Mechanisms for decadal scale variability in a simulated Atlantic meridional overturning circulation
- 1.2k Downloads
Variability in the Atlantic Meridional Overturning Circulation (AMOC) has been analysed using a 600-year pre-industrial control simulation with the Bergen Climate Model. The typical AMOC variability has amplitudes of 1 Sverdrup (1 Sv = 106 m3 s−1) and time scales of 40–70 years. The model is reproducing the observed dense water formation regions and has very realistic ocean transports and water mass distributions. The dense water produced in the Labrador Sea (1/3) and in the Nordic Seas, including the water entrained into the dense overflows across the Greenland-Scotland Ridge (GSR; 2/3), are the sources of North Atlantic Deep Water (NADW) forming the lower limb of the AMOC’s northern overturning. The variability in the Labrador Sea and the Nordic Seas convection is driven by decadal scale air-sea fluxes in the convective region that can be related to opposite phases of the North Atlantic Oscillation. The Labrador Sea convection is directly linked to the variability in AMOC. Linkages between convection and water mass transformation in the Nordic Seas are more indirect. The Scandinavian Pattern, the third mode of atmospheric variability in the North Atlantic, is a driver of the ocean’s poleward heat transport (PHT), the overall constraint on northern water mass transformation. Increased PHT is both associated with an increased water mass exchange across the GSR, and a stronger AMOC.
KeywordsDecadal variability Atlantic Meridional Overturning Circulation Deep water formation Water mass transformations Convection
The Atlantic Meridional Overturning circulation (AMOC) is wind- and density driven with northward flowing surface water in the North Atlantic Current, buoyancy loss and sinking in the north, and southward flowing North Atlantic Deep Water (NADW) at depth. The circulation is closed as NADW is gradually brought to the surface by low latitude diapycnal mixing as well as wind-driven upwelling in the Southern Ocean (Kuhlbrodt et al. 2007). Further below there is a weaker and reversed overturning cell associated with the northward spreading of the denser Antarctic Bottom Water, which gradually mixes with the southward flowing NADW above (e.g., Ganachaud and Wunsch 2000; Johnson 2008).
While surface water is flowing from low to high latitudes, it gradually looses buoyancy due to cooling. The effect is partly compensated by freshening from river runoff, precipitation and ice melt. The lowering of the centre of mass represents a loss in potential energy. Without energy input, the deep ocean would turn into a stagnant pool of dense water within the order of thousand years (Munk and Wunsch 1998). The energy needed to drive the AMOC comes from winds and tides gradually mixing deep and dense water masses with lighter waters above and thus increasing the potential energy (Munk and Wunsch 1998; Wunsch 2002; Gade and Gustafsson 2004; Kuhlbrodt et al. 2007). Using wind and satellite altimetry products (e.g., Egbert and Ray 2000) the spatial variations in the energy input can be estimated, but little is known about how this energy input is varying on decadal and longer time scales, and to what extent and on which time scales such variations can influence the overturning.
Due to its linkages with the northward heat transport and the climate of the North Atlantic region (e.g., Vellinga and Wood 2002; McManus et al. 2004; Rhines et al. 2008), AMOC variability is a key constraint on observed or projected climate change. The majority of the state-of-the-art climate models show a weakening in the AMOC throughout the 21st century when they are forced with increasing atmospheric greenhouse gas concentrations (e.g., Gregory et al. 2005; Meehl et al. 2007; Medhaug and Furevik 2011). In the models warming and freshening in the north reduce the buoyancy loss and weaken the rate of water mass transformation. While the models agree on the sign of the changes they do not agree on the relative importance of the change in heat and freshwater fluxes for the weakening of the AMOC. There are at present no observations indicating whether such a decrease in overturning rate is already taking place (e.g., Cunningham et al. 2007; Cunningham and Marsh 2010). On the contrary, high-resolution modelling (Biastoch et al. 2008) and combined satellite altimetry and in situ observations (Willis 2010) hint to a weak upward trend in overturning circulation during the last decades.
From direct current measurements and water mass properties, Dickson and Brown (1994) estimated that around one-third of the total NADW originates from water spilling over the Greenland-Scotland Ridge, one-third is due to entrainment of ambient water when this water is cascading downwards south of the sill, and the final third originates in the Labrador Sea.
Due to its potential great importance for North Atlantic climate, improved knowledge of the mechanisms for the variability in AMOC will improve the understanding and simulation of present and upcoming climate variability. This includes detection and attribution of anthropogenic climate change, the origins for the discrepancies between models and observations, and the construction of observational schemes for initialising future decadal climate prediction systems.
The aim of this work is to identify mechanisms for the low frequency changes in the Atlantic Meridional Overturning Circulation (AMOC) in the Bergen Climate Model (Otterå et al. 2009, 2010) and compare with what is known from observations. Previously suggested candidates for AMOC variability are oceanic response to aggregated atmospheric white noise forcing at high northern latitudes where dense water is produced (e.g., Dickson et al. 1996; Häkkinen 1999; Delworth and Greatbatch 2000; Eden and Willebrand 2001; Deshayes and Frankignoul 2008), a pure oceanic mode associated with advection of density anomalies (e.g., Delworth et al. 1993; Jungclaus et al. 2005), or a coupled atmosphere–ocean mode which in some models includes sea ice variability (e.g., Holland et al. 2001; Bentsen et al. 2004; Biastoch et al. 2008).
The layout of this paper is as follows. Section 2 contains a brief model description with the formulation of mixed layer and convection dynamics, and the statistical methods used in the analysis. The dense water formation regions are identified and the driving mechanisms for the decadal to multidecadal changes of the AMOC are presented in Sect. 3 and discussed in Sect. 4. Summary and concluding remarks are given in Sect. 5.
2 The coupled Bergen climate model
2.1 Model description
The model output being used for this study is a 600-year long simulation after the spin-up phase of the pre-industrial control climate with the Bergen Climate Model (BCM), a fully coupled atmosphere–ocean-sea ice general circulation model. A general description of the model can be found in Furevik et al. (2003), with more recent updates given in Otterå et al. (2009). Only a brief overview of the model system will be given here.
The model consists of the ocean model MICOM (Miami Isopycnic Coordinate Ocean Model; Bleck et al. 1992) coupled with the atmospheric model ARPEGE/IFS (Action de Recherche Petite Echelle Grande Echelle/Integrate Forecast System; Déqué et al. 1994). A dynamic-thermodynamic sea ice model (GELATO; Salas-Mélia 2002) is included. The model uses no flux correction, and is therefore free to evolve its own climatology. The only constraint is the incoming solar radiation at the top of the atmosphere.
ARPEGE is configured with a spectral truncation of wave number (TL63) on a linear grid. The physical resolution is approximately 2.8° along the Equator and a total of 31 vertical levels are used, ranging from the surface to 0.01 hPa. The horizontal distribution of continental and marine aerosols, aerosols from desert dust and black carbon, are held constant at their respective default values. Concentrations of tropospheric sulphate aerosols and the atmospheric CO2 concentration and held fixed at pre-industrial level, and a solar constant of 1,370 W m−2 is used.
The horizontal ocean grid in MICOM is almost regular with a grid spacing approximately 2.4°latitude × 2.4°longitude. To better resolve tropically confined dynamics, latitudinal grid spacing is gradually decreasing to 0.8° near the Equator. The ocean model consists of 34 isopycnic layers below the non-isopycnic mixed layer. MICOM uses potential density with reference pressure at 2,000 decibar (db) as vertical coordinate (σ2-coordinate), whereas the previous version of BCM (Furevik et al. 2003) used 0 db (σ0-coordinate) as reference pressure. The potential density relative to 2,000 db ranges from σ2 = 30.119 kg m−3 to σ2 = 37.800 kg m−3 in the isopycnic layers. The pressure gradient force is computed as the gradient of the geopotential on pressure surfaces and the geopotential is found by an accurate integration of the hydrostatic equation using in situ density.
2.2 Mixed layer dynamics
On top of the isopycnic layers there is a non-isopycnic mixed layer (ML), providing the connection between the atmosphere and the subsurface water. The density of the ML varies horizontally and temporally. Three mixing processes determine the mass exchange across the interface between the ML and the interior isopycnic layers. These are (1) diapycnal mixing, i.e. mixing across the density interfaces (2) the mass exchange caused by changes in mixed layer depth (MLD) determined by a Kraus-Turner type parameterization, and (3) convective adjustment (Kraus and Turner 1967; Gaspar 1988; Bleck et al. 1992). The diapycnal mixing is generally small compared to the other two processes, but in areas with dense plumes of water flowing down steep bottom slopes, the diapycnal mixing is considerable. To incorporate the shear instability and gravity current mixing a Richardson number dependent diffusivity is included (see Orre et al. 2009).
The rate of which the water is detrained from the mixed layer due to instability is used as an indicator of where the convection in the model occurs, and where dense water is convected to increase the thickness of the intermediate layers.
In the model, more than 80% of the annual convection in the Nordic Seas (including Greenland, Iceland and Norwegian seas) and more than 90% of the annual convection in the Labrador Sea happens during the winter (herein November–April). This period has therefore been used when calculating convection rates and for calculating the atmospheric forcing variables used in the convection analysis in Sect. 3. For all other purposes annual values are used.
Since the focus of this paper is on multidecadal variability, the inter-annual variability has been removed using a centred third order Butterworth low-pass filter with a cut-off period of 5 years. In the correlation/regression analyses the time series have further been high-pass filtered using the same filter but removing time scales longer than 100 years. The latter is to avoid spurious statistical relationships due to model drift or other low-frequency changes.
For significance testing a student t test is used together with Chelton’s (1983) method for estimating the effective number of degrees of freedom by taking into account the cross- and auto-covariance of the two time series. All correlations given in the text are significant above the 99% confidence level.
3.1 Atlantic meridional overturning circulation (AMOC)
In this study the AMOC index is defined as the maximum overturning stream function value in the latitudinal band north of 20°N for each model year (Fig. 3b). The position of the maximum is found to vary between 22 and 45°N. For decadal and longer time scales a latitudinal varying index has been shown to capture the basin-wide spinup/spin-down of the NADW cell, while the physical interpretation of this index on interannual time scales is shown to be potentially problematic (Vellinga and Wu 2004). The mean modelled AMOC index is 18.8 Sv. From observational estimates the mean overturning circulation is in the range of 13–24.3 Sv, estimated from hydrographic observations at 24°N (Ganachaud and Wunsch 2000; Lumpkin and Speer 2003), 26.5°N (Cunningham et al. 2007), and 48°N (Ganachaud 2003), and from estimated NADW formation rates (Smethie and Fine 2001; Talley et al. 2003). In the model, the long-term mean at 26°N and 40°N are 17 Sv and 17.8 Sv, respectively. Hence, the modelled overturning strength is well within the observational range.
There is little observational basis for constraining the longer term variability of the AMOC or the amplitude of fluctuations in general. The power spectrum of the modelled AMOC index shows power resembling a theoretical red noise spectrum. An exception is the increased power at periods around 40–70 years, where the energy at 45 years periodicity is found to be significant above the red noise (Fig. 3c). The time scales are similar to what is found in many other models (e.g., ~50 years in Delworth et al. 1993; 70–80 years in Jungclaus et al. 2005; ~60 years in Zhu and Jungclaus 2008), and also similar to the 65–70 year variability in the North Atlantic surface climate suggested by observations (e.g., Schlesinger and Ramankutty 1994). Similar periods as for the maximum overturning north of 20; 70–80°N are also found when using maximum streamfunction fixed either at 26°N or 48°N, or by using the first principal component of the streamfunction in the North Atlantic.
3.2 Convective mixing regions
A mixture of water from these two gyres is flowing across the Greenland-Scotland Ridge (GSR) into the Nordic Seas, where it circulates in a cyclonic direction, gradually releasing heat to the atmosphere, increases in density and sinks. In the model, the western part of the Nordic Seas cyclonic circulation cell is shifted east compared to observations (Fig. 4). This is most likely due to too zonal westerlies bringing fewer storms into the Nordic Seas (Otterå et al. 2009), resulting in a large Nordic Seas sea ice cover in the model.
The strongest convection occurs over the continental slope southwest of Svalbard, with an average sinking rate of water from the mixed layer to the layers below reaching 30 m day−1 in the winter months (Fig. 4, colour shading). Some open ocean convection also occurs in the eastern part of the Greenland Sea. Although the sinking rate may seem very large, horizontal advection rates responsible for exporting the denser water from the formation region are at least two orders of magnitudes larger. In the northwestern part of the Labrador Sea, the modelled sinking from plume convection reaches 12 m day−1 while smaller values are seen in the Irminger Sea. The bulk part of this convection occurs on the continental shelves or on the shelf slopes.
The corresponding leading EOF of the convection in the Labrador Sea (Fig. 5c) shows that most of the variability in the convection occurs in the northwestern part. In contrast to the Nordic Seas the mode shows a monopole structure. This pattern explains 19% of the variance in the convection time series. The corresponding PC1 for this pattern can be seen in Fig. 5d. The sea ice extent in the Labrador Sea does not reach the main convective region. Decadal scale variability similar to that of the Nordic Seas is also found for PC1 in the Labrador Sea.
The EOF for the full domain (including both the Nordic Seas and the Labrador Sea) is dominated by the variability in the Nordic Seas, due to the larger fluctuations in the Nordic Seas convection (Fig. 5a) compared to that for the Labrador Sea (Fig. 5c). The reason for the larger fluctuations in the Nordic Seas is that the density difference between the isopycnic layers is smaller for the higher densities, making it less energy consuming to detrain water from the mixed layer in the Nordic Seas.
3.3 Atmospheric forcing: heat, freshwater and momentum fluxes in the Nordic Seas and Labrador Sea
In order to investigate where the buoyancy forcing (heat and freshwater fluxes) is increasing the local surface density and contributing to convection, and where the forcing has a stabilizing effect on the water column, regressions between the convection and each of the forcing terms have been calculated (Fig. 7d–f). Regions of positive (negative) regressions show where the forcing has a positive (negative) contribution to the convection. More heat loss, less freshwater input or more wind (i.e., more vertical mixing) generally acts to increase surface density and thus the convection. The results show that the strongest response in convection (i.e., detrainment rate) is due to changes in heat and freshwater fluxes. The freshwater flux is acting to reduce convection in the Fram Strait region, since it is negatively correlated with convection. The reason is that convection driven by more heat loss from more open water implies less sea ice freezing. Here the negative contribution from sea ice melt is counteracting the positive contribution from decreased precipitation. In contrast, in the Norwegian Sea a positive freshwater flux anomaly (more sea ice and evaporation and/or less precipitation) is contributing to convection. In the Labrador Sea all three contributions have a positive effect on the convection.
As similar to the Nordic Seas, the PC1 for convection in the Labrador Sea is found to have maximum correlation with all three atmospheric forcing parameters at zero lag (Fig. 8b). The strongest relation is between the convection and the heat flux, with a correlation of 0.51, followed by the freshwater flux (0.45) and the momentum flux (0.41). The average Labrador Sea atmospheric fluxes (for ice free areas) all have a significant correlation with the positive phase of the NAO index, giving increased heat loss, less freshwater and stronger winds. The correlations between the NAO index and the fluxes are 0.69, 0.42 and 0.31, respectively at zero time lag.
3.4 Volume transport across the Greenland-Scotland Ridge and Newfoundland section
The water crossing the Greenland-Scotland Ridge (GSR) is usually classified into three different water masses (e.g., Hansen and Østerhus 2000; Eldevik et al. 2009); A warm and saline northward flow of Atlantic water (AW) with the Norwegian Atlantic Current (NwAC), a cold and fresh southward flow of polar water (PW) with the East Greenland Current (EGC), and cold and dense southward flow of deep water (DW) with the overflows (Overflowing Deep Water: ODW).
Water mass definitions
Water mass flow
Norwegian Atlantic Current
S > 34.7 psu or T > 8°C
East Greenland Current
S ≤ 34.7 psu & T ≤ 8°C
Overflowing Deep Water
Labrador Sea Water flow
S < 35.3 psu
Entrained Overflowing Deep Water
Sa < 35.3 psu
Average water mass properties
σ2 (kg m−3)
The ODW exits the Nordic Seas through two openings, the Faroe-Shetland Channel and the Denmark Strait, which on average contribute with 3.8 and 1.9 Sv, respectively. The overflows in the two sections are anti-correlated, with a correlation coefficient of −0.47. The variability of the ODW is mainly controlled by the Denmark Strait overflow, where the correlation between the two is 0.82.
Long-term annual mean transports in and out of the Nordic Seas calculated relative to a reference salinity of 34.9 psu and a reference temperature of 0°C
Salt (kt s−1)
Barents Sea opening
The described exchange of the three water masses across the GSR essentially carries the advective convergence of ocean heat in the Nordic Seas and Arctic Ocean. This net poleward heat transport (PHT) is generally understood to scale with the strength of AMOC (Gregory et al. 2005; Kuhlbrodt et al. 2007) and in this model PHT appears as a constraint on AMOC as PHT variability leads AMOC variability by 3 years. Furthermore, as PHT is the net heat given up by the ocean to the atmosphere, it is the overall thermal constraint on northern water mass transformation. Deep convective mixing is a part of this transformation and it thus appears as a lagged response to PHT (but not a controlling factor for AMOC in this model).
The increased volume transport associated with Nordic Seas deep water can be seen in the Newfoundland section (Fig. 10a) as Entrained Overflowing Deep Water (EODW) around two years later (not shown). The water entrained into the ODW downstream of the GSR is defined as the difference between the EODW (8.7 Sv) and the ODW (5.7 Sv) 2 years earlier, where the two years give the time from a signal is found on the ridge until the same signal is found in the Newfoundland section. This gives an average entrainment of 3 Sv. Dickson and Brown (1994) estimated that the dense overflows double in volume transports due to entrainment, which is somewhat more efficient than found in this model. Observational estimates of EODW (8.9 Sv; Schott et al. 2004) are similar to the modelled values.
In addition to the contribution from the Nordic Seas, the convection in the Labrador Sea contributes on average 4.9 Sv to the NADW, similar to observational estimates of 4.5 Sv (Yashayaev 2007). The time series for the volume transports in the Newfoundland section are shown in Fig. 10b. There is quite large variability of both the LSW flow and the EODW. The variability of the EODW is mainly due to the entrainment since the ODW is relatively constant in comparison (see Fig. 9b). On average the flow of NADW (EODW + LSW: gray line; 13.6 Sv) is 5.2 Sv lower than the AMOC (18.8 Sv). If we also include the Mediterranean water, the sum is 17.5 Sv. Additional dense water is found in layer 16. This water can not be attributed to any specific source region since it also consists of re-circulated water, hence is omitted from the further analysis. The AMOC co-vary with the flow of NADW (R = 0.45), where the LSW flow dominates the variability of NADW (R = 0.68), for NADW leading by 1 year.
In the previous section we described the water masses and circulation in the North Atlantic. The main regions of deep convection are found in the Nordic Seas and in the Labrador Sea. The convection in the Labrador Sea has a direct connection with the deep North Atlantic circulation, while the links between the Nordic Seas convection and exchanges across the ridge are more complicated due to the barrier of the GSR. The deep water formed in these two regions make up the bulk part of the North Atlantic Deep Water constituting the lower limb of the AMOC.
Admittedly, course resolution general circulation models may have some generic weaknesses by not being able to resolve the smaller scale features of the circulation, e.g., energetic fronts and ocean eddies. Recent findings indicate that synoptic surface winds and small scale ocean eddies have much more important roles in the circulation than what has been the traditional view, and that the various components of the overturning circulation are varying both spatially and temporally in contrast to what has been the perception from studies of ocean tracers or from coarse resolution climate models (see Lozier 2010, and references therein).
4.1 Water mass transformation in the Nordic Seas
In the Bergen Climate Model convection occurs both in the Nordic Seas and in the Labrador/Irminger seas in contrast to many earlier model studies showing deep convection mainly in the Labrador Sea (e.g., Deshayes et al. 2007; Zhu and Jungclaus 2008). For models that have deep convection in the Nordic Seas, convection is mainly associated with the central Greenland Sea (Bentsen et al. 2004; Dong and Sutton 2005; Jungclaus et al. 2005). The convection in this version of the Bergen Climate Model occurs more towards the eastern rim of the Nordic Seas. This is consistent with the concept first introduced by Mauritzen (1996), where the inflowing Atlantic Water gradually looses buoyancy and sinks as it circulates around the basin. This is due to heat loss (while freshwater forcing partly compensates), as has been shown in several other observational studies (Rudels et al. 1999; Segtnan et al. 2011).
To elucidate the water mass transformation in the Nordic Seas, volume, heat and freshwater budgets, including both the model advection and eddy diffusion of tracers, are calculated for the region (Table 3). The volume transport budget for the Nordic Seas is closed, with estimated transports close to observations (cf., Østerhus et al. 2005; Skagseth et al. 2008). In the following, all heat transports/fluxes are given relative to 0°C. Vertical heat fluxes of 170 TW (1 TW = 1012 W) is lost to the atmosphere within the Nordic Seas. This is in good agreement with recent estimates of 197 TW (Segtnan et al. 2011). A modest storage anomaly term of −0.05 TW is calculated from the total temperature drift of −0.19°C over the entire 600-year period.
Looking at the freshwater budget, all transports are calculated relative to a salinity of 34.9 psu. Based on the vertical fluxes, the net freshwater input from evaporation minus precipitation (E-P), runoff and sea ice melting/freezing is 66 mSv (1 mSv = 103 m3 s−1), which is equivalent to removing salt at a rate of 2.3 kt s−1 (1 kt s−1 = 106 kg s−1). The seeming imbalance of 1 mSv in the freshwater budget between ocean transports and the vertical fluxes is due to round-off errors. The external freshwater forcing does not add any volume to the budget, but is rather a “virtual salt flux” accounted for by adjusting the salinity according to the forcing. The external freshwater input in the model compares favourably with recent observational-based estimates of around 55 mSv (Dickson et al. 2007; Segtnan et al. 2011). Note that the model includes river runoff to the Baltic Sea and the North Sea, which is not included for in observational-based estimates. The change in storage is 0.02 kt s−1 (0.03 mSv of freshwater), calculated from the total salinity increase of 0.06 psu over the entire 600-year period.
From the freshwater and heat budgets, there is an increase in the Nordic Seas density of around 0.07 kg m−3 over the entire period, which is partly due a net heat loss and partly due to a net freshwater export from the Nordic Seas resulting in colder and more saline water. In the model, it is not found that a density difference of this magnitude in the Nordic Seas has any significant consequence for the dynamics. Furthermore, the change in density is not affecting the intensity of the overflow as one should expect (e.g., Curry and Mauritzen 2005). In the model there is a decrease in the steric sea surface height in the Nordic Seas compensating for the change in density (not shown). A similar mechanism has been discussed by Olsen et al. (2008).
Based only on the advective heat and freshwater budgets, the water mass transformation from the warm and saline inflowing AW to the two distinct outflowing water masses, PW and DW can be described. Most of the heat transported across GSR by the NwAC (307 TW, Table 3) is lost to the atmosphere (170 TW) within the Nordic Seas. This can be seen from Fig. 12, where the AW with a temperature of on average 10.5°C (Table 2) transforms into DW and PW with temperatures of 1.4 and 3.3°C, respectively. The modelled heat transported into the Nordic Seas crossing the GSR compares well with the observational estimates of 313 TW from Østerhus et al. (2005). The EGC and ODW remove 23 and 27 TW of heat from the Nordic Seas. The southward positive heat transport of the EGC is not seen in observations, as observed temperatures are lower than in the model, where the PW is −1.5°C and DW 0.5°C (Rudels et al. 1999; Eldevik et al. 2009).
The circulation in the Nordic Seas can be divided into a horizontal estuarine and a vertical overturning part. To a good approximation, there is a volume balance on GSR both in observations (Hansen et al. 2008) and in the model. Hence there is a strong and significant correlation between the strength of the inflow and outflows at zero lag. The correlation coefficients between the NwAC and the EGC, and the ODW are 0.86 and 0.65, respectively. The weaker correlation between the ODW and EGC (0.44) indicates that the volume transports of the two water masses are depending on two factors: firstly, increased inflow will tend to increase transports in both (positive correlation); whilst, secondly, increased heat loss will tend to increase ODW and decrease EGC transports (negative correlation). Indeed, correlation between the EGC and ODW volume transports when the variability of the NwAC is removed through linear regression gives a significant negative correlation of −0.33. The increase in EGC relative to ODW occurs for negative anomalies in the Nordic Seas density as should be expected (not shown).
4.2 Mechanisms controlling AMOC variability on decadal to multidecadal scale
Labrador Sea convection is forced by heat loss and, to a slightly lesser degree, a negative freshwater flux (mainly due to less precipitation) associated with a positive phase of the North Atlantic Oscillation (NAO+) at zero time lag (Fig. 8b). During NAO+ northerly air masses blow over the Labrador area, where it is found to be colder and dryer than normal (Hurrell 1995), giving favourable conditions for deep convective mixing. In addition, an increased Labrador Sea sea-ice extent is also found at this time, contributing additionally to the convection through brine release.
The convection in the Labrador Sea is found to be related to a positive phase of AMOC (Fig. 6). As similar to the Labrador Sea, the heat and freshwater fluxes, forced by the NAO, contributes to the variability in the Nordic Seas convection. Here the increased air-sea fluxes are connected to a negative phase of NAO, when fewer storms are bringing warm and moist air masses into the Nordic Seas (Hurrell 1995; Furevik and Nilsen 2005).
So far the results are in agreement with the traditional view that NAO is responsible for the deep water formation, and hence AMOC (e.g., Dickson et al. 1996; Curry et al. 1998; Häkkinen 1999; Eden and Willebrand 2001; Bentsen et al. 2004; Deshayes and Frankignoul 2008), where in this model there is a correlation of 0.4 between NAO and AMOC. However, there is one important difference in this study: For the Nordic Seas the convection does not determine the water mass exchange across the GRS but is rather a result of it. An increased water mass exchange leads to an increased net poleward heat transport (PHT); the total heat available for northern water mass transformation.
The PHT is a measure of the water mass transformation actually taking place within the Nordic Seas, contributing to the ODW across the GSR. It has a correlation with AMOC of 0.42, where PHT is leading AMOC by 3 years. Thus, the deep convective mixing is not necessarily an ideal measure of the total water mass transformation that is taking place in the Nordic Seas (Fig. 4), since some of the deep convection occurring will end up below the sill depth of the GSR, and therefore not contribute directly to the ODW.
The two convective regions’ influence on AMOC can be further understood through their interaction via the Subpolar Gyre (SPG). In an accompanying paper by Langehaug et al. (2011) a more detailed assessment of North Atlantic/Arctic exchanges including the influence on, and their interaction within the SPG can be found.
5 Summary and conclusions
An increased understanding of the atmospheric and oceanic climate variability is needed for prediction of future climate, where the response to altered air-sea fluxes might play an important role in the Atlantic oceanic heat transport. In this study we have investigated the mechanism for decadal to multidecadal Atlantic Meridional Overturning Circulation (AMOC) variability in a multi-century, pre-industrial control simulation, using the Bergen Climate Model.
The modelled AMOC is found to be within the observed range of Atlantic overturning, and has increased energy in a 40–70 year frequency band. A novelty with this study is that convective mixing is directly diagnosed in the model, opposed to most previous model studies. Deep-water formation is found both in the Labrador Sea and in the Nordic Seas, but the linkages to the AMOC differ substantially.
The water mass exchange, and hence poleward heat transport (PHT) on the Greenland-Scotland Ridge (GSR) is driven by increased northerly winds associated with the Scandinavian Pattern, the third mode of the sea level pressure in the Atlantic sector. The PHT sets the mode of variability of the convection in the Nordic Seas through the sea ice extent. For high PHT the sea ice edge retracts, resulting in more open ocean convection in the Greenland Sea and less in the Norwegian basin. On average most of the Nordic Seas convection occurs in the Norwegian basin, and a reduction in the Norwegian basin convection is concurrent with an overall decrease in the total Nordic Seas deep-water formation.
Air-sea fluxes, related to opposite phases of the North Atlantic Oscillation (NAO), are contributing to the convection in the Labrador Sea and in the Nordic Seas. For a positive phase of NAO cold and dry air masses contribute to favourable conditions for convection, i.e., stronger wind, increased heat loss and less precipitation, in the Labrador Sea. In the Nordic Seas the same effect of the air-sea fluxes are found during a negative phase of NAO, when there are fewer storms bringing warm and moist air masses into the region.
The Nordic Seas contributes with most of the North Atlantic Deep Water originating in the high northern latitudes (two-thirds when entrainment of ambient water downstream the GSR is included), while the rest is a result of deep convection in the Labrador Sea. The variability in the Labrador Sea convection is forced by the local air-sea fluxes related to NAO, where the convection is directly related to the AMOC. The Scandinavian Pattern, is a driver of the ocean’s PHT, the overall thermal constraint on northern water mass transformation. Increased PHT is both associated with an increased water mass exchange across the GSR, and a stronger AMOC.
This work has been supported by the Research Council of Norway through the NorClim (IM and TF) and BIAC (HRL and TE) projects. It also contributes to the DecCen project (TF). This publication is no. A 345 from the Bjerknes Centre for Climate Research. Thanks to OH Otterå for providing the model data, and to OH Otterå, H Drange, PB Rhines, SM Olsen and J Mignot for discussions and comments. Thanks to the two anonymous reviewers for comments helping to improving the manuscript.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
- Cunningham SA, Marsh R (2010) Observing and modeling changes in the Atlantic MOC. Wiley Interdiciplinary Reviews: Climate Change 1:180–191. doi:10.1002/wcc.22
- Furevik T, Nilsen JEØ (2005) Large-Scale Atmospheric Circulation Variability and its Impacts on the Nordic Seas Ocean Climate - a review. In: Drange H, Dokken T, Furevik T, Gerdes R, Berger W (eds) The Nordic Seas: An Integrated Perspective, AGU Monograph 158. American Geophysical Union, Washington DC, pp 105–136Google Scholar
- Gade HG, Gustafsson KE (2004) Application of classical thermodynamic principles to the study of oceanic overturning circulation. Tellus 56A:371–386Google Scholar
- Gregory JM, Dixon KW, Stouffer RJ, Weaver AJ, Driesschaert E, Eby M, Fichefet T, Hasumi H, Hu A, Jungclaus JH, Kamenkovich IV, Levermann A, Montoya M, Murakami S, Nawarath S, Oka A, Sokolow AP, Thorpe RB (2005) A model intercomparison of changes in the Atlantic thermohaline circulation in response to increasing atmospheric CO2 consentration. Geophys Res Lett 32:L12703. doi:10.1029/2005GL023209 CrossRefGoogle Scholar
- Hansen B, Østerhus S, Turrell WR, Jónsson S, Valdimarsson H, Hátún H, Olsen SM (2008) The inflow of Atlantic Water, heat, and salt to the Nordic Seas across the Greenland-Scotland Ridge. In: Dickson RR, Meincke J, Rhines P (eds) Arctic-Subarctic Ocean fluxes, Springer Netherlands, Chap. 1, pp 15–43. doi:10.1007/978-1-4020-6774-7_2
- Kuhlbrodt T, Griesel A, Montoya M, Levermann A, Hofmann M, Rahmstorf S (2007) On the driving processes of the Atlantic Meridional Overturning Circulation. Rev Geophys 45:RG2001. doi:10.1029/2004RG00166
- Langehaug et al (2011) Arctic/Atlantic exchanges via the Subpolar Gyre. J Clim (submitted)Google Scholar
- Meehl GA, Stocker TF, Collins WD, Friedlingstein P, Gaye AT, Gregory JM, Kitoh A, Knutti R, Murphy JM, Noda A, Raper SCB, Watterson IG, Weaver AJ, Zhao ZJ (2007) Global Climate Projections. In: Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, Tignor M, Miller HL (eds) Climate Change 2007: The Physical Science Basis Contribution of Working group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USAGoogle Scholar
- Rhines P, Häkkinen S, Josey SA (2008) Is oceanic heat transport significant in the climate system? In: Dickson RR, Meincke J, Rhines P (eds) Arctic-Subarctic Ocean fluxes, Springer Netherlands, chap 4, pp 87–109. doi:10.1007/978-1-4020-6774-7_5
- Segtnan et al (2011) Heat and freshwater budgets for the Nordic Seas computed from atmospheric reanalysis and ocean observations. J Geophys Res (submitted)Google Scholar
- Skagseth Ø, Furevik T, Ingvaldsen R, Loeng H, Mork KA, Orvik KA, Ozhigin V (2008) Volume and heat transports to the Arctic Ocean via the Norwegian and Barents Seas. In: Dickson RR, Meincke J, Rhines P (ed) Arctic-Subarctic Ocean fluxes, Springer Netherlands, chap 2, pp 45–64. doi:10.1007/978-1-4020-6774-7_3