Changes in the global heat transport and eddy-mean flow interaction associated with weaker thermohaline circulation



This study investigates the anomalous patterns of the oceanic and atmospheric heat transport, transient waves and the interaction between transient waves and the mean zonal flow. The evaluation is based on simulations performed with a coupled model forced by an increase of 1 Sverdrup of freshwater flux into the North Atlantic Ocean. It is found that an increase of freshwater flux in the North Atlantic leads to a weakening in the Northern Hemisphere (NH) oceanic heat transport by up to 1 Petawatt (1015) PW but results in an intensification of the Southern Hemisphere (SH) total heat transport. This inter-hemispheric seesaw leads to substantial changes in transient wave activity which is associated with anomalous meridional temperature flux and eddy kinetic energy (EKE). During winter in the NH, weakening of the thermohaline circulation induces an increase in the storm track activity. However, a reduction in storm tracks is found over the extratropical regions of the SH. The correspondence between the anomalies of storm track intensity and Eady growth rate lead to the conclusion that changes in the transient eddy activity are mostly generated by changes in baroclinic conditions. Furthermore, calculations of the E vector show that interaction between transient and mean flow is most pronounced over the North Atlantic where stronger storm tracks enhance the mean westerlies, thus minimising the effect of changes in the oceanic heat transport. Copyright © 2011 Royal Meteorological Society

1. Introduction

The global thermohaline circulation, or Atlantic meridional overturning circulation (MOC), is the transport of ocean water masses caused by differences in the sea water density due to variations in temperature and salinity. However, the atmospheric circulation has the potential to accelerate/decelerate the MOC (Rahmstorf, 1994; Oka et al., 2001; Timmermann et al., 2005a). The MOC is responsible for part of the heat transport to Europe which contributes to relatively milder winters over the western region of this continent when compared to the same latitudes of Canada (Vellinga and Wood, 2002). It should be noted, however, that deviations from zonal symmetry of winter temperatures are essentially caused by the atmospheric circulation interacting with the oceanic mixed layer, as demonstrated by Seager et al. (2002).

Climate changes associated with MOC perturbations have been inferred from paleo-climatic records (e.g. McManus et al., 2004) as well as from model simulations forced by an increase of freshwater into the ocean (e.g. Timmermann et al., 2005a; Vettoretti et al. 2009). These studies indicate that the global anomalous pattern involves both atmospheric and oceanic pathways to communicate Northern Hemisphere (NH) high-latitude events to both the mid-latitudes and the tropics. Moreover, weaker MOC through air–sea interaction mechanisms may result in a southward shift of the Intertropical Convergence Zone (ITCZ) over the Atlantic and Pacific Oceans, and weakened Indian and Asian summer monsoons (e.g. Zhang and Delworth, 2005). It is important to state that changes to atmospheric circulation patterns forced by thermodynamic changes on the tropical ocean surface can produce drastic climatic anomalies in the Southern Hemisphere (SH), as proposed by Pezzi and Cavalcanti (2001). Recent efforts have been made to evaluate the impacts of modified MOC on the atmospheric baroclinic structure. Brayshaw et al. (2009a, 2009b) argued that changes in the NH meridional thermal gradient intensify and elongate storm tracks toward the poles which is primarily linked to changes in near-surface baroclinicity. Jacob et al. (2005) also demonstrated the substantial influence of the MOC on the NH climate. These authors found stronger maritime cold air advection over Europe when the MOC is weakened when compared to the case where there are no MOC changes.

Studies focused on storm tracks and baroclinic activity for the Last Glacial Maximum interval (LGM) (Kageyama et al., 1999; Kageyama and Valdes, 2000; Justino et al., 2005; Li and Battisti, 2008), have led to the conclusion that storm tracks over the North Atlantic shifted eastward during the LGM. This shift has been attributed to the presence of vast ice sheets and distinct patterns of sea surface temperature (SST) and sea ice extension in the NH. Atmospheric dynamic evaluation during the Younger Dryas (YD) demonstrated that the jet stream over the North Atlantic was considerably strengthened, causing enhanced cyclonic activity over the Eurasian continent (Renssen et al., 1996). The YD was characterized by lower SST than today in the North Atlantic, but higher in some regions of the SH polar ocean. These climate conditions, to some extent, are similar to those considered in the present study.

It is quite surprising that despite the lively scientific debate and increased knowledge on the role of MOC on leading climate changes in the NH, its impact on the baroclinic structure of the SH atmosphere has been partly overlooked. In this sense, the present study provides additional insight on the impact of MOC weakening, focusing mainly on heat, salt and Sverdrup transport, baroclinic activity and storm tracks, as well as on the oceanic and atmospheric patterns which involve the interaction between transient waves and mean atmospheric flow. Therefore, the analyses discussed herein are a useful complement to other studies on this subject, such as those by Knutti et al. (2004), Fluckiger et al. (2008), and van Meerbeeck et al. (2009), which have used similar modelling approaches. The primary focus of these previous investigations was not on transient atmospheric waves, baroclinic activity and the eddy-mean flow interaction.

Specifically, Knutti et al. (2004) showed that the amount of freshwater discharge into the North Atlantic Ocean has a direct effect on Southern Ocean temperature. Fluckiger et al. (2008) evaluated the linear response of temperature and precipitation change patterns for different intensifications of the Atlantic MOC. Van Meerbeeck et al. (2009), on the other hand, focused their analyses on the role of boundary conditions on driving climate changes during the Marine Isotope Stage 3 (MIS3). Justino and Machado (2010) investigated the impact of a weaker MOC on global climate, but under the LGM background state which differs from the present study. Van der Meerback et al. (2009) also highlighted that simulations conducted under distinct boundary conditions lead to distinct climate responses.

The paper is organized as follows. Section 2 provides a description of the model used in this study and methodological details employed. Section 3 evaluates oceanic and atmospheric circulation anomalies as a function of MOC weakening. This section also shows the changes caused on the total salt and heat transport for the ocean and atmosphere. Finally, Section 4 presents the concluding remarks.

2. Numerical model and experiments design

The atmospheric component Earth Climate Model De Bilt ECBILT (Opsteegh et al., 1998) of the coupled system LOch-Vecode-Ecbilt-CLio-agIsm Model. LOVECLIM (Driesschaert et al., 2007) is a 3-layer model with a quasi-geostrophic (QG) adiabatic core (Marshall and Molteni, 1993) and a set of physical parameterisations for the hydrological cycle (Held and Suarez, 1978; Opsteegh et al., 1998) with a simplified radiation scheme. It is a spectral model with T21 triangular truncation, corresponding to an approximate resolution of 5.6° in both latitude and longitude. It should be noted that the other components of LOVECLIM, namely the ice sheet model (AGISM), the oceanic carbon and biogeochemical cycle models (LOCH) are turned off. Owing to its formulation, LOVECLIM can be included in the group of climate models referred to as Earth Models of Intermediate Complexity (EMIC).

The coupled ocean sea-ice model (CLIO) (Goosse et al., 1999; Goosse and Fichefet, 1999; Goosse et al., 2003) is based on the primitive equations and employs a free surface for the ocean component and thermodynamic/dynamic assumptions for the sea-ice component. The vertical mixing parameterisation (Goosse et al., 1999) used in this model is a simplification of the Mellor and Yamada 2.5-level turbulence closure scheme (Mellor and Yamada, 1982). Furthermore, the ocean model CLIO includes mixing along isopycnals in order to include meso-scale eddy impacts on transport (Gent and McWilliams, 1990), as well as the flow of dense waters down to topographic features (Campin and Goosse, 1999). The CLIO model's horizontal resolution is 3°. In the oceanic vertical coordinate (z) there are 20 unevenly spaced levels. The individual models are coupled through flux exchanges of momentum, freshwater and heat. Simulations are performed using weak freshwater flux corrections in order to avoid large climate drifts (Marotzke and Stone, 1995).

Some limitations of the EMIC modelling approach pursued here originated from the atmospheric model set-up, which is based on the QG approximation and uses only three atmospheric layers. Strictly speaking, the QG assumption holds for small Rossby numbers (Ro ≪ 1) and for small topographic gradients. Despite this theoretical limitation, the atmospheric response (e.g. tropical SST anomalies) is qualitatively well reproduced by this numerical system, as shown in Justino et al., 2005. However, the amplitude of the response is diminished in comparison with higher-resolution primitive equation models. This is predominantly due to the low resolution of the atmospheric model rather than the QG assumption.

Furthermore, it was found that the presence of the ice-sheet topography leads to an improved atmospheric circulation as well as stationary and transient wave activities with respect to a simulation which neglects orographic forcing, e.g. for LGM conditions (Timmermann et al., 2004; Justino et al., 2005, 2006). As will be shown in the next section, the comparison of our results with those delivered by more complex models leads us to believe that the approach pursued here of increasing the fresh water input in the North Atlantic Ocean is justified insofar as the state of this perturbed climate system is concerned.

The first experiment is the control simulation (CTR). This was run for 500 years where the last 200 years were assumed as the pre-industrial climate. The second simulation, named the freshwater experiment (FW), was performed by adding 1 Sv (1 Sv = 106 m3s−1) of freshwater into the North Atlantic Ocean between 50°N and 70°N during a period of 500 years. This approach indicates that each grid box of the model had a freshwater increase of 0.88 × 10−7 Sv. In order to evaluate the FW simulation, the years from 200 to 400 have been assumed as the mean climate. After the 500-year interval, the simulation was run for an additional 500-year period with no freshwater forcing. The amount of freshwater added in the first 500 years of simulation is sufficient to induce a substantial MOC slowing. For the CTR and FW experiments, present-day values were employed for some parameters such as atmospheric concentration of greenhouse gases, albedo, vegetation, orbital forcing and topographic boundary conditions.

The reason for performing a long numerical simulation is because of the need to reach a quasi-steady state and evaluate the simulated climate, when the variables of interest to be analysed are in supposed equilibrium according to the external forcing. As stated in Danabasoglu et al. (1996) and according to their approach, the solution of the present study is defined as quasi-steady state when the simulated seasonal and annual cycles became cyclic. This means that the analysed variables show few variations between cycles.

As discussed in Pezzi and Souza (2005), the exact spin-up time for a given integration is not easily found because it depends mainly on the timescale characteristics, type and location of the phenomena being studied (Bryan, 1984), and remains a perennial topic of debate among the scientific community. Supposing that the interest is on oceanic equatorial surface fields, this spin-up time can be achieved with a few years of integration. However for mid-latitudes and deep waters, this time can be much longer, reaching decadal to centennial time scales.

To simulate the CTR experiment, the spin-up time was approximately 200 years (Figure 1(a)), while for the FW experiment the spin-up was 100 years (Figure 1(b)). It was also observed for the FW simulation that when 1 Sv is added, the spin-up time after 500 years of failure was about 250 years to regain the equilibrium of pre-industrial levels. Furthermore, Figure 1(a) shows the overall annual average air temperature at 2 m (T 2 m) obtained from the control simulation (15.3 °C), while Figure 1(b) shows the T 2 m annual average from the 1 Sv simulation, between 50°N and 70°N (FW simulation). In the latter case, the freshwater input leads to a drop in local temperature of 5 °C in the region between latitudes 50°N and 70°N. Similar results have been reported in previous studies found in the literature (e.g. Timmermann et al., 2005a; Stouffer et al., 2006; Vettoretti et al., 2009).

Figure 1.

Annual averaged air temperature at 2 m height (° C). (a) CTR experiment (globally averaged), and (b) FW experiment between 50°N and 70°N

3. Results and discussion

3.1. Oceanic changes

3.1.1. Thermohaline circulation

Figure 2 shows the Atlantic Ocean MOC stream function for both CTR and FW simulations. It should be noted that in the present study, the MOC also includes the transport of water masses due to atmospheric flow, the wind-driven part. These periods were selected for analysis because the MOC is at quasi-steady state for both simulations. It can thus be observed in the CTR simulation (Figure 2(a)) that there is a southward water transport ranging from 18 to 12 Sv where maximum values are found in the North Atlantic Deep Water (NADW) formation region. This is observed between 40°N and 60°N at depths ranging from 500 to 2000 meters. The negative values are related to the formation of Antarctic Bottom Water (AABW). Despite the low resolution of LOVECLIM, one should note that it properly simulates the water transport associated with MOC when compared to observations (Talley et al., 2003; Kanzow et al., 2010, 18 Sv), Global Climate Models (GCMs) (Gent, 2001) and other EMICs (Rahmstorf et al., 2005.)

Figure 2.

MOC annual average (Sv) in the Atlantic Ocean. (a) CTR experiment, and (b) FW experiment. This figure is available in colour online at

The MOC in the FW case is shown in Figure 2(b). Note that the addition of 1 Sv of fresh water to the North Atlantic Ocean leads to a total disruption of the NADW formation due to the decreasing density of seawater, resulting in values less than 1 Sv from 40°N. Moreover, there is an increased formation of AABW. Thus, the reduction of NADW transport generates an intensification of AABW transport with an increase of up to 6 Sv between 3500 and 4000 m depth.

3.1.2. Inter-hemispheric seesaw

One of the most important findings in the study of millennial-scale climate events is the out-of-phase climate response of the two hemispheres, the inter-hemispheric seesaw. Several inter-hemispheric seesaw studies propose that the atmospheric response to weaker MOC is related to the equator-to-pole temperature gradient and production of baroclinic eddies (e.g. Kageyama and Valdes, 2000). Therefore, investigations focusing on changes of the polar ocean (e.g. sea ice), as discussed here, may contribute to understanding the atmospheric anomalous pattern linked to the inter-hemispheric seesaw.

Broecker (1998) and Rind et al. (2001) suggested that for rapid climate changes to be initiated there must be a trigger for a sudden ‘switching off’, or a strong decrease in the rate of deep water formation either in the North Atlantic or the Southern Ocean. This must be due to the reduced density difference between subtropics and high-latitude regions (Nilsson et al., 2003). It should be noted that modelling studies revealed the importance of the wind forcing the establishment of large-scale density gradients and hence, the MOC (Oka et al., 2001; Timmermann and Goosse, 2004).

Knutti et al. (2004) provided an in-depth discussion on the mechanisms associated with the inter-hemispheric seesaw characterisation which may be applied to the EMIC approach conducted herein. These authors proposed a mechanism invoking a fast ‘wave-adjustment’. This adjustment in the Atlantic Ocean leads to a southward cross-equatorial oceanic current, which transports heat from north to south during periods of massive ice-sheet discharge. Moreover, it has been argued that a large part of temperature changes which are signalled by the SH is due to this current, and that the remaining part is due to the thermal-seesaw effects and changes on the large-scale MOC.

In the present study, the seesaw is analysed using an FW simulation time series. The sea-ice volume for both hemispheres (Figure 3(a)), the air temperature in Greenland and in the Weddell Sea (Figure 3(b)), and the Atlantic Ocean heat transport at 30°S (Figure 3(c)) are shown. The results indicate a gradual increase in sea-ice volume in the NH (black line) as freshwater is added to the North Atlantic (Figure 3(a)). Conversely, the volume of sea ice in the SH suffers a sharp decline, reaching values close to zero in the period when additional freshwater forcing is applied in the North Atlantic. As a result, the air temperature at 2 m increases up to 10 °C above its mean value in the Weddell Sea, and decreases 10 °C in Greenland as compared to the CTR run (Figure 3(b)). However, it is demonstrated that after the anomalous period, the ice volume in both hemispheres tends to return to its initial volume. These changes to the sea-ice volume, area and thickness, as well as the surface air temperature in the Weddell Sea and the Nordic Seas regions are associated with changes in the Atlantic Ocean northward heat transport at 30°S (Figure 3(c)). Northward heat transport is about 0.5 PW at the beginning of the simulation. The MOC shutdown due to the freshwater input into the North Atlantic Ocean modifies the heat transport which attains negative values of about − 0.5 PW (i.e. southward) from the year 100 to 500 (Figure 3(c)). As might be expected, reduced heat transport into the NH causes an increased sea-ice volume and an SST decrease. Regarding the SH, heat gain leads to sea-ice volume reduction and growing of positive SST anomalies. When the MOC is restored the oceanic heat transport also tends to return to the initial conditions prior to freshwater input in the system.

Figure 3.

Time evolution of FW experiment. (a) Sea-ice volume (103 Km3) in the Northern Hemiphere (black line) and the Southern Hemisphere (red line), (b) surface temperature (° C) in Greenland (red line) and the Weddell Sea (black line) and, (c) northward heat transport in Atlantic Ocean at 30°S (PW), where negative values indicate heat transport toward the south. This figure is available in colour online at

3.1.3. Thermal and haline density flux

In this section, surface density anomalies are analysed in order to identify the thermal and haline contributions to MOC changes. As proposed by Schmitt et al. (1989) and Speer and Tziperman (1992), surface density anomalies can lead to MOC changes. In order to diagnose thermal and haline contributions to the density changes in the FW experiment, their density fluxes (kg−2s−1) are computed. The surface density flux is defined as in Schmitt et al. (1989)

equation image(1)

with the thermal expansion (equation image) and the haline contraction coefficient (equation image). Ft and Fs are the heat and salt fluxes, respectively. In these expressions, CP, ρ(S, T), p, T and S are specific heat, density, pressure, SST and salinity, respectively. Q, E and P represent net heat fluxes, evaporation and precipitation, respectively.

The thermal component of the density fluxes, as well as the total density flux due to thermal and salt fluxes are shown in Figure 4. The two regions of greatest density gain are in the western region of the North Atlantic Ocean (Figure 4(a) and (b)), where cold and dry continental air masses are blown over relatively warmer waters of the Gulf Stream and the North Atlantic Current. The second region of density gain is in the Nordic Seas where the model simulates negative net heat fluxes associated with strong cooling of surface waters. Contribution of the haline density fluxes to the total density is smaller (not shown). The density losses occur around the equatorial region and correspond to areas of heat gain and excess precipitation, where this last situation is clearly linked to the ITCZ activity.

Figure 4.

Time-averaged annual density fluxes in CTR (10−6× kg/m2s−1). (a) thermal contribution in CTR experiment, (b) thermal contribution in FW experiment, (c) thermal anomalous contribution between the FW experiment and CTR, (d) thermal and haline contribution in CTR experiment, (e) thermal and haline contribution in FW experiment, and (f) contribution of thermal and haline anomalies between FW experiment and CTR. This figure is available in colour online at

Regarding the investigation of density flux anomalies between the CTR and FW simulations, it is found that salinity dominates the density fluxes at the ice–water interface, but it does not dominate in open-ocean fluxes at higher latitudes. The sum of thermal and haline flux contributions in the CTR experiment (Figure 4(d)) demonstrates the importance of the thermal contribution to the density flux. The anomalies in the thermal and total fluxes between FW and CTR experiments (Figure 4(c) and (f)) indicate strong thermal flux reductions in the North Atlantic. Negative anomalies of thermal density fluxes result from the weakening of the vertical air–sea temperature gradient and, thus, the heat loss from the ocean to the atmosphere.

This decrease in heat losses to the atmosphere as a consequence inhibits convective mixing in the main sites of NADW formation. Moreover, the role of thermal fluxes on density anomalies is highlighted by the similarities between the thermal and total density flux (Figure 4(c) and (f)). This result was not anticipated since one would expect prominent changes occurring in the haline density flux due to the link between salinity and the freshwater discharge flux.

3.1.4. Sverdrup transport

Intensification of the meridional wind shear stress and the oceanic subtropical gyre associated with a sharpened meridional thermal gradient (e.g. Brayshaw et al., 2009a,b), may lead to advection of saltier water from the subtropics into mid-latitudes. This will further increase the surface salinity, reducing the freshwater forcing effect. Moreover, warmer subtropical water advection may accelerate the atmospheric jet in accordance with the thermal wind assumption. Thus, a complex interaction involving air–sea feedbacks may be identified by evaluating the Sverdrup relation (e.g. Justino and Peltier, 2008).

The Sverdrup transport calculation is a diagnostic method that can be successfully employed on evaluation of polar and subtropical oceanic gyres (Oka et al., 2001; Timmermann and Goosse, 2004). Thus, it may assist in understanding of MOC shutdown and recovery.

Sverdrup transport is defined as equation image where ψ(y) represents the Sverdrup transport, β is the meridional derivative of the Coriolis parameter, ρ is the mean density of seawater and τx is the zonal component of wind stress. The integral is computed from the eastern (xe) to the western (x) boundary in the North Atlantic and North Pacific Oceans using simulated atmospheric wind stress data.

Sverdrup transport found in the CTR experiment (Figure 5(a)) is reasonably well simulated when compared to results based on wind stress from Hellerman and Rosenstein (1983) and computed by Tomczak and Godfrey (2003). However, the front that separates the subtropical and polar gyres is located a little northwards due to the coarse resolution of the atmospheric model employed here.

Figure 5.

Time-averaged annual mean Sverdrup transport (Sv). In the North Atlantic: (a) CTR and (b) anomalies between FW and CTR experiments. In the North Pacific: (c) CTR and (d) anomalies between the FW and CTR experiments. Positive and negative values denote clockwise and anticlockwise circulations, respectively. This figure is available in colour online at

The Sverdrup transport anomalies between FW and CTR sensitivity experiments (Figure 5(a) and (b)) clearly show an overall increase of mass transport in the subtropical gyre (30°N–50°N), but a small reduction in the polar gyre region (60°N–70°N) of the North Atlantic. This Sverdrup transport intensification of about 3–4 Sv is resultant from a more vigorous wind stress simulation over the former region. Wind stress intensification is also indirectly associated with the stronger meridional thermal gradient between high latitudes and the tropics (Figure 6). Furthermore, enhanced wind-driven circulation may induce the advection of saltier equatorial water to mid-latitudes in order to compensate the freshening of water masses at higher latitudes (Figure 11(c)).

Figure 6.

Distribution of the zonally averaged annual mean temperature (° C). (a) CTR experiment, (b) anomalies between the FW and CTR experiments. (c) and (d) are the same as for (a) and (b) panels, but for zonal wind (m/s). This figure is available in colour online at

Similar evaluation for the North Pacific shows a distinct picture (Figure 5(c) and (d)). Reduction of the mean wind stress curl over the western Pacific leads to a spin-down of the subtropical and sub-polar gyres. This favours a surface cooling in the Kuroshio area and a warming in the eastern North Pacific via changes of the meridional heat transport by the subtropical gyre. Furthermore, the cold water supply from the sub-polar area to the eastern North Pacific is decreased, thereby favouring a slight warming in this area.

3.2. Atmospheric changes

3.2.1. Atmospheric circulation

As proposed by Bjerknes (1964), changes of the poleward heat transport, perhaps associated with distinct NADW formation rate as discussed above, lead to an anomalous meridional thermal gradient between equatorial and extra-tropical latitudes. Shaffrey and Sutton (2006) argue that the energy transport in the ocean is associated with MOC fluctuations, which leads to a reduced equator-to-pole surface temperature gradient, consequently resulting in a weakened atmospheric transient energy transport.

Figure 6(a) shows the annual mean zonally averaged air temperatures calculated by the CTR simulation. The expected distribution which is highlighted by the rapid decrease of temperature with height in the troposphere, is in agreement with the atmospheric NCEP Reanalyses dataset. For the sake of brevity, this is not discussed here in detail since it has been previously addressed in other studies in the literature such as those of Opsteegh et al., 1998, Justino et al., 2005, 2006, and Menviel et al., 2008. One may note, however, that higher temperatures are identified in the equatorial region when compared to NCEP Reanalyses, because LOVECLIM does not properly simulate the Pacific cold tongue. This is not a particular characteristic of this specific model and has been previously addressed by other authors (e.g. Stockdale et al., 1993; Pezzi and Richards, 2003).

Comparing changes in air temperature between the FW and CTR runs (Figure 6(b)), it is possible to see the strong cooling over the North Atlantic, with temperature values dropping down to approximately–15 °C. The strong cooling exhibited by the FW experiment over the North Atlantic and adjacent areas is primarily a result of the increasing sea-ice volume (Figure 3(a)) that insulates the atmosphere from the underlying warmer ocean. Furthermore, the reduced specific humidity also plays a key role in cooling the air due to a reduced greenhouse capacity of the dry atmosphere (not shown).

The induced NH cold anomalous pattern extends vertically up to the mid-troposphere (500 hPa) where positive temperature anomalies are predicted to occur (Figure 6(b)). This boreal vertical distribution indicates that weakening of MOC increases atmospheric stability when compared to conditions with the unperturbed MOC. As will be further discussed in this work, the enhanced temperature fluxes associated with intensified transient waves play an important role in high and mid-troposphere warming.

In contrast with temperature response in the NH, the lower troposphere in the SH warms up as shown in Figure 6(b). The tropical region warms slightly, principally due to weaker southeast trade winds (Figure 6(d)) and, consequently, reduced upwelling and Wind-Evaporation-SST (WES) feedback. One should note the well-defined inter-hemispheric seesaw from the high to mid-troposphere characterized by the opposite signal as compared to the anomalous pattern in the lower troposphere.

The model which zonally averaged vertical wind (Figure 6(c)) shows a reasonable degree of similarity when compared to the NCEP reanalyses results (Timmermann and Goosse, 2004). Nevertheless, the simulated jet stream maximum is weaker and poleward shifted by approximately 10° in the NH. This jet weakening is, in part, a consequence of simulated weaker transient and stationary eddies mainly due to model resolution. The zonal wind weakening is also observed in the SH around 45°S. In the tropical region, the atmospheric model simulates an artificially strong zonal component of the trade winds as a result of dynamical simplification adopted in the model. In the extra-tropics, the model can accurately reproduce the shape and position of the westerlies, although their strength is slightly weaker (e.g. Justino et al., 2005, 2006; Menviel et al., 2008).

Figure 6(d) shows the zonally averaged wind anomalies calculated by the CTR and the FW simulations. One may note an intensification of the subtropical jet stream in the NH, while in the SH the jet stream weakens. From the surface to the upper levels at 60°N, the steeper meridional thermal gradient in the FW experiment (Figure 6(b)) induces an intensification of the extratropical jet with a vigorous core at 500 hPa. It should be noted that the stronger vertical wind shear anomalies also occur in the mid-troposphere, which is in agreement with the anomalous temperature pattern. Regarding the SH analyses, the results reveal a weakening of trade winds and the westerly flow at lower atmospheric levels, as well as a deceleration of the subtropical jet around 30°S. However, increased winds are simulated by the FW in the upper troposphere as compared to the CTR simulation.

3.2.2. Baroclinic instability

Large-scale temperature and atmospheric circulation changes also have the potential to modify the statistics of transient waves. These are predominantly generated by baroclinic instabilities from the background flow. The majority of studies focusing on atmospheric changes as part of a glacial background state, which include storm track dynamics and eddy-mean flow interaction indicate an increase in the storminess (e.g. Kageyama et al., 1999; Dong and Valdes, 2000). This has been attributed to enhanced meridional thermal contrast due to the presence of glacial ice sheets. However, Li and Battisti (2008) have found for the LGM (Last Glacial Maximum, 21 000 BP) a weak transient eddy activity in the Atlantic sector compared to the current climate. These conflicting results also reiterate the need for additional evaluations on the impact of enhanced equator–pole temperature contrast in atmospheric transient waves.

In order to investigate changes to atmospheric baroclinicity induced by the anomaly of freshwater forcing, the Eady growth rate (σB1) is computed in the mid-troposphere at 500 hPa level (Figure 7) for December–January–February (DJF) and June–July–August (JJA) mean conditions. This is a simplified measure of the atmospheric baroclinicity that can be employed to quantify the potential for instability and cyclone growth (Hoskins and Valdes, 1990; Paciorek et al., 2002). The Eady growth rate estimates baroclinic instability from the vertical wind shear and the static stability of the atmosphere. It is defined as equation image where f is the Coriolis parameter, N the Brunt-Väisälä frequency, Z the upward vertical coordinate, and V the horizontal wind speed.

Figure 7.

Seasonal averaged Eady growth rate (day 1) based on CTR simulation for DJF (a) and for JJA (b). (c) and (d) are anomalies calculated between FW and CTR experiments for DJF and JJA. This figure is available in colour online at

The CTR simulation suggests that the long ‘tail’ of σB1 over eastern Asia originates from the strong vertical wind shear (Figure 7(a)) rather than from atmospheric stratification. Static stability contributes to σB1 amplitude but does not control its spatial distribution. The comparison of Figure 7(a) in the present study with Figure 1a of Paciorek et al. (2002) reveals that the simulated σB1 is reasonably in good comparison with the NCEP reanalysis. One may note that regions presenting a strong meridional temperature gradient such as over eastern North America and Asia exhibit higher baroclinic activity. In the SH, the Eady growth rate is maximum over the border between the Antarctic continent and the Indian Ocean.

In the CTR run during the NH summer, simulated σB1 shows a higher growth rate over North America and northern Eurasia than during the NH winter. In Eurasia, the larger meridional thermal gradient between the warmer continent and the colder Arctic Ocean enhances the wind shear. However, in the storm tracks region (i.e. western North Atlantic and Pacific), σB1 is extremely reduced. In part, this reduction of synoptic activity can be attributed to weaker meridional temperature contrasts related to the retreat of the sea-ice margin, in particular, in the Labrador Sea and in the northwestern Pacific. A remarkable feature during JJA in the SH is the σB1 split in two parts: one centred at about 35°S and the second along the edge of the Antarctica sea ice. This agrees with the dual westerlies structure during the SH winter as discussed by Trenberth (1991).

Differences between the FW and CTR experiments in DJF and in JJA are shown in Figures 7(c) and (d). In the NH, there is a remarkable intensification (weakening) of baroclinic activity over the Greenland–Iceland–Norway seas (the Arctic Ocean). As expected, the cooling over the North Atlantic and northern Europe induced by enhanced freshwater forcing increases the meridional thermal gradient and the vertical wind shear, which subsequently produces stronger baroclinic activity (higher σB1) over these regions. On the other hand, the general cooling over the central Arctic Ocean associated with increased ice thickness that isolates the atmosphere from the underlying ocean, enhances the atmospheric stability leading to negative Eady growth rate anomalies (Figure 7(c) and (d)).

In the SH, a weaker MOC leads to a decrease in baroclinic activity over the frontier between the Antarctic continent and sea water/ice. These changes are perhaps associated with the reduced thermal contrast along the SH extratropics (40°S–80°S) which further lead to a weaker wind shear. Furthermore, it may be noted that changes in the meridional temperature gradient and an associated modulation of the atmospheric baroclinicity drives anomalous transient eddy fluxes that feed back into the zonal mean circulation, as demonstrated by Hoskins and Valdes (1990). This interaction between transient eddy fluxes and the zonal mean circulation is addressed in the next section.

3.2.3. Storm tracks

Changes in the atmosphere baroclinic structure are closely linked to anomalies in the mean wind, stationary and transient eddies, more commonly known as storm tracks. A storm track is often defined by the regions where there is a maximum variance of geopotential height in the upper and mid-troposphere, arising from disturbances with periods less than approximately one week (e.g. Hoskins and Valdes, 1990; Justino et al., 2005). Moreover, as shown in many studies (e.g. Jeffreys, 1926), transient eddies may feed back to the zonal and time mean-flow circulation because the momentum balance in the atmosphere requires a two-sided interaction between eddies and the mean circulation (e.g. Kuo, 1956; Hoskins et al., 1983).

In order to obtain an accurate estimate of storm track intensity the statistics of the transient waves are investigated. Next, transient eddy activity is analysed in terms of the mid-troposphere eddy heat flux (EHF) EHF = vT and eddy kinetic energy (EKE) equation image. The transient eddies are extracted from daily model output data and have been temporally filtered using a high-pass filter to include only systems with growth and decay within timescales smaller than six days. The analyses presented here are computed at the atmospheric level of 500 hPa.

Eddy heat flux patterns simulated over the seasons extending from DJF and JJA in the CTR (Figure 8(a) and (b)) reproduces observations which are quite satisfactory. An exception is the overestimation over the North Pacific in DJF, as a result of the strong simulated zonal wind component. There are several reasons that may explain the link between the zonal mean wind and the transient eddy activity. A strong jet is often associated with increased baroclinic and barotropic instabilities resulting in an enhanced transient wave activity. Furthermore, Trenberth (1991) showed that stronger jets also advect transient wave activities further to the east.

Figure 8.

Eddy heat flux (K ms-1). (a) CTR experiment in DJF, (b) CTR experiment in JJA, (c) anomalies between FW and CTR experiments in DJF, and (d) anomalies between FW and CTR experiments in JJA. This figure is available in colour online at

The LOVECLIM atmospheric component of the model underestimates storm track activities during the summer season in the respective hemisphere. It may be stressed, moreover, that the simulated storm tracks display a zonal and uniform structure that differs from previous investigation (e.g. Hoskins and Valdes, 1990; Paciorek et al., 2002).

EHF anomalies between the FW and CTR experiments are shown in Figure 8(c) and (d). During the NH winter, weakening of the MOC induces increases in storm track activity. However, the weakening of MOC causes a reduction of EHF activity over the extratropical regions in the SH (Figure 8(c)). It is important to state that positive EHF anomalies, performed for FW and CTR simulations (Figure 8(b) and (d)), indicate southward heat transport. As previously shown, MOC weakening generates warming in the SH and a strong cooling in the NH. This is associated with jet stream deceleration in the SH, while the opposite occurs in the NH (Figure 9).

Figure 9.

Same as Figure 8, but for EKE (m2/s2). This figure is available in colour online at

This oceanic–atmospheric feature involving the storm tracks and MOC is crucial for reorganisation of the climate system in the sense that when the two hemispheres are out of phase, continuous freezing of the NH and warming of the SH is avoided. The importance of heat transport due to transient waves may be highlighted by the seesaw pattern in the high and mid-troposphere, as highlighted by the temperature anomalies between FW and CTR simulation in Figure 6(b). Intensified (weakened) storm track activity as shown by the EHF anomalies at 500 hPa in the NH (SH) is well corresponded with positive (negative) temperature anomalies.

Another useful storm track diagnostic is the EKE. The EKE values and positions obtained in CTR simulation for the periods of DJF and JJA (Figure 9(a) and (b)) are well placed when compared with results found by previous analyses. However, the EKE magnitude presented here is smaller as compared to studies based on higher-resolution models (e.g. Trenberth, 1991; Hall et al., 1994; Dong and Valdes, 2000). A MOC slowdown leads to alteration of the EKE in both hemispheres. In the NH, the EKE is strengthened and shifted northward during the DJF season (Figure 9(c)).

Over the North Atlantic, this reflects the enhancement of the jet stream steady components. This strong jet stream plays an important role in reducing the seasonal zonal and meridional thermal contrast in mid-latitudes by coupling the large oceanic heat reservoir to a much smaller terrestrial heat reservoir (Schneider, 1996). It should be noted, moreover, that the land–sea contrast in North America and Asia is another source for EKE intensification, as proposed by Brayshaw et al. (2009a). For the SH, the MOC weakening leads to an EKE reduction at higher latitudes for both seasons, but especially in JJA (Figure 9(c) and (d)). As shown for the EHF, jet stream weakening in the SH associated with the reduction of the meridional temperature gradient plays a significant role in defining the storm track anomalous pattern.

However, the argument for explaining the impact of the EHF and EKE for defining the troposphere temperature distribution is not straightforward. The storm tracks' anomalous pattern should play a prominent role in coupling extra-tropical to polar regions, serving as an efficient pathway to transport heat between these regions.

3.2.4. Eddy-mean flow interaction

It can be shown that the zonal mean circulation is decelerated/accelerated by the meridional convergence/divergence of the zonally integrated eddy momentum fluxes (e.g. Kuo, 1956; Hoskins et al., 1983). The meridional and zonal wind components of transient eddies are highly correlated which leads to a net meridional transport of zonal momentum in the atmosphere. In turn, a modified mean circulation has the potential to modify the eddy characteristics both in a barotropic and baroclinic framework. Here, an attempt was made to quantify this interaction for the FW experiment.

Hoskins et al. (1983) proposed a diagnostic framework to interpret the relationship between eddies and the mean flow in terms of the E vector, E = (v2u2, uv. It quantifies the shape, propagation and mean flow interaction of transient eddies. Furthermore, it also measures the propagation of meridionally elongated eddies relative to the westerly mean flow (Hall et al., 1994). In regions where E is divergent (convergent), there is a forcing of the mean horizontal circulation consistent with a tendency of the storm tracks to accelerate (decelerate) the westerly mean flow. Figure 10 shows the 500 hPa high-pass E superimposed on the seasonal mean 500 hPa zonal wind. The E pattern in the CTR is predominantly zonal during the winter season, with maximum downstream in the storm tracks.

Figure 10.

(a) DJF zonal wind at 500 hPa (m/s) with 500 hPa E-vector for CTR simulation. (b) JJA zonal wind at 500 hPa (m/s) with 500 hPa E-vector for CTR simulation. (c) and (d) are the same as (a) and (b), but for FW simulation. This figure is available in colour online at

The E vector over the North Atlantic shows a divergence downstream of the jet maximum and this pattern is maintained until converging as it reaches central Europe (Figure 10(a)). Over the North Pacific, the E vector diverges from the middle to the end of the storm track. In general, the E vector during the NH winter does not show a predominant structure in the CTR experiment, but it is divergent near areas of maximum westerly flow, such as over the North Atlantic and the North Pacific (Figure 10(b)). In the SH winter, it is evident that the transient and mean flow act upon each other in the sense that the E vector is always divergent, accelerating the mean flow between 30°S and 60°S. This interaction is highlighted over the Indian Ocean.

During evaluation of the FW simulation in DJF (Figure 10(c)), one may note that the E vector is very similar to that simulated by the CTR but with larger magnitude over the North Pacific. Over the western North Atlantic, the FW experiment exhibits weaker divergence of E and a strengthening in the downstream flank over the eastern Atlantic and Europe, thereby favouring the jet stream intensification over the North Atlantic. According to the relationship between eddy momentum flux divergence and zonal mean flow acceleration, the northward anomalies of E over the North Atlantic are associated with jet stream deceleration and acceleration of the mean flow to the south (Figure 10(c)).

Slowdown of the zonal mean circulation in the SH during the winter season is originated from several contributions (Figure 10(d)). Primarily, the weaker meridional thermal gradient in the southern oceans leads to a more gentle vertical wind shear which decelerates the upper level westerlies. In turn, this leads to reduced baroclinicity and EKE which consequently, is associated with a weaker zonal component of the E vector in the southern oceans (Figure 10(d)).

3.3. Heat and salt transport

Changes in sea surface and atmospheric circulation which includes the mean and transient flows are crucial to understanding climate conditions. Accordingly, it is important to investigate the impact of freshwater forcing on the oceanic and atmospheric heat transport. The total poleward heat transport of the atmosphere–ocean system in each latitudinal band is computed from the difference between net shortwave radiation and outgoing longwave radiative flux at the top of the atmosphere (Miller and Russell, 1989; Peixoto and Oort, 1992; Trenberth and Caron, 2001). It can be expressed as

equation image(2)

Htotal, HATM and HOCE are the total, atmospheric and oceanic heat transports. Where a is the radius of the Earth, φ is the latitude, STOA is the zonally net shortwave radiation, and LTOA is the zonally averaged outgoing longwave radiation. Both fluxes are computed at the top of the atmosphere.

The simulated present-day total heat transport is reasonably well simulated as compared to calculations based upon observed data (Peixoto and Oort, 1992; Trenberth and Caron, 2001). The maximum poleward heat transport appears in both hemispheres around 45°N and 45°S and attains a maximum value of 5.2 PW (Figure 11(a)). By comparing the simulations of CTR and FW (Figure 11(a)), it can be observed that increasing freshwater in the North Atlantic leads to a weakening in the total NH heat transport by up to 1.5 PW. However, there is an intensification of the SH total heat transport of about 0.5 PW.

Figure 11.

(a) Time-averaged total heat transport (atmosphere + ocean) for the CTR and FW experiments. (b) The oceanic heat transport in Atlantic Ocean (PW). (c) Poleward salt transport (Sv) in the Atlantic Ocean. This figure is available in colour online at

The oceanic poleward heat transport in the Atlantic is mainly driven by a MOC cell, with the northward branch flowing warm surface waters in the western boundary current, and a southward flow of cold deep waters in the deep western boundary. This feature is modified in the FW simulation to show a southward oceanic heat flux (Figure 11(b)). For instance, the substantial reduction (increase) of the northward (southward) heat transport is evident in the FW experiment at 30°N. In comparison with CTR simulation, the oceanic heat transport is reduced by approximately 1 PW in the FW simulation on a global basis (Table I).

Table I. Average heat transport at 30°N for the CTR and FW experiments. The atmosphere heat transport is calculated as the difference between total heat transport and the oceanic contribution. Units are in PW
ExperimentsTotal heatGlobal oceanAtl.Indi./Pac.Atmosphere

For the SH, on the other hand, there exists an increase in the oceanic heat transport at 30°S by up to 0.8 PW (Table II). Moreover, it is demonstrated that the South Atlantic plays a leading role in increasing heat transport in the SH (Figure 11(b), Table II). Moreover, the reduction in the heat transport is also predicted to occur in the Pacific Ocean for both hemispheres (Tables I and II). However, the primary contribution to the total NH oceanic heat transport is provided by the Pacific Ocean.

Table II. Average heat transport at 30°S for the CTR and FW experiments. Negative values indicate transport from north to south. Units are in PW
ExperimentsTotal heatGlobal oceanAtl.Indi./Pac.Atmosphere
CTR− 5.430.050.42− 0.42− 5.38
FW− 5.61− 0.77− 0.42− 0.35− 4.07

In evaluation of the atmospheric heat transport, one may note the increase (reduction) in the atmospheric component at 30°N (30°S) by up to 1.2 (1.4) PW. This drop in the atmospheric heat transport corroborates with changes in the storm tracks. It is herein argued, therefore, that transient waves play a significant role on defining the distribution of heat transport primarily via atmospheric changes.

Changes to atmospheric and oceanic circulations also exert a great impact upon the oceanic salt transport due to induced wind-driven circulation anomalies. As compared with the CTR simulation, the FW experiment shows strong northward salt transport in the North Atlantic (Figure 11(c)). The maximum northward salt transported in the CTR experiment attains values of 7 psuSv at 45°N, although it increases up to 10 psuSv in the FW experiment. Furthermore, one may note the reduction in the southward salt transport. These salt transport anomalies oppose the North Atlantic freshening induced by the freshwater forcing. According to Bjerknes's assumption (Bjerknes, 1964), modifications of the meridional thermal gradient in the atmosphere (ocean) should lead to a reorganisation of oceanic (atmospheric) conditions to compensate the reduction in the poleward energy transport.

4. Summary and concluding remarks

Despite the limitations of the modelling approach adopted in this study, such as its lower resolution, the experiments conducted herein allowed for investigation of the complex interactions behind this important seesaw mechanism of the inter-hemispheric gradients. Moreover, the seesaw impacts on storm track activities and their interaction with the zonal mean flow, heat transport and salt transport of both hemispheres have been addressed.

It was found that the MOC weakening induces lower values of SST and air temperature in the NH, whereas in the SH, strong warming is accompanied by an increase in the AABW formation. This contributes to the so-called bipolar seesaw effect. Nevertheless, this anomalous climate pattern (i.e. warmer SH/colder NH) is confined to the low and mid-troposphere. From the high to mid-troposphere, the air temperatures for the FW and the CTR simulations exhibit an opposite feature with a colder SH and warmer NH.

It is demonstrated, moreover, that the MOC slowdown is associated with substantial modification of atmospheric and oceanic heat transport. In the Atlantic Ocean, the northward heat transport decreases by up to 1 PW, causing a large part of the oceanic heat to be transported to the South Atlantic and, consequently, higher near-surface air temperatures.

This atmospheric–oceanic inter-hemispheric seesaw leads to substantial modifications of transient waves and their interaction with the mean zonal flow. The zonal wind shear weakening (strengthening) leads to a reduction (increase) in the baroclinic activity over higher latitudes of the SH (NH). Correspondence between anomalies in storm track intensity and the anomalies of the Eady growth rate lead to the conclusion that changes in the transient eddy activity are primarily generated by changes in the baroclinic conditions further upstream. Furthermore, calculations of the E vector show that interactions between eddies and mean flow, in the FW simulation, are most pronounced over the North Atlantic where stronger storm tracks enhance the mean westerlies.

Finally, it is suggested that experiments be conducted with a more sophisticated coupled model using a higher spatial resolution in order to more precisely quantify the feedback of the wind stress modulation to the NADW. However, the results of this study do demonstrate the potential importance of coupling between the MOC, and that the atmospheric circulation and such effects must be considered in studies of the SH climate system itself and climate changes. This is still a matter for further investigations by means of using more complex numerical models such as fully coupled sea-ice atmosphere GCMs.


We would like to thank the LOVECLIM development team for providing the model source code. Thanks also goes to the two anonymous reviewers in providing helpful comments that have led to an improvement in the presentation of the results. This work was supported by the Conselho Nacional de Pesquisas—CNPq Brazil under Grants 305980/2009-2 and 304887/2009. We are also greateful to the Fundação de Pesquisa do Estado de Minas Gerais (FAPEMIG) for the Grant PPM-00020-11.