Formation and export of deep water in the Labrador and Irminger Seas in a GCM
The influence of changes in the rate of deep water formation in the North Atlantic subpolar gyre on the variability of the transport in the deep western boundary cur-rent is investigated in a realistic hind cast simulation of the North Atlantic during the 1953-2003 period. In the simulation, deep water formation takes place in the Irminger Sea, in the interior of the Labrador Sea and in the Labrador Current. In the Irminger Sea, deep water is formed close to the boundary currents. It is rapidly exported out of the Irminger Sea via an intensified East Greenland Current, and out of the Labrador Sea via increased southeastward transports. The newly formed deep water, which is advected to Flemish Cap in approximately one year, is preceded by fast propagating topographic waves. Deep water formed in the Labrador Sea interior tends to accumulate and recirculate within the basin, with a residence time of few years in the Labrador Sea. Hence, it is only slowly exported northeastward to the Irminger Sea and southeastward to the subtropical North Atlantic, reaching Flemish Cap in 1 to 5 years. As a result, the transport in the deep western boundary current is mostly correlated with convection in the Irminger Sea. Finally, the deep water produced in the Labrador Current is lighter and is rapidly exported out of the La-brador Basin, reaching Flemish Cap in few months. As the production of deep-water along the western periphery of the Labrador Sea is maximum when convection in the interior is minimum, there is some compensation between the deep water formed along the boundary and in the interior of the basin, which reduces the variability of its net transport. These mechanisms which have been suggested from hydrographic and tracer observations, help to understand the variability of the transport in the deep western boundary current at the exit of the subpolar gyre.
The oceanic Meridional Overturning Circulation (MOC) and the associated poleward heat transport contribute substantially to the present energy balance of the Earth (Trenberth and Caron 2001). One fundamental driver of the MOC is the formation of intermediate to deep water masses in the northern North Atlantic resulting from the winter densification of the surface water.
The variability of the MOC can be related to changes in the rate of deep water formation in the northern North Atlantic in simulations with oceanic General Circulation Models (GCMs) (Eden and Willebrand 2001, Bentsen et al. 2004, Mignot and Frankignoul 2005). Coupled climate GCMs suggest that the MOC is highly sensitive to surface salinity perturbations in the regions of formation of intermediate to deep water (e.g., Manabe and Stouﬀer 1995, Stouﬀer et al. 2006): an anomalous input of freshwater will reduce the deep water formation rate, resulting in a weakening of the Atlantic MOC. This will tend to reduce the poleward heat transport and have therefore considerable impact on, in particular, the Atlantic-European climate (e.g., Vellinga and Wood 2002). However, the link between changes in the rate of deep water formation and the variability of the MOC is not clearly established. For instance, Mauritzen and H¨akkinen (1999) suggested that changes in the rate of deep water formation might not aﬀect the MOC but influence the characteristics of the water masses at depth. The role of the wind-driven circulation is also unclear, and Straneo (2005) suggested that the overturning and the poleward heat transport might vary because of changes in the wind-forced circulation even if the rate of deep water formation remains unchanged.
The deep water formed in the North Atlantic is mostly carried southward by the Deep Wes-tern Boundary Current (DWBC), which constitutes the lower limb of the Atlantic MOC. The DWBC originates from the southward slopes of the Greenland-Scotland Ridge, and in the Irmin-ger and Labrador Seas (Dickson and Brown 1994, see the location of these geographical features in Fig. 3.23). Once formed, the DWBC follows the topography along the western boundary of the North Atlantic basin. The densest water masses are formed in the Nordic (Greenland, Ice-land and Norwegian) Seas and in the Arctic Ocean, and overflow through the deepest sills of the Greenland-Scotland Ridge. The lighter water masses are formed in the Labrador Sea (therefore called Labrador Sea Water, LSW) and in the Irminger Sea (Pickart et al. 2003, Bacon et al. 2003). Although the characteristics of the DWBC and its variations on the monthly time scale have been described (e.g. Lazier and Wright 1993, Fischer et al. 2004), its interannual variability remains poorly documented due to the scarceness of tailored and long-term observations, and so does its link with changes in the rate of deep water formation. The repeated observations of Schott et al. (2004) reveal that the DWBC transport in the region of the Grand Banks hardly varied in two surveys 6 years apart, although the convection activity had changed a lot du-ring that period. Downstream of the Grand Banks, the transport in the DWBC seems to be mostly influenced by the Gulf Stream (Bower and Hunt 2000), topographic Rossby waves and recirculation cells (Pickart and Watts 1990).
In order to understand the connection between deep water formation and the transport in the DWBC, it is essential to clarify the time scale and pathways of the export of the newly formed deep water out the subpolar gyre. Straneo et al. (2003) have shown, using an advective-diﬀusive model based on lagrangian drifter data in the Labrador Sea, that the mean interior flow field in the Labrador Sea plays an important role in the export of LSW. The interior circulation in the Labrador Sea consists of large recirculation cells adjacent to the cyclonic boundary currents (Lavender et al. 2000, Fischer et al. 2004), which are mainly forced by the wind (K¨ase et al. 2001, Spall and Pickart 2003). The residence time of LSW in the Labrador Sea is estimated to be 4 to 5 years, which suggests that the Labrador Sea acts as a reservoir for newly formed LSW, while LSW formed in the Irminger Sea seems to be directly exported out of the subpolar gyre by the intense boundary currents (Cuny et al. 2002). The time scale of the LSW export thus depends on whether convection occurred in the Labrador or Irminger Seas.
Deep water formation has also been observed in the boundary current of the Labrador Sea (Pickart et al. 1997, Cuny et al. 2005). The newly formed water masses are lighter than those produced in the interior and are called upper LSW (uLSW), by opposition to the classical LSW (cLSW) formed in the Labrador Sea interior and in the Irminger Sea. Studies based on CFC-11 inventories showed that uLSW is an important constituent of the DWBC (Smethie and Fine 2001), and that it can also be produced in the central Labrador Sea when the formation of cLSW is reduced (Stramma et al. 2004). Indeed, Kieke et al. (2006) found that the lack of cLSW formation in the late 1990’s was compensated by the formation of uLSW.
In this paper we use a hind cast simulation of the circulation in the North Atlantic from a state-of-the-art oceanic GCM to explore the link between the formation and export of deep water in the Labrador and Irminger Seas and the variability of the DWBC transport. The model and the simulation are presented in section 2, and the characteristics of deep water formation are described in section 3. In section 4, we discuss the pathways and time scales of the export of deep water formed in the Irminger and Labrador Seas, and the variability of the DWBC at the exit of the Labrador Sea. The results are summarized and the mechanisms explaining the link between changes in the rate of deep water formation and the variations of the transport in the DWBC are discussed in section 5.
Description of the simulation
The oceanic GCM used in this study is the Nansen Center (Bentsen et al. 2004, Drange et al. 2005) version of the Miami Isopycnic Coordinate Ocean Model MICOM (Bleck et al. 1992). The regional model covers the North Atlantic from 30◦N to 78◦N, and it has a horizontal resolution of about 20 km in the region of interest. In the vertical, the model has 26 density layers. The upper model layer represents a vertically uniform mixed layer (ML), with variable temperature and salinity, and hence density. The 25 layers below the ML have constant potential density (given in table 3.1) but variable thickness and temperature. The simulated layer salinity is then diagnosed based on the simplified equation of state by Friederich and Levitus (1972). Sea ice dynamics and thermodynamics are represented by the model of Harder (1996) and Drange and Simonsen (1996), respectively. Isopycnal diﬀusive velocities for layer interface, momentum and tracer dispersion are respectively 0.02 ms−1, 0.025 ms−1 and 0.015 ms−1, yielding isopycnal diﬀusivities of the order of 103 m2s−1. Diapycnal mixing is parameterized as a function of stratification by Kd = 3 × 10−7N−1 (m2s−1), where N is the Brunt-V¨ais¨al¨ frequency.
At the lateral boundaries, the regional model is relaxed toward a global 40 km version of the same model by means of one way nesting. The global model fields are read once a week and interpolated in time to specify the relaxation boundary conditions for the regional model at each time step. The global model was spun up for 85 years by applying monthly climatological (for 30 years) then daily (from 1948 to 2003) NCEP/NCAR reanalysis fields, using the scheme of Bentsen and Drange (2000). The global simulation at the end of the spin up was interpolated and used as the initial condition for the regional simulation. The regional and the global simulations were then both forced with the daily mean NCEP/NCAR reanalysis fields from 1948 to 2003. Although there was no other restoring than that inherent to the turbulent heat flux formulation in the global model, the ML temperature and salinity fields of the regional model were linearly relaxed towards the monthly mean climatological values of Levitus and Boyer (1994) and Levitus et al. (1994), respectively, as described in Bentsen et al. (2004). The relaxation scheme has been shown to be suﬃciently weak to allow for seasonal to interannual variations in the simulated hydrography in the northern North Atlantic (H´at´un et al. 2005 a, b).
Influence of the boundary conditions
The mean circulation and the hydrography in the global model are reasonably realistic, except that convection in the Labrador Sea is too shallow and convection in the Irminger Sea is overestimated. This seems to be due to too low salinity in the interior of the Labrador Sea and too low temperature in the interior of the Irminger Sea in winter. These biases are transmitted to the regional model, but they are largely compensated by a too cold temperature in the interior of the Labrador Sea, which yields deeper convection in that region, and too low salinity in the interior of the Irminger Sea, which reduces convection there. Hence, horizontal resolution does not seem to influence directly the depth of convection, consistent with Treguier et al. (2005). On the other hand, the circulation of the nested model is improved by the higher resolution, except in the subtropical gyre (see below).
Near the boundaries of the regional model, the variability of temperature and salinity are largely constrained by the global simulation. The latter may not have reached equilibrium in the Nordic Seas, as there is a slow increase of salinity below 2000 m depth. Near the southern limit of the regional model, salinity and temperature tend to decrease in the upper 500 m and increase below, while further east a freshening develops at depth after 1973. Again, this may be in part due to model drift. However, it does not seem to strongly aﬀect the subpolar gyre, where salinity tends to decrease at depth and increase near the surface, which corresponds to the observations (Curry and Mauritzen 2005), as will be discussed elsewhere. It is therefore not expected that the low frequency variability of the subpolar gyre, which is the focus of this study, is degraded by model drift issues.
c) Mean circulation
In this paper, we focus on the circulation in the subpolar North Atlantic (approximately from 47◦N to 67◦N) from January 1953 to December 2003. The surface circulation (Fig. 3.24, a) compares well with the observations (Reverdin et al. 2003). However, the subtropical gyre is too weak and the Gulf Stream too broad, following the 1000 m isobath as far north as Flemish Cap before leaving the coast. Hence, the circulation is not realistic near the Grand Banks, reflecting the relatively low resolution of the model. Another discrepancy is that the North Atlantic Current flows eastward at approximately 48◦N, which is more to the south than in observations.
In the simulation, a direct connection exists between the West Greenland Current and the Labrador Current near 59◦N, which does not clearly appear in observations, although it has been suggested by the propagation of surface drifters (Cuny et al. 2002). The Labrador Current is formed of two branches near the Hamilton Bank. One closely follows the coast as a continuation of the Baﬃn Island Current. The other follows the 1000 m isobath until a recirculation cell forms, creating a northward current at 55◦N, 48◦W that either turns west toward the interior of the Labrador Sea, or east toward the interior of the Irminger Sea (Fig. 3.24, a). This can be seen as a wide recirculation cell which goes from the interior of the Labrador Sea to the interior of the Irminger Sea. Spall and Pickart (2003) have shown that this recirculation cell is mainly influenced by wind-forcing and topography.
The transport of dense water across the Greenland-Scotland Ridge is much weaker than in observations: the total transport is 1.8 Sv for σ0 ≥ 27.86 kgm−3, to be compared to 5.6 Sv for σθ ≥ 27.80 kgm−3 (Dickson and Brown, 1994). As a result, the circulation of dense water coming from the Nordic Seas is not realistic and will not be studied further.
The recirculation from the Labrador Current to the West Greenland Current and the Ir-minger Sea is also present at 2000 m depth (Fig. 3.24, b), which seems realistic as Talley and McCartney (1982) determined that LSW is advected northeastward into the Irminger Sea. In the circulation scheme obtained by Lavender et al. (2000) at 700 m depth, a counter-current flows parallel and opposite to the Labrador Current, the West Greenland Current and the East Greenland Current. Based on Lavender et al. (2000), the northward part of the counter-current in the Labrador Sea is located near 59◦N, 55◦W, while in the simulation it is found near 55◦N, 48◦W. Fischer and Schott (2002), using in situ measurements, also describe a weak anticyclonic recirculation adjacent to the southeastward Labrador Current (their Fig. 4, b).
In the simulation, the DWBC originates in the Irminger Sea, hence does not carry dense water coming from the eastern part of the North Atlantic through the Charlie Gibbs Fracture Zone as in the observations (Dickson and Brown 1994). It flows along the eastern topography of Greenland and along the topography around the Labrador Sea as far as Flemish Cap (Fig. 3.24). The observations show that the deep water masses in the North Atlantic Current diﬀer from those in the DWBC (Schott et al. 2004). Because the Gulf Stream remains too close to the coast in the simulation, the North Atlantic Current only detaches from the coast at the Flemish Cap. Hence most of the DWBC flows northeastward at Flemish Cap, toward the Mid-Atlantic Ridge. On its western flank, part of this current turns southward and forms an intense jet along the topography that eventually turns southwestward. Part of the northeastward current goes through the Charlie Gibbs Fracture Zone and forms an intense jet on the eastern flank of the Mid-Atlantic Ridge. P-ALACE floats have been reported to pass through the Mid-Atlantic Ridge at 700 m (Lavender et al. 2000) and 1500 m (Fischer and Schott 2002, Schott et al. 2004) depth, and the presence of LSW in the eastern North Atlantic has been observed repeatedly (e.g., Cunningham and Haine 1995). The observed circulation (e.g., Paillet et al. 1998) appears less intense than in the simulation, but the southward pathways for subpolar deep water at 30◦W and 20◦W have been suggested from observations (Schott et al. 2004).
The spatial distribution of eddy kinetic energy at the surface corresponds well to the observed one (Reverdin et al. 2003), but its amplitude is about one order of magnitude too small (not shown), likely because of the limited horizontal resolution of 20 km and the lack of vertical shear in the ML, which prevents baroclinic instabilities (e.g., Eldevik 2002). The weak eddy kinetic energy results in too weak turbulent mixing and reduces the influence of the warm and salty subtropical water advected by the boundary currents, on the interior of the Labrador and Irminger Seas. This may explain why the surface isotherms and isohalines are parallel to the boundary currents, and why the interior of the Labrador and Irminger Seas are 1◦ to 2◦C too cold and 0.3 too fresh (Fig. 3.25). The eddy kinetic energy at depth is also one order of magnitude smaller than in observations (not shown). The ice cover in the simulation is restricted to the northern Irminger Sea and to an area along the Labrador coast (Fig. 3.25, b), in general accordance with available observations.
At depth, the winter hydrography shown along WOCE section AR7W (Fig. 3.26 left, see loca-tion in Fig. 3.24) is in general agreement with snapshot observations (e.g. Pickart et al. 2002) and Hydrobase-2 climatology (from http://www.whoi.edu/science/PO/hydrobase, R. Curry, WHOI) except that the temperature is approximately 1.5◦C too cold and the salinity 0.2 too low in the simulation. However, the patterns are similar. Note that the warm and salty water masses carried by the West Greenland Current are hardly visible in the Labrador Current due to mixing with fresh and cold water masses coming from the Baﬃn and Hudson Bays, consistent with obser-vations (e.g. Cuny et al. 2002). Along WOCE section A25 (Fig. 3.26, right), the temperature is again 1.5◦C too cold and the salinity 0.2 below the observed values (Krauss 1995, Hydrobase-2), but the patterns are similar to observations, except at depth presumably due to failures in the representation of the overflow water masses coming from the Nordic Seas.
Although there are biases in the mean temperature and salinity distributions, this may not influence the variability. At the beginning of the simulation, the salinity anomalies along sections AR7W and A25 diﬀer from the observed ones, presumably because the model simulation starts in 1948 (with the availability of atmospheric reanalysis fields). The comparison improves with time and by the late 1960s the salinity anomalies become more realistic. Salinity and temperature slowly decrease at depth since 1973, the anomalies being largest from 1500 m to 3000 m, while they increase near the surface and in the boundary currents, mostly since 1990. Finally, density slightly increases since 1973 from the surface to 2500 m depth in the interior the Labrador and Irminger Seas but slightly decreases in the boundary currents and above the Reykjanes Ridge and below 2500 m depth presumably due to a drift in the simulation.
Table of contents :
1.1 Transport méridien de chaleur et circulation méridienne moyenne dans l’océan
1.2 La MOC dans l’Océan Atlantique Nord
1.3 La MOC et le climat terrestre
1.4 Variabilité interannuelle à multidécennale de la MOC
1.5 Influence de la formation d’eau profonde sur la variabilité de la MOC
1.6 Objet de la thèse
2 Variabilitéde la MOC induite par la formation d’eau profonde
2.1 Mécanismes de variabilité
2.2 Spectral characteristics of the response of the MOC to deep-water formation
2.3 R´esultats et perspectives
3 Formation et exportation d’eau profonde dans l’Atlantique Nord
3.1 Les observations
3.2 Utilisation d’une simulation réaliste de la circulation dans l’Atlantique Nord
3.3 Formation and export of deep water in the Labrador and Irminger Seas in a GCM
3.4 Résultats et perspectives
4 Variabilité de la circulation dans l’Atlantique Nord
4.1 Influence de la variabilité atmosphérique sur la circulation dans l’Atlantique Nord
4.2 Variability of the circulation in the North Atlantic from 1953 to 2003
4.3 Perspective : Variabilité de la gyre subpolaire de 1995 à 2003