3.1 Sea surface height
Due to the scarcity of direct measurements on ocean currents and the rarity of extensive time series data, we use satellite SSH products to derive changes in ocean circulations (Hill et al. 2011; Behrens et al. 2020; Graham and De Boer 2013; Rühs et al. 2022). We first examine the similarity between observed and simulated SSH in the Southern Hemisphere. Figure 1 shows the time-averaged SSH patterns during 1993–2020 from the observation (Fig. 1a), the CanESM5 ensemble (Fig. 1b), the CESM ensemble (Fig. 1c), and the multi-model ensemble (Fig. 1d). It can be seen that all model simulations faithfully reproduce the structures of the South Pacific Gyre (SPG), the South Indian Gyre (SIG), and the South Atlantic Gyre (SAG), as well as their interconnections, and that the three ocean gyres are connected by the Tasman and Agulhas Leakages. Since satellites only measure variability at the sea surface, we also examine the similarity between Argo-derived and simulated velocity fields. The similarity of vertically averaged (0–2,000 m) horizontal flow fields again confirms the reliability of the models (Fig. 1). However, in contrast to the detailed representation in the model data, Argo does not adequately sample the narrow Western Boundary Currents (WBCs) due to the limited sampling and heavy smoothing used in the Argo product (Send et al. 2010; Riser et al. 2016; Chandler et al. 2022).
Since the center of the gyre is identifiable as the SSH maximum, we focus on regions of high SSH (black boxes in Fig. 1a) and compare the change in each ocean basin (Fig. 2). In the South Indian and South Atlantic Oceans, the observation shows significant and increasing trends in SSH from 1993–2020, and the three large ensemble means show essentially the same upward trends from 1955–2020 (CanESM5 & multi-model/CESM historical simulation before 2014/2005 and then Shared Socioeconomic Pathway 585 (SSP585)/Representative Concentration Pathway 8.5 (RCP8.5) thereafter until 2020 to ensure time consistency with the observation). However, the SPG center has been remarkably stable in recent decades, with the growth rate being small and insignificant from both observations and model simulations. We conduct a correlation analysis on the historical SSH trends of the three model data from 1955–2020 and the observed data from 1993–2020 (Supporting information Fig. S1). The correlation coefficients between these data results indicate significant consistency, suggesting that all types of models are effective in simulating the changes of SSH in the Southern Hemisphere, which is consistent with previous studies (Landerer et al. 2014; Ferrero et al. 2021).
The use of satellite and Argo float data displays the trend of subtropical gyres in the Southern Hemisphere, but they have inherent limitations that affect their usefulness for certain aspects of oceanic research. Satellite data are restricted to surface observations, Argo floats sample primarily the upper 2,000 meters of the global ocean, but are limited near ocean boundaries and in narrow currents. In addition, the time span covered by both is relatively short. Since model simulations can faithfully reproduce the major patterns and temporal evolution of observations, the following analysis will primarily focus on analyzing model results to comprehensively understand the SHSG variability and to investigate the influence of various external forcings since the 1920s.
3.2 Streamfunction
To quantify the response of the SHSG to different climate forcings, we examine the total mass transport streamfunction. As shown in Fig. 3a, the historical simulation in the CanESM5 ensemble captures the mean pattern of the SHSG well. Figures 3b&c show the time series during 1920–2014 and the linear trends during 1955–2014 of the change in the zonal mean southern boundary of SHSG in the historical simulation and the contributions of each individual forcing. Here, the zero mass transport streamfunction line is utilized as the southern boundary of the SHSG (black line in Fig. 3a). The historical simulation reveals a remarkable poleward shift of the southern boundary of the SHSG at a rate of about 0.11° per decade since the 1950s (black line and bar in Figs. 3b&c). This movement is primarily attributed to the influence of GHG (red line and bar in Figs. 3b&c), while the OD forcing makes a secondary contribution (yellow line and bar in Figs. 3b&c). Under AER forcing, however, there is a minor equatorward shift in the southern boundary of the SHSG (blue line and bar in Figs. 3b&c).
Similar results are found in the CESM ensemble and the multi-model ensemble (Supporting information Figs. S2 and S3). Taken together, these results show that the SHSG exhibits a significant trend of poleward shift in the historical simulations primarily due to GHG forcing, with a secondary influence from OD forcing.
3.3 Inter-ocean connection
The Tasman and Agulhas Leakages are the two major components of the SHSG that connect the ocean basins. In this subsection, we examine how the water transport through these two leakages changes as the SHSG shifts toward the polar regions.
The WBC in the South Pacific, known as the East Australian Current (EAC), is a component of the SPG and bifurcates into two distinct pathways at 30°-34°S (Cetina-Heredia et al. 2015; Bull et al. 2018; Oke et al. 2019). After departing from the northern coast of New Zealand, a portion of the water flows westward from the Tasman Front, traverses the Tasman Sea, and forms the East Auckland Current. At the same time, another portion of the water flows along the Australian coast and continues until it reaches the eastern side of Tasmania Island. This is referred to as "the EAC extension", which follows the Tasmania coastline and ultimately enters the southeastern Indian Ocean through the Tasman Leakage (Ridgway and Dunn 2003; Cai et al. 2005; Sutton and Bowen 2014; Chiswell et al. 2015; Sloyan et al. 2016). This unique process allows water to flow westward along the coast and contributes to the formation of the SHSG in the southeastern Indian Ocean. Building on this understanding of the complex pathways of the EAC, the following analysis leverages Argo and model data to capture these oceanic pathways.
To the west of Tasmania, the main path of the Tasman Leakage is quite stable. Based on the 141–143°E section, the Tasman Leakage is defined as the meridional integral of the westward flow in the upper 2,000 m between 40°S and 50°S from Argo observations (Fig. 4a). Similarly, the pattern of zonal velocity along the same section in the CanESM5 ensemble shows a resemblance to the observations (Fig. 4b). Recent studies suggested that the time-averaged transport of the Tasman Leakage is between 8 and 11 Sv (Ridgway and Dunn 2007; Qu et al. 2019; Behrens et al. 2020), in our study the transport is about 9.7 Sv in Argo and 8.9 Sv in the CanESM5 ensemble. The CanESM5 ensemble results provide insights into the time series of the Tasman Leakage transport during 1920 to 2014, showing a significant upward trend since the mid-1950s at a rate of about 0.3 Sv per decade (black line in Fig. 4c). In the GHG-only simulation, the Tasman Leakage transport also exhibits an increasing trend (red line in Fig. 4c), the magnitude of which is largely consistent with the historical simulation. Conversely, in the AER-only simulation, the Tasman Leakage transport shows a slight weakening, comparable to the subdued enhancement in the OD-only simulation (blue and yellow lines in Fig. 4c).
The water exchange between the Southern Indian Ocean and the Southern Atlantic Ocean is primarily attributed to the Agulhas Leakage, with the primary transport mechanism being the Agulhas rings, anticyclonic formations resulting from the retroflection of the Agulhas Current (AC; Gordon and Haxby 1990; Schouten et al. 2000; Rouault et al. 2009; Dencausse et al. 2010). The highly nonlinear and complex system involves how the WBC in the South Indian Ocean can either exit eastward through the Agulhas Return Current and reconnect to the Indian Ocean subtropical gyre, or westward into the Atlantic Ocean through the Agulhas Leakage (Biastoch et al. 2009; Dencausse et al. 2010; Beal et al. 2011; Durgadoo et al. 2013; Zhang et al. 2023). In contrast to the Tasman Leakage, this intricate pathway predominantly involves meso- to small-scale features (Reason et al. 2003; Van Sebille et al. 2009), which are difficult to capture comprehensively.
We select the mean zonal velocity across the section between 15–17°E, located west of the Agulhas Current retroflection point at 19°E, which exhibits a rapid decline in eddy-driven transport and provides a stable measurement of the Agulhas Leakage (Van Sebille et al. 2010). The mean transport of the Agulhas Leakage obtained from Argo is defined as the meridional integral of westward flow in the upper 2,000 m of the 15–17°E section, spanning latitudes 33–44°S (Fig. 4d). Similarly, the CanESM5 ensemble successfully captures this inter-ocean connection of the SHSG (Fig. 4e) in the same section. Figure 4f shows the time series of the Agulhas Leakage transport in the historical simulation and three climate forcing simulations. Under GHG forcing, there is an increasing trend since the 1950s that is significant at the 95% confidence level (red line in Fig. 4f), similar to the historical simulation (black line in Fig. 4f). AER and OD forcings have relatively minor contributions (blue and yellow lines in Fig. 4f).
Therefore, it is the GHG forcing that plays a dominant role in driving the historical changes of transports at both the Tasman and Agulhas Leakages, while the AER and OD forcings have much smaller effects. Significantly, this increase in inter-ocean connections coincides with the southward shift of the SHSG, which is also primarily attributed to the GHG forcing. Such consistency suggests that a underlying coherent process is at work, possibly related to changes in the wind stress curl (Cai et al. 2005; Bostock et al. 2006; Hill et al. 2008; Oliver and Holbrook 2014)
In summary, both inter-ocean connections within the SHSG show the increasing transport trend since the 1950s, and the consistent patterns are also reproduced in the CESM ensemble and the multi-model ensemble (Supporting information Figs. S4 and S5). The climate model analyses indicate that the GHG forcing primarily induces the spin-up of SHSG, while the influences of OD and AER are less significant. Next, we will perform a comprehensive analysis of the internal ocean basin circulation to better understand the influences of different external forcings and discuss the mechanisms behind these changes.
3.4 Internal ocean basin circulation
To better understand the variability of the inter-ocean connection, in this subsection we examine the circulation within each ocean basin. We focus first on the northern limbs of the subtropical gyres, which turn southward after reaching the western boundaries and supply water to the Tasman and Agulhas Leakages in the Pacific and Indian Oceans, respectively.
Previous studies have indicated that the South Equatorial Current (SEC) bifurcates around 17°S-18°S in the western ocean basins, with its southern branch transporting water poleward to close the subtropical gyres and contribute to inter-basin exchange, and its northern branch moving equatorward to close the tropical gyres (Rodrigues et al. 2007; Chen and Wu 2015; Yamagami and Tozuka 2015). Following Qu et al. (2019), in this study the northern limb of the subtropical gyre is represented by the southern branch of the SEC, which is defined by integrating the meridional westward flow between 20°S and 40°S in the upper 2,000 m in the 170°W section of the Pacific Ocean, the 60°E section of the Indian Ocean, and the 30°W section of the Atlantic Ocean (Fig. 5). The results indicate that in both the Indian and Atlantic Oceans (black lines Figs. 5a&c), transports in the northern limbs of the subtropical gyres show increasing trends since the 1950s, contributing additional water to the higher latitudes, and these trends are driven by the GHG forcing (red lines in Figs. 5a&c). In contrast, the AER forcing drives decreasing trends in transports (blue lines in Figs. 5a&c), but its magnitudes are significantly smaller than the GHG forcing. The OD forcing does not produce significant changes in transports for the northern limbs of the subtropical gyres (yellow lines in Figs. 5a&c).
However, the transport through the northern limb of the SPG does not show a clear upward trend (Fig. 5b). The EAC bifurcates into a flow along the Tasman Front and a flow that continues south as the EAC extension. Although the EAC does not have an increasing trend in transport, the EAC extension supplies more water to the Tasman Leakage due to a reduction in transport through the flow along the Tasman Front (black line in supporting information Fig. S6a), which is defined as a meridional integral eastward flow in the upper 2,000 m between 30°S and 40°S across the 165°E section. The strengthening of the EAC extension - a part of the SHSG - facilitates an increased inflow of water from the southwest Pacific Ocean into the southeast Indian Ocean. Similar to the increase in transport through the Tasman Leakage (black line in Fig. 4c), the decrease in flow along the Tasman Front is also due to GHG forcing (red line in supporting information Fig. S6a). The above results remain consistent across the CESM and multi-model ensembles (Supporting information Figs. S6b&c and S7).
Figure 6 examines the intermodal correlations between the SEC and WBC transport trends in different ocean basins. There is a strong positive correlation for the Indian and Atlantic Oceans, with correlation coefficients of 0.78 for the former and 0.73 for the latter, emphasizing that the strengthened southward WBCs in the Indian and Atlantic Ocean basins are closely related to the acceleration in the northern limbs of the subtropical gyres. However, there is a much weaker correlation (r = 0.15) for the Pacific Ocean and it does not pass the 95% test, which is consistent with the above analysis.
In summary, the increasing inter-basin water exchange in the connecting regions of the SHSG between 35°S-45°S is primarily driven by changes in transport at lower latitudes. The GHG forcing leads to a strengthening of the northern limbs of the subtropical gyres in the Indian and Atlantic Oceans and a weakening of the flow along the Tasman Front, supporting the strengthening of the WBCs as part of the SHSG. Since the 1950s, all three ocean basins have shown upward trends in the WBCs, consistent with increased transports through the Tasman and Agulhas Leakages, providing evidence for the spin-up of the SHSG.
3.5 Wind stress curl
The flows across the ocean basins are primarily driven by wind stress, which has been extensively studied (Huang 1991; Cai et al. 2005; Bostock et al. 2006; Ridgway and Dunn 2007; Yang et al. 2007b; Hill et al. 2008; Wang et al. 2014). According to Sverdrup balance, large-scale ocean circulation patterns are mainly determined by the spatial distribution of wind stress curl (Godfrey 1989). Increased positive wind stress curl in the subtropics of the Southern Hemisphere leads to an enhancement in Ekman pumping and a spin-up of the SHSG (Rintoul and England 2002; Yang et al. 2007a; Duan et al. 2016). The transports through the Tasman and Agulhas Leakages are also influenced by the wind (Hill et al. 2008; Sasaki et al. 2008; Biastoch et al. 2009; Rouault et al. 2009; Durgadoo et al. 2013; Oliver and Holbrook 2014; Tim et al. 2019). In the following, we will focus on changes in the wind stress curl to illustrate the impact of single forcing on the SHSG.
Figure 7 shows the climatological zonal average of the wind stress curl from 10°S to 60°S and its trends from 1955 to 2014 under the historical and single forcing simulations in the CanESM5 ensemble. Climatologically, the positive wind stress curl extends from 18°S to 50°S (cyan shading in Fig. 7) and the zero wind stress curl line is located at 50°S. Since the 1950s, there has been a significant positive anomaly in the wind stress curl between 36°S and 57°S (black line in Fig. 7), resulting in an increase in the positive wind stress curl over the southern SHSG and a southward migration of the zero wind stress curl line.
These changes in the wind stress curl are primarily attributed to the GHG forcing (red line in Fig. 7). The OD forcing also produces a positive anomaly in wind stress curl, but it appears to be smaller and further south and thus contributes less to the increase in the positive wind stress curl over the southern SHSG than the GHG forcing (yellow line in Fig. 7). In contrast, the wind stress curl changes induced by the AER forcing are much smaller (blue line in Fig. 7). The strength of the SHSG is more significantly impacted by the wind field within the positive wind stress curl region, rather than the entire westerly jet zone. Thus, the GHG forcing is the main driver for the spin-up of the SHSG. For the poleward shift of the SHSG, the southward migration of the zero wind stress curl line is driven by both GHG and OD forcings, with the GHG forcing showing a more pronounced trend of southward shift compared to the OD forcing (Supporting information Fig. S8). The result is consistent with previous studies (Cai 2006; Cai and Cowan 2007; Wang et al. 2014). The zero wind stress curl line is situated south of the southern boundary of the SHSG (i.e., the zero mass transport streamfunction line) and is more easily influenced by the OD forcing. In contrast, the zero mass transport streamfunction line is less responsive to the OD forcing than the zero wind stress curl line. Therefore, the poleward shift of the SHSG is found to be primarily influenced by the GHG forcing, with the OD forcing playing a secondary role. While both the GHG and OD forcings contribute almost equally to the southward migration of the westerlies, it is the former that plays a primary role in increasing the positive wind stress curl over the southern SHSG and thus driving the poleward shift and spin-up of the SHSG.
In the Pacific basin, the increasing transport of the Tasman Leakage from the EAC extension is attributed from the weakening of flow along the Tasman Front rather than the northern limbs of the SPG. The strengths of the flow along the Tasman Front and the EAC extension are anticorrelated, and both are consistently influenced by the wind stress curl over the South Pacific (Hill et al. 2008). Changes in the wind stress curl in the South Pacific (20–50°S, 180–280°E) drive these changes in the EAC system. According to the CanESM5 ensemble, the trend of wind stress curl in this region has increased over recent decades, primarily due to the influence of GHG forcing (Supporting information Fig. S9). This strengthening of basin-wide wind stress curl drives a southward expansion of the SPG (Sasaki et al. 2008; Hill et al. 2011). As the gyre shifts south, the EAC extension pathway receives more water at the expense of the Tasman Front, resulting in increased Tasman Leakage, as discussed in Section 3.4.
In conclusion, the poleward shift and spin-up of the SHSG can be attributed to the positive anomaly of the wind stress curl in the mid-latitudes of the Southern Hemisphere, driven primarily by the GHG forcing. The effect of OD on the winds is further south and thus plays a secondary role in the changes of the SHSG since the 1950s. The effect of AER on the changes in wind stress curl is much smaller.
3.6 Validation with the partially coupled experiments
To verify that changes in the SHSG are indeed driven by GHG-induced changes in wind stress, we employ the partially coupled experiments, in which the ocean circulation changes in response to CO2 quadrupling are decomposed into a dynamic component driven by wind stress and a thermodynamic component induced by buoyancy forcing. Comparing Figs. 8a with 3a and 8b with 7, it can be seen that the patterns of mean ocean circulation and wind stress curl changes simulated in the CESM CO2 quadrupling experiments are in good agreement with those in the CanESM5 ensemble. In particular, the CO2 quadrupling experiment simulates an increase in the positive wind stress curl over the southern SHSG region, except that the magnitude of this change is greater due to its higher CO2 concentration compared to the GHG-only simulation (black lines in Figs. 7 and 8b).
The difference between the two lines in Fig. 8a clearly shows the poleward shift of the southern boundary of the SHSG under the wind stress effect of CO2 quadrupling, which is similar to the change in the GHG-only simulation (Fig. 3). In addition, the wind stress effect of CO2 quadrupling leads to a significant increase in transport through both the Tasman Leakage (~ 5.52 Sv increase) and the Agulhas Leakage (~ 1.06 Sv increase). Furthermore, a comparison of the patterns between Figs. 8c, 8d, and 8f suggests that the circulation changes in the SHSG under CO2 quadrupling are primarily due to its wind stress effect, while its buoyancy forcing effect plays only a secondary role.
Therefore, the partially coupled experiments confirm that it is the GHG forcing that induces to a southward shift of the zero wind stress curl line and an increase in the positive wind stress curl over the southern SHSG region, which in turn causes a strengthening and poleward shift of the SHSG.