Subduction, water mass transformation, and transport rates in the Indo-Pacific Ocean are diagnosed in a recent version of the Canadian Centre for Climate Modelling and Analysis coupled model. It is found that the subduction across the base of the winter mixed layer is dominated by the lateral transfer, particularly within the relatively dense water classes corresponding to the densest mode and intermediate waters. However, within lighter densities, including those characterizing the lighter varieties of mode waters, the vertical transfer has a strong positive input to the net subduction. The upper-ocean volume transports across 30°N and 32°S are largest within the density classes that correspond to mode waters. In the North Pacific, the buoyancy flux converts the near-surface waters mostly to denser water classes, whereas in the Southern Ocean the surface waters are transformed both to lighter and denser water classes, depending on the density. In response to a doubling of CO2, the subduction, transformation, and transport of mode waters in both hemispheres shift to lighter densities but do not change significantly, whereas the subduction of intermediate waters decreases. The area of large winter mixed layer depths decreases, particularly in the Southern Hemisphere. In the low latitudes, the thermocline water flux that enters the tropical Pacific via the western boundary flows generally increases. However, its anomaly has a complex structure, so that integrated estimates can be sensitive to the isopycnal ranges. The upper part of the Equatorial Undercurrent (EUC) strengthens in the warmer climate, whereas its lower part weakens. The anomaly in the EUC closely follows the anomaly in stratification along the equator. The Indonesian Throughflow transport decreases with part of it being redirected eastward. This part joins with the intensified equatorward thermocline flows at the western boundaries and contributes to the EUC anomaly.
There is a growing interest in the behavior of water mass formation and subduction under global warming. By transferring fluid from the mixed layer into the subsurface ocean (Stommel 1979; Woods 1985), subduction effectively connects atmosphere to oceanic interior. Annual rates of subduction have been estimated for different parts of the ocean using observational, mainly climatological data (e.g., Marshall et al. 1993; Qiu and Huang 1995; Karstensen and Quadfasel 2002). There are also several studies that have addressed the impact of climate change on the subduction rates using different climate models and focusing mostly on the Southern Ocean regions. In particular, Goes et al. (2008) analyzed the rates of subduction in the South Atlantic using the Community Climate System Model, version 3; Banks et al. (2002) employed the Third Hadley Centre Coupled Ocean–Atmosphere General Circulation Model to examine the rates of mode water formation in the Indian Ocean; Downes et al. (2009) diagnosed changes in the Southern Ocean subduction rates using the Commonwealth Scientific and Industrial Research Organisation Mark version 3.5. One common conclusion from these studies is that as the climate becomes warmer, the subduction shifts to lighter densities, with its rates being essentially unaffected. The only significant change is reported by Downes et al. (2009), who project a decrease in the subduction of intermediate waters. One reason to think that there could be changes in the Southern Ocean subduction in a warmer climate is because climate models predict a shoaling of the mixed layer and a strengthening of wind stress (e.g., Sen Gupta et al. 2009). However, the overall effect of these and other changes on the Southern Ocean subduction is not easy to foresee. In the North Pacific, Luo et al. (2009a) concluded, by analyzing changes in potential vorticity and surface fluxes, that in a warmer climate, mode waters are produced on lighter surfaces and significantly weaken in terms of their formation and evolution (although they did not present estimates for the rates of subduction). Given the importance of subduction for transferring the upper-ocean properties into the subsurface interior, including those associated with anthropogenic effects, one of our purposes here is to evaluate its changes in a 2 × CO2 climate, focusing on the Indo-Pacific Ocean.
Changes in the extratropical subduction and water mass formation have the potential to affect tropical oceans. Therefore, our second purpose is to examine the changes in the tropical Pacific dynamics. Using observations, McPhaden and Zhang (2002) show that the shallow meridional overturning circulation in the Pacific, that connects the tropical and extratropical oceans, has been slowing down since the 1970s. Climate models seem to suggest that, as the climate warms, the total transport from the subtropics to the tropics in the Pacific Ocean will not change significantly, although its partitioning between the western boundary pathway and that through the interior ocean does change (Luo et al. 2009b). In addition, climate models project changes in the equatorial current system, including a shallowed (Wang et al. 2008) and a significantly enhanced Equatorial Undercurrent (EUC) (Luo et al. 2009b). It is, however, less clear what causes these changes.
In the next section, we briefly describe the model we use and the design of model experiments. In section 3, the methods to calculate annual subduction, water mass transformation, and transport are presented. Main results are described in section 4, where in the extratropics we mainly focus on mode waters [for a recent review, see Hanawa and Talley (2001)]. In the tropics, the focus is on the pathways (boundary versus interior) for the mass exchange with the subtropical oceans and on the role of the Indonesian Throughflow (ITF). This is followed by a discussion and conclusions in section 5.
2. The model and experimental design
We employ a developmental version of the Canadian Centre for Climate Modelling and Analysis coupled climate model, the initial version of which is described in Flato et al. (2000). The oceanic component of the model is a version of the Geophysical Fluid Dynamics Laboratory Modular Ocean Model. Here, the horizontal resolution is 1.41° × 0.94° (longitude × latitude) and there are 40 vertical levels, with 10 levels in the upper 100 m. The ocean model employs anisotropic viscosity (Large et al. 2001) and eddy transport (Gent and McWilliams 1990) parameterizations. Vertical mixing driven by buoyancy and shear is parameterized by the K-profile parameterization (KPP) scheme (Large et al. 1994), where both vertical viscosity and diffusivity depend on the gradient Richardson number. The model also accounts for the effect of the dissipation of internal tides on deep ocean vertical mixing, implemented in a very similar fashion to that described in Simmons et al. (2004).
The atmospheric component is a general circulation spectral model (Scinocca et al. 2008). Here, it has T63 resolution (≈2.8°) and 31 vertical levels. Sea ice thermodynamics are governed by an energy balance model, whereas sea ice dynamics use a cavitating-fluid rheology (Flato et al. 2000). The coupled model is run without flux adjustments. It is capable of reproducing many features of the observed climate. Some aspects of the model evaluation are given in Saenko and Lee (2010), where they illustrate (their Fig. 2) that the simulated heat transports closely follow their observational estimates, within the climate system as a whole as well as separately in the ocean and atmosphere. In particular, the total (i.e., atmosphere plus ocean) heat transport in the model is close to that obtained from the Earth Radiation Budget Experiment data (Trenberth and Caron 2001); the heat transports in the model atmosphere and ocean agree well with those obtained based on the atmospheric reanalysis products and recent hydrographic sections presented in Trenberth and Caron (2001) and Ganachaud and Wunsch (2003), respectively. It is also illustrated in Saenko and Lee (2010, their Fig. 1) that the employed model simulates reasonably well the observed position of sea ice in both hemispheres. In addition, an older version of the model was employed by Saenko et al. (2005), where the simulated wind stress and its curl are compared to their observational estimates. In sections 3 and 4, we address the model’s ability to reproduce the observed structure of potential vorticity in the ocean, as well as the interior volume transports within the subtropical gyres.
We analyze climate states in two model runs. The first run is the control run, aimed at modeling the present-day climate, where the atmospheric CO2 was set to 348 ppm (hereafter 1 × CO2). It was run for 400 yr, starting from the climatological distribution for ocean potential temperature and salinity and from the atmospheric state obtained via a preliminary spinup of the model’s atmospheric component using observed sea surface temperature (SST) and sea ice. The second run is an idealized climate change experiment, wherein the atmospheric CO2 was abruptly doubled, that is, set to 696 ppm. This value is close to that in the A1B scenario [from the Intergovernmental Panel on Climate Change (IPCC)’s Special Report on Emissions Scenarios (SRES)], adopted for model runs in a stabilization mode (with fixed CO2). Our climate change experiment (2 × CO2) branched from the control at year 300, and then continued for another 100 yr, keeping the CO2 level fixed. This way, the upper-ocean thermal structure came close to equilibrium with the 2 × CO2 forcing, so that the mean position of the upper-ocean isopycnals does not change significantly, as required by the methods chosen to analyze the results. The evolution of the global-mean SST in both runs for the last 100 yr is shown in Fig. 1. In what follows, the results are based on the mean monthly fields obtained by averaging over the last 20 yr of the corresponding runs.
Here we define several diagnostics that are employed in section 4 to quantify the simulated changes in the ocean ventilation in response to the applied radiative forcing. In addition, we will consider maps of potential vorticity (PV) on isopycnals. These characterize the pathways of fluid in the limit of weak mixing and friction. Maps of PV are also useful to identify mode waters, which usually occur within or near the top of the permanent pycnocline and are characterized by layers of low PV (e.g., Hanawa and Talley 2001). With the squared buoyancy frequency being N 2 = −(g/ρθ)∂σθ/∂z, the large-scale PV can be written as
where g is the acceleration due to gravity and f is the planetary vorticity. The use of σθ (=ρθ − 1000) is justified since, for the most part, we will deal with the processes in the upper kilometer of the ocean.
Sections of simulated and observed |PV| are shown in Fig. 2. Essentially, the regions of low |PV| in the tropics and midlatitudes are due to PV dependence on f and N 2, respectively. Overall, the model does a good job in capturing the |PV| minimum associated with the penetration of mode waters from the midlatitudes toward the subtropics, in both the Pacific and Indian Oceans. In particular, the simulated position of the lowest |PV| values in the extratropics is at about the same depths and latitudes as those observed, with the simulated minimum |PV| being somewhat smaller. In the Southern Hemisphere, the observed low |PV| values of mode waters penetrate to the deeper ocean.
A transfer of fluid from the mixed layer into the oceanic interior is represented by subduction. Of particular interest is the subduction that takes place during a relatively short time window, from late winter to early spring, since the fluid subducted during this period penetrates into the main thermocline (Stommel 1979; Marshall et al. 1993; Karstensen and Quadfasel 2002). Thus, as suggested by Stommel (1979), a good approximation to the annual subduction into the main thermocline is provided by a volume flux across the base of the winter mixed layer. The annual subduction Sann is estimated locally as (e.g., Marshall et al. 1993; Karstensen and Quadfasel 2002) by
where uh and wh are the horizontal and vertical velocities, respectively, at the base of the winter mixed layer z = −h. Following Luo et al. (2009a), the winter mixed layer depth (MLD) is defined as the depth at which σθ differs by 0.1 from the surface value, using March (for the Northern Hemisphere) or September (for the Southern Hemisphere) densities. A similar criterion is typically used when diagnosing the rates of subduction based on observational data. For example, the corresponding values used in Karstensen and Quadfasel (2002) for the Southern Ocean are 0.075, 0.125, and 0.2, whereas the value used by Marshall et al. (1993) for the North Atlantic is 0.125. The total subduction (e.g., Goes et al. 2008) is then an integral of the local positive subduction over the area defined by the adjacent isopycnals σ − Δσ/2 and σ + Δσ/2, where Δσ = 0.1.
When diagnosing subduction rates from observational data, the velocity vectors for the lateral transfer (−uh · ∇h) are typically obtained using the thermal wind equations along with an assumption about the level of no motion (e.g., Marshall et al. 1993). In turn, the vertical transfer (−wh) is estimated from the linear vorticity balance. Here, we employ a general circulation model, so that both uh and wh are readily available. In general, these include contributions due to the eddy-induced transport of Gent and McWilliams (1990), which can be important in regions with large slopes of isopycnals.
b. Water mass transformation and transport
where the delta function is used to sample the buoyancy flux only for surface water of density σ and B incorporates the net heat and freshwater fluxes to the ocean, with positive B indicating buoyancy gain. In practice, this computation is done using discrete density bins, set here to Δσ = 0.1, so that F is scaled accordingly (e.g., Karstensen and Quadfasel 2002). As pointed out by Marshall et al. (1993), the transformation due to surface buoyancy fluxes cannot be directly compared to the subduction (the link between subduction and transformation is addressed in Marshall et al. 1999). Rather, in the limit of weak diffusive buoyancy flux, the rate of transformation can be related to the net advective volume flux across isopycnals (e.g., Speer et al. 2000).
The advective transports can be represented by the overturning streamfunction, Ψ(σ) in isopycnal coordinates. For meridional flows, this streamfunction is computed by integrating zonally along isopycnals (e.g., Park and Bryan 2000) as follows:
so that the net meridional volume transport between any 2 isopynals, σ1 and σ2, is ΔΨ ≡ Ψ(σ1) − Ψ(σ2). For individual ocean basins, such as the Indo-Pacific and Atlantic Oceans, the difference Ψ(σsurface) − Ψ(σbottom) should balance the transport through the Bering Strait (our ocean model uses the rigid-lid approximation).
a. Extratropical North Pacific
The penetration of low PV mode waters into the oceanic interior along some isopycnal surfaces is shown in Fig. 3. On the lighter isopycnals (top panels), these mode waters can be associated with the North Pacific Subtropical Mode Water (NPSTMW) (Masuzawa 1969) and the North Pacific Eastern Subtropical Mode Water (NPESTMW) (Hautala and Roemmich 1998). In models, including in the one employed in this study, the isopycnals along which the low PV signal penetrates into the subsurface ocean may not always be the same as those in observations. Nevertheless, we find it satisfactory that the distribution of PV in our model closely resembles that based on an ensemble mean from 11 climate models in Luo et al. (2009a).
On denser surfaces (Fig. 3, bottom panels), the low PV tongue corresponds to the North Pacific Central Mode Water (NPCMW) (Nakamura 1996; Suga et al. 1997). Also shown in Fig. 3 are the contours of the winter MLD. Compared to observational estimates, the model tends to overestimate somewhat the MLD in the region west of about 170°W and between 30° and 40°N, where the winter MLD standard deviation is also large (Suga et al. 2004; Ohno et al. 2009). The total area where the winter MLD is deeper than 200 m is 6.7 × 1012 m2 in 1 × CO2; it decreases to 5.6 × 1012 m2 in 2 × CO2—a reduction of more than 15%. The corresponding mean PV values at the base are 1.7 × 10−10 and 2.2 × 10−10 m−1 s−1 in 1 × CO2 and 2 × CO2, respectively. In agreement with Xie et al. (2000) and with the multimodel analysis of Luo et al. (2009a), mode waters tend to form in the regions where large gradients of the winter MLD intersect with the outcropping line. As implied by Eq. (2), such regions are favorable for an enhanced lateral mass exchange between the mixed layer and the interior, provided that the flow does not exactly follow the contours of the mixed layer base (i.e., uh · ∇h ≠ 0).
Comparing the right and left panels in Fig. 3, we find that the high PV signal in the 2 × CO2 climate becomes stronger, whereas the low PV signal is weaker, consistent with Luo et al. (2009a). While important, the weakening of the low PV signal in the oceanic interior could be due to the stronger stratification and higher PV values at the mixed layer base and it may not necessarily imply a weakening of mode water subduction and transport. To conclude about the latter, these quantities need to be estimated more directly. In particular, it follows from Eq. (2) that changes in the base of the mixed layer can strongly influence the lateral component of subduction.
In the 1 × CO2 climate, the model simulates at least two distinct peaks in the North Pacific subduction (Fig. 4a). A broad maximum around 25.4 σθ corresponds in the model mostly to NPSTMW. In the observational estimates of Qiu and Huang (1995), a similar maximum is found around σθ = 25.1. There are also two secondary subduction peaks—at σθ = 25.9−26.0 and around σθ = 26.2 (Fig. 4a)—that correspond to NPCMW found by Suga et al. (1997). Tsujino and Yasuda (2004), using a regional ocean model with the horizontal resolutions of ¼° and ⅙° (in the zonal and meridional directions, respectively), found that the annual subduction in the subtropical North Pacific has major peaks at 3 density classes of σθ = 25.5, 26.2, and 26.5. In our model, there is also a maximum at σθ = 26.8 (Fig. 4a), which roughly corresponds to North Pacific Intermediate Water (NPIW) (Talley 1993). Overall, given that we use an observationally unconstrained (including in terms of the surface fluxes) fully coupled model, the simulated annual subduction in the North Pacific appears to be satisfactory.
The net subduction between σθ = 24.8–25.7, which combines the contributions due to the subduction of NPSTMW and NPESTMW, is 31.7 Sv, of which 19.8 Sv is due to the lateral transfer. Thus, within these relatively light density classes, more than one-third of the annual mode water subduction in the North Pacific is due to the vertical transfer. This is expected, since the light water within the subtropical gyres is generally pumped down by the Ekman flux. For the denser water classes, the subduction is dominated by the lateral transfer (Fig. 4a). While the area of winter MLD becomes smaller in the subpolar Pacific, the product (−uh · ∇h) increases, so that the integrated lateral transfer remains relatively large. We also note that the contribution due to the eddy-induced velocity is relatively small, except for the density classes corresponding to the Kuroshio Extension (Fig. 4a). The eddy contribution should, however, be revisited using a model with higher resolution.
In the 2 × CO2 climate, the peak of subduction associated with the intermediate waters in the North Pacific significantly decreases (Fig. 4b; it should be noted, however, that this global model is not intended to simulate accurately the processes within the Sea of Okhotsk, where some NPIW is thought to form; e.g., Shcherbina et al. 2003). Within the density classes representing the mode waters of North Pacific, the subduction shifts to the lighter densities (Fig. 4b). However, the net rate of subduction does not decrease much. Taking σθ = 24 as the upper limit for the light waters that can still move southward across 20°N in the Pacific (Fig. 6), and integrating between σθ = 24.0–26.5, the subduction in the North Pacific (Fig. 4b) within this rather wide density range is 58.9 Sv in 1 × CO2 and 57.3 Sv in 2 × CO2. However, the subduction generally decreases for σθ > 25.2, in part because of the decrease in the subduction of NPCMW, and increases for the less dense waters (Fig. 4b).
The rate of diapycnal transformation due to the surface buoyancy flux is mostly positive in the North Pacific, indicating a conversion from lighter to denser water classes (Fig. 5). In the 2 × CO2 climate, the peak of this conversion shifts to lighter densities; however, the model does not predict a significant change in the net diapycnal transfer in the North Pacific because of the surface buoyancy flux. From Fig. 5, 15–20 Sv are transformed into denser classes between σθ = 24.5–26.0 in 1 × CO2 and between σθ = 24.0–25.5 in 2 × CO2. In the absence of a strong diffusive flux, this should match the corresponding volume fluxes. From Fig. 6, about 15 Sv move southward cross 20°N between σθ = 24.5–26.0 in 1 × CO2 (Fig. 6, top) and between σθ = 24.0–25.5 in 2 × CO2 (Fig. 6, bottom). This is supplied by the diapycnal volume flux from the light to the denser waters, which, across the densities around σθ = 24.5, occurs largely between 25° and 35°N, both in 1 × CO2 and in 2 × CO2 (Fig. 6). Thus, the overturning circulations in Fig. 6 are largely compatible with the corresponding transformations in Fig. 5, implying that the tropical surface waters lose their buoyancy as they move north. Upon conversion to the denser classes, the water returns to the tropics near adiabatically in the subsurface interior (Fig. 6).
Talley (2003) provides observational estimates for several components of volume transport, including those associated with shallow overturning circulation in all ocean basins. In the North Pacific, her estimate for the interior transport across 24°N is close to that simulated (Table 1). This is encouraging, given that one of our aims here is to estimate the potential changes in the shallow overturning. A detailed (in density bins of Δσ = 0.1) structure of the volume transports across 30°N in the Pacific, that is, about 10° to the south from the Kuroshio Extension, is shown in Fig. 7. For the density classes of mode waters, the largest southward transports are around σθ = 25.5. In 2 × CO2, the transport weakens for denser waters, but it becomes stronger for waters between σθ = 24.5– 25.3 (Fig. 7, top). For the density classes between σθ = 24.5–25.0, the transport is southward in 2 × CO2 but northward in 1 × CO2. The shift to the lighter density classes, essentially because of the upper-ocean warming under the 2 × CO2 forcing, is further illustrated in the bottom panel of Fig. 7 for waters with σθ < 24 (note that, within these lightest densities, the interior flows outside of the western boundary do not compensate for the strong transport within Kuroshio, presumably because the zonal wind stress is weak near 30°N; therefore, the net transport in the bottom panel of Fig. 7 is northward).
b. Extratropical southern Indo-Pacific
The penetration of low PV waters into the Indo-Pacific interior, along representative isopycnal surfaces for the 1 × CO2 and 2 × CO2 climates is illustrated in Fig. 8. Also shown are the regions within the 2 climates where the winter MLD is deeper than 300 and 400 m, respectively Using proxy data for the winter MLD (the depth of 95% oxygen saturation), Hanawa and Talley (2001) show that the MLD can be deeper than 300 m over vast areas in the Indian and Pacific sectors of the Southern Ocean, which is consistent with the simulated pattern of the MLD (Fig. 8, top). The Southern Ocean MLD can locally reach 700 m in both the model and the observations.
In the warmer climate (Fig. 8, bottom), the area of large winter MLD becomes notably smaller. In particular, the area where winter MLD is deeper than 300 m in the Indo-Pacific is 2.2 × 1013 m2 in 1 × CO2, whereas it is only 1.5 × 1013 m2 in 2 × CO2—a reduction of more than 30%. The total volume contained within the winter MLD (where the MLD is deeper than 50 m) decreases by about 10%, which is within the range of the decrease projected by the IPCC Fourth Assessment Report (AR4) models under the SRES A1B scenario (Sen Gupta et al. 2009). The corresponding mean PV values at the mixed layer base are 1.2 × 10−10 and 1.5 × 10−10 m−1 s−1 in 1 × CO2 and 2 × CO2, respectively. The associated low PV mode water is known as the Subantarctic Mode Water (SAMW) and is observed within σθ = 26.5–27.1 (McCartney 1977). A variety of SAMW in the Indian Ocean is called the Southeast Indian Subantarctic Mode Water (SEISAMW). In observations, it is associated with a vast region of the southeast Indian Ocean where the winter MLD can be deeper than 300 m (Hanawa and Talley 2001); the model also captures this feature (Fig. 8, top). The SEISAMW has a major impact on the ventilation of the Indian Ocean’s subtropical gyre thermocline through subduction and northward advection (e.g., Hanawa and Talley 2001). The SAMW is thought to be formed via a surface buoyancy loss, which is largely a result of the northward Ekman transport of cool Antarctic surface waters across the Subantarctic Front (SAF). Therefore, the relative importance of these two processes, that is, the surface buoyancy loss and northward Ekman transport across the SAF remains unclear (Rintoul and England 2002; Hanawa and Talley 2001), and it cannot be easily separated in a coupled system.
Within the density classes corresponding to the SAMW, the model simulates cooling and freshening on isopycnals in response to the 2 × CO2 forcing (Figs. 9a,c). This is consistent with Downes et al. (2009). In the water column where the ratio Rρ = αθz/βSz > 1 (typical of the main thermocline), such a cooling and freshening on isopycnals in the subsurface ocean can result from warming (or freshening) in the subduction regions at the surface (Church et al. 1991; Bindoff and McDougall 1994). The interior cooling on the same isopycnal (σθ = 25.9) in response to the positive radiative forcing is illustrated in Figs. 9b,d.
The salinity minimum associated with the core of Antarctic Intermediate Water (AAIW) shifts to a lighter density class in the 2 × CO2 climate and becomes warmer and fresher in both the Indian and Pacific Oceans (Figs. 9a,c) (note that in observational studies, this salinity minimum is typically taken as the σθ = 27.2; the model tends to simulate this minimum at somewhat lighter densities). Within the density classes that can be associated with the denser variety of AAIW in the model (27.1 < σθ < 27.4), the subsurface waters become warmer and more saline on isopycnals (Figs. 9a,c). This suggests that changes in the patterns of wind stress and/or in the rate of renewal play a role. In particular, within this density range, the simulated vertical motions become more upward in the warmer climate (the associated impact on the rate of subduction is quantified below). Climate models, including the one employed here, typically simulate significant changes in the Southern Ocean wind stress and wind stress curl in response to a positive radiative forcing (e.g., Sen Gupta et al. 2009; Saenko et al. 2005).
The model simulates maximum subduction around σθ = 26.2 (Fig. 10a). For denser water classes (26.2 < σθ < 27.3), the annual subduction is dominated by lateral transfer, much as in the North Pacific. If fact, for densities around σθ = 27.0–27.2, the vertical transfer tends to weaken the net subduction (Fig. 10a), as might be expected in the case of generally upward motions. It is also between σθ = 26.9–27.3 where the (parameterized in the model) mesoscale eddies contribute significantly to the net subduction (Fig. 10a), generally counteracting the Ekman upwelling. This is consistent with Sallée et al. (2010), who estimated the overall transports (i.e., the transports toward lighter and denser waters).
Within the densities lighter than σθ = 26.2, the vertical transfer becomes essentially as important for the net subduction as the lateral transfer (Fig. 10a). This might be expected in the case of predominantly downward motions. Observational estimates of Karstensen and Quadfasel (2002) also point to the important role of vertical transfer for subduction within light density classes. In contrast, Downes et al. (2009) argue for little input from the vertical term for subduction of SAMW. As pointed out by Downes et al. (2009), the criterion they used to define MLD was different from that used by Karstensen and Quadfasel (2002). We have repeated our calculations using the MLD criterion adopted by Downes et al. (2009), but we still found a significant contribution of the vertical transfer to the net subduction for 25.0 < σθ < 26.0 (not shown). We note that, in general, it seems reasonable to expect the vertical transfer to play a role in the subduction north of SAF, that is, north of 0 wind stress curl, up to σθ = 24.8.
Much like in the North Pacific, the peak of mode water subduction in the Southern Ocean shifts to lighter density classes in the 2 × CO2 climate (Fig. 10b). The rate of net subduction is also affected but not much. Integrated between σθ = 25.5–26.5, about 82.3 and 77.4 Sv are subducted in 1 × CO2 and 2 × CO2, respectively, in the southern Indo-Pacific. Most of this change is due to the lateral transfer, which within this density range decreases from 56.6 Sv in 1 × CO2 to 52.1 Sv in 2 × CO2, whereas the vertical transfer remains essentially unaffected.
Within the water classes that can be associated with the densest SAMW in the model and with AAIW, the annual subduction decreases (Fig. 10b). Integrated between σθ = 26.5–27.3, about 32.6 Sv are subducted in 1 × CO2, whereas only 23.4 Sv are subducted in 2 × CO2. A fraction of this 9.2-Sv decrease (about one-third) is due to the vertical subduction becoming more negative (stronger upward motion) in the warmer climate within this density range. It should be noted, however, that although the subduction of AAIW is projected to decrease significantly in a warmer climate both in this study and in Downes et al. (2009), one would have to employ a model with an appropriate resolution to make such projections more reliable, given the important role of mesoscale eddies within the Antarctic Circumpolar Current (ACC).
The rate of water mass transformation in the Southern Ocean (Fig. 11) has a structure similar to that in Speer et al. (2000). Essentially, with the exception of the densest water classes corresponding to the formation of bottom waters, the transformation is negative within the dense waters classes, indicating surface buoyancy gain. This is associated with the upwelling of deep dense waters in the Southern Ocean and their conversion to lighter waters, with the corresponding net transformation reaching roughly 20 and 25 Sv in 1 × CO2 and 2 × CO2 climates, respectively (Fig. 11, right). Thus, according to the model, the net surface buoyancy gain and the associated dense-to-light water conversion is likely to increase in the Southern Ocean in a warmer climate, in particular in the climate with an atmospheric CO2 level twice as large as it is today. At steady state, this would have to be balanced by the corresponding increase in the net light-to-dense water conversion elsewhere.
Around σθ = 26, the transformation changes sign and becomes positive at the lighter densities. This indicates buoyancy loss and conversion from light to denser waters, which creates favorable conditions for the deep mixed layer on the northern side of the ACC. In the 2 × CO2 climate, this change of sign happens at somewhat lighter densities (Fig. 11), which is consistent with the overall shift of subduction to lighter densities. From Fig. 11 (left), the buoyancy flux in the Indo-Pacific converts some 25–30 Sv to denser water classes between σθ = 25.0–26.0 and σθ = 24.5–25.5 in 1 × CO2 and 2 × CO2, respectively. This is largely consistent with the corresponding volume transports in the Indo-Pacific (Fig. 12), where about 20 Sv move northward across 20°S between σθ = 25.0–26.0 in 1 × CO2 (Fig. 12, top) and between σθ = 24.5–25.5 in 2 × CO2 (Fig. 12, bottom). The corresponding circulation cell, particularly its upper branch, shifts to lighter densities in the warmer climate (Fig. 12). However, the rate of this circulation is not strongly affected, much like that in the North Pacific.
It should be noted that the model tends to overestimate the interior transports associated with the shallow overturning circulation in both the Pacific and Indian Oceans (Table 1). This must be kept in mind when examining the projected changes for the warmer climate. The main focus here is, however, on the net transport in the upper Indo-Pacific basin. This includes the waters carried not only in the interior of the gyres but also within the western boundary currents. In addition, we combine the transports in the two ocean basins, which effectively excludes the contribution of the ITF. Therefore, the discrepancy between the net simulated and observed transports in the upper Indo-Pacific Ocean (i.e., above roughly σθ = 27) is expected to be smaller, as compared to the case where the meridional transports are considered separately for each of the ocean basins.
A detailed (in density bins of Δσ = 0.1) structure of the volume transports across 32°S in the Indo-Pacific is shown in Fig. 13 (for densities larger than σθ = 27.0). In the 1 × CO2 climate, the maximum northward transport is around σθ = 26.1, with most of the water transported northward between σθ = 25.9–26.4 (Fig. 13, top). This corresponds to the density range where the so-called net surface buoyancy input, given by (Marshall et al. 1993)
increases (becomes less negative) by up to 0.5 × 10−5 kg m−1 s−3 in the 2 × CO2 climate [in Eq. (5), UEk and ρm are the Ekman transport vector and mixed layer density, respectively). The positive change in Bnet occurs despite the second term in Eq. (5) (i.e., the Ekman transport across isopycnals) tending to offset the increase in the buoyancy gain because of the surface flux, particularly between σθ = 25.9–26.4 (not shown). The maximum transport across 32°S shifts to 25.8 σθ (Fig. 13, top), where Bnet essentially does not change and subduction increases (Fig. 10, bottom). Furthermore, at lower densities, between σθ = 25.4–25.7, Bnet becomes more negative, by about 0.3 × 10−5 kg m−1 s−3. For densities between σθ = 26.6–27.0, the transport across 32°S is directed southward in both climates (Fig. 13, top). Within the lightest densities (Fig. 13, bottom), the transport is also mostly southward, indicating that at these latitudes (around 32°S), the western boundary flows are not totally compensated by the interior flows within these density classes.
c. Pacific subtropical–tropical thermocline pathways
Water exchange between the Pacific subtropical and tropical oceans is mainly due to a shallow, wind-driven meridional circulation (e.g., Fine et al. 1987; McCreary and Lu 1994; Liu et al. 1994). The surface light water is converted to denser water in the subtropics and flows equatorward near adiabatically (Figs. 14a,c; see also Figs. 6, 12), where it is converted back to lighter water and moves poleward. As the lower branches of these meridional cells, the Pacific thermocline water mass transport plays an important role in conveying the signals of the ocean–atmosphere extratropical interaction to the tropics, and thereby it can influence the variability of equatorial Pacific SST (Gu and Philander 1997; Lohmann and Latif 2005; Merryfield and Boer 2005).
The model predicts that in the 2 × CO2 climate, these circulation cells do not change significantly (Figs. 14b,d). In particular, in both climates, there is roughly the same net convergence of thermocline waters in the tropical Indo-Pacific and their conversion to lighter densities between 10°S and 10°N (50–55 Sv). (Note that, as can be concluded from Fig. 14, the upwelling in the tropical Indo-Pacific is somewhat stronger if depth, rather than density, is used as the vertical coordinate. This indicates that some portion of the equatorial upwelling is adiabatic). However, the streamfunctions shown in Fig. 14 represent the zonally averaged (along density or depth surfaces) transports, whereas there still can be important local changes to the thermocline water pathways (Luo et al. 2009b; McPhaden and Zhang 2002; Wang et al. 2008), which is our main subject in this subsection.
Three pathways can be identified for the thermocline transport (e.g., Gu and Philander 1997). In the North Pacific, for example, the thermocline water flows southwestward from the northeast, essentially along isopycnal surfaces (e.g., see Fig. 1 in McPhaden and Zhang 2002). Then, upon reaching the western boundary, part of it moves northward (first pathway), with the rest turning southward toward the equator (second pathway). In addition, the thermocline water can reach the tropics through the interior, that is, without encountering the western boundary, including by breaking a strong PV constrain around 10°N (third pathway). This PV constraint has been shown to become stronger under the global warming scenarios, making it even more difficult for the subtropical thermocline water to take the third pathway. In particular, Luo et al. (2009b) investigated the response of tropical–subtropical transport pathways to global warming. Using the IPCC AR4 model simulations, they suggested that global warming will lead to an intensified thermocline transport through the western boundary route (second pathway) and less transport through the interior route (third pathway).
Figures 15a,c show the structure of |PV| and circulation in the western part of the tropical Pacific on σθ = 25.4, which is representative of the pycnocline in 1 × CO2 (cf. Fig. 1 in McPhaden and Zhang 2002). As expected, the water tends to follow PV contours in the interior, crossing them mostly at the western boundaries. Nevertheless, there is a significant mass convergence of thermocline waters through the interior (third pathway). Integrated between σθ = 22.5– 26.0 (the positions of these isopycnals are shown in Fig. 16), a total of 14.2 and 38.7 Sv enter the tropical Pacific (across 7°N and 7°S) through the interior and the western boundary, respectively.
In 2 × CO2, the high |PV| zones, particularly the one around 10°N, extend farther west (Figs. 15a,b), consistent with Luo et al. (2009b). The equatorward boundary flows, in the region equatorward of 10°N and 10°S, generally intensify (Fig. 15d). However, as illustrated in Fig. 16 (top and middle panels on the left), the response of the tropical boundary flows to a doubling of CO2 has a complex structure. In particular, at 7°N, the southward boundary flow gets stronger between 70 and 250-m depth, whereas the northward boundary flow at 7°S intensifies only below 120-m depth (between about 120 and 360 m). Integrating between two isopycnals (e.g., Luo et al. 2009b)—such as, for example, σθ = 22.5–26.0 in 1 × CO2 (Fig. 16) and selecting the corresponding isopycnal limits for 2 × CO2 that have the same mean position (which are σθ = 21.5–25.75)—the net convergence through the boundary transports at 7°S and 7°N increases only by 2 Sv. From Fig. 16 (top and middle panels on the left), it can be seen that, within the chosen σθ ranges, the equatorward boundary flows not only strengthen in the warmer climate but also weaken. Therefore, such estimates, including those in Luo et al. (2009b), should be viewed in the context of the density classes they represent, since these integrated transports can be sensitive to the isopycnals selected as the limits of integration. This also applies to the interior flows, where we find that the net convergence within the indicated density layer essentially does not change in the 2 × CO2 climate. However, as illustrated in Fig. 16 (top and middle panels on the right), there are regions where the interior flows both strengthen and weaken in the warmer climate.
We next consider changes in the ITF transport, which has long been a focus of research for its impact on regional and global ocean circulation and climate (e.g., Gordon 2001; England and Huang 2005). Using the model results from the IPCC AR4 simulations, Wang et al. (2008) found that the ITF transport weakens. In our simulations, consistent with Wang et al. (2008), the ITF transport decreases from 19.1 Sv in 1 × CO2 to 16.3 Sv in 2 × CO2. Most of this weakening happens in the upper ocean, although some penetrates to the deeper ocean (Fig. 16, bottom panels). Within the density layers of σθ = 22.5–26.0 and σθ = 21.5–25.75 in 1 × CO2 and 2 × CO2, respectively, the transport decreases from 12.3 to 10.3 Sv. At least some of this decrease is redirected to the east (Fig. 15d), feeding the upper EUC. In turn, consistent with Wang et al. (2008) and Luo et al. (2009b), we find the model simulates the EUC anomaly (i.e., the upper part of the EUC intensifies, whereas its lower part weakens; Fig. 17, left panels). Luo et al. (2009b) suggest that the strengthening of the upper part of the EUC may be linked to the enhanced boundary transports (Figs. 15d, 16, top and middle left panels), whereas Wang et al. (2008) suggest that the flow redirected from the ITF may feed the EUC anomaly (Fig. 15d). While it is likely that both these processes contribute, we note that the ITF transport is too strong (by about 4 Sv) in the 1 × CO2 model compared to the recent estimate of 15 Sv from full-depth velocity measurements reported by Sprintall et al. (2009).
The EUC anomaly closely follows the anomaly in stratification along the equator (Fig. 17). The increase in the buoyancy frequency, centered at approximately 100-m depth, tends to inhibit the vertical transfer of momentum, making the upper EUC even stronger and weakening it at its lower edge. However, despite the relatively large changes in stratification along the equator, the main parameters of the EUC are only slightly affected. This seems to be consistent with the inertial scalings (Pedlosky 1987), according to which the thickness H and velocity U of the EUC can be estimated from
where τ, L, ρ, and g′ = gΔρ/ρ are the characteristic values of wind stress, zonal scale, density, and reduced gravity, respectively. These scalings imply, in particular, that an increase in the vertical density contrast Δρ between the core of the EUC and the deeper ocean by 10%–15%, that is, what is simulated under the 2 × CO2 forcing, would result in a decrease of the current thickness by only 5%–7%, whereas its velocity would essentially remain unaffected (the corresponding scaling for the current width indicates that it is even less sensitive to stratification and wind stress). Thus, near 150°W—that is, where the EUC is strongest in the model—H decreases by about 5%, whereas the maximum velocity increases by less than 3%, from 88.4 to 90.9 cm s−1.
Changes in the wind stress, including those outside the tropics, also could affect the EUC. Liu and Philander (1995) suggest, based on numerical experiments, that a reduction of the subtropical wind stress curl by a factor of 2 can increase the equatorial stratification by 30%. Climate models, including the one employed here, do simulate changes in the low latitude wind stress and wind stress curl in response to CO2 increase, but generally these changes are small on the zonal mean (e.g., Saenko et al. 2005).
5. Discussion and conclusions
Using a climate model, we diagnose subduction, water mass transformation, and transport rates in the Indo-Pacific Ocean and their changes in response to a doubling of atmospheric CO2. There are several conclusions from this study. In particular, it is found that subduction across the base of the winter mixed layer is dominated by the lateral transfer, particularly within the relatively dense water classes corresponding to the densest mode and intermediate waters. Furthermore, within some density classes, the vertical transfer contributes negatively to the net subduction, as may be expected for regions with predominantly upward vertical motions. However, within the relatively light densities, including those characterizing the lighter varieties of mode waters, the vertical transfer is found to have a strong positive input to the net subduction. This may be expected for regions with predominantly downward vertical motions, such as subtropical gyres. It is also found that, much like subduction, the volume transports across 30°N and 32°S in the subsurface upper ocean are largest within the density classes that correspond to mode waters. Since the along-isopycnal transport in the interior is supplied, at least in part, by the cross-isopycnal convergence within the uppermost ocean, its sign and even the magnitude can be in some cases inferred from the surface buoyancy flux. In the North Pacific, the buoyancy flux converts the near-surface waters mostly to denser water classes, whereas in the Southern Ocean the surface waters are transformed both to lighter and to denser water classes, depending on the density (cf. Speer et al. 2000).
In response to a doubling of CO2, the subduction, transformation, and transport of mode waters in both hemispheres shifts to lighter densities, but they do not change significantly. In contrast, the subduction of intermediate waters decreases. This result is consistent with some previous studies (e.g., Downes et al. 2009), but it needs to be verified using models that “permit” mesoscale eddies. The regions of maximum mixed layer depths decrease, particularly in the Southern Hemisphere where the total area with winter MLD deeper than 300 m decreases by 30%. This implies, in particular, a decrease in the effective heat capacity of the mixed layer that creates a positive feedback onto the climate warming.
Moving toward the low latitudes, there is a slight increase in the thermocline water that enters the tropical Pacific with the western boundary flows. This is favored by the structure of PV response and is consistent with some previous studies (e.g., Luo et al. 2009b). We find, however, that the structure of meridional velocity response in the Pacific at 7°N and 7°S is complex, so that such transport estimates can be sensitive to the selected isopycnal ranges. In addition, it is found, consistent with Luo et al. (2009b) and Wang et al. (2008), that the mean position of the EUC shifts upward, and so does the position of maximum buoyancy frequency along the equator. Despite the relatively large changes in the vertical density contrast, the main parameters of EUC, such as its thickness and maximum velocity, are only slightly affected, which appears to be consistent with the inertial scalings. The transport through the ITF decreases by 2.8 Sv, mostly in the upper ocean. At least a fraction of this decrease is redirected eastward and, jointly with the stronger inflow of thermocline waters at the western boundaries, feeds the upper EUC. By affecting the vertical transfer of momentum, the change in stratification further contributes to the vertical structure of EUC anomaly, tending to make the upper part of the EUC stronger and its lower part weaker. Thus, if these predictions were to be realized, one implication would be that the stronger upper EUC, combined with the increase in the vertical temperature contrast, would increase heat transport from the west to the eastern equatorial Pacific, between about 50 and 150-m depth. This, as well as the increase in stratification, may have important implications for the tropical ocean wave dynamics and climate variability, including that associated with ENSO.
We thank the climate model development group at the Canadian Centre for Climate Modelling and Analysis. We also wish to thank the anonymous reviewers for their comments, which helped to improve the original manuscript. X-YY was supported by the Government of Canada Program for International Polar Year (IPY) project Variability and Change. MHE was supported by the Australian Research Council.
Corresponding author address: Oleg A. Saenko, Canadian Centre for Climate Modelling and Analysis, University of Victoria, P.O. Box 3065 STN CSC, Victoria BC V8W 3V6, Canada. Email: email@example.com