JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 86, NO. C5, PAGES 4193-4197, MAY 20, 1981
Seasonalityof SouthernOcean Sea Ice
ARNOLD L. GORDON
Lamont-DohertyGeological Observatory of ColumbiaUniversity, Palisades, New York 10964
The seaice coverof the SouthernOcean undergoes a rapid decreasefrom mid-Novemberto mid-Janu- ary. The atmosphere-to-oceanheat flux is insufficientto accountfor thismelting, even in the presenceof the high percentageof open water characteristicof the SouthernOcean seaice. It is estimatedthat sea-air heatexchange in the 60ø to 70øSzone can account for roughly50% of the requiredspring heating. The remaindermust be suppliedby the relativelywarm deep water, residing below the SouthernOcean pyc- nocline.Deep-to-surface water heat flux is accomplishedby upwellingof the pycnoclinedue to the re- gionalEkman divergenceof the surfacelayer and by cross-pycnoclinemixing. It is estimatedthat the re- quiredvertical mixing coefficient for the pycnoclineis 1.5 cm2/s.This valueis consideredas realisticin view of the relativelyweak pycnocline which is a consequenceof low levelsof freshwater input to the surfacelayer. It is suggestedthat the magnitudeof freshwater input is a majorfactor in determiningthe degreeof seasonalityof SouthernOcean sea ice.
1. INTRODUCTION new ice. In this way new seaice forms in the interior of the sea The sea ice surrounding Antarctica undergoeslarge sea- ice fieldsrather than being accretedto the outer edge.While sonalchanges in areal extent.This is evidentfrom both histor- some addition at the edge may be expected,the northward ical observations and recent satellite sensing [Mackintosh, migration of the ice edge may be primarily governedby the 1946, 1972;Heap, 1964;Zwally and Gloersen,1977; S. F. Ack- Ekman velocity. ley, unpublishedmanuscript, 1980]. The Zwally et al. [1979] As ice is carried northward, the local heat balance would review of the sea ice areal cover as observedby satellitesfor induce melting, and a steady-stateice cover is attained, as 1973-1975 for greater than 15% and greater than 85% (repro- may be the casefrom August to October. During this period duced as Figure 1) show a sea ice maximum cover in Septem- ice forming in the south is balancedby melting in the north. ber and Octoberof approximately20 x 106km 2, with a mini- Using Ekman drift velocitiespresented by Taylor et al. [1978], mum cover in February of approximately 3 x 106 km2. the sea ice northward transport across60øS amountsto 9.62 x Therefore,17 x 106km 2 of seaice, an arealarger than Antarc- 10•3 W polewardheat flux, for eachmeter thicknessof ice. tica, is within the seasonalsea ice zone. As Zwally et al. (Fig- Ice melting in the springwould also be influencedby Ek- ure 1) indicate,much of this cover(about two thirds) has con- man divergence.The atmosphere-to-oceanheat flux would be centrations between 15% and 85% of full ice concentration. If greatestwithin the low albedo leadsgenerated by the diverg- the ice were compactedto 100%concentration, its area would ence. The leads act as heat collectors, which accelerate melt- be approximatelythree fourthsof the presentcover with leads ing of the surroundingpack. Langleben[1972] also stresses the (C. Parkinson,personal communication, April 1980). importanceof radiational heatingwithin leadsin regardto sea The growth and decayof the seaice covershown in Figure ice melting in the Arctic, but points out that radiational heat- 1 for both the 15%and 85% concentrationcurves are not sym- ing of the sea ice should not be neglected. metric about the maximum ice cover. The growth rate is less The ability of the sea-airheat flux to accountfor the spring than the decayrate by nearly a factor of two. The major por- retreatof the seaice is now tested.The ice chartsof the Navy- tion of the decay occursin the period mid-November to mid- NOAA ice center indicate that during the mid-November to Januarywith a decreasefrom 17.5x 106km 2 to 6.5 x 106km 2. mid-Januaryperiod sea ice virtually disappearsover the deep Decay of the greaterthan 85% coveroccurs earlier (mid-Octo- ocean.The additional ice melt from mid-Januaryto mid-Feb- ber to mid-November) and is even more spectacularin areal mary is mainly over the continental margins of Antarctica. decrease,retreating from the near annual maximum cover to The major exceptionis the 0.8 x 106km 2 (aboutone third of only 25% of that in the one month period. which is over the deep ocean)field of perennial seaice east of the Antarctic Peninsula. Walsh and Johnson[1979] area versusmonth plot for Arctic Southern Ocean seasonal sea ice is estimated at 1.0 to 1.5 m sea ice cover (their Figure 3) showsthat the ice growth rate is faster than the ice retreat rate. This contrast with the Southern [Untersteiner,1966]. The heat requiredto melt the averagesea Ocean situation suggeststhat fundamental differences exist ice thicknessof 1.25 m with a latent heat of fusionof 292 J/g between the heat budgetsof Arctic and Antarctic sea ice. (Table 4.3 from Neumannand Pierson[1966]) is 3.3 x 108J/ During the growth period for Southern Ocean sea ice a m2. Sincemost of the openocean ice is removedin the 60-day simple Ekman divergencemodel yields realistic results[Gor- period from mid-November to mid-January,the daily heat re- don and Taylor, 1975; S. F. Ackley, unpublishedmanuscript, quirementis 552 x 104J/m 2 d or 64 W/m 2. This representsa 1980]. The conceptis as follows: The cyclonicwind field in- minimum heat requirement since a higher value would be ducesa divergentEkman transportin the surfacelayer of the necessaryif significant warming of the surface water occurs ocean. Sea ice on the ocean also diverges, thus generating before mid-January or if the heat required to melt icebergsis open water or leads.In the winter when the ocean-atmosphere taken into account.While somewarming is observed,and ice- heat flux is directedtowards the atmospherethe leads fill with berg melting in early summeroccurs, both are neglectedin the calculations.The question is: Can the atmosphere-to-ocean heat flux provide enough heat to explain the sea ice decay Copyright¸ 1981by the AmericanGeophysical Union. even if no other demands are placed on the available heat?
Paper number 80C 1623. 4193 0148-0227/81/080C- 1623501.00 4194 GORDON: SEASONALITYOF SOUTHERN OCEAN SEA ICE
25. tion. Trenberth[1979], on the other hand, determinesa 100 x 10•3 W flux across60øS, closer to the open oceanvalue and i• 20 about twice that for the realistic sea ice condition (Table 4). f 15- Newton[1972, Appendix C] gives29.3 x 10•3 W oceanheat _o _ flux across60øS, approximatelytwice Hastenrath'svalue, half of the Table 4 value for realistic ice cover, and one third of Trenberth's values. The one order of magnitude range in heat flux estimates across60øS indicatessome of the difficultiesin determining O FMAMJ d A SOND this importantnumber. Before it is possibleto identify the pri- MONTH mary mechanismresponsible for meridional heat flux in the Fig. l. Areal extent of Southern Ocean waters with sea ice con- Southern Ocean, it is necessaryto establish reliable sea-air centration at least 15% and with sea ice concentration at least 85% in heat flux estimates as a function of latitude for the Southern 1973, 1974, and 1975 [Zwally et al., 1979]. Ocean. In this presentstudy the mode in which the heat flux into the 60ø-70øSzone is accomplishedis important only in that it is assumed the annual heat loss of the surface water to 2. SEA-AIR HEAT EXCHANGE IN THE BAND 60ø-70øS the atmosphereis balancedby vertical heat transferacross the Estimates for sea-air heat exchange for the 60ø-70øS lati- pycnocline,rather than lateral poleward heat transferwithin tude band are made usingmonthly climaticdata (Table 1) for the surfacelayer itself. Since the Ekman drift is directed to- ice free (Table 2), full ice and partial ice cover(Table 3). Ad- wards the equator, this assumptionis justified. mittedly, the errors associatedwith these calculations are 3. DISCUSSION large, due to inadequate meteorological data set, use of monthly meansrather than shorterterm means,uncertainty in It is unlikely that the atmosphere-to-oceanheat flux given the exchangecoefficients for the non-radiative heat flux terms in Tables 2 and 3 is adequateto accountfor the massiveice and from uncertaintyin the exact amount of seaice insulation melt from mid-November to mid-January. For the realistic of the ocean(the R term in Table 3) and ratio of ice cover to ice coverthe averageheat gain for thisperiod is 34 W/m2, open ocean (P value in Table 3). slightlyover 50% of the requiredminimum heat flux. Even the The requiredpoleward oceanic heat across60øS for balance higherestimate of 47 W/m 2 for a no-icecondition is not suf- of oceanheat lossbetween 60ø-70øS (16.8 x 1016cm 2) is de- ficient and the full ice covervalue of 7 W/m 2 is grosslyin- termined for each of the ice concentrationsituations (Table sufficient.This suggeststwo things:leads induced by the gen- 4). Oceanic heat flux across60øS estimated by Hastenrath eral Ekman divergence must be an important factor by [1980]is 13 x 1013W. Hastenrath'sheat flux estimateacross increasing ocean warming to a point where atmosphere-to- 53øS,which marks the mean circumpolarposition of the polar ocean heat flux can play more than a minor role in the spring front, is 35 x 1013W (his Figure 9), similarto the value sug- ice cover retreat and that an additional source of heat is gestedby Gordon[1975]. It is surprisingthat his 60øSvalue is needed to balance the surfacewater heat balance during the so much lower than his 53øSvalue. It implieslarge oceanheat springmelting as well as on an annual basis(both valuesare lossin the 53ø-60øSzone (13 W/m 2) where other estimates about 30 W/m2). [see Taylor et al., 1978] actually suggesta small annual heat Heat flux into the surfacelayer from the sub-pycnoclinewa- gain by the oceanin this zone. His heat flux value across60øS ter is the probablesource of this heat. Sub-pycnoclinewater is is low by a factor of 4 comparedto the value given in Table 4 significantlywarmer than the surfacewater (Figure 2) and for realistic ice cover, but is close to the full ice cover condi- forms an effectiveheat (and salt) reservoir.Deep water heat can be introduced into the surface layer by the general up- welling of the pycnoclineexpected from the climatic Ekman TABLE 1. Meteorological Parameters Used to Calculate Sea-Air divergenceand by cross-pycnoclineheat flux by turbulence Heat Exchange Within 60ø-70øS and possiblyconvective processes, which can be parameter- Month T•, øC Ta,øC Ta,øC W,m/s C ized by a K• (vertical mixing coefficient). The Ekman induced upwelling in the 60ø-70øS band esti- January I 0 -3 3 0.75 mated by Taylor [1978](also see Gordon et al. [1977], Taylor et February 1 - 1 -4 3 0.75 March 0 -2 -5 3 0.75 al. [1978])is 1 x 10-4 cm/s. Sincethe deepwater is approxi- April 0 -5 -8 4 0.75 mately 3øC warmer than the winter surface water, Ekman May -2 -8 -10 4 0.75 pumpingcould provide about 13 W/m 2. Sinceboth the total June -2 -10 -13 4 0.75 annual as well as average mid-November to mid-January July -2 -12 -14 4 0.75 deep water contributionto the surfacelayer is expectedto be August -2 -12 -14 4 0.75 September -2 -11 -13 4 0.75 31 W/m 2, about 18 W/m 2 must be derived from the vertical October -2 -7 - 11 4 0.75 diffusiveand/or convectiveheat flux (pCpK/X T/ AZ). This re- November -2 -4 -7 4 0.75 quiresa pK, valueof 1.5 cm2/s(Using a pycnoclinethermal December 0 -1 -5 3 0.75 gradientof 3 x 10-4 øC/cm). T•, sea surface temperature (summer values from historical data; The annual averagevertical flux of heat by Ekman pump- values during sea ice months taken as nearestwhole degreeto freez- ing and vertical mixing effectswould provide on a daily basis, ing). Ta, air temperature(average at 65øSfrom Taljaard et al. [1969]). about 50% of the required mid-November to mid-January T•, dew point temperature (average at 65øS from Taljaard et al. [1969]). W, wind speed(average at 65øS from Jenneet al. [1971]). C, heat. This amount togetherwith the atmosphereto oceanheat cloud cover (75% cover all year, from Table 2.2 of $asamori et al. flux can satisfythe heat requiredto explain the springsea ice [1972]). melting. GORDON: SEASONALITYOF SOUTHERN OCEAN SEA ICE 4195
TABLE 2. AverageMonthly Heat Flux Into Ocean60ø-70øS for Ice-Free Conditions(Values in W/m 2 andcal/cm 2 d) Month Qs Qb Qn Qe Q,o January 162[335] 35 [72] 7 [15] 39 [81] 81 [167] February 109[225] 36 [74] 15[31] 47 [98] 11[22] March 56 [116] 36 [74] 15[31] 44 [91] -39 [-80] April 21 [43] 36 [75] 50 [104] 72 [149] -138 [-285] May 12[24] 36 [74] 60 [124] 63 [131] -148 [-305] June 4 [8] 36 [74] 80 [165] 66 [137] -178 [-368] July 2 [5] 36 [75] 100[207] 72 [149] -206 [-426] August 38 [78] 36 [75] 100[207] 72 [149] -171 [-353] September 76 [156] 36 [74] 90 [186] 68 [140] -118 [-244] October 106[219] 36 [74] 50 [103] 69 [143] -49 [-101] November 128[265] 35 [73] 20 [41] 60 [124] 13[27] December 142[294] 36 [75] 8 [16] 52 [107] 47 [97]
Average 71 [147] 36 [74] 49 [102] 61 [125] -75 [-154] -56.2 kcal/cm2 yr Q•,incoming solar radiation absorbed by ocean 60ø-70øS from values given by Sasarnorietal. [1972, Tables 2.6 to 2.9] interpolatedto eachmonth with an oceansurface albedo of 10%for the direct radiation and6.6% for thediffuse radiation. Qt,, net back radiation where cloud cover (in tenths)is assumedto be middlelevel clouds (equation from Reed[1976]). Qn, sensible heat exchange (equation from Budyko [1963]).Qe, evaporative heat exchange (equation from Budyko[1963]). Qto, net sea-airtransfer for ice- free ocean: Qto-- Qs- Q•- Qn- Q•.
A pKz value of 1.5, though large for an oceanicmain pycno- high end of the envelope,but in view of the greater shearex- cline, may be appropriatefor the weak stability characteristic pectedwithin the shallowSouthern Ocean pycnocline relative of the Southern Ocean pycnocline south of the polar front. to the abyssalocean, the 1.5K• value is not unreasonable,par- Sarmiento et al. [1976] using radon distribution within the ticularly if convectiveevents occur [Gordon, 1978; Killworth, abyssalocean show that vertical mixing coefficientbears an 1979]. inverse relationship to the buoyancy gradient, suggestinga The above calculationssuggest that sea ice melting in the constant buoyancy flux. The Southern Ocean pycnocline, mid-November to mid-Januaryperiod is facilitatedby Ekman while strongerthan the vertical densitygradient of the abyssal layer generationof leadswhich act as radiative heat collectors ocean, is markedly weaker than the main pycnoclineof the and by deep water to surfacewater heat flux due to Ekman world ocean. The maximum buoyancy gradient below the upwellingand cross-pycnoclinemixing processes.While it is temperatureminimum layer, which representsthe baseof the possible,in view of the uncertainty in the sea-air heat ex- winter mixed layer, around Antarctica at 65øS (determined changecalculations that the lead generationmechanism alone from the data given in the SouthernOcean Atlas by Gordonet is sufficientto accountfor the melting, the possibilityof a sig- al. [1981] falls in the range of 350 to 3400 x 10-8 s-2 with an nificant role of cross-pycnoclineheat flux must be considered averageof 1300x 10-8 s-'•.This rangewith a Kz of 1.5,plotted particularly sincethe requiredK• value is not unreasonable.It on the Sarmiento et al. relation (Figure 3) falls toward the is expectedthat cross-pycnoclineprocesses would be most im- portant at the end of the ice growth period when (1) the cu- mulative effectsof winter sea ice formation induce the yearly TABLE 3. Average Monthly Heat Flux Into Ocean for Full and PartialIce Cover(Values in W/m: andcal/cm: d) maximum in surfacewater salinity (and density),and (2) the insulatingeffect of the sea ice may act to trap the deep water Month R Qti P Qt heat flux in the surfacelayer [Welander, 1977]. January 0.36 29 (60) 0.3 65 (135) Parkinson and Washington [1979] require a substantial February 0.28 3 (6) 0.1 10 (20) ocean to ice heat flux to induce spring melt back. They use a March 0.20 -8 (-16) 0.2 -32 (-67) value of 25 W/m 2, aboutone ordermagnitude higher than the April 0.23 -32 (-66) 0.4 -95 (-197) value used in the Arctic. Use of the Arctic value for the South- May 0.17 -25 (-52) 0.5 -87 (-179) june 0.12 -21 (-44) 0.6 -84 (- 174) ern Ocean did not result in the spring melt back. Use of a July 0.12 -25 (•51) 0.7 -79 (-164) value of 40 W/m: did not allow for winter sea ice formation August 0.09 -16 (-32) 0.7 -62 (-128) (C. Parkinson, personal communication, June 1980). Thus September 0.07 -8 (- 17) 0.8 -30 (-62) modelling attempts have difficulty in producing the intense October 0.15 -7 (-15) 0.7 -20 (-41) November 0.14 2 (4) 0.6 6 (13) seasonalityof Southern Ocean sea ice, without resorting to December 0.27 13 (26) 0.4 33 (68) vertical flux of the deep ocean heat.
Average -8 (-16) -31 (-65) 5.8 kca!/cm:yr -24 kcal/cm:yr TABLE 4. Annual AverageHeat Flux Across60øS to Balance60 ø- R, ratio of heat flux with full ice cover to heat flux for no ice cover. 70øS Ocean Heat Lossto the Atmosphere Determinedfrom data presented• in Figures A-8 andA-10 of Fletcher [1969]. (See sectionIII-B of Hibler [1979] for a discussionof the ther- Condition Watts Cal/s mal processesin sea ice.) Qti, net sea-air heat transfer for full seaice 1. Ice free 130 X 1013 31 X 1013 cover.P, proportionof 60ø-70øScovered with seaice (estimatedfrom 2. Total ice cover 13 X 1013 3 X 1013 the Na•/Y-NOAA charts).Qt, net sea-airheat flux for realisticsea ice- 3. Realistic ice cover 54 X 1013 13 X 1013 open oceanratio or the 60ø-70øS belt. 4196 GORDON:SEASONALITY OF SOUTHERNOCEAN SEA ICE
TEMPERATURE (øC) antarcticbottom water are not included in Stommel's results. -5 0 5 I0 15 20 25 ;50 0 Perhapsuse of the hydrographicdata south of 60øS,particu- 2OO 400 larly in the Weddellregion, might have revealed this regime. 6OO 8OO 4. ROLE OF PRECIPITATION IOO0 I I I .. The weak SouthernOcean pycnocline is a consequenceof SALINITY (*/**) the relativelyß high salinityof tlie surface water due to the 33.0 33.5 34.0 34.5 35.0 35.5 36.0 36.5 :57.0 small fresh water input. Estimatesof precipitation•minus i -- i i i i 20 - - ---' evaporation(P- E) summarizedby Newton[1972, Figure 400•- 9.14]for 60ø-70øS are 0.075 cm/d, or 27cm/yr. Bau•ngartner Oo1: . 8oob : and Reich'el•[1975] arrive at a valueof 32 cm/yr. Coatmental i000 L I I I I I runoff(in theeform of icebergs)contribution to the60 ø to 70øS SIGMA-T oceanarea can be taken as 10 cm/yr (since t•e ocea•narea is 25.0 25.5 26,0 .26.5 27.0 27.5 28.0 28.5 29.0 about15% larger than Antarctica; which has • meanprecipi- '• 200O• I I i I-- , I I tationof 11to 19cm/yr [VanLoon, 1972]); so the totalfresh = 400 •. waterfiui (F) is taken about 40 cm/•r. This value, when used •- 600•- withestimates 9focean heat loss to the atmosphere, 31W/•n 2 "' 800•- i000 L • - I I I I I (Table3), canbe used to determinethe path followed in tem- perature-salinityspace by a parcelof • waterexposed to theat- Fig.2. An exampleof thermohalinestratification in the seasonal mosph•re. The AT/AS slopeis givenby (sois theinitial salin- seaice zohe of theSouthern Ocean. The plot shows the mean tempei'- a{ure(degrees Centigrade), salinity (per mil), andsigma-t (density ity of the water pa:rcel) anomaly)versus depth at 65øS,20øE. The envelopearound the cen- tral curveindicates range in parameters.This plot is takenfrom the AT_Qr/pC v= 18.0oc/%ø microfichesupplement to the SouthernOcean Atlas by Gordonet al. AS SoF [1981]. Sincethe AT/AS slopeof an isopycnalin the -2 to +2øC Stommel[1980], using vertical t•mperature-salinity curves temperaturerange is •5.4øC/%o,the thermohaline alteration betweenAustralia and Antarctica (35ø-60øS) and the esti- of deepwater whichbecomes incorpoi'ated into the surface matesof meridional11gat and freshwater flux [Hastenrath,layer would follow a. T/S pathwhich would increase its den- 1980;Baumgartner and Reichel, 1925], determines the required sity with time. Therefore, thermohaline alteration in the 60ø•- watermass flux across latitudes, as well as the spreading pat- 70øS belt inducesan unstableboundary between deep water tern. Across60øS Stommel finds a required poleward flux of and surface water or, at best, a marginally stable stratification. warm, saltydeep water of 18 x !06 rn3/s,with a circulation It ispossible that fresli water input would be further in deficit patternsimilar to that proposedby Deacon[1937]: deep water if seaice is removedfrom the 60ø-70ø•1 bfind at a morerapid convertedby sea-airenchange into antarcticsurface water re- rate than the surfacewater by Ekman drift. tumsnorthward at depthsshallower than the deepwater pole- In view of the errorsassociated with the heat and fresh•va- ward migration.Using the larger 60øS heat flux value pre- terflux, it cabbe concluded that low stability of theS0ut.h6rn sentedin this paper(about 2.5 largerthan the valueused by Oceanpycnocline is a consequenceof a •'elativelylow fresh Stommel),the requireddeep water flux would increasecorre- waterinput which nearly balances the thermal buoyancy flux. spondinglyand the mode of the exchangewould be more iso- Thisis in str0ng.contrasttothe high fresh water input i•to the pycnal.It is noted that the southernend of the Deaconpat- Arctic, which inducesa strongpycnocline: ' tern, in which the altered deep water spreadsnorthward as Variationsin P- E havea strongimpact on thepycnoclin6 stabilityand hence on the K, valueand vertical heat flux. K;=(cm2/sec) creasedP - E wouldlead tO higher vertical stability with less springice melt, while decreasedP - E would increasevertical I0,000 I0 I00 I000 processesand lead to morerapid ice removalin the spring. Thusthe seasonalityof SouthernOcean sea ice may depend on precipitationin additionto the obviousdependence on the I000I i ] heatbudget. Perhaps the significant variations in thedegree of seasonalityof the SouthernOcean sea ice whichhave oc- / curredin thegeological past [Cooke and Hayes, 1980] are consequenceof pycnoclinestability variations. Another manifestationof the cross-pycnoclineheat flux is ] the Weddell Polynya.The polynyais most likely the con-