On the Response of Subduction in the South Pacific to an Intensification of Westerlies and Heat Flux in an Eddy Permitting Ocean Model

The South Pacific is a vital region for the transformation and ventilation of water masses (McCartney, 1977; Huang and Qiu, 1998; Donners et al., 2005; Sallèe et al., 2006; Liu and Wu, 2012; Cerove\vcki et al., 2013). In austral late winter, the Subantarctic Mode Water (SAMW) and Antarctic Intermediate Water (AAIW) are formed by subduction from the water masses in the mixed layer (Walin, 1982; Woods and Barkmann, 1986; Marshall et al., 1999; Kwon et al., 2013). Thereafter, SAMW and AAIW spread eastwards and northwards in the Southern Ocean, transporting dissolved gases, nutrients, fresh water and heat into the interior of the global ocean (Mikaloff Fletcher et al., 2006; Le Quèrè et al., 2007). SAMW and AAIW are also recognized as major components of the upper cell of the global overturning circulation, balancing the southward flow of the North Atlantic Deep Water (Mazloff et al., 2013).

Extending from the surface to a depth of about 600 m in the formation region, the SAMW is defined as a homogeneous layer with a minimum in potential vorticity (McCartney and Talley, 1982). The properties of SAMW are closely associated with the fluctuant Subantarctic Front (SAF) (Rintoul and England, 2002). For example, the warmest SAMW is located in the western South Atlantic where the SAF is farthest north (Provost et al., 1999), and the coldest SAMW is to the west of Drake Passage where the SAF is farthest south (Talley, 1996). In contrast to SAMW, AAIW fills the layer at depths of 800-1000 m in the subtropical and tropical oceans, with only minor changes of its properties (Georgi, 1979; Piola and Georgi, 1982; Wong, 2005; Iudicone et al., 2007). Previous studies have already proved that the formation and erosion of SAMW and AAIW are closely linked to interactions at the air-sea surface (Sloyan and Rintoul, 2001, Banks et al., 2002; Goes et al., 2008; Saenko et al., 2011), suggesting that changes in atmospheric forcing, such as the strength of westerlies and associated heat flux, could strongly impact the production and characteristics of SAMW and AAIW in the South Pacific.

Since 2001, the global average surface air temperature has remained relatively steady——a period commonly known as the global warming hiatus (Trenberth and Fasullo, 2013). In the slowing down of global warming, the subduction of SAMW and AAIW may play a positive role. Based on results from the Southern Ocean State Estimate project, (Mikaloff Fletcher et al., 2006) suggested a close relationship between the uptake of anthropogenic CO₂ and the formation of SAMW. Based on observational estimates, (Sallèe et al., 2012) also revealed a mechanism responsible for carbon sequestration in the Southern Ocean, and proved that the subduction of SAMW and AAIW is closely associated with the absorption of anthropogenic CO₂. Previous studies highlight the potential of SAMW and AAIW for slowing down the growth of atmospheric CO₂ (Le Quèrè et al., 2007). Therefore, there is a need to investigate the response of the production and characteristics of the SAMW and AAIW to different climate scenarios.

Although previous studies have been limited by the coarse spatial resolution of climate models and the poor spatial and temporal coverage of observations, the critical roles of SAMW and AAIW have nonetheless been certified (Sloyan and Kamenkovich, 2007; Downes et al., 2009; Downes et al., 2010; Liu and Wang, 2014). In recent years, scientists have focused on finer resolution model studies of the Southern Ocean. For example, utilizing an eddy permitting coupled climate model (0.25^°× 0.25^°), the impact of wind stress on the subduction of SAMW and AAIW was analyzed by (Downes et al., 2011). To access the sensitivity of the SAMW to horizontal resolution, relatively higher resolution simulations have been carried out in the Southern Indian Ocean (Spence et al., 2009; Lee et al., 2011; Lee and Nurser, 2012). However, the response of SAMW and AAIW to changes in westerlies and associated heat flux in the South Pacific has not been quantitatively assessed with a high resolution model. In this study, we aim to understand the response of SAMW and AAIW to stronger westerlies and stronger heat flux in the South Pacific, with the aid of an eddy permitting resolution model.

This paper is organized as follows. In section 2, we describe the model setup. In section 3, a detailed examination of the mixed layer depth (MLD) and eddy kinetic energy (EKE) is presented. The response of the properties and subduction of SAMW and AAIW is analyzed in section 4. In section 5, we diagnose the mechanism responsible for changes in the MLD through analysis of the buoyancy budget. Section 6 summarizes the results with discussion.

The model employed in this study is the Massachusetts Institute of Technology General Circulation Model (MITgcm), version 6.1. Based on primitive equations, this model applies Boussinesq approximation and the hydrostatic assumption. Our experiment is placed on a cube-sphere grid projection, which has a relatively even grid spacing throughout the global ocean without polar singularities. The model has a horizontal resolution of about 18 km to partly resolve mesoscale eddies. In the vertical coordinate, there are 50 levels, with intervals ranging from 10 m near the surface to 450 m at the bottom. There are no parameterizations of eddy induced transport. The MITgcm sea ice model is coupled to the ocean model (Losch et al., 2010), and the momentum equations are solved through the parallel implementation solver (Zhang and Hibler, 1997).

The products of the 40-yr European Centre for Medium-Range Weather Forecast Re-Analysis (ERA-40) are used as the atmospheric boundary condition, including downward longwave and shortwave radiation, humidity, 2-m air temperature, 10-m surface wind, and river runoff. The forcing data are called every six hours at a spatial resolution of 1.125^°× 1.125^°. The initial conditions and bathymetry are provided by a blend of the Polar Hydrographic Climatology, the World Ocean Circulation Experiment Global Hydrographic Climatology, and the Estimating the Circulation and Climate of the Ocean phase II: high resolution global-ocean and sea-ice data synthesis (ECCO2). In order to obtain good approximations of the initial conditions, the ECCO2 state estimate is improved in the optimized values of uncertain parameters (Menemenlis et al., 2008).

The changes in the westerlies and associated heat flux under different climate scenarios in the Southern Ocean have already been clearly identified (Russell et al., 2006; Yang et al., 2007). The different thermal gradients of heat from the pole to midlatitudes may result in different strengths of westerlies, e.g., local Antarctic cooling and a global warming could give rise to stronger zonal winds around Antarctica. In order to investigate the response of SAMW and AAIW to two representative climate scenarios, we integrate the model with two annual repeating atmospheric forcings according to the Southern Annular Mode (SAM). The SAM is an index of the atmospheric oscillation that results from changes in the westerlies and other air-sea surface fluxes over the Southern Ocean. A positive phase of the SAM is associated with a strengthening and poleward shift of the westerlies, and vice versa (Thompson and Wallace, 2000). According to the SAM's historical records, the forcing in the years 1992 (typical negative SAM) and 1998 (typical positive SAM) are introduced into our experiments. The 1992 radiative forcing, with weaker westerlies and heat loss from the ocean to the atmosphere, is applied to one of the simulations, named "92R". Figure 1a depicts the spatial distribution of zonal wind stress in 92R derived from ECCO2, with strong prevailing westerly winds between 40^°S and 60^°S. The 1998 radiative forcing, with stronger westerlies and heat loss, is applied in the other simulation, named "98R". Figure 1b shows that a clearly significant intensification of the westerlies occurs in the Antarctic Circumpolar Current (ACC) region, with a maximum greater than 0.1 N m^-2. Since the properties of SAMW and AAIW are primarily determined in late winter, the heat flux at the air-sea interface in late winter is presented in Fig. 1c. Basin-scale heat loss to the atmosphere is identifiable in 92R in late winter (Fig. 1c). In particular, the differences in the heat flux between the two simulations (Fig. 1d) coincide closely with the differences in subduction. The regions with more heat loss overlap with the regions of increased subduction (see below). In this study, late winter denotes September. Each of the two simulations is integrated for 28 years and saved with five-day sampling. We use the averaged outputs for the 24-28 years in this study. The extent of the model domain covers the global ocean, and our analysis focuses on the South Pacific.

Figure 1.  (a) Zonal wind stress (units: N m^-2) diagnosed from 92R. (b) Differences in zonal wind stress between the two simulations (98R minus 92R) (eastwards is positive). (c) Net heat flux (units: W m^-2) diagnosed from 92R (downwards is positive). (d) Differences in net heat flux between the two simulations (98R minus 92R).

The topography superposed with the Subtropical Front (STF), SAF and Polar Front (PF) derived from 92R is presented in Fig. 2a. The formation area of SAMW is bounded by the STF and SAF and concentrated closer to the SAF, and the formation area of AAIW is bounded by the SAF and PF. In this study, the positions of the STF, SAF and PF follow the definitions of (Orsi et al., 1995). The STF is defined as the latitudes of the 10^°C isotherm at a depth of 100 m; the SAF is defined as the latitudes of the 4^°C isotherm at a depth of 400 m; and the PF is defined as the latitudes of the 2.2^°C isotherm at a depth of 800 m. The potential vorticity, which is calculated by the same method as employed in (Liu and Wu, 2012), clearly shows the vertical distribution of SAMW across 135^°E, with a typical value less than 1× 10^-10 m^-1 s^-1 (Fig. 2b). Compared with water masses at the same depth, the AAIW to the south of the SAF is characterized by relative lower salinity (Fig. 2b).

Figure 2.  (a) Topography of the Southern Pacific (units: m), with the STF, SAF and PF from 92R superimposed (red solid lines from north to south) and the arrows denoting the formation regions of the SAMW (bounded by the STF and the SAF) and the AAIW (bounded by the SAF and the PF). (b) Annual mean potential vorticity (color shading) across the section of 135^°E in 92R (units: 10^-10 m^-1 s^-1), in which the black lines denote the annual mean isohaline with a contour interval at 0.1 psu, and the position of the STF, SAF and PF are shown as green lines. Note that the SAMW and AAIW are characterized by lower potential vorticity and lower salinity, respectively.

Figure 3.  (a) MLD in austral late winter from WOA09 (units: m). (b) Last 5-year mean MLD in late winter from 92R (units: m) and corresponding fronts (red lines denoting the STF, SAF and PF from north to south). (c) As in (b), but for 98R. (d) Differences between the two simulations (98R minus 92R) (units: m), in which the STF, SAF and PF in 92R and 98R are superimposed with grey solid lines and dark dashed lines, respectively, from north to south.

Figure 4.  (a) EKE from 92R (units: m² s^-2). (b) Differences between the two simulations (98R minus 92R).

Since the late winter MLD contributes greatly to the formation of SAMW and AAIW, we use the World Ocean Atlas 2009 (WOA09) dataset to evaluate the simulated MLD in MITgcm. The MLD is defined as the depth at which the potential density is 0.03 kg m^-3 denser than that at the surface (Downes et al., 2009). Figure 3a depicts the MLD distribution derived from WOA09. Two zonal northwest-southeast slanting bands of deep MLD are apparent in the South Pacific. The first is located to the south of Australia (40^°-55^°S), extending from 120^°E to 170^°E, with a typical value of about 350 m. The second is located to the southeast of New Zealand (45^°-60^°S), extending from 170^°W to 70^°W, with a maximum greater than 500 m. Compared to the MLD derived from WOA09, the model reproduces the main features of the MLD in 92R (Fig. 3b), but with a significant increase in 98R (Fig. 3c). The simulated MLD in the ACC region is a little noisier than that in WOA09, and this may result from the relatively higher spatial resolution in MITgcm than the 0.5^° spatial average in WOA09. Figure 3d shows the differences in MLD between 92R and 98R. In contrast to 92R, the MLD in 98R deepens substantially between the STF and the PF, from the east of New Zealand to Drake Passage, with maximums of about 500 m at (45^°-50^°S, 200^°-100^°W). Some previous studies have suggested that the SAF and PF would shift polewards with intensification of the westerlies. In our simulation, the poleward shift of the major fronts in 98R is inconspicuous in the eastern South Pacific, and this might be due to the influence of the topography (Graham et al., 2012). Just as (Graham et al., 2012) suggested, the fronts with the ACC are nearly barotropic, and thus they are more dependent on the changes in the underlying bathymetry rather than the westerlies. In the western South Pacific, the major fronts diverge quite a bit in 98R from those in 92R.

In order to diagnose the eddy activity, we computed the EKE as follows: \begin{equation} \label{eq1} {\rm EKE}=\frac{1}{2}({u}'^2+{v}'^2) , (1)\end{equation} where u' and v' are the eddy turbulent components of the horizontal velocities. The surface EKE calculated from 92R is presented in Fig. 4a. The spatial distribution of EKE indicates that the most intense eddy activities are located along the path of the ACC, which compares favorably with observations from altimeter data (not shown). The East Australia Boundary Current also spawns eddies due to the baroclinic instability (Nilsson and Cresswell, 1980). The differences of EKE between the two simulations are shown in Fig. 4b. The areas with significant change in EKE are detected in both the path of the ACC and downstream of the East Australia Boundary Current, with maximums greater than 0.08 m² s^-2. The intensification of EKE in 98R is likely due to barotropic and baroclinic instabilities induced by more momentum flux exported into the ocean interior from stronger westerlies.

Because of the short effective period of subduction in late winter (Huang and Qiu, 1994; Yu et al., 2015), the properties of upper-limb water masses are synonymous with the bulk SAMW and AAIW. We assess the sea surface temperature (SST) and sea surface salinity (SSS) in austral late winter. The simulated SST and SSS in 92R (Figs. 5a and b) agree reasonably well with the observed fields derived from WOA09 (not shown). There are two significant cooling centers in Fig. 5c. The first is located around (55^°S, 150^°W), and the second is located in the central subtropical gyre around (40^°S, 240^°), with minimums less than -2^°C. In contrast to the differences in SST, a basin-scale saltier pattern appears at high latitudes, except a slightly fresher band extending from (65^°S, 160^°E) to (50^°S, 130^°W) (Fig. 5d).

As mentioned above, the formation area of SAMW in the South Pacific is bounded by the STF and the SAF and concentrated closer to the SAF, and the formation area of AAIW is bounded by the SAF and PF. Thus, we average the SST, SSS and potential density over all of the surface cells bounded by the STF and SAF to diagnose the characteristics of SAMW. The same method is applied to the surface cells bounded by the SAF and the PF to diagnose the properties of AAIW. The average temperature and salinity of SAMW is about 7.58^°C and 34.26 psu in 98R——saltier and warmer than in 92R (7.04^°C and 34.05 psu). Its average density is 26.71 kg m^-3 in 98R, which is 0.11 kg m^-3 denser than that in 92R (Table 1). The average temperature and salinity of AAIW, which is characterized by relatively lower salinity, is about 2.03^°C and 33.88 psu in 98R——saltier and warmer than in 92R (1.40^°C and 33.66 psu). The average density of AAIW in the South Pacific has a typical value of 27.06 kg m^-3 in 98R, which is 0.12 kg m^-3 denser than that in 92R (Table 1).

The widely used kinematic approach for diagnosing subduction (Marshall et al., 1993) is introduced in our study to quantify the formation rate of SAMW and AAIW. From this kinematic perspective, by following a particle released at the base of the mixed layer, the Lagrangian subduction rate can be calculated throughout the year (Huang and Qiu, 1994; Xie et al., 2011). The annual subduction rate S _total is defined as follows: \begin{equation} \label{eq2} S_{\rm total}=-\frac{1}{T}\int_{t_1}^{t_2}w_{\rm mb}dt+\frac{1}{T}(h_{\rm m,t_1}-h_{\rm m,t_2}) ,(2) \end{equation} where h _m denotes the thickness of the mixed layer, and w _mb denotes the vertical velocity of the water parcel at the base of the mixed layer (positive meaning downwards). By integrating from the end of the first winter t₁ to the end of the second winter t₂ over one year T, the S _total can be calculated. On the right-hand side in Eq. (3), the first term represents the contribution of vertical pumping at the base of the mixed layer. The second term, which is usually named as lateral induction, denotes the contribution from the slope of the mixed layer base. Based on this Lagrangian subduction rate, we estimate the effective detrainment of the SAMW and AAIW. It means that the mixed layer waters are irreversibly transferred into the main pycnocline, without considering the seasonal cycle of the MLD.

Figure 5.  (a) Five-year mean SST from 92R (units: ^°C). (b) As in (a) but for SSS (units: psu). (c) Differences in SST between the two simulations (98R minus 92R) (units: ^°C). (d) As in (c) but for SSS (units: psu).

Figure 6.  (a) Five-year mean subduction rate from 92R (units: m yr^-1). (b) Differences between the two simulations (98R minus 92R). The potential density is superimposed as black solid lines with a contour interval of 0.5 kg m^-3. The thick black solid lines denote the coastal lines, and figures in below are the same.

The subduction of the SAMW and AAIW mostly occurs in the central and east South Pacific, bounded by the isopycnals of 26.5 kg m^-3 and 27.0 kg m^-3, with a maximum greater than 200 m yr^-1 around Drake Passage (Fig. 6a). Compared to the subduction in 92R, a significant intensification is detected to the north of the 27.0 kg m^-3 isopycnal in 98R (Fig. 6b), extending from 170^°W to 80^°W and with maximums greater than 200 m yr^-1. The surface isopycnals in 98R shift somewhat equatorwards, whereas the areas with higher subduction rate are still bounded by isopycnals of 26.5 kg m^-3 and 27.0 kg m^-3. As expected, the changes in the spatial pattern of the subduction are very consistent with the changes in the MLD (Fig. 3d).

Figure 7.  Subduction volume (Sv) in the South Pacific for each 0.1 kg m^-3 interval: (a) total subduction volume; (b) contribution of the lateral induction term; (c) contribution of the vertical pumping term.

By averaging over 60 predetermined potential density classes for five years, the net subduction volume and the correlated terms can be obtained (Fig. 7). The integration is applied to each cell bounded by the isopycnals in the entire South Pacific (30^°-70^°S, 120^°-60^°W), with predetermined potential density classes ranging from 23 kg m^-3 to 28 \(kg\;m^-3\). The subduction in 92R apparently occupies a large amount of water mass with a density range of 24.7-27.3 kg m^-3. Compared to 92R, the most active subduction in 98R is distributed at the density of 26.9 kg m^-3, with only one maximum greater than 40 Sv (1 Sv=1× 10⁶ m³ s^-1). As suggested by (Liu and Wu, 2012), the lateral induction term still plays a dominant role in the formation of the SAMW and AAIW in the South Pacific (Fig. 7b). We also find that the vertical pumping term contributes more than the lateral induction term at fresher density classes (25.0-26.0 kg m^-3), implying a different mechanism responsible for the subduction in the subtropical area (Fig. 7c). The subduction volume of SAMW is calculated by integration of the subduction rate over cells bounded by the STF and SAF, and the subduction volume of AAIW is obtained by integration of the subduction rate over cells bounded by the SAF and PF. The total subduction volume of the SAMW and AAIW increases from 20.5 Sv and 3.7 Sv to 37.1 Sv and 6.9 Sv, respectively (Table 1). It is worth noting that the increase in subduction for 98R is a little larger. This might result from the repeating atmospheric forcing with larger differences between 92R and 98R to deliberately enlarge the responses of subduction.

Figure 8.  Buoyancy flux derived from 92R (units: 10^-5 N m^-2 s^-1): (a) total buoyancy flux; (b) heat flux term; (c) freshwater flux term; (d) Ekman drift term (downwards is positive). Note the different scale in the color bars for (c) and (d). The potential density in September from 92R is superimposed as black solid lines with a contour interval of 0.5 kg m^-3.

Figure 9.  Differences in buoyancy flux (98R minus 92R) (units: 10^-5 N m^-2 s^-1): (a) total buoyancy flux; (b) heat flux term (the STF, SAF and PF in 92R and in 98R are superimposed using grey solid lines and dark dashed lines, respectively, from north to south); (c) freshwater flux term; (d) Ekman drift term (downwards is positive). Green lines in (a) denote the shoaling of the MLD with a contour interval of 100 m. Grey lines in (a) denote the deepening of the MLD with a contour interval of 200 m. Note the different scale in the color bars for(c)\,and(d).

Figure 10.  Winter buoyancy flux in potential density classes derived from 92R (blue lines) and 98R (green lines) (units: 10⁸ N s^-1) for the total buoyancy flux (stars), the heat flux term (squares), the freshwater flux term (circles), and the Ekman drift term (diamonds) (downwards is positive).

Previous studies have suggested that the stratification of the upper ocean in the South Pacific is primarily influenced by the buoyancy flux (Downes et al., 2009; Downes et al., 2010). Since the formation rate and properties of the SAMW and AAIW are closely related with the MLD in late winter, we further analyze the buoyancy flux, which plays a dominant role in the deepening of the MLD in late winter. The method used in this study is the same as employed in (Liu and Wang, 2014): \begin{equation} \label{eq3} B_{\rm net}=\frac{g\alpha}{C_w}H+g\beta SW+\frac{g}{\rho f}{k}{\tau}\cdot\nabla\rho , (3)\end{equation} where a positive B _net value indicates buoyancy gain (i.e., decreased density of surface water) and a negative value indicates buoyancy loss; g denotes the gravitational force; and ρ denotes the sea water density at the surface. The first term on the right-hand side in Eq. (4) is the buoyancy flux induced by the surface heat flux H. C_w denotes the heat capacity, and α denotes the thermal expansion coefficient. The second term represents the buoyancy induced by the freshwater flux W. S denotes the salinity in the mixed layer, and β denotes the haline contraction coefficient. The third term is induced by Ekman drift, where k is the unit vertical vector, τ is the wind stress, and f is the Coriolis parameter.

Since the seasonal cycle of MLD is determined by the seasonality of buoyancy flux, we diagnose each term in Eq. (4) in September. Figure 8 depicts the buoyancy flux derived from 92R in late winter, and a strong basin-scale buoyancy loss is shown in Fig. 8a. The minimum buoyancy loss occurs around the Antarctic, and this may result from the isolation of the sea surface from the atmosphere by the cover of sea ice. Negative cores associated with the deep MLD are found along the path of the ACC, implying an association between the MLD and the buoyancy flux. The surface heat flux term also displays a basin-scale heat loss in the South Pacific (Fig. 8b), and bears a close resemblance to the total buoyancy flux in magnitude and spatial pattern. The term induced by freshwater flux tends to make the surface water lighter in the ACC region, with a typical value less than 2.5× 10^-5 N m^-2 s^-1 (Fig. 8c). The term associated with Ekman drift contributes much less than the other terms (Fig. 8d), which is in agreement with previous studies (Downes et al., 2010). As a result, the term induced by surface heat flux plays a dominant role in the stratification of the Southern Ocean in wintertime.

Figure 9a depicts the changes in the buoyancy flux in our simulations. The spatial pattern of the response of the MLD compares very well with the changes in the buoyancy flux (Fig. 9a), especially in the central South Pacific. Essentially, the changes in the heat flux term also resemble the changes in the total buoyancy flux (Fig. 9b). The terms induced by both the freshwater flux (Fig. 9c) and the Ekman drift (Fig. 9d) contribute much less to the change in buoyancy flux. The implication, therefore, is that the change in the surface heat flux should still be a dominant term in adjusting the change of the MLD. The heat flux term shows a zonally asymmetric response in different SAM events. This is consistent with previous studies that highlighted the intrusion of low pressure in the eastern South Pacific (e.g., Sallèe et al., 2010). The high spatial correspondence between the changes in the MLD and buoyancy flux indicates that a deepening of the MLD is associated with intensified buoyancy flux loss due to the intensification of heat loss, and vice versa.

As noted previously, the properties and subduction of SAMW and AAIW are typically associated with wintertime deep convection, and thus also to the spatial distribution of the MLD (Iudicone et al., 2007). As such, they are also closely consistent with the change in buoyancy flux. Similar to the analysis of subduction volume reported in section 4, Fig. 10 shows the accumulative buoyancy flux and its terms in different potential density classes in the two simulations. The integration is also applied to each surface cell bounded by the isopycnals in the entire South Pacific (30^°-70^°S, 120^°-60^°W), with predetermined potential density classes ranging from 23 kg m^-3 to 28 kg m^-3. The distribution of buoyancy flux exhibits a strikingly similar pattern to the subduction in different potential density classes (Fig. 10). The strongest buoyancy loss in 98R occurs at the density of 27.0 kg m^-3, with a minimum of about -4.5× 10⁸ N s^-1; whereas, the buoyancy flux in 92R peaks at 26.9 kg m^-3, which is 0.8× 10⁸ N s^-1 larger than that in the 98R. Thus, the wintertime convection can be enhanced and result in a more active subduction in 98R.

In this study, the response of the SAMW and AAIW to two distinct atmospheric forcings in the South Pacific is quantitatively estimated with the aid of MITgcm. Forced by two different atmospheric conditions, significant differences are detected in the properties and subduction of the SAMW and AAIW. Based on observations, a previous study revealed that, since around 2001, the warming of global average surface air temperature has been interrupted by a marked hiatus (Guemas et al., 2013). Several mechanisms have been proposed to explain this recent phenomenon, including increased ocean heat uptake through ventilation at higher latitudes (Meehl et al., 2013). As the most active area of ventilation, the South Pacific may play an important role in this slowing down of global warming, due to effective CO₂ uptake through subduction (Sallèe et al., 2012). In early studies, the austral westerlies were found to intensify under global warming (Cai et al., 2005; Russell et al., 2006). Thus, our study focuses on quantifying the response of SAMW and AAIW to different westerlies and associated heat flux in an eddy permitting ocean general circulation model——the aim being to preliminarily understand the response and potential effect of the subduction to different westerlies and associated heat flux at high latitudes. Through two representative experiments with significantly distinct atmospheric forcings, we find that the subduction of the SAMW and AAIW substantially increases in a typical positive SAM year, which is characterized by stronger westerlies and heat loss from the ocean to the atmosphere. Thus, it implies that the South Pacific has the potential for slowing down global warming by sequestering more CO₂ through intensified subduction. The lateral induction processes play a dominant role in the subduction, due to the deepening of the MLD in late winter. Based on an analysis of the buoyancy flux budget, the heat flux term is found to contribute most to the buoyancy loss in late winter, which results in the deepening of the MLD. Thus, the potential role of the subduction of the SAMW and AAIW to the heat uptake and slowing down of global warming may be preliminarily understood. Furthermore, two cooling centers of SST in 98R are separated by a narrow warmer band (Fig. 5c), extending from (40^°S, 180^°E) to (55^°S, 120^°W). This warmer band coincides with the intensified EKE in that region. Previous studies have suggested that enhanced eddy activity can hinder the formation of a deep MLD in the North Pacific (Qiu et al., 2007; Xu et al., 2014; Xu et al., 2016). However, the relative role of mesoscale processes in the formation of SAMW and AAIW in the South Pacific are still not fully understood. We intend to address this knowledge gap in future work.