Jun 15, 1999 - Institute for Marine Research, University of Hamburg, Hamburg, Germany ... The great Siberian rivers that drain .... Some smaller corrections.
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104, NO. C6, PAGES 13,431-13,448,JUNE 15, 1999
Modeling the seasonalvariability of hydrography and circulation in the Kara
Sea
I. H. Harms Institutefor Marine Research,Universityof Hamburg,Hamburg,Germany
M. J. Karcher• FederalMaritime and HydrographicAgency,Hamburg,Germany
Abstract. In the frame of a projecton transportof contaminants in the Arctic, the Hamburgshelf oceanmodel (HAMSOM) is appliedto the Kara Sea. The HAMSOM systemconsistsof a threedimensional,barocliniccirculationmodelcoupledto a thermodynamicanddynamicseaice model. The Kara Seamodelis forcedwith climatologicalwinds,atmosphericheatfluxes,river runoff, and tides.The obtainedresultsrevealno typical Kara Sea circulationthat prevailsthroughoutthe year. Instead,the modelshoweda strongseasonalvariabilityin circulationandhydrographydue to winds,freshwaterrunoff,andice formation.The circulationis weakestin springwhenthe wind speedsare low andhorizontaldensitygradientsare small.Freshwaterfrom the rivers spreads towardthe northand northwestratherthan forming a coastallytrappedcurrentthat flows to the east. In autumn,the circulationis significantlyenhancedbecauseof increasingwind speedsand stronghorizontaldensitygradients.Good agreementwas foundbetweenmodel resultsand recent observations. The "classical"cycloniccurrentpatternin the southernKara Sea, however,was not reproducedby the model.
1. Motivation
Over the past 5 years, several publications reported on increasingenvironmentalproblemsin the Arctic Ocean and the Kara Sea (overview is given by Nilsson [1997]). One prominent example of Arctic pollution is the dumpingof radioactive waste by the former Soviet Union [Yablokov et al., 1993] which provided the motivation for intense experimental fieldwork [Joint Russian-Norwegian Expert Group, 1996] and model studies on radionuclide transportin Arctic Shelf seas [Scott et al., 1997' Harms, 1997b]. The great Siberian rivers that drain huge land areasand industrialzonesmay also contribute to the input of pollutants into the Arctic environment. In addition, intense oil and gas exploitation which is planned for the Kara Sea and finally the operation of the Northern Sea Route may causeenvironmentalproblems[Moe et al., 1997]. In 1994, a joint project was initiated by the German Federal Maritime and Hydrographic Agency and the University of Hamburg (Institute for Marine Research)to investigatetransport and dispersionof contaminantsin the Arctic Ocean. The major goal of this work was to improve the oceanographicknowledge of the Arctic and the Kara Sea with respect to pathways and transittimesof possiblecontaminants.Severalnumericalmodels were used in this project to evaluate present and future environmentalproblemsof that region. The present model study is part of this project and investigatescirculation and hydrographyof the Kara Sea with
specialemphasison the seasonalvariability.Section2 describes the modelusedand section3 presentsthe results.A summary and discussionare given in section4.
2. Kara
Sea Model
The KaraSeamodelis basedon thecodingof the Hamburg shelf oceanmodel (HAMSOM) which is briefly describedin section2.1 and 2.2. A moredetaileddescriptionof HAMSOM withapplications to coastalwatersandArcticShelfseasis given by Stronachet al. [1993], Harms [1994], andHarms [1997a]. Specialemphasisis given to the applicationof the forcingdata which is described in section 2.3.
2.1. HAMSOM
system
HAMSOM is a three-dimensional, baroclinic circulation model, developed at the Institute of Marine Research
•NowatAlfred-Wegener-Institute forPolarandMarineResearch,
(UniversityHamburg)for investigations of shelf sea processes [Backhaus,1985]. The level-type model is basedon nonlinear primitive equations of motion, invoking the hydrostatic approximationand the equation of continuitywhich servesto predictthe elevationof the free surfacefrom the divergenceof the depth mean transport. The numerical scheme of the circulation model is semi-implicit, and the equations are discretizedasfinite differenceson an ArakawaC-grid. Vertical sub-gridscaleturbulenceis parameterized by means of a turbulentclosureapproach,proposedby Kochergin[1987] and later modifiedby Pohlmann[1996]. The schemeis closely
Bremerhaven,Germany
related to a Me!lot and Yamada [1974] level-2 model where
Copyright1999by the AmericanGeophysicalUnion.
verticaleddy viscositycoefficientsdependon stratificationand verticalcurrentshear.Convectiveoverturningis parameterized
Paper number 1999JC900048.
by vertical mixing: an unstable stratification is turned into a
0148-0227/99/1999JC900048509.00
neutralstatethroughartificialenlargementof the verticaleddy 13,431
13,432
HARMS AND KARCHER: MODELING THE KARA SEA
viscositycoefficient. The horizontal diffusion of momentum is calculatedusinga constant,isotropiceddy viscositycoefficient. The circulation model includes an Eulerian transport algorithmfor temperature,salinity, and passivetracersbasedon the advection-diffusionequation and an upstream scheme. Vertical eddy diffusivity coefficientsare calculatedin the same way as vertical eddy viscosity coefficients, depending on stratification and vertical current shear. Horizontal eddy diffusion is neglectedbecauseof numericaldiffusion stemming from the advection
scheme. This artificial
horizontal
diffusion
case to
0.2
m.
observed
ice thickness
and concentrations
thinner
than
this
value
is
Sea
treated
new ice lbrmation.
The dynamic part of the ice model is kept simple and does not include an ice theology. lnstead, a "free drift" algorithm of ice thickness
and ice concentration
due to wind and water stress at the surface or the bottom
of the
ice [Bruno and Madsen. 1989]. This approachwas chosen in
2.2. Model Configuration
The HAMSOM-systemis appliedwith highspatialresolution to the centralKara Sea (Figure 1). The domainincludesvery shallow areas (< 50 m) along the Siberian coast, the river
estuaries of Ob andYeniseias well as the deepNovayaSemlya Trough (> 400 m). The topographyis basedmainly on the ETOPO 5 data set [Hirtzle•; 1985]. Some smaller corrections
were madeaccordingto sea chartsof that region[Perry and Fleming,1986; Cherkiset al., 1990] and topographic data sets from other model applications[Kowalik and Proshutinsky,
1993].The KaraStrait'(tothe BarentsSea)andthe Vilkitsky Strait (to the Laptev Sea) representopen boundariesto the model.The northernpartsof the KaraSea,namely,theSvyataya Anna Trough, were excludedsince theseareasare strongly influenced by the Barents Sea outflow toward the Arctic. The grid of the model is equidistantwith a meshsize of 9.4 km. The
verticalscaleis resolvedwith 12 layershavingboundariesin 5, 10, 15, 25, 50, 75, 100, 150, 200, 300, 400, and500 m depths.
!
0
in the Kara
constant for thin and thick ice is set in
Ice
heat fluxes. Latent and sensible heat fluxes at the
for advection
In order to smooth
remarkablywell.
ocean-or ice-atmosphereinterfacesare calculatedwith standard bulk formulae [Maykut, 1986] on the basisof air temperatures, humidity, and cloud cover. Additional heat fluxes encompass the conductiveheat flux through the ice, the turbulent heat flux under the ice, the long-wave radiation, and the incoming shortwave radiation. Salt fluxes due to brine releaseand ice melting are proportional to thermodynamicice growth [Lemke et ai. 1990]. A constant and isotropic ice salinity of 10 practical saliniy units (psu) is usedwhich is related mainly to thin ice and
accounts
However, a "numerical switch" between normal ice movement
(depth > 20 m) and no ice movement(depth < 20 m) caused rather unrealisticdiscontinuitiesnot only in the ice velocities
regionswith more than 50 m depth, where the ice drift is free and regionswith lessthan 20 m depth, where the ice drift is set to zero. This interpolationalgorithmis applied only in winter (i.e. January,February, and March). It will be shown in section 3.1.1. that this rather simple dynamic approachreproduces
thermodynamicallyas open water. Thermodynamic changes in ice thicknessand concentrationdepend principally on the sum of the involved
suppressedin very shallow regions with depth < 20 m.
this effect, we decided to define a "transition zone" between
The circulation model is coupled to a thermodynamicand dynamic sea ice model which calculates space and time dependentvariationsof ice thicknessand ice concentration.The basic configurationtbllows Hibler's [1979] one-layer sea ice model. The classification
orderto account for landfastice,thee icedriftin wintershouldbe
but also in ice thickness and concentration.
is
related to the advectionvelocity, the averagegrid sizesand the time step.For small and moderatevelocities(< 0.3 m/s), which are typical for the simulatedKara Sea circulation,the artificial diffusionremainsbelow 103m:/s.
our
order to reducethe computertime for tracer simulationsover long time spans The calculationof the wind stressat the ice or water surfacefollows a second-orderapproachwith individual wind drag coefficientsfor water and ice [McPhee, 1979]. In
20
40
60
depth [m] 50
80
1 O0
1 O0
120
200
Figure 1. Domain and topographyof the Kara Sea model
140
300
160
HARMS AND KARCHER: MODELING THE KARA SEA
13,433
the agreement is very good in the deeper central and southwesternparts of the Kara Sea, where the tidal influence is points. strong(see below). In very shallow northeasternparts, however, If not mentionedotherwise, a zero-gradientcondition is appliedat all openboundaries to (1) the horizontalvelocity the comparisonrevealed some smaller differences (20 - 30%) components foriceandwatermovement, to (2) temperature and betweencomputedand observedamplitudesor phases(Table 1). salinity,andalsoto (3) iceconcentration andicethickness. The In two cases,the differenceswere higher than 40%. This can be attributedto the northeasternopen boundaryof the model which sea surfaceelevationat open boundariesis prescribed.The calculation of boundary values encompassesthe inverse unfortunately cuts through an amphidromic point west of barometriceffectcomingfrom the air pressure, the geostrophic SevernayaSemlya. The tidal solution proved to be sensitiveto adjustment of thebaroclinic field,andthetides(seesection 2.3). prescribed amplitudes and phases there. Another reason for differences might be the model topography which may 2.3. Applied Forcing Data reproduce the very shallow, complex bottom relief insufficiently. The applicationof numericalmodelsto Arctic Shelf seasis Simulated M 2 tidal elevations are dominated by an verydifferentfrom applications to well-knownareassuchas the amphidromicpoint in the westernKara Sea which is forcedby a North Sea or the Baltic Sea. The quality and quantityof forcing 180ø phase lag of the incoming tidal wave between the Kara data is usuallycritical in Arctic Shelf regions,and a validation Strait and the open boundary to the Arctic Ocean. Simulated of model resultsis a difficult task. The intention of the paper is amplitudes (Figure 2a) usually remain below 20 cm except in thereforealso to describethe forcing data in order to evaluate two areas:the southern "Baydaratskaya"Bay area, where tidal their role in affectingthe seasonalvariabilityof circulationand hydrography. The Kara Sea modelis forcedwith (1) M2-tides, resonancecausesamplitudesof more than 70 cm, and the area north of Yamal peninsulaaroundBelyy Island (30 - 35 cm). (2) monthly mean climatologicaldata for wind speed and Simulated maximum tidal currents(Figure 2b) are strongest direction, (3) monthly mean climatological data for air in the BaydaratskayaBay and near the river estuary of the Ob temperature, cloudcover,and humidity,(4) 10-daymeanriver (30 cm/s). In the small strait between Yamal and Belyy Island, runoff data, (5) 10-day mean Kara Strait throughflowdata, and tidal currentsmay exceed50 cm/s due to considerablehorizontal (6) initial temperature andsalinityfields. gradients in tidal elevation. Preliminary tracer simulations The following five paragraphs describe the main [Karcher et al., 1997] showedthat theseareasare significantly characteristicsof the applied forcing in more detail. Section 2.3.1. summarizesbriefly the results from tidal simulations influencedby tidal mixing. Other regions,in particular those to the east of Novaya Semlya and in the easternKara Sea (along which were achievedusinga previousbarotropic/no-iceversion the Siberiancoast)are only weakly influencedby the M2 tides. of the Kara Sea model. 2.3.1. Tides. The Kara Sea Model accounts for the Residual currents can usually be tbund in shallow areas where tidal currentsare strong.Again, this holdstrue for the Ob dominant M2 tidal constituent. Amplitudes and phases were taken from a tidal model of the Arctic Ocean [Kowalik and estuary, the area north of Yamal peninsula (around Belyy Island), and the BaydaratskayaBay. However, the velocitiesare Proshutinsky,1994] and applied to the open boundaries.A comparison with observations andothermodelresults[Kowalik generallysmall and remainbelow 2 cm/s.Around Belyy Island, residualcurrentsform a weak anticycloniccirculation. and Proshutinsk).,, 1994; Gjevikand Straume,1989] showedthat
The sizeof the matrixis 85 by 170, resultingin 50,051 wet grid
Table 1. ComparisonBetweenObservedand SimulatedAmplitudesand Phasesof the M2 tide in the Kara Sea. Observed
Station
Name
Simulated
Amplitudecm
Phasedeg
Amplitudecm
Phasedeg
186
Yugorskiy Shar
22
351
26
355
188
Ytsi Kara
21
48
29
50
189
Maaresaale
11
18
9
20
190
Harasavey
17
24
26
30
191
Belyi
23
129
23
150
197
Zjelania
16
150
16
165
199
Uedinenia
10
195
6
! 95
200
IzvestiyCtik Island
11
195
7
210
201
Isachenko Island
8
180
7
205
Island
202
PravdyIsland
18
201
15
235
204
Taymir Island
19
159
19
240
205
RusskyiIsland
16
159
14
240
208
Krasnoflotskie
17
159
13
275
222
Dikson
10
318
7
320
223
Leskin Island
17
345
17
345
224
ObskayaGuba
32
357
16
360
Island
The stationnumbersreferto locationsin Figure2.
13,434
HARMSANDKARCHER: MODELINGTHEKARASEA
Figure 2. (a) Simulatedamplitudes(5-cm intervals)and phases(20ø intervals)of the M2 tide in the Kara Sea.(b) SimulatedM2 tidal currentsover one tidal cycle (tidal ellipses).Stationnumbersrefer to locationsdescribedin Table
1.
2.3.2. Atmospheric forcing. The Kara Sea model is forced with monthly mean climatologicalwind stressvalues,deduced by Trenberthet al. [1989] from EuropeanCentre for MediumRange Weather Forcastdata [1988] for the period 1980-1989. The obtainedwind fields (Figure 3) show strong,monsoon-like variability due to the seasonalair pressuredistributionover the
considerableand rangesfrom +5øC over the southwestern parts in Augustto -33øC over easternpartsin January.Comparedto thesestrongvariances,the relative humidity and the cloud cover remain quite stable during the whole year. The humidity is usually > 80%, and the cloud cover is between60% and 80% in
Arctic.
2.3.3. River runoff. The Kara Sea is dominatedby large freshwaterrunoff in spring[Pavlov et al., 1993]. The freshwater supply to the Kara Sea is mainly through the Ob and Yenisei
In winter, the Kara Sea is in a transitionzone betweena large and stable high- pressurecell over central Siberia and low-
summer and > 80% in winter.
pressureridgesover the Barentsand NorwegianSeas.Monthly meanwind speedsare high and very stablethen, with maximum velocities of 8.5 m/s in February. From October to March, a cyclonic curl prevailswith main wind directionsfrom southto
rivers
southwest
1995].
for the central Kara Sea.
In summer,the air pressuredistributionis reversed.The Siberian high-pressurecell decreases,and a high-pressurecell develops over the central Arctic. Horizontal air pressure gradientsare much weaker, and the wind speedsare lower (2-4 m/s). The wind directions are more changeable,but north or
which
drain
a catchment
area in Siberia
and Russia
of
more than 5 million km2. The total annualamountof freshwater
inputintotheKaraSeaequals roughly1200km3/yr, of which 80% is dischargedin spring(May - June) [Pavlov and Pfirman, Information
on the river runoff
to the Kara
Sea was taken
from an unpublisheddata review on the hydrology of Arctic rivers (O.F. Vasiliev et al., "Estimate of river water inflow to the Karsk Sea", Institute for Water and Environmental Problems,
Siberian Branch, Russian Academy of Sciences,Novosibirsk,
northeast winds dominate.
1995). This review includes data of observed freshwater runoff
Monthly mean climatologicalair temperature,humidity and cloud cover data were deduced by Aukrust and Oberhuber [1995] from ECMWF data for the period 1985-1990. The seasonalair temperaturevarianceover Kara Sea waters is
rates from Siberian rivers for certain years. Apart from strong seasonal variations, the runoff values also showed interannual
variationsbetween "wet" and "dry" years.Tables 2 and 3 give
yearlymeanrunoffratesin m3/sfordry,wetandaverage years
HARMS AND KARCHER: MODELING THE KARA SEA
for the Ob and Yenisei rivers, respectively.In order to be consistentwith the climatologicalatmosphericforcing, it was decidedto use a realisticyear with an averagerunoff for both rivers. This was found to be 1980, when the Ob and Yenisei
rivershadyearlymeanfreshwater discharges of 12,700m3/sand 17,300m3/s, respectively. In the Kara Sea model, the river runoff rates from the Ob,
Yenisei and Pyasina are prescribedas 10-day mean volume
fluxesin m3/s(Figure4). TheOb tributaries, Taz andPur,are includedin the Ob runoff.The salinityof the rivers at the point wherethey enterthe modeldomainwassetto 5 psu,whereasthe temperaturewas prescribeddependingon the season.Highest river temperaturein August was set to 3.8øC; the lowest was during winter at the freezingpoint (-0.3øC).
13,435
2.3.4. Kara Strait throughflow. The Barents Sea inflow
throughthe Kara Straitis of majorimportancebecauseit brings heat and salt into the Kara Sea. To account for these effects, the
net inflow or outflow is prescribedas 10-day mean volume fluxes(Figure 5). Thesedata were deducedfrom a couplediceocean isopycnicgeneralcirculationmodel for the Arctic and sub-Arctic domains [Karcher and Oberhuber, 1997]. Temperatureand salinity in the Kara Strait were prescribed accordingto the Gorshkov[ 1980] dataatlas. The simulated time series in Figure 5 suggeststhat the variability of the flow throughthe Kara Strait is mainly wind driven. Inflow from the Barents Sea is most enhanced during winter, when strongsouthto southwesterly winds prevail. The Table 3. Selected yearly mean dischargerates of the Yenisei
Table 2. Selectedyearly meandischargeratesof the Ob river. Year 1987 1980 1979
Conditions mostdry average most wet
river.
Dischargerate 8,490 12,700
Year
Conditions
Dischargerate
1968 1980
mostdry average
15,500 17,300
18,200
1995
most wet
20,900
Values aregivenin m3/s. Source isanunpublished report"Estimate of
Values aregivenin m3/s. Source isanunpublished report"Estimate of
river waterinflow to the Karsk Sea",by O.F. Vasiliev et al., Institutefor Water and EnvironmentalProblems,SiberianBranch, RussianAcademy
river water inflow to the Karsk Sea",by O.F. Vasiliev et al., Institutefor Water and EnvironmentalProblems,SiberianBranch, RussianAcademy
of Sciences, Novosibirsk, 1995.
of Sciences, Novosibirsk, 1995.
13,436
HARMS AND KARCHER' MODELING THE KARA SEA 2.3.5. Temperature and salinity data. Becauseof severe climateconditionsbut alsobecauseof political reasons,the Kara Sea was relatively inaccessibleduring the last decades.This still seemsto be the case,and resultsin a generallack of data for validationand forcingof numericalmodels.With respectto the high resolutionof the applied model grid and in order to avoid the application of insufficiently resolved temperature and salinity data sets, the model ]s not constrained to any climatological data. Instead, it was decided to use reasonable
100,000--
80,000
60,000
initial
40,000 + Pur
20,000
0
vI
vii viii IX
x xI
xII
month
Figure 4. Applied 1O-daymeanriver runoff from Ob (including tributariesTaz and Pur), Yenisei, and Pyasinarivers.
fields
for
the
start
and
to run
the
model
in
a full
prognosticmode. Several test runs showed that the small grid size allows for a very detailedspaceresolutionof, for example, frontal structuresin the vicinity of the river plume which are foundto be poorly resolvedin availableclimatologicaldata sets. The modelis startedat restin late winter, April 1. There is no ice at the sea surface. The initial temperatureis set to the freezing point in the whole model domain following the description of Pavlov et al. [1993]. The initial salinity distributionis prescribedaccordingto the springdata of Levitus [1982]. Applyingthesestartconditions,the modelneedsa spinup of at least one summer and one winter seasonto achieve reasonableresultsfor circulationand hydrography. 3. Results
maximum transport rateintotheKaraSeais 0.65 Sv duringthe firsthalfof January. In summer, whennortherly windsprevail, the inflow is reduced,showinga slightnet outflow(-0.07 Sv) towardthe BarentsSea in July and August.The meaninflow averagedover 1 yearis 0.3 Sv.
The computed time seriesof volumeflux throughthe Kara Strait is in good agreementwith observationsof Pavlov and
Pfirman [1995]. They estimated the yearlyaveragedinflow towardtheKaraSeato be in therangeof 0.04 Sv up to 0.6 Sv. However,the seasonal variabilityof the throughflow is almost unknown, except for some single observations:"Historic" circulationpatterns,deducedfrom hydrographic observations and directcurrentmeasurements, suggesta waterexchange throughthe strait in both directions[Pfirmaneta!., 1997]. According to thesesurfaceflow charts,thenorthern partof the straitis dominatedby an outflowfrom the Kara Sea(the "Litke Current")whereas the southern partis dominated by an inflow
The Kara Seamodelwasrun with the completesetof forcing functionsin a prognosticmode for severalyears.After 3 years, the model reachedan almost cyclic stationarystate.This holds true, in particular, for the circulation, the sea surfaceelevations, and the temperatureswhich behave quite robustly. A more sensitivevariable is the surfacesalinity. Simulationsfor time spanslonger than 3 yearsrevealeda small drift for the average surfacesalinity which is due to an imbalancebetweenfreshwater runoff, ice formation,and ice export. The following model resultsare taken from the third year of simulation. We first describe the results concerning the 0.7--
_
from the Barents Sea. This bidirectional flow, which also
0.5--
appearsin laboratorysimulations(T.A. McC!imans,personal communication, 1997),hasbeenrecentlysupported by satellite • imagesfrom the Kara Strait,but only for the summerseason. > .co. Advancedvery high resolutionradiometer(AVHRR) datafrom August 1988 and 1994
reveal considerable horizontal
Strait throughflow
--
•
averageinflow •
•
KaraStrait: •
0.3--
temperature gradientsacrossthe straitwhichmightbe due to horizontalcurrentshear[Pfirmanet al., 1997]. Sincethereis no informationon the winter situation,it remainsunclearif this is a permanent or a seasonal feature.
!2).1--
Theobserved bidirectional flowin theKaraStraitis thought to be forcedby northerlywindsdrivingsurfacewatersto the westandby thePechorarunoffwhichresultsin a density-driven coastalcurrent,flowingto theeast.Bothfeatures do notappear in winter; the runoff is absent and the winds are from south to
-0.1 '
southwest. It is thereforeunlikelythatthe observed bidirectional
I I I I I I I I I I I I I 30
flow is alsopresent duringwinter.Lookingat thewinddirection
90
150
210
270
330
day of the year
and speedover the Kara Strait, it is more reasonableto assume
that the BarentsSeainflow is muchstrongerfrom Octoberto Figure5. Prescribed KaraStraitthroughflow, deduced froman Marchthanduringtheremaining partof theyear.Thecomputed Arctic isopycnicgeneralcirculationmodel [Karcherand time seriesof volume fluxes in Figure 5 supportsthese Oberhuber, 1997]. The total fiver runoff is shown for assumptions.
comparison.
HARMS AND KARCHER: MODELING THE KARA SEA
13,437
Figure 6. Simulatedseasurfacetemperaturesfrom May to October.
hydrography(section3.1) followed by the circulationresultsin section 3.2. Where possible, the results are compared to observations or oceanographic descriptions such as the comprehensivesummaryon the Barentsand Kara Seasgiven by Pavlov et al. [ 1993].
3.1. Hydrography The following Section describes the simulated seasonal variability of the hydrography, focusing on sea surface temperatures(3.1.1), sea ice (3.1.2), sea surfacesalinity (3.1.3), and vertical stratification (3.2.4).
3.1.1. Sea surface temperature (SST). Becauseof a closed ice cover, the SST in the Kara Sea (Figure 6) is at the fleezing point until April. In spring,the incomingshortwave radiationis mainly used for ice melting which starts in May in the river estuariesand along the Siberian coast. In these ice-free areas, the inflowing river water mixes with melt water creating, togetherwith atmosphericwarming, a thin but very pronounced, warm and low salinesurthcelayer. The springice melting is also strongalong the east coast of Novaya Semlya where prevailing northerly winds drive the ice cover offshore. This allows the short wave radiation to penetrateinto surfacewaters and to heat them. In July, August, and September,the southernKara Sea is ice free. The highestsurfacetemperaturesoccur in Augustnear the Kara Strait and in the BaydaratskayaBay (6 ø- 7øC). These values agree well with observed average summer temperatures in the southern Kara Sea given by Pavlov et al. [1993] and Burenkovand Vasil'kov [1995]. Becauseof a small remaining ice shield between Novaya Semlya and Severnaya Semlya, the SST doesnot rise significantlyabovethe freezing pointsthere. Freezing startsalready in Septemberin the easternparts and the surface temperaturesdrop rapidly during October and
November in the whole Kara Sea. In the eastern parts of the Kara Sea, the ice cover establishesquite fast because of a pronouncedhalthe stratification(river and melt water) which inhibits vertical mixing and heat transfer from below. In the western parts, however, where the Barents Sea inflow is dominating, the halthe stratification is much weaker. Here vertical mixing and upward heat transfer are more intense, which hampersa rapid closing of the ice cover and keeps the autumnSST significantlyhigher than in the easternparts. Until December, an almost closed ice cover is established also in the
southernparts,and the SST is at (or closeto) the freezingpoint. 5OO
400
3OO
o
2OO
.w
lOO
IIIIIIII year 2
[
year 3
Figure 7. Simulated seasonalcycle of the ice volume in the Kara Sea model domain
13,438
HARMS AND KARCHER:MODELING THE KARA SEA
2.0
Figure 8. Comparison between observed (left) and simulated (right) ice thickness in the Kara Sea, during August/September(minimum) and March/April (maximum).
model results.For this purpose,mean ice thicknessdistributions in Siberian Shelf seas [Romanov, 1995] during minimum and maximum ice seasons (August-September and March-April, (average) ice thicknessof 1.7 m. In August-September, respectively) were averaged over the years 1972-1981. The however,the Kara Sea is almosttotally ice free. Only a small ice comparison with model results (Figure 8) reveals a slight
3.1.2. Sea ice. The model reproducesa strongseasonal varianceof the ice cover. In January,February,and March, the Kara Sea is completely covered by ice with a maximum
shield remainsbetweenNovaya Semlyaand SevernayaSemlya
overestimation of the simulated ice thickness in winter, whereas
with an averageice thickness of 0.5 m. The seasonal varianceof
the simulated
ice volumein the modeldomainis depictedin Figure 7' the ice
alono
volumedecreases fromalmost500 km3 in Aprildownto less than50 km3in September.
summer ice thickness between Novaya Semlya and Severnaya Semlya. It has to be mentioned, however, that the interannual varianceof the original data ( 1972-1981 ) is considerable.
Simulated ice thicknessand concentrationsare comparedto
Mean
climatological data (Figures8 and 9) in orderto evaluatethe
summer
the Siberian
ice
ice thickness
coast.
is somewhat
Instead,
concentrations
for
ß .½::::!? -
underestimated
the model
similar
shows lar•er
(minimum
q•
....... ......-.-'3-:-::½:. .*', (,-.........
!::
