Formation and Maintenance of a Polynya in the Weddell Sea

Ralph Timmermann

Alfred-Wegener-Institut fur Polar- und Meeresforschung, Bremerhaven, Germany

Peter Lemke

Institut fur Meereskunde an der Universitat Kiel, Germany

Christoph Kottmeier

Institut fur Meteorologie und Klimaforschung, Universitat Karlsruhe, Germany

; accepted Received Short title:

Abstract. A dynamic-thermo dynamic sea ice-mixed layer mo del for the Weddell

Sea is complemented by a simple, diagnostic mo del to account for lo cal sea ice -

atmosphere interaction. To consider the atmospheric in uence on the o ceanic mixed

layer, the pycno cline upwelling velo city is calculated using the theory of Ekman

pumping. In several exp eriments, formation and conservation of a p olynya in the

Weddell Sea are investigated. Intrusion of heat into the lower atmosphere ab ove the

p olynya area is assumed to cause a thermal p erturbation and a cyclonic thermal wind

eld. Sup erp osed with daily ECMWF surface winds this mo di ed atmospheric forcing

eld intensi es o ceanic upwelling and induces divergent ice drift. Simulation results

indicate that in case of a weak atmospheric cross p olynya ow the formation of a

thermal wind eld can signi cantly extend the life-time of a large p olynya. The rep eated

o ccurrence of the Weddell Polynya in the years 1974 to 1976 thus app ears to b e an e ect

of feedback mechanismsbetween sea ice, atmosphere and o ceanic mixed layer.

Intro duction

During the Southern Hemisphere winter up to 20 million square kilometres of the

Antarctic o cean are ice-covered. Scattered patches of op en water within the sea ice

region consist of narrow leads of 1 to 10 km length and wide op enings called p olynyas.

Although they comprise only a few p ercent of the ice-covered area, leads and p olynyas in

the winter pack ice are imp ortant for a wide range of physical and biological pro cesses.

Here, the relatively warm water at freezing temp erature is in contact with the cold air

ab ove, thus causing large upward turbulent uxes of heat and water vap our.

Coastal p olynyas are a few hundred meters to 100 km wide and can b e explained

by prevailing winds and o cean currents at the margins of the Antarctic continent

Kottmeier and Engelbart, 1992. In the Weddell Sea, they contribute signi cantly to

the formation of deep and b ottom water which is the ma jor source of the b ottom water

of the world o cean Fahrbach et al., 1995.

O shore p olynyas are far less frequent. The most striking eventwas observed in

the Weddell Sea for the entire winters of 1974 { 1976 and is called \Weddell Polynya".

Within an area of ab out 300 000 km between 64 S and 69 S near the Greenwich

meridian, the ice concentration did not exceed 15 while the surrounding area was

nearly completely ice covered. According to the analysis of remote sensing data collected

by the Electrically Scanning Microwave Radiometer ESMR, the p olynya did not

emerge from an already existing ice cover; the ice cover rather grew around the p olynya

nally creating a secluded op en water area Zwally et al., 1981.

Di erent explanations for the o ccurrence of such an extended and long-lived p olynya

were given. The simplest mechanism which might cause the formation of a large area of

op en water would b e the exp ort of ice due to a divergent wind eld. However, the rate

of ice exp ort required to maintain nearly ice free conditions in the p olynya area would

b e of the order of meters p er second. This would require wind sp eeds averaging over 50

ms which is clearly to o large to b e reasonable Martinson et al., 1981.

From 1974 to 1976, the p olynya shifted westward with a mean velo city of 1 cm/s

which is consistent with the mean o cean current Carsey, 1980. From this p oint,

the phenomenon app ears to b e related to o ceanic pro cesses. Comparison of water

temp eratures within 200 to 2700 m depth b etween the years 1973 and 1977 shows

substantial co oling after the p olynya's o ccurrence. Gordon 1982 explained this by

extensive heat loss to the atmosphere, connected with low stability of the water column

and deep convection.

Martinson et al. 1981 prop osed that when surface co oling and ice formation

reduce the temp erature and increase the salinity of a shallow mixed layer, static

instability with intense vertical mixing can o ccur. The upwelled warm, salty deep water

can supply enough heat to melt the ice or prohibit its formation. To induce a shallow

mixed layer, Martinson et al. 1981 suggested that upwelling may raise the pycno cline

until convection can o ccur. Actually, Gordon and Hub er 1984 detected large warm

subsurface eddies which travelled from the east into the Weddell Sea. These warm water

cells lifted the mixed layer base signi cantly and increased the pycno cline temp erature,

leading to more intense entrainmentofwarm and saltywater.

Using a one-dimensional, purely thermo dynamic, sea ice - mixed layer mo del,

Lemke 1987 indicated that b oth a divergent drift of sea ice and the pycno cline

upwelling prop osed by Martinson et al. 1981 can induce the formation of a p olynyain

three subsequentyears:

A divergent ice drift reduces the mean ice thickness and the ice concentration

during the season of sea ice formation. This leads to a higher freezing rate and

thus to increased brine rejection.

Increased upwelling leads to enhanced entrainmentofwarm, salty deep water into

the mixed layer. While heat is lost to the atmosphere, salinity of the mixed layer

is preserved if the e ects of di usion and advection are neglected.

In b oth cases, increased density of the mixed layer induces deep convection with

entrainment heat uxes of the order of 150 W/m Lemke, 1987. To reduce the

formation of sea ice signi cantly, it is essential that during the p erio d of increased

upwelling the o cean is not yet ice-covered. In case of an existing ice cover, melting of

sea ice due to entrained heat creates a fresh water ux which stabilizes the upp er o cean

and prohibits convection.

In a similar simulation using a dynamic-thermo dynamic sea ice-mixedlayer mo del

Lemke et al., 1990 the system's resp onse was limited to the rst winter. Apparently,

advection of sea ice into the ice free area and subsequent melting i.e. stabilization of

the water column reduces the lifetime of a p olynya signi cantly. In order to maintain a

p olynyaover a longer p erio d a few years two prerequisites seem to b e necessary:

1. a suciently weak o ceanic strati cation, and

2. no in ow of ice into the p olynya.

This pap er aims to address the second p oint through examination of the interactions

between sea ice, o ceanic mixed layer and lower atmosphere which can explain the

formation and conservation of an op en o cean p olynya similar to the 1974 { 1976

observations. As already mentioned, the upward heat uxes over a p olynya far exceed

the values over a solid ice cover. Assuming that this causes lo cal warming of the lower

atmosphere, a low pressure system asso ciated with cyclonic thermal wind will b e forced

to develop. This will havetwo e ects on sea ice conditions Fig. 1:

1. The surface wind stress will induce a divergent ice drift.

2. The curl of the surface wind stress will according to the theory of Ekman

pumping enhance the o ceanic upwelling velo city.

Through the exp ort of ice out of the p olynya, this lo cal wind system may b e able

to maintain itself - as long as the surrounding area is ice-covered and the large-scale

pressure gradients are relatively small.

To fo cus on this item, a state-of-the-art sea ice mo del is coupled to a diagnostic

mo del of the thermal wind system whichisintro duced in the following section. Results

of the coupled simulations are presented and consequences for the maintenance of the

p olynya are discussed.

Mo del description

Sea ice-mixed layer mo del

The sea ice dynamics is based on the dynamic-thermo dynamic sea ice mo del

presented by Hibler 1979. Thermo dynamic ice growth rates are derived from a surface

energy balance Parkinson and Washington, 1979 using the Semtner 1976 zero-layer

approach for heat conduction. The incoming longwave and shortwave radiation uxes

are treated with standard formulations based on Idso and Jackson 1969 and Zillmann

1972. A prognostic snowlayer Owens and Lemke, 1990 considering the e ect of

snow/ice conversion Lepparanta, 1983; Fischer, 1995 is included.

To determine the o ceanic heat ux prognostically, the mo del is coupled to the

one-dimensional o ceanic mixed layer-pycno cline mo del of Lemke 1987, as describ ed

by Lemke et al. 1990. For this study, a diagnostic calculation of the pycno cline

upwelling velo city is added using the theory of Ekman pumping. If A denotes the ice

concentration, then the surface stress on top of the o cean is

~ =1 A~ + A~ ; 1

a iw

where ~ and ~ represent the atmospheric surface stress and the shear stress b etween

a iw

ice and water, resp ectively. The upwelling velo city is calculated from curl ~ using the

standard formula. Salt ux divergence in the upp er o cean is parameterized by restoring

to a constantvalue 33.8 psu, taken from the Southern Ocean Atlas with a time

constant of 1 a. Ocean surface currents remain xed. A more detailed description of the

coupled mo del is given by Timmermann 1997.

The mo del domain represents the Atlantic sector of the Southern Ocean including

the Weddell Sea Fig. 2. The spatial resolution is 1:6875 in zonal and 0:5625 in

meridional direction. A daily time step is applied. At continental b oundaries a no-slip

condition is prescrib ed, whereas ice mayenter or leave the mo del domain due to

advection at the o ceanic b oundaries of the mo del.

The atmospheric forcing is derived from daily values of air temp erature, relative

humidity and wind elds obtained from seven years 1986 to 1992 of the global

analyses of the Europ ean Centre for Medium Range Weather Forecasts ECMWF.

Spatially varying cloudiness is interp olated from monthly climatological means of the

International Satellite Cloud Climatology Program ISCCP; Rossow et al., 1991.

Precipitation is prescrib ed constant in time and space according to Parkinson and

Washington 1979. Prescrib ed o cean currents are constant in time and taken from

estimates based on observed sea ice drift Kottmeier and Sellmann, 1996 and mo del

simulations of Ross 1996. The o cean b oundary conditions for temp erature and salinity

b elow the pycno cline are taken from the Hydrographic Atlas of the Southern Ocean.

In order to reach an equilibrium cycle the mo del is integrated for seven years

starting with an ice-free o cean and using the daily forcing data of 1986 - 1992. The ice

conditions of Decemb er 31st, 1992 are used to restart the mo del. Within the following

seven integration years the o ccurrence of a p olynyaissimulated.

Lower atmosphere mo del

To allow for an atmospheric feedback, a simple description of a cyclonic wind

system induced by a large upward heat ux ab ove a p olynya has b een develop ed. The

thermal wind is derived from analytical functions describing the atmospheric wind and

temp erature p erturbation. Intro duced by Smith 1979 to describ e the in uence of

mountains on the atmosphere, the concept has b een used to derive a quasi-geostrophic

ow solution for the circulation over Antarctica Kottmeier and Stuckenb erg, 1986 as

well as for a paleo-climate study of the circulation over the Scandinavian ice sheet during

the last glacial maximum Kottmeier and Meyer, 1988. In the study presented here,

the mo del is extended to a description of cyclonic motion whichmayvary in time. The

mo di cations preserve the assumptions of hydrostatically balanced and quasigeostrophic

ow and consequently the thermal wind relation b etween the temp erature and ow

elds.

The temp erature pro le in the atmosphere's basic state is sp eci ed by the p otential

temp erature at the surface z = 0 and the Brunt-Vaisala-frequency N by

z = + z 2

G 0G

where is a mean p otential temp erature and g =9:8 m/s . Eq. 2 follows from the

de nition of N and assumes a stable basic state, i.e. > 0. Perturbation of the basic

thermal state is describ ed bya vertical displacement function x; y ; z of p otential

temp erature surfaces. Assuming geostrophic and hydrostatically balanced ow and

using the Boussinesq approximation, the thermal wind equations can b e written as

Smith, 1979

@ @u

= N 3 f

@z @y

@ @v

f = N 4

@z @x

In these equations f represents the Coriolis parameter; u and v are the geostrophic wind

comp onents in x- and y-direction, resp ectively. In the region of a warm temp erature

p erturbation, e.g. over a p olynya, isentropic surfaces are displaced downward. The

displacement at a certain x,y-p osition decreases with height; at distances far from the

sea surface, the displacement approaches zero. According to equations 3 and 4 this

is asso ciated with a cyclonic motion decaying with height Fig. 3.

Intro ducing a stream function , from which the velo cities are derived using

u = 5

v = ; 6

and integrating equations 3 and 4 over x and y , resp ectively, leads to

@ N

= x; y ; z : 7

@z f

According to Smith 1979 and Kottmeier and Stuckenb erg 1986 is chosen as the

function of a p oint source:

; 8 x; y ; z :=

where S represents the source strength and

2 2 2

z + z 9 r = x x +y y +

0 0 0

gives the distance from the lo cation x ;y ;z of the p oint source. For p ositive z the

0 0 0 0

p oint source is b eneath the sea surface. The vertical co ordinate z is stretched by

the factor N=f to takeinto account that the vertical scale is much smaller than the

horizontal scale of the ow. The vertical displacement function x; y ; z results as

x; y ; z = z + z : 10

4f r

Assuming that the lowest -surface is b ell-shap ed as given by

; 11 x; y ; 0 =

with h as the maximum displacement and

2 2 2

r := x x +y y 12

0 0

2 2

z ; 13 r :=

0 0

2 f

the source strength is obtained as

: 14 S = 4 h z

From 5, 6, 8 and 14, the velo city comp onents can b e derived as

h N y y h N x x

m 0 m 0

u = ; v = : 15

3 3

2 2

r r

z z

h h

+ + +1 +1 r r

2 2

0 0

r z r z

0 0

The values of N and f and the co ordinates x ;y of the p oint source need

0 0

to b e adequately chosen. The equations additionally contain the maximum vertical

displacement h of the -surfaces and the vertical co ordinate z of the p oint source of

m 0

the stream function as unknown parameters, where the value of z determines the width

of the b ell-shap ed -surfaces. The analytical mo del thus contains only very few free

parameters.

In extension of the earlier applications, h is calculated diagnostically from the

turbulent heat ux in the p olynya region. Assuming the conservation of heat, warming

of an atmospheric volume of thickness dz bya vertical heat ux H is describ ed by

dt = c T; 16

where T is the change of temp erature within the time interval t . The air densityis

denoted by , and c is the sp eci c heat at constant pressure. The warming of the lower

atmosphere is related to the surface heat ux by spatially integrating this equation:

z x

Z Z Z

x; y ; z T x; y ; z dz dy dx ; 17 H F t = c

0 p

x y

1 1

where H is the average surface heat ux in the time interval t . F represents the area

of turbulent heat exchange { the p olynya{ so H F t is the heat transfered to the

atmosphere. Using equations 2, 10 and 14, the equation of state of an idealized gas

and the de nition of p otential temp erature as well as the exp onential decay of pressure

with height, an equation for h as a function of H and t can b e derived Timmermann,

m 0

1997 from equation 17 as

0 1

max

m h z

Z Z

B C

+1 e

2RH Ft

B C

h = B r dr dz C ; 18

m h h

@ A

r c p

z p G0

G h

0 0

+ z + +1

g r z

1 1

where R is the sp eci c gas constant for dry air R = 287 J kg K , p is the sea

level pressure in the basic thermal state and z is the scale height of pressure decay.A

-surface intersecting the p ointx; y ; z in the disturb ed state is displaced from the

level z according to

z x; y ; z = z x; y ; z 19

G G

h +1

= z : 20

h G

+ +1

r z

From this implicit relation z is calculated using the Newton scheme of iteration

at each p ointx; y ; z of the integral 18, which is also calculated numerically. The

equations are combined using a simple scheme of iteration, starting with an estimated

new old

value of h and recalculating eq. 18 until jh h j < 10 m. The horizontal

m m

max

integration limit is chosen as R = 1000 km, which is larger by a factor of 1.5 than

the maximum radius of the p olynya. According to simulations of Law et al. 1992,

using a general circulation mo del, moist convection south of 50 S is restricted to a

maximum height of 3000 m. This value is applied as the limit z for vertical integration.

With these assumptions the distribution of heat emerging from the p olynyainto the

atmosphere is prescrib ed.

As the thermal wind system is induced by a temp erature p erturbation in the lower

trop osphere, the mo del is forced by the heat ux anomaly at the o cean - atmosphere

interface. Instead of taking the absolute value of the o cean - atmosphere heat ux, only

the latent and sensible heat ux di erence b etween the p olynya and the standard run

has to b e taken into account. The di erences in the long- and shortwave radiation ux

divergences b etween standard and p olynya run are negligible compared to the turbulent

heat uxes.

Now, only the value of t remains to b e xed. It corresp onds to the characteristic

residence time of an air mass over the p olynya and to the e ect of advection by the

mean \background" ow. For a prescrib ed heat ux at the o cean surface, t determines

the absorb ed heat and thus the magnitude of the temp erature p erturbation in the

lower atmosphere. The in uence of di erentvalues for t is investigated in the p olynya

simulations.

Adjusting the width of the b ell-shap ed -surfaces and thus of the thermal wind

system to the maximum size of the p olynya, the value of z was xed to 6000 m

corresp onding to r 600 km. The Brunt-Vaisala-frequency N is chosen as 0.01 s .

Equation 15 gives the thermal wind at each p oint x,y,z of the mo del, related to a

ctitious motionless height z = 1. Cho osing the upp er limit of the vertical integration

in 18 as 3000 m, it is assumed that the thermal p erturbation is con ned to the lower

3000 m of the atmosphere. The e ect of the thermal wind has to b e restricted to this

layer. The geostrophic wind at 10 m height consequently is calculated as the di erence

between the thermal wind comp onents at z = 10 m and z = 3000 m.

This simple atmospheric mo del provides the thermal wind as a function of the

vertical heat ux in the p olynya region, but it do es not account for e ects of the

large scale atmospheric ow. The investigated p olynya, however, is lo cated in the

region of the Circump olar Trough where migrating low pressure cells are linked with a

characteristic ow pattern. To derive a forcing dataset with realistic daily variability,

the calculated thermal wind is combined with daily means of the ECMWF wind eld by

sup erp osition of the geostrophic winds. As only the ECMWF 10 m-winds were available

for this study,we deduced the geostrophic winds from those. The description of the

planetary b oundary layer included in the ECMWF assimilation scheme could not b e

used for this purp ose as some of the variables used there are not available in our simple

approach. Instead, the basic relationship b etween ECMWF 10 m and geostrophic winds

was obtained by calculating geostrophic winds from the ECMWF sea surface pressure

eld and comparing these with the ECMWF 10 m-wind at the same p osition Fig. 4.

Geostrophic and 10 m-wind sp eed ju~ j and ju~ j and the deviation angle b etween the

g 10

10 m-wind ~u and the geostrophic wind ~u can b e reasonably related by

10 g

ju~ j

ju~ j = = +12 : 21

g ~u ~u

0:6

The geostrophic winds derived with these equations are sup erimp osed by the thermal

wind comp onents 15. From the resulting geostrophic wind eld, mo di ed 10-m winds

are obtained again using 21.

Generally the atmospheric variability is large on time scales shorter than those

relevant for sea ice. With resp ect to computing time this is the ma jor problem in

investigating the interaction of atmosphere and sea ice. The simple mo del applied

here is not able and not meant to replace an atmospheric circulation mo del but gives

an estimate on the velo city of a thermal wind system induced by a surface heat ux

anomaly. It contains a purely diagnostic description of the atmospheric resp onse and

is computationally cheap which easily allows an integration of the coupled mo del over

several years.

Standard exp eriment

In contrast to Lemke et al. 1990 who used spatially and temp orally constant

b oundary conditions for temp erature and salinity b elow the pycno cline, in this study the

observed spatial patterns of temp erature and salinity in the deep o cean are taken from

the Hydrographic Atlas of the Southern Ocean Olb ers et al., 1992. As an example for

the standard exp eriment the sea ice thickness and the entrainment heat ux in a typical

winter month are displayed in Fig. 5. The results indicate that the present mo del setup

yields a realistic description of the conditions in the Weddell Sea.

Polynya exp eriments

Similar to the study of Lemke et al. 1990, the p olynya exp eriments are initiated

by mo delling the entrainmentofwarm saltywater induced bywarm subsurface eddies

at a time when the mixed layer tends to deep en. For ftydays in the rst year of

integration, the pycno cline upwelling velo city is set to a value of 1.3 m/d whichisabout

ten times larger than the mean upwelling velo city in the standard exp eriment and the

mixed layer mo del b ottom temp erature T is increased by1 C. The preconditioned area

extends over 300 000 km b old marks in Fig. 2. This p erturbation is applied starting

from Julian day 52. In the annual cycle, this is the p erio d when the mixed layer depth

increases due to co oling and later brine rejection from new ice growth. At Julian day

52, the mixed layer depth has just exceeded 20 m and the o cean is not yet ice-covered

- which is essential to achieve the formation of a p olynya. In case a solid ice cover has

already formed at the time of the p erturbation, vertical heat ux would melt part of

the sea ice cover; the asso ciated fresh water ux would stabilize the strati cation of the

upp er o cean thus suppressing further entrainment heat ux.

Using the sea ice-mixed layer mo del without any atmospheric feedback the e ect on

the simulated sea ice distribution is quite similar to the p olynya exp eriment of Lemke

et al. 1990. Due to the increased upwelling, warm and saltywater is entrained into

the mixed layer. While salt remains in the o cean, heat is lost to the atmosphere. As a

result of the reduced mixed layer depth, mixed layer density increases rapidly.Thus,

deep convection is induced nally leading to a mixed layer depth of almost 500 m. The

increased entrainment heat ux up to 500 W m for a short time interval prevents the

formation of sea ice for the entire freezing p erio d. The extent of the emerging p olynya

Fig. 6 shrinks in the course of winter due to the advection of sea ice but it keeps b eing

op en until the sea ice cover of the whole area retreats. In the second year of integration,

a smaller p olynya o ccurs, but at Julian day 205 the area of op en water is closed due to

the advection of sea ice. After veyears of integration the system has again reached the

annual cycle of the reference exp eriment.

At this p oint, the question arises whether the atmospheric feedback is able to

prohibit the imp ort of sea ice into the p olynya. As we will see, the answer dep ends on

the assumptions concerning the residence time t .Four exp eriments were p erformed

with t = 7 d, 10 d, 12 d and 14 d which represents a mean atmospheric cross-p olynya

owbetween 1.2 and 0.6 m/s, resp ectively.

In the simulations with t = 7 d and t = 10 d, the lifetime of the p olynya do es not

di er signi cantly from the p olynya exp eriment without any atmospheric feedback. In

contrast to the ice-mo del-only integration, however, the spatial extent of the p olynya

remains constant or in case of t =10deven grows during the course of winter, as

the observed Weddell p olynya did. The cyclonic thermal wind system with typical

maximum velo cities of 10 m/s apparently is not able to prevent the imp ort of sea ice. As

the net growth rate within the p olynya area indicates, thermo dynamic sea ice formation

in the p olynya area starts some time after the grid cells have b een covered by imp orted

sea ice. This proves that advection is the main feature reducing the lifetime of the

p olynya.

Fig. 7 displays time series of the p olynya area de ned as the area of grid cells

with an ice concentration b elow 15 , the ice thickness in the p olynya center and the

maximum thermal wind sp eed taken from the exp eriment with t = 12 d. As an e ect

of the atmospheric feedback, the p olynya o ccurs in three subsequentyears. The typical

maximum sp eed of the cyclonic thermal wind system is 15 m/s in the rst twoyears

and 6 m/s in the third year. App earently, in the latter case the cyclonic owistooweak

to prevent the imp ort of sea ice.

Setting t = 14 d, the typical maximum sp eed of the cyclonic wind system rises

to 18 m/s and thus is of the same order of magnitude as the cyclones contained in the

large scale ECMWF-wind eld. The turbulent uxes of latent and sensible heat which

are resp onsible for its existence exceed those of the reference exp erimentby a factor

of 10 and are consistent with measurements over Arctic leads Andreas et al., 1979.

Latent and sensible heat uxes are of the same order of magnitude Fig. 10

In each winter p erio d of this seven year integration, a p olynya with an areal extent

of of 200 000 to 300 000 km o ccurs. In the fth winter, ice cover grows up to a thickness

of 40 cm; in all the other years the o cean remains virtually ice-free.

Besides the mean ice concentration in Septemb er which is the time of maximum

ice extent of the seventh year of integration, Fig. 8 top displays the mean 10 m-wind

sp eed. In addition to the typical large scale pattern, a distinct cyclonic eddy is visible.

This leads to a divergent ice drift Fig. 8 { b ottom and { via Ekman pumping and

increased entrainment { to an enhanced vertical o ceanic heat ux g. 9. Together,

these features prevent b oth the freezing and the imp ort of sea ice.

Fig. 9 and 10 present the evolution of the p erturb ed system. During the winter

p erio d, the thermal wind system is strong enough to keep the preconditioned o cean area

uncovered. As an implication of the entrainmentofwarm, salty deep water, not only

the heat but also the salt content of the o ceanic mixed layer is increased. As a result,

the mixed layer salinity exceeds the values of the reference exp eriment and those of

the not-preconditioned vicinityby ab out 0.1 psu { even in the summer p erio d. At the

b eginning of the following winter, the mixed layer density in the p olynya area is still

increased so that atmospheric co oling alone is able to induce intensi ed entrainment

which delays the formation of sea ice. By this time, the undisturb ed surrounding of the

preconditioned area is already ice-covered, so that a thermal wind system forms again.

Tokeep this pro cess running, the mean atmospheric cross-p olynya ow has to

be weaker than ab out 1 m/s during the entire winter season. One would exp ect that

this situation should b e correlated with small surface wind velo cities. Under such

conditions, lo cal anomalies like the thermal wind system induced by the increased

turbulent heat ux in the p olynya area would b e more signi cant compared with the

large-scale \background" wind system. On the other hand, turbulent heat uxes at the

o cean surface are assumed to b e prop ortional to the 10 m-wind sp eed. Reduced wind

sp eed consequently leads to reduced turbulent heat ux and thus to a reduced thermal

p erturbation of the lower atmosphere. These e ects will b e discussed in the next section.

Sensitivity exp eriments

Reduced ECMWF wind sp eed

Tocheck whether weak surface winds would promote or complicate the conservation

of a p olynya, two p olynya exp eriments with a mo di ed ECMWF wind eld were

p erformed and compared with the \normal" p olynya exp eriment. To construct a

ctitious surface wind with realistic distribution but reduced sp eed, the ECMWF wind

was multiplied with a factor of 0.75 and 0.5, resp ectively.

Whereas in the rst case factor 0.75 no signi cant alteration is to b e recorded,

p olynya extent and lifetime are reduced signi cantly in the second case factor 0.5.

As an e ect of the reduced ECMWF-wind sp eed, the strength of the thermal wind

system and its impact on the p olynya are reduced. In addition to that, the increase of

the o ceanic mixed layer density at the b eginning of winter is smaller. Consequently,

in this exp eriment, the mixed layer depth in the preconditioned area do es not exceed

190 m and the entrainment heat ux exceeds 100 W/m only for short time intervals.

Hence, lifetime and extent of the p olynya are reduced even compared with the p olynya

exp eriment without any atmospheric feedback.

Increase of p olynya area

As the absorb ed amount of heat in the atmosphere is, according to Equation 17,

prop ortional to t as well as to the area F of enhanced turbulent heat exchange, one

would exp ect that an expansion of the preconditioned area would lead to a longer

lifetime of the p olynya. When the extent of the preconditioned area is doubled, however,

the extent of the p olynya do es increase as well as the strength of the thermal wind

system, but the p olynya's lifetime changes only by a few days. This leads to the

conclusion that the extended lifetime of the p olynya in the exp eriments with t =12d

and t = 14 d is not only a result of the intensi ed cyclonic wind but also of the

increased resp onse time of the lower atmosphere which is caused by the longer time

interval over which the surface heat ux is averaged.

This is con rmed by a further sensitivity exp eriment. An arbitrary increase of the

cyclonic wind by 50 leads to an increased p olynya area but not to a longer p olynya

lifetime.

Mo di ed initialization

As the ECMWF wind is the only consistent forcing eld with interannual variations,

it seems to b e p ossible that the results of the exp eriment dep end on the year of the

p olynya initialization. Several exp eriments with di erentyears of initialization as well as

mo del integrations with the wind eld of the same year rep eated seven times, however,

reveal no signi cant di erence.

Parameterization of turbulent heat uxes

Toinvestigate the sensitivity concerning di erent parameterizations of the turbulent

heat exchange b etween o cean and atmosphere, several exp eriments with di erentvalues

for t were p erformed setting the exchange co ecients, which in the exp eriments so

far were C = C =1:75 10 according to Parkinson and Washington 1979, to

s l

C = C =1:3 10 . As describ ed ab ove, reduced heat exchange b etween o cean and

s l

atmosphere leads to a smaller thermal p erturbation and thus to a weaker thermal wind.

Consequently, the result is a reduced impact on the sea ice cover. In this case t has to

b e 14 d or larger to achieve a signi cant prolongation of the p olynya lifetime.

Width of the cyclonic wind system

As already explained, z is a measure of the width of the b ell-shap ed isotherms in

the thermally p erturb ed lower atmosphere. Higher values of z lead to larger extent

and decreased strength of the cyclonic wind system. The exp eriments describ ed ab ove

were p erformed with z = 6000 m R 600 km. Varying z between 5000 and 8000 m

0 0 0

reveals no signi cantchanges. For z > 8000 m R > 800 km, however, the weakening

0 0

of the cyclonic wind system reduces its impact on the sea ice cover.

Conclusions

Considering the interaction b etween sea ice and atmosphere in a simpli ed way,

exp eriments with a dynamic-thermo dynamic sea ice - mixed layer mo del indicate that

the development of a self-conserving low pressure system is able to promote the rep eated

o ccurence of a large p olynya in the Southern Ocean.

Initialized by a preconditioning of the o ceanic mixed layer, the p erturbation

develops in time bywarming the lower atmosphere ab ove the p olynya during the ice

covered season and by partly conserving an increased mixed layer salinity during the

summer p erio d.

To allow for this pro cess, the mean - \background" - atmospheric cross-p olynya

ow apparently has to b e smaller than ab out 1 m/s t > 10 d. This situation, which

can b e regarded as a p ersistent preconditoned atmosphere, is not related to slow surface

winds: The sensitivity exp eriments indicate that weak surface winds do not favour the

o ccurence and maintenance of a p olynya.

Whereas at rst glance this seems to b e a contradiction in itself, ECMWF analyses

reveal that the vector mean of surface wind velo cities south of 65 S in the winter season

is only a few meters p er second Kottmeier et al., 1996. The 10 m monthly mean wind

sp eed ranges from 0 m/s to ab out 7 m/s and the mean vector generally has an eastward

comp onent north of 65 S and a westward comp onent south of 70 S in the region of

interest. This re ects that the depressions passing along have opp osite wind directions

in the di erent sectors, which partly average out in monthly means. The transition

zone b etween the strong, p ersistent mean westerly ow at ab out 62 S and the coastal

easterlies south of 68 Smay b e a place where the advection may b ecome rather weak

from time to time - and this actually is the place where the Weddell Polynya o ccurred.

Assuming a situation like this, increased vertical heat uxes in the op en water area

induce a lo cal warming of the lower atmosphere. The corresp onding cyclonic thermal

wind system mo di es the sea ice drift into a divergent motion and increases the vertical

o ceanic heat ux due to Ekman pumping. In case a cyclonic thermal wind of ab out

15 m/s or higher is induced, a large scale p olynya is maintained over the entire winter

p erio d.

As an e ect of Ekman pumping, not only heat but also salt is injected into the

o ceanic mixed layer. As time scales of transp ort in the o cean are long, the increased

mixed layer salinity is nearly p ersistentover the summer p erio d. Hence, at the b eginning

of the following winter the mixed layer density still is high so that even atmospheric

co oling leads to intense entrainment. This e ect maintains the p erturbation over several

years - despite the short time scales that are relevant for the atmosphere.

A further requirement for the formation of a secondary low pressure system is

a suciently large p olynya area. Evidently, the quantitiy of heat injected into the

atmosphere ab ove an area of op en water grows with its extension. In the case when

the p olynya area is to o small, thermal p erturbation of the lower atmosphere in the

resp ective region will b e weak, so that no signi cant in uence on the sea ice distribution

is to b e exp ected. In the p olynya exp eriments, the \critical extension" is of the order

of 100 000 km ; as the interaction strongly dep ends on the dynamic state of the

atmosphere, it is imp ossible to give an exact value.

In contrast to the observations of 1974 to 1976, the simulated p olynya do es not

move eastward in the course of time. This can easily b e explained, as the onedimensional

mixed layer mo del do es not contain any horizontal advection. The movement of the

preconditioned area with the mean o ceanic ow therefore cannot b e considered. As

the area of increased turbulent heat ux cannot move either, the lo cation of the

atmosphere's thermal p erturbation in the mo del is xed.

If the numerical description of b oth the mixed layer and the atmosphere were

mo di ed in a suitable way,we exp ect that the motion of the observed Weddell p olynya

could b e describ ed as well: In reality, sea ice and o cean move with di erent sp eed in

di erent directions but, as the observations indicate, the lo cation of the sea ice anomaly

is given by the p osition of the preconditioned region of the o cean. As the lo cation of

the atmospheric thermal p erturbation is connected with the op en water area, lo cal

interaction b etween atmosphere, o cean and sea ice would not change. Concerning the

pro cesses which are relevant for formation and maintenance of a p olynya, ma jor changes

are not exp ected as long as the time scale of di usion do es not change.

The p olynya lifetime after b eing initialized bya warm eddy dep ends on the large

scale atmospheric circulation: Only in case the mean cross-p olynya owisunusually

small b elow 1 m/s and thus distinctly smaller than in the quasi-climatology of the

ECMWF-analyses, heat injection through an area of op en water is able to create a

suciently large thermal p erturbation in the lower atmosphere.

The question arises, whether a large p olynya itself can create these conditions by

mo difying the general atmospheric circulation. An answer could b e found replacing the

simple diagnostic mo del used here by an atmospheric circulation mo del with suciently

ne resolution. A more complete numerical description of the o cean including horizontal

advection would allow a more detailed investigation of the o ceanic part of p olynya

formation and maintenance.

Acknowledgements

The authors would like to thank W. Cohrs, H. Fischer, M. Harder, M. Kreyscher

and C. Lichey for preparing the sea ice mo del forcing elds. L. Sellmann provided

the geostrophic and 10 m-wind data whichwere used to derive the empiric relation

between geostrophic and surface winds. Atmospheric forcing data were derived from the

ECMWF analyses whichwere recieved via the German Weather Service. This is AWI publication no. xxx.

References

Andreas, E. L., C. A. Paulson, R. M. Williams, R. W. Lindsay, and J. A. Businger,

The turbulent heat ux from Arctic Leads, Boundary Layer Meteorol., 17:57-91, 1979.

Carsey, F. D., Microwave observations of the Weddell p olynya, Mon. Wea. Rev.

108, 2032-2044, 1980.

Fahrbach, E., G. Rohardt, N. Scheele, M. Schroder, V. Strass, and A. Wisotzki,

Formation and discharge of deep and b ottom water in the northwestern Weddell Sea,

Journal of Marine Research, 53, 515-538, 1995.

Fischer, H., Vergleichende Untersuchungen eines optimierten dynamisch-

thermo dynamischen Meereismo dells mit Beobachtungen im Weddellmeer, Berichte zur

Polarforschung 166, ISSN 0176-5027, Alfred-Wegener-Institut AWI, Bremerhaven,

1995.

Gordon, A., Weddell deep water variability, J. Mar. Res., 40:199-217, 1982.

Gordon, A. L., and B. A. Hub er, Thermohaline strati cation b elow the Southern

Ocean sea ice, J. Geophys. Res., 89:641-648, 1984.

Hibler, W. D., I I I., A dynamic thermo dynamic sea ice mo del, JPO, 94:815-846,

1979.

Idso, S. B., and R. D. Jackson, Thermal radiation from the atmosphere, JGR,

74:5379-5403 , 1969.

Kottmeier, Ch., and H.-U. Stuckenb erg, A Quasi-Geostrophic Flow Solution for the

Circulation over Antarctica, Beitr. Phys. Atmosph. Vol. 59, No. 4, 1986.

Kottmeier, Ch., and H. H. Meyer, Ein einfaches analytisches Mo dell zur Darstellung

der Zirkulation in Europa im Weichsel-Ho chglazial, Erdkunde, Band 42, 1988.

Kottmeier, Ch., and D. Engelbart, Generation and atmospheric heat exchange of

coastal p olynyas in the Weddell Sea, Boundary Layer Meteorology, 60:207-234, 1992.

Kottmeier, Ch., and L. Sellmann, Atmospheric and o ceanic forcing of Weddell Sea ice motion, JGR, 101, C9, 20809-208 24 , 1996.

Kottmeier, Ch., S. Ackley, E. Andreas, D. Crane, H. Ho eb er, J. King, J. Launiainen,

D. Limb ert, D. Martinson, R. Roth, L. Sellmann, P.Wadhams, and T. Vihma, Wind

and ice motion statistics in the Weddell Sea, World Meteorological Organisation,

Technical Do cument, 1/1997, 1997.

Law, R., I. Simmonds, and W. F. Budd, Application of an atmospheric tracer

mo del to high southern latitudes, Tellus, 44B:358-370, 1992.

Lemke, P., A coupled one-dimensional sea ice - o cean mo del, JGR, 92C12:

13164-1317 2, 1987.

Lemke, P., W. B. Owens, and W. D. Hibler, I I I, A coupled sea ice - mixed layer -

pycno cline mo del for the Weddell Sea, JGR, 95C6:9513-95 25, 1990.

Lepparanta, M., A growth mo del for black ice, snow ice, and snow thickness in

subarctic basins, Nordic Hydrology, 14:59-70, 1983.

Martinson, D. G., P. D. Killworth, and A. L. Gordon, A Convective Mo del for the

Weddell Polynya, J. Phys. Oceanogr. 11, 466-487, 1981.

Olb ers, D., V. Gouretski, G. Seiss, and J. Schr"oter, Hydrographic Atlas of

the Southern Ocean. Alfred Wegener-Institut f"ur Polar- und Meeresforschung,

Bremerhaven, 1992.

Owens, W. B., and P. Lemke, Sensitivity studies with a sea ice-mixed layer-pycno-

cline mo del in the Weddell Sea, JGR, 95C6:9527-9538 , 1990.

Parkinson, C. L., and W. M. Washington, A large-scale numerical mo del of sea ice,

JGR, 84C1:311-337, 1979.

Ross, H., Ein massenerhaltendes hydrographisches Inversmo dell mit

verallgemeinerter Vertikalko ordinate angewandt auf den Atlantischen Sektor des

Sudlichen Ozeans, Ph.D. theses, Univ. Bremen, Germany in preparation, 1996.

Rossow, W.B. et al., ISCCP Do cumentation of Cloud Data, WMO/TD-No. 266,

1991.

Semtner, A. J., Jr., A mo del for the thermo dynamic growth of sea ice in numerical

investigations of climate, JPO, 63:379-389, 1976.

Smith, R. B., The in uence of mountains on the atmosphere, Advances in

Geophysics 21, S. 87-230, 1979.

Timmermann, R., Entstehung und Erhaltung einer Polynja im Weddellmeer,

Berichte aus dem Fachb ereichPhysik, Rep ort 77, Alfred-Wegener-Institut AWI,

D-27515 Bremerhaven, 1997.

Zillman, J. W., A study of some asp ects of the radiation and heat budgets of

the southern hemisphere o ceans, in: Meteorological study,vol. 26, 526pp., Bureau of

Meteorology, Dept. of the Interior, Canb erra, Australia, 1972.

Zwally, H. J., J. C. Comiso, and C. L. Parkinson, Satellite-Derived Ice Data Sets

No.1: Antarctic Monthly Average Microwave Brightness Temp eratures and Sea Ice

Concentrations 1973-1976, NASA Technical Memorandum 83812, 1981. 26

L ug

ρ Ue

Figure 1. Section through a secondary low pressure system over a southern hemisphere

p olynya, including the geostrophic wind ~u , the Ekman transp ort U in the o ceanic and

g e

the atmospheric b oundary layer and the resulting vertical velo cities w . e 27

-20 0 -40 20 -60

-80

-45

-65

-85

Figure 2. Mo del domain. Bold marks indicate the area of preconditioned mixed layer. 28

Θ n+1 Θ n

Figure 3. Cyclonic circulation due to lo cal warming of the lower atmosphere southern hemisphere. 29

ECMWF Wind at (-67.5, -7.875), Aug.94 20

5 10m ECMWF Wind Speed [m/s]

0 0 10 20 30 40 Geostrophic Wind Speed [m/s], calculated from ECMWF sea surface pressure field

180

-90 10m ECMWF Wind Direction [deg]

-180 -180 -90 0 90 180

Geostrophic Wind Direction [deg], calculated from ECMWF sea surface pressure field

Figure 4. Comparison of sp eed top and angle b ottom of the ECMWF-10m-wind

and the geostrophic wind calculated from the ECMWF sea surface pressure eld at the

same p osition for the analyses of August 1994. As an e ect of friction in the planetary

b oundary layer, wind sp eed slows down by a factor 0.6 and wind direction turns 12 to

the right. 30

20˚W 0˚

40˚W 20˚E

60˚W

40˚E

50˚S

60˚S 2.8 2.4 2.0 70˚S 1.6 1.2

80˚S 0.8 0.4 [m] 0.0 20˚W 0˚

40˚W 20˚E

60˚W

40˚E

50˚S

60˚S 50 40 30 70˚S 20 10 80˚S 0

[W/m2] -10

Figure 5. Ice thickness top and entrainment heat ux at the mixed layer base b ottom

in August of the second year of the reference integration. Contour intervals are 0.4 m

and 10 W/m , resp ectively. 31

20ÊW 0Ê

40ÊW 20ÊE

60ÊW 40ÊE

50ÊS

60ÊS 100 80 60 70ÊS 10 cm/s 45 30 80ÊS 15

Figure 6. Ice concentration and mean drift in July of the rst year of integration in a

p olynya exp eriment without atmospheric feedback. 32

a) Polynya Area 6·105

4·105

[km2] 2·105

0 0 365 730 1096 1461 1826 2191 2557 b) Mean Ice Thickness in Polynya Center 1.5

1.0 [m] 0.5

0.0 0 365 730 1096 1461 1826 2191 2557 c) Maximum Thermal Wind Velocity 20

10 [m/s] 5

0 0 365 730 1096 1461 1826 2191 2557

Days of Integration

Figure 7. Time series of a the p olynya area, b the ice thickness in the p olynya

center and c the maximum thermal surface wind taken from the p olynya exp eriment

with atmospheric feedback and t := 12 d solid lines. The dashed line in a indicates

the p olynya extent in the p olynya exp eriment without any atmospheric feedback. The

dotted line in b indicates the ice thickness at the resp ective p osition in the reference

exp eriment. 33

20ÊW 0Ê

40ÊW 20ÊE

60ÊW 40ÊE

50ÊS

60ÊS 100 80 60 70ÊS 10 m/s 45 30 80ÊS 15 0 20ÊW 0Ê

40ÊW 20ÊE

60ÊW 40ÊE

50ÊS

60ÊS 100 80 60 70ÊS 10 cm/s 45 30 80ÊS 15

Figure 8. Mean ice concentration and 10 m-wind sp eed top, mean ice concentration

and sea ice drift b ottom in Septemb er of the seventh year of integration taken from a

p olynya exp eriment including atmospheric feedback assuming a mean atmospheric cross

p olynya ow of 0.6 m/s. t = 14 d. 34

a) Polynya Area 6·105

4·105

[km2] 2·105

0 0 365 730 1096 1461 1826 2191 2557 b) Mean Ice Thickness in Polynya Center 1.5

1.0 [m] 0.5

0.0 0 365 730 1096 1461 1826 2191 2557 c) Upwelling Velocity 5 4 3 2 [m/d] 1 0 -1 0 365 730 1096 1461 1826 2191 2557 d) Entrainment Heat Flux 600

400

[W/m2] 200

0 0 365 730 1096 1461 1826 2191 2557

Days of Integration

Figure 9. Time series of a p olynya extent, b ice thickness in the p olynya center, c

pycno cline upwelling velo city and d entrainment heat ux into the o ceanic mixed layer

taken from the p olynya exp eriment with atmospheric feedback and t := 14 d solid

lines. The dashed line in a indicates the p olynya extent in the p olynya exp eriment

without any atmospheric feedback. The dotted lines in b to d indicate the resp ective

values of the reference exp eriment. 35

a) Turbulent Vertical Heat Flux 400 200 0 -200

[W/m2] -400 -600 -800 0 365 730 1096 1461 1826 2191 2557 b) Sensible Heat Flux 100 0 -100 -200 [W/m2] -300 -400 0 365 730 1096 1461 1826 2191 2557 c) Latent Heat Flux 0 -50 -100 -150 [W/m2] -200 -250 0 365 730 1096 1461 1826 2191 2557 d) Maximum Thermal Wind Speed 25 20 15

[m/s] 10 5 0 0 365 730 1096 1461 1826 2191 2557

Days of Integration

Figure 10. Time series of a vertical turbulent heat ux negativevalues denote a

ux from the o cean to the atmosphere, b sensible and c latent heat ux and d

the maximum thermal wind sp eed taken from the p olynya exp eriment with atmospheric

feedback and t := 14 d solid lines. The dotted lines in a to c indicate the resp ective

values of the reference exp eriment.