Global Teleconnections in Response to Freshening over the Antarctic Ocean (original) (raw)

1. Introduction

There is emerging evidence showing that the southern high-latitude ocean has undergone significant changes during recent decades, including rising temperatures between 700 and 1500 m, general salinity declining in the intermediate water, sea level elevating at a relatively slow pace, and so on. (e.g., Gille 2002; Boyer et al. 2005; Jacobs 2006; Mayewski et al. 2009). Also, a reduction of sea ice extent has been found in the Ross and Bellingshausen Seas in recent years (Jacobs 2006). Such changes, especially the freshening of the Antarctic Ocean, have been receiving great attention, as they may be related to elevated greenhouse gases (Boyer et al. 2005). Several Antarctic ice shelves appear to break up suddenly during recent years, and the Larsen B Ice Shelf has lost 60% of its icebergs since 1995 (Doake et al. 1998). Based on statistical analysis, Rignot et al. (2008) indicated the West Antarctic has suffered great ice melting; ice sheet loss in the Amundsen and Bellingshausen Seas had increased by 59% within 10 yr, and the loss rate in 2006 even reached up to as high as 132 ± 20 Gt yr−1. Moreover, geological evidence demonstrated that during major deglaciations, large amounts of meltwater discharged into both the North Atlantic (e.g., Bond and Lotti 1995; Broecker 1998; Peltier and Solheim 2004) and the Southern Oceans (e.g., Shackleton et al. 2000; Blunier and Brook 2001), triggering the meltwater pulse (MWP) event, which could induce significant impacts on global climate (e.g., Broecker 2000; Clark et al. 2002; Rahmstorf 2002; Weaver et al. 2003). Since the freshwater storage of the Greenland Ice Sheet is limited, which is not able to release enough freshwater to cause the major MWP event (Clark et al. 1996), studies based on sediment record and modeling analyses tend to attribute the primary source of freshwater to the melting of the Antarctic Ice Sheet (Kanfoush et al. 2000; Bassett et al. 2005; Huang and Tian 2008). In the South Atlantic, the ice-rafted detritus (IRD) was found to increase in abundance during the MWP-1A event, which can be comparable to the Heinrich events in the North Atlantic, and suggested the massive breakup of the Antarctic Ice Sheets in that period (Kanfoush et al. 2000).

So far several mechanisms for the teleconnection from the southern high latitudes to the tropics and the Northern Hemisphere have been proposed. At short time scales, recent studies proposed a rapid teleconnection associated with oceanic wave adjustment from the southern high latitudes to the tropics. With an idealized model, Ivchenko et al. (2004, 2006) found signals generated by anomalies in the Antarctic sea ice cover/salinity distribution can propagate to the tropics in the form of fast-moving barotropic Rossby waves. With a coupled model, Richardson et al. (2005) found a significant warming in the eastern Pacific a few months after a reduction of salinity over the southern high latitudes, and attributed to the oceanic wave mechanism proposed by Ivchenko et al. (2004). This result seem to contradict the coupled model results by Blaker et al. (2006), who found significant warming in the eastern tropical Pacific shortly after an increase of salinity over the southern high latitudes, although they both attributed this warming to the same wave mechanism. This discrepancy in the tropical Pacific response may be either due to a small ensemble size used in these coupled modeling studies, or it suggests that the warming in the tropical Pacific may not be a robust response.

In addition to the fast oceanic wave transmissions, studies also suggest a fast atmospheric teleconnection from the southern high latitudes to the tropics and the Northern Hemisphere. Observations show statistically significant correlation between the Antarctic and tropical climate variability (e.g., White and Peterson 1996; Yuan et al. 1996; Peterson and White 1998; Yuan and Martinson 2000; Venegas and Drinkwater 2001), which are usually attributed to the origin of the teleconnection to the tropics. In contrast, in a coupled model simulation, Blaker et al. (2006) appeared to find a fast atmospheric teleconnection associated with salinity forcing from southern high latitudes to the tropics. However, in their study it remains unclear how the SST anomaly induced by local sea surface salinity (SSS) change affects the southern atmospheric circulation, or whether the atmospheric teleconnection is triggered by the tropical SST change induced by the oceanic wave processes.

At longer time scales, the teleconnection is associated with changes of global meridional overturning circulation (MOC). Proxy records and coupled ocean–atmosphere model simulations show that the freshwater input over the southern high latitudes can change the balance between the North Atlantic Deep Water (NADW) and the Antarctic Bottom Water (AABW) formation, a key factor affecting the global thermohaline circulation (e.g., Broecker 1991; Stocker 1998; Weaver et al. 2003; Seidov et al. 2001, 2005; Stouffer et al. 2007). The alteration of this balance may lead to changes of cross-equatorial oceanic heat transport, inducing a bipolar seesaw (e.g., Stocker 1998; Seidov and Maslin 2001). Broecker (1998, 2000) suggested that a weakening of AABW formation can lead to an increase of the Atlantic meridional overturning circulation (AMOC) and vice versa. Using a simple coupled model, Weaver et al. (2003) found that a freshening over the southern high latitudes can lead to a switch of the AMOC from an “off” to an “on” state. However, recent coupled general circulation model (GCM) studies suggest that the freshening in the southern high latitudes can spread over the global ocean within several decades, inhibit the formation of NADW, and may not be able to lead to an increase of the AMOC (Seidov et al. 2005; Stouffer et al. 2007).

This study examines extratropical–tropical and interhemispheric teleconnections in response to the Antarctic freshening from a coupled ocean–atmosphere perspective. We conduct a series of coupled GCM experiments based on the “modeling surgery” strategy developed by Wu et al. (2003) to explicitly assess the effects of key teleconnective processes, complementary to most of coupled model studies that are usually based on the diagnosis of one single perturbation experiment. The modeling surgery is effective in distinguishing the roles of oceanic and atmospheric teleconnections in transmitting climatic anomalies, which has been demonstrated in a recent water-hosing study to examine the global climate teleconnection in response to a shutdown of the AMOC (Wu et al. 2008).

The paper is constructed as follows. Section 2 briefly describes the coupled model and sensitivity experiments. Sections 3 and 4 discuss local and remote responses, respectively. Section 5 is a summary with further discussions.

2. Model description and experiment design

We use the Fast Ocean Atmosphere Model (FOAM), version 1.5, a fully coupled global model developed jointly at the University of Wisconsin—Madison and the Argonne National Laboratory. This is the improved version of the original FOAM (version 1.0), which is described in detail in Jacob (1997). The atmospheric model is a parallel version of the National Center for Atmospheric Research (NCAR) Community Climate Model, version 2 (CCM2), but the atmospheric physics is replaced with those of CCM3. The ocean model was developed following the Geophysical Fluid Dynamics Laboratory (GFDL) Modular Ocean Model (MOM). The FOAM used here has an atmospheric resolution of R15 with 18 vertical levels and an oceanic resolution of 1.4° latitude × 2.8° longitude with 32 vertical levels. The coupled model has a thermodynamic sea ice component. Without flux adjustment, the fully coupled control simulation has been integrated for more than 1000 years, without an apparent climate shift. FOAM captures the major features of the observed climatology (Jacob 1997). FOAM also produces reasonable climate variability, such as ENSO (Liu et al. 2000), North Atlantic climate variability (Wu and Liu 2005) as well as the AMOC (Wu et al. 2008).

To investigate the global climatic response to freshening over the Antarctic Ocean, 1 Sv (1 Sv ≡ 106 m3 s−1) freshwater flux is added uniformly to the south of 60°S (the intensity of freshwater forcing is equal to about 156 cm yr−1) as a virtual salt flux. This rate is fixed for 400 model years. This experiment is named the Southern Ocean Water-Hosing (SOWH) experiment. A parallel experiment starting from the same initial condition is also conducted for 400 yr with no freshwater anomaly added and this control run is referred to as CTRL hereafter. The differences of the last 100 years’ average between these two experiments are taken as the modeling responses.

3. Local ocean–atmosphere–sea ice responses

The injection of the freshwater flux induces significant coupled responses of ocean–atmosphere–sea ice in the Antarctic region. The input of freshwater flux poleward of 60°S leads to a decrease of global SSS with the surface advection of regional SSS anomalies (Fig. 1a) and a corresponding rise up of the global sea level (Fig. 1b). Moreover, southern high-latitude freshwater perturbation triggers an intensive regional surface cooling due to a stabilization of the water column, which inhibits deep convection and thus the formation of the AABW. The transport of the Antarctic meridional overturning circulation (AnMOC), defined as the minimum streamfunction of the meridional overturning circulation south of 60°S, is reduced from 15 to about 5Sv within the initial 50 yr and reaches a new equilibrium state (Fig. 1c).

Although freshwater forcing is limited to the south of 60°S, anomalous local cooling stimulates a significant global climatic response (Fig. 2a). The cooling is more intense in the Pacific sector, with the maximum reaching −6°C near the Ross Sea. The intensive cooling is accompanied by downward heat flux anomalies from the atmosphere to the ocean, which plays a damping role to mitigate the surface cooling (Fig. 2b). Moreover, the cooling is coupled with an intensification of the westerly wind within the zonal band of 40°–60°S (Fig. 2b), corresponding to the presence of positive phase of the southern annular mode (SAM) (Fig. 2c). A heat budget analysis of the upper 100-m water column is conducted to reveal the dynamic processes responsible for the anomalous cooling occupying the Antarctic Circumpolar Current (ACC) region. The thermodynamic equation can be written as

i1520-0442-24-4-1071-eq1

i1520-0442-24-4-1071-eq1

Here ρ and Cp denote density and specific heat at constant pressure, respectively. Variables with prime and overbar are anomaly and the control run climatology, respectively. Terms from left to right on the right-hand side represent the contribution of mean zonal advection, anomalous zonal advection, mean meridional advection, anomalous meridional advection, mean vertical advection, anomalous vertical advection, surface heat flux, and the residual, respectively. The residual term usually reflects high-order dynamic processes, predominantly referring to the mixing process. For upper-layer heat budget analysis, we need to conduct a volume integration of the above equation (vertical depth is taken by 100 m here as the depth of mixed layer) for a specific region.

Our analysis result indicates that the cooling over the ACC region is predominantly due to the intensification of surface heat loss associated with the strengthening of the westerly winds. In addition, meridional advection of temperature gradient due to both the mean (−ρCp υ_∂_T ′/∂y) and anomalous (−ρCpυ_′∂_T/∂y) Ekman flow also contributes to transmitting cooling anomalies equatorward (Fig. 2d).

The freshening also leads to significant changes of both sea ice and atmospheric circulation. The cooling over the southern high latitudes induces a thickening and an expansion of sea ice (Fig. 3a). The atmospheric circulation demonstrates an equivalent barotropic low anomaly capping the South Pole with a magnitude of about 80 and 40 m at 250 and 850 mb, respectively (Figs. 3b,c).

The surface freshening induces substantial changes of subsurface ocean temperature and salinity (Fig. 4). The cooling is largely trapped in the upper 50 m, with substantial warming underneath. The warming can extend to depths of more than 3000 m, with a maximum of 6°C located at 250-m depth, along the Antarctic. This baroclinic response is associated with a reduction of deep convection, which prevents the warm subsurface water from entraining into the mixed layer. Moreover, we note that the ocean deeper than 3000 m is covered by cold anomalies, which may be associated with the change of the AMOC and will be discussed later. The intensification of the westerly winds also accelerates the northward Ekman transport north of 60°S, and the Antarctic divergence, leading to a modest strengthening of the meridional overturning cell north of 60°S (not shown).

In general, freshwater forcing over the Antarctic leads to a regional coupled response of ocean–atmosphere–sea ice. The freshening leads to a strong regional cooling and northward extension of sea ice margin, coupled with an acceleration of the westerly winds. The strengthening of the westerly winds plays an important role in conveying cold anomalies from the southern high latitudes to the entire ACC regions by enhancing both the surface heat loss and northward Ekman advection. Here we highlight the important role of the coupled response in transmitting the cooling from the southern high latitudes to the ACC region.

4. Remote responses

a. Southern extratropical and tropical responses

The freshening over the Antarctic region also induces significant responses over the southern subtropical and tropical oceans. The model demonstrates a significant cooling dominating the entire southern subtropical and tropical oceans (Fig. 2a), coupled with an acceleration of the southeasterly trades (Fig. 2b). In the following, we will explore the teleconnection mechanisms involving southern extratropical and tropical responses.

Since the subtropical South Pacific, South Atlantic, and Indian Oceans are all consistently dominated by cooling (Fig. 2a), here we will mainly focus on the South Pacific. To demonstrate the different processes involved in the development of cooling over the subtropical and tropical Pacific, a heat budget analysis of the upper 100-m water column is conducted for the subtropical (40°–10°S, 180°–280°E) and tropical (10°S–0°, 195°–280°E) oceans (see the two boxes in Fig. 2a over the Pacific sector). Over the subtropical South Pacific, the cooling is predominantly associated with the upward (ocean to atmosphere) heat flux and anomalous zonal advection (−ρCpu_′∂_T/∂x) (Fig. 5a). The intensification of the oceanic heat loss is caused by the acceleration of the southeasterly trades, which enhance the oceanic turbulent heat loss (Fig. 2b). The acceleration of the southeasterly trades also intensifies the zonal flow, which can push relatively cold water from the South American coast to the west. The cooling is partly offset by the anomalous vertical advection (−ρCpw_′∂_T/∂z), which tends to redistribute the cold anomalies down to the deep ocean. In addition, the enhancement of the mixing (which includes convection and diffusion, and is indirectly estimated through the imbalance between the changes of the oceanic heat content, advection, and external forcing contributions) associated with strengthened southeasterly trade winds also helps subsurface warm water to entrain into the mixed layer. Overall, the heat budget analyses indicate the cooling in the subtropical South Pacific is largely controlled by the surface process.

In the tropical Pacific, the heat budget analyses indicate that the cooling is associated with both subduction (−ρCp υ_∂_T ′/∂y) and an intensification of the subtropical–tropical cell (STC) (−ρCpυ_′∂_T/∂y), and damped largely by the mixing process, downward heat flux, and mean vertical advection (−ρCp w_∂_T ′/∂z) (Fig. 5b), which suggests that tropical cooling is not dominated by atmospheric forcing but largely determined by oceanic dynamics; thus, the downward surface heat flux is necessary to damp cold SST. This can be understood as follows: the intensification of the southeasterly trades strengthens the STC in the upper ocean (McCreary and Lu 1994; Liu et al. 1994), which can bring more cold water from the subtropics to the tropics to sustain the tropical cooling. The STC of the Southern Hemisphere is intensified shortly after Antarctic freshening (Fig. 5c); meanwhile, cold anomalies in the subtropics can also subduct to the equatorial region, seemingly following isopycnals, to enhance the tropical cooling (Fig. 5d). The diapycnal vertical mixing induced by enhanced trade winds also helps to bring deep warm water to move upward through the entraining process. Moreover, subsurface warm anomalies originated from the southern high latitudes, presumably through coastal Kelvin–Rossby wave adjustment (Ivchenko et al. 2004), can be upwelled to the surface in the tropics to offset the surface cooling (−ρCp w_∂_T ′/∂z). Besides, it should be noted that enhanced southern trade winds can accelerate the subtropical gyre, intensifying coastal upwelling in the western coast of the South America, which also contributes to southern extratropical cooling. Also, for the annual mean response, a narrow region in the western coast of the South America is occupied by anomalous downward heat flux and enhanced alongshore winds (Fig. 2b), implying the intensification of coastal upwelling.

Here we argue that the changes of the surface winds in the subtropical South Pacific is primarily caused by the surface ocean–atmosphere coupled process, namely, wind–evaporation–SST (WES) feedback. This mechanism was originally proposed by Xie and Philander (1994) and was invoked to explain the equatorial annual cycle by Liu and Xie (1994). Recent studies have found that the WES mechanism can be an effective way to transmit the extratropical influence to the tropics (e.g., Xie 1999; Chiang and Bitz 2005; Wu et al. 2007b; Chiang et al. 2008b). That is, the initial midlatitude surface cooling can give rise to anomalous sea level pressure (SLP) gradients and drive anomalous southeasterlies to enhance the background easterly trades, and hence increase evaporative cooling, pushing the initial cold SST farther equatorward (Chiang and Bitz 2005).

Seasonal development of cooling in the subtropical–tropical ocean provides evidence for the surface ocean–atmosphere coupled processes. In boreal spring, the cooling is largely limited to the south of 20°S (Fig. 6a). However, in the eastern subtropical South Pacific, southeasterly wind anomalies have developed in the equatorward flank of the SST anomalies, thus enhancing oceanic turbulent heat loss (Fig. 6b). This leads to an equatorward and westward extension of eastern subtropical cooling in the following season (Fig. 6c), triggering southeasterly trades (Fig. 6d). As the cooling largely occupies the eastern equatorial Pacific in boreal autumn (Fig. 6e), it sets up both zonal and meridional pressure gradient anomalies, inducing easterly anomalies over the equator and southwesterlies north of the equator (Fig. 6f). The enhanced equatorial easterlies lead to an amplification and westward propagation of equatorial SST anomalies, as a result of “Bjerknes feedback” (Bjerknes 1969). The cooling peaks in boreal winter (Fig. 6g) and decays afterward because of the surface damping. A similar propagation in the South Pacific has also been simulated in some recent modeling study with an idealized SST anomaly in the eastern subtropical South Pacific (Wu et al. 2007a; Matei et al. 2008). Besides, the characteristic of seasonal evolution of SST anomalies in the equatorial eastern Pacific, with earlier onset and later decay, has also been identified in a coupled model study to investigate the impact of interhemispheric thermal gradient on tropical Pacific climate (Chiang et al. 2008b).

To further assess how the extratropical cold anomaly propagates to the tropics in the Southern Ocean, we carried out 10-member ensemble experiments with each forced by the same freshwater flux over the Antarctic and integrated for 10 yr with different initial conditions from the control simulation. The anomalies discussed are the differences between the perturbed ensemble mean and the mean of the 10 “unperturbed ensembles” in the control run.

Shortly after the freshwater injection, an intensive surface cooling develops over the entire ACC region and extends rapidly to about 40°S (Fig. 7a). It can be seen that southeasterly wind anomalies and negative surface heat flux have developed farther north of the cold SST anomalies (Fig. 7a). This subsequently leads to an equatorward extension of the cold anomalies and further development of southeasterly wind and negative heat flux anomalies in the lower latitudes in the following year (Fig. 7b). The cooling has occupied the eastern equatorial Pacific at the fourth year, followed by a substantial development of equatorial easterly anomalies (Fig. 7c). The fast development of cooling in the tropics suggests an important role of the surface processes in conveying cooling anomalies from southern high latitudes to the tropics. It is noted that the entire equatorial Pacific is covered by cooling anomalies within only 10 yr (Fig. 7d).

The acceleration of the southern STC due to the intensification of the southeasterly associated with the WES process can be also demonstrated in this spinup experiment (Fig. 8). The transport of the southern STC is increased by about 3 Sv during the first 2–3 years (Fig. 8a) and about 5 Sv after about a decade (Fig. 8b). Associated with the acceleration of the southern STC, tropical Pacific cooling is eventually amplified (Figs. 8c,d).

Here we suggest that the surface air–sea coupled process plays a key role in the Antarctic–tropical Pacific teleconnection. To explicitly demonstrate the relative roles of different processes, we carried out two “partial blocking” (PB) experiments (Wu et al. 2003), which are similar to the SOWH experiment but with a sponge wall placed in the latitudinal band (45°–35°S). In the PB region, the oceanic temperature and salinity are restored to the model climatological values, and in this way, the sponge wall can effectively damp the anomalous temperature and salinity. In the first PB experiment, the sponge wall is placed from the surface to the bottom, while in the second one, the sponge wall is placed from 50-m depth to the bottom. These two experiments are named PB-ALL and PB-LOW, respectively. The PB-LOW experiment can essentially disable the subsurface oceanic teleconnection from the Antarctic to the lower latitudes, while the PB-ALL can disable both the surface and subsurface oceanic teleconnections. Each experiment is run for 200 yr, and the last 50 years’ integration is used for analysis. To prevent the model climatology drift due to the PB strategy, for each PB experiment a parallel experiment is also conducted, which is configured as the corresponding PB experiment but with zero freshwater anomaly added.

Both PB experiments demonstrate similar local responses as those in the SOWH, characterized by a coupled development of cooling and westerly wind anomalies at the surface and significant warming in the subsurface (Fig. 9). Outside the Antarctic Ocean, PB experiments demonstrate different responses from each other. In the PB-ALL experiment where the oceanic teleconnections, including both subsurface and surface coupled pathways, are disabled, the responses outside the Antarctic virtually disappear, although some weak cold anomalies exist in the subtropics (Figs. 9a,c). This readily suggests that teleconnection involving ocean dynamics and surface ocean–atmosphere coupling are critical in global responses, whereas the atmosphere alone cannot effectively mediate this teleconnection without interacting with the ocean. The PB-LOW experiment further isolates the role of surface coupled pathway versus subsurface oceanic teleconnection. Without the subsurface oceanic teleconnection from the Antarctic to the tropics, the subtropical and tropical Pacific demonstrate similar responses as those in the SOWH, with substantial cooling in the subtropical–tropical South Pacific coupled with an intensification of the southeasterlies and the equatorial easterlies (Fig. 9b). This further suggests that the surface coupled pathway plays a crucial role for the Antarctic–tropical teleconnection. However, without subsurface oceanic teleconnection, the tropical cooling is intensified because of the elimination of the subsurface warming from the Antarctic (Fig. 9b versus Fig. 2a). This is consistent with heat budget analyses, which demonstrate a damping role of tropical mean vertical advection (−ρCp w_∂_T ′/∂z) in the surface cooling. In addition, it is noted that warming over both the North Pacific and the North Atlantic disappears when the subsurface warming is eliminated because of a blocking of oceanic transmission from the southern high latitudes, indicating a potential role of deep ocean teleconnection in promoting warming in the Northern Hemisphere (Fig. 9b), which will be discussed in detail in the next section.

Generally speaking, the teleconnection processes in the South Atlantic and Indian Oceans are broadly similar to the South Pacific. Heat budget analyses in the South Atlantic and Indian Oceans also identify the dominant role of surface heat flux in controlling subtropical cooling, while tropical cooling appears to be associated with an intensification of the upper-ocean meridional overturning circulation and subductive process in the Southern Hemisphere (not shown). Additionally, the PB-ALL and PB-LOW experiments also reveal the important role of the surface coupled process in transmitting the Antarctic climatic influence to the low latitudes over the Atlantic and Indian Ocean sector, which is similar to the South Pacific (Figs. 9a,b).

Over the tropical Atlantic, a notable interhemispheric dipolelike SST anomaly occurs in summer, accompanied by a C-shape wind anomaly characterized by southeasterly wind anomalies in the south tropical Atlantic, southerly cross-equator wind anomalies, and southwesterly anomalies in the north tropical Atlantic (Figs. 6c,d). This coupled pattern largely indicates the tropical WES feedback (e.g., Chang et al. 1997; Xie et al. 1999). The meridional pressure gradient associated with the SST dipole can push the mean ITCZ farther northward in this season.

In summary, both the heat budget analyses and sensitivity experiments suggest the WES feedback plays an important role in the Antarctic–tropical teleconnection. The WES mechanism also intensifies the STC in the Southern Hemisphere to enhance the tropical cooling. In addition, subduction of the subtropical cold anomalies also helps to sustain the tropical cooling.

Our study here identifies a new mechanism for the Antarctic–tropical teleconnection, namely, a joint WES–STC relay teleconnection mechanism. The surface WES coupled process transmits high-latitude SST anomalies to the tropics and enhances the southeasterly trade winds, which modulate the STC to further affect the equatorial thermocline and SST. This mechanism has also been identified in a recent study of extratropical–tropical teleconnection over the North Pacific (Wu et al. 2007b). Therefore, the WES–STC relay mechanism appears to be a robust pathway for the extratropical–tropical teleconnection.

b. Northern extratropical responses

For the Northern Hemisphere, the most prominent feature is that both the North Pacific and the North Atlantic are broadly occupied by anomalous warming with the maximum reaching 0.5°C, although there appears some modest cooling over the high latitudes of the North Atlantic (Fig. 2a). The warming anomalies over both the North Pacific and the North Atlantic take about 200 years to reach equilibrium, indicating an important role of deep ocean teleconnection (Fig. 10). Indeed, within the first several decades, both the North Pacific and the North Atlantic are dominated by anomalous cooling, which is more significant over the North Pacific. A persist warming occurs after 70 yr for the North Pacific and even later for the North Atlantic (Fig. 10). Here, we will analyze dynamic mechanisms responsible for the initial cooling and subsequent warming.

Over the initial several years, cold SST anomalies in the North Pacific and the North Atlantic are both located near the west boundary current extension region (Fig. 11a). To demonstrate the roles of different processes involved in the initial cooling, a heat budget analysis of the upper 100-m water column is conducted for the North Pacific (10°–60°N, 120°E–180°) and the North Atlantic (30°–55°N, 285°–345°E).

Over the North Pacific, the cooling is predominantly associated with anomalous meridional advection (−ρCpυ_′∂_T/∂y) (Fig. 11b), indicating a dominant role of oceanic dynamics. The mechanisms can be understood as follows: over the Pacific, tropical cooling can reduce the Hadley circulation and the Aleutian low, and weaken the midlatitude westerly through the atmospheric teleconnection (e.g., Alexander et al. 2002), which produces positive wind stress curl anomalies over the subtropical and the midlatitude basins (Fig. 11f). This leads to anomalous southward boundary flow (Fig. 11d), which reduces the warm advection of the subtropical gyre and thus causes a development of cold anomalies over the western part of the basin. Moreover, it should be noted that the reduction of the Hadley cell also reduces poleward heat transport to compensate for the warming effect due to the weakened midlatitude westerly as a dynamic response to the La Niña–like cooling in the tropics, so there does not appear to be significant SST changes in the central North Pacific.

Over the North Atlantic, however, anomalous cooling is largely associated with the surface heat flux, while anomalous meridional advection and mixing play a damping role (Fig. 11c). In contrast to the North Pacific, although the positive wind stress curl anomalies dominate the subtropical North Atlantic, the western boundary current is intensified, which can be mainly attributed to the intensification of the AMOC (refer to the discussion section) (Fig. 11e). This warming effect, however, is overwhelmed by an intensification of the surface heat loss, ultimately leading to the surface cooling (Fig. 11c). In short, the initial cooling over the North Pacific and the North Atlantic are both in connection with atmospheric teleconnections originated from the tropical cooling. The major difference is that atmospheric teleconnection forces the North Pacific SST change through a modulation of the oceanic gyre circulation, while in the North Atlantic, atmospheric teleconnection directly triggers the surface cooling through the surface heat flux.

To further assess the role of tropical cooling in the northern ocean response, a “partial coupling” (PC) experiment (Wu et al. 2003) is carried out, in which tropical SSTs are replaced by climatological values to drive the atmospheric circulation above so air–sea coupling is deactivated in the global tropics (10°S–10°N); in this way, the atmospheric teleconnection from the tropics to the extratropics is turned off. The PC experiment is run for 50 yr and the last 20 yr are used. To prevent the model climatology drift due to the PC strategy, a parallel experiment is also carried out with no freshwater forcing added over the Antarctic to configure as the corresponding PC control run. The PC experiment explicitly shows that without air–sea coupling in the tropics, the cooling over both the North Pacific and North Atlantic virtually disappears and is replaced by anomalous warming (Fig. 12). This tropical PC experiment readily suggests that the tropical SST plays an important role in relaying the Antarctic influence to the Northern Hemisphere.

Concerning the distinct warming over both the North Pacific and the North Atlantic only appears several decades after the freshwater injection, the persist warming can be attributed to a slow oceanic teleconnection. To explore the mechanism involving in the Northern Hemisphere warming in detail, the evolution of zonal-averaged temperature and salinity anomalies in the initial 100 yr for the Pacific and Atlantic are tracked, respectively (Fig. 13). Over the Pacific sector, the initial subsurface warming in the southern high latitudes gradually spreads northward and arrives in the northern extratropical region 70 yr later. Meanwhile, when subsurface warm anomalies reach the northern subsurface ocean, the initial northern surface cooling is gradually replaced by warm anomalies because of the entrainment of subsurface warm water into the mixed layer, which is consistent with the time series of North Pacific SST (Fig. 10). A similar process can be also seen in the Atlantic sector, except over the northern high latitudes, where substantial cooling dominates both the surface and deep ocean. The latter is associated with the weakening of the deep convection in this area and will be discussed later.

Our modeling studies here demonstrate a collaborative role of the upper-ocean WES–STC relay teleconnection and the deep ocean in setting up the interhemispheric seesaw pattern in response to a freshening over the Antarctic Ocean. While the upper-ocean teleconnection conveys the cooling to the global ocean quickly within a few decades (Figs. 9b,d), heat accumulated in the deep Antarctic basin because of a shutdown of local deep convection acts as a reservoir, which eventually leads to Northern Hemispheric warming through vertical mixing. The warming effect due to the deep ocean heat reservoir induced by a shutdown of high-latitude convection has been recently invoked to explain the climate shift from the Last Glacial Maximum (LGM, about 21 000 years ago) to the Bølling–Allerød warm interval (about 14 500 years ago) (Liu et al. 2009).

So far we have not discussed the roles of the AMOC changes in leading to the bipolar seesaw. Previous studies indicate a weakening of AABW formation due to the surface freshening may lead to an increase of the AMOC, the meridional heat transport, and thus a bipolar seesaw (e.g., Weaver et al. 2003). However, the response of the AMOC to the Antarctic freshwater forcing can be more complicated. The change of the AMOC is predominantly determined by two factors—the interaction between the NADW and AABW—that play a competitive role (Broecker 1998; Weaver et al. 2003) and the northward spread of Antarctic fresh anomalies (Seidov et al. 2005; Stouffer et al. 2007). Within the initial several decades, the AMOC is slightly enhanced because of the strengthened NADW induced by the weakened AABW (Fig. 14b), and it starts to decrease after 50 yr when the fresh anomalies spread to the source region of the NADW (Fig. 14a). The cooling over the subpolar North Atlantic from the surface down to the deep ocean signifies the reduction of both local deep convection and the AMOC (Fig. 13, right panel).

5. Summary and discussion

In this paper, global teleconnections in response to freshening over the Antarctic Ocean are investigated in a series of sensitivity experiments (including the partial coupling and partial blocking experiments). A schematic diagram describing the major teleconnective process is shown in Fig. 15 and the main findings can be summarized as follows:

  1. A surface freshening over the Antarctic Ocean stabilizes the water column regionally, inhibiting deep convection, leading to intensive surface cooling and subsurface warming, and an intensification of the westerly winds. Both atmospheric teleconnection and upper-ocean advection associated with Ekman transport help to convey cold anomalies from southern high latitudes to the entire ACC regions.
  2. In the southern subtropics, the cooling subsequently propagates toward the tropics, predominantly through the coupled WES–STC relay teleconnective process. The WES mechanism accelerates the STC in the Southern Hemisphere, which further substantiates the cooling in the tropics.
  3. Over the northern extratropical ocean, some modest cooling is generated in the initial several decades, which is associated with tropical atmospheric teleconnection and subsequently replaced by significant warm anomalies induced by entrainment of subsurface warming from the Antarctic. The persistant warming in the Northern Hemisphere and cooling in the Southern Hemisphere forms an interhemispheric SST seesaw.
  4. The AMOC is slightly intensified in the first several decades of the Antarctic freshwater forcing due to the suppression of the AABW, but it is eventually weakened because of the propagation of surface fresh anomalies to the North Atlantic.

Our modeling study here identifies an interhemispheric seesaw similar to that found by Weaver et al. (2003) as a response to the Antarctic freshening, but with different mechanisms. In their study, the interhemispheric seesaw is associated with the acceleration of the AMOC due to a suppression of the AABW, which enhances the northward heat transport. Our study suggests the acceleration of the AMOC due to a suppression of the AABW is overwhelmed by the northward spread of the Antarctic fresh anomalies, which ultimately leads to a significant weakening of the AMOC, and the interhemispheric seesaw originates primarily from the upper-ocean WES–STC relay teleconnective process and the deep ocean teleconnection. In our case, the changes of the AMOC tend to offset the interhemispheric seesaw. This effect can be seen in the Pacific, where the meridional overturning circulation is much weaker than that of the Atlantic; however, the interhemispheric seesaw remains more robust than the Atlantic (Fig. 2a). Our study here appears to indicate two different adjustment phases of the AMOC to the Antarctic freshening: early acceleration regime and late deceleration regime. Recent modeling studies also indicate that the AMOC is sensitive to the intensity of Antarctic freshwater forcing (Swingedouw et al. 2008). Thus, the response of the AMOC to freshwater perturbation over the Antarctic may involve multiple processes.

Recently, the North Atlantic water-hosing studies suggest important roles of the WES mechanism (e.g., Chiang et al. 2008b) and oceanic nonlinear process associated with the interplay between the AMOC and the North Atlantic STC (Chang et al. 2008) in tropical Atlantic responses to a shutdown of the AMOC. Compared with the North Atlantic water-hosing, the freshwater forcing in the Antarctic has more direct impacts on the global tropical oceans. Since the AnMOC is much deeper than the AMOC without direct impacts on the STC, the tropical responses are largely controlled by the joint relay effect of the WES and STC adjustment proposed by Wu et al. (2007a). This joint WES–STC pathway appears to be an effective way in transmitting southern high-latitude anomalies equatorward as an important addition to the Antarctic–tropics teleconnective pathways discussed in previous studies (e.g., White and Peterson 1996; Yuan et al. 1996; Peterson and White 1998; Yuan and Martinson 2000; Venegas and Drinkwater 2001; Ivchenko et al. 2004, 2006; Richardson et al. 2005; Blaker et al. 2006). Furthermore, the interaction of the AnMOC and the AMOC is complicated by a global spreading of the surface fresh anomalies from the southern high latitudes, which does not appear with freshwater forcing in the subpolar North Atlantic.

The present study also suggests a possibility that the Antarctic freshening may exert significant impacts on the tropical climate variability mode. Over the tropical Pacific, both the warm pool and the cold tongue are enhanced (Fig. 2a), which intensify the zonal SST gradient and thus the Walker circulation. It is conceivable that the induced changes of the tropical mean climate should impact ENSO, which will be explored in the future.

Acknowledgments

This work is jointly supported by the Chinese National Science Foundation Outstanding Youth Program (Grant NSFC40788002), National Key Basic Research Program (Grant 2007CB411800), and National Creative Research Group Project (Grant NSFC40921004). Discussions with Drs. Ping Chang, Wenju Cai, and Zhaomin Wang were of great help. We wish to thank Dr. Chun Li and Ms. Chunxue Yang for conducting some numerical experiments. The authors are grateful to the two anonymous reviewers for their constructive comments, which improved the original manuscript in many aspects. All calculations were carried out on a SGI supercomputer at Ocean University of China.

REFERENCES

Fig. 1.

Fig. 1.

Fig. 1.

(a) Annual mean (psu) anomalies. (b) Annual mean sea surface height (SSH, m) anomalies. Note that the color scales in (a) and (b) are nonlinear. Values in (a) and (b) are significant at the level of 95% statistical significance using a t test. (c) Time series of minimum streamfunction (Sv) south of 60°S representing the AnMOC. The black solid line and gray dashed line denote the SOWH and the CTRL experiment, respectively.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 2.

Fig. 2.

Fig. 2.

(a) Annual mean SST (°C) anomalies. Solid black lines and dashed gray lines denote positive and negative values, respectively. The contours stand for −7°, −6°, −5°, −4°, −3°, −2°, −1°, −0.5°, −0.3°, −0.2°, −0.1°, 0.1°, 0.2°, 0.25°, 0.3°, 0.4°, 0.5°, 0.6°, 0.7°, 0.8°, and 1.0°C. (b) Annual mean surface heat flux (contours, W m−2) and surface wind (vectors, m s−1) anomalies. Positive (solid black lines) and negative (dashed gray lines) values represent downward and upward heat fluxes, respectively. Contours stand for −30, −20, −15, −12, −9, −7, −5, −3, −2, −1, 1, 3, 5, 10, 20, and 30 W m−2. Values over the shaded areas in (a) and (b) exceed the 95% and 90% statistical significance level, respectively, using a t test. Plotted arrows in (b) are significant at the level of 90% statistical significance. (c) Time series of SAM index in the SOWH case [hPa; the index of SAM is defined as zonal mean SLP gradient between 65° and 40°S (Gong and Wang 1998)]. (d) Heat budget terms (PW) for the upper ocean (0–100 m) water column in the ACC regions (60°–40°S).

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 3.

Fig. 3.

Fig. 3.

(top) Annual mean sea ice response, and geopotential height anomalies (gpm) at (left) 250 and (right) 850 hPa. In (a), black contours and shaded area represent ice thickness and ice fraction in the SOWH experiment, respectively. White line stands for northern boundary of sea ice in control run. In (b) and (c), values over shaded area exceed 95% statistical significance level using a t test.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 4.

Fig. 4.

Fig. 4.

Latitude–depth plot of zonally averaged annual mean temperature (°C, contours) and salinity anomalies (psu, shaded area). Logarithmic coordinates are adopted here in the vertical direction.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 5.

Fig. 5.

Fig. 5.

(top) Heat budget terms (PW) for the upper ocean (0–100 m) water column in (a) subtropical South Pacific and (b) tropical Pacific (see the two boxes over the subtropical South Pacific and tropical Pacific in Fig. 2a). Physical meanings of terms 1–8 as in Fig. 2. (bottom) (c) Anomalous meridional overturning circulation (Sv) for the upper ocean (0–400 m) in the Pacific for annual mean of years 30–50. (d) Latitude–depth plot of zonally averaged temperature anomalies (light contours, °C) and mean potential density (thick contours, kg m−3) in the Pacific for the mean of years 30–50. Solid black lines and dashed gray lines denote positive and negative values, respectively. The light contours represent −1.2°, −1°, −0.5°, −0.3°, −0.2°, −0.1°, 0.1°, 0.2°, 0.25°, 0.3°, 0.5°, 1°, 2°, 3°, 4°, and 5°C. The thick contours represent 22, 23, 24, 25, 26, 27, 28, 29, 30, and 31 kg m−3. Note that logarithmic coordinates is adopted in the vertical direction.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 6.

Fig. 6.

Fig. 6.

Seasonally stratified anomalies. Left panels show SST [°C; solid black lines and dashed gray lines denote positive and negative values respectively (contours represent −7°, −6°, −5°, −4°, −3°, −2°, −1°, −0.5°, −0.3°, −0.2°, −0.1°, 0.1°, 0.2°, 0.25°, 0.3°, 0.4°, 0.5°, 0.6°, 0.7°, 0.8°, and 1.0°C)]. Right panels show surface heat fluxes (W m−2) and surface wind [vectors, m s−1; positive (solid black lines) and negative (dashed gray lines) values represent downward and upward heat fluxes]. Contours in right panels represent −30, −20, −15, −12, −9, −7, −5, −3, −2, −1, 1, 3, 5, 10, 20, and 30 W m−2. Values over the shaded areas in left and right panels exceed the 95% and 90% statistical significance level using a t test, respectively. Plotted arrows in right panels are significant at the level of 90% statistical significance.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 7.

Fig. 7.

Fig. 7.

SST (contours, °C), heat flux (contours, W m−2), and surface wind (vectors, m s−1) anomalies in the ensemble-mean spinup experiments. Left panels show SST (solid black lines and dashed gray lines denote positive and negative values respectively; the contours represent −8°, −7°, −6°, −5°, −4°, −3°, −2°, −1°, −0.7°, −0.5°, −0.3°, −0.2°, −0.1°, 0.1°, 0.2°, 0.3°, and 0.5°C). Right panels show surface heat flux [positive values (solid black lines) and negative values (dashed gray lines) represent downward and upward heat flux, respectively] and surface wind. The contours in right panels represent −4, −3, −2, −1, −0.2, 5, 10, 20, 30, and 40 W m−2.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 8.

Fig. 8.

Fig. 8.

Anomalies of ensemble-mean spinup experiments in the Pacific. (a) and (b) Anomalous meridional overturning streamfunction (Sv) for the upper ocean (0–400 m) over years (a) 2–3 and (b) 9–10. (c) Time series of the STC index [averaged over the box in (a)]. (d) Tropical Pacific SST anomalies (averaged over the box in Fig. 2a, °C). Black thick line in (c) and (d) denotes the trend.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 9.

Fig. 9.

Fig. 9.

(top) Annual mean SST (contours, °C) and surface wind (vectors, m s−1) response in (a) PB-ALL and (b) PB-LOW experiments. The contours in both plots stand for −7°, −6°, −5°, −4°, −3°, −2°, −1°, −0.5°, −0.3°, −0.2°, and −0.1°C. All the plotted values have exceeded 95% statistical significance level using a t test. Black boxes represent the position of sponge wall in PB-ALL and PB-LOW experiments. (bottom) Latitude–depth plots of zonally averaged temperature (contours, °C) and salinity (shaded areas, psu) anomalies in (c) PB-ALL and (d) PB-LOW experiments. The contours in both plots stand for −1°, −0.5°, −0.3°, −0.2°, 0.1°, 0.2°, 0.3°, 0.5°, 1°, 2°, 3°, 4°, 5°, 6°, 7°, and 8°C. Note that scale of shaded regions is nonlinear and logarithmic coordinates are adopted here in vertical direction.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 10.

Fig. 10.

Fig. 10.

Time series of SST (°C) anomalies averaged over the North Pacific and the North Atlantic regions (see boxes in Fig. 2a).

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 11.

Fig. 11.

Fig. 11.

Annual mean response in the initial 70 yr. (a) SST (°C) anomalies. The contours stand for −7, −6, −5, −4, −3, −2, −1, −0.5, −0.3, −0.2, and −0.1°C. Values over the shaded area exceed 95% statistical significance level using a t test. (b) and (c) Heat budget analysis (PW) for the upper ocean (0–100 m) water column in the North Pacific and North Atlantic cooling regions (see the two boxes in Fig. 9a). Physical meanings of terms 1–8 as in Fig. 2. (d) and (e) Vertical-averaged velocity anomalies in the upper 100 m for the Northwest Pacific and Atlantic (m s−1). (f) Wind stress curl anomalies over the North Pacific and North Atlantic (10−6 m s−2). Solid black lines and dashed gray lines denote positive and negative values, respectively. The contours represent −0.3, −0.2, −0.1, −0.05, −0.01, 0.01, 0.05, 0.1, 0.2, and 0.3 (×10−6 m s−2).

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 12.

Fig. 12.

Fig. 12.

SST (°C) response in tropical PC experiment. Solid black lines and dashed gray lines denote positive and negative values, respectively. The contours stand for −7°, −6°, −5°, −4°, −3°, −2°, −1°, −0.5°, −0.3°, −0.2°, −0.1°, 0.1°, 0.2°, 0.3°, 0.4°, and 0.5°C, respectively. Values over the shaded area exceed the 95% statistical significance level using a t test. The black box with solid thick margin stands for the PC domain.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 13.

Fig. 13.

Fig. 13.

Evolution of zonally averaged temperature (contours, °C) and salinity (shaded area, psu) anomalies in the initial 100 yr for (left) Pacific and (right) Atlantic, respectively. Solid black lines and dashed gray lines denote positive and negative values, respectively. The contours stand for −5°, −4°, −3°, −2°, −1°, −0.5°, −0.2°, −0.1°, 0.1°, 0.3°, 0.5°, 0.7°, and 1.0°C. Logarithmic coordinates are adopted in vertical direction. Note that the salinity scale is nonlinear.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1

Fig. 14.

Fig. 14.

Fig. 14.

(a) Time–latitude plot of zonally averaged SSS anomalies (psu) over the Atlantic basin for the initial 100 yr. (b) Time series of the AMOC transport (Sv) in CTRL and SOWH. The AMOC index is defined as the averaged transport of the AMOC in the North Atlantic subtropical region (20°–40°N, 1000–2000 m). The solid and dashed curves denote the SOWH and the CTRL experiment, respectively.

Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3634.1