal wind trends, respectively, using intervals of (top) 0.2 and (bottom) 0.1 m s−1 decade−1. Contour lines and color shading intervals are staggered about 0 (i.e., ±0.1, ±0.3, … and ±0.05, ±0.15, …). Gray shading in (d) and (h) denotes the 95% confidence levels of the residual wave drag trends; the absence of shading in all but a few regions indicates that the residual trends are not statistically significant.

The latitudinal shifts in wave drag seen in Fig. 13 may, instead, be a direct response to the mean-flow changes in the lower stratosphere induced by the ozone hole. Thorncroft et al. (1993) showed that in idealized baroclinic life cycles, the proclivity of baroclinic waves to break equatorward or poleward was sensitive to the horizontal shear in the initial zonal mean flow, with anticyclonic shear favoring equatorward breaking (which Thorncroft et al. termed LC1 events), and cyclonic shear favoring poleward breaking (which they termed LC2 events). Shepherd (2002) argued that this sensitivity of upper-tropospheric wave breaking to the zonal flow provided a mechanism for strong stratosphere–troposphere coupling, whereby a change in lower-stratospheric winds could change the likelihood of LC1 versus LC2 events and thereby induce latitudinal shifts in wave breaking that would drive changes in surface winds. From Fig. 4b, the lower-stratospheric zonal wind changes induced by the ozone hole are cyclonic poleward of about 50°S and anticyclonic equatorward of about 50°S, which following this reasoning would imply a poleward shift in wave breaking poleward of 50°S, and an equatorward shift equatorward of 50°S, which is exactly what is seen in Fig. 13b. This hypothesis is also supported by the baroclinic life cycle experiments conducted by Wittman et al. (2007), who found that increased vertical shear of the stratospheric zonal flow increased the likelihood of LC2-type breaking, thereby inducing a poleward shift of synoptic-scale wave drag at high latitudes.

Note also that the latitude–height structure of the ODS-induced changes is quite different from the GHG-induced changes (cf. Figs. 13b and 13g) despite their more similar appearance at 250 hPa (Figs. 10b and 10f). The far-right panels show the trends and their uncertainties (gray shading) computed from the sum of the time series from the ODS and GHG simulations minus the REF-B2 time series. The fact that the residual trends are not statistically significant over most of the plotted domain indicates that the wave drag response is additive to within the statistical uncertainty throughout the SH troposphere, a finding that is consistent with the additivity of the zonal mean zonal wind trends.

4. Summary and conclusions

The Canadian Middle Atmosphere Model (CMAM) is used to assess the separate effects of ozone depletion (and recovery) and climate change on the circulation of the SH troposphere. The version of CMAM that is used here is coupled to an ocean, making it unique among the models that participated in CCMVal2. Three sets of simulations extending from 1960 to 2099 are analyzed in detail: 1) the REF-B2 simulation, in which the concentrations of both GHGs and ODSs vary in time; 2) the ODS simulation, in which GHGs are held fixed at their 1960 levels and ODSs vary in time; and 3) the GHG simulation, in which ODSs are held fixed at their 1960 levels and GHGs vary in time. Since each set of simulations comprises an ensemble of three members, a good signal-to-noise ratio is obtained, enabling the responses to the ODS and GHG forcings to be characterized with a fairly high degree of precision. The ODS and GHG simulations permit the separate effects of ozone depletion–recovery and climate change to be examined; the three together permit the question of additivity to be addressed. The tropospheric diagnostic quantities that are analyzed include zonal mean zonal wind, extratropical tropopause pressure, Hadley cell boundary, jet location, SAM index, high-latitude precipitation, Antarctic surface temperature, and wave drag, as well as phase speed spectra of the eddy momentum flux convergence and meridional heat flux.

The response to the GHG and ODS forcings of all of the quantities analyzed is found to be additive; that is, past (1960–99) and future (2000–99) linear trends computed from the sum of the time series from the ODS and GHG simulations equal the REF-B2 trend, within the specified level of statistical uncertainty (95% confidence level). Strictly speaking, by “additive” we mean that we are unable to detect nonadditivity. The fact that additivity is found in the wave drag trends over the entire region of the SH troposphere and lower stratosphere, as well as for trends in the eddy-flux phase-speed spectra at specific heights in the troposphere, is perhaps somewhat surprising, but nevertheless consistent with our finding that the trends in the tropospheric zonal mean zonal wind, which evolves in conjunction with those fluxes, are also additive.

The impacts of the ODS forcing occur mainly in austral summer, with the exception of the tropopause pressure, which also shows statistically significant trends in spring and fall. The impacts of the GHG forcing, on the other hand, are more evenly spread out over the year. As expected, past summertime trends are dominated by the ODS forcing. Future summertime trends, on the other hand, are made up of nearly equal contributions from the ODS and GHG forcings. However, because the future trends from the ODS forcing are opposite in sign to the past trends, there is strong cancelation with the GHG-induced trends, which explains the weak future summertime trends in the REF-B2 simulation.

The ODS forcing induces surface cooling over eastern Antarctica in the past during ozone decline and warming in the future during ozone recovery. This is obscured when a 75°–90°S polar-cap average is used, since the largest ODS-induced temperature trends occur over the portion of eastern Antarctica that is equatorward of 75°S. That, in conjunction with some isolated regions of warming in western Antarctica, results in near-zero polar-cap average trends in the ODS simulation. The spatial pattern of past surface cooling due to the ODS forcing agrees qualitatively with results from models forced by ozone depletion (Gillett and Thompson 2003; Sigmond and Fyfe 2010). A comparison of the past surface temperature trends from the ODS simulation also reveals quantitative agreement (i.e., within our 95% confidence levels) over most of the Antarctic continent with those from the REF-B1 simulation in which CMAM is constrained by observed SSTs.

A comparison to the AR4 model results of Son et al. (2009) reveals good quantitative agreement of trends in lower-stratospheric polar-cap temperature, tropopause pressure, jet location, Hadley cell boundary, SAM index, and high-latitude precipitation. The AR4 models without ozone depletion and recovery agree well with the results from the GHG simulation, with both exhibiting no change in sign of the trends between the past and future, and past and future trends of similar strength. The AR4 models with ozone depletion and recovery are consistent with the REF-B2 simulation, with both exhibiting stronger past trends and weaker future trends, which are of opposite sign for some of the quantities. The ODS simulation shows the contribution from only ozone depletion and recovery, which is missing in the AR4 model intercomparison in Son et al. (2009) and also in the CCMVal2 intercomparison in Son et al. (2010).

A brief comparison of the Son et al. (2009) tropospheric diagnostics to the REF-B1 simulation of the recent past shows good quantitative agreement with REF-B2 (within the error bars), indicating that there is no noticeable impact on these diagnostics of the observed SSTs and additional radiative forcings (volcanic aerosols and solar variability) used in REF-B1. Note, however, that the uncertainties of the past trends are quite large.

Phase speed spectra of momentum flux convergence in the upper troposphere for the REF-B2 simulation are shown to be in good qualitative agreement with the reanalysis data presented in Chen and Held (2007), which exhibits an increase in time of the angular phase speed and a poleward (southward) shift in the subtropical maxima, which they argue may be responsible for the poleward shift in the surface westerlies. The corresponding spectra for the sensitivity experiments reveal the relative contributions of the two forcings to those trends. In the future, the trends for the ODS simulation reverse, while those for the GHG simulation remain of the same sign. Similar results are found for meridional heat flux in the lower troposphere.

The fact that the past momentum flux spectral trends for the ODS simulation are very similar in structure and strength to those for REF-B2 indicates that it is the changes in the stratosphere that initiated the changes in the troposphere, a conclusion that could not be drawn from the results of Chen and Held (2007), which included the effects of both increasing GHG concentrations and declining ozone concentrations. On the other hand, the latitude–height cross sections of changes in Eliassen–Palm flux divergence suggest that for both ODS and GHG forcing, important changes in wave drag occur not only in the subtropics but also on the poleward flank of the tropospheric jet. A poleward shift of wave drag on the poleward side of the jet is inconsistent with an increase in angular phase speed and suggests instead the role of an LC1–LC2 transition, as proposed by Shepherd (2002) and illustrated by Wittman et al. (2007). This feature, as well as the striking difference between the ODS- and GHG-induced changes in wave drag, provides interesting material for future investigation.

A forthcoming paper by Polvani et al. (2011), which was submitted not long before ours was, also examines the separate impacts of ozone depletion and GHG forcing on the SH tropospheric summertime circulation. In contrast to our study, they use a set of time-slice simulations and consider only the recent past. As with a number of the studies mentioned in the introduction, theirs suffers from many of the same limitations, namely, prescribed zonal mean ozone changes, a model without a well-resolved stratosphere, and the use of observed SSTs. Like us, they conclude that the past effects of ozone depletion are considerably larger than those of GHG forcing. They also explore the question of the additivity of the response to the two forcings, and find that the response for some quantities is additive, but is nonadditive for others. However, because they did not take into account statistical significance, it is possible that the nonadditive responses they found are not robust. Furthermore, as stated earlier, since the ODS forcing either dominates over the GHG forcing in the past or is negligible in comparison, additivity is less relevant in this time period.

Acknowledgments

The authors thank the four anonymous reviewers for their constructive comments on the submitted manuscript. CM thanks Seok-Woo Son for kindly providing the AR4 model data and Isla Simpson for helpful discussions. The authors also thank Norman McFarlane for comments on an earlier version of the manuscript. This work was supported by the Canadian Foundation for Climate and Atmospheric Sciences through the C-SPARC project.

REFERENCES

REFERENCES
Arblaster
,
J. M.
, and
G. A.
Meehl
,
2006
:
Contributions of external forcings to southern annular mode trends.
J. Climate
,
19
,
2896
2904
.
Austin
,
J.
, and
Coauthors
,
2010
:
Chemistry-climate model simulations of spring Antarctic ozone.
J. Geophys. Res.
,
115
,
D00M11
.
doi:10.1029/2009JD013577
.
Chen
,
G.
, and
I. M.
Held
,
2007
:
Phase speed spectra and the recent poleward shift of Southern Hemisphere surface westerlies.
Geophys. Res. Lett.
,
34
,
L21805
.
doi:10.1029/2007GL031200
.
de Grandpré
,
J.
,
S. R.
Beagley
,
V. I.
Fomichev
,
E.
Griffioen
,
J. C.
McConnell
,
A. S.
Medvedev
, and
T. G.
Shepherd
,
2000
:
Ozone climatology using interactive chemistry: Results from the Canadian middle atmosphere model.
J. Geophys. Res.
,
105
,
26475
26491
.
Eyring
,
V.
, and
Coauthors
,
2008
:
Overview of the new CCMVal reference and sensitivity simulations in support of upcoming ozone and climate assessments and the planned SPARC CCMVal Report.
SPARC Newsletter, No. 30, Stratospheric Processes and Their Role in Climate, World Climate Research Programme, 20–26
.
Fu
,
Q.
,
C. M.
Johanson
,
J. M.
Wallace
, and
T.
Reichler
,
2006
:
Enhanced mid-latitude tropospheric warming in satellite measurements.
Science
,
312
,
1179
.
Fyfe
,
J.
,
G.
Boer
, and
G.
Flato
,
1999
:
The Arctic and Antarctic oscillations and their projected changes under global warming.
Geophys. Res. Lett.
,
26
,
1601
1604
.
Gerber
,
E. P.
, and
Coauthors
,
2010
:
Stratosphere–troposphere coupling and annular mode variability in chemistry-climate models.
J. Geophys. Res.
,
115
,
D00M06
.
doi:10.1029/2009JD013770
.
Gillett
,
N. P.
, and
D. W. J.
Thompson
,
2003
:
Simulation of recent Southern Hemisphere climate change.
Science
,
302
,
273
275
.
Gong
,
D.
, and
S.
Wang
,
1999
:
Definition of Antarctic oscillation index.
Geophys. Res. Lett.
,
26
,
459
462
.
Houghton
,
J. T.
,
Y.
Ding
,
D. J.
Griggs
,
M.
Noguer
,
P. J.
van der Linden
,
X.
Dai
,
K.
Maskell
, and
C. A.
Johnson
,
Eds.
2001
:
Climate Change 2001: The Scientific Basis.
Cambridge University Press, 881 pp
.
Kalnay
,
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
Kushner
,
P. J.
, and
L. M.
Polvani
,
2004
:
Stratosphere–troposphere coupling in a relatively simple AGCM: The role of eddies.
J. Climate
,
17
,
629
639
.
Kushner
,
P. J.
,
I. M.
Held
, and
T. L.
Delworth
,
2001
:
Southern Hemisphere atmospheric circulation response to global warming.
J. Climate
,
14
,
2238
2249
.
Langematz
,
U.
,
M.
Kunze
,
K.
Krüger
,
K.
Labitzke
, and
G. L.
Roff
,
2003
:
Thermal and dynamical changes of the stratosphere since 1979 and their link to ozone and CO2 changes.
J. Geophys. Res.
,
108
,
4027
.
doi:10.1029/2002JD002069
.
Marshall
,
G. J.
,
A.
Orr
,
N. P. M.
van Lipzig
, and
J. C.
King
,
2006
:
The impact of a changing Southern Hemisphere annular mode on Antarctic Peninsula summer temperatures.
J. Climate
,
19
,
5388
5404
.
McLandress
,
C.
,
A. I.
Jonsson
,
D. A.
Plummer
,
M. C.
Reader
,
J. F.
Scinocca
, and
T. G.
Shepherd
,
2010
:
Separating the dynamical effects of climate change and ozone depletion. Part I: Southern Hemisphere stratosphere.
J. Climate
,
23
,
5002
5020
.
Perlwitz
,
J.
,
S.
Pawson
,
R. L.
Fogt
,
J. E.
Nielsen
, and
W. D.
Neff
,
2008
:
Impact of stratospheric ozone hole recovery on Antarctic climate.
Geophys. Res. Lett.
,
35
,
L08714
.
doi:10.1029/2008GL033317
.
Polvani
,
L. M.
,
D. W.
Waugh
,
G. J. P.
Correa
, and
S.-W.
Son
,
2011
:
Stratospheric ozone depletion: The main driver of twentieth-century atmospheric circulation changes in the Southern Hemisphere?
J. Climate
,
24
,
210
227
.
Randel
,
W. J.
, and
I. M.
Held
,
1991
:
Phase speed spectra of transient eddy fluxes and critical layer absorption.
J. Atmos. Sci.
,
48
,
688
697
.
Randel
,
W. J.
, and
Coauthors
,
2009
:
An update of observed stratospheric temperature trends.
J. Geophys. Res.
,
114
,
D02107
.
doi:10.1029/2008JD010421
.
Reichler
,
T.
,
M.
Dameris
, and
R.
Sausen
,
2003
:
Determining the tropopause height from gridded data.
Geophys. Res. Lett.
,
30
,
2042
.
doi:10.1029/2003GL018240
.
Scinocca
,
J. F.
,
N. A.
McFarlane
,
M.
Lazare
,
J.
Li
, and
D.
Plummer
,
2008
:
The CCCma third generation AGCM and its extension into the middle atmosphere.
Atmos. Chem. Phys.
,
8
,
7055
7074
.
Seidel
,
D. J.
, and
W. J.
Randel
,
2006
:
Variability and trends in the global tropopause estimated from radiosonde data.
J. Geophys. Res.
,
111
,
D21101
.
doi:10.1029/2006JD007363
.
Shepherd
,
T. G.
,
2002
:
Issues in stratosphere–troposphere coupling.
J. Meteor. Soc. Japan
,
80
,
769
792
.
Shine
,
K. P.
, and
Coauthors
,
2003
:
A comparison of model-simulated trends in stratospheric temperatures.
Quart. J. Roy. Meteor. Soc.
,
129
,
1565
1588
.
Sigmond
,
M.
, and
J. C.
Fyfe
,
2010
:
Has the ozone hole contributed to increased Antarctic sea ice extent?
Geophys. Res. Lett.
,
37
,
L18502
.
doi:10.1029/2010GL044301
.
Simmonds
,
I.
, and
K.
Keay
,
2000
:
Variability of Southern Hemisphere extratropical cyclone behavior, 1958–97.
J. Climate
,
13
,
550
561
.
Simpson
,
I. R.
,
M.
Blackburn
, and
J. D.
Haigh
,
2009
:
The role of eddies in driving the tropospheric response to stratospheric heating perturbations.
J. Atmos. Sci.
,
66
,
1347
1365
.
Son
,
S.-W.
, and
Coauthors
,
2008
:
The impact of stratospheric ozone recovery on the Southern Hemisphere westerly jet.
Science
,
320
,
1486
1489
.
Son
,
S.-W.
,
N. F.
Tandon
,
L. M.
Polvani
, and
D. W.
Waugh
,
2009
:
Ozone hole and Southern Hemisphere climate change.
Geophys. Res. Lett.
,
36
,
L15705
.
doi:10.1029/2009GL038671
.
Son
,
S.-W.
, and
Coauthors
,
2010
:
The impact of stratospheric ozone on Southern Hemisphere circulation change: A multimodel assessment.
J. Geophys. Res.
,
115
,
D00M07
.
doi:1029.2010JD014271
.
Song
,
Y.
, and
W. A.
Robinson
,
2004
:
Dynamical mechanisms for stratospheric influences on the troposphere.
J. Atmos. Sci.
,
61
,
1711
1725
.
SPARC CCMVal
,
2010
:
SPARC report on the evaluation of chemistry-climate models.
V. Eyring, T. G. Shepherd, and D. W. Waugh, Eds., SPARC Rep. 5, WCRP-132, WMO/TD-1526, 434 pp. [Available online at http://www.atmosp.physics.utoronto.ca/SPARC]
.
Thompson
,
D. W. J.
, and
J. M.
Wallace
,
2000
:
Annular modes in the extratropical circulation. Part I: Month-to-month variability.
J. Climate
,
13
,
1000
1016
.
Thompson
,
D. W. J.
, and
S.
Solomon
,
2002
:
Interpretation of recent Southern Hemisphere climate change.
Science
,
296
,
895
899
.
Thompson
,
D. W. J.
,
J. M.
Wallace
, and
G. C.
Hegerl
,
2000
:
Annular modes in the extratropical circulation. Part II: Trends.
J. Climate
,
13
,
1018
1036
.
Thompson
,
D. W. J.
,
J. C.
Furtado
, and
T. G.
Shepherd
,
2006
:
On the tropospheric response to anomalous stratospheric wave drag and radiative heating.
J. Atmos. Sci.
,
63
,
2616
2629
.
Thorncroft
,
C. D.
,
B. J.
Hoskins
, and
M. E.
McIntyre
,
1993
:
Two paradigms of baroclinic-wave life-cycle behaviour.
Quart. J. Roy. Meteor. Soc.
,
119
,
17
55
.
Waugh
,
D. W.
,
W. J.
Randel
,
S.
Pawson
,
P. A.
Newman
, and
E. R.
Nash
,
1999
:
Persistence of the lower stratospheric polar vortices.
J. Geophys. Res.
,
104
,
27191
27202
.
Wittman
,
M. A. H.
,
A. J.
Charlton
, and
L. M.
Polvani
,
2007
:
The effect of lower stratospheric shear on baroclinic instability.
J. Atmos. Sci.
,
64
,
479
496
.
WMO
,
2007
:
Scientific assessment of ozone depletion: 2006.
Global Ozone Research Monitoring Project Rep. 50, WMO, Geneva, Switzerland
.

Footnotes

Corresponding author address: Charles McLandress, Dept. of Physics, University of Toronto, 60 St. George St., Toronto, ON M5S 1A7, Canada. Email: charles@atmosp.physics.utoronto.ca