JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. C2, PAGES 2995-3011, FEBRUARY 15, 1998 Sea level variability and surface eddy statisticsin the Mediterranean Sea from TOPEX/POSEIDON data Daniele Iudicone and Rosalia Santoleri Istitutodi Fisica dell'Atmosfera,ConsiglioNazionaledelleRicerche,Rome,Italy Salvatore Marullo and Paolo Gerosa CentroRicercheCasaccia,Ente per le Nuove Tecnologie,l'Energia e 1'Ambiente,S. Maria di Galeria (Rome), Italy Abstract. Two yearsof TOPEX/POSEIDON altimeterdatahave successfullybeenusedto studythe mesoscalefield in the Mediterraneanandto investigatethe seasonalandyear-toyear variability of the sealevel andeddy statisticsin this basin.The mesoscalefield described by TOPEX/POSEIDON revealeda strong,but subbasindependent,seasonalsignal.Year-toyear variationsare alsoevidentin termsof bothintensityandpositionof the main mesoscale features.Crossoveranalysisindicatedthe existenceof a meridionaltransportof eddy momentumaway from the Algeriancurrentdueto the northwardmigrationof mesoscale eddies.A comparisonbetweenmesoscale featuresdetectedby the altimeterand contemporaneous featuresobservedusingadvancedvery highresolutionradiometer (AVHRR) seasurfacetemperature hasbeenmadefor the followingMediterraneansubbasins: theAlgerianbasin,the Levantinebasin,andtheTyrrhenianSea.The resultsdefinitivelyprove the direct relationbetweensealevel anomaliesand the Mediterraneaneddy field. only 10 cm. Second, the coarse separationbetween subsatellite tracks(240 km at 40øN)in respect to thetypical Earth-observing satellites are providing a significant dimensions of theMediterranean mesoscale (30-100km)may quantityof high-qualitysynopticdatafor the investigation of significantly reducethe advantage of the high-frequency !. Introduction oceanographicphenomenaon spatial and temporal scales repeatcycle (-10 days). unattainable with in situ measurements. Altimeter data have alreadyprovedto be able to resolveseasurfacevariabilityin regionscorresponding to the majorwesternboundarycurrents (Gulf Stream, Kuroshio, Malvinas, and Angulhas) and the Antarctic Circumpolar Current. On the contrary,very few works have been done on studies of ocean areas characterized by smallsealevel slopesor on smallsemienclosed seaslike the Mediterranean.This lack of researcheffort could probably be due to the poor precisionof radar altimetersoperating before ERS-1 and TOPEX/POSEIDON. The launch of the TOPEX/POSEIDON (T/P) satellite Howeverthesizeof Mediterranean Sea,although relatively small,is sufficient,so that circulation maybe governed by large-scale oceandynamics. TheMediterranean is amongthe mostinteresting semienclosed seasbecause of thegreatrange of processesoccurringwithin it [Robinsonand Golnaraghi, 1994]. Most of the physical processesthat characterizethe global generaloceancirculation,many of which are not well known or understood, occur analogously in the Mediterranean.All major forcing mechanisms,surfacewind, buoyancyfluxes, and lateral massexchange,are present.Airsea interaction is vigorous, and both deep and intermediate water massesare formed. Increaseddensity due to saltiness causedby intensesurfaceevaporationis importantfor water massformationand for the main thermohalinecell in analogy to the global conveyorbelt. usheredin a new era in satellite oceanographyby greatly improvingthe accuracyof the altimetermeasurements. The orbits for T/P are estimatedto have a radial accuracyof 3-4 cm root-mean-square (RMS) [Tapley et al., 1994]. For this reason,T/P is the most accuratealtimetersatelliteflown up to The Mediterranean is located between 30 ø and 45øN and is date. The main goal of this work is to use T/P data to connectedto the Atlantic by the shallowStrait of Gibraltar.It contributeto the descriptionof the subbasinand mesoscale is composedof two similar-sizebasins(the westernand the dynamic features of the Mediterraneancirculationand to easternMediterranean Seas) separatedby the shallow and comparethe resulting picture against other observational narrow Strait of Sicily (Figures 1 and 2). In the Strait of evidences. Gibraltar,at the surfacea comparativelyfresherAtlantic water Thereare manyreasonsthat suggestthatthe Mediterranean flows into the MediterraneanSea to replace both the water Sea is a basin where T/P performancesare under critical that evaporatedbecauseof the dry winter winds and the conditions.First, sealevel anomalies(SLA) are of the order of denser,saltier Mediterraneanwater flowing out at depth into the Atlantic. The incoming Atlantic water forms a surface water Copyright1998by theAmericanGeophysical Union. Papernumber97JC01577. 0148-0227/98/97 JC-01577 $09.00 mass 100-200 m thick that flows eastward. The Atlanticwater,movingeastward,is heatedby and mixeswith the saltier surface Mediterranean water becoming the modified Atlantic water (MAW). 2995 2996 IUDICONE ET AL.: SEA LEVEL VARIABILITY IN TIlE MEDITERRANEAN SEA z 40 I> 35 I> 30 I>,, 5 Iong.W Figure 1. Map of theMediterranean Sea.A pictorialview of the surfacecirculationin theMediterranean Sea hasbeensuperimposed; permanent featuresare solidlinesandrecurrentfeaturesaredashedlines.(Modified fromMillot [1987] andPOEM Group[1992]).Acronyms usedareasfollows:westernAlborangyre(WAG), easternAlboran gyre (EAG), Almeria Oran front (AO), Balearicfront (BF), LigurianProvencalCatalan current(LPCC), northemTyrrhenianeddy(NTE), southernTyrrhenianeddy(STE), Algerianeddy(ALE), IonianAtlanticstream(IAS), Pelopseddy(PE),Cretangyre(CG),Iera-Petra anticyclone (IPA), Rhodesgyre (RG), AsiaMinor current(AMC), Ciliciancurrent(CC), western Cyprusgyre(WCG),Mid-Mediterranean jet (MMJ), Shikmonaeddy(SE), andMersa-Matruhgyre(MMG). In the Levantine basin, Levantine intermediatewater (LIW) of relativelyhigh temperatureand salinityis formedduring winter. This water mass circulatesthroughboth the eastern and western Mediterraneanbasin in a generally cyclonic fashionbefore arriving to the Atlantic throughthe Strait of Gibraltar.LIW is usuallyobservedbetween200 and 800 m of depth.Deep wateris producedfor the westernbasin in the Gulf of Lions and for the eastern basin in the south Adriatic and northeasternLevantinebasin [Robinsonand Golnaraghi, 1994]. Since the middle of the 1980s several international are the Westem MediterraneanCirculation Experiment (WMCE) [WMCE Consortiumof investigators,1989] the PhysicalOceanography of the EasternMediterranean (POEM) [POEM Group, 1992], and the GibraltarExperiment[Kinder and Bryden,1987]. In the frameworkof theseprograms, numerousoceanographic campaignswere made.The resultof thisinternational effortis a completely changed pictureof the Mediterraneancirculationcharacterized by a more complex generalpatternthan obtainedfrom earlier studies.The newest concepts of the Mediterranean circulation focus on the existence of three scales of motion, i.e., the basin scale researchprogramswere performedin thisbasin;amongthem defined by the overall thermohalinecells, the subbasinscale 4'5N/I I I I I I I I I' I I I 1 I I I I I I /I • I 1 1 I I I I I I l"l I I I 1 1 I I I I I I / • •_ ,ION N• 3,5 6W - ß •)•0 5E ) , 10E MAX DEPTH 16E = 4636.8 20E M 26E Figure 2. Mediterranean bottomtopography. 30E IUDICONE ET AL.: SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA 2997 defining the gyres of the main thermoclineconnectedby intensejets and meanderingcurrents,and the ubiquitousand very energetic mesoscale eddy field modulating the two previousscalesthroughmultiple and nonlinearinteraction.A pictorial view of the Mediterraneancirculation,as resulting from WMCE and POEM experiments,is given in Figure 1. The contrastsin sea surfacetemperatureand the small cloud cover make the satelliteinfrared imageryan appropriatetool to studythis basin. In the last yearsseveralstudieswere carriedout usingsea surface temperature (SST) data. These works revealed the the MediterraneanSea. They found that T/P inversebarometer correctionsometimefails in this basin, occasionallyyielding SLA shiftsover singletracks.They overcomethis problemby adjusting the SLA bias of these tracks with respectto the background.They demonstratethe importanceof the steric circulation. Sea,startingfrom November22, 1992 (i.e., from cycle7), wereusedfor this study(Figure3). Previouscycleswerenot usedbecauseof attitudeproblems[Neremet al. 1994]. Datawereinitiallyeditedon thebasisof qualityflagsand effect on the mean Mediterranean sea level variation and show the potentiality of the T/P altimeter data to catch the main features of the circulation. The purpose of this paper is to examine the sea level mesoscalevariabilityof the MediterraneanSeaduringthe first 2 yearsof the T/P mission.The paperis organizedas follows. richness of mesoscale features in the basin and indicated the Section 2 describesthe processing.Section 3 deals with importanceof thesefeaturesfor the Mediterraneandynamics repeat-trackanalysisand large-scalecirculationas observed [e.g., Philippe and Harang, 1982; La Violette et al., 1990, on T/P. Cross-overanalysisand Reynoldsstressare discussed Millot, 1991]. In particular,recently,IR imageswere used to in section4. SLA along some particularlyinterestingtracks infer the seasonaland interannualvariability of this basin that cross the main features of the Mediterranean circulation [Santoleri et al., 1994; Marullo et al., 1998]. These works are analyzed and discussedin section 5. The results are demonstratethat a strongseasonalsignal is presentin the summarized in the conclusions. whole basin with different regional characteristicsand is mainly due to the seasonalvariability of the main mesoscale 2. Altimeter Data Processing structures.They showan interannualvariabilityof seasurface temperaturepattern causedby the variability of the surface Two years of T/P altimeter data over the Mediterranean The altimeterdata are particularlyuseful for studiesof the ocean circulation [Wunschand Goposchkin,1980] because the measurementof the surfacetopographycan be converted into surface geostrophiccurrent velocity. Therefore the analysisof altimeterdata of the Mediterraneancan be useful to studythe variabilityof the circulation on the global scale and identify its main characteristics. B6hm et al. [1992] first tried to use altimeter data in this basin. Their analysis of Geosat data was limited by the capability of the Geosat altimeter and by the lack of knowledgeof the tidal correction to apply in the Mediterraneanbasin. Nevertheless,they demonstratethe effectiveness of altimeterdata to investigate the Mediterranean circulation. parameter ranges as recommended in the user handbook [AVISO, 1992]. A secondqualitycontrol(outliersdetection and elimination)basedon a procedurespecializedfor the Mediterranean casewas alsoappliedafterenvironmental and geophysicalcorrections.The following correctionswere appliedto bothTOPEX and POSE]DONdata:watervapor from the onboard radiometer,TOPEX dual-frequency ionosphericcorrection,dry troposphere, invertedbarometer, electromagnetic bias,centerof gravity,solidEarthtide,pole tide, and net instrument correction. The ocean tide model Recently, Larnicol et al. [1995] used T/P data to investigatethe circulationand the meansealevel variationsin ß: .',.:.• "," .:,5:'::;:..::•""!,'!:;:::: ;'.,i'?.',:'•':•::..:";•..:;"i• , ,.( suggestedby the user handbookfor the MediterraneanSea (P. Canceilet al., Barotropictidesin the MediterraneanSea '::.' :- j ': .... • .: .,., ' '•.:,:•.'";'"'"; :.'.•i/;,..,' ,.?-.••:•5'' ß::(•- 35 •> .... .% 15 Iong.W10 5 0 5 10 15 20 25 Figure3. TOPEX/POSEIDON(T/P) groundtracksin theMediterranean Sea. 30 Iong. E 35 2998 IUDICONE ET AL.: SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA The SLA obtained using the conventional repeat-track using a finite element numerical model, submittedto the Journal of GeophysicalResearch,1994) wasusedto subtract analysismethod were used to computesea level variability the main ocean tide constituents within the Mediterranean (SLV) mapsof the entireMediterranean.The RMS of the sea Sea, while outside the Mediterranean, "modified enhanced level residualwas computedat eachT/P grid point usingall Schwidersky"tides [AVISO, 1992] have been used. Finally, the SLA data. the JointGravity Model (JGM-2) preciseorbit wassubtracted. Eddy kinetic energy(EKE) for the entire data set was also In order to evaluatethe mean sea level, altimeter passes computedat eachgrid point alongthe T/P track.Maps of were interpolatedto a fixed grid definedby eachfirst satellite SLV andEKE werecomputed by objectiveanalysis usingthe track free of gaps.The resultingmean grid size was -6 km. correlationfunction(1). As a consequence of the combined Interpolatedaltimeterdata were usedto computea mean sea effect of the data pointsdistributionand the usedcovariance level that was subtracted from each single altimeter pass function,the interpolationerror rangesfrom a minimumof 8% along tracks to a maximum of 75% in the centerof the obtainingSLA data. Becauseof the very low sea level signalof oceanographic T/P grid. featuresin the MediterraneanSea a particularattentionhas Apart from these considerationsabout the statistical been paid to the outliersdetectionphase.We adoptedan ad interpolationerror,the large spatialgapsbetweenthe tracks hoc algorithm based on the objective analysis theory (dueto samplingcharacteristics of the T/P orbitrepeatcycle) [Brethertonet al., 1976]. Objective analysis(OA) has been imply that small-scalefeaturesalwaysexistingnearthe center widelyusedto interpolatein spaceandtime sparsedataontoa of T/P mashescannotbe detectedby this altimeter.Therefore regular grid but rarely applied to individuate outliers as any interpolationmethodthat can revealtheir presencedoes suggested by Breterthonet al. [1976]. FollowingBretherton not exist. This is particularlyimportantin the Mediterranean, et al. [1976], we use the discrepancybetween OA estimate where mesoscaleeddiesare characterized by a diameterof and the measuredvalue as an indicatorof whethera particular few tensof kilometers. OA horizontalmapsshouldsimplybe observationis grosslyincorrect.We optimallyinterpolatedthe consideredas a two-dimensional representation of the along SLA along the satellitetrack over measurement points,and track signals,which on the contraryhave enoughspatial we calculated the difference between the estimated and the measured SLA. Measurements that deviate from the estimate for more than 3 times the standard deviation of the differences were eliminated. The covariance function used for the objectiveinterpolation wasdirectlydeduced fromthedataand is F(x)= A1e-(x/B• ) + A2e-(X/B 2) where A•=0.825, A2=0.175, B•=90 km, B2=20 km, and x is the distance. The objective analysismethod for outliers detection,even if rather time-consuming,worked extremely well in all the cases.A visual inspectionof data samplesconfirmedthat the algorithm is really able to individuate the sameoutliersthat a "manual editor" would individuate but can definitively individuatemore quickly (and more objectively). Finally, slope and bias have been subtractedfrom each track in order to separatethe variability, causedby steric effectsinducedby seasonalwarming, and seasonalvariations, caused by large-scale quasi-stationary currents, from mesoscalevariability.The removalof slopeand bias, even if not strictly necessaryowing to preciseorbit determinationof T/P, also contributed to remove the relative bias found resolution. In thefollowingdescription of thehorizontalmaps of SLV and EKE, we will focus on features observed in areas of relativelysmallinterpolation error. 3. Variability of the Dynamic Ocean Topography The map of SLV for the first 2 yearsof the T/P mission (from cycle 2) have alreadybeenproducedby Larnicol et al. [ 1995]. We computethe samemap (seePlate 1) using2 years of T/P data(from cycle7) andafterbiasandslopeadjustment. Besidethe different analysismethod and the almost2 month shift betweenthe two data setsthe two SLV patternsare very similar. In our case the bias adjustment applied to T/P data considerablyreducedthe meanSLV by removingthe seasonal signal due to the steric effect and other basin scaleeffects [Larnicol et al. 1995]. The SLV ranges from a minimum value of 1-2 cm, located in the area south of Cyprus, in several small cells in the Ionian Sea, and in the area offshore the Gulf of Lions, and maximum values of 8-10 cm southeast of Crete, mainly causedby the variability of Iera-Petra(IPA, Figure 1), southwestof the islandof Sardiniabecauseof the Algerian current eddies (AE, Figure 1) along the north between TOPEX and POSEIDON altimeters (for instance, African coast, and in the Alboran Sea because of the Alboran Nerem et al. [1994]) and to eliminate biasesdue to occasional gyressystem(WAG, EAG, AO, Figure 1). failure of the inverse barometer correction [Larnicol et al., Note that the mentionedfeaturesare crossedby T/P tracks 1995]. The residual sea surface velocities can be calculated from along track sea surfaceslope using the geostrophicequation; that is, for ascendingtrack _Va =gdh f dx (2) The componentof the geostrophicvelocity perpendicularto each track were estimatedfrom the SLA slopes.Slopeswere obtainedusinga linear fit over sevenT/P datapoints. (superimposed on Plate 1) where the interpolationerror is small. In order to investigate the seasonal and interannual variability of the SLV and EKE fields, maps were computed by dividing our data set into four parts, each of one half a year. From the meteorologicalpoint of view, spring and autumn in the Mediterraneanarea are very short and are characterizedby a rapid sequenceof summerand winter situations[Meteorological Office, 1962]. Recent works by Marullo et al. [1998] and Santoleri et al. [1994] suggested IUDICONE ß. ET AL.' SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA 2999 / ß Plate 1. Sea level variability (SLV) for the period from November 1992 to November 1994 Contour intervalsare 0.5 cm. T/P tracksare superimposed. that also in the sea, intermediateseasonsare short(1 month). slightlystrongerin winter-springthanin summer.This means Therefore we have decided to divide the year into the that the differencein the eddy kinetic energyand sea level following two periods:"winter" from Novemberto May and variability, observedduring 1994, should only be due to a "summer" from June to October. The time series of the 2 different level of intermittenceof the two-gym (sometimes main seasonsfor the 2 years(Plate 2) revealsevidentseasonal three-gyre)systemof the AlboranSea.Unfortunately,we do vailabilities as well as differencesbetween corresponding not have data to supportthis hypothesis.Someanswercould seasonsin the 2 years. be providedby a systematicanalysisof the infraredimagesof Seasonal variability is more evident in those areas the 1993-1994 period.A large set of imagesis not availableat characterizedby a higher energy level. In the western present,and suchanalysisis out of the scopeof this work. Mediterranean Sea the Algerian basin clearly shows the East of the Alboran Sea, a seasonalchangeis more evident higher seasonal variability, the winter being generally in 1994 rather than in 1993 with an absolute maximum of characterizedby a higher level of EKE and SLV. In 1993 the activity in summer 1994ß In the Ligurian-Provencalbasin winter to summerchangeis evidentin termsof extensionand (northwest part of the western Mediterranean Sea) the intensityof the high-energyarea,while in 1994 the seasonal seasonality is markedby a deeperwinterminimumin the Gulf variability is limited to a decreaseof the energylevel in the of Lions. cell southwestof Sardinia. It is worth noting that SLV and In the TyrrhenianSea the seasonalsignalis not so evident. EKE are more intense in 1994 than in 1993. A more detailed investigationabout the seasonalityin this In the Alboran Sea, while energy levels are fairly basinwill be donein section5.3 analyzingthe time evolution equivalentduring the summerand winter of 1993, in 1994 we of SLA alongsingletracks. observean increasedenergy level in summerrather than in In the Ionian basin we always observea very low and winter. On the basis of our data we cannotgive a definitive patchysignal with a minimum during summer.Immediately explanationof this particularbehavior.The variabilityof the southwest of Greece, where two T/P tracks cover the area of sea level should,in any case,be related to the variability of Pelops (PE, Figure 1), a more pronouncedvariability is water mass exchangesbetween the Atlantic Ocean and the observedin 1994 rather than in 1993. In 1993, Pelops Mediterranean Sea throughoutthe Strait of Gibraltar and to produceda small signal (SLV) only in winter, southwestof the intermittent presenceof the two-gym system in the Greece.During 1994 a maximumof SLV appearssouthof Alboran Sea. In situ sea level measurements[Bormans et al, Greecein winter and westof Greecein summer,while a peak 1986] suggestthat flux intensitiesand flux variability are of EKE is visibleonly in winter.The 1994 mapssupportthe 3000 IUDICONE ET AL.: SEA LEVEL VARIABILITY Eddy Ki-'neticF•nergy IN THE MEDITERRANEAN SEA St,and,xrd Deviation ., ! 0.6 3 55 '75 95 1.3 .I •,.8 3.6 > 43 :':'' .3 Plate2. Seasonal mapsof eddykineticenergy(EKE)(leftpanel)andSLV (rightpanel).Unitsarein centimeters for SLV andcm2/s 2 for EKE. 5 IUDICONEET AL.: SEALEVEL VARIABILITY IN THE MEDITERRANEANSEA 3001 that the originaltime seriesof altimetervelocities hypothesis thatPelopscanbe foundin twodifferentpositions Assurning givesnearlyindependent estimates, eachcrossover pointwill depending ontheseason [Theocharis et al., 1993]. The area southeast and east of Crete shows a strong seasonal and interannual variation of SLV and EKE. This have a maximum of 74 independentvelocity estimates meridional front [Marullo et al., 1998]. The existenceof the Mersa-Matruh gyre (MMG, Figurel; POEM Group, [1992]) is revealedduring wintertimeby the is (degrees of freedom)overthe 2 yearperiod.On the basisof effectis mainlydueto theIera-Petraanticyclone (IPA, Figure the numberof degreesof freedoma statisticalF-testcan be 1) activityas alreadypointedout by Larnicolet al. [1995] usedto estimatehow closewe are to the expectedvarianceof analyzingmeanseasonal SLA maps.A detailedanalysisof an infinite numberof samples.On the averagewe have 60 degreesof freedomat eachcrossover point.This meansthat the T/P data in this area will be done in section 5. In the Levantinebasin,alongthe Africancoasts,the north with a 95% confidence level we are within 32% of the variancefor aninfinitetimeseries.In thecaseof no Africancurrentmarksits presence by enhanced variabilityand expected EKE in proximityof the Libyancoastsonly duringwinter datagapsthispercentwouldbe 29%. The magnitude andthedirectionof theeddyvariabilityare 1993.Duringsummer,waterexchanges betweenthe Ionian by thevariance ellipses. Thedirection of theaxis basin and the Levantine basin are stronglyreducedby the represented counterclockwise fromeast increaseddirnensionof the west Cretangyre and by the 22øE of principalvariabilitymeasured tanO = 0.11<u'u'> presenceof an area of increasedvariabilityalongthe African coast south of the Rhode gyre (RG, Figure 1) area. This is (4) more intense in 1993 rather than in 1994. whereO']]isthevariance alongthemajoraxisandisgivenby 4. Crossover Analysis 0'11 =•(<u'u'> +<v'v'> The EKE maps discussedabove are obtainedunder the (5) assumption thatvariancesalongthe ascending anddescending tracksare similar. To quantify anisotropy,we computedthe whilethe variancealongtheminoraxisis principalaxesof variancesof thevelocityfluctuations. Following Morrow et al. [1994], the componentsof the geostrophic velocityperpendicular to each track (ascending (6) 0'22 --(< u'u'>2 + < v'v'>2 )--0'11 anddescending)were calculatedat eachcrossoverpoint.Then they are interpolatedlinearly to commontime (every 5 days) Anisotropicflow is represented by an elongatedellipsis,with and used to derive the zonal (u') and meridional (v') the principaldirectionof the velocity variancealignedwith components of the geostrophic velocityfluctuations the directionof the major axis. Even if the spatialdistributionof crossover pointsfor T/P (-V' a+V'd ) in the Mediterranean basin is very coarse, this type of 2cosgo analysisgivesan indicationof the anisotropy of the examined (3) area. The principalaxes of the variance(Figure4) show a +i(<u'u'>-< v'v'>) 2+4<u'v'> 2 v,=(Va-Vd ) stronganisotropic natureof theregion.Thereis a tendency of 2sintp the varianceellipsisto be alignedwith the bottomtopography whereq0is the anglebetweengroundtrackand the north (Figure2) especiallyalongthe basinboundaryeitherin the meridianandV,,andVaarethecomponents of thegeostrophic MediterraneanSea or in the Atlantic Ocean. High values of velocity perpendicular to theascending anddescending track, eddyvariabilityare foundalongthe Algeriancurrentand in the crossovereast of Crete, where high valuesof SLV were respectively. observed. The crossoverat 0øE 37øN has the major axis The covariancesof the velocity fluctuations,<u'u'>, <v'v'>, and<u'v'>, are thencomputed from the 2 yeartime oriented from NW to SE indicating a strong and highly seriesof 5 day interpolatedu' and v' velocitycomponents. variabletransportfrom SpaintowardtheAfricancoast. 45 '• 35 3O , i -10 ......... i 0 ......... i ......... 10 i 20 , i 30 Lon (degrees E) Figure4. Principalaxesellipsesfromcrossover variance. 4O 3002 IUDICONE ET AL.: SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA • .,o..,..,.. ;•.:,,,,,,,, •,;;.,,3,:,,,..o. v,.•' .:•:: ,,..,';.! ;•..1'..:.•,.,..... ß.V';"?:'.:.: :.¾ :'•&.'.:., ::;'•,'•7:'.'o•, ":,.";d ½,'."•' e';.;, •.::•.: ;•.'::' •' ?v ,' •,vq.~ ,•,*•.,';,,;,,v',,.• ":,',%',,,.. ';,"• ••.•.':':, ,'"'•.'d":."/: :"!.':½i:'2 •-•',,,".,q';':.',',",•'..,'?'.' '"•.•'• '."."•.•?;,'"...•:; .... .: ;-'• :: '•2 .-.l')[".\,.. -........:,,,...-,,.•,,:, ..:•.., ?.,.... •: ', '-,.....ß.' •., ..... v,. '":':'. ;",'.'• .,•.'•7.":' t::.'. •"'•,,;';.';'%" )'i7';.'! ,'•6',•'.!;,',,,".,;,• •,':q'.':',;.'!',',;,' •'.'.?:,•':•,•.' ..... . •:'• :.i•,-•3" :.• .... ',.2',":.'::,'.:•.",,'' '.' ,'.7:".'.'%'.' '., .'~•: ß!•'•.•....., ;)v.'-';.-::',.::•;,•::.,,?.,q:.:',. ,'-'..':':".."..' ß.,',•x :,....•,; .•.,• :'.".•:!:": ', '.'..', • • ." ',.: 3:;; .... ",.",•;;• •'",q:•5::;!',':• :•,'.:"." .:.:?: ':' '•," ':'.";:' . ß. 'r'.:,,::';;•,.,•,,;. ?,:y;;...,: i:..¾ ,..•....y.,,) ,.)... , '?...•,':.,•,•,:v., :2.'.: 2 '.:".":.:. •.:;i..: . •' ß '. ':,:•.:-,::. :'. ':.•,;'..4•.'•-•-'..?, .....';'.':•: •:'.'•';:?. •d, ß•':.•:':'5.'k": " '"::'?:;i:•: .• . -3'7.1 . ' ."i:. ....:',........ - , '.":.;:•,'.:.;•, ;.•,•i..,...;.•.•,,,•l'.•,:./.:•.:2...V".'?;';,: i;. •"•' •.'.•.'.•,",".;' ,.';•4•( ,..•.':",•"'d "& :'vø' _• ;•,."1 42i'•?.?:':.:t.•'.•,:"..•.':•/•:',:'.'::v".'•",,. '.:'';: ß • ::1..'t'"-'::!•.!, [:•3:•. ,:,•,?.:'.r;-..•.'•,'5;•.V•.:•'..."; -' ' .',.'7 •. ,;,•:.,.•,:•?:.,...,: •.?,.. -.':!','.:'.:;:'.'!. :;'."O';;.•.•.,'.:'.:..': .... :: :.½•;. v,.:•.';.c, ,....:,,; :?,,: ;,:.•':.7 ,,:...': ...... '....ß', o.1 -12.3 1,4.. •'-.." •'"'•"' '•";'•;":'":"i Dv•,•,'.q,•.•;..'•,' .... ,.,:;' •'.. :•,•"•,";'.',•.'•.'.•" . ::,.'::'i%' ',',,?:•2.•,;:3'2•-.'77;•...• ,..:',•?,';' v':'2. c;•::..,' !'i::'.':.bi":';;.;',,,•,'".•.",:,•: ,4',•'f',•,ii,'•;.':c. ;.""'".'•'•:.:4•'.:. ,',:• ' -'.••""•':.:- •.e• :....,.,'• •.;". .',,'.,•,':..:'..:,•,'•:'• ;.::: ,,, .'•:',:'"m•a,•7," ,:,'.,.',,.."..',..'3";,.": ß ' :-•':.-;",:' L'œ' . '.•':i;•?!:.:?(•:ti: •..'.:;.:;;• ;':. ':.... • ":"•'•;'".:'.,':' ' :.".",•"".L'.•:•'.• " -' ':':,:'..•.:"' ß ß"',.... :''•;:,-" ß:"'%'-•'::,;'!•,,.". '.::':::•'• "3"•' ß ".':-';,"'*:i'•":," :•:,'" ,:,q':.. '•;' ' '.',C3",,":• . ,.•':c.'•.;'.'. ,::'..;..,..':,•',.•. /.::.'.,.: '...'..',::•..' .,,!;:'(.':, ',:.' :,..:, _ .•1•34.7 .3 17.1 .... -D..6 2•• .':?!:?;t?;!47::?.:•.:i.".•;',,:..'.:::;•:: ,?,..:. . o 27.6 o '•,.,' .' ß ß -1 ß•.: ,; 1.0 ß •{,-,'.,'.....•:,... ' ..... ;(..... ','. • ',:'.•.• :•:;.,,:,;, ,'......'',.';.'::.•'..'-'.•:.:.'.'.,, ,'.'•.;•. '•...•,-..•"3:. •r . v:•3;'g' '• :•:,?:,'- '• q: "".' ß .:•-:.½-: '-'.'.'..? ' •,:'•';;.:,;,:,.'•,'..,'.•:i •;,'(,: ::.'•. • •,'",'.; •'•. ,::., ,,..•;...,t•: ',.:b::;'-.'•,,.•;, ;73.•i•.. :i:'.,i?,,;,•;•: ....... ;..•, ........3':/ ß " ' ..... •.....,: :,:.:,:.,t: .... ..-.'. . ... ...:....,•.... •l, ,': ..'•. .'•7, ][ - t,.,.,, • 34.6 -51.3 ':!7i!",,'):!: ..... •:•;" -67:5,..., 73.9 -28.3 5,4-33.8 .... .... ,,98 .?:•,*,'.:;: (:;':;!;..:, , ..:','!. ,:•:.'...',,."?;•.' ,'" t" ..... '.:.:_,i/.'..}',',:.'tr'.'('?.•.•,.•:.:',. 'i$.!;': "" !'"•;::':"'t' "";•"q: ..... '":"'"" "; ........ ';?. :.': '• :,::'."c ?':' :.•.'.?:(. ? i::¾':•!: •.•;,:,•il; '.a, •8ß7 0.9ß -4ß4 '"q•laL•.8 ,•. ø '; .'.:"".,: ....'.......... :'¾,:,.;•9': -•5%:,.,:.,.i.c,•.::'-',;.'.;',,½•.::?;,.':',•;.':,i,;t,:•q.:' ß ',g?•:.'.??'.'!:'"'•')'-)?..;:::::%'.'-.-:.•.•,•':i.'.•,; -',.'.,':.'•.;'..:,.t'5,'I::%'.',¾".':•..%?."' ß ß .t.. .•. ,• ,.,. o,. •...; "."' ',' .' .• ' ".' ....... • •.•,.:'.-ß , ..... . • • •'".' ß :. .... .... ß, ,' •. *',•::'.,,':?..'•;"• ....; -14.2 ß ?.t•,'• ?':.?: ':,::.'v ':'..Q•i: ¾;',', :7,. :/'.:"'•.•.. •i'•::':,' •."',;',t;Y'•'•,'7,'"3•,?;'?:O•"z,?;•'bt•'.?;:"•;5) ' '":•":. I: : ß'•'. . t,::...',3 ,: •.'.:•':-•' .. 4;,:'' '?'--,:','F,.'.'. '•',,•-"'0. ': ,-'.'3•", i., ' '.' ':•/;,t'..,ß",::."'•..:".';':.' '.&'•"k,,? 0 11,t3 -6_. ,, ' •:':i':'"::"ß•""?!!k);'!i:•i'"'""'"?•ßf;"' '"::'r'"?'•""'"':"':':":•':"':"'"'t•:•':"":"';":""""'• ..... .' • ':'!';::'".':. ', ::' '•,:.•'t',:.'.,:;:',--::,' . ..•,...• ,'•.,.v... .:,., , t•.,','.Y:" ,'"'.: )'?•"':" ",':,":,..::,.:'v, ..::...: :..: •.•.,, '.,,,•. :'.c:::., :'•..",',:, '?.';.;'. ßß ß ..',;;;.%c '.. :(:...t., •.•':;: ,:,-,,;, ...':• :,.•,,' : ...'.:.:.'. 3':v* '; ,.. ,:',-.,....'::..,:.., '•. :,'.'.:•;.,......•':'vi:'?-;: '.:•'.ß.":•:.ti...:t?.'",•:,.'• ß '.;::.-':, .-:'i•:;2•:,'":ci.. - .:J:.'•',:.•': .?:) 'i• ..•..)!i." '!':: ?:'.%",'.:" ..',:':.': ? "(:. , ,, ß •' ß .- ,;•-.";,'+ :',",,t: ..... ".'.l,t5 '"•,; Figure5. Mapof <u'v'>distribution. Values arein cm2/s 2. In the TyrrhenianSeathe ellipsesfollow the directionof 5. A Closer Look at the Tracks the main current. This seems to indicate that the main A way to better understandlinks betweenSLA derived from T/P data and dynamicalfeaturesof the circulationis to makea comparison with othercontemporary datasets. We will utilize advanced very high resolution radiometer (AVHRR) East of Crete, in an area betweenIera-Petraand the Rhode gyre the orientationof the major axis in the northwest data to compareSLA anomaliesto thermal signaturesof subbasins: quadrant impliesthatthe<u'v'> is negative andindicates the dynamicalfeaturesin the followingMediterranean the Algerian basin, the Levantine basin, and the Tyrrhenian direction of the residual flow along the oscillating front circulationis responsible for the eddyfield advection.In the Ionian Sea, all the ellipseshave similar intensity,and the north-southorientationindicateseddyflux in thatdirection. Sea. These areas have been chosen on the basis of the results between Iera-Petra and the Rhode gyre. The computed Reynoldsstresses canbe usedto inferthe energytransferbetweenthe meanflow andthe eddyfield. Neglectingcontributions due to vertical velocities,the exchange termof kineticenergybetween theeddiesandthe meanflow canbe writtenas [e.g.,WilkinandMorrow, 1994] 3U 3V 3U 3V of the SLV analysis (section 3) that stressedan intense mesoscaleactivity in the first two areasand a connectedcell in the north TyrrhenianSea off the Bonifacio Strait. The National Oceanic and Atmospheric Administration (NOAA) satellite passescorrespondingto the days of T/P measurementsover the Mediterranean were acquired. Sea surface temperature (SST) and channel 4 brightness temperature maps of the selected areas were produced. Clearly, the cloud cover often partially obscuredthe view of surface temperature field of the three selected basins. Nevertheless, the typical cloud cover condition of the Mediterraneanpermittedus to have enoughclear or partially clear imagesto analyze. The comparisonhas been done by superimposingalong track sea level anomalies and corresponding residual geostrophicvelocities on the images. Positive values are to the fight (east side) and negative values are to the left (west side) of ground tracks plotted over the images. In the following sectionthe most representativeexamplesof these imagesare shown. <u'u'> 3x +<u'v'>-b--ffx +<Uv>3y +<v'v'> 3y(7) wherecapitallettersrefer to meanflow. When this term is negativea transferof kinetic energyfrom the mean flow to the eddiesoccurs.For a pure zonal flow, only the third term of the equationis differentfrom zero. Consideringthat the Algerian currentis mainly a zonal flow, the third term gives the major contributionto the energytransfer.In the Algerian basin, <u'v'> is positive (Figure 5), and (assumingthe maximumvelocityat the coast)3U/3y is negative.It follows that the current tends to releasekinetic energy toward the eddy field by means of the main current leaving eddies. Moreover, consideringthat the typical velocity of the current is around50 cm s-• andthe currentis -50 km wide,we can estimatethe horizontal meridional eddy viscositycoefficient using 3U - <u'v'>= Ay3y andweobtain avalue ofAyof106_ 107 cm 2s-•. (8) 5.1. Algerian Basin In April the Algerianbasinimagesrevealedthe appearance of a very complexfeaturein the southeastern part of the basin. The image of April 19, 1993, shows an organizedstructure coveringthe area between6.5ø-8.5øEand 37.2ø-38.5øN (Plate 3). This area is delimited southby the African coastand east IUDICONE ET AL.' SEA LEVEL VARIABILITY •, 7 •' • IN THE MEDITERRANEAN SEA 3003 ' ß,.• - . '-{ ; .. -,.". ',.' o ':'."::"-:.. •)-•.'•......... ::• •: • ...' •: •" .... .' ß .. ..'•:.•;;.:...'._--'•ß .•-.. ........... • ..•..,-'..... •".,.:•. ..." •... ?'. .• ". ' .•-• .......•.-. :"• ....'.• '...-'....'"" .... .':'..'::' No, ß '.., :.:' .'L ,:'.::•:f ",• ,' .:.-'•.':' "'• 'J?• •. -•:-. ....... ..:•.:......... "::::.:•?•::'":'•?::•..:: .......... •?;...• ..• ' '::: '"-..'-:- ' • • •'"': '•. :'"' •' .. *'•:' ' • •3 • •.1.. : "'-:' "-'*'.'.•.%;..: .... •:•?:.:.::•'•' ..::".-"::•q . 1 ,-•- "".-.-?.'% .'.:-.•:'• ....• :'., .'. ..:.::..::: a'D , ' ' ':'"', ..... :.'..::•:•...-:'• .;.:.:.'• '• . P . •-, .'...:.----.:., ..'...' ;.•:.:.•' ....... .. ?-:' • :.: .• • • .".-' ....• ......:("•..• .... ' :• ........ "..'") .'."..••' , .... • .::•.;": ..... ...-:2: ':'. ", 'i:-:• '•' ..':,'•, :[' •• •-•'.*.:":•'::.-.:.:: •.:L:.-:: ::'•.:•.•: "•,. .•' .'..: "'?-': .... ',:".?*':L...:.':."-¾ ??.'..'• ••. ,-: ................ ';:":..-•";':k . ::.. ::,:,-:::.::.::.7:'.'.'.•..'.::.::• ,.•.....::..:.•.: ............... j .... • ....... ... .... . * . ... . ....: ............ •.,,,-:.-•:..-...... ...... . .' ' - ' .' '..:: .•'". -".:-:.::-'•...':•."::....:-..-'-:---:-:--•---.. • ß ':'. ' ...•--":'•-::--.."' , -.. :.-'-....... ......... '-' ...... :":.•:..:•.'..... '. • • ß : ..... ....... ........... .........•..:.... , .......:.•'--.•' .:.•:'__ .• ............. ......... •......... .:•3(:.... ....... ........ ....... ;..j" ...--:';:" • ...7 . .:......_'" '• ,. "•'•. '..•-'-•: -'.:...•:... '.5'......2..":'.:."':':::::::." ':j.... '...... . (. •:'--: • .... " :-'.•,•.. •, ......:.:: . It '71 , . •5 .. 3•. •. . . 4 _5 -,•ll • ß2 -- •,•a . ...... • ...... ....$:......•- • J 12 _•2 , . . .... •6 4,7 k, ..... $ • ''iS • 2 •. 9,'2 i•, . . --.... :. . "t. • Figure 6. Temporalevolutionof SLA alongtrack 146 (upperpanel),alongtrack094 (centerpanel),and alongtrack 161 (lowerpanel). bytherisingof bottom topography in theSardinian channel circulation for B. The system appears as two coupled (Figure 2). This structureseemsstrictlyconnectedwith the mushroomssharingthe middle eddy (B). North of these features,a hammer-likestructureis present. eastwardflowing Algerian current, constitutedby MAW SLA along track 146 (cycle 21, April 15) shows a incomingfrom the Straitof Gibraltar.It hasthe shapeof three of the C eddyanda highlyconvolutedspiraleddies,two of themzonallyaligned minimumof--12 cm in correspondence (A centeredat 38.0øN, 7.0øE and B centeredat 38.1øN, 7.9øE, maximumof 14 cm for the B eddy (Plate 3). The SLA also that the anticyclonic circulationis largerthen the respectively)and the third one (C centeredat 37.5øN, 8.1øE) suggest located southeast of the first two. The eddies have an almost spiralingfeaturerevealedby the SST field reachingthe at 38.7øN.Evenif uniform SST of 14.7øCwith somecolderpatchesof entrained northernboundaryof the warmstructure water. Their mean radius is about 30 km, even if A and, thealtimeter onlyprovides theanomaly relative to themean, SLA quiteresembles whatwe wouldexpectfor above all, C are elliptical. Their spiralingshapessuggesta the observed by the SST field. cyclonic circulation for A and C and an anticyclonic the dipolesuggested 3004 IUDICONE ET AL.: SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA 8O 60 40 .2o 0 Jun Sept Dec Mor' Jun 1992 Sep Dec Mor Jun 1993 Sep Dec 1994 Figure7, Wind datain the westMediterranean (ECMWFsquared windintensity, averaged on 1 month). Unitsarein m2/s2. 38. o January 38 I 24.8 38. o 28.8 Apr i i . 32.8 . ß 34 24.8 38.8 38. 38. M•y M•rch 1 0 34 1 38 28.8 32.8 24.8 28,8 32.8 June I 38.0 I 38 24.8 38.0. 30. February 38 3o 34. ß 28.8 July . ' I 32.8 ' 24.8 38.8 28.8 32.8 August 1 24.8 28.8 32.8 September I 38.8 o o 24.0 38 - 28.8 32.8 24.8 28.8 24.8 32.8 28.8 Decemben November Oc { ober 38.8 32.8 38.0 38 8 i 34.0 38ø 38 38.8 24.• 28.8 32.0 24 . ltl 28.0 32.8 ß 24.0 28.0 32.• Figure8. Tenyear(1983-1992)monthlyaveraged SST mapsof theRhodegyreareaandT/P groundtracks. IUDICONE ET AL.' SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA 3005 ß ,. 40N ' -' 5E '7E 9E Plate 3. Advancedvery high resolutionradiometer(AVHRR) image of April 19, 1993 (Channel4, NOAA 11). Superimposed are the 146 and 161 T/P groundtracks,the sealevel anomalies(SLA) (black), and vectors (white) for the estimatedgeostrophicvelocityorthogonalto the trackfor cycle 21. The maximum velocity is betweenB and C, suggestinga The time evolution of the SLA along track 146 is shownin westward jet-likeflowof -50 cm s-•. Thevelocity for the Figure 6 (upper panel). The picture presentsmany striking othersideof C is 40-50cms-l, whilefor thenorthern sideof features.First, the featureformerlydiscussedas anticycloneB B thevelocityis -30 cms-•. Thislastfeaturehasalsotwo in 1993 has a lifetime of severalmonths.After lasting in the plateausof uniform velocity,symmetricallydisplaced-20 km south and north of the center, like a ring, that alters the expectedfield. This is probablythe result of the entrainment of outer water, as seen in the image, not yet dynamically adjusted.The asymmetryin the velocityfield seemsto suggest that the smaller cyclone (C) is more energetic than the anticylonic structure (B); therefore C acceleratesB. The areafor a periodof 4 months,at the end of March it movesto thenorthwest ata quiteconstant phase speed of -3 kmday -• until it reaches39øN. The phasespeedobtainedfrom the T/P data is coherent either with the in situ observations [Millot, 1991] or with numerical estimatesof the phase speed for a baroclinicinstabilityof a coastalflow similarto the Algerian current (Mortier, 1992). Once the eddy reaches39øN, it relative vorticity estimated fromthedataare-0.810-5s'• forB almostsuddenlychangesits trajectoryto a more westward by the SLA of the adjacenttracks and 1.3 10-5s-• for C, aboutonetenthof the localCoriolis motionas also suggested parameter. SLA along track 161 give someadditionalinformationon the easternboundaryof the complex structuresvisible in the image. The southern part of the track clearly shows the eastwardflow of the coastalcurrentwith residualgeostrophic andby the analysisof AVHRR images.The lack of spatial resolutionof the groundtracksof the satellitecannotpermit us to follow it longerwith enoughreliability.CycloneC has insteaddisappeared from the T/P data after a few cycles (Figure6, upper panel), as expectedby the anticyclonic velocityof 10-15cm s-•. Immediately north,the westwardrotation flow is associatedto the local cyclonic tendency of the circulationsuggestedby the AVHRR image. of the whole structure. Sincethe beginningof 1994 we againobservefeatures with a behavior similar to that found for the structure 3006 IUDICONE ET AL.' SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA o_- ,/ .:,2H Plate 4. AVHRR imageof the Levantinebasinof June26, 1993 (Channel4, NOAA 11). Superimposedare the 094 and 185 T/P groundtracks,the SLA (blue),and vectors(white) for the estimatedgeostrophicvelocity orthogonalto the track for cycle 28. discussed above.They have the samepropagation speedand all have left the track around39øN. What is changedis the intensity,which is much higher in this secondperiod.This interannualvariability, already seen in the EKE and SLV maps(section3), is also presentin the wind data (seeFigure 7) that showan absolutemaximumjust beforethe beginning of the moreactiveperiodsuggesting a link betweenthe wind and the observationof energeticeddies at the southeastern comer of the basin. A possible mechanismcould be the secondarycirculationdue to the curl of the wind (mistral)that producesa circulationwith a main flow crossingthe basin from north-west to southeast[Herbaut, 1994]. This would constrain the eddies to move close, or closer than in absence fed for a longer time by the currentand, at last, forced to moveeastwardas far as topographypermits. 5.2. Eastern Basin The absolutemaximumin the mesoscale activityof the Mediterranean, as derived from the T/P data, is located southeast of Crete in the Levantine basin near the crossover point betweentracks094 and 185 (see section3). The time evolutionof the SLA alongtrack094 is shownin Figure6 (center panel). At the latitude between 34ø and 35øN a sequenceof positiveand negativeSLA is clearly visible, suggestingthe seasonalappearanceof a coherentstructure. of thiswind, to the coastin their initial mainlyeastwardpath, Indeed,in the 2 yearT/P dataseta negativeSLA characterizes IUDICONE ET AL.: SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA /////.///// //,//// cms 3007 Ocm / //// 17 19 21 23 Plate 5. AVHRR imageof theLevantinebasinof November26, 1994 (SST, NOAA 11). Superimposed are the 185 T/P groundtrack,the SLA (black),and vectors(black)for the estimatedgeostrophic velocity orthogonalto the trackfor cycle 80. the first half of each year and a positiveone characterizesthe second half. To understand the cause of a similar behavior for the SLA time evolution, we compared the altimetric data with the AVHRR climatology of the Levantine basin reported by Marullo et al. [1998]. In Figure 8 the SST monthly climatology (1983-1992) of the AVHRR-derived SST field are shown together with the altimetric ground tracks. In autumnand in winter a well-organizedRhodegyre structureis mainly localized in the centerof a meshof the T/P grid. The warmer Iera-Petra anticyclone is evident south of the crossover,and it is separatedfrom the Rhode gyre by a sharp front. In summera cold plumecomingfrom the AegeanSeais presentin the area crossedby tracks094 and 185, while the Iera-Petraanticyclonehasshiftedwestwardandis not crossed by the tracks. To further investigatethe relationshipbetween SLA and SST featuresin thesetwo scenarios,we give two examplesof 3008 IUDICONE ET AL.: SEA LEVEL SEFTEMBER 18 VARIABILITY 30 1993 2O ß"'"": .....:':": :" : IN THE MEDITERRANEAN 10 cm/s SEA 10 cm 22 24 ' ? i ....... " Plate 6. AVHRR imageof theTyrrhenianSeaof September 30, 1993(SST,NOAA 11). Superimposed are the 161and044 T/P groundtracks,theSLA (blue),andvectors(white)for theestimated geostrophic velocity orthogonalto the trackfor cycle38. the SLA and residual geostrophicvelocities for June and November superimposedon contemporaryAVHRR images. In Plate 4, SLA and residual geostrophicvelocitiesof tracks 185 and 94, cycles 28, are superimposedon the AVHRR image of June 26, 1993. In Plate 5, SLA and residual geostrophic velocities of tracks 185, cycle 80, are superimposedon the AVHRR imageof November26, 1994. The image of June 26, 1993, (Plate 4) indicates the presenceof a plumeof relativelycold watercomingfrom the AegeanSea east of Crete and elongatingfor 220 km in the southeast direction.At its southeastern limit, part of the cold water is spiraling, entrained by two anticyclonic twin structures of 100 km of diameter each. The negativeanomaliesobservedin boththeascending and descending tracksat about35øNcorrespond quitewell to the areaof relativelycold water. The valuesof the minima are of about10 cm, and the resultinggeostrophic velocitiesdescribe the cyclonic circulation associatedto the cold area with IUDICONE ET AL.: SEA LEVEL VARIABILITY maximum velocities of 20 cm s't. The same value is associatedto the west of the two anticyclonic features, crossedby track 094. The SST image of November 26, 1994 (Plate 5), reveals the presenceof both Iera-Petraand the Rhode gyre, southeast and east of Crete, respectively.The warm structure (IeraPetra) has an almost circular shapewith a diameterof -150 km and has a temperature of .•23øC. The cold structure (Rhode gyre) has an elliptical shape zonally oriented with major and minor axes of -220 km and -120 km, respectively.It hasa seasurfacetemperatureof-• 17øC. Track 185 crossesthe center of Iera-Petra, while the Rhode gyre is not crossedby any T/P track. The SLA alongtrack 185 exhibit a regular dome, with an amplitude of more than 25 IN THE MEDITERRANEAN SEA 3009 and anticyclonicgyre.Plate 6 illustratesa particularcaseof comparison. On September30, 1993, in the northernpart of the Tyrrhenian Sea an almostelliptic. al cold patchof waterof about 150 km of major axis and lessthan 100 km of minor axisis clearlyvisible.The featurehassomemeandersand is elongating alongthe continental shelf.At the southeast tip of Sardinia a filamentous hammer-like structure extends in the northeast direction for about 100 km. Along track 161 the SLA has a domingwith a maximum of 10 cm in goodcorrespondence with the warmwater,just southof the cold feature,and a negativetrendgoingtoward the northernend of the track. The SLA along track 44 better reproduces the relative minimumconnectedto the cold cm, exactly in correspondence of the warm structure.The structurebut, not having an optimal intersectionwith the residualgeostrophicvelocitiesdescribethe clockwiserotation anticyclone,showsa reducedmaximumconnectedto the of theanticyclone withmaximum values of -50 cms'•. The warmpatch.Thevelocityfieldalongtrack161showsa 20 cm velocityfield is asymmetricwith highervaluesand a stronger horizontalgradientin the northernflank near the Rhodegyre. This asimmetry could be due to the interactionof Iera-Petra andthe Rhodegyre itself. The whole of the comparisonsbetweenthe SST fields and -1 s westwardflow connectedto the southernpart of the warm areaand an almostequalreturnflow northof it, still in the warm area. In the cold area the estimated velocities are weaker than in the warmer ones. This asymmetrycan be explainedby two differenthypotheses. First, the cyclonic circulationis a quasi-steady characteristic of the northern 1995] that the observedSLA variability is essentiallydue to TyrrhenianSea [Astraldi and Gasparini, 1994], weakly circulationis the seasonal appearance of Iera-Petra. Nevertheless, the detectedby T/P SLA. Second,the anticyclonic availabledata set is not enoughto reject the hypothesisthat reinforcedby the generalcirculationfield of the southern of a zonaljet between 40øand the seasonalshift of the westernboundaryof the Rhodegyre basinthatimpliestheexistence is alsosupported can also contributeto the SLA variability. Structureslike the 41øN [Perilli et al., 1995]. This hypothesis cold patch observedin June 1993, and presentin the SST byin situdata[Moen,1984]thatshowin thesamemonthof a climatology (Figure 8), should also contributeto the SLA different year the presenceof a southernanticyclonic circulationstrongerthanthe northerncyclonicone [seealso variability. Astraldi and Gasparini, 1994]. A similar asymmetric T/P measurements seems to confirm [see Larnicol et al., 5.3. Tyrrhenian Case circulationfield hasbeenalreadyobservedby Trasvinaet al. [1995] in the Gulf of Tehuantepec,where a symmetric The northernpart of the TyrrhenianSea is characterized by the presenceof a well-known cyclonicstructure[e.g., Marullo et al., 1994] linked to the funneling of the wind in the Bonifaciomouths.The historicalanalysisof AVHRR images showedthat in autumnthis gyrehasa seasonalintensification, causedby the setupof the mistralwind and the .strongupper offshorewind producedmainly a large anticyclonicwarm :ore eddywith only a weakcycloniccounterpart. Interpolatingthe EuropeanCentre for Medium-Range layerstratification. The commonideais thatit is coupledwith of the wind intensityis similarto that of the westernbasin (Figure7), themistralbeingthemainforcingfor bothbasins andthewinddatashowinga clearinterannual variability,with an anticyclonicstructureimmediatelysouthof it, againcaused by the curl of the northwesterlywind [Moen, 1984]. The propernatureof the structuremakesit almostinvisiblein the AVHRR images,and the observations comefrom few in situ measurements[Artale et al., 1994]. T/P tracks over the TyrrhenianSea are well locatedto intersectthe anticyclonic gyreandthencanbe usedto studyits time evolution. The time series of the SLA along track 161 is shown in Figure 6 (lower panel). At the beginning of the T/P measurements a quite strongcoupleof positiveand negative anomaliesoccupiedthe areabetween39.5ø and41øN. Then a positiveanomalyappearsin the datain thesecondhalf of July 1993 at 40.5øN and disappearsat the end of the year. This anomalyseemsconnectedby a weak positivesignal to the next positive anomaly, startingagain in July 1994 but at Weather Forecast(ECMWF) wind data in the lee of the islands,in front of the BonifacioStrait, we obtain a raw estimateof the windflowingon thebasin.The time evolution an absolute maximum in winter 1994. This maximum corresponds well to thepositivemaximum between 40.5ø and 39.5øN observedin the secondyear for track 161 SLA time series(Figure6, lowerpanel).Duringsummer, reduced values of SLA are relatedto the reduceddynamicalactivityobserved in the TyrrhenianSea by severalauthors[Astraldiand Gasparini,1994'Marulloet al., 1994]as a consequence of the isolationof the basinduringthatperiod. 6. Conclusions Two yearsof T/P altimeterdata were analyzed to studythe sealevel variabilityof the MediterraneanSea.The analysisof of samepositionof our first cycle positivemaximum.More at the SLV andEKE mapsclearlyrevealsthe mainstructures north (-41øN) in the periodJune-September 1994, thereis a the Mediterraneancirculation,indicatingthat T/P altimeter is able to catch a large part of the Mediterraneanmesoscale traceof a weaknegative-positive anomaly. The comparisonbetweenAVHRR imagesand altimeter circulation.The analysispointsout two areasof strongsea data shows a good correspondencebetween negative level signal in the easternpart of the Algerian basin and anomaliesand cyclonicgyre and betweenpositiveanomalies southeastof Crete. SeasonalSLV and EKE mapsstressedthe 39.5øN. Then it moves to the north until it reaches about the 3010 !UDICONE ET AL.: SEA LEVEL VARIABILITY IN THE MEDITERRANEAN SEA seasonal variation of the Mediterranean circulation and also suggested a differentbehavior in eachsinglesubbasin. The comparisonbetween1993 and 1994 datahas clearly shownan evident year to year changein the SLV and EKE values, 1994 values being definitively larger than the 1993 ones.This interannualvariationcan be explainedin termsof interannual variability of the atmosphericforcing. The analysisof ECMWF winds confirmedthis hypothesis.From the meteorologicalpoint of view, 1993 is consideredan anomalousyearsincewinterwascharacterized by high values of atmospheric pressure, especially in the western Mediterranean,where in Januarythe pressurefield was 12 mbar higher than the mean of the last 30 years (M. Conte, private communication,1996). information onthetimeevolution of thiscomplex system is of greatinterestfor understanding the role of the interactionof theIera-Petra anticyclone andtheRhodegyrein theformation andespeciallyin the spreadingof the Levantineintermediate water. Acknowledgments. The presentworkwasfundedby Agenzia SpazialeIraliana (ASI). The image processing softwarewas developedby O. Brown, R. Evans, J. Brown, and A. Li at the Universityof Miami with Office of Naval Research funding. Continuing support of theMiamigroupis gratefully acknowledged. We thankE. B6hmfor theconstructive discussion andfor providing theAVHRR imageof November26, 1994. References of SatelliteOceanographic The crossoveranalysishighlightedthe stronganisotropic Archiving,ValidationandInterpretation data (AVISO), AVISO User Handbook: Merged behaviorof the mesoscale variability.The principalaxisof TOPEX/POSEIDON Products, 2nd ed., Cent. Nat. D'Etud. varianceis alignedwith the bottomtopographymainly along Spatiales,Toulouse,France, 1992. the basin boundary.In the case of the Algerian basin, in Artale,V., M. Astraldi,G. Buffoni,andG. P. Gasparini,Seasonal which the main current is mainly zonally oriented, this variabilityof gyre-scalecirculationin the northernTyrrhenian Sea,J. Geophys.Res.,99(C7), 14127-14137,1994. analysissuggeststhat the eddy momentumflux is from the of the main currentto the eddyfield andpermitsan estimationof the Astraldi,M., andG. P. Gasparini,The seasonalcharacteristics circulation in the Tyrrhenian Sea, in Seasonaland interannual horizontaleddy coefficient. variability qf the western Mediterranean Sea, Coastal and A specific analysisof the time evolutionof the SLA was EstuarineStud.,vol. 46, editedby P.E. LaViolette,pp. 115-134, donefor someparticularly interesting areaswith strongand/or AGU, Washington,D.C., 1994 coherentSLV signals.Theseareaswere investigatedlooking B6hm, E., S. Marullo, G. M. R. Manzella and R. Santoleri;Studio della variabilith della topografia superficiale nel Mar at the time seriesof along track sealevel anomalies,AVHRR Mediterraneo Occidentale mediante l'uso di dati Geosat, data were usedto betterunderstandwhich particulardynamic Proceedingsof the Workshopon Calibrazionee validazionedei structureis responsiblefor the observedaltimetersignal.This modellinumericicon i risultatisperimentali in Oceanogra. fia Fisica,editedby IstitutoUniversitario Navaledi Napoli,pp. 133combinedanalysisalso allowed the integrationof the two141, 1992. dimensionalhorizontalinformationprovidedby AVHRR with Bormans, M., C. Garrett,andK. R. Thompson, Seasonal variability the one-dimensionalvertical view offered by the altimeter. of the surfaceinflow throughthe Strait of Gibraltar, Oceanol. This work successfully demonstrated the capabilityof T/P to Acta, 9, 403-414, 1986. F. P., R. E. Davis,andC. B. Frandry,A technique for correctly detect dynamical information over the Bretherton, objectiveanalysisand designof oceanographic experiments Mediterranean Sea, even if the signal to noise ratio is appliedto MODE-73, Deep SeaRes.,23, 559-582, 1976. relativelylow. Herbaut,C., Etude de la circulationoceaniqueen Mediterran6e Moreover,this comparisongives us the opportunityto occidentale,Ph.D. thesis,Univ. ParisVI, Paris, 1994. understandthat the intensificationof the SLV in the Algerian Kinder, T. H., and H. L. Bryden,The 1985-1986Gibraltar experiment: Datacollection andpreliminary results, EosTrans. basin in winter is due to the occurrenceof a complex AGU, 68 (40), 786-787, 793-794, 1987. structureof interactingmesoscaleeddies appearingduring Larnicol, G.,P.-Y.Le Traon,N. Ayoub, andP, DeMey,Meansea spring of both years.The time evolutionof the SLA along levelandsurface circulation variability of theMediterranean Sea track 146 (Figure 6 upper panel) indicatesthat the structure from2 years of TOPEX/POSEIDON altimetry, J. Geophys. Res., has a lifetime of several months and tends to moves to the northwest ata quiteconstant phase speed of-3 kmday'• until it reaches39øN. The phasespeedobtainedfrom the T/P data is coherent either with the in situ observations or numerical models. ! 00(C 12), 25163-25177, 1995. La Violette,P. E., J. Tintor6,andJ. Font,The surfacecirculation of theBalearic Sea,J. Geophys. Res.,95(C2),1559-1557, 1990. Marullo, S., R. Santoleri,and F. Bignami , The surface characteristics of the TyrrhenianSea: Historicalsatellitedata analysis, in Seasonal andInterannual Variability of theWestern Mediterranean Sea,CoastalandEstuarine Stud.,vol.46, edited T/P measurementsover the Tyrrhenian Sea permitted for byP.E.LaViolette, pp.135-154, AGU,Washington, D.C.,1994 the first time the remote detection of the presenceof the Marullo, S., R. Santoleri, P. Malanotte Rizzoli, and A. Bergamasco, anticyclonicgyre southof Bonifacio togetherwith its cyclonic The sea surfacetemperature field in the easternMediterranean northern counterpart often observed by AVHRR images. fromAVHRRdata,I, Seasonal variability, J. Mar.Syst.,inpress 1998. Moreover, the analysisof the SLA time seriesshowsa clear P. Malanotte Rizzoli,andA. Bergamasco, interannualvariability of positionof the anticyclonethat in Marullo,S.,R. Santoleri, The sea surfacetemperaturefield in the easternMediterranean turn induces the enhancedvariability observedin the SLV maps. In the Levantine basin the comparisonbetween SLA time series, satellite climatology, and single AVHRR images allows us to understand that the strong altimetric signal observedsoutheastof Crete is mainly due to the fall-winter appearanceof the Iera-PetraanticyclonealongT/P tracks185 and 094. The Rhode gyre is generallysitedat the centerof the ground track mesh. The strong frontal area separatingIeraPetra and the Rhode gyre can occasionallycontributeto the sea level variability.The possibilityof usingT/P data to get fromAVHRRdata,II, Interannual variability, J. Mar. Syst.,in press,1998. Meteorological Office, Weatherin the Mediterranean, in General Meteorology, vol. I, 362 pp., Her Majesty'sStn.Off., Norwich England, 1962. Millot, C., Circulation in the westernMediterranean Sea,Oceanol. Acta, 10, 143-150, 1987. Millot, C., Mesoscale andseasonal variability of thecirculation in the westernMediterranean, Dyn. Atmos.Oceans,15, 179-214, 1991. MoenJ.,Variability andmixingof thesurface layerin theTyrrhenian sea:MILEX-80finalreport,SACLANTEN Rep.SR-75,128pp., SACLANTEN,La Spezia,1984. IUDICONE ET AL.: SEA LEVEL VARIABILITY Morrow, R., R. Coleman, J. Church, and D. Chelton, Surface eddy momentum flux and velocity variancesin the SouthernOcean from Geosataltimetry,J. Phys. Oceanogr.,24(10), 2050-2071, 1994. Mortier, L., Les instabilitesdu courantalgerien,Ph.D. thesis,Univ. Aix-Marseille II, Marseille, France, 1992. Nerem, R. S., E. J. Scharma,C. J. Koblinsky, and B. D. Beckley, A preliminary evaluation of ocean topography from TOPEX/POSEIDON mission,J. Geophys.Res., 99(C12), 2456524583, 1994. Perilli, A., V. Rupolo, and E. Salusti, Satellite investigationsof a cyclonic gyre in the central Tyrrhenian Sea (western MediterraneanSea),J. Geophys.Res.,100(C2), 2487-2499, 1995. Philippe, M., and L. Harang, Surface temperaturefronts in the IN THE MEDITERRANEAN Tapley, B. D., et al., SEA Precision orbit 3011 determination for TOPEX/POSEIDON, J. Geophys.Res., 99(C12), 24383-24404, 1994. Theocharis,A., D. Georgopoulos,A. Lascaratos,and K. Nittis, Water masses and circulation in the central region of the eastern MediterraneanIonian, South Aegean and Northwest Levantine, 1986-1987, Deep Sea Res.,40(6), Part I, 1121-1142, 1993. Trasvina, A., E. D. Barton, J. Brown, H. S. Velez, P.M. Kosro, and R. L. Smith,Offshorewind forcingin the Gulf of Tehuantepec, Mexico:The asymmetric circulation, J. Geophys. Res.,100(ClO), 20649-20663, 1995. Wilkin, J., and R. A. Morrow, Eddy kinetic energyand momentum flux in the SouthernOcean: Comparisonof a global eddy- resolving model with altimeter, drifter and current-meterdata, J. Geophys.Res., 99(C4), 7903-7916, 1994. Mediterranean Sea from infrared satellite imagery, in Hydrodynamicsqf Semi-EnclosedSeas,editedby J. C. J. Nihoul, WMCE Consortium of Investigators, Western Mediterranean pp. 91-128, Elsevier,New York, 1982. CirculationExperiment:A preliminaryreview of results,Eos, Trans.AGU, 70(31), 746, 756-758, 1989. POEM Group, General circulation of the eastern Mediterranean, Earth Sci. Rev., 32, 285-309, 1992. Wunsch,C., andE. M. Goposchkin, On usingsatellitealtimetryto Robinson, A. R., and M. Golnaraghi, Physical and dynamical determine thegeneralcirculation of theoceanwith application to geoidimprovements, Rev.Geophys., 18, 725-745,1980. oceanographyof the MediterraneanSea, in Ocean Processesin Climate Dynamics: Global and MediterraneanExamples,edited P. Gerosaand S. Marullo, Centro Ricerche Casaccia,ENEA, via by P. Malanotte-Rizzoliand A.R. Robinson,NATO ASI Ser., 4, 19, 255-306, 1994. Anguillarese301, 00060 S. Maria di Galeria(Rome), Italy. (e-mail: Santoleri, R., E. BOhm, and M. E. Schiano, The sea surface salva•gorgona.casaccia.ena.it) D. Iudicone and R. Santoleri, Istituto di Fisica dell'Atmosfera, temperatureof western Mediterranean Sea: Historical satellite thermal data in Seasonaland Interannual Variability o.f the CNR, Via del Fossodel Cavalieres.n.c.,00133 Rome,Italy. Western Mediterranean Sea, Coastal and Estuarine Stud., vol. 46, editedby P. E. LaViolette,pp. 155-176, AGU, WashingtonD. C., 1994. (ReceivedJuly 6, 1996;revisedMay 16, 1997; acceptedMay 29, 1997.)