Changes of Southern Hemisphere westerlies in the future warming climate

The Southern Hemisphere westerlies (SHWs) play a key role in regulating global climate and ocean circulation, but their future changes under low to high greenhouse gas forcings remain unclear. This study investigates the long-term trends in strength and position of the SHWs and their linkage with human activities, based on the ERA5 reanalysis and model simulations from the Coupled Model Intercomparison Project phase 6 (CMIP6). The results show that the SHWs have intensified and shifted poleward in the recent decades, and are projected to experience divergent trends in strength and position during the 21st century under different Shared Socioeconomic Pathway (SSP) scenarios. Forced by SSP245, 370


Introduction
The Southern Hemisphere (SH) westerlies are the strongest timeaverage surface winds on the Earth (Russell et al., 2016).They are strongly located south of 40 • S and are traditionally known for their fierce storms over the southern oceans (Trenberth, 1991).Changes in the strength and position of the SH westerlies (SHWs) could exert huge impacts on global climate, regional weather extremes, ocean circulation, and the carbon uptake and heat transport in the SH oceans.For example, a poleward shift of moisture-bearing SHWs may lead to reduced rainfall and more frequent droughts in Southeast Australia (Holgate et al., 2020), stronger winds in New Zealand (Safaei Pirooz et al., 2019), warmer ocean surface and higher sea levels in the Tasman Sea (Duran et al., 2020), faster ice retreat in Patagonia (Boex et al., 2013), demise of the sub-Antarctic glaciers (Bakke et al., 2021), and increased warm and salty water transport from the Indian Ocean to the South Atlantic via the so-called Agulhas leakage (Biastoch et al., 2009).Paleoclimate records have further suggested that the changes in SHWs could regulate the pace of global climate change (Toggweiler et al., 2006;Fletcher and Moreno, 2011).That is, the intensification of SHWs may enhance the upwelling of carbon-rich deep water in the southern oceans (Waugh et al., 2013;Tamsitt et al., 2017;Ferrari, 2014), releasing more CO 2 into the atmosphere and thus causing faster global warming (Anderson et al., 2009;Skinner et al., 2010;Saunders et al., 2018).Therefore, it is important to study the past and future changes in the SHWs and to better understand their underlying causes.
Previous studies have shown that the SHWs have strengthened and shifted poleward in the last several decades due to global climate change (Toggweiler, 2009;Swart and Fyfe, 2012;Thomas et al., 2015).The trends in the strength and position of SHWs show zonal asymmetries: large strengthening and weak equatorward shift appear over the Pacific, in contrast to the weaker strengthening and significant poleward shift over the Atlantic and Indian Ocean sectors (Waugh et al., 2020).The trends in SHWs also exhibit seasonal differences in their magnitudes, with the most prominent strengthening trends during the austral warm seasons (Swart and Fyfe, 2012).The intensification of SHWs in the recent decades has been mainly attributed to the stratospheric ozone depletion over the Antarctica (Son et al., 2008) and the greenhouse gas (GHG) induced warming (Cai and Cowan, 2007).The strengthening of SHWs is likely to persist in the coming decades, given the ongoing trend in anthropogenic carbon emissions (Brown and Caldeira, 2017).In particular, GHG warming is expected to play a more crucial role during the 21st century, as the stratospheric ozone losses have recently stabilized due to the implementation of the Montreal Protocol (Banerjee et al., 2020).Moreover, the changes in atmospheric circulation in the polar and tropical regions could also affect the strength and position of the SHWs (Mindlin et al., 2020).For instance, the Southern Annular Mode (SAM), which is the most distinguished mode of atmospheric variability in the SH (Fogt and Marshall, 2020), is proposed to play a key role in modulating the changes in SHWs (Swart et al., 2015).The variations of tropical Pacific climate are also suggested to affect the strength and position of the SHWs (Yang et al., 2020).
Although it is widely agreed that the SHWs have experienced a poleward shift in the last decades (Toggweiler, 2009;Swart and   2012; Thomas et al., 2015), their future changes and associated causes are not well understood due to the relatively large uncertainties in model simulations.The Coupled Model Intercomparison Project (CMIP) phase 3, phase 5, and phase 6 models have been used to simulate the changes in the SHWs.In general, the three generations of models have correctly simulated the signs of trends in the strength and position of SHWs over the historical period (Goyal et al., 2021).However, the SHWs simulated by the CMIP3 and CMIP5 models are on average weaker and more equatorward shifted relative to observations, with equatorward shift bias of approximately 4 • of latitude in the CMIP3 models (Kidston and Gerber, 2010) and 3.3 • of latitude in the CMIP5 models (Bracegirdle et al., 2013).The SHWs trends during the 21st century show even larger uncertainties in the CMIP3 and CMIP5 projections: the projected poleward shift of SHWs ranges between 0.5 • and 10 • latitudes and the projected intensification of SHWs is estimated to be from 10% to over 300% (Goyal et al., 2021).In contrast, there are substantial reductions in the equatorward bias and the strength of the SHWs in CMIP6 (Bracegirdle et al., 2020).These evaluation studies suggest that the CMIP6 models are more reliable for studies of the future projections of the SHWs.
In this study, we will be focusing on the long-term changes in the SHWs, using the state-of-the-art ERA5 reanalysis data and CMIP6 model simulations.In particular, this study is an important addition to Goyal et al. (2021), which has previously analyzed the historical and projected changes in the SHWs.Compared to Goyal et al. (2021), this study not only comprehensively illustrates the SHWs changes on seasonal basis and regional scale, but also explores the possible underlying causes.Different from previous studies, most of which have suggested a future strengthening of the SHWs, our study reveals that there exists a possibility to interrupt and even reverse the ongoing trends in the SHWs under a low GHG emission scenario in the future.
The remainder of this paper is organized as follows: Section 2 describes the data and methods.Analysis results are presented in Sections 3 and 4. A summary of our main findings and concluding remarks are given in Section 5.

Reanalysis data
The SHWs are located over the southern oceans, where station observations are sparse.Global atmospheric reanalysis products are thus useful in the studies of SH oceanic winds, due to their advantages of consistent time series and complete spatial coverage.Previous studies have evaluated the skills of six recent global reanalysis datasets in representing the 10-m wind speed in the SH, i.e. the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) (Gelaro et al., 2017), the Japan Meteorological Agency 55-Year Reanalysis (JRA-55) (Kobayashi et al., 2015), the Climate Forecast System Reanalysis (CFSR) (Saha et al., 2014), the National Centers for Environmental Prediction-U.S.Department of Energy (DOE) Reanalysis 2 (NCEP-R2) (Kanamitsu et al., 2002), the European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis (ERA-Interim) (Dee et al., 2011), and the ECMWF fifth-generation reanalysis (ERA5) (Hersbach et al., 2020).The ERA5 reanalysis shows the best performances for the monthly magnitude and interannual variability of the 10- m wind speed in the high latitudes from 1979 onward (Dong et al., 2020).Therefore, in this study, the historical trends in strength and position of the SHWs are evaluated based mostly on the ERA5 reanalysis, with a horizontal resolution of 0.25 • × 0.25 • for the period of 1979-2020, which is available in the Climate Data Store at website of (https://cds.climate.copernicus.eu/#!/search?text=ERA5&type=dataset).

CMIP6 model simulations
The simulation datasets are from the outputs of the CMIP6 (Eyring et al., 2016), which can be downloaded through the portal of the Lawrence Livermore National Laboratory (https://esgf-node.llnl.gov/search/cmip6/; last access on 1 July 2021).The CMIP6 historical experiments are designed to simulate the observed variables in the climate system from 1850 to 2014 and to project the changes in the climate system after 2015.Twenty CMIP6 models used in this study are listed in Table 1.They all have produced historical simulations for 1850-2014 and are extended for 2015-2099 under the Shared Socioeconomic Pathway (SSP) scenarios (O'Neill et al., 2016).Note that although most models have produced datasets to 2100, only a few (such as the CAMS-CSM1-0) run to the year 2099.Therefore, when calculating the future changes in wind speed, relative to the historical period 1981-2010, we simply define the future period as 2070-2099, instead of 2071-2100.In addition to the wind speed datasets, precipitation (in mm), 2-m air temperature (in • C), SST at a single level (in • C), and 3dimensional winds (in m s − 1 ) at multiple pressure levels in CMIP6 simulations and projections are also analyzed.1g and h.
Compared to the Representative Concentration Pathways scenarios that are used in CMIP5, the SSP scenarios provide more comprehensive assumptions about social, technical, cultural, educational, and economic developments during the 21st century (O'Neill et al., 2016).In this study, four SSP scenarios are selected, including SSP126, SSP245, SSP370, and SSP585.Briefly, SSP126 describes a green road that leads to sustainable development, using environmentally friendly technologies and renewable energy, under which would have low challenges to mitigation and adaptation.SSP245 represents a scenario in which the current social, economic, and technological trends continue, which poses moderate challenges to mitigation and adaptation.The SSP370 leads to a rocky road, with rapid population growth, resurgent nationalism, and regional conflicts, posing high challenges to mitigation and adaptation.By comparison, SSP585 is driven by fossil-fueled development, implying global adoption of resource and energy-intensive lifestyles, with higher challenges to mitigation.Note that the Paris Agreement's goal is to limit future global warming well below +2 • C, preferably to +1.5 • C, compared to the pre-industrial level (Rogelj et al., 2016).To achieve this goal, countries need to reach the peak of global greenhouse gas emissions as soon as possible to achieve a climateneutral world by the middle of the 21st century.Among the four SSPs analyzed in this study, SSP126 is the only pathway where the "well below +2 • C" goal of the Paris Agreement could be met.

Methods
The twenty CMIP6 models produced datasets with various spatial resolutions.To facilitate a comparison, all of the model simulations were interpolated onto 2 • × 2 • (180 × 91) grid resolution using a bilinear interpolation method (Mastyło, 2013).The magnitude of the wind speed is calculated as , where u and v are the zonal and meridional components of the 10-m winds (in m s − 1 ), respectively.The values of annual and seasonal mean wind speeds were derived by averaging the monthly results.The austral seasons, i.e. winter, spring, summer, and autumn are defined as three month average of JJA (June-July-August), SON (September-October-November), DJF (December-January-February), and MAM (March-April-May), respectively.
The strengths of SHWs are measured by averaging the 10-m wind speed over 60 • S to 50   (Marshall, 2003;Ho et al., 2012).The tropical precipitation time series are derived by the average of precipitation over 5 • S-5 • N/0 • -358 • E.

Changes in strength and position of the SHWs
Fig. 1 shows the annual mean 10-m wind speed in the SH.The multimodel ensemble mean has realistically reproduced the distribution and magnitude of the wind speed climatology over 1981-2010 (Fig. 1a and  b).Wind speeds greater than 9 m s − 1 are found in the latitudes south of 40 • N and are accurately simulated by the models.Stronger winds appear over the Indian and Atlantic Ocean basins in contrast to the Pacific Ocean, which is also correctly simulated by the models.Further examinations of the individual model simulations reveal that the multi-model ensemble mean can substantially improve the simulation skills for the SH wind speed climatology (Fig. S1).
Figs. 1c to 1f further display the spatial patterns of 10-m wind speed changes between future (2070-2099) and historical (1981-2010) periods, driven by the four SSP scenarios.In general, positive (negative) wind speed changes are found on the southern (northern) flank of the SHWs axis, suggesting that the SHWs will continue to shift poleward during the 21st century.Consistent with the multi-model ensemble mean, the majority of the CMIP6 models have simulated poleward intensified SHWs over the historical period (Fig. S2) and projects a further southward shift in the future (Fig. S3).
Fig. 1g and h present the time series of strength and position of the SHWs during 1979-2099, respectively.For the historical period, i.e., prior to 2015, the SHWs strengthened and moved southward.However, after 2015, divergent trends in the SHWs occur, depending on the choice of future SSP scenarios.If the world follows the pathways with uneven development (SSP245), resurgent nationalism and regional rivalry (SSP370), and fossil-fueled development (SSP585), the SHWs are projected to strengthen and shift poleward in 2015-2099.The higher level the SSPs, the larger the trends.By comparison, if the world shifts toward a more sustainable road with major carbon reductions (SSP126), the SHWs are projected to weaken and shift equatorward.
The projected changes in the SHWs are basically consistent among the three ocean basins (Fig. 2).Relative to 1981-2010, the SHWs during 2070-2099 will strengthen distinctly (Fig. 2a) and move closer to the Antarctica (Fig. 2b) in the Pacific, Atlantic, and Indian Oceans, under middle to high GHG forcings.This is particularly marked under the SSP585 scenario, when the SHWs will intensify by 1 m s − 1 and shift southward by 2 • -3 • of latitude by the end of the 21st century.As seen from Figs. 2c to 2e, positive (negative) trends in wind speed are observed in latitudes south (north) of 50 • S under SSP245, SSP370, and SSP585 scenarios, implying that the SHWs will experience poleward displacement.
By comparison, wind speed trends forced by the SSP126 are opposite to those forced by the higher SSPs, indicating an equatorward displacement of the SHWs in the future lower GHG forcing.These results show that SSP126 is the only scenario among the four that could trigger opposite trends in the SHWs during 2015-2099, especially after the mid-21st century.In other words, if global socioeconomic policies follow the Paris Agreement and drastically cut carbon emissions, we would have a chance to interrupt and even reverse the ongoing trends in the SHWs.Otherwise, continued strengthening of the SHWs would be likely inevitable throughout the 21st century.Note that most of previous studies reported an unanimously intensified and poleward trend in the SHWs.Thus, this discovery of weakened and equatorward shifted SHWs under a low emission scenario is an important addition to the existing academic literature, which also provides important implications for global climate change and its mitigation.
The trends in the strength and position of the SHWs show seasonal differences in their magnitudes and uncertainties (Fig. 3).During the historical period, the trends in SHWs strength are basically consistent among the four seasons.Nevertheless, the trends in position of the SHWs over the historical period are less robust in austral cold seasons than in austral warm seasons.In particular, during JJA and SON, the uncertainties of the trends in the SHWs exceed their magnitudes (see subplots of Fig. 3), implying that the remarkable changes in intensity and position of the SHWs mainly occur in the austral warm seasons.During future period 2015-2099, the projected trends in strength and position of the SHWs are tied to the choices of SSPs.Similar to the annual mean, the seasonal SHWs forced by SSP245, SSP370, and SSP585 will continue to intensify and shift poleward in all seasons, while the SHWs forced by SSP126 will weaken and move equatorward, which is especially true for the austral warm seasons.

Possible causes for projected changes in the SHWs
Further examination shows that the changes in the SHWs are closely linked to global warming intensity.As shown in Fig. 4, under SSP245, SSP370, and SSP585, the increases in global mean 2-m air temperature will exceed +2.0, +3.0, and + 4.0 • C, respectively, by the end of the 21st century.Correspondingly, the SHWs will intensify by more than +0.2, +0.3, and + 0.4 m s − 1 during austral warm seasons (e.g., DJF) and more than +0.1, +0.2, and + 0.3 m s − 1 during austral cold seasons (e.g., JJA).In contrast, under SSP126, the increases in global mean 2-m air temperature correspond to slight decreases in the strength of SHWs, which are particularly prominent during DJF.These results suggest that anthropogenic warming could have regulated and will influence the changes in the SHWs.
The GHG-induced surface warming is unevenly distributed in the SH (Fig. 5): under middle to high GHG forcing, the largest warming exists in the sub-Antarctica coastal regions (around 60 • S latitude); and the continent land areas seem to warm more rapidly than the surrounding oceans.As a result, the equator-to-pole temperature gradient in areas south (north) of 60 • S latitude will likely increase (decrease) under SSP245, SSP370 and SSP585, which could cause strengthened (weakened) westerly wind speed and result in the poleward displacement of the SHWs.Under SSP126, however, strong warming mainly exists in the Antarctic continent, leading to reduced equator-to-pole temperature gradient in lower latitudes and thus causing weakened SHWs.We note that under middle to high SSP scenarios the largest warmings appearing in the sub-Antarctica instead of the pole regions could be linked to the local atmosphere -sea ice -SST interactions, simply like those in the Arctic (Screen and Simmonds, 2010), where such feedback mechanisms are absent in the Antarctica continent due to the thick glaciers and lack of SST.Under SSP126, however, the surface temperature warmings are much smaller than higher SSPs, which could be insufficient to melt sea ice in the sub-Antarctica regions and thus fail to initiate the sea ice -SST air temperature feedbacks there.
Moreover, thermal contrast between the continent land areas and the surrounding oceans will also affect SHWs via modulating the sea level pressures.In particular, thermal contrast between the Antarctic continent and the mid-latitude oceans is stronger than in other regions, and this will lead to most pronounced anomalies of sea level pressure in the Antarctica and the mid-latitude oceans.As shown in Fig. 6, under SSP245, SSP370 and SSP585, sea level pressures over the Antarctica (mid-latitude oceans) will experience decreasing (increasing) trends; however, under SSP126, these trends will be interrupted due to low GHG forcing and weak thermal contrast.
The changes in sea level pressures are reminiscent of the Southern Annular Mode (SAM) or the Antarctic Oscillation, which is the most distinguished mode of atmospheric variability in the SH (Marshall, 2003;Ho et al., 2012).As shown in Fig. 7, the SAM has strengthened over the historical period, which is correctly simulated by the CMIP6 models.Moreover, the SAM is projected to intensify rapidly throughout the 21st century under SSP245, SSP370 and SSP585, and to weaken slowly under SSP126.The trends in SAM are largest in MAM and weakest in JJA, which are consistent with the changes in SHWs.In fact, during the austral warm seasons, the changes in SAM show a nearly linear correlation with the changes in the SHWs, while during the austral cold season such linear correlation are absent (Fig. 8).
In addition to the SAM, atmospheric circulation changes driven by tropical warming may also influence the SHWs.As shown in Fig. 9, tropical precipitation is projected to increase in all seasons under SSP245, SSP370, and SSP585 during the 21st century, with the largest trends forced by SSP585.In contrast, under SSP126, the trends in tropical precipitation are much smaller.The largest trends in tropical precipitation occur in DJF and MAM, consistent with the SHWs changes.Further analysis reveals that the future increases in tropical precipitation are unevenly distributed, being mainly concentrated over the Pacific (Fig. S4).The enhanced tropical Pacific precipitation is accompanied by stronger uplift over this region, and by anomalously descending motion over the SH higher latitudes (Fig. S5).In particular, anomalous rising air will be forced over the equator under the high GHG forcing, accompanied by alternative upward and downward motions over the extratropical regions (see subplots of Fig. 9).Note that the interactions between tropical and extratropical atmosphere via the meridional overturning circulation (i.e., the Hadley cell) have been widely reported (Nguyen et al., 2018;Xia et al., 2020;Lucas et al., 2021), which can readily explain the projected intensification of the SH Pacific trade winds (Fig. 2c).
The changed meridional atmospheric circulation, driven by SAM and tropical convection, converges over the mid-to-high latitudes in the SH.The strongest interactions between tropical and polar atmospheric circulations occur during DJF and MAM (see subplots of Fig. 9), because the tropical Pacific convection is mainly located over the Northern Hemisphere during JJA and SON and the changes in the SAM are smaller in these seasons than in MAM and DJF.The enhancements of the meridional atmospheric circulation due to increased GHG emissions in SSP245, SSP370, and SSP585 are likely to cause strengthened and southward displaced SHWs.Conversely, under SSP126, the SAM phases will be reversed and the tropical precipitation increases will slow down in the future, both of which could contribute to weakened meridional atmospheric circulations in the SH and thus result in the decelerated and equatorward shifted SHWs.

Summary and discussion
This study investigates the past and future changes in SHWs based on the ERA5 reanalysis and simulations from 20 CMIP6 models.The major conclusions were summarized as follows.(1) The SHWs have intensified and shifted poleward in the past several decades.They are projected to strengthen and shift poleward during 2015-2099 under the SSP245, SSP370 and SSP585 scenarios, but will eventually weaken and shift equatorward under the SSP126 scenario.(2) The changes in SHWs are basically consistent in the SH ocean basins, but show seasonal differences in magnitudes.The largest and most robust trends in the strength and position of the SHWs exist in the austral warm seasons (DJF and MAM).(3) Human activities may affect the SHWs by modulating the atmospheric circulation in the SH.In particular, the SAM and tropical Pacific convection could play critical roles in regulating the strength and position of SHWs.
This study is built largely based on Goyal et al. (2021), which has previously analyzed the historical and projected changes in the SHWs.Compared with Goyal et al. (2021), this study presents a few similarities and differences.First, both have investigated the changes of SHWs in different ocean basins and for the four seasons.Second, both have proposed that the SHWs will continue to intensify and shift poleward in a persisting warming future.Nevertheless, different from previous literatures, our study reveals that there exists a possibility to interrupt and even reverse the ongoing trends in the SHWs under a low GHG emission future.This discovery has not been reported before and would have strong implications for climate change and its mitigation.We also have explored the possible causes for the projected changes in SHWs, especially discussing the roles of tropical and high-latitude atmospheric circulations in regulating the SHWs.
Our study depicts future developments of SHWs depending on SSP scenarios.It also suggests a possible way through which human activities may influence the pace of climate change.On the one hand, if the world follows the socioeconomic development pathways with middle to high GHG emissions (SSP245, SSP370, and SSP585), global warming will persist and intensify during the 21st century, which will cause enhanced SAM and increased tropical Pacific precipitation, leading to strengthened SHWs.The strengthened SHWs can enhance the upwelling of carbon-rich deep water in the southern oceans, releasing more CO 2 into the atmosphere and causing faster global warming (e.g., Toggweiler et al., 2006;Fletcher and Moreno, 2011;Waugh et al., 2013).As a result, a positive feedback cycle involving GHG warming, stronger SHWs, increased atmospheric CO 2 concentration, and stronger GHG warming, could become established in the SH, seriously threatening global climate stability with a serious impact on the SH.On the other hand, if the world follows the more conservative pathway (SSP126), which requires that we follow the Paris Agreement recommendations and limit future warming below +2 • C (ideally to +1.5 • C), the proposed feedback mechanism would be interrupted and the SHWs will gradually decelerate.
Finally, we emphasize that the SHWs could also be affected by the changes in stratospheric ozone, which plays a key role in the dynamics of atmospheric circulation in the Antarctica.In historical period, the Antarctica ozone losses have recently stabilized due to the implementation of the Montreal Protocol (Banerjee et al., 2020); in future warming climate, the Antarctica ozone is projected to gradually recover in all CMIP6 SSP scenarios (Shang et al., 2021).It is expected that the future changes in Antarctica ozone would likely further influence the atmospheric circulation in the SH, which deserves further monitoring.However, in this research, we mainly explore the projected changes in the SHWs and their possible linkage to anthropogenic GHG emissions.Our choices of socioeconomic pathways will determine the magnitude of GHG emissions and the strength of trends in the SHWs, thus exerting a possible impact on climate stability and resilience in the future.Global collaboration and enormous efforts are essential if we are to cut carbon emissions and avoid the irreversible consequences of global warming.

Fig. 1 .
Fig. 1.Changes in the strength and position of the annual SHWs.(a)-(b) Reanalyzed and simulated climatological patterns of annual mean 10-m wind speed during 1981-2010.(c) to (f) Differences in simulated wind speed (2070-2099 minus 1981-2010), under the SSP126, SSP245, SSP370, and SSP558 scenarios, respectively.(g) Strength of the annual SHWs, defined as zonal wind speed averaged over 60 • S to 50 • S. (h) Position of the annual SHWs, defined as latitudes of the strongest zonal mean 10-m wind speed.All of the time series are expressed as the anomalies from the climatology (1981-2010).Contours in a-f indicate the 9 m s − 1 threshold of climatological 10-m zonal wind speed.Shadings in g-h denote the model spreads, representing one standard deviation of the individual models' departures from the twenty-model mean.A 10-yr running mean was applied to the reanalyzed time series.The trends in strength and position of the SHWs during 1979-2014 and 2015-2099 as well as their uncertainties are shown respectively in the subplots within (g) and (h).

Fig. 2 .
Fig. 2. Changes in the annual SHWs.(a) Strength of SHWs averaged over the historical period 1981-2010 and future period 2070-2099.(b) Similar to (a) but for the position of the SHWs.Shadings represent the model spread defined as one standard deviation of the individual models' departures from the twenty-model mean.(c)-(e) Zonal mean of the projected wind speed trends (multi-model ensemble mean) for 2015-2099 over the Pacific Ocean (150 • E-70 • W), Atlantic Ocean (70 • W-25 • E), and Indian Ocean (25 • E-150 • E) basins in the SH, respectively, driven by SSP126, SSP245, SSP370, and SSP585 scenarios.Dashed black curves denote the climatology of zonal mean 10-m wind speed.

Fig. 3 .
Fig. 3. Interannual variations of the strength and position of seasonal mean SHWs.(a)-(b), (c)-(d), (e)-(f) and (g)-(h) are similar to Figs. 1(g)-(h), but for MAM, JJA, SON, and DJF, respectively.The trends in strength and position of the SHWs during 1979-2014 and 2015-2099 are shown in the corresponding subplots.Shadings and error bars denote the model spreads of one standard deviation.The time series for ERA5, historical all forcing, and future SSP scenarios are drawn in different colors, which are the same with those in Fig. 1g and h.
• S and 0 • to 358 • E. Note that although the SHWs exist in latitudes between 70 • S and 40 • S, they are strongest and mostly located over the oceans from 60 • S to 50 • S, and a small change of the latitudes will not affect the main conclusions in this study.Besides, the positions of SHWs are identified by the latitudes of peak magnitudes of the SH zonal mean 10-m wind speed.When assessing the future projected changes in 10-m wind speed, we calculate the differences in the wind speed between 2070 and 2099 and 1981-2010 for the annual mean and each season.The time series of strength and position of the SHWs during 1979-2014 for historical all forcing and during 2015-2099 for the four SSP scenarios, are computed separately for the twenty models.Then, the multi-model ensemble mean was applied to all time series to obtain more robust results.The inter-model spreads are assessed based on one standardized deviation of the individual models' departures from the twenty-model mean.Similar to Goyal et al. (2021), we also examine the SHWs changes in different ocean basins.The Pacific, Atlantic, and Indian Ocean basins in the SH are roughly divided by 150 • E-70 • W, 70 • W-25 • E, and

Fig. 4 .
Fig. 4. Relationships between the strength of SHWs and global warming levels.(a), (b), (c), and (d) are for MAM, JJA, SON, and DJF, respectively.The historical simulations and future projections were driven by historical all forcing over 1979-2014 and the SSP scenarios over 2015-2099.Changes in the SHWs' strength and global mean 2-m air temperature are expressed as anomalies from the base 1981-2010.The changes in SHWs' strength per 1 • C global warming under the SSP585 scenario are drawn in each panel.

Fig. 6 .
Fig. 6.Projected trends in sea level pressure over the SH during 2015-2099.(a), (b), (c) and (d) indicate the trends in sea level pressure (unit: Pa dec − 1 ) for MAM, JJA, SON, and DJF, respectively.The first to fourth columns denote the trends driven by SSP126, SSP245, SSP370, and SSP585, respectively.The latitude range is 90 • S-30 • S, with an interval of 30 • of latitude.

Fig. 7 .
Fig. 7. Historical and projected changes in seasonal mean SAM.(a) Time series are for SAM indices in MAM for 1979-2099, driven by historical forcing (1979-2014) and four SSP scenarios (2015-2099).(b)-(d) Similar to (a) but for JJA, SON, and DJF, respectively.Note that the time series have been normalized using their means and standardized deviations over 1981-2010.The subplots display the future projected changes in sea level pressure (shading, unit: Pa) and 850-hPa horizontal winds (vectors, unit: m s − 1 ) between 2070 and 2099 and 1981-2010, driven by the SSP585 scenario.

Fig. 8 .
Fig. 8. Relationships between the SHWs strength and the SAM.(a), (b), (c) and (d) are for MAM, JJA, SON and DJF, respectively, driven by SSP126 and SSP585 over 2015-2099.Changes in SHWs' strength are expressed as the anomalies from the base 1981-2010, and the SAM indices are normalized using their historical mean and standardized deviation.

Fig. 9 .
Fig. 9. Historical and projected changes in the seasonal mean tropical precipitation and meridional atmospheric circulation.(a) Time series are for the tropical precipitation averaged over 5 • S-5 • N/0-358 • E for MAM over 1979-2099, driven by historical forcing (1979-2014) and four SSP scenarios (2015-2099).(b)-(d) Similar to (a) but for JJA, SON, and DJF, respectively.All time series are expressed as the anomalies from the climatology over 1981-2010.In the subplots, vectors (omega multiplied by a factor of 100, unit: m s − 1 ) denote the differences in zonal-mean vertical and meridional winds (2070-2099 minus 1981-2010), forced by the SSP585 scenario; and shadings denote the changes in omega, where negative values indicate upward motions.

Table 1
Fyfe, Model name, country, nominal resolution, and variant label of twenty CMIP6 global climate models analyzed in this study.