The storm tracks and the energy cycle of the Southern Hemisphere: sensitivity to sea-ice boundary conditions

The storm tracks and the energy cycle of the Southern

Hemisphere: sensitivity to sea-ice boundary conditions

C. G. Menendez, V. Serafini, H. Le Treut

To cite this version:

C. G. Menendez, V. Serafini, H. Le Treut. The storm tracks and the energy cycle of the South-ern Hemisphere: sensitivity to sea-ice boundary conditions. Annales Geophysicae, EuropeanGeosciences Union, 1999, 17 (11), pp.1478-1492. <hal-00329145>

HAL Id: hal-00329145

https://hal.archives-ouvertes.fr/hal-00329145

Submitted on 1 Jan 1999

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinee au depot et a la diffusion de documentsscientifiques de niveau recherche, publies ou non,emanant des etablissements d’enseignement et derecherche francais ou etrangers, des laboratoirespublics ou prives.

https://hal.archives-ouvertes.fr

https://hal.archives-ouvertes.fr/hal-00329145

The storm tracks and the energy cycle of the Southern Hemisphere:sensitivity to sea-ice boundary conditions

C. G. MeneÂ ndez1, V. Sera®ni2, H. Le Treut2

1 Centro de Investigaciones del Mar y la AtmoÂ sfera/CONICET-UBA, Ciudad Universitaria, PabelloÂ n 2, Piso 2, (1428) Buenos Aires,Argentina2 Laboratoire de MeÂ teÂ orologie Dynamique/CNRS, UniversiteÂ P. et M. Curie, Tour 15-25, 4 place Jussieu, 75252 Paris Cedex 05, France

Received: 18 August 1998 / Revised: 9 June 1999 /Accepted: 30 June 1999

Abstract. The e�ect of sea-ice on various aspects ofthe Southern Hemisphere (SH) extratropical climate isexamined. Two simulations using the LMD GCM areperformed: a control run with the observed sea-icedistribution and an anomaly run in which all SH sea-iceis replaced by open ocean. When sea-ice is removed, themean sea level pressure displays anomalies predomi-nantly negatives near the Antarctic coast. In general, themeridional temperature gradient is reduced over most ofthe Southern Ocean, the polar jet is weaker and the sealevel pressure rises equatorward of the control ice edge.The high frequency ®ltered standard deviation of boththe sea level pressure and the 300-hPa geopotentialheight decreases over the southern Paci®c and south-western Atlantic oceans, especially to the north of the iceedge (as prescribed in the control). In contrast, over theIndian Ocean the perturbed simulation exhibits lessvariability equatorward of about 50°S and increasedvariability to the south. The zonal averages of the zonaland eddy potential and kinetic energies were evaluated.The e�ect of removing sea-ice is to diminish theavailable potential energy of the mean zonal ¯ow,the available potential energy of the perturbations,the kinetic energy of the growing disturbances and thekinetic energy of the mean zonal ¯ow over most ofthe Southern Ocean. The zonally averaged intensityof the subpolar trough and the rate of the baroclinicenergy conversions are also weaker.

Key words. Air-sea interactions á Meteorology andatmospheric dynamics (climatology; ocean±atmosphereinteractions)

1 Introduction

There has been a statistically signi®cant increase intemperature since the mid-1940s in some regions of theAntarctic periphery (Harangozo et al., 1994; Turneret al., 1997). The reason for this observed warming isnot fully understood at present, but it may be related tochanges in the sea-ice extent and in the regionalatmospheric circulation. A comprehensive understand-ing of the interaction between the sea-ice and theatmospheric circulation is therefore necessary. Currentclimate models diverge in their estimation of the climatechange scenarios at the high latitudes of the SouthernHemisphere (e.g. Carril et al., 1997). This may be in partdue to the complexity of sea-ice physics and dynamicsthat is parameterized in a di�erent manner in eachmodel. It seems likely that many physical processesoccurring in those regions are not e�ciently representedin the coupled models.

In order to understand the model sensitivity and toassess the possible impact of a climatic change it isuseful to study how the melting of sea-ice, and thecorresponding warming of the surface temperature,may a�ect the extratropical circulation of the South-ern Hemisphere (hereafter SH). These circulationchanges are related to basic characteristics such asthe atmospheric energy cycle and the storm tracks. Wetry to estimate the role of sea-ice on the SHatmospheric circulation, as a limited approach forunderstanding and simulating high-latitude climatechange. In particular, although the interaction betweensea-ice characteristics and the cyclonic activity israther complex, the aspect we wish to address is theimportance of changes in the sea-ice cover for theclimatological high-frequency behaviour in the south-ern extratropics.

Most of the Antarctic sea-ice melts in summer andforms again in winter, covering up to 18 ´ 106 km2.Compared with the Arctic, the percentage of open wateris greater and the ice thickness is smaller. Other sea-ice

Correspondence to: C. G. MeneÂ ndezE-mail: [email protected]

Ann. Geophysicae 17, 1478±1492 (1999) Ó EGS ± Springer-Verlag 1999

morphology di�erences between the SH and the North-ern Hemisphere are related to the latitude domain andthe location relative to the continents. The presence ofsea-ice ®rst a�ects the vertical transfer of energy betweenthe atmosphere and the surface. Turbulent heat ¯uxesare substantially reduced over sea-ice. Surface roughnessvalues in the standard parameterizations are larger oversea-ice than over the ocean a�ecting the surfaceexchange of momentum. In addition, the high re¯ectiv-ity of sea-ice reduces the absorption of solar radiationby the surface.

However the Antarctic sea-ice and the neighbouringocean is also a region of high cyclone frequency andcyclogenesis at the synoptic and mesoscale (Bromwich,1991; Carleton, 1992). Thus, one might expect arelationship between surface conditions and cyclonebehaviour. This connection between sea-ice extent,cyclonic activity and features of the circulation is notcompletely clari®ed. Carleton (1989) and Streten andPike (1980), for example, found little observationalevidence to support this relationship. However, insu�-cient or poor data in southern high latitudes a�ect theobservational studies. Simmonds and Wu (1993) (here-after SW93) discuss the ambiguous nature of thepresently published studies concerning the response ofthe atmosphere to sea-ice perturbations. Ambiguity canarise from the, usually, short periods of data being used,the particular epoch being considered and the highlyvariable character of the synoptic circulation in thesouthern extratropics. In a recent observational analysis,Godfred-Spenning and Simmonds (1996) have foundthat a sea-ice±cyclone link is not apparent over theentire sea-ice zone on a seasonal time scale, supportingthe conclusion of Carleton (1989) and Streten and Pike(1980).

Sensitivity studies using general circulation models(GCMs) therefore constitute an important approach tounderstand the climate of these high-latitude regions.Previous numerical studies have analyzed the atmo-spheric response to reductions in the Antarctic sea-iceextent or concentration (Simmonds, 1981; Mitchell andHills, 1986; Simmonds and Dix, 1986; Mitchell andSenior, 1989; Simmonds and Budd, 1991; SW93). Inmost cases, these GCM experiments (winter simula-tions) only report some evidence that the sea-ice forcesthe atmospheric circulation in the circumpolar troughregion. A local decrease in surface pressure was gener-ally found, especially in the region from which sea-icewas removed. SW93 have studied the sensitivity ofcyclone behaviour to prescribed changes in winterAntarctic sea-ice concentration. They show an increasein the number of cyclones near the Antarctic coast(particularly over most of the Weddell and Ross seas)and a decrease in the cyclone frequency to the north, inresponse to the reduction of sea-ice concentration.However the intensity of these extratropical cyclonesbecomes weaker. SW93 suggest that the climatologicaldecrease in pressure results essentially from an increasein cyclone numbers near the coast as opposed to adeepening of existing cyclones. Recently, MeneÂ ndezet al. (1999) (hereafter MSLT99) performed a sensitivity

experiment in which all southern sea-ice was removedwith a relatively high-resolution atmospheric globalmodel in order to study the e�ect of sea-ice on thepoleward transport of heat and momentum by eddiesand the mean eddy statistics. MSLT99 found thatthe removal of ice in the Southern Ocean a�ects thebaroclinic structure of the atmosphere. In general,the meridional wind variance, the poleward transienttemperature ¯ux and the eddy ¯ux convergence ofwesterly momentum were weaker over the SouthernOcean when sea-ice was removed. Le Treut and Kalnay(1990) also suggest that the simulated cyclone life cycleand the cyclone tracks may, in part, be dependent on theprescriptions of the boundary conditions (sea-surfacetemperature and sea-ice extent), especially in the highsouthern latitudes.

In this study two simulations of an atmosphericclimate model are compared: a control run with theobserved sea-ice distribution and an anomaly run inwhich all SH sea-ice was replaced by open ocean.Section 2 presents the model, the experiment design anda brief evaluation of the basic climatology. Section 3examines the e�ect of sea-ice on various aspects of thesouthern extratropical climate. The impact on somebasic characteristics is described, together with the cir-cumpolar trough and its semi-annual oscillation, thehigh frequency variability and the storm tracks, and theenergy cycle. A general discussion is given in sect. 4.

2 Model description, experiment designand model evaluation

2.1 The model

The model being used is the general circulation model ofthe Laboratoire de MeÂ teÂ orologie Dynamique du CNRS(LMD GCM). This model is formulated in ®nitedi�erences, in its dynamical part, and contains compre-hensive physics. Sadourny and Laval (1984) ®rst de-scribed the original version of the model. The presentversion, called Cycle 5, contains a number of newfeatures. The cloud scheme is a prognostic scheme basedon a budget equation for cloud water (Li and Le Treut,1992). Convection is parameterized by a combination ofa Kuo (1965) scheme and a moist adiabatic adjustment.The radiation scheme is the same as used operationallyat ECMWF, with a representation of the transfer ofsolar radiation following the scheme of Fouquart andBonnel (1980). The infrared radiation is parameterizedfollowing Morcrette (1990). Some re®nement of thesurface parametrization and a description of the mainclimatology of the model are available in Harzallah andSadourny (1995). Note however that our simulations aredi�erent from those of Harzallah and Sadourny (1995)because they include a di�erent sea surface temperature(SST) climatology. An open seawater fraction in allpolar sea-ice is considered in the model. Sea-ice isallowed to cover a fraction of the grid only, but thecomputation of all ¯uxes is carried out using mixedcoe�cients, which correspond to an average over the

C. G. MeneÂ ndez et al.: The storm tracks and the energy cycle of the Southern Hemisphere 1479

whole grid area. The surface albedo (in particular thesea-ice albedo) parametrization is described in Chalitaand Le Treut (1994). The model resolution correspondsto a grid of 64 points in longitude, 50 points in latitudeand 11 vertical layers. Note that the version of the LMDGCM used in this work is similar to the model used inMSLT99, but with lower resolution (96 ´ 72 ´ 19 inMSLT99).

2.2 General design of the experiment

In the present work we consider as control simulation astandard simulation using the Cycle 5 LMD GCM andclimatological sea surface temperatures (SST) for theperiod 1970±1988, with no interannual ¯uctuations. Asecond simulation (hereafter referred as ``NOICE'') wasmade, in which all Antarctic sea-ice was removed. Byremoval we mean setting again the albedo and surfacedrag to the values of seawater, and imposing aprescribed SST of )2 °C (freezing point of seawater),as it has been used in previous ice-free experiments (e.g.MSLT99). This type of imposed surface forcing signi®esthat we assume that the ocean has enough heat storageto be able to provide the upward heat ¯ux needed toprevent the sea-ice from reappearing. Certainly, this isnot a very realistic supposition since it implies in®niteoceanic heat capacity. However, the present experimentis only a preliminary sensitivity experiment with ratherunreal surface forcing.

The climates of the control and anomaly runs wereestimated from the analysis of two 5-y simulations withfull seasonal cycle. The duration of our experiment, asmany previous, similar sensitivity experiments for theremoval of Antarctic and Arctic sea-ice found in theliterature (e.g. Mitchell and Senior, 1989 averaged theirresults over nine months, Royer et al., 1990, analyzedthe second winter after an annual cycle model integra-tion) is relatively short, even if we take into account theabsence of interannual variability of the SST. As thedynamical ®elds in the southern extratropics are highlyvariable on an interannual time scale, a longer than 5 ysimulation would be desirable to describe fully thevariability of the circulation. This is why, in thisexploratory work which is also using a low resolutionof the model, we restrict the analysis to basic elements ofthe SH climate, and to the mean zonal energetics of theatmosphere.

A large amount of forcing (with no interannualvariations) was chosen for the sake of enhancing thesignal-to-noise ratio and reducing the need to test thestatistical signi®cance of the results. From the physicalconsistency of the results, we will be able to infer thatthe sea-ice-atmosphere associations are genuine and didnot merely arise through a coincidence of anomalies inthe analyzed simulations. In this sense, the extremesurface anomaly in NOICE should help to make aclearer representation of the interactions between thedi�erent physical and dynamic processes. However, thislarge change in sea-ice concentration (from observed to0% ice coverage and without interannual variations) is

not realistic, so caution must be exercized beforegeneralizing our results to the more moderate andhighly variable perturbations in ice condition that areactually observed.

2.3 Simulated SH climate

The control climate of this version of the LMD GCM issimilar to those reported for the Atmospheric ModelIntercomparison Project (AMIP) (e.g. Le Treut et al.,1995). The tropospheric circulation in the SH presentssome remarkable characteristics such as a deep circum-polar trough all the year round, a double westerly jet inwinter and a semiannual oscillation of sea level pressureand other quantities. The ability of the model toreproduce these features represents a good test of themodel's ®delity. The observational-based climatology isestimated using the ECMWF operational analyses. Theobserved mean sea level pressure for summer (Fig. 1a)presents a circumpolar trough near 65°S, a steepmidlatitude meridional gradient and a subtropical ridgewith centres in the eastern part of the oceans. In winter(Fig. 1b) the observed subAntarctic trough is somewhatdeeper than in summer and the subtropical ridgeintensi®es and shifts northward. In the simulation (lowerpanels), in common with other low resolution GCMs(Xu et al., 1990), the subpolar trough is too weak andlies too far north throughout the year, although it tendsto be better simulated during winter. The subtropicalhighs are too zonal and somewhat underestimated, butthe equatorward shift of the ridge from summer towinter is captured by the model. The meridionalpressure gradient is too weak especially in summer.

The tropospheric circulation over the SouthernOcean displays a marked half-yearly cycle. This semi-annual oscillation (SAO) produces a twice-yearly pole-ward migration and intensi®cation of the circumpolartrough with equinoctial maxima in baroclinity and in thestrength of westerly winds in the troposphere south ofabout 50°S. Several authors (e.g. Schwerdtfeger andProhaska, 1956; Schwerdtfeger, 1960; van Loon andRogers, 1984; Meehl, 1991; Simmonds and Jones, 1998)have documented this phenomenon and its variability.To evaluate the simulation of this half-yearly wave,Fig. 2 shows the mean annual cycle of the secondharmonic of the zonally averaged intensity of thecircumpolar trough determined through Fourier analy-sis. Xu et al. (1990) emphasize that most of thelow-resolution GCMs fail to simulate this large-scalecharacteristic of the annual cycle in the southern mid-and high latitudes. Other studies (e.g. Tzeng et al., 1993)manifest improvements in simulating some features ofthe SAO. The amplitude of the second harmonicsimulated by this version of the LMD GCM agreeswell with the oscillation derived from the analyses.However, the phase is shifted by about one month.

The observed and simulated mean zonal wind at500 hPa for summer and winter is shown in Fig. 3. Inwinter (Fig. 3b) the subtropical jet reaches its maximumintensity to the east of Australia. Near South America,

1480 C. G. MeneÂ ndez et al.: The storm tracks and the energy cycle of the Southern Hemisphere

the axis of the jet begins to spiral in toward the pole,producing a double jet over part of the hemisphere. Themain features of the simulated ®elds (lower panels) aresimilar to the observational climatology but the modeltends to produce a jet that is too zonal, especially insummer. The polar branch of the westerlies and thezonal wind minimum over the southern Paci®c arepoorly captured by many GCMs (e.g. Boville, 1991;Tzeng et al., 1993; Katzfey and McInnes, 1996). How-ever, the wintertime split jet structure is apparent in theLMD GCM control simulation. The zonal wind mini-mum is quite well captured near New Zealand. Ingeneral, the subtropical jet tends to be stronger and thepolar jet weaker than in the analysis.

3 Results

3.1 General features

As mentioned the most immediate e�ect of removingsea-ice is to raise the surface temperature. In ourexperiments this e�ect is prescribed. Warming of theoverlying atmosphere is mainly con®ned near thesurface. The atmospheric disturbances distribute thiswarming over a range of latitudes. Figure 4 shows theannual cycle of temperature averaged between 55°S and70°S in the model's lowest sigma level (r = 0.991). Thenear-surface temperature augmentation is important inwinter, but is negligible between November and April.

Fig. 1a±d. Mean sea level pressure from observations (upper panels) and the model (lower panels), for summer (a, c) and winter (b, d). Contourinterval is 5 hPa


The largest temperature increase in winter is locatedover the Weddell Sea at about 20°W and between 160°Eand 120°W in the Paci®c sector (Fig. 5a). Noted that wecomputed our diagnostics over a three-month period(austral winter ± June, July, August ± or summer ±December, January, February) instead of a one-monthperiod (as in most previous similar sensitivity studies).The use of a longer period has the advantage ofreducing the noise level of the results and consequentlyincreases their signi®cance and reliability. In winterlarge surface heat ¯uxes into the atmosphere occur tothe north of the ice edge. The higher ¯uxes areproduced by the outbreaks of cold dry air from theAntarctic continent. The location of the maximum forthe sensible heat ¯ux into the atmosphere shiftspoleward in the NOICE simulation. There is a decreasein the ¯ux near the location of the ice edge, asprescribed in the control experiment, and a largeincrease in the regions corresponding to the largesttemperature increase (Fig. 5b). Surface heating intensi-®es the mean upward motion over the sea-ice (notshown). This dynamical response is compensated byrelative descent northward of the ice edge and overAntarctica. Figure 2c shows the changes in mean sealevel pressure. It displays both negative and positiveanomalies near the Antarctic coast. The largest pressuredecrease near Antarctica is found at the regions oflargest increase in surface temperature (i.e. over theWeddell and Ross seas). Equatorward of the control iceedge there are pressure rises over the southern oceans.

3.2 Zonal wind

Sea surface temperatures and temperature gradients inthe lower troposphere are strongly correlated in the SH(van Loon, 1972). The meridional temperature gradientis closely linked to the zonal winds through the thermal

wind relationship. The presence of the Antarcticcontinent and its associated pack ice provides a heatsink throughout the year, a�ecting the meridionaltemperature gradients. The polar jet is in¯uenced bythe strong temperature gradient that appears in thelower troposphere adjacent to Antarctica. Removingsea-ice may a�ect the meridional temperature gradientand thus the SH jet structure. Figure 6 displays thechanges in the mean zonal wind at 500 hPa. Themeridional temperature gradient (not shown) is reducedover most of the southern Atlantic and Indian oceans,probably as a consequence of the changes in thedistribution of the surface heat ¯uxes. Consequentlythe polar jet is weaker in NOICE and the double jetstructure is less evident. In contrast, the subtropical jetis more zonally elongated in NOICE. The zonallyaveraged mean zonal winds (not shown) are decreasedthroughout the troposphere southward of 30°S, exceptin a shallow boundary layer over the sea-ice anomaly.Northward of this, there is a latitude band betweenabout 10°S and 30°S where winds are increased in theNOICE case.

3.3 SAO of the subpolar trough

As noted previously, an important feature of the SHclimatology is the SAO of the circumpolar trough. Thetrough is more intense and is located farthest south inMarch-April and September-October in the observa-tional climatology (see Fig. 2). The circumpolar troughis in¯uenced by the collective tracks of individualbaroclinic disturbances: depressions that develop withinthe circumpolar trough and move eastwards or thosethat move into the Antarctic coastal area from the northor northwest (Jones and Simmonds, 1993). As theannual cycle of this transient eddy activity is alsomodulated by the SAO (Meehl, 1991), it is interesting toexamine the model sensitivity from this point of view. Itsmajor cause is the di�erence in the seasonal heating andcooling between middle and high latitudes. An intensi-®cation of the temperature gradient between the oceaniclatitudes and the Antarctic latitudes is produced twice ayear in the midtroposphere. This behaviour is dependenton the heat budget of the oceanic upper layers, which isa�ected by the annual cycle of sea-ice. To illustrate thee�ect of sea-ice on the SAO, the mean annual cycle ofthe second harmonic of the zonally averaged intensity ofthe circumpolar trough for the control and NOICEsimulations are compared in Fig. 7. The phase ismaintained in the anomaly run. But the amplitude issubstantially reduced in NOICE. Besides the e�ect onthe semi annual component, the trough tends to beweaker between May and July when sea-ice is removed(not shown here).

3.4 Filtered standard deviation of the sea level pressure

We study here the impact of the sea-ice on the simulatedhigh frequency transient features of the circulation. The

Fig. 2. Mean annual cycle of the second harmonic of the zonallyaveraged intensity of the subpolar trough from observations and themodel (in hPa)


variability in the synoptic time scale was investigatedutilizing a ®lter based on the procedure formulated byMurakami (1979). This technique has the advantagethat we can specify at our convenience the maximum inthe response curve, and its bandwidth (as shown inMurakami, 1979). In this sensitivity experiment, thelargest e�ect can be expected where sea-ice extent is thelargest (i.e. in western Antarctica). Synoptic systems inthis region tend to displace and develop more slowlythan elsewhere in the Southern Ocean (Berbery andVera, 1996). The phase speeds of the synoptic waves aresmaller in the South Paci®c Ocean than in the mainstorm track region (i.e. the South Indian Ocean).

According to Berbery and Vera (1996), their dominantperiod in the South Paci®c sector is about 8 days (incomparison with a period of about 4 days in the IndianOcean). For this reason, the bandpass ®lter applied herecovers a period range from about 3±9 days with thecentre at day 6. The distribution of the ®ltered standarddeviation of the sea level pressure for both simulations isrepresented in Fig. 8. As shown by Simmons andHoskins (1978) waves with periods up to 10 days havea variance related to baroclinic instability.

In winter, the control experiment shows a high levelof synoptic scale variability over the high-latitudeoceans, particularly over the Atlantic Ocean between

Fig. 3a±d. Mean zonal wind at 500 hPa from observations (upper panels) and the model (lower panels), for summer (a, c) and winter (b, d).Contour interval is 5 ms)1


40°S and 60°S and over the Paci®c Ocean in theBellingshausen/Amundsen Sea. It can be seen that thearea of the Bellingshausen/Amundsen Sea has a largerstandard deviation than any other area of the circum-polar trough. When comparable ®elds are producedusing UK Meteorological O�ce, US NCEP and theAustralian Bureau of Meteorology analyses, the largeststandard deviation is also present over this area (Turneret al., 1997). This region is characterized by a largestandard deviation in the number of individual cyclonesand also in the cyclone density, which takes account ofthe number and speed of the individual lows. Turneret al. (1997) reported that an atmosphere-only versionof the Hadley Centre climate model also succeed inreproducing this pattern. According to these authorsthe fact that the circulation variations in this region arewell reproduced by atmosphere-only GCMs with pre-scribed surface conditions indicates that is not associ-ated with inter-annual variability in the oceancondition. In summer, there is a belt of high variabilityaround the hemisphere at about 40°S±60°S with max-ima over the three oceans. Compared to winter, thesimulated summer synoptic variability is weaker andwith more zonal symmetry.

Not surprisingly, in the NOICE simulation themajor anomalies are observed in winter. The wintertimesea level pressure variability over the southern SouthPaci®c, particularly over the Bellingshausen/Amundsensector, is decreased and the maximum over the Atlanticis more elongated eastward. These di�erences can bebetter observed in Fig. 9 which shows the anomalous(i.e. change from the control) geographical distributionof the ®ltered standard deviation of sea level pressurefor winter. The general tendency apparent in this ®gureis a decrease of the short-term variability in theneighbourhood of Antarctica as the sea-ice is eliminat-ed, mainly in the Paci®c sector and over the WeddellSea. The decrease is especially evident over andnorthward of the ice edge (as prescribed in the control).That is, in the NOICE case the amplitude or the density

of the high frequency perturbations over this region ofthe sub-Antarctic trough is diminished. Another regionof decreased variability is observed over the South

Fig. 4. Mean seasonal variation of the temperature in the model'slowest sigma level, averaged between 55°S and 70°S, in °C

Fig. 5a±c. Meanwinter anomalies (NOICEminus control): a changesin temperature at the model's lowest sigma level, every 5 °C. bChanges in sensible heat ¯ux, every 30 Wm)2. c Changes in mean sealevel pressure, every 2 hPa. The zero contours are not shown. Thedashed line around Antarctica indicates the mean sea-ice extent forwinter in the control simulation. The northern latitude is 40°S


Atlantic at about 45°S. Nevertheless, there are somesectors which show up as regions of increased synopticvariability when sea-ice is removed. This is the case overthe Antarctic coasts, especially over the eastern conti-nental sector. Another region with increased highfrequency variability is found over South America near40°S. The longitudinal di�erences in the anomaly maybe related to the sea-ice distribution in the control run: ithas removed more ice in the western than in the easternsector. Nevertheless, these longitudinal di�erences mayalso suggest that perturbations are in¯uenced by otherfactors besides the sea-ice (e.g. baroclinic and barotropicprocesses, the interaction of the high topography ofAntarctica with the large-scale ¯ow, etc.).

3.5 The storm track

The eddy activity associated with the SH storm tracks ismostly related to baroclinic systems (Trenberth, 1991).In the SH winter large weather systems propagate intothe upper level subtropical jet while at low levelsbaroclinic waves are maintained in the subpolar jet(Hoskins et al., 1983). The SH storm tracks are alsoin¯uenced by a strong barotropic (height independent)component in the mean westerlies (James and Gray,1986). This barotropic component (not analyzed here) isquite well captured by the model. However, the eddyenergy generation is mainly due to baroclinic processeswhile barotropic processes are weaker. Largest values ofbaroclinicity occur along the regions of maximum winds(as shown in Berbery and Vera, 1996 their Fig.1, thegeographical distribution of the jet streams is similar tothe Eady growth rate).

We describe here the impact of sea-ice on the ®lteredvariance of geopotential height in the upper troposphere(300 hPa). We use the same ®lter as described previous-ly. This variance arises from the higher frequencytransient activity (i.e. disturbances with periods lessthan about 10 days). The major axis of the storm tracksapproximately corresponds to the paths commonlyfollowed by active depression centres. As we haveshown before, the sea-ice removal a�ects the surfaceheat ¯ux distribution, the temperature gradient and thusthe baroclinicity and the polar jet. Furthermore, theweakness of the polar jet in the NOICE case isimportant because any perturbation is transported moreslowly, thereby limiting the local standard deviation ofgeopotential height. In the SH there is a major stormtrack which begins in the eastern South Atlantic andgenerally extends towards the seas south of Australia,although its eastward extent is rather variable. In theobservational analyses (Fig. 10a) the strongest section

Fig. 6a, b. Changes (NOICEminus control) in mean zonal wind at 500 hPa for a summer and b winter. Contours every 1 ms)1 (a) and 2 ms)1 (b)

Fig. 7. Mean annual cycle of the second harmonic of the zonallyaveraged intensity of the subpolar trough for the control and theNOICE simulations (in hPa)


of the storm track is located in the southern IndianOcean, between 40°E and 90°E. The maximum intensityis similar in both seasons, but in summer the storm trackis more zonally elongated and in winter the activityextend over broader latitudes. In order to determine therole of sea-ice on the storm track, Fig. 10 shows also the3±9 days bandpass ®ltered standard deviations of300 hPa geopotential height for the control (Fig. 10b)and NOICE (Fig. 10c) simulations and the anomaliesNOICE minus control (Fig. 10d). The observed geo-graphical distribution in winter of the high-frequency

variability seems to be reasonably reproduced in themodel, but the standard deviation is underestimated. Inthe NOICE case the maximum intensity in the IndianOcean is similar to the control, although the main stormtrack shifts slightly poleward and is less elongatedeastward when sea-ice is eliminated. In general over theIndian Ocean the NOICE simulation exhibits lessvariability equatorward of about 50°S and increasedvariability to the south (see Fig. 8d). Over the southernAtlantic and Paci®c Oceans the anomalies are predom-inantly negatives (i.e. the variability is diminished in

Fig. 8a, b. Filtered standard deviation 3±9 days of mean sea level pressure for summer and winter: a control and bNOICE. The contour intervalis 1 hPa


NOICE), especially near 50°W and 170°W. In summer(not shown), the changes in the storm track are ratherdi�cult to see. It is worth noting that the structure of thechanges in variance at the surface and 300 hPa are verysimilar. A similar ``barotropic'' response was found byMeneÂ ndez (1994) in a sensitivity study, performed with aregional model, of the synoptic evolution of subAntarc-tic storms to di�erent sea-ice conditions.

3.6 Energy cycle

The atmospheric circulation is maintained against thedissipative e�ects of friction by converting potentialenergy into kinetic energy of atmospheric motionthrough baroclinic instability. In the midlatitudes, thebaroclinic instability is manifested through the genera-tion of active cyclones and anticyclones. These eddiescarry heat poleward and upward, with the main centresof heat ¯uxes concentrated into the storm track.According to SW93 and our results, sea-ice may a�ectthe cyclone behaviour in the Antarctic periphery. Thecollective e�ect of individual cyclones in the circumpolartrough can be linked to circulation changes overextensive areas of the SH (Meehl, 1991).

We now examine some aspects of the simulatedenergy cycle and its changes from the NOICE to thecontrol case. Figure 11 shows the zonal averages of thezonal and eddy potential and kinetic energies for winter,for both the control and NOICE simulations. The meanzonal available potential energy (Fig. 11a) depends onthe north-south variance of the temperature at eachlevel, and is smaller in the NOICE case, south of about

60°S. The eddy available potential energy (Fig. 11b) isproportional to the east-west variance of temperature,and it is also weaker in NOICE. The main di�erencesfrom the control to the NOICE case are observedbetween 50°S and 70°S. Some of the eddy availablepotential energy is transformed into kinetic energy of theperturbations (Fig. 11c), through baroclinic instability.The e�ect of removing sea-ice is to diminish the kineticenergy of the growing disturbances, mainly poleward ofabout 50°S and also in the region of the subtropical jet(near 30°S). The zonal mean kinetic energy (Fig. 11d)exhibits a rise in the zone of the subtropical branch ofthe westerlies and a decrease poleward of 30°S, espe-cially evident in the latitudes of the polar jet and thestorm track.

Figure 12 shows the annual march of the averagebetween 50°S and 65°S of the rate of conversion fromeddy potential energy to eddy kinetic energy. Thisconversion is associated with the sinking of colder airand rising of warmer air at di�erent longitudes, and canbe considered as a measure of the mean intensity ofbaroclinic processes between the region of the stormtrack and the subpolar trough. The control curve hastwo maxima in April-May and October, as a manifes-tation of the SAO in the model. In the NOICEsimulation, the rate of the baroclinic energy conversionsis weaker from March to October and tends to bestronger in summer. These variations are consistent withthe SW93 and MSLT99 results concerning the intensityof the perturbations. They found that cyclones becomeweaker in winter when the sea-ice concentration isreduced. We recognize that the limited length of thesimulations may also a�ect these results, but theirphysical consistency has to be noted and tends tobalance this statistical weakness.

4 Summary and conclusions

The high latitudes of the SH are a sensitive region of theglobe where perturbation experiments to large-scalemodi®cations of the climatic boundary conditions showa major response. Sea-ice is often thought to play adominant role in this high-latitude response. The role ofsea-ice is primarily thermodynamical, through a mod-i®cation of the surface energy balance, but the associ-ated changes in the atmospheric dynamics may also beimportant. This importance of the sea-ice on theatmospheric dynamics was studied here, using a versionof the LMD GCM, through a sensitivity experiment inwhich all Antarctic sea-ice was replaced by open ocean.The use of this rather arti®cial and large surfaceanomaly has the advantage of reducing the noise levelof the results and thus increases their reliability andsigni®cance. However, it must be emphasized that inorder to make a realistic simulation of the climateconditions over an ice-free Southern Ocean it will benecessary to couple the atmospheric model with oceanicmodels, including sea-ice.

The potential for sea-ice to modify the transientcirculation is a question worthy of consideration. It is

Fig. 9. Changes (NOICE minus control) in the 3±9 days ®lteredstandard deviation of mean sea level pressure for winter. Contoursevery 0.4 hPa; negative contours are dashed. The zero contours areomitted


now well established that the sea-ice characteristicsutilized in climate models has considerable impact uponthe simulated climate (e.g. Mitchell and Senior, 1989;Simmonds and Budd, 1991) as well as the climatology ofthe model's cyclonic systems (SW93). Apart from amodulating e�ect on latent and sensible heat ¯uxes,anomalous sea-ice distribution can also alter the posi-tion and strength of baroclinic zones in the highsouthern latitudes (Simmonds and Budd, 1991). Sea-ice also acts as a mechanical forcing. The replacement ofsea-ice by open ocean is accompanied also by a

substantial reduction in surface roughness. Mitchelland Senior (1989) found that the changes in surface dragcontribute substantially to the response to reduced sea-ice extent. Any of these factors have the potential toalter cyclone characteristics. Alteration of extratropicaltransient behaviour may well have broader e�ects, giventhe primary role of the synoptic systems in the polewardheat and momentum transport in the SH (MSLT99).

The simulation of the atmospheric dynamics andthermodynamics may be sensitive to resolution. Ac-cording to previous studies, even the coarse-resolution

Fig. 10a±d. Filtered standard deviation (3±9 days) of 300 hPa geopotential height for winter: a ECMWF analysis, b control simulation, cNOICEsimulation and d anomaly. The contour interval is 10 m in a, b and c and 4 m in d


global models can represent many aspects of synopticweather systems with useful accuracy. For example,Tzeng et al. (1993) found that a low-resolution GCM (intheir case the CCM1 with R15 resolution, equivalent toa grid mesh of 4.5° latitude by 7.5° longitude) can wellsimulate, to some extent, the dynamics of Antarcticclimate not only for the synoptic scale, but also for somemesoscale features (mesoscale cyclogenesis). Obviouslyincreased resolution would improved the simulation ofthe transient high-frequency eddies, but the suggestion isthat low-resolution GCMs are capable of providingguidance in the study of the synoptic circulation in thesouthern extratropics. Moreover, according to SW93 theuse of low-resolution models in sensitivity experimentssuch as this (rather than a more demanding forecasting)

is not a serious handicap to the interpretation of theresults. The baroclinicity is tied up with cyclogenesis,and cyclones in turn with energy and momentumtransports. Given that these transports are higher ordermoments, their simulation would be expected to berather sensitive to resolution. However, the high-reso-lution sensitivity experiment described in MSLT99exhibits changes in eddy activity qualitatively consistentwith our results. As a consequence, the results of thisstudy are unlikely to be an artefact of an inadequatemodel resolution.

The sea-ice has a strong e�ect on temperatures alongthe Antarctic coasts because the layer of ice insulates thesurface of the ocean. The di�erences between the controland perturbed simulations display some longitudinal

Fig. 11a±d. Zonally and vertically integrated basic forms of energyfor winter, in 106 Jm)2, for the control (solid line) and the NOICE(dashed line) simulations: a available potential energy of the mean

zonal ¯ow, b available potential energy of the perturbations, c kineticenergy of the perturbations, and d kinetic energy of the mean zonal¯ow


asymmetry, partly due to the geography of Antarcticaand to the sea-ice distribution around the continent. Inthe NOICE case, the zones with larger surface heat¯uxes displace poleward and the upward ¯uxes becomestrong at the coast of Antarctica. As a consequence,there is an atmospheric warming over the ice-free oceansurface, mainly con®ned to the high latitudes polewardof 60°S and to the lower levels of the troposphere. Whilesea level pressure shows a large decrease in some sectors,other neighbouring areas experience compensating in-creases. Thus, the zonal mean in the latitude of thecircumpolar trough is similar in both simulations. Ingeneral, the surface pressure response agrees with someprevious studies (e.g. Simmonds and Dix, 1986, removedall SH sea-ice, ®nding both rises and falls in surfacepressure in the region from which sea-ice was removed).

Many of the anomalies presented here show aconsiderable contrast in the north-south direction. Forexample, an opposition in the zonal wind responsebetween mid- and subpolar latitudes in the Paci®c sectorhas been found in winter. In the NOICE simulation thepolar jet is weaker and the subtropical jet is stronger. Notsurprisingly, the changes are much weaker in summer.This might suggest that the changes in temperaturetaking place along the Antarctic coast in winter do notjust a�ect the high-latitude processes, but also thecirculation in the mid-latitudes areas of the SH. It isworth noting that recent observational research suggeststhat anomalies in sea-ice could be associated with large-scale adjustments in the SH circulation. For example,Turner et al. (1997) found a signi®cant correlationbetween observed temperature anomalies on the Ant-arctic Peninsula and 500 hPa height anomalies across theSouth Paci®c between 30°±50°S during winter.

According to our results, it seems likely that anom-alous conditions in the cyclonic activity are related tochanges in the sea-ice extent. Extratropical storms areobserved to evolve in the Antarctic periphery within thecircumpolar trough (e.g. Jones and Simmonds, 1993).This zone is characterized by a strong meridionalthermal contrast linked with the large heat ¯uxes fromthe ocean. When sea-ice is eliminated, the wintertimeshort scale perturbations exhibit less surface pressurevariability near and to the north of the control ice edge.The most a�ected areas are found in the western sectorof the circumpolar trough, over the Bellingshausen/Amundsen and Weddell seas. Over the Antarctic coaststhe changes are lesser. However, a longitudinal sectorover eastern Antarctica shows up as a region ofincreased variability. Another region with enhancedvariability appears over South America at about 40°S.In winter the main storm track over the Indian Oceanshifts poleward and is less elongated eastward and thePaci®c and Atlantic Oceans exhibit less synoptic vari-ability in the NOICE simulation. The anomalies insummer are much weaker.

The simulated energy cycle was evaluated. In generalthe e�ect of removing sea-ice is to diminish the zonaland eddy potential and kinetic energies, in the middle tohigh latitudes. The zonal mean of all the potentialenergy of the mean zonal ¯ow, the available potentialenergy of the perturbations, the kinetic energy of theperturbations and the kinetic energy of the mean zonal¯ow tend to be weaker in NOICE. The zonal meanpatterns of the simulated energy cycle reveal also thatthe baroclinic processes in the NOICE simulation are ingeneral weaker between the latitudes of the storm trackand the circumpolar trough. The weaker eddy activity inNOICE is also manifested through a weakness of thehalf-yearly cycle of the circumpolar trough and of thebaroclinic processes.

In general, the changes in zonal wind, eddy activityand energy cycle are consistent with the study ofMSLT99 who used a higher resolution version of thismodel in a 10-y experiment with full seasonal cycle. Wecould infer that our ®ndings would not be associatedneither with de®ciencies in the resolution nor with theinherent variability of the simulated climate in highlatitudes. The results also con®rm some features ob-tained previously by SW93. These authors have shownthat, over most of the sea-ice area, the cyclonic systemsbecame weaker as the sea-ice concentration was re-duced. They also found a tendency for more cyclonesover the Antarctic coastal zone and fewer cyclones tothe north. Simulations of subAntarctic storms withregional models can be useful to indicate what kind ofsystems should be expected over an ice-free ocean.MeneÂ ndez (1994) reported on a simulation of two latewinter storms with a high-resolution regional model (theCIMA version of the LAHM/GFDL model). The roleof sea-ice was evaluated by performing a series ofsensitivity tests in which surface heat and momentum¯uxes were a�ected. The regional model also producedweaker storms with a kinetic energy reduction through-out the troposphere when conditions were set to that

Fig. 12. Seasonal variation of the rate of conversion between eddypotential energy and eddy kinetic energy, averaged between 50°S and65°S, in 0.1 Wm)2


appropriate to the open ocean, supporting the conclu-sion of SW93 and our results.

We emphasize here that the role of the atmosphericcirculation needs to be considered when studyingclimate sensitivity in the SH. Of course, there are non-linear interactions within the climate system, and thusthere is a limit to the extent to which high-latitudeatmospheric anomalies can be explained wholly in termsof the sea-ice conditions. The divergence of the variousmodels in response to climate perturbation is oftenattributed to thermodynamic feedbacks (i.e. surfacealbedo, clouds) or to the role of the ocean. However,atmospheric dynamics can a�ect and be a�ected by thesea-ice distribution and concentration, and may beresponsible of other important feedback e�ects. Forexample, Turner et al. (1997) propose a possible positivefeedback on the western side of the Antarctic Peninsula.This feedback relates periods of northerly (southerly)¯ow with negative (positive) sea-ice anomalies. Arelationship between winter season ice extent anomaliesand the meridional component of the geostrophic ¯owseems evident (Harangozo, 1994). MSLT99 proposedother feedback mechanisms regarding the meridionaltransport of heat and other properties by the atmo-sphere. Further investigation of these processes usingdi�erent climate models is clearly needed.

Acknowledgements. The authors would like to thank Dr. IanSimmonds and an anonymous reviewer for their helpful comments.The research for this paper was supported by the EuropeanCommission under contract CT94-0111 and the ANPCyT, Argen-tina (BID 802/OC-AR-PICT 583).Topical Editor D. Webb thanks I. Simmonds and P. Valdes for

their help in evaluating this paper.

References

Berbery, E. H., and C. S. Vera, Characteristics of the SouthernHemisphere winter storm track with ®ltered and un®ltered data,J. Atmos. Sci., 53, 468±481, 1996.

Boville, B. A., Sensitivity of simulated climate to model resolution,J. Clim., 4, 469±485, 1991.

Bromwich, D. H., Mesoscale cyclogenesis over the southwesternRoss Sea linked to strong katabatic winds, Mon. Weather Rev.,119, 1736±1752, 1991.

Carleton, A. M., Antarctic sea-ice relationships with indices of theatmospheric circulation of the Southern Hemisphere, Clim.Dyn., 3, 207±220, 1989.

Carleton, A. M., Synoptic interactions between Antarctica andlower latitudes, Aust. Meteorol. Mag., 40, 129±147, 1992.

Carril, A. F., C. G. MeneÂ ndez, and M. N. NunÄ ez, Climate changescenarios over the South American region: an intercomparisonof coupled general atmosphere-ocean circulation models, Int. J.Climatol., 17, 1613±1633, 1997.

Chalita, S., and H. Le Treut, The albedo of temperature and borealforest and the Northern Hemisphere climate: a sensitivityexperiment using the LMD GCM, Clim. Dyn., 10, 231±240,1994.

Fouquart, Y., and B. Bonnel, Computations of solar heating of theEarth's atmosphere: a new parametrization, Beitr. Phys.Atmos., 53, 35±62, 1980.

Godfred-Spenning, C. R., and I. Simmonds, An analysis ofAntarctic sea-ice and extratropical cyclone associations, Int. J.Climat., 16, 1315±1332, 1996.

Harangozo, S. A., Interannual atmospheric circulation-sea-iceextent relationships in the Southern Ocean: an analysis for thewest Antarctic Peninsula region, Sixth Conf. on ClimateVariations, A. M. S., January 23±28, Nashville, Tennessee,364±367, 1994.

Harangozo, S. A., S. R. Colwell, and J. C. King, Interannual andlong-term air temperature variability in the southern AntarcticPeninsula from a reconstructed record for eastern AlexanderIsland, Sixth Conf. on Climate Variations, A. M. S., January 23±28, Nashville, Tennessee, 250±251, 1994.

Harzallah, A., and R. Sadourny, Internal versus SST-forcedatmospheric variability as simulated by an atmospheric generalcirculation model, J. Clim., 8, 474±495, 1995.

Hoskins, B. J., I. N. James, and G. H. White, The shape,propagation, and mean-¯ow interaction of large-scale weathersystems, J. Atmos. Sci., 40, 977±988, 1983.

James I., and L. Gray, Concerning on the e�ect of surface drag onthe circulation of baroclinic planetary atmosphere, Qt. J. R.Meteorol. Soc., 112, 1231±1250, 1986.

Jones, D. A., and I. Simmonds, A climatology of SouthernHemisphere extratropical cyclones, Clim. Dyn., 9, 131±145,1993.

Katzfey, J. J., and K. L. McInnes, GCM simulations of easternaustralian cuto� lows, J. Clim., 9, 2337±2355, 1996.

Kuo, H. L., On formation and intensi®cation of tropical cyclonesthrough latent heat release by cumulus convection, J. Atmos.Sci., 22, 40±63, 1965.

Le Treut, H., and E. Kalnay, Comparison of observed andsimulated cyclone frequency distribution as determined by anobjective method, AtmoÂsfera, 3, 57±71, 1990.

Le Treut, H., Z. X. Li, O. Boucher, G. Seze, and M. Doutriaux,Some aspects of the LMD GCM veri®cation and development:clouds and radiative budget, Abstr. First International AMIPScienti®c Conf., Monterey, California, 1995.

Li, Z. X., and H. Le Treut, Cloud-radiation feedbacks in a generalcirculation model and their dependence on cloud modellingassumptions, Clim. Dyn., 7, 133±138, 1992.

Meehl, G. A., A reexamination of the mechanism of the semiannualoscillation in the Southern Hemisphere, J. Clim., 4, 911±926,1991.

MeneÂ ndez, C. G., Impacto del hielo marino sobre dos ciclonessubantarticos, Meteorologica, 19, 43±51, 1994.

MeneÂ ndez, C. G., V. Sera®ni, and H. Le Treut, The e�ect of sea iceon the transient atmospheric eddies of the Southern Hemi-sphere, Clim. Dyn., in press, 1999.

Mitchell, J. F. B., and T. S. Hills, Sea-ice and the antarctic wintercirculation, a numerical experiment, Qt. J. R. Meteorol. Soc.,112, 953±969, 1986.

Mitchell, J. F. B., and C. A. Senior, The antarctic winter;simulations with climatological and reduced sea-ice extents,Qt. J. R. Meteorol. Soc., 115, 225±246, 1989.

Morcrette, J. J., Radiation and cloud radiative properties in theECMWF operational weather forecast model, J. Geophys. Res.,96, 9121±9132, 1990.

Murakami, M., Large-scale aspects of deep convective activity overthe GATE area. Mon. Weather Rev., 107, 994±1013, 1979.

Royer, J. F., S. Planton, and M. DeÂ queÂ , A sensitivity experiment forthe removal of Arctic sea ice with the French spectral generalcirculation model, Clim. Dyn., 5, 1±17, 1990.

Sadourny, R., and K. Laval, January and July performance of theLMD general circulation model, in, New perspectives in climatemodelling, Ed. A. Berger and C. Nicolis, Elsevier, Amsterdam,1984.

Schwerdtfeger, W., The seasonal variation of the strength of thesouthern circumpolar vortex, Mon. Weather Rev., 88, 203±208,1960.

Schwerdtfeger, W., and F. Prohaska, The semiannual pressureoscillation, its cause and e�ects. J. Meteorol., 13, 217±218, 1956.

Simmonds, I., The e�ect of sea-ice on a general circulation model ofthe southern hemisphere, in, Sea level, ice, and climate change.


Proc. IUGG Canberra Symp., December 1979. IAHS Publ. 131,193±206, 1981.

Simmonds, I., and W. F. Budd, Sensitivity of the SouthernHemisphere circulation to leads in the Antarctic pack ice, Qt.J. R. Meteorol Soc., 117, 1003±1024, 1991.

Simmonds, I., and M. Dix, The circulation changes induced by theremoval of antarctic sea-ice in a July general circulation model,Proc. Second International Conference on Southern HemisphereMeteorology, Wellington, New Zealand, December 1986.American Meteorological Society, pp 107±110, 1986.

Simmonds, I., and D. A. Jones, The mean structure and temporalvariability of the semiannual oscillation in the southern extra-tropics, Int. J. Climatol., 18, 473±504, 1998.

Simmonds, I., and X. Wu, Cyclone behaviour response to changesin winter southern hemisphere sea-ice concentration, Qt. J. R.Meteorol. Soc., 119, 1121±1148, 1993.

Simmons, A. J., and B. J. Hoskins, The life cycles of some nonlinearbaroclinic waves, J. Atmos. Sci., 36, 414±432, 1978.

Streten, N. A., and D. J. Pike, Characteristics of the broad scaleAntarctic sea-ice extent and the associated atmospheric circu-

lation, 1972±1977, Arch. Meteorol. Geophys. Biokl., 29, 279±299, 1980.

Trenberth, K. E., Storm tracks in the Southern Hemisphere,J. Atmos. Sci., 48, 2159±2178, 1991.

Turner, J., J. King, S. Harangozo, W. Connolley, and G. Marshall,Major changes in the climate of the Antarctic Peninsula duringthe last 50 years, Seventh Conference on Climate Variations,A. M. S., February 2±7, Long Beach, California, J88±J91, 1997.

Tzeng, R. Y., D. H. Bromwich, and T. R. Parish, Present-dayAntarctic climatology of the NCAR Community ClimateModel Version 1, J. Clim., 6, 205±226, 1993.

Van Loon, H., Meteorology of the Southern Hemisphere, Meteor.Monogr., 35, Ed. C. W. Newton, American MeteorologicalSociety, pp 59±99, 1972.

Van Loon, H., and J. C. Rogers, Interannual variations in the half-yearly cycle of pressure gradients and zonal wind at sea level inthe Southern Hemisphere, Tellus, 36A, 76±86, 1984.

Xu, J.-S., H. Von Storch, and H. van Loon, The performance of fourspectralGCMs in the SouthernHemisphere: the January and Julyclimatology and the semiannual waves, J. Clim., 3, 53±70, 1990.


The storm tracks and the energy cycle of the Southern Hemisphere: sensitivity to sea-ice boundary conditions

Documents