ocean general circulation modelling of the nordic …

OCEAN GENERAL CIRCULATION MODELLING OF THE NORDIC SEAS

Helge Drange, Rüdiger Gerdes, Yongqi Gao, Michael Karcher, Frank Kauker, Mats Bentsen

From The Nordic Seas: An Integrated Perspective H. Drange, T. Dokken, T. Furevik, R. Gerdes and W. Berger (Eds.)

AGU Monograph 158, American Geophysical Union, Washington DC, pp. 199-220.

The official version of the paper is available from AGU (https://www.agu.org/cgi-bin/agubookstore?memb=agu&topic=..GM&book=OSGM1584238)

https://www.agu.org/cgi-bin/agubookstore?memb=agu&topic=..GM&book=OSGM1584238

Drange, Gerdes, Gao, Karcher, Kauker, and Bentsen (2005): Ocean General Circulation Modelling of the Nordic Seas, in The Nordic Seas: An Integrated Perspective (Drange, Dokken, Furevik, Gerdes and Berger, Eds.),

AGU Monograph 158, American Geophysical Union, Washington DC, pp. 199-220. Official version available from AGU.

NumeriSeas an d to assess Atlanti MOC), the cycThe fie t years due to a s, the intrimprov s. Increasibeen emSeas harealistiobserva ns of the o models. Still, m y of the ttempor d the inhentire A

oceans,barocliconseqatmosp al to interanguided

Numrange oof the cNordicthat are e.g. Hatun et calculatime hiThis apcharact o e.g. greatmosp 48 [e.g., Kapproacscale st

OCEAN GENERAL CIRCULATION MODELLING OF THE NORDIC SEAS

Helge Drange1-4, Rüdiger Gerdes5, Yongqi Gao1,2,4, Michael Karcher5,6, Frank Kauker5,6, Mats Bentsen1,2

1Nansen Environmental and Remote Sensing Center, Bergen, Norway

2Bjerknes Centre for Climate Research, University of Bergen, Bergen, Norway 3Geophysical Institute, University of Bergen, Bergen, Norway

4Nansen-Zhu International Research Centre, Beijing, China 5Alfred Wegener Institute for Polar and Marine Research, Bremerhaven, Germany

6O.A.Sys, Ocean Atmosphere Systems GbR, Hamburg, Germany

The complexity of the state-of-the-art Ocean General Circulation Models (OGCMs) has increased and the quality of the model systems has improved considerably over the last decades. The improvement is caused by a variety of factors ranging from improved representation of key physical and dynamical processes, parallel development of at least three classes of OGCM systems, accurate and cost-effective numerical schemes, an unprecedented increase in computational resources, and the availability of synoptic, multi-decadal atmospheric forcing fields. The implications of these improvements are that the current generation of OGCMs can, for the first time, complement available ocean observations and be used to guide forthcoming ocean observation strategies. OGCMs are also extensively used as laboratories for assessing cause-relationships for observed changes in the marine climate system, and to assess how the ocean system may change in response to, for instance, anomalous air-sea fluxes of heat, fresh water and momentum. The Nordic Seas are a particularly challenging region for OGCMs. The challenge is caused by characteristic length scales of only a few to about ten km, a variety of complex and interrelated ocean processes, and extreme air-sea fluxes. In the paper, an overview of the status of the prognostic modelling of the Nordic Seas marine climate system is given. To exemplify the status, output from two widely different, state-of-the-art OGCM systems are used. The paper also addresses processes that are still inadequately described in the current generation of OGCMs, providing guidelines for the future development of model systems particularly tailored for the Nordic Seas region.

1. INTRODUCTION cal modelling of the climate of the Nordic

d the adjacent waters is an important methoand predict the influence of the region on the c Meridional Overturning Circulation (Aling of fresh water, and the ocean productivity. ld has made considerable progress in recen larger number of contributing modelling group

oduction of higher spatial resolution, and ed representation of important processengly realistic atmospheric forcing data have ployed although their reliability in the Nordic

s still to be assessed. The combination of c forcing and the existence of high-quality in situ tions and continuous time series at key locatioceanic circulation have lead to validated any problems remain, caused by the complexitopography, the small and short spatial and al scales of the main processes in the region, anerent coupling between the Nordic Seas and the tlantic and Arctic Oceans. As in all high latitude

baroclinic adjustment processes mediated by nic Rossby waves are very slow. This has uences for the ocean circulation response to heric variability on time scales from seasonnual. The response is more barotropic and more by topography than in mid-latitude oceans. erical ocean-sea ice models address a broad f scientific questions regarding the development irculation and water mass distribution in the Seas. They aid the interpretation of observations by their nature sparse in space and time [

al., this issue]. Usually, results of hindcast tions are used for this purpose where a realistic story of atmospheric forcing data is prescribed. proach can, in principle, be used to assess the eristics of decadal variability modes compared tenhouse gas forced trends. Available heric forcing fields cover the period since 19alnay et al., 1996], meaning that the hindcast h is currently somewhat restricted to decadal-udies.

of 27



Com

providetemporin itselocean-skey im r conducting reliable seasonal to decadal c Anotheexperimexperim with th scenari

No dmodell mesoscprocess [1991],Heburn nt modell The most re [1994], adocean m

To li is heav OGCMatmospThe ob differendegree n compardemongenerat o the Nordic Seas region. In this respresent Seas.

The t construAtlanti uses levels o such a s, wherea of the M(MICOvertica e spin-upresolut the mixed lay el parame ed results planned

The availabl es of how th dy the ma ion 3, a com re

present variabiNordicseasona(SST) f and the also ma quantit e particu modelling of the Nordic Seconclus

2. PRO Num

groupeprognomodell observamodellevoluti space dand an atmospare bui and pro data-assimilation system therefore feeds available observations into the prognostic model so that the model

ughout the integratio [e.g. Bennett, 1992; Evensen, 2003; Stammer et al., 200 tion systemdecadastate isclimate

Progprimitiv comprimomen of state. Tapproxismallerdepth iOnce thstate, itpoint bconsistsea bou stress, h nt the atm . The most co mean clavailab(http://w MWF

bined with observational data, models can a detailed state description of the system and its al development. The state description is valuable f but can also be utilized to initialise coupled ea ice-atmosphere models. This approach is of

portance folimate predictions [e.g., Collins et al., 2005].

r important branch of modelling is numerical entation that acts as a substitute for the physical ents that are usually not feasible (nor desirable)

e climate system. Idealized experiments ando calculations belong to this category. edicated review exists on the sea ice-ocean ing of the Nordic Seas. However, a description ofale sea ice-ocean models and sea ice-ocean studies in general is covered by Häkkinen whereas Hopkins [1991], Stevens [1991] and et al. [1995] introduced and reviewed releva

ing of the Nordic Seas until the mid 1990’s. levant model review is by Mellor and Häkkinen

dressing the development of coupled ice-odels with particular focus on the Arctic Ocean.

mit and focus the scope of the paper, this reviewily based on the output from two state-of-the-arts, both of which are driven by realistic, daily heric forcing fields for the period 1948-2002. jectives are to explicitly illustrate similarities andces between the two models, to address the of realism in the simulated fields based oison with observations, and by that to

strate strengths and weaknesses of the present ion of OGCMs applied t

pect the paper is more than a review as it s the first multi-model comparison for the Nordic

wo model systems used differ substantially byction: The Alfred-Wegner Institute North c/Arctic Ocean-Sea-Ice-Model (NAOSIM)f fixed depth as the vertical co-ordinate and is as

representative of the classical family of OGCMs the Nansen Center model, which is a derivative

iami Isopycnic Coordinate Ocean Model M), uses surfaces of constant density as the l co-ordinate. Also the applied model domain, th procedure, the horizontal and vertical grid

ion, the details of the atmosphere forcing, er dynamics, and several of the mod

terisations differ, implying that the presentshould not be interpreted as the output from a model intercomparison project.

review starts with a brief overview of the e model systems and gives some exampl

e various model systems have been used to sturine climate system of the Nordic Seas. In Sect

parison between MICOM and NAOSIM a

ed. For this comparison, the mean state and thelity of the volume fluxes into and out of the Seas, the horizontal circulation pattern, the l sea surface salinity (SSS) and temperature ields, and the thickness of the upper mixed layersea ice extent are considered. Comparisons are de to observed values of several of these

ies. The review continues with a discussion of thlar challenges for the ocean

as (Section 4), and is ended by discussion and ion sections.

GNOSTIC OCEAN MODELLING

erical ocean model systems can in general be d into three main categories; diagnostic, stic and data-assimilation systems. In diagnostic ing, the state of the ocean is directly derived fromtions [e.g. Engedahl et al., 1988]. In prognostic

ing, which is the topic of this review, the on of the model system is governed by time andependent continuity and momentum equations, equation of state (see below), and prescribed heric forcing fields. Data assimilation systems lt in response to the fact that both observationsgnostic model systems are imperfect: The

state remains close to the observed state thron3a,b]. For climate research, data-assimila

s are of fundamental importance to seasonal to l climate prediction assessments as the initial of key importance for the evolution of the system [e.g., Collins et al., 2005]. nostic model systems are fully governed by the e equations [e.g., Müller and Willebrand, 1989]

sed of coupled time and space dependent tum and continuity equations and an equation he primitive equations are regarded as good mations provided that vertical motion is much than horizontal motion, and that the fluid layer s small compared to the radius of the sphere. e prognostic model is started from an initial

will compute, time-step by time-step and grid-y grid-point, a complete set of internally ent dynamic and thermodynamic fields. The air-ndary conditions are prescribed fields of windeat and fresh water fluxes. These fields represe

ospheric forcing of the ocean-sea ice systemmmon atmospheric forcing products are monthlyimatological fields, and daily varying fields le from, for instance, the NCAR/NCEP

ww.cdc.noaa.gov/cdc/reanalysis/) and EC

of 27



(http://wwprojectcommo f past, prstates.

Progback to [1984]. Ocean ions by Semcontinucode, th cs. The lat l. [2004]. withou d OGCM ple, a derivat

The known OGCM as the verdepths.vertica t the versystemordinatdevelop t of thes as been applied toArctic Mellorand Hä

A thipotentiordinat c co-ordi latter choice n transpoof cons Oberhuisopycn ce, Aukrust tic configurawhile H rhuber [2002] circula s betwee tively. Today, widelyNERSCMICOM

The isopycnadvantahybrid he most w o-

ordinatwhich t ic co-ordina d with terrain-layer islimit ofgrid res M concep tem [Semtn

For a in geneare alsoprojectincludeco-ordi up [1997]; and WillebrModel present[2004] AOMIPhttp://fi

3. MOD AND OBSERVATION-MODEL COMPAR

This

addressMICOM is useful athe senappliedadequa(indicaThe sec en simulatassess t l-model ly extensi l., 2004; H een restricted y large-sc

The fcirculat c (or vert ean and temseasonaminimuand seaSeptemare give

3.1 Sim

w.ecmwf.int/research/era/) reanalyses s. By its nature, a prognostic model is the nly used model system to study the dynamics oesent and future atmosphere and ocean climate

nostic 3-dimensional ocean modelling extends the pioneering works of Bryan [1969] and CoxThis model system, now known as the Modular

Model (MOM), was extended to the polar regtner [1976a,b]. The model system has ously developed through improvements of the e physical parameterisations, and the numeri

est state of MOM is documented in Griffies et aThe MOM system and derivations thereof is

t comparison the most frequently and widely use system of today. NAOSIM is, as an examive of the MOM system. Bryan-Cox (or MOM) model system is also as a level or geopotential co-ordinatetical discretisation is based on layers at fixed There are two alternative formulations for the l discretisation in OGCMs: One approach is to letical co-ordinate follow topography, and this is known as terrain-following (or sigma) co-e OGCM. The Princeton Ocean Model (POM) ed by Blumberg and Mellor [1987] was the firs

e models. A derivative of this system h the North Atlantic, the Nordic Seas and the

by, in particular, Sirpa Häkkinen [Häkkinen and , 1992; Häkkinen, 1995, 1999, 2002; Mauritzen kkinen, 1999]. rd approach is to treat surfaces of constant al density, or isopycnals, as the vertical co-e. Such a model system is known as an isopycninate OGCM. The motivation behind the of vertical discretisation is that the oceart and mixing are mainly directed along surfaces tant density. The OPYC model developed byber [1993] played a pioneering role in the use of al models for high latitude oceans. For instan and Oberhuber [1995] used an Atlantic-Arc

tion with a grid focussed in the Nordic Seas olland et al. [1996] and Karcher and Obe

applied OPYC to study the mixed layer tion and the exchanges of different water massen the Arctic and the subarctic seas, respec however, MICOM [Bleck et al., 1992] is the only used isopycnic co-ordinate model system. The

model used in this review is a derivative of .

geopotential, the terrain-following and the ic co-ordinate OGCMs have all inherent ges and weaknesses. As a consequence of this, co-ordinate OGCMs have been developed. Tell-known hybrid OGCM is the HYbrid C

e Ocean Model (HYCOM; Bleck [2002]) in he open stratified ocean is treated with isopycn

tes, the shallow coastal regions are treatefollowing co-ordinates, and the upper mixed treated with geopotential co-ordinates. In the infinite numerical resources – and hence infinite olution – the basic features of the various OGCts will converge into one consistent model syser, 1995]. thorough review of prognostic ocean modellingral, see Griffies et al. [2000]. Of high relevance the results of several model intercomparison

s. One such project for the North Atlantic that s geopotential, terrain-following and isopycnic nate OGCMs is documented by DYNAMO Gro Böning et al. [2001]; New et al. [2001];and et al. [2001]. Another is the Arctic Ocean Intercomparison Project (AOMIP) with resultsed in Proshutinsky et al. [2001], Steiner et al. and Uotila et al. [2005]. More information about is available at sh.cims.nyu.edu/project_aomip/overview.html.

EL-MODELISONS

section is split into two parts: The first part es similarities and differences between the

and NAOSIM systems. Such a comparison s it points towards robust model responses (in

se that the models respond similarly to the forcing, indicating that the model physics is tely represented) and model uncertainties ting where the model systems need to improve). ond part contains a comparison betweed and observed fields as this is the only way to he degree of realism of the models. Both modeand observation-model analyses become quickve [e.g., Karcher et al., 2003; Steiner et aátún et al., this issue], so this review has b to display and briefly discuss some of the ke

ale features of the Nordic Seas. ields to be addressed are: The annual mean ion at 150 m depth, the annual mean barotropiically integrated) circulation, the long-term mporal variability of the volume transport, the l cycle of SST and SSS, and the mean, m, and maximum thickness of the mixed layer ice concentration, both for March and ber. Short descriptions of the two model systems n in the appendix.

ulated circulation in the Nordic Seas region

of 27



The

climate of the Nordic Seas are the volume, heat and salt fluxes into and out of the region. The pole-ward

d nutrients into the region and further in the Arctic Ocean [e.g., Hansen and Østerhus,SkjelvaflowingGreenlkey imdynam[e.g., D t al., 200

Figuthe 199[Jakobcircula 2. The modriftersSeas taFaroes, inflow compoStrait (Water cro the NoNorwaNorweAtlantiinto thebranchcircula e drifter

How am Strait region dif s between the models, with NAOSIM

current system with northward flowing water on the SpitzbergGreenlnorthwstrait inthe coa One branchDenmaeastwa (EIC). IrmingdirectioSeas intopogra structureason linked t ooth topography in the models.The internal c odels but wit

details. of the GNAOSIcentral located that it is tooMICOMNAOSI current

It is a that NAOSIthan Mdoubled s. 40 km), bucharact on and the n 3.5).

Figuintegraflow at indicating the top[this isssurface topogra wn for a loby Leg Oberhu 3.2 SimNordic

More

the volnorthwDenma , betweeBarentsadditio the English

Therthe volNordic by a fac1996; HØsterhuobserva[Østerh on into twpresentguideli nt meter o

We acurrentheat and salt fluxes is far from solved. Lateral

most fundamental bulk properties of the marine system

flow of Atlantic Water transports significant amounts of heat, salt, carbon an

to 2000; Blindheim and Østerhus, this issue;

n et al., this issue], whereas the southward cold and fresh Polar waters off the coast of

and, and the dense GSR overflow waters, are of portance for the hydrographic, and likely the ic, state of the North Atlantic climate system ickson et al., 2002; Curry et al., 2003; Hansen e4].

re 1 shows the observed surface circulation from 0s based on analyses of drifter trajectories

sen et al., 2003] and the mean simulated tion at 150 m depth for the period 1948-200del systems show, in accordance with the , that the inflow of Atlantic Water to the Nordic kes place just east of Iceland, southeast of the and along the Scotish slope, with a minor nent located on the eastern side of the Denmark the Irminger Current). Furthermore, the Atlantic

ssing the Iceland-Scotland Ridge flows intorth Sea and continues close to the coast of y or in an outer branch following the outer gian continental slope. At about 70ºN, the c Water splits into two branches, one meandering Barents Sea close to Northern Norway, another

heading towards the Fram Strait. The simulated tion fields show a general agreement with thfield for all of these features. ever, the circulation pattern in the Fr

ferclosest to the drifter data. NAOSIM produces a two-

en side and southward flowing water on the and side of the strait, whereas almost all the ard flowing water apparently recirculates in the MICOM. Further south, the Polar Water follows st of Greenland towards the Denmark Strait. of the Polar Water continues through the rk Strait, whereas the other branch flows rd north of Iceland as the East Icelandic CurrentThe EIC is also fed by the northward flowing er Current.Both models simulate the two-nal flow in the Denmark Strait. In the Nordic terior, both models fail to simulate the phically steered, north-westward directed, flow

re seen in the central basin in the drifter data. The for this failure is not known, but it could be o a too sm

irculation is rather similar between the mh NAOSIM showing sharper gradients and more

A profound difference, however, is the locationreenland Gyre which is more to the west in M. It will be shown in Section 3.5 that the Greenland Sea gyre is too intensive and is slightly too far to the west in NAOSIM, and diffusive and is located too far to the east in . The latter difference also implies that

M produces a more prominent eastward-directednorth of Jan Mayen. pparent from this and the following figure M produces stronger gradients and more details ICOM. This is likely caused by the essentially horizontal resolution in NAOSIM (28 km vt can also be attributed to the quite different eristics of the winter-time mixing in the regisomewhat different extent of sea ice (see Sectio

re 2 shows the barotropic (or vertically ted) flow field. Many of the features from the 150 m are present in the barotopic field,

the weak stratification and the importance ofography (see Fig. 1 in Blindheim and Østerhus ue]) in guiding the circulation, including the flow, in the region. The controlling role of phy on the surface circulation has been knong time, but was first demonstrated in OGCMs utke [1991], and later confirmed by Aukrust andber [1995].

ulated and observed volume fluxes in the Seas region

quantitative results are obtained by examining ume transports into and out of the region, i.e., the ard, southward and net flow through the rk Strait, across the Iceland-Faroe Ridgen the Faroe Islands and Scotland, through the Opening and through the Fram Strait. In

n, there is a small component of flow through Channel. e are numerous observation-based estimates of ume, heat and fresh water transports through the Seas. Unfortunately, some of the estimates varytor two to three [e.g., Simonsen and Haugan ansen and Østerhus 2000; Blindhein and s, this issue]. Accurate current meter tions are only available from around 1994 us et al., 2005]. We therefore split the discussio parts; the full integration period 1948 to , essentially without reliable observational nes, and the period from 1994 with direct currebservations. re aware, however, that even with accurate meter observations, the derivation of volume,

of 27



movemgaps in g. due to instrument loss or failure) o xes. This ha . [2004] Strait. T er of the net fluxes no synoptieliminastrait w

3.2.1 The

the Nor ed in Fig. flowingand 9.2fluxes respectare 9.5GSR, a MICOM) for

e two Arctic sections. The flow through the English els.

fluxes, and even more the individual uxes for the sections in Fig. 3 and Table 1, show som s. There aall, thein extethrough longer,BeringMICOM gh the No ig. 3 and Table 1). This residual is balanced e Bering gh the Can residual flow thcorrespArchipvolumeexpecte some o

Secoto quiteatmospis the c a certain degree bility. This internal variability originates ial state. D ularly hydrog s in the N lar gyre an rait,

have ththe regHaak et

Repesystem that the y differ willustra rts given b he ocean s n shown in Figs g in the F the long-term mean northward flow in th second integration is reduced from 3.8 to 3.2 Sv, and thsecond MICOM integration (MICOM*) is only included here for istates m not be u

To suglobal OGCMinitial sof aboucomparMICOMsystemConseqsay 1 S y explained by differences in the modelor the udifferenpropertmodel f resolvenumeri

3.3 Comvolume

The

comparestimatmeasurcompil h simulatobservaOctobeestimat DenmarespectNAOSI r with 1.9 Sv. For the Faroe-Scotland openings,observe rt across transpo

ents of strong fronts, eddy activity and spatial the coverage (e.

ften hamper a reliable estimate of these flus recently been discussed by Schauer et alfor flux estimates from observations in the Fram hey estimate the error as being of the ord

volume fluxes through the strait. The fact that the are variable in time and that usually there is c cover of all straits with observations also tes the closure of missing information from one ith observations from the others.

Mean volume transports 1948-2002

simulated net volume transports into and out of dic Seas for the period 1948-2002 are provid3. It follows that the total amount of water northward across GSR is 9.3 Sv in NAOSIM Sv in MICOM, whereas the corresponding into the Arctic Ocean are 5.9 Sv and 7.9 Sv, ively. For the net southward flow, the numbers Sv (NAOSIM) and 8.6 Sv (MICOM) across nd 5.8 Sv (NAOSIM) and 7.3 Sv (

thChannel is small and is about 0.1 Sv in the two mod

The above volume fle differences between the two model system

re several reasons for these differences. First of version of MICOM used in this review is global nt, meaning that any residual flow of water the Nordic Seas region, averaged for a month or

is balanced by volume transports through the Strait and the Canadian Archipelago. In

, there is a net northward flow of 0.6 Sv throurdic Seas (F

by a pole-ward flow of 1.0 Sv through th Strait and a net southward flow of 1.6 Sv throuadian Archipelago. There is near zero rough the Nordic Seas in NAOSIM, with a onding near zero net flow through the Canadian elago due to the closed Bering Strait. Therefore, transport differences of about 0.5 Sv can be d between the two model systems, at least forf the transport routes. ndly, even OGCMs that are forced – and by that a degree constrained – by prescribed heric momentum, heat and fresh water fluxes, as ase for MICOM and NAOSIM, have of internal varia

from the essentially unknown ocean initifferences in the initial ocean state, particraphic differences in the weakly stratified regionordic Seas and in the North Atlantic sub-po

d the flux of fresh water through the Fram St

e potential to moderate the ocean circulation in ion on time scales from years to decades [e.g., al., 2003; Bentsen et al., 2004]. ated model simulations with identical model s but with different ocean initial states indicate major volume fluxes in the Nordic Seas maith a few tenths of Sv to about 0.5 Sv. This is

ted in Fig. 4, showing the net volume transpoy MICOM for an integration continuing from ttate at the end of the MICOM realisatio. 1-3. Significant differences may be noticed, e.ram Strait where

e e southward flow from 5.3 to 4.4 Sv. The

llustrating the effect different ocean initial ay have on the simulated ocean climate, and willsed further. m up: The combined effect of regional versus

model domain, and internal variability of any system based on the essentially un-known ocean tate, may produce volume transport differences t 0.5 Sv (this figure is based on the presented ison between the NAOSIM, MICOM and

* realisations, and may be larger if more model s or realisations are included in the analyses). uently, volume transport differences exceeding, v, cannot be simpl

domain (i.e., global versus regional domains) n-known ocean initial state. In this case model ces can only be attributed to the intrinsic

ies of the models like horizontal and vertical resolution, formulation and parameterisation od and un-resolved ocean processes, and the cal implementation of the governing equations.

parison between observed and simulated mean transports

simulated volume transports in Fig. 3 can be ed with available observation-based transport es. As already mentioned, reliable velocity ements are only available from 1994 onwards. A ation of available literature values, together wited values for the same time period as for the tions, is presented in Table 2. For the period r 1994 to August 2000, the observation-based e of the northward flow of Atlantic Water in therk Strait is 0.75 Sv in Østerhus et al. [2005]. The ive MICOM value is 0.8 Sv, whereas that of M is highe

both MICOM and NAOSIM are close to the d transports, but with the MICOM transpo

the Iceland-Faroe Ridge and the NAOSIM rt across the Faroe-Scotland opening 0.8 Sv too

of 27



high. WAtlantirange frSv [Østthe tim

A strapparen k Strait wThis is e Icelandcontrast ly based estimatesthe IceluncertaDenmaSv of d [Blindh w outflowØsterhsurfaceArchipestimatSeas.

For tare bas ajor inflow 1997 to these oWater addCoasta es with anMICOM , flow es recircu

Wheobserva that several openinopenin e transpoFram Sgap. Alare basmassesintegrawithouvarious about these compli ].

In conclusion, an encouraging correspondence

arly for the inflow of Atlantic Water between Iceland and Scotland. Also the simulated heat transobserva

3.4 Mo l variabi

The

marinevariabiinto anmatch otranspoone wo ies should e correspquantitby the aprocess r conversshow a d, in a dirthat the

Figusouthwsystems ffigure canomal t follows flow acdeviatioNAOSIcorrelasituatiobetweeOpenin e two modelsScotlan correlapositive s producthree se uld be noted tsystembe too s nces in the circulation in the reg

The given in the r producDenmabetweeBarents

In geanomalsystemlarger a

The the simulat alistic can only be assessed based on comparison with continuous and high-

e may note, however, that the estimates for the c Water inflow across the Iceland-Faroe ridge om 3.3 Sv [Hansen and Østerhus, 2000] to 3.8 erhus et al., 2005] depending on the method and e interval. iking difference between the two models is t in the southward transports through Denmarith 7.5 Sv for NAOSIM and 4.0 Sv for MICOM. partly balanced by the southward flow across th-Scotland ridges of 2.2 Sv for NAOSIM in to 5.0 Sv for MICOM. Observational

for the mostly deep southward flows across and-Scotland ridges are 2.2-2.9 Sv with large inties [Hansen and Østerhus, 2000]. For rk Strait an observationally based estimate of 3 eep overflow to the south is rather well acceptedeim and Østerhus, this issue], while the shallo with the EGC is very uncertain. Hansen and

us [2000] calculated the combined outflow at the through Denmark Strait and the Canadian elago to be 3 Sv, based on the residual from es for all other flows in and out of the Nordic

he Barents Sea Opening, most recent estimates ed on current meter observations from the msection between Bear Island and Norway from 2001. Ingvaldsen et al. [2004] calculated frombservations 1.5 Sv for the inflowing Atlantic

ing up with 0.5-1 Sv for the Norwegian l Current to a total of 2-2.5 Sv. This compar inflow of 2.7 Sv for NAOSIM and 4.2 Sv for

in the 1995-2000 period. For the Fram Straittimates are uncertain because of the very stronglation in the strait. n evaluating the differences of model and tionally based estimates we have to considerof the water masses which pass the Nordic Sea

gs are subject to intense recirculation at the gs which complicates the interpretation of thrts at fixed sections: this hold especially for the trait, Denmark Strait and the Faroer-Scotland so the fact that different observational estimates ed on different definitions for the passing water may lead to confusion. Here we use vertically ted total water column transports for the models t perfoming a detailed comparison along the water mass definitions. A discussioncations can be found in e.g. Nilsen et al. [2003

between the observed and simulated volume transports are obtained, particul

ports (Table 2) are consistent with the tion-based values.

del-model comparison of the interannuality of the volume transports

major variability modes of the Nordic Seas climate system can be explored by examining the lity of the amount and properties of water flowing d out of the region. Irrespective of the actual r mismatch between the simulated mean

rts through the different sections shown in Fig. 3, uld expect that the simulated flow anomalshow similarities over the integration period. Thondence should be particularly clear for ies that are directly and to a large degree driven pplied atmospheric forcing fields, or for es that are properly resolved by the models. Oely, it is likely that simulated quantities that high degree of co-variance over time are forceect way, by the applied atmospheric fields, and governing ocean dynamics is appropriate. re 5 shows the simulated northward and ard volume transport anomalies for both model

or the five open ocean sections in Fig. 3. In the aption, the linear correlation of the transport ies between MICOM and NAOSIM are given. I from Fig. 5 that the variability in the northwardross the Denmark Strait is weak (standard n of 0.16 Sv and 0.13 Sv for MICOM and M, respectively), and that there are no

tion between the two simulated time series. The n is opposite for the northward flow anomalies n Iceland and Scotland and across the Barents g. Here the mean standard deviations of th are 0.37 Sv (Iceland-Faroe), 0.45 Sv (Iceland-d) and 0.33 Sv (Barents opening), and the giventions are significant. It is interesting, and a result, that two widely different model system

e so consistent flow anomalies through these ctions for such a long time period. It sho

hat the correlation between the two model s breaks down for the Fram Strait, which may not urprising given the differe

ion (cfr. Figs. 1-2). southward volume transport anomalies are ight panels of Fig. 5. The two model systemse quite consistent transport anomalies in the rk Strait, across the Iceland-Faroe Ridge, n the Faroes and Scotland and through the Opening, but not for the Fram Strait. neral, the magnitude of the volume transport ies is comparable between the two model s, although MICOM tends to produce slightly nomalies than NAOSIM.

extent to which the temporal evolution of ed anomalies in Fig. 5 is re

of 27



quality cuSec. 3. ted transpoFaroe-S to Decem oth MICOMobservevariabitime seseries aclose mpotentimodel modes of the mavailab to extend,e.g. Há .

3.5 Sim

In th

addressSeas. W an fresh w way from the ice edge where strong salt fluxes might occur, SSS is far less influenced by the re, salinity can be mixing of he major water masses in the region.

The upp n of SSS in Mresolutisharper d in the E ll the way ile MICOM and the Iceland evident MICOM .

The modelslower s ral part of t in the cent

Obse f temper g the cent The obflowing A500 m 6 ºC ar In the c ridian, temper the wes

Greenla constitu

Both e return A . The po narrowMICOMclose tohas a toNAOSI n AtlantiWater othe menable to distribuyears (sappend

Figu a ice concecorrespfor the a ice structur per left pan as docume e integra sport in the Is Odden on in NAOSImaximobserve ar of 1881 [B d Østerhus, this issue]. It should be kept in mFig. 9 apoints t and that the el has actinstanc he Greenladetailed th that sim l. [2003].

The thickne . MICOMwith an d on the GasNAOSIparame result of convstratificthickne he density t is there n of the tof deep is quite similar with rather deep mixing south of GSR, along the path of the

rrent meter observations (cfr. the discussion in 2). Here, as an example, observed and simularts are compared for the northward flow in the hetland Channel for the period October 1998

ber 2002 (Fig. 6). Consistent with Table 2, b and NAOSIM are, on average, close to the

d northward volume flux. The monthly mean lity is, however, larger in the observation-based ries. By subtracting the mean values of the time nd by normalising the resulting anomalies, a atch is obtained. This finding illustrates the

al for combining observations and prognostic systems to better understand the variability

arine climate system in the region, and that le model systems have the potential to be used in time and space, available observations (see tún et al. [this issue])

ulated thermodynamic surface properties

e following, a model-model comparison es the thermodynamic properties of the Nordic e start with SSS or equivalently the upper oce

ater content. A

atmospheric forcing than SST. Therefoconsidered as a tracer for the transport and

ter panels in Fig. 7 displays the March distributio

ICOM and NAOSIM. Due to the differences in on, NAOSIM describes more details and accepts gradients than MICOM. This is especially pronounceGC that is visible as a fresh and cold boundary layer a to the southern tip of Greenland in NAOSIM wh produces a front between the polar waters of the EGC

subpolar Atlantic water aligned with the Greenland-Ridge. The northward flowing Atlantic Water is off the coast of Norway. NAOSIM is fresher than in the central and eastern part of the Nordic Seas

March SST is quite similar between the two in the east (lower panels in Fig. 7), but with urface temperatures in MICOM in the centhe Nordic Seas, and with a very cold regionral Greenland Sea in NAOSIM. rved and simulated vertical distributions oature in July 1999 along 75 ºN, and thus crossinral Greenland Sea basin, are provided in Fig. 8.

servations clearly show the warm poleward tlantic Water towards east, extending below

eastward of 9ºE. At the surface, waters exceeding e found in the upper 50 m eastward of about 3ºE.entral Greenland Basin near the prime meatures below 0 ºC is found below about 50 m. Int, the return Atlantic Water flows along the

nd continental slope, whereas cold Polar Watertes the main water mass on the Greenland shelf.

the poleward flowing Atlantic Water and thtlantic Water are clearly seen in the simulations

leward flowing Atlantic Water is, however, too in both models, and then particularly in

. The mixed layer temperature in MICOM is the observed temperature, whereas NAOSIM o warm layer at the surface. Towards west, M produces too strong and prominent retur

c Water. Both models capture the cold Polar n the Greenland shelf. It is encouraging, despite tioned differences, that both model systems are

produce realistic sub-surface temperature tions after a total integration time of about 100 pin-up plus the reanalysis integration, see ix). re 9 displays the simulated mean and extreme se

ntrations for the period 1948-2002. The onding observed sea ice edge position is given period 1978-2002. The Is Odden – the see extending into the Greenland Sea in the upel of Fig. 9 – is very pronounced in NAOSIMnted in the mean sea ice concentration over th

tion period. Sea ice formation and tranare important processes for deep convecti

M [Gerdes et al., this volume]. The NAOSIM um sea ice extent seems to even exceed the d sea ice extent in the extremely cold yelindheim an

ind, however, that the ice extent extrema of re derived from time series at individual grid hroughout the integration period 1948-2002, map does not correspond to a state the modually occupied during the integration. For e, high sea ice cover in the Barents Sea and in tnd Sea virtually never occur simultaneously. A comparison of observed sea ice variability wiulated by NAOSIM is given in Kauker et a

combination of winter SSS and SST governs the ss of the upper well-mixed layer in the ocean

and NAOSIM treat the mixed layer differently, explicit mixed-layer model in MICOM basepar [1988] bulk representation, whereas M does not employ an explicit mixed layer terisation. A deep winter mixed layer is aective mixing as a result of unstable ation. For the diagnostic, the mixed layer ss in NAOSIM is based on the depth where t is 0.02 kg m-3 higher than the surface density. Ifore difficult to perform a one-to-one comparisowo mixed layer fields. Nevertheless, the structure mixing in March

of 27



s,

and withi he Greenland Basin. The history of deep convectiosimulat volume

Thermodel convec . 10. Under duces very sh C while Mparts o s and maintains relatively deep convectio d (not shstronglmaxim n Fig. 10Greenlproducuniform the No is caused comparexperimch et

9; y

4. PARMODE

4.1 Mo

In th

the excocean b exchan e Nordiccircula sses within t nd the pro ntic and Arctic

For botproper Rossby of thumb;dynam rid spacing tion radius 1-2 kmparts o it of whabeyond t

ocean mresoluticomponmisreprform th lt in the s th Atlantic becoming the main source for northe t has bee at this canfluctuatwater in in the Lanomalintermedistanc the pot from the larg tic energy present

Enhaof regioresolutiregionaclimatoscale mfeedingOGCMcan be s of defosystem ies (i.e., tehave, inboundarie artefactinformaof modMartinJensen t al. [1997];

The o R is assocmodelsthe othmodelsfluid duparame e been deproblempotentimodificresolveoverflo layer th

Botto Ms are rev illworth [2003], and various schemes

Atlantic Water in the eastern parts of the Nordic Sean tn in the Nordic Seas and the Labrador Sea as

ed in NAOSIM is the topic of Gerdes et al. [this]. e are substantial differences between the two systems when interannual variability of tive winter mixing is considered, see Figminimum mixing conditions, NAOSIM proallow convection that is restricted to the NwAICOM shows shallow convection over large

f the Nordic Sean in the Irminger Basin and south of Icelan

own). NAOSIM convective mixing depth y exceeds the mixing depth in MICOM at um mixing in the Greenland Sea (lower panels i). In fact, NAOSIM mixes to the sea floor in the and Sea at maximum mixing, whereas MICOM es less deep but a much more extended and mixing along the eastern and northern rims of

rdic Seas. A likely reason for this difference by the very weak stratification in NAOSIM ed to MICOM (Fig. 8). Dedicated model ents incorporating passive tracers like

lorofluorocarbons [Schlosser et al., 1991; Bönischal., 1997] or sulphur hexafluoride [Watson et al. 199Olsson et al., 2005; Eldevik et al., this isse] are probablrequired to properly address the mean state and the variability of the simulated mixed layer.

TICULAR CHALLENGES FOR THE LLING OF THE NORDIC SEAS

del specific challenges

e previous section, particular focus was put on hanges of the Nordic Seas with the adjacent asins. We regard proper representation of these

ges as the foremost challenge for modelling th Seas as a key region of the global ocean tion. This is particularly the case since procehe Nordic Seas influence the exchange rates a

perties of the waters exported to the Atlaoceans.

h the exchanges and the interior processes, resolution of the ocean eddies, or the baroclinic Radius, of O(10 km) is important. As a rule an OGCM will properly describe ocean ics on a horizontal scale of about 5 times the g. Proper model representation of the deforma

would therefore require an ocean grid spacing of . Such a fine grid mesh for the Nordic Seas and f the neighbouring oceans is currently on the limt is computationally feasible, and certainly the computational resources available for mos

odelling groups. Models with insufficient on – like virtually all of the current ocean ents of coupled climate models – typically esent the properties of the deep overflows that e dense constituents of NADW. This can resuubpolar Nor

rn hemisphere dense water in the models. In demonstrated [Döscher and Redler, 1997] th result in a false sensitivity of the models to ions in atmospheric forcing and anomalous fresh flux to the North Atlantic. The deep convectionabrador Sea is much more susceptible to ies than the more robust formation of diate waters north of GSR occurring over long es and through several processes. Furthermore,entially important process of energy transfer e-scale potential energy reservoir to the kineof the large circulation is not captured by the parameterizations. nced grid resolution can be obtained by nesting nal, high-resolution models into larger, coarser on models. One variant of this approach are l models that receive boundary conditions from logy [e.g., Gerdes et al., this issue] or larger odels [e.g., Hátún et al., this issue] without back into these 'parent' models. Regional s have the advantage that the spatial grid mesh fine, possibly resolving the first baroclinic radiurmation. The disadvantage with regional model s is that water mass fluxes and their propertmperature, salinity and tracer concentrations) general, to be prescribed at the lateral

s. Technical problems can lead to boundarys propagating into the model domain. For more tion about open boundary conditions and nesting

el systems with different resolution, see sen and Engedahl [1987]; Ginis et al. [1998]; [1998]; Palma and Matano [2000]; Perkins e Heggelund and Berntsen [2002]. verflow of dense water masses across the GSiated with excessive mixing in many ocean [Gerdes, 1993; Roberts and Wood, 1997]. On er hand, it has been a problem for isopycnal to include sufficient entrainment of ambient ring the overflow [Roberts et al., 1996]. New terisations [Hallberg, 2000; Shi et al., 2001] havveloped and implemented to reduce this . However, according to Gerdes [1993],

al vorticity constraints make mixing or frictional ation of the flow inevitable in models that don't the baroclinic Rossby Radius and when the wing water masses experience large changes inickness. m boundary layer parameterisations in OGC

iewed by K

of 27



have beenDöscheThese sdown tentrainlarge-simprov gh latitude rs et al. [200hydrau he Denmamodel.

The bounda ed by continu with a of Atla the oceaffectin t is an im t releaseal., 200 cluded in ocean-s ric fields. eat supply 2003; Gedge in2003].

The branching of the Atlantic water into the two ter as

an Mayen Current a andic Current are heavily

graphy. The bathymetry of an OG M is usually computed as the mean value of the real o ll. This imchannel s therefo of the bathymspacing hed bathymparticu .

Smousually els are, at lea ed by artificia[BiastoAgain,basic re odels. Howev e furtherbetweevery se f barotroNAOS

represeRelief ( nd Wrightboundachanne ar challen

The ions where o[Marsh to influenmasses e downw [1998],scales wfuture bparame to inclu[PaluskMuch mincorposcale v s. Linked of brine w e-scale mdynamicsand Ba in OGChighly

Final the outfeedingcan aff cially the confresh w ait into the n of both the subWater u region,of GreeNorway ater and thechallen

4.2 Obs

The

model ses and defic the ana betweethroughexampl f OGCMmarine time and spa

devised, among others, by Beckmann and r [1997] and Killworth and Edwards [1999]. chemes provide a pathway for dense waters

he topographic slopes and avoid excessive ment. The schemes have been employed in many cale ocean models where they, to some degree, e overflows and sloping convection from hi shelf seas. A new approach is due to Köste5] who introduce a parameterization based on

lic control theory to describe the strength of trk Strait Overflow in a large-scale Atlantic overflows and other exchanges across the ries of the Nordic Seas are strongly linkity. A weaker overflow will thus be associated

weaker inflow of Atlantic waters. Weaker inflowntic waters will lead to reduced heat release froman to the atmosphere in the Nordic Seas, in turn g the properties of the return Atlantic Water thaportant contribution to the overflows. The hea

can also affect Arctic sea ice volume [Goose et 4], although this is an effect not inea ice models forced by prescribed atmosphe

More directly, the Atlantic inflow affects the hto the Atlantic layer of the Arctic [Karcher et al., erdes et al., 2003] and the position of the sea ice

the Barents and Greenland seas [Kauker et al.,

Arctic contributaries and the return Atlantic Wawell as the branching of the EGC into the J

nd the East Icelinfluenced by details of the topo

Ccean depth underlying each horizontal grid ce

plies that bathymetric features like ridges and s are smoothed. The degree of smoothing i

re governed by the horizontal length-scaleetric features compared to the actual grid . In addition, some OGCMs require a smootetry to avoid numerical instabilities. This has larly been the case for terrain-following OGCMsothed (or artificially reduced) height of ridges is not adjusted in OGCMs, whereas deep chann

st in climate modelling, commonly adjustl widening or deepening of the channels ch et al., 2003; Beismann and Barnier, 2004]. proper resolution of these features is one of the quirements for Nordic Seas circulation mer, horizontal and vertical grid spacing hav, less obvious effects. The energy transfer n baroclinic and barotropic modes seems to be nsitive to resolution as the much larger energy opic flows in a higher resolution version of IM indicates [K. Fieg, pers. comm.]. Clearly, the

ntation of the Joint Effect of Baroclinicity and JEBAR) [Sarkisyan and Ivanov, 1971; Mertz a

, 1992] in a region of strong water mass ries and characterized by deep and narrow ls, steep slopes and complex ridges is a particulge for ocean modelling. central Greenland Sea is one of few locatpen ocean convective mixing takes place all and Schott, 1999]. The process is believedce the ventilation of dense, sub-surface water on climate time scales. Except for the large-scalard vertical advection proposed by Budeus et al. most vertical mixing processes take place on hich are not and can not in the foreseeable e explicitly resolved by OGCMs. Several trisation schemes have therefore been proposedde vertical sub-grid scale processes iewicz and Romea, 1997; Canuto et al., 2004]. ore work is, however, needed to properly

rate a physically consistent description of small-ertical mixing processes in climate-type OGCM to open ocean convection is also the treatmentaters generated during freezing of sea ice. Finodelling has been carried out to describe the

of brine waters released from sea ice [Kämpf ckhaus, 1999], but the incorporation of this effectMs is typically ignored or incorporated in a

simplified way. ly, the Atlantic inflow into the Arctic is linked toflow of much fresher waters near the surface, the EGC. The fresh water carried by the EGC ect the interior of the Nordic Seas and espevection in the central Greenland Sea. Most of the ater, however, is carried through Denmark Str subpolar Atlantic. Proper representatioduction of the northward flowing saline Atlantic nder the fresh polar water in the Fram Strait

the dynamics of the fresh water along the coast nland (and, similarly, along the coast of ), and the frontal mixing between the fresh w

more saline open ocean surface waters are ges for OGCMs.

ervation-based evaluation of OGCMs

only way to proceed from plain comparison of results and by that identifying model weaknes

iencies are to actively include observations inlyses. Figure 6, showing a one-to-one comparisonn observed and simulated northward transport the Faroe-Scotland Channel, provides an e of direct observation-based evaluation os. Unfortunately, available observations of the climate system are, in general, scattered ince. It is therefore difficult, and in many cases

of 27



imposs of a mo e scales, mentioregionscoverag revik et al., 200 f availab ded to generate the best pevaluatsimulatobservalocatioevaluat

For d tracers like chradiocaeffectivventilawaters Maier-2003]. radiois sing plants h o address the inter-annual transport,Atlanti ., 1998; K n shorter cer experimassessitagged deliber entral Greenlsubseqwithin ; Eldevik

5. CON

Num

the No of similar lling efforts. Tthat aremotionAtlantiinfluenis demo of the No , and thaobservabe less come news bimprov in

large-sc led climate

Modvariabi ds into a larger scale perspectiv with th oceanicmodel ng fluctuatheir pa e fate of passage of the wat g into difwithin t of oceaidentifi ith the 199the lastlocationBlindheinvaluafor dram tem of the r

Desp to be ov rovements to be realize nt model differenThese d e Nordicexisted erences are bas e would s have tathe ArcprocessFram S terior of the G polar North Abetter oprocesshorizon en implem

The wcontinu s calls fohas beestate. Ttransfer re climatebiogeoc t is therefore foreseen that available tailoredsignific

ible, to adequately address the quality, or realism,del system on climate (i.e., multi-decadal) timas on the time scale of Fig. 5. It should be ned, however, that the Nordic Seas is one of the in the World Ocean with best observational e [Blindheim and Østerhus, this issue; Fu2]. Acquisition, quality check and synthesis ole observations are therefore nee

ossible observation-based background for ing the mean state and the variability of ed ocean states. Needless to say, continuous tions of key ocean parameters at key ocean

ns are of paramount importance for any ion of OGCMs. ecadal and longer time scales, the use of

lorofluorocarbons (CFC-11 and CFC-12) and rbon (14C) has turned out to be useful and cost-e in assessing the integrated (or net) effect of

tion of the basin-scale and World Ocean surface [e.g., Toggweiler et al., 1989; England and Reimer, 2001; Dutay et al., 2002; Gao et al. Furthermore, tracers from point sources like otopes from the European nuclear re-procesave been useful t

mixing and age properties of, for instance, the c Water in the Nordic Seas region [Nies et alarcher et al., 2004; Gao et al., 2004, 2005]. O

time and smaller spatial scales, dedicated traents have been found to be of great use for

ng small-scale mixing and transport of explicitly water masses. A unique example here is the ate release of sulphur hexafluoride in the cand Sea in 1996 [Watson et al. 1999], and the uent observation of the spreading of the tracer and out of the Nordic Seas [Olsson et al., 2005 et al., this issue].

CLUSION

erical modelling of the ocean-sea ice system of rdic Seas and adjacent areas has reached a statematurity to that of Atlantic and Arctic mode

his is the case despite some unique challenges posed by the small inherent scale of oceanic , the large water mass contrast between inflowing c and Arctic water masses, and the large ce of fresh water flux fluctuations. Particularly, it nstrated that the mean exchanges into and out

rdic Seas agree with available observed estimatest interannual variations do correlate with tions, although the simulated amplitudes tend to

than the observed ones. The progress is welecause it will lead, over some time, to an ed representation of Nordic Seas processes

ale ocean-sea ice models and eventually coup models. els have recently been used to put decadal lity and observed tren

e, connecting individual observed time seriese large-scale atmospheric forcing fields and the conditions in adjacent basins. For instance,

experiments were important tools in identifyitions in the fresh water flux out of the Arctic and thways through the Nordic Seas. Similarly, thAtlantic water entering the Nordic Seas in the s through GSR, the subsequent modification er masses in the Nordic Seas, and the branchinferent paths entering the Arctic and recirculating he Nordic Seas have been the subject of analysisn-sea ice hindcast experiments. This led to the cation of multiple Arctic warming events w0s warming as an outstanding event for at least 50 years. Long-term observations at key s that have been taken in the Nordic Seas [e.g. im and Østerhus, this issue] have proven ble for model evaluation and also as indicators

atic developments in the marine climate sysegion. ite this recent progress, many difficulties remainercome and important model imp

d. As has been demonstrated here, differesystem like NAOSIM and MICOM provide t results under similar atmospheric forcing. ifferences pertain to important fluxes through th

Seas and would be of climatic importance if they in the real ocean. The cause for the diffically unknown and clarification of the causrequire systematic intercomparison efforts aken place for the Atlantic (e.g. DYNAMO) and tic (AOMIP) ocean basins. Clearly, some es like the exchanges with the Arctic through trait, the interaction of the EGC with the inreenland Sea and the overflows to the subtlantic require enhanced resolution and perhaps

r new parameterisation schemes of small scale es. New models with significantly higher tal and vertical resolution have recently beented or are planned for the immediate future. ealth of existing historical data and the

ous stream of data from observational programr a more systematic use of models. This paper n limited to studies of the present day climate he presented discussion is, however, directly able and applicable to studies of past and futu states, as well for studies of the marine hemical cycles. I

and forthcoming OGCMs, even OGCMs for the small Nordic Seas region, will antly contribute to improved understanding of

of 27



the varand fut

6. APP

6.1 The l (MICO

The

mixture1992], f the Hib Harder and Simonsen [1992].

Gao et al. [2005] o examine the transit time of the pole-ward flowing A a modific 002], Nilsen the obs eas, the var , and the inteAtlantithe FurBentsen f the parame the latter bet al. [2identic[2003] al resolut

Speckm in t eas region, es divided by the , momen salt) dispersi -1, respect -

1) is pa Brunt-Vtracer d s 2 and 6, respe[2002],

The mentio , i.e., daily atmospheric re-analysis fields fromNCARThe resnumbe here cycle twend of s initiate two.

A 20output

review.the 20 k

6.2 The(NAOS

AWI

modelsSea Iceare des Köberl re taken frof the missue]. air temprecipitation, wind speed, and surface wind stress. For

in-up, a climatological mean seasonal cycle based on the period 1979-1993 with added typRöske, consist EP re-anal1996]. include -off from into thewith an ded to the surfThe resmean csimilarlflux onOcean

Ackn

Climate e OGCMproject CTION (contract EVK-CT-2000-00058), D(EVK2 00-00080) n additioFounda der the Res sk Oceanc f MinistefundedtechnolLD 004via the throughProject vided valuablMads H1. Com

iability and stability properties of past, present ure climate states.

ENDIX

Miami Isopycnic Co-ordinate Ocean ModeM)

NERSC model used in this study is based on a of version 2.7 and 2.8 of MICOM [Bleck et al., ully coupled to a sea-ice module consisting ofler [1979] rheology in the implementation of [1996], and the thermodynamics of Drange

The model system is identical to that used in ttlantic Water in the Nordic Seas, and

ation of the model used by Furevik et al. [2et al. [2003] and Bentsen et al. [2004] addressingerved and simulated salinity in the Nordic Siability of the volume transports across GSRr-annual to decadal-scale variability of the c MOC, respectively. The main modifications to evik et al. [2002], Nilsen et al. [2003] and et al. [2004] studies are reduced strength oterised isopycnal and diapycnal mixing rates, ased on the CFC model evaluation study by Gao 003]. The applied model grid configuration is

al to that in Furevik et al. [2002], Nilsen et al. and Gao et al. [2005], but has doubled horizontion compared to that in Bentsen et al. [2004]. ifically, the horizontal grid resolution is about 40 he Northern North Atlantic and the Nordic S and the diffusive velocities (diffusivitisize of the grid cell) for layer interface diffusiontum dissipation, and tracer (temperature andon are 0.015 m s-1, 0.01 m s-1 and 0.0025 m s

ively. The diapycnal mixing coefficient Kd (m2 srameterized as Kd = 5·10-8 / N, where N (s-1) s the

äisälä frequency. Consequently, the value of ispersion and diapycnal mixing are factorctively, below those used in Furevik et al. Nilsen et al. [2003] and Bentsen et al. [2004].

applied forcing is identical to all of the above-ned studies

1948 to present provided by the /NCEP re-analyses project [Kalnay et al., 1996]. ults presented here are based on integration cycle r two and three with NCAR/NCEP forcing, w

o is initialised with the full ocean state at the cycle one (the spin-up cycle), and cycle three id with the full ocean state at the end of cycle km version of the model system is available, but from this model version is not used in this

See Hátún et al. [this issue] for an example of m version of MICOM.

North Atlantic/Arctic Ocean-Sea-Ice-Model IM)

maintains a hierarchy of coupled sea ice-ocean called NAOSIM (North Atlantic/Arctic Ocean- Model). Models from the NAOSIM hierarchy cribed in some detail in Karcher et al. [2003] ande and Gerdes [2003]. Results for this paper weom an experiment with a quarter degree version odel that is described in Gerdes et al. [this

The model was forced with daily mean 2-meter perature, dew point temperature, cloudiness,

the first 50 years of sp

ical daily variability [OMIP-climatology, 2001] is used. After the spin-up, the forcing s of daily mean atmospheric data from the NCysis for the period 1948-2001 [Kalnay et al., Fresh water influx from rivers is not explicitlyd. To account for river run-off and diffuse run the land, as well as to include the effect of flow

Arctic through the Bering Strait, a restoring flux adjustment time scale of 180 days is adace freshwater flux [Gerdes et al., this issue]. toring flux is calculated in reference to annual limatological surface salinity data, constructed y to the initial data. The effect of the restoring the surface salinity for this and other Arctic models is documented in Steele et al. [2001].

owledgement. The European Union DG-XII and Environment Programme has supported th activities at NERSC and AWI through the s CONVE

YNAMITE (contract 003903), PREDICATE -CT-1999-00020) and TRACTOR (EVK2-20. The model development at NERSC has in received support from the G. C. Rieber tions, the RegClim and ProClim projects unearch Council of Norway, and the Vestnordilima programme under the Nordic Council ors. The ocean modelling at AWI has also been

by the German ministry for research and ogy under the DEKLIM program (contract 01 7) and by the Deutsche Forschungsgemeinschaft SFB 512. Additional funds were made available the Arctic Ocean Model Intercomparison

(AOMIP). Dr. J. E. Ø. Nilsen (NERSC) proe help with some of the figures and tables, and vid Ribergaard provided the drifter data in Fig. ments from Torben Schmidt and three

of 27



anonymThis is for Clim

REFER

Aukr Greenlandice-mix4771-47

Beck oved represengeopote591.

Beism meridiosensitiv , 92-106,

Bennoceanogpp.

BentsSimulatcirculat

BiastSensitivForcing10.1175

Bleckframed Modelli

Bleckdriven tforced i ic. J. Phys. O

Blumthree-diDimens p. 1-16, Americ Washington, DC, 1987.

Bönin2001. Ssubtrop ion to obser

BöniWallacesalinity, Sea. J. G

Bryaclircula .

Bude ter convect d Sea. J. G

CanuDubovi cean deep co

ColliLatif, M , R., Terrpredicta multi-perfect-model-ensemble study. J. Climate, cond accepted

4: A primitive equation three-dimensional . Report, GFDL Ocean Group, NOAA,

Princeton Univ., Princeton, NJ. n

Nature 426, 826-829. Dickson, B., Yashayaev, I., Meincke, J., Turell, B., Dye, S.,

Holfort, J.,Ocean o

DöscnorthernAtlantic1894-19

Drang fluxes in rt 125, Nansen Griegsv. 3

Dutayventilatimodels.

DYN ., Böning,Dieteric Kröger,A. L., O 1997. DSimulatBerichte -Albrech

EngeProductNordic S

Engla al tracers t

Even l formula53, 343–

FurevKorable surface sdoi:10.1

Gao Y . Tracer-dNordic S

Gao, Y.diapycnthe Nort55B, 83

Gao, Simulat h Atlantic-Arctic rdoi:10.1

Gasp r ocean. J

Gerdemodel u 2 applic , 14703-1

GerdeCauses events.

Ginisof a muWea. Re

Goosse Influencon sea-iClimate

GriffiGerdes, , D., 2000. DModelli

Griffi ., 2004. ADynami

HaakFormatiGeophy 0, DOI 10.1029/2003GL017065.

Häkk olar oceanogrPergam

ous reviewers are also highly acknowledged. publication No A 97 from the Bjerknes Centre ate Research.

ENCES

ust, T., Oberhuber, J.M., 1995. Modeling of the, Iceland and Norwegian Seas with a coupled sea

ed layer-isopycnal ocean model. J. Geophys. Res. 100, 89. mann, A., Döscher, R., 1997. A method for imprtation of dense water spreading over topography in ntial-coordinate models. J. Phys. Oceanogr. 27, 581-

ann, J.-O., Barnier, B., 2004. Variability of thenal overturning circulation of the North Atlantic: ity to overflows of dense water masses. Ocean Dyn. 54 DOI 10.1007/s10236-003-0088-x. ett, A.F., 1992. Inverse methods in physical raphy. Cambridge University Press, Cambridge, 346

en, M., Drange, H., Furevik, T., Zhou, T., 2004. ed variability of the Atlantic meridional overturning ion, Clim. Dynam., doi: 10.1007/s00382-004-0397-x. och, A., Käse, R.H., Stammer, D.H., 2003. The ity of the Greenland–Scotland Ridge Overflow to Changes. J. Phys. Oceanogr. 33, 2307–2319, doi: /1520-0485(2003)033. , R., 2002. An oceanic general circulation model

in hybrid isopycnic-cartesian coordinates. Ocean ng 4, 55-88. , R., Rooth, C., Hu, D., Smith, L.T., 1992. Salinity-

hermocline transients in a wind- and thermohaline-sopycnic coordinate model of the North Atlantceanogr. 22, 1486-1505. berg, A.F., Mellor, G.L., 1989. A description of a mensional coastal ocean circulation model, in Three-ional Coastal Ocean Models, N. Heaps (Ed.), pan Geophysical Union,g, C.W., Dieterich, C., Barnier, B., Jia, Y.,

easonal cycle of meridional heat transport in the ical North Atlantic: a model intercomparison in relatvations near 25°N. Prog. Oceanogr. 48, 231-253.sch, G., Blindheim, J., Bullister, J.L., Schlosser, P., , D.W.R., 1997. Long-term trends of temperature, density, and transient tracers in the central Greenlandeophys. Res. 102, 18553-18571.

n, K., 1969: A numerical model for the study of tion of the world oceans. J. Comput. Phys. 4, 347-376us, G., Schneider, W., Krause, G., 1998. Winive events and bottom water warming in the Greenlan

eophys. Res. 103, 18513-18527. to, V. M., Howard, A., Hogan, P., Cheng, Y., kov, M.S., Montenegro, LO.M., 2004. Modeling onvection. Ocean Modelling 7, 75-95. ns, M., Botzet, M., Carril, A., Drange, H., Jouzeau, A., ., Otterå, O.H., Pohlmann, H., Sorteberg, A., Suttonay, L., 2005. Interannual to decadal climate bility: A

itionallyCox, M.D., 198

model of the ocean

Curry R., Dickson, R.R., Yashayaev, I., 2003. A Change ithe Freshwater Balance of the Atlantic Ocean over the Past four decades.

2002. Rapid freshening of the deep North Atlantic ver the past four decades. Nature 416, 832-837.

her, R., Redler, R., 1997. The relative importance of overflow and subpolar deep convection for the North Thermohaline circulation. J. Phys. Oceanogr. 27, 02. e H, Simonsen, K., 1996. Formulation of air-sea the ESOP2 version of MICOM. Technical RepoEnvironmental and Remote Sensing Center, Edv.

A, N-5037 Solheimsviken, Norway. , J.-C., et al., 2002. Evaluation of ocean model on with CFC-11: comparison of 13 global ocean Ocean Modelling 4, 89-120. AMO Group (Barnard, S., Barnier, B., Beckmann, A C.W., Coulibaly, M., de Cuevas, D., Dengg, J., h, C., Ernst, U., Herrmann, P., Jia, Y., Killworth, P.D., J., Lee, M.-M., LeProvost, C., Molines, J.-M., New, schlies, A., Reynaud, T., West, L.J., Willebrand, J.),YNAMO - Dynamics of North Atlantic Models. ion and assimilation with high resolution models. aus dem Institut für Meereskunde an der Christiants-Universität Kiel, 294, 334 pp. dahl H, Ådlandsvik, B., Martinsen, E.A., 1998. ion of monthly mean climatological archives for the eas. J. Mar. Syst. 14, 1-26. nd, M., Maier-Reimer, E., 2001. Using chemico assess ocean models. Rev. Geophys. 39, 29–70. sen, G., 2003: The Ensemble Kalman Filter: theoretication and practical implementation. Ocean Dynamics 367, DOI 10.1007/s10236-003-0036-9 ik, T., Bentsen, M., Drange, H., Johannessen, J., v, A., 2002. Temporal and spatial variability of the seaalinity in the Nordic Seas. J. Geophys. Res. 107, 029/2001JC001118. ., Drange, H., Bentsen, M., Johannessen, O.M., 2005erived transient time of the eastern waters in the eas. Tellus, in press.

, Drange, H., Bentsen, M., 2003. Effects of al and isopycnal mixing on the ventilation of CFCs in h Atlantic in an isopycnic coordinate OGCM. Tellus 7-384. Y., Drange, H., Bentsen, M., Johannessen, O.M., 2004. ing the transport of radionuclides in the Nortegion. J. Environ. Radioactivity 71, 1-16, 016/S0265-931X(03)00108-5. ar, P., 1988. Modeling the seasonal cycle of the uppe. Phys. Oceanogr. 18, 161–180. s, R., 1993. A primitive equation ocean circulation sing a general vertical coordinate transformation, Partation to an overflow problem. J. Geophys. Res. 984726. s, R., M. Karcher, F. Kauker, and U. Schauer, 2003.

and development of repeated Arctic Ocean warming Geophys. Res. Lett. 30, doi: 10.1029/2003GL018080. , I., Richardson, R.A., Rothstein, L.M., 1998. Design ltiply nested primitive equation ocean model. Mon. v. 126, 1054-1079.

H., Gerdes, R., Kauker, F., Koeberle, C., 2004. e of the exchanges between the Atlantic and the Arctic ce volume variations during the period 1948-1997. J. 17, 1294-1305. es, S.M., Böning, C., Bryan, F.O., Chassignet, E.P., R., Hasumi, H., Hirst, A., Treguier, A.-M., Webbevelopments in ocean climate modelling. Ocean ng 2, 123-192. es, S.M., Harrison, M.J., Pacanowski, R.C., Rosati, A technical guide to MOM4, NOAA/Geophysical Fluid cs Laboratory, Princeton, USA, 337pp. , H., Jungclaus, J., Mikolajewicz, U., Latif, M., 2003. on and propagation of great salinity anomalies. s. Res. Lett. 3inen, S., 1991. Models and their applications in paphy, in Polar Oceanography (Ed. W.O. Smith),

on Press.

of 27



Häkk e

Greenlasurface eophys. Res. 100, 4761-4770.

Häkkinen, S., 1999. Variability of the simulated meridional .

., Mellor, G.L., 1992. Modeling the seasonal variability Arctic ice-ocean system. J. Geophys. Res. 97, 20

HäkknorthernRes. 107

Hallb n and Ricsoordin

Hans Sea Exc

HansAlready 10.1126

HardMeereisfür Pola pp.

HebumesoscaSeas. J.

Hegg g nesting t ns. Int. J. N

Hible Model.

HollaNumeriArctic O

HopkphysicaEarth-S

Ingvacycle in2001. C

JacobT., Hugnorthernvariabil . 108, 32

Jense ocean m

Johan ite evidenc nce 286, 1937-1939

Kalnay,analysis

Kamduring sthree di /7), 1335-13

KarcArctic w event in . Res. 108

KarcGH.E., 99 in the Nmodel r185-198

Karcmodific rctic Ocean.

KaukJ.L., 20 ice: A combin om

1978 to 10.1029

KillwboundarPhys. O

Killwlayer in Process13th 'Ah an 2003, H185.

Köbe rmining the variabil16, 2843

Köste tner, A., Herrmann, P., 2005. The effect nmark Strait Overflow on the Atlantic meridional overturning circulation. Geophys. Res. Lett., in

cal investigation of the circulation in the Greenland and Norwegian Seas. J. Phys. Oceanogr.

Marsobserva

Martitesting of conditio -627.

MaurbetweenOvertur . 46, 877-

Mello ice-ocean m Role in Shaping the Globa .

Mertzterm. J.

MüllemotionsVolumeBerlin

New, 1. On the o Seas: in

Nies,KuhlmanBackhau1998. A the Arctic Oce Hydrog50, 313

NilseM., 2003exchangLett. 30

Oberhcirculatigeneral cOceano

OlssoEldevikWatsonSea in th ulphur hexafluo

ØsterMeasure Arctic Mdoi:10.1

Palm open bothree-dim

Palusmodel focean. D

inen, S., 1995. Simulated interannual variability of thnd Sea deep water formation and its connection to forcing. J. G

heat transport in the North Atlantic for the period 1951-1993J. Geophys. Res. 104, 10991-11007.

Häkkinen, S of a coupled ,285-20,304.

inen, S., 2002. Surface salinity variability in the North Atlantic during recent decades. J. Geophys. , 10.1029/2001JC000812. erg, R., 2000. Time integration of diapycnal diffusiohardson number-dependent mixing in isopycnal ate ocean models. Mon. Wea. Rev. 128, 1402-1419. en, B., Østerhus, S., 2000. North Atlantic - Norwegianhanges. Prog. Oceanogr. 45,109-208. en, B., Østerhus, S., Quadfasel, D., Turrell, W., 2004. the day after tomorrow? Science 305, 953-954, DOI: /science.1100085. er, M., 1996: Dynamik, Rauhigkeit und Alter des es in der Arktis. PhD thesis Alfred-Wegener-Institut r- und Meeresforschung, Bremerhaven, Germany. 124

rn, G.W., Johnson, C.D., 1995. Simulations of the le circulation of the Greenbland-Iceland-Norwegian

Geophys. Res. 100, 4921-4941. elund Y., Berntsen, J., 2002. A method for analysinechniques for the linearized shallow water equatioumer. Meth. Fl. 38, 163-185. r, III W.D., 1979. A Dynamic Thermodynamic Sea IceJ. Phys. Oceanogr. 9, 815–846. nd, D.M., Mysak, L.A., Oberhuber, J.M., 1996. cal simulation of the mixed-layer circulation in the cean. J. Geophys. Res. 101, 1111 - 1128. ins, T.S., 1991. The GIN Sea – a synthesis of its l oceanography and literature review 1972-1985. ci. Rev. 30, 175-318. ldsen, R.B, Asplin, L., Loeng, H., 2004. The seasonal the Atlantic transport to the Barents Sea during 1997-ont. Shelf Res. 24, 1015-1032. sen, P.K., Ribergaard, M.H., Quadfasel, D., Schmith, hes, C.W., 2003. Near surface circulation in the North Atlantic as inferred from Lagrangian drifters:

ity from the mesoscale to interannual. J. Geophys. Res51. n, T.G., 1998. Open boundary conditions in stratifiedodels. J. Mar. Sys. 16, 297-322. nessen O.M., Shalina, E.V., Miles, M., 1999. Satell

e for an Arctic Sea Ice cover in transformation. Scie.

E., et al., 1996. The NCEP/NCAR 40-year re- project. Bullet. Amer. Meteorol. Soc. 77, 437-471. pf, J., Backhaus, J.O., 1999. Ice-ocean interactions hallow convection under conditions of steady winds: mensional numerical studies. Deep-Sea Res. 46(655.

her, M.J., Gerdes, R., Kauker, F., Köberle, C., 2003. arming: Evolution and spreading of the 1990s warm

the Nordic Seas and the Arctic Ocean. J. Geophys, 3034, doi:10.1029/2001JC001265.

her, M.J., Gerland, S., Harms, I.H., Iosjpe, M., Heldal, Kershaw, P.J., Sickel, M., 2004. The dispersion of Tcordic Seas and the Arctic Ocean: A comparison of esults and observations. J. Environm. Radioactivity 74, .

her, M.J., Oberhuber, J.M., 2002. Pathways and ation of the upper and intermediate waters of the AJ. Geophys. Res. 107, 10, 1029/2000JC000530. er, F., Gerdes, R., Karcher, M., Köberle, C., Lieser, 03. Variability of Arctic and North Atlantic seaed analysis of model results and observations fr

2001. J. Geophys. Res. 108, 3182. doi: /2002JC001573. orth, P.D., Edwards, N.R., 1999. A turbulent bottom y layer code for use in numerical ocean models. J. ceanogr. 29, 1221-1238. orth, P.D., 2003. Inclusion of the bottom boundary ocean models, In: Muller, P. (Ed.) Near-Boundary es and Their Parameterization, Proceedings of the a Huliko'a Hawaiian Winter Workshop, 21-24 Jonolulu HI, University of Hawaii SOEST, 2003, 177-

rle, C., Gerdes, R., 2003. Mechanisms deteity of Arctic sea ice conditions and export. J. Climate -2858. rs, F., Käse, R.H., Schmit

of De

press. Legutke, S., 1991. A numeri

21, 118-148. hall, J., Schott, F., 1999. Open-ocean convection: tions, theory, and models. Rev. Geophys. 37, 1-64. nsen, E.A., Engedahl, H., 1987. Implementation and

a lateral boundary scheme as an open boundary n in a barotropic ocean model. Coastal Eng. 11, 603

itzen, C., Häkkinen, S., 1999. On the relationship dense water formation and the “Meridional ning Cell” in the North Atlantic Ocean. Deep-Sea Res894. r, G.L., Häkkinen, S., 1994. A review of coupled odels. In: The Polar Oceans and their

l Environment. Geophysical Monograph, 85, 525pp, G., Wright, D.G., 1992. Interpretations of the jebar Phys. Oceanogr. 22, 301–305. r, P., Willebrand, J., 1989. Equations for oceanic . In Landolt-Börnstein, Group V, Oceanography, 3b, pages 1.14. J. Sündermann (Ed.), Springer Verlag,

A.L., Barnard, S., Herrmann, P., Molines, J.M., 200rigin and pathway of the saline inflow to the Nordic

sigths from models. Prog. Oceanogr 48, 123-161. H., Harms, I.H., Karcher, M.J., Dethleff, D., Bahe, C.,

, G., Kleine, E., Loewe, P., Oberhuber, J.M., s, J.O., Matishov, D., Stepanov, A., Vasiliev, O.F.,

nthropogenic radioactivity in the Nordic Seas and an: Results from a joint German project. Deutsche

raphische Zeitung, (German Journal of Hydrography) -343. n, J.E.Ø., Gao, Y., Drange, H., Furevik, T., Bentsen, . Simulated North Atlantic-Nordic Seas water mass

es in an isopycnic coordinate OGCM. Geophys. Res. , doi: 10.1029/2002GL016597. uber, J.M., 1993. Simulation of the Atlantic on with a coupled sea-ice mixed layer-isopycnal irculation model, Part I: Model description. J. Phys.

gr. 23, 808-829. n, K.A., Jeansson, E., Anderson, L.G., Hansen, B., , T., Kristiansen, R., Messias, M.-J., Johannessen, T., , A.J., 2005. Intermediate water from the Greenland e Faroe Bank Channel: spreading of released sride. Deep Sea Res. I 52, 279-294.

hus, S., Turrell, W.R., Jónsson, S., Hansen, B., 2005. d volume, heat, and salt fluxes from the Atlantic to theediterranean. Geophys. Res. Lett. 32, L07603,

029/2004GL022188. a, E.D., Matano, R.P., 2000. On the implementation ofundary conditions for a general circulation model: the

ensional case. J. Geophys. Res. 105, 8605-8627.kiewicz, T., Romea, R.D., 1997. A one-dimensional or the parameterization of deep convection in the yn. Atm. Ocean 26, 95-130.

of 27



Perkins,

1997. Aequation

Prosh A., M. Steele, J. Zhang, G. Holloway, N. Steiner, Hä ., Karcher,Hibler, ies DifferenAmeric

Robe 97. Topographic Sensitivity Studies wiOceano

RöskEC

Sarkisyabaroclin dynamiOceanic

Schauer 2004. Arctic wtranspor es. 109, C0

SchlossReducti a during the 1980 56.

SemtArctic O ceanogr. 6, 409-425.

Semtnerthermod s of clima

SemtnerScience

Shi, X the Denmar es. 106, 22

SimoArctic Mparamet 01, 6553-65

StamHaimba Adcroft, A., Hill, C.N., Marshall, J., 2003a. circulat model c(WOCE) ddoi:10.1

Stam iering, R., Eckert, C., Haimbach J., 2003estimatemodel. es. 108, 3118, doi:10.102

Steel ., HollandSteiner, the Beaufort Gyre: A model i Lett 28, 2935-2838.

Steiner,HollandProshutmodeled sArctic Odoi:10.1

Steve an circulation model of the Norwe -402.

Togg s of radiocar l, 1, Steady state prebomb distributions. J. Geophys. Res. 94, 8217–8242.

UotilHäkkinen, S., Holloway, G., Karcher, M., Kauker, F., Steele,

M., Yakovenergy-Arctic mdoi:10.1

WatsonJohanne ., Ledwell , 1999. M yre studied by tracer release. Na

WilleKillwor ., New, A.L., 2001 g models H. Dran Center

A.L., Smedstad, L.F., Blake, D.W., Heburn, G.W., new nested boundary condition for a primitive ocean model. J. Geophys. Res. 102, 3483-3500. utinsky,

kkinen, S., Holland, D., Gerdes, R., Koeberle, C M., Johnson, M., Maslowski, W., Walczowski, W.,

W., Wang, J., 2001.Multinational Effort Studces Among Arctic Ocean Models. EOS Transactions,

an Geophysical Union, 82, 637, 643, 644. rts, M.J., Wood, R.A., 19

th a Bryan–Cox-Type Ocean Model. J. Phys. gr. 27, 823-836. e, F., 2001. An atlas of surface fluxes based on the

MWF Re-analysis: A climatological dataset to force global ocean general circulation models. Max-Planck-Institut fur Meteorologie Report 323, 31 pp.

n, A. S., Ivanov, V.F., 1971. Joint effect of icity and borrom relief as an important factor in the

cs of sea currents. Izv. Acad. Sci. USSR, Atmos. Phys. 7, 173-178.

, U., Fahrbach, E., Østerhus, S., Rohardt, G.,arming through the Fram Strait - Oceanic heat t from three years of measurements. J. Geophys. R6026, DOI: 10.1029/2003JC001823.

er, P., Bönisch, G., Rhein, M., Bayer, R., 1991. on of deepwater formation in the Greenland Ses: Evidence from tracer data. Science 251, 1054-10

ner, A. J. Jr., 1976a. Numerical simulation of the cean circulation. J. Phys. O

, A. J. Jr., 1976b. A numerical model for the ynamic growth of sea ice in numerical investigationte. J. Phys. Oceanogr. 6, 379-389.

, A. J. Jr., 1995. Modeling ocean circulation. 269, 1379-1385.

.B., Rød, L.P., Hackett, B., 2001. Variability ofk Strait overflow: A numerical study. J. Geophys. R277-22294. nsen, K., Haugan, P.M., 1996. Heat budgets of the editerranean and sea surface heat flux

rizations for the Nordic Seas. J. Geophys. Res. 176. mer, D., Wunsch, C., Giering, R., Eckert, C., ch, P., Marotzke, J.,

Volume, heat and freshwater of the global oceanion 1993-2000, estimated from a general circul;ationonstrained by World Ocean Circulation Experiment

ata. J. Geophys. Res. 108, 3007, 029/2001JC001115. mer, D., Wunsch, C., G

, P., Marotzke, J., Adcroft, A., Hill, C.N., Marshall,b: Global ocean circulation during 1992-1997, d from ocean observations and a general circulation

J. Geophys. R9/2001JC000888.

e, M., Ermold, W., Holloway, G., Häkkinen, S, D.M., Karcher, M., Kauker, F., Maslowski, W., N., Zhang, J., 2001. Adrift inntercomparison. Geophys. Res.

N., Holloway, G., Gerdes, R., Häkkinen, S., , D.M., Karcher, M., Kauker, F., Maslowski, W., insky, A., Steele, M., Zhang, J., 2004. Comparing

treamfunction, heat and freshwater content in the cean. Ocean Modelling 6, 265-284, doi: 016/S1463-5003(03)00013-1. ns, D.P., 1991. A numerical oce

gian and Greenland Seas. Prog. Oceanog. 27, 365

weiler, J., Dixon, K., Bryan, K., 1989. Simulationbon in a coarse-resolution world ocean mode

a, P., Holland, D.M., Morales Maqueda, M.A.,

lev, N., Zhang, J., Proshutinsky, A., 2005: An diagnostics intercomparison of coupled ice-ocean odels. Ocean Modelling, 016/j.ocemod.2004.11.003, in press.

, A.J., Messias, M.-J., Fogelquist, E., Scoy, K.V., ssen, T., Rey, F., Tanhua, T., Carse, F., Simonsen, K, J., Jansen, E., Cooper, D., Kruepke, J., Guilyardi, E.ixing in the Greeland Sea g

ture 401, 901-904. brand, J., Barnier, B., Boning, C., Dieterich, C., th, P.D., le Provst, C., Jia, Y., Molines, J.-M

. Ciculation characteristics in three eddy-permittinof the North Atlantic. Prog. Oceanogr. 48, 123-161.

ge, sing Nansen Environmental and Remote Senand B47, 50helge

R. GerdReseaGerm

Y. Gao, and B47, 50

M. Karc Marine ResearcGerm

F. Kauk e ReseaGerm

M. BentCenteThormmats.

.

jerknes Centre for Climate Research, Thormhølensgt. 06 BERGEN, Norway (email:

[email protected]) es, Alfred Wegner Institute for Polar and Marine rch, Bussestrasse 24, 27570 BREMERHAVEN, any (email: [email protected]) Nansen Environmental and Remote Sensing Centerjerknes Centre for Climate Research, Thormhølensgt. 06 BERGEN, Norway (email: [email protected]) her, Alfred Wegner Institute for Polar and

h, Bussestrasse 24, 27570 BREMERHAVEN, any (email: [email protected]) er, Alfred Wegner Institute for Polar and Marinrch, Bussestrasse 24, 27570 BREMERHAVEN, any (email: [email protected]) sen, Nansen Environmental and Remote Sensing r and Bjerknes Centre for Climate Research, hølensgt. 47, 5006 BERGEN, Norway (email:

[email protected])

of 27



Table 1. Simulated northward, southward, and net northward (northward-southward) volume (Sv), heat flow (TW) and fresh water flux (0.01 Sv) through the Denmark Strait (DS), across the Iceland-Faroe Ridge (IFR), between the Faroes and Shetland (FS), across the Barents Opening (BO) and through the Fram Strait (FS) in MICOM and NAOSIM. The heat flows lated online (i.egiven relat o S=34.8. Negathe h 4.8) are direerr .2 Sv. σV is sta

FS

and fresh water fluxes are computed and accumuive to T=0 ºC and the fresh water fluxes relative t

., time step by time step). The heat flows are tive heat and fresh water fluxes means that

cted opposite to the flow. Due to round-off eat and fresh water transports (relative to T=0 ºC and S=3ors, the volume transport budgets balance within 0.1 to 0 ndard deviation.

Section DS IF BO FS

Northward volume transports (Sv) FV σV FV σV FV σV FV σV FV σV NAOSIM 2.0 .13 3.5 .33 MICOM 1.1 .16 4.5

3.8 .45 2..42 3.6 .46 4.1

e transports (σ σV FV σV F

7 .34 3.2 .49 .33 3.8 .63

Sv) V σV FV σV

Southward volum

F FV V V NAOSIM 7.4 .49 1.2MICOM 3.7 .48 2.4

.21 0.9 .07 0.

.32 2.5 .26 2.

d) volume traFV σV F

7 .13 5.1 .60 0 .37 5.3 .62

Net (northward – southwar FV σV FV σV

nsports (Sv) V σV FV σV

NAOSIM -5.4 .46 2.3 .36 MICOM

2.9 .48 2.1.2 .69 2.1

transports (TWFH σH F

0 .40 -1.9 .33 -2.7 .49 2.0 .37

Northward heat

FH σH FH σH

.32 -1.5 .36

) H σH FH σH

NAOSIM 26 3.5 101 10MICOM 26 3.4 106 16.1

.8 137 16.1 49136 15.6 86

transports (TW

H FH σH F

8.7 28 6.7 8.6 25 2.8

)Southward heat FH σH FH σ

H σH FH σH

NAOSIM 22 7.2 29 5.3 MICOM 7 1.2 37 6.5

14 1.5 5 24 5.5 21

ard) heat trans

H FH σH F

1.5 -5 2.7 4.3 24 5.2

ports (TW) H F

Net (northward – southw F F σH σH H σH H σH NAOSIM 4 6.1 72 11.6MICOM 19 3.8 69 12.1

Northward salt F σ F

123 16.5 44112 19.4 65

transports (0.01σS FS σS F

8.5 33 6.6 6.8 1 4.2

Sv) S σS FS σS S S S

NAOSIM -0.6 0.3 3.6MICOM -0.4 0.3 3.4 0.

0.4 5.1 0.6 1.28 4.8 0.7 0.

ansports (0.01σS FS σS F

0.3 -0.2 0.4 2 0.5 2.0 1.6

Sv) S σS FS σS

Southward salt tr

FS σS FS NAOSIM -9.2 1.8 1.1 0.2 MICOM -4.4 1.0 0.9 0.5

0.8 0.1 -0.1 1.6 0.3 -0

ard) salt transpF σ FS σS FS σS F

0.2 -6.7 1.6 .2 0.3 -8.1 2.6

Net (northward – southw

orts (0.01 Sv) S σS FS σS S S

NAOSIM 8.6 1.6 2.5MICOM 4.0 0.8 2.5 0.5

0.4 4.4 0.6 1.3.2 0.8 0.

3 0.2 6.5 1.5 4 0.4 10.2 3.0

of 27



Table 2. Observed [Østerhus et al., 2005] and simulated properties of the Atlantic inflow to the Nordic Seas. Heat flow is relative to 0 oC.

A

D

F

tlantic inflow Period Average fluxes

Volume Heat

Sv TW

Observed enmark Strait 1994-2000 0.75 19

NAOSI

MICOM

01 Observ

M 1.9 26

0.8 24

Iceland-Faroe Jun 1997 – Jun 20 ed 3.5 124

NAOSIM

MICOM

Observ

3.4 98

4.3 95

aroe-Scotland 1994-2000 ed 3.2 149

NAOSIM

MICOM

4.0 147

3.7 137

of 27



Figure 1. Upper panel: Observation-based n during the 1990s [Jacobsen et al., 2003]. Lower panels: Simulated velo ty at 150 m ag v e time period 1948-200 anel) and MICOM.

ear-surface circulation derived from drifters ci aver ed o er th 2 by NAOSIM (left p

of 27



of 27

Figure 2. Simulated barotropic velocity field averaged ove e time perio 2 by NAOSIM (left panel) d MICOM. r th d 1948-200 an



es (Sv) for the period 1948-2002 (upper panels) and for 1995-19ght. Throughout the manuscript, the transports from MICOM are calculated on

Figure 3. Simulated mean northward and southward volume flux 99 (lower panels). NAOSIM to left and MICOM to ri line,

e. time step by time step, whereas weekly output of velocity, tem e and salinity are used to diagnose the transports from i. peraturNAOSIM.

of 27



Figure 4. As Fig. 3, but for a realisation with MICOM, labelled MICOM*, starting in 1948 from the ocean state at the end of 2002 of the version of the model shown in Figs. 1-3.

of 27



Figure 5. Simulated annual mean volume transport anomalies (Sv) for NAOSIM (thin lines) and MICOM. Left (right) column shows anomalies in net northward (southward) transports. The panels represent, from top, the Denmark Strait (r = 0.05/0.46), the Iceland-Faroe Ridge (r = 0.51/0.45), the Faroe-Scotland section (r = 0.59/0.63), the Barents Opening (r = 0.28/0.35), and the Fram Strait (r = -0.05/0.04). The numbers in the parentheses give the (linear) correlation coefficient between the two model time series for the northward and southward volume fluxes, respectively, with the bold numbers identifying 95% or higher significance. The sections are indicated in Fig. 3.

of 27



Figure 6. Observed and simulated northward volume transport through the Faroe-Shetland Channel. Upper panel: Observed transport (Sv) based on current meters [Østerhus et al., 2005] in dashed line, simulated NAOSIM (MICOM) transport (Sv) in thin (thick) solid line. Mid panel: Corresponding normalised volume flux anomalies, where normalization is done with respect to the standard deviation of the respective time series. Lower panel: The number of days with observations per month. The correlation between the observed time series and MICOM is 0.43 (significant at 99% confidence level). The corresponding figure for NAOSIM is 0.26 at 90% confidence level (serial correlations are included in the analysis).

of 27



Figure 7. Simulated salinity (psu) (upper panels) and temperature (ºC) (lower panels) in March averaged over the uppermost 150 m of the water column for 1948-2002. NAOSIM at left and MICOM at right.

of 27



Figure 8. Observed (upper panel) and simulated temperature along 75 ºN in the Nordic Seas (NAOSIM to left, and MICOM to right). The observations are from cruise ARK XV/1 with RV Polarstern from 23 June-19 July 1999 [G. Budeus, pers. comm.]. The simulated fields are July monthly means. The same contouring is used for all panels.

of 27



aximum) concentration is given in the lower left (right)

panels. Observations and NAOSIM are shown with thin lines and shaded contour levels of 10%, 40% and 70%, whereas MICOM is shown with thick solid lines with similar contour levels.

Figure 9. Mean sea ice concentration in March for 1978-2002 from observations (upper left panel) [Johannessen et al., 1999] and for1948-2002 from the simulations (upper right panel). Simulated minimum (m

of 27



Figure 10. Simulated mean (upper panels) and maximum (lower panels) mixed layer depth (m) in March for the period 1948-2002. NAOSIM to left and MICOM to right.

of 27

ocean general circulation modelling of the nordic …

Documents