Mechanisms for the variability of dense water pathways in the Nordic Seas

Mechanisms for the variability of dense water pathways

in the Nordic Seas

Rolf H. Kase,1 Nuno Serra,1 Armin Kohl,1 and Detlef Stammer1

Received 16 May 2008; revised 15 August 2008; accepted 30 October 2008; published 29 January 2009.

[1] Interannual changes in simulated flow fields of the Nordic Seas are analyzed withrespect to their dynamic causes and consequences regarding the flow of dense water fromthe Nordic Seas into the subpolar North Atlantic across the Greenland-Iceland-ScotlandRidge. The simple case of pure density-driven outflow with closed northern boundariesshows that dense water mainly originates in the northern Lofoten Basin and flowssouthward in three branches, namely along the Norwegian continental slope, along theMohn and Jan Mayen Ridges, and a weak current along the east Greenland continentalslope. Adding variable exchange through Fram Strait shows a strengthening of themost western branch and strong recirculations that may reverse the other two branches.For this case, we find in-phase modulation of the Denmark Strait overflow (DSO) by achanging Fram Strait supply and a Faroe-Shetland transport that is in opposite phase.The scaling of this relation provides a potential explanation of recently observed DSOchanges. However, details of the changes in the simulated pathways suggest, in accordwith the size of the prescribed varying Fram Strait supply, basin-wide wind stress curl andlocal convection, which feeds water from different source regions into the outflowpathways, as the primary cause for the upstream flow field reorganizations.

Citation: Kase, R. H., N. Serra, A. Kohl, and D. Stammer (2009), Mechanisms for the variability of dense water pathways in the

Nordic Seas, J. Geophys. Res., 114, C01013, doi:10.1029/2008JC004916.

1. Introduction

[2] The Nordic and Arctic Seas are important elements ofthe global circulation because warm and saline Atlanticsurface waters are there transformed to cold dense watermasses returning to the North Atlantic at depth. In this loop,the Greenland, Iceland and Norwegian (GIN) Seas consti-tute an effective buffer (storage). It is there where trans-formed water masses from the North Atlantic and waterfrom the Arctic Mediterranean are finally mixed to formthose water masses that represent the core of dense waterswhich, once entering the subpolar North Atlantic, becomethe lower limb of the global meridional overturning circu-lation (MOC [e.g., Swift, 1984; Schmitz and McCartney,1993]). Any change in the dense water supply to thesubpolar North Atlantic via the Greenland-Iceland-ScotlandRidge (GISR) system may therefore have a profound impacton the variability of the MOC [e.g., Bacon, 1998; Kohl andStammer, 2009].[3] The GISR consists of two main passages (Figure 1):

the Denmark Strait (DS) and the Faroe Bank Channel(FBC). Water masses exported through either of those twopassages have quite distinct water mass properties andtherefore also distinct source regions. Early reports charac-terize the DS overflow as consisting of Greenland Sea Deep

Water (GSDW) and the FBC overflow as consisting ofNorwegian Sea Deep Water (NSDW [Swift et al., 1980]).However, there seems to be a consensus that, nowadays,returned (and cooled) Atlantic Water (RAW), as well asArctic and Nordic Seas Intermediate Waters, form the coreof the outflowing water masses [Crease, 1967; Borenas andLundberg, 1988; Mauritzen, 1996; Hansen and Osterhus,2000; Borenas et al., 2001]. This suggests that significantshifts in the properties and possible creation mechanisms ofoutflowing water masses have occurred over the last decades.[4] To understand the impact of the GISR overflow water

and its related fluctuations on a changing MOC, it isimportant to consider the total export of the Nordic Seasacross the entire GISR, because transport changes in indi-vidual passages may compensate each other. Moreover, wehave to understand changes in the pathways of the densewaters and variations in their source mechanisms if we wantto understand the cause of GISR overflow changes and theirpredictive skill on the MOC variability. In this context onemay consider the entire Nordic Seas-Subpolar North Atlan-tic system as modulated by a large-scale wind-drivencyclonic gyre surrounding Iceland, that, if modified, canenhance the DS overflow, while simultaneously reducingthe FBC overflow, or vice versa [Biastoch et al., 2003].[5] Macrander et al. [2005] observed for the first time a

reduction of the DS overflow by 20–30% during 4 yearsand argued that during this time frame the FBC overflowwas increasing while the net overflow seemed to be almostunchanged. In the context of Nordic Seas variability, theestablishment of a new additional DS overflow path from

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114, C01013, doi:10.1029/2008JC004916, 2009

1Institut fur Meereskunde, Zentrum fur Meeres- und Klimaforschung,Universitat Hamburg, Hamburg, Germany.

Copyright 2009 by the American Geophysical Union.0148-0227/09/2008JC004916

C01013 1 of 14

the Iceland Sea to DS by Jonsson and Valdimarsson [2004]raises the additional question where this flow has its origin.Is it a result of stronger localized convection in certainbasins or increased export of dense water from the Arcticbasins that produces temporarily enhanced flows on certainpathways? Are decadal/interdecadal changes in the curl ofthe wind stress responsible for the variability? Clearly, toexplain changes in the properties of the GISR overflowwater, we need to understand not only the processeschanging the water properties in the vicinity of the ridge,but also those changing the route of dense waters toward thepathways and in addition also the water mass formationproperties on a basin scale.[6] It has been shown that the hydraulic character of the

outflow through the passages of the GISR theoreticallysmoothes the transport variations [Kase, 2006]. Accordingto this, variability on subdecadal scales are predominantlyseen in local storage variations and less so in the outflow.Considering changes in the overflow properties, one there-fore has to think about water mass formation changesupstream of the passages.[7] This paper is concerned with temporal variability of

dense water pathways in the Nordic Seas and its impact onoverflow transports. Accordingly, our focus will be on the

upstream origin of changes that have been observed andmodeled for the DS by Kohl et al. [2007]. We will show thatcirculation changes occurring within the Nordic Seas,feeding different water masses to the DS and the FBC,are correlated with the basin-wide wind stress curl, theArctic export of dense water through Fram Strait and withchanges in convection regimes in the Nordic Seas.[8] The study is based on and follows from the numerical

simulations reported by Kohl et al. [2007], who investigatedcauses of changes in the DS overflow. The time averagedcirculation of dense water (denser than sq = 27.8 kg m�3) inthe high-resolution Kohl et al. [2007] model is depicted inFigure 2. Large-scale cyclonic gyres are present in theGreenland and Iceland Seas and in the Norwegian andLofoten Basins, the latter featuring an embedded strongquasipermanent anticyclonic eddy [Kohl, 2007]. The mostimportant pathways of dense water upstream of DS andFBC reside along the east Greenland slope and along theMohn and Jan Mayen Ridges. Two other pathways aresignificant, the first along the north Iceland slope and thesecond along the Norwegian slope. In this work we willinvestigate changes in these dense water pathways andexplain them according to forcing agents.

Figure 1. Model domain showing the main topographic and geographic features of the Nordic Seas andsubpolar North Atlantic. The inset shows in detail the Faroe-Bank Channel region and gives animpression of the model resolution.

C01013 KASE ET AL.: NORDIC SEAS DENSE WATER PATHWAYS

2 of 14

C01013

[9] The structure of the remaining paper is as follows.Section 2 will describe the methodology used in our studyand section 3 discusses changes of the dense water circu-lation of the Nordic Seas produced by different forcingmechanisms. Section 4 analyzes transport balances andwater mass conversions. Section 5 finally provides a dis-cussion and concluding remarks.

2. Methodology

[10] Our study is based on the regional model describedin detail by Kohl et al. [2007]. The model uses the MITgcmcode and is implemented for the eastern subpolar NorthAtlantic and Nordic Seas from 51�N to 78�N and from46�W to 17�E (see Figure 1). The realistic bottom topog-raphy is extracted from the Smith and Sandwell [1997] dataset. In the vertical, the model has 30 levels with thicknessvarying from 10 m near the surface to 500 m at the largestdepth. Between 600–900 m (which is the range of thedeepest GISR passages), the vertical resolution is about100 m. The model includes a parameterization for verticalmixing by the KPP scheme of Large et al. [1994] anda dynamic/thermodynamic sea ice model [Zhang andRothrock, 2000].[11] In our study, the above mentioned model is used for

several new experiments but now run in idealized settingswith 1/4 horizontal resolution and with idealized initial and

boundary conditions. Background coefficients of verticaldiffusion and viscosity are 10�6 m2 s�1 and 10�4 m2 s�1,respectively. Horizontally, biharmonic diffusion and vis-cosity represent unresolved eddy mixing. Coefficients ofhorizontal diffusion and viscosity are both 4 � 1011 m4 s�1.[12] The idealized experiments, which are listed in Table 1,

have no surface heat and freshwater forcing and wereperformed to separate the effects of different forcing mech-anisms upon the dense water pathways and overflowvariability in the Nordic Seas. Specifically, we investigatethe role of (1) pure density-driven cross-ridge exchange,(2) reaction to Arctic export, and (3) varying wind stresscurl. The lower resolution is justified since the focus is noton overflow hydraulic effects, but on large-scale upstreamcirculation pathways.[13] The initial and boundary conditions for our experi-

ments are specified as follows. In the first experiment(DAM BREAK), we used an idealized two water massdistribution as specified by Kase and Oschlies [2000]. Inthis case, spatially uniform density values south and north ofthe GISR are associated, respectively, with temperatures of5�C and �1�C at constant salinity of 35. In a secondexperiment (LEVITUS) initial conditions were taken fromthe January state of the WOA2001 climatology, i.e., spatiallyvarying. In LEVITUS WIND the mean wind stress of either1998–1999 or 2001–2002 are additionally applied. Forthese experiments, the lateral boundaries are closed. Inexperiments FRAM DEEP and FRAM SHALLOW, themodel is driven laterally by idealized settings of prescribedclimatological values of temperature, salinity and velocity.Temperature and salinity are from the WOA2001 climatol-ogy and velocity is specified as barotropic inflow andoutflow. Figure 3b (top) shows the resulting cumulativetransports at the northern boundary split into northward andsouthward transports. Figure 3b (bottom), showing theaverage temperature across Fram Strait, motivated theselection of the transport distribution. Clearly visible isthe warm Atlantic water inflow into the Arctic on the righthand side and the cold return southward flow over theGreenland shelf on the western side. In FRAM DEEP andFRAM SHALLOW experiments, the eastern and westernboundaries are closed, while at the southern boundary, aninflow-ouflow of 8 Sv is imposed with temperature andsalinity prescribed (at inflow points) from the WOA2001climatology.

3. Sensitivity of Dense Water Pathways

[14] A central question for our analysis is the cause ofchanges in the characteristics of water masses leaving bothstraits of the GISR. To that end, we will use the abovesensitivity experiments to investigate potential causes forchanges in the dense water pathways bringing different

Figure 2. Time-averaged vertically integrated volumefluxes (m2 s�1) of the dense water (sq � 27.8 kg m�3) inthe Kohl et al. [2007] numerical simulations. Vectors arecolored according to transport magnitude: black, between 5and 15 m2 s�1; red, between 15 and 100 m2 s�1; blue, largerthan 100 m2 s�1. The thick black line corresponds to the350-m isobath.

Table 1. Summary of Model Configurations for the Idealized Experiments

Experiment Northern Boundary Condition Initial Condition Forcing

DAM BREAK Closed Kase and Oschlies [2000] NoneLEVITUS Closed WOA02001 NoneLEVITUS WIND Closed WOA02001 1998–1999, 2000–2001 WindFRAM DEEP Steady deep exchange WOA02001 NoneFRAM SHALLOW Varying shallow exchange WOA02001 None


3 of 14

C01013

water masses toward the sill. We will first look at the resultsof experiment DAM BREAK, addressing a purely density-driven exchange in an idealized closed basin with only twodistinct water masses north and south of the GISR that arerepresentative of the density contrast between the NordicSeas and the subpolar North Atlantic.

3.1. Density-Driven Exchange

[15] The DAM BREAK experiment consisted of a runwhich starts from two homogeneous densities north andsouth of the GISR and was integrated for 10 years. Duringthis period the near surface density front situated initially

Figure 3. (a) Model bottom topography (in meters) for the study area. Dense water pathways arehighlighted by white areas. Specific transport sections shown are discussed in the text: FSW, Fram StraitWest; FS, Fram Strait; FSE, Fram Strait East; DS, Denmark Strait; EGC, East Greenland Current; IS,Iceland Sea; MOHN, Mohn/Jan Mayen Ridge Current; LOF, Lofoten Basin; FSC, Faroe ShetlandChannel. (b, top) Cumulative total transport for two different inflow scenarios with Levitus initialconditions. The deep inflow case (FRAM DEEP, black) has all southward inflow in region FS while thesecond case (FRAM SHALLOW, red) has additional inflow in the shallower western Fram Strait (FSW).Note that there is no net transport through Fram Strait in all the idealized runs. (bottom) Modeltemperature section (in �C) at the northern boundary (78�N).


4 of 14

C01013

over the GISR, separating the Nordic Seas dense water fromthe warm Atlantic water, moves northward. The water massfront, shown in Figure 4a for the initial time as the transitionbetween the shaded and unshaded areas, is captured subse-quently by the 4.0�C and 4.1�C isotherms. After 1 year the

front still resides in the vicinity of the GISR; however, theenhanced inflow from the Atlantic is already indicated by amore poleward position in the eastern part. This shapecontinues during the following years. The fact that the frontmoved northward results from the baroclinicity of the flow,moving light water northward near the surface, while thedeep dense water is flowing, steered by the topographicslopes, in the opposite direction. The front reaches thenorthern wall in 6 years on the Norwegian side (around10�E in longitude).[16] At depth, the colder dense water (specified here by

water colder than 0�C) is propagating southward throughthe channels (Figure 5a). We note that one part of the flowalong the Jan Mayen Ridge ‘‘hits’’ the Iceland slope andcontinues directly toward Denmark Strait while the otherpart joins the Lofoten flow and feeds the Faroe-Shetlandchannel. Apparently, there is some loss of water colder than0�C, seen from the continuously diminishing values flowingthrough both channels. We also note that the largest south-ward transport close to the northern wall agrees with themost extreme poleward position of the 4�C isotherm in theupper layer, suggesting that water mass formation is favoredin that region due to pathways of deep flow (pull of upperlayer water).[17] The volume of warm water that has moved north-

ward in 5 years (when the northern wall is reached) is about0.8 millions of cubic kilometers corresponding to a netnorthward flow of 5 Sv. In more detail, we seem to find thedense water mainly originating in the northern LofotenBasin. It then splits into three branches: (1) one thatcontinues southward along the Mohn Ridge, (2) the secondwhich is stronger and flows along the Norwegian continen-tal slope, and (3) a smaller one in the west flowing towardand further downstream along the east Greenland continen-tal slope.[18] After 10 years, at 69.5�N, the Lofoten branch carries

2.7 Sv, the Mohn/Jan Mayen branch 2.2 Sv and the eastGreenland branch only 0.1 Sv of volume transport (Figure 6),adding up to the 5 Sv previously estimated from the volumeflux of warm water. We note that while individual pathwaysare adjusted after about 5 years, the total GISR overflowneeds at least 10 years to reach some equilibrium for apredefined density level. In fact, there will be no equilibriumwithout production and the exchange will spin downcontinuously.

3.2. Influence of Arctic Export Through Fram Strait

[19] During the second experiment (LEVITUS), a realis-tic and spatially varying initial density distribution based on

Figure 4. The adjustment of the surface front defined bythe isotherms 4.0�C and 4.1�C as function of time for threeexperiments with different initial and/or boundary condi-tions: (a) idealized water masses (DAM BREAK experi-ment), (b) Levitus initial condition with inflow prescribed atthe northern boundary through sections FS (FRAM DEEPexperiment), and (c) same as Figure 4b but with anadditional inflow at section FSW (FRAM SHALLOWexperiment). Shaded area in Figure 4a corresponds to theinitial conditions; in Figures 4b and 4c, the shaded areacorresponds to the respective water mass front in theLevitus data set.

5 of 14

C01013 KASE ET AL.: NORDIC SEAS DENSE WATER PATHWAYS C01013

the Levitus climatology was used instead of the twohomogeneous water masses of the DAM BREAK experi-ment. Results lead to findings comparable to the DAMBREAK experiment, except for a smaller overflow magni-tude caused by the weaker initial density contrast in theupper layers, and are therefore not shown here. However,adding inflow from the Arctic had a significant impact asdescribed next.[20] Current meter measurements obtained in Fram Strait

over a 3-year period [Schauer et al., 2004] established aview of the variability of exchange of mass and heatbetween the Nordic Seas and the Arctic. Over the 3-yearperiod reported, southward transports varied between 12and 13 Sv and northward transports between 9 and 10 Sv.Since the exact value of net throughflow is not differentfrom zero within error bars, we have postulated a balancedflow at Fram Strait and investigated the response to a steadyor varying inflow/outflow of 15 Sv amplitude with a periodof 2 years. Two experiments were conducted (FRAM DEEPand FRAM SHALLOW experiments), with deep and shal-low inflows, respectively (see Figure 3b).[21] The FRAMDEEP experiment is similar to LEVITUS,

but has prescribed steady inflow of dense water from theArctic through Fram Strait (through section FS in Figure 3a)and a mass balanced export of light water in the east (throughsection FSE). The FRAM SHALLOW experiment has inaddition a shallow inflow (in section FSW of Figure 3a),which is compensated again in section FSE. This flow ismodulated by a sinusoidal oscillatory signal that bringsdown the southward flow to zero or to double amplitudewith a 2-year period.[22] Figure 4b shows the time development of the warm

surface front for the FRAM DEEP case. Instead of reachingthe Greenland Sea within 5 years as in Figure 4a, the front isstill south of Jan Mayen. This is due to the blocking by theincrease of deep circulation in the Greenland Sea generatedby the deep inflow in section FS.[23] Figure 4c shows the time development of the warm

surface front for the FRAM SHALLOW case. As can beinferred from Figure 4, the thermal front does not movepoleward, especially on the western side, but stays close toits initial climatological position because warm water isallowed to exit Fram Strait and cold water spreads along theeast Greenland shelf. Since this exchange is periodicallychanging, the frontal response is mainly an east/westundulation close to Iceland and west of Spitsbergen. Years1 and 5, labeled in Figure 4c, correspond to a minimumFram Strait inflow while Year 2 correspond to a maximumphase. The general frontal displacement from the initialposition in the northwest results from our simplification of aclosed Barents Sea entrance.[24] The deep flow from both experiments with inflow

through Fram Strait is shown in Figures 5b and 5c,respectively. In experiment FRAM DEEP (Figure 5b), themean flow of dense water reveals a more pronounced EastGreenland Current than found during the density-drivencases. However, most of the deep inflow is recirculating inthe Greenland Sea basin. Qualitatively, the dense waterbranches are the same, only their strength is different. Thisis not surprising because the topographic steering favorsregions of strong slope currents. Adding the shelf inflow insection FSW (FRAM SHALLOW), markedly leads to a

Figure 5. Instantaneous vertically integrated fluxes of thewater layer with temperatures colder than 0�C after 10 yearsfrom (a) the DAM BREAK, (b) the FRAM DEEP, and(c) the FRAM SHALLOW experiments (in this case, thenorthern inflow is at its maximum phase). Thick vectorscorrespond to transports larger than 20 m2 s�1 in Figure 5aand larger than 10 m2 s�1 in Figures 5b and 5c. Vertical andhorizontal lines correspond to the transport sectionsintroduced in Figure 3.


6 of 14

C01013

strong shallow western boundary current (Figure 5c) that ismodulated periodically and that progresses toward DenmarkStrait.[25] Transports on the individual sections for the FRAM

SHALLOWexperiment are depicted in Figure 7. Noteworthyis the direct in-phase modulation of the DSO by thechanging Fram Strait supply. The FSCO is in oppositephase resulting from the behavior of the Lofoten and Mohnbranches that act to compensate the supply in order tomaintain the zero throughflow condition at the northernboundary. The mean transport through DS is about 3 Sv inthis case and the amplitude of change is 2 Sv. If we scale theinflow according to the observed magnitude by Schauer etal. [2004], which has an amplitude of roughly 1.5 Sv, theresulting DSO change would be 0.5 Sv. This compares wellto an observed variability of amplitude 0.4 Sv at DS[Macrander et al., 2005]. Note that the EGC has a largeramplitude (about 4 Sv) than the total Fram Strait flow, sincepart of it does not exit through DS. Instead, it recirculatesvia the IS branch north of Iceland. If there was no mixing atall, the sum of EGC, MOHN, and LOF would exactly equalFRAM. This is only approximately true during the first fewyears. After 10 years there is a difference of about 1 Sv thatis lost from the dense to lighter water. This is not surprisingsince there is no renewal due to local surface buoyancyfluxes in this idealized setting.

3.3. Influence of Wind on Dense Water PathwaysVariability

[26] The simulated maximum GISR overflow reported inKohl et al. [2007] occurred at a particularly high North

Atlantic Oscillation (NAO) state in 1998–1999 (NAOindex +0.98 and +1.83) and turned to lower values until2001–2002 (NAO index �0.5 and +0.79, see Osborn[2006]). However, other than the former sections wouldsuggest, the prescribed inflow through Fram Strait arelowest during high NAO while the signal of the shallowinflow is less than 0.5 Sv and is thus not able to explain thesimulated transport changes. Due to the northward shift ofthe westward storm tracks over the North Atlantic duringhigh NAO conditions, the strength of the wind stress curlover the Nordic Seas is changing with the NAO index. Thehigh NAO phase 1998–1999 coincides also with observedlarge DSO [Macrander et al., 2005]. The circulation ofdense water in the high resolution model is illustrated inFigure 8, which shows the vertically integrated transport in thelayer with potential density sq larger than 27.879 kg m�3.Strong southward flow directed to the GISR system ispresent in Figure 8 along the east Greenland shelf break andalso along the eastern flank of the Jan Mayen Ridge. Northof 70�N, northward flow can be found west of the MohnRidge and along the Norwegian shelf break. Superimposedis the dense layer topostrophy, T, which is a parameteruseful to diagnose the existence of rim currents (eithercyclonic or anticyclonic). The topostrophy is here defined asT = k � (~vh � rH), where ~vh is the vertically integrated (inour case below sq = 27.879 kg m�3) flow field and rH isthe gradient of the bottom topography (compare definitionwith Holloway et al. [2007]). T is positive for cyclonic flow(i.e., the deep basin is to the left of the flow) and negativevice versa. Figure 8 therefore suggests that during highNAO conditions, the circulation of dense water in the

Figure 6. (top) Dense water volume transports on specific upstream sections (see locations in Figure 3).(bottom) The time dependence of the overflow at Denmark Strait (DSO) through the Faroe-ShetlandChannel (FSCO) and the sum of both for the density-driven exchange DAM BREAK experiment.


7 of 14

C01013

Nordic Seas feeding the GISR overflow is dominated bytwo cyclonic gyres separated by the Mohn and Jan MayenRidges and superimposed to local circulation features thatpresumably are topographically steered. Although thecyclonic circulation (reddish color in the topostrophy)dominates the Nordic Seas, there is a remarkable deviationfrom this cyclonic dominance near the entrance of the twomain GISR passages. This holds near the DS, north ofIceland, and in the Faroe-Shetland Channel, north ofScotland. Both regions are seen as negative (or blue tonecolored) topostrophy.[27] The circulation changes associated with the transition

from high to low NAO are displayed in Figure 9, whichshows the topostrophy difference (2001–2002 minus1998–1999) of the dense water flow for the high-resolutionexperiment. Negative topostrophy difference is clearly vis-ible near the eastern and western margins of the NordicSeas, indicating a reduced East Greenland Current that feedsthe DSO in association with larger transports along theLofoten branch feeding the FBC overflow. The overallstructure is that of a basin wide gyre change which wasalready described by Kohl et al. [2007] in the context of theassociated sea level changes. To investigate further theirhypothesis of a mainly wind driven change and to investi-gate the influence of this change on the dense water path-ways in the Nordic Seas, the experiment LEVITUS WINDwas performed, which adds wind-forcing to the previousLEVITUS experiment. Here two cases were considered,namely the time averaged wind stress forcing from 1998 to1999 and from 2001 to 2002, representing relatively highand low NAO conditions, respectively. We then compare

the flow differences between low and high NAO states inthe LEVITUS WIND experiment with the correspondinghigh resolution model results. Figure 9b displays the top-ostrophy difference (case 2001–2002 forcing minus case1998–1999 forcing) of the dense water flow in the idealizedrun. Note that the absolute value of topostrophy is resolution-dependent due to larger topographic slope in a model withhigher horizontal resolution. Qualitatively, the differenceplots are comparable with one exception: the flow alongthe Jan Mayen Ridge at about 8�W and between 65�N and70�N shows an opposite (positive) sign in the realistic run.This corresponds to an enhanced southward flow duringthe phase of decreased DS overflow in the realisticsimulation. We therefore cannot explain this feature bywind stress changes alone. However, wind stress wasshown to be an active agent in determining large-scalechanges in the flow field.

3.4. Changes on Pathways as a Result of Convection

[28] Finally, convection, mixing and air-sea interactionsin the Nordic Seas have to be considered. Several mecha-nisms participate in the water mass formation processoccurring in the Nordic Seas. Among those are the changesin the inflow properties of water entering from the Atlantic.It was shown by Orvik and Skagseth [2003] that fluctua-tions in the warm Norwegian Atlantic Current (NwAC)correlated well with wind stress changes near 55�N15 months earlier. Fluctuations in the water mass exchangebetween the Nordic Seas and the Arctic Ocean likewiseimpact the water mass formation in the Nordic Seas[Blindheim and Rey, 2004].

Figure 7. As in Figure 6, but for the case with Levitus initial fields and a varying inflow at Fram Straitwith a 2-year period (FRAM SHALLOW experiment).


8 of 14

C01013

[29] According to Gerdes et al. [2003], strong convectionin the Greenland Sea appears to be correlated with negativephases of the NAO. Over the last decades, the NAOsystematically increased from negative values in the sixtiesand early seventies to positive values in the nineties,resulting in a substantial warming of the Arctic and areduced deep convection in the Greenland Sea. Neverthe-less, Dickson et al. [1999] found no indication of a changingstrength of the DS overflow in the Irminger Basin, suggest-ing that substantial inflow of Arctic dense water into theNordic Seas through Fram Strait (FS) must have occurred tocompensate for the reduced water mass formation in theNordic Seas.[30] In contrast, Hansen et al. [2001] reported a decreas-

ing thickness of NSDW near the ocean weather ship M forseveral decades and reported a decrease of the FBC over-flow for almost 5 years from direct current measurements.This finding is not really plausible because the weather shipM location is not by itself an indication of the dense waterreservoir. It has been shown [Borenas et al., 2001;Mauritzen,1996] that the FBC outflow includes North Iceland winterwater that flows along the slope of the Iceland-Faroe Ridgetoward the Shetland Channel. Interestingly, Olsen et al.[2008] do not support the Hansen et al. [2001] results in anew model analysis and by comparisons with current meterobservations.[31] NSDW is regarded as a mixture of GSDW with

Eurasian Basin Deep Water (EBDW) which is exportedvia Fram Strait into the Greenland Sea along the EastGreenland continental slope but also along the GreenlandFracture Zone [Blindheim and Rey, 2004]. This mixtureleaves the Greenland Basin via the Jan Mayen channel andfills the deep Norwegian and Lofoten Basins. One mightspeculate, therefore, that the diminished Greenland Seaconvection is partly responsible for the reduced outflow ofNSDW across the Iceland-Scotland Ridge. On the other

hand, Fahrbach et al. [2001] reported also a significantchange in heat transport through Fram Strait during the1990s.[32] From recent model results, Eldevik et al. (personal

communication) argue, however, that ‘‘the commonly pre-sumed causality between convective mixing in the Green-land Sea and changes in the overflows is neither evident inthe observations nor in the model’’. Here we take a look at theoutcrop region of the density surface sq = 27.879 kg m�3,which is shown in Figure 10a for the strong outflow phasein DS (1998–1999, shaded) and the strong outflow phase inFBC (2001–2002, hatched). The reduced cyclonic vorticityin the Greenland Sea leads to a spin-down of the interiorgyre, thereby reducing the doming of isopycnals toward thesurface. This process leads to less convection as seen inFigure 10 of the outcrop area. The extension of the outcroparea is not only smaller in 2001–2002, but also shifted fromthe central Greenland Sea to the region southwest ofSpitsbergen.[33] A rough estimate of the volume produced by con-

vection within the Nordic Seas can be obtained by deter-mining the summer depth of the isopycnal that defines theupper limit of the dense water (sq = 27.879 kg m�3) andthen multiplying it with the winter (here February) outcroparea. The respective time series is shown in Figure 10b(thick line). The time series shows a significant correlation(r2 = 0.71 with r2 = 0.55 being significant on the 95%significance level) with the annual mean of the DS overflow(dashed line). This is compatible with the view that thetemporal variability of the dense water formed in the NordicSeas may control the overflow through the GISR straits viachanging the interface height upstream of the sills.[34] In our experiment, density variations are mainly

controlled by (or at least correlated with) temperaturevariations and the region of temperatures between �1.8�Cand 3�C is found to be colocated with the outcropping area.

Figure 8. Dense water topostrophy (units, m2 s�1; see text for explanation) for the period of high DSoverflow transport (1998–1999) from the high-resolution model of Kohl et al. [2007]. Dense water (sq �27.879 kg m�3) transport vectors are superimposed.


9 of 14

C01013

Temporal variations of the area defined by the temperaturecriteria (asterisks) are therefore highly correlated (r2 = 0.87)with the outcrop area time series. If the temperature isprominently controlled by local heat fluxes, the heat fluxover the outcrop area should also be a good proxy for thecreation of dense water. Indeed, the time series of thespatially averaged heat flux (Figure 10b, triangles) is foundto be significantly (r2 = 0.62) correlated with our estimate ofdense water production.

4. Volume and Heat Balances in the Nordic Seas

[35] Oliver and Heywood [2003] applied an inversemodel to determine transports of mass, heat and freshwaterin nine different water mass classes from a hydrographic

section across the Nordic Seas. Based on data from summer1999, they estimated the southward volume transport acrossa line from Norway to Greenland to be 6.7 Sv for theirdense water classes 4–9 (potential densities larger than sq =27.879 kg m�3). The transport of the light waters in classes1–3 (sq < 27.879 kg m�3) added up to 5.4 Sv, but allestimates are associated with large uncertainties. Theauthors also estimated a net heat transport, relative to theirsection mean, of 0.2 ± 0.08 PW toward the Arctic. We note,however, that the section was made during the high DSoverflow reported by Macrander et al. [2005] and that it isstill unclear how variable in time such estimates would be.[36] We use results from the sensitivity experiment

FRAM DEEP and the high-resolution experiment to com-pute meridional volume transports within the same den-

Figure 9. (a) Topostrophy difference 2001–2002 minus 1998–1999 (units, m2 s�1) as results from thehigh-resolution model of Kohl et al. [2007]. (b) Topostrophy difference produced by applying the meanwind stress of 2001–2002 and that of 1998–1999 to the Levitus initial condition case (LEVITUSexperiment).


10 of 14

C01013

sity classes as specified by Oliver and Heywood [2003,Figure 11]. The meridional transports shown were integratedfrom the bottom up to three different density levels: sq =27.879 kg m�3, sq = 27.80 kg m�3 and sq = 27.65 kg m�3.Although Oliver and Heywood [2003] did not consider sq =27.80 kg m�3 a water mass boundary in the Nordic Seas, wenevertheless include it here because many authors define theDS overflow as having densities larger than this value.

[37] As can be seen in Figure 11a, the mean southwardtransport between Fram Strait and the DS is 4 Sv in theidealized experiment. South of the latitude of DS, mixingentrains surrounding water and the transport increasesalmost by a factor of 2, but looses its identity furtherdownstream. The layer between sq = 27.879 kg m�3 andthe bottom has a slightly larger transport because water upto sq = 27.80 kg m�3 is also part of the northward flowing

Figure 10. (a) The outcrop region of the density surface sq = 27.879 kg m�3 for the strong outflowphase in Denmark Strait (1998–1999, shaded) and the strong outflow in Faroe Bank Channel (2001–2002, hatched). (b) Time series of (thick line) the preceding summer depth of the sq = 27.879 kg m�3

isopycnic multiplied by the February outcropping area in Figure 10a, (asterisks) the outcropping areadiagnosed from the area of the January SST between �1.8�C and 3�C, (triangles) the February heat fluxintegrated between 70�N and 78�N, and (dashed line) the transport of water denser (sq = 27.8 kg m�3)through Denmark Strait filtered with a 1-year running mean. All curves are normalized by their standarddeviation after removing the mean.


11 of 14

C01013

Atlantic water. This water mass is, however, diluted byentrainment so that at 60�N it is almost absent. The layerdenser than sq = 27.65 kg m�3 is less exported throughFram Strait and has a southward transport of only less than3 Sv. This is due to northward flowing light water reducingthe net transport. The large increase south of 66�N againemphasizes the role of further entrainment.[38] Concerning the results from the eddy-resolving run,

we encounter a main difference to our sensitivity result inthe region north of the sill. In the idealized run no localwater mass production is happening due to missing surfaceheat and freshwater fluxes, so the transports are almostunchanged from Fram Strait to DS. On the other hand, thetransport of water denser than sq = 27.8 kg m�3 in the highresolution run increases by 4 Sv from Fram Strait toDenmark Strait at 66�N. It is noteworthy that all curves forthe three different density limits intersect at 66�N, indicating

a transition of the overturning maximum to lighter watermasses due to entrainment. However, the behavior down-stream is rather different. While the entrainment of waterdenser than sq = 27.65 kg m�3 is 6–8 Sv at differentperiods (3 Sv at sq = 27.80 kg m�3), the water denser thansq = 27.879 kg m�3 disappears completely reaching thelatitude of 60�N.[39] The variability of the water mass divergence can be

analyzed in more detail by looking at three different 2-yearperiods, 1994–1995, 1998–1999 and 2001–2002 for thethree density limits (Figure 11b): sq = 27.789 kg m�3, sq =27.8 kg m�3 and sq = 27.65 kg m�3. All lines closelyintersect at 66�N, the location of the DS sill. We note thatbetween 72�N and 74�N there is only little temporal volumechange for all chosen density classes, an indication of theimportance of processes south of the latitude of Jan Mayen.A remarkable feature is further that stronger flow north ofDS leads to lower transport downstream in the same densityclass, suggesting that mixing is proportional to the strengthof the flow.[40] Figure 12a presents the heat balance between 61�N

and 69�N for the FRAM SHALLOWexperiment during onefull inflow/outflow cycle. In the absence of surface heatfluxes, the heat storage (i.e., the time derivative of the boxintegrated heat content) is balanced by the net heat fluxthrough the northern and southern boundaries. The heatbudget follows from

Hstor þr �~F ¼ D;

where Hstor is the heat storage, D is the diffusion and

r �~F ¼ Fnorth � Fsouth þ Feast � Fwest þ Ftop � Fbot

is the heat flux divergence.[41] In the idealized run, there is no flux through eastern

and western walls and through the top or bottom. As we seein Figure 12a, the balance holds within reasonable errormargins. The meridional distribution of zonally integratedheat flux for the extreme inflow/outflow phases of theFRAM SHALLOW experiment are shown in Figure 12b.The extreme positive and negative deviations from the meannorth of 69�N are rather uniform (0.18 PW and 0.06 PW,respectively). A large divergence occurs southward of 69�Nand is mainly balanced by local storage.[42] The corresponding heat balance between 61�N and

69�N for the case of the high-resolution model can be seenin Figure 13. We show the sum of storage and surface heatflux between 61�N and 69�N as a dashed line and themeridional heat flux at the northern boundary (69�N) as asolid line in Figure 13 (bottom). The heat transport at thesouthern boundary (61�N) is the dashed line in Figure 13(top) and the dotted line represents the sum of storage,surface heat flux and the transport at 69�N. Withoutdiffusion out of the regarded open box, the latter two linesshould perfectly match. This is not the case, but themismatch is on the order of 0.02 PW and therefore inagreement with the diffusive transport shown in Figure 12a.It is interesting to note that on longer timescales the DSoverflow (drawn multiplied by 1/19 as a solid line inFigure 12a, top) is highly correlated with the heat transport

Figure 11. Time-averaged meridional transport integratedfrom the bottom up to three different isopycnics (sq =27.65 kg m�3, sq = 27.8 kg m�3, and sq = 27.879 kg m�3)from (a) the FRAM DEEP experiment and (b) the high-resolution model. In Figure 11b, colors correspond todifferent time periods and line styles to different boundaryisopycnics as follows: solid lines, sq = 27.65 kg m�3;dashed lines, sq = 27.8 kg m�3; long dashed lines, sq =27.879 kg m�3.


12 of 14

C01013

at 61�N. If, in agreement with hydraulic control, the heattransport is seen as a consequence of the DSO, the shownsum has also to be, because of the smallness of the diffusivetransport, regarded as a consequence of DSO. The possibil-ity of a following northward propagation of this heatanomaly is then indicated by multiyear delay. Figure 13indicates a 1.5-year lag of the heat transport at 69�N to thetransport at 61�N which is in agreement with the lagbetween these components of about 14 months from FRAMSHALLOW shown in Figure 12a. This is remarkable andshows a generic response of the system, since the former isdriven by wind stress changes and the latter by changes intransport through Fram Strait.

5. Discussion and Concluding Remarks

[43] We have run a numerical ocean model with severalidealized forcing scenarios to look into the question ofchanges in the transport of dense water upstream of theGISR. It was shown that the density-driven exchange

delivers water along preferred pathways east of the Mohnand Jan Mayen Ridges. This pathway lacks direct confir-mation from current meter measurements, however, it hasbeen predicted based on dynamical reasoning by Orvik[2004]. There it is suggested that the deep counterflowunderneath the Arctic front amounts to about 1 Sv. This is inagreement with our model results. Additional supply fromthe Arctic Ocean through Fram Strait takes different routes,depending on the location of the prescribed inflow to theNordic Seas, enhancing or reducing the transport found inthe unforced case. Noteworthy is this: without convectionand surface momentum stress, the density difference acrossthe GISR drives a baroclinic circulation north of the sillsthat transports warm and saline Norwegian Atlantic Waterinto the Norwegian and Lofoten Basins, replacing thesouthward flowing dense water. The bulk of the overflowpasses primarily through the Faroe Shetland Channel andexits via the Faroe Bank Channel. About 1/3 is carried viathe Iceland Sea and exits through Denmark Strait.[44] Adding inflow to the Greenland Sea at the deep parts

of the East Greenland slope, without affecting the transportson the shelf and shelf break, changes the throughflow onlymarginally but leads to a stronger cyclonic gyre in thecentral Greenland Sea. The situation changes markedly ifinflow at the shallower regions is included. Then, the EastGreenland Current carries a large amount of dense water. Itspart above the DS sill depth flows into the Irminger Sea andits deeper part is topographically steered along the Icelandbranch and follows then back north along the Jan MayenRidge.[45] By adding wind-forcing typical of NAO+ and

NAO� regimes, we find that its effect is similar to theshallow Fram Strait inflow. Comparing the idealized casewith the fully forced run of Kohl et al. [2007], changes indense water production suggest that this difference is theresult of convection being shifted toward the Lofoten Basinduring the early years of 2000. The effect of this enhancedconvection in the Lofoten/Spitsbergen area is to increase thedense transport along the Mohn and Jan Mayen Ridge pathand consequently the westward flow in the Iceland branch.However, this cannot compensate for the larger reduction of

Figure 12. (a) Heat balance between 61�N and 69�Nduring a full cycle of periodic forcing at Fram Strait (FRAMSHALLOW experiment): thin, flux at 69�N; thin dashed,flux at 61�N; thick, storage; crosses and dots, diffusion ateach latitude; squares, residual. (b) The meridional heat fluxas function of latitude for the extreme inflow/outflow phasesof the experiment.

Figure 13. Heat balance between 61�N and 69�N from thehigh-resolution model (see enclosed explanation of eachcurve).


13 of 14

C01013

the EGC which is due to the reduced cyclonicity of the windfield during NAO�.[46] We conclude that the pattern of changes in the dense

water circulation in the Nordic Seas can be explained by acombination of varying inflow from the Arctic, wind stressand heat flux changes superimposed on the basic basin-widedensity-driven circulation. The full model of Kohl et al.[2007] can be partly regarded as a hindcast of the NordicSeas circulation, whereby the part of circulation change dueto changed transports though FRAM Strait is not simulatedbut represented by in-phase changes due to wind stresschanges. Dickson et al. [2008] have discussed the observedfluctuations in the DS overflow [Macrander et al., 2005]and found a general slowdown after 2000 further down-stream in the VEINS Angmassalik current meter array aswell as in the model simulations of Olsen and Schmith[2007]. The increase of the FBC overflow during NAO�phases seems to be present also in direct measurements[Osterhus et al., 2008]. However, all direct long termcurrent meter observations of the overflows span only arelatively short time compared to the multidecadal time-scales seen in the hydrographic data of weather ship M.

[47] Acknowledgments. This study is supported in part through DFGgrant SFB 512 project TP E1 and BMBF Nordatlantik project WP 4.1. N.S.also acknowledges the financial support by the Portuguese Foundation forScience and Technology grant SFRH/BPD/12472/2003. We thank CarmenUlmen for preparing Figure 1. This is a contribution to the ‘‘IntegratedClimate System Analysis and Prediction (CliSAP)’’ effort.

ReferencesBacon, S. (1998), Decadal variability in the outflow from the Nordic Seasto the deep Atlantic Ocean, Nature, 394, 871–874.

Biastoch, A., R. H. Kase, and D. Stammer (2003), The sensitivity of theGreenland-Scotland Ridge overflow to forcing changes, J. Phys. Ocea-nogr., 33, 2307–2319.

Blindheim, J., and F. Rey (2004), Water-mass formation and distribution inthe Nordic Seas during the 1990s, ICES J. Mar. Sci., 61, 846–863.

Borenas, K. M., and P. A. Lundberg (1988), On the deep water flowthrough the Faroe Bank Channel, J. Geophys. Res., 93, 1281–1292.

Borenas, K. M., I. L. Lake, and P. A. Lundberg (2001), On the intermediatewater masses of the Faroe Bank Channel overflow, J. Phys. Oceanogr.,31(7), 1904–1914.

Crease, J. (1967), The flow of Norwegian Sea water through the FaroeBank Channel, Deep-Sea Res., 12, 143–150.

Dickson, R., J. Meincke, I. Vassie, J. Jungclaus, and S. Osterhus (1999),Possible predictability in overflow from the Denmark Strait, Nature, 397,243–246.

Dickson, R., et al. (2008), The overflow flux west of Iceland: Variability,origins and forcing, in Arctic-Subarctic Ocean Fluxes: Defining theRole of Northern Seas in Climate, edited by R. Dickson, J. Meincke,and P. Rhines, pp. 443–474, Springer, New York.

Fahrbach, E., J. Meincke, S. Osterhus, G. Rohardt, U. Schauer, V. Tverberg,and J. Verduin (2001), Direct measurements of volume transports throughFram Strait, Polar Res., 20, 217–224.

Gerdes, R., M. Karcher, F. Kauker, and U. Schauer (2003), Causes anddevelopment of repeated Arctic Ocean warming events, Geophys. Res.Lett., 30(19), 1980, doi:10.1029/2003GL018080.

Hansen, B., and S. Osterhus (2000), North Atlantic-Nordic Seas exchanges,Prog. Oceanogr., 45, 109–208.

Hansen, B., W. R. Turrell, and S. Osterhus (2001), Decreasing overflowfrom the Nordic Seas into the Atlantic Ocean through the Faroe BankChannel since 1950, Nature, 411, 927–930.

Holloway, G., et al. (2007), Water properties and circulation in ArcticOcean models, J. Geophys. Res., 112, C04S03, doi:10.1029/2006JC003642.

Jonsson, S., and H. Valdimarsson (2004), A new path for the DenmarkStrait overflow water from the Iceland Sea to Denmark Strait, Geophys.Res. Lett., 31, L03305, doi:10.1029/2003GL019214.

Kase, R. H. (2006), A Riccati model for Denmark Strait overflow varia-bility, Geophys. Res. Lett., 33, L21S09, doi:10.1029/2006GL025870.

Kase, R. H., and A. Oschlies (2000), Flow through Denmark Strait,J. Geophys. Res., 105, 28,527–28,546.

Kohl, A. (2007), Generation and stability of a quasi-permanent vortex in theLofoten Basin, J. Phys. Oceanogr., 37, 2637–2651.

Kohl, A., and D. Stammer (2009), Variability of the meridional overturningin the North Atlantic from the 50 years (GECCO) state estimation,J. Phys. Oceanogr., 38(9), 1913–1930.

Kohl, A., R. H. Kase, D. Stammer, and N. Serra (2007), Causes of changesin the Denmark Strait overflow, J. Phys. Oceanogr., 37, 1678–1696.

Large, W. G., J. McWilliams, and S. C. Doney (1994), Ocean verticalmixing: A review and a model with a nonlocal boundary layer parame-terization, Rev. Geophys., 32, 363–403.

Macrander, A., U. Send, H. Valdimarsson, S. Jonsson, and R. H. Kase(2005), Interannual changes in the overflow from the Nordic Seas intothe Atlantic Ocean through Denmark Strait, Geophys. Res. Lett., 32,L06606, doi:10.1029/2004GL021463.

Mauritzen, C. (1996), Production of dense overflows feeding the NorthAtlantic across the Greenland-Scotland Ridge: part 1. Evidence for arevised circulation scheme, Deep-Sea Res., 43, 769–806.

Oliver, K. I. C., and K. J. Heywood (2003), Heat and freshwater fluxesthrough the Nordic Seas, J. Phys. Oceanogr., 33, 1009–1026.

Olsen, S. M., and T. Schmith (2007), North Atlantic-Arctic Mediterraneanexchanges in an ensemble hindcast experiment, J. Geophys. Res., 112,C04010, doi:10.1029/2006JC003838.

Olsen, S. M., B. Hansen, D. Quadfasel, and S. Østerhus (2008), Observedand modelled stability of overflow across the Greenland-Scotland Ridge,Nature, 455(7212), 519–522.

Orvik, K. A. (2004), The deepening of the Atlantic water in the LofotenBasin of the Norwegian Sea, demonstrated by using an active reducedgravity model, Geophys. Res. Lett., 31, L01306, doi:10.1029/2003GL018687.

Orvik, K. A., and O. Skagseth (2003), The impact of the wind stress curl inthe North Atlantic on the Atlantic inflow to the Norwegian Sea towardthe Arctic, Geophys. Res. Lett. , 30(17), 1884, doi:10.1029/2003GL017932.

Osborn, T. J. (2006), Recent variations in the winter North Atlantic Oscilla-tion, Weather, 61, 353–355.

Osterhus, S., T. Sherwin, D. Quadfasel, and B. Hansen (2008), The over-flow transport East of Iceland, in Arctic-Subarctic Ocean Fluxes: Definingthe Role of Northern Seas in Climate, edited by R. Dickson, J. Meincke,and P. Rhines, pp. 443–474, Springer, New York.

Schauer, U., E. Fahrbach, S. Osterhus, and G. Rohardt (2004), Arctic warm-ing through the Fram Strait: Oceanic heat transport from 3 years of mea-surements, J. Geophys. Res., 109, C06026, doi:10.1029/2003JC001823.

Schmitz, W. J., and W. S. McCartney (1993), On the North Atlantic circu-lation, Rev. Geophys., 31, 29–49.

Smith, W. H. F., and D. T. Sandwell (1997), Global seafloor topographyfrom satellite altimetry and ship depth soundings, Science, 277, 195–196.

Swift, J. H. (1984), The circulation of the Denmark Strait and Iceland-Scotland overflow waters in the North Atlantic, Deep-Sea Res., 31,1339–1355.

Swift, J. H., K. Aagaard, and S. Malmberg (1980), The contribution of theDenmark Strait overflow to the deep North Atlantic, Deep-Sea Res., 27,29–42.

Zhang, J. L., and D. Rothrock (2000), Modelling Arctic sea ice with anefficient plastic solution, J. Geophys. Res., 105, 3325–3338.

��R. Kase, A. Kohl, N. Serra, and D. Stammer, Institut fur Meereskunde,

Zentrum fur Meeres- und Klimaforschung, Universitat Hamburg, Bundesstr.53, D-20146 Hamburg, Germany. ([email protected]; [email protected]; [email protected]; [email protected])


14 of 14

C01013

Mechanisms for the variability of dense water pathways in the Nordic Seas

Documents