generation mechanisms for mesoscale eddies in the gulf of...

Ocean Dynamics (2011) 611587ndash1609DOI 101007s10236-011-0482-8

Generation mechanisms for mesoscale eddies in the Gulfof Lions radar observation and modeling

Amandine Schaeffer middot Anne Molcard middotPhilippe Forget middot Philippe Fraunieacute middot Pierre Garreau

Received 17 December 2010 Accepted 22 July 2011 Published online 8 September 2011copy Springer-Verlag 2011

Abstract Coastal mesoscale eddies were evidencedduring a high-frequency radar campaign in the Gulf ofLions (GoL) northwestern Mediterranean Sea fromJune 2005 to January 2007 These anticyclonic eddiesare characterized by repeated and intermittent occur-rences as well as variable lifetime This paper aims atstudying the link between these new surface observa-tions with similar structures suggested at depth by tradi-tional acoustic Doppler current profiler measurementsand investigates the eddy generation and driving mech-anisms by means of an academic numerical study Theinfluence of the wind forcing on the GoL circulationand the eddy generation is analyzed using a numberof idealized configurations in order to investigate theinteraction with river discharge buoyancy and bathy-metric effects The wind forcing is shown to be cru-cial for two different generation mechanisms A strongnortherly offshore wind (Mistral) generates a vortexcolumn due to the bathymetric constraint of a geo-strophic barotropic current which can surface after the

Responsible Editor Aida Alvera-Azcaacuterate

This article is part of the Topical Collection on Multiparametricobservation and analysis of the Sea

A Schaeffer (B) middot A Molcard middot P Forget middot P FraunieacuteLSEET Universite du Sud Toulon-Var BP 2013283957 La Garde Francee-mail amandineschaefferlseetuniv-tlnfr

A SchaefferIFREMER LER PAC BP 330 83500La Seyne sur Mer France

P GarreauIFREMER DYNECOPHYSED BP 7029280 Plouzane France

wind relaxes a southerly onshore wind generates afreshwater bulge from the Rhocircne river discharge whichdetaches from the coast and forms a well-defined sur-face anticyclonic eddy based on buoyancy gradientsThese structures are expected to have important conse-quences in terms of dispersion or retention of biogeo-chemical material at local scales

Keywords Mesoscale eddy middot HF radar middot Modeling middotGulf of Lions

1 Introduction

Despite recent progress in ocean modeling and devel-opment of observation platforms the study of meso-scale dynamics (20ndash100 km) is still a challenge Theinvestigation of eddy structures is crucial in coastalareas because of their physical and biogeochemicalimpact on ecosystems as they contribute to water trans-port vertical mixing and possibly trapping of biolog-ical materials Eddies can be observed by differentmeans such as Lagrangian floats (Griffa et al 2008)or scientific satellites providing the monitoring of seasurface temperature anomalies of sea surface heightor biological properties (Lavrova and Bocharova 2006Zamudio et al 2008 Henson and Thomas 2008) High-frequency (HF) radars appear to be a powerful toolby providing high spatial and temporal resolution dataThe HF technique is nowadays quite widely used by theoceanographic community to measure ocean surfacecurrents in coastal areas They constitute a tremendouspotential for a number of applications like transportstudies (Kaplan et al 2005 Shadden et al 2009) iden-tification of optimal release locations (Coulliette et al

1588 Ocean Dynamics (2011) 611587ndash1609

2007 Molcard et al 2009) support for safe navigationand search and rescue operations (Gurgel et al 2002Ullman et al 2006) as well as description of coastal cir-culation (Shay et al 1998 2003 Paduan and Rosenfeld1996) The assimilation of surface current data (Paduanand Shulman 2004 Barth et al 2008) into numericalocean model may also considerably improve the fore-casts Quantitative information on eddy characteristicscan be captured like their diameter elongation ve-locities vorticity lifetime or trajectories (Bassin et al2005 Parks et al 2009)

A few physical processes are known that can gener-ate such mesoscale structures McWilliams (1985) andRobinson (1983) detailed some of them for mesoscalecoherent vortices barotropic or baroclinic instabilitiesof the current leading to eddy formation with verticaland horizontal dimensions comparable to the currentwidth and the characteristic Rossby radius topographicboundary effects (frictional drag when flowing aroundtopography) or topography variation under a flow(stretching or compression of the water mass creatingvorticity according to the conservation of potential vor-ticity) mixing and adjustments in an unstable frontalregion and the β-effect based on the latitudinal vari-ation of the Coriolis parameter Furthermore Orlicet al (1994) and Estournel et al (2003) described theinfluence of the wind curl on eddy formation

The Gulf of Lions (GoL) is a particularly relevantarea to study mesoscale structures as it is a continen-tal shelf characterized by a microtidal regime suitablefor the generation and evidence of eddy structuresIndeed this quite shallow gulf (90 m mean depth) is

bounded by a complex coastline and subject to majorforcing mechanisms (Fig 1) The GoL circulation isstrongly influenced by the along-shelf northern current(NC) or Liguro-Provenccedilal current which is the north-ern branch of the general surface cyclonic circulationin the northwestern Mediterranean Sea (Astraldi andGasparini 1992) The NC is known as a density currentsubject to geostrophic equilibrium and often exhibitsa noticeable thermal signature in winter as it carrieswarmer water from the Ligurian basin Its high seasonalvariability was evidenced by Millot (1990) Conan andMillot (1995) and Alberola et al (1995) From spring toautumn the NC flows relatively far from the coast andremains shallow (150 m deep) and wide (40ndash60 km)while in winter it gets deeper (more than 200 m) andnarrower (30 km) appears to be closer to the coast(Petrenko 2003) and to be more intense with a trans-port ranging from 15 to 2 Sv

Far from being a stable current the NC exhibitsan important mesoscale activity (Alberola et al 1995Sammari et al 1995 Rubio et al 2009b) Meanders aregenerated from baroclinic instabilities related to pertur-bations of the stratification (with periods of 75 daysand wavelengths of 60 km according to Flexas et al(2002)) or from barotropic instabilities due to bot-tom topography gradients with shorter periods (around35 days according to Flexas et al (2005)) Althoughthe NC generally adjusts itself above the 1000ndash2000-misobaths (Flexas et al 2005) it sometimes intrudes intothe shelf on the eastern part of the GoL under specificcondition of wind and stratification (Millot and Wald1980 Echevin et al 2002 Petrenko 2003) In contrast

Fig 1 Schematic circulationin the Gulf of Lions includingthe Northern Current themain freshwater output(Rhocircne river) predominantwind regimes and bathymetry(isobaths 50 100 200 5001000 2000 m) Radarstations are located at points1 and 2 with the coveragerepresented by dashed lines

Ocean Dynamics (2011) 611587ndash1609 1589

intrusions in the central and western parts of the GoLare less frequent (Estournel et al 2003 Petrenko et al2008)

The shelf dynamics is mostly driven by intense at-mospheric forcings hence the barotropic circulationof the GoL is strongly correlated to wind directionand curl (Estournel et al 2003) Strong cold and drycontinental winds blow from the North (Mistral) chan-neled by the Rhocircne valley and from the northwest(Tramontane) inducing localized upwellings along thenortheastern coast (Millot 1979 Hua and Thomasset1983 Millot 1990 Andre et al 2005) and dense waterformation in winter on the southern part of the GoL(Ulses et al 2008 Dufau-Julliand et al 2004) Thepredominant onshore wind regimes are southerly andeasterly winds They both notably affect the currentintrusion mechanism into the shelf

Buoyancy forcings mainly originate from the Rhocircneriver outputs bringing the largest amount of freshwaterinto the GoL with a mean discharge of 1700 m3s Thecorresponding river plume position and the inducedcirculation have been investigated via observations andmodeling by Broche et al (1998) Marsaleix et al(1998) Estournel et al (2001) Reffray et al (2004)Ulses et al (2005) and Gatti et al (2006)

In this region mesoscale activity is often correlatedto the NC meandering (Echevin et al 2003) Howeverseveral persistent mesoscale eddies linked to othermechanisms have recently been observed and doc-umented Through acoustic Doppler current profiler(ADCP) measurements Estournel et al (2003) evi-denced a large anticyclonic eddy located at the centerpart of the GoL under the mixed layer Modeling inves-tigation concluded to a generation mechanism relatedto the local continental wind curl More recently Huet al (2009) computed a summer mesoscale anticy-clonic eddy on the western part of the GoL compa-rable to some structures observed on SeaWifs imagesshowing an average area around 1200 km2 and a life-time of 60 days However the generation processeshave not been yet investigated Using remote sensingand in situ measurements Rubio et al (2005 2009a)described a well-developed anticyclonic eddy over theCatalan continental shelf characterized by horizontaland vertical length scales of 45 km and 100 m surfacevelocities of 05 ms and a lifetime of at least 2 weeksWith numerical modeling the generation of this eddywas then related to a flow separation off Creus Cape(southwest of the GoL) during intense northwesterlywind events

A 17-month coastal radar campaign has been per-formed in the eastern part of the GoL from June 2005to January 2007 as reported by Allou et al (2010) From

this large and unique data set of surface currents atvery high temporal (1 h) and spatial (5 km) resolutiona particular mesoscale eddy has been observed Thiseddy was characterized by repeated and intermittentoccurrence as well as a variable lifetime up to 25 days

Using four current meters data made available fromDecember 2005 to June 2006 Allou et al (2010) de-veloped a vortex detection algorithm and were able tocorrelate some eddy events to HF surface current mea-surements The generation of the eddy by a sheddingmechanism was suggested For specific wind episodesa current instability off Sicieacute Cape would be followedby the detachment of a vortex and by its westward driftHowever such an explanation was not confirmed by anumber of previous modeling experiments on the siteThe dynamics of the NC after its detachment down-stream the Sicieacute Cape and its offshore distance from thecoast definitely need more investigation but vorticeswere observed from computations at the same locationwithout any vortex shedding in the area Moreoversuch vortices were rarely observed in high-resolutionoperational runs contemporary to experiments and es-pecially in the surface layer

The goal of this paper is to analyze the GoL regionaldynamics by means of process-oriented numerical stud-ies in order to investigate the generation and drivingmechanisms of coastal eddies as observed several timesby HF radar current measurements in the eastern partof the GoL While the study by Allou et al (2010)was strictly observational here the modeling strat-egy is adopted to highlight possible local generationmechanisms

The radar setup is described in Section 2 and theresulting observations are presented in Section 3 ev-idencing the variable occurrence and lifetime of thestructure The wind features expected to play a crucialrole in the eddy generation measured from a buoyplatform and simulated by a meteorological forecastingmodel and are also analyzed and linked to the eddyevents in Section 3 Part 4 is devoted to the process-oriented study with the description of the idealizednumerical configurations and of two different wind-driven generation mechanisms of the eddy The resultsare then summarized in part 5

2 Setup of the HF radar system

Surface circulation dynamics have been observed in theeastern part of the GoL by two HF radars that were op-erated from June 2005 to January 2007 The experimentwas designed to monitor shelfopen sea water exchangeat the eastern entrance of the GoL


To measure surface currents at least two radar sta-tions are required each of them measuring the ra-dial components of the current field within a circularsector centered at the radar location Radial currentcomponents are estimated from the difference betweenthe Doppler frequency of the received signal and theDoppler frequency that is expected for the gravitywaves (λB) which are responsible for the backscatteredsignal at first order (Bragg mechanism) Vector currentmaps are derived from the combination of the radialvelocities estimated by the radars HF radars are ableto estimate the ocean surface currents up to hundredkilometers offshore resulting in a unique mappingof surface currents at very high resolution in spaceand time

A dual station of high-frequency Wellen Radars(WERA) (Essen et al 2000 Gurgel et al 1999) wasdeployed in the eastern part of the GoL the first one onthe Frioul island (site 1) and the other one at Salins deGiraud (site 2) 40 km apart (Fig 1) WERA is a radartransmitting frequency modulated continuous chirpsThe radars operated on a 1-MHz wide bandwidth cov-ering the radio spectrum around 162 MHz (wavelengthof λ = 185 m giving λB = 925 m) At these transmittedfrequencies the velocities measured by the radar repre-sent the current at an effective depth of approximately74 cm (Stewart and Joy 1974) The receiving arraysconsisted of 4 and 12 linearly spaced antennas at sites1 and 2 respectively The limited number of antennasespecially on site 1 due to local constrain resultedin a low performance in azimuthal resolution whenusing the standard beam-forming method for radar sig-nal processing The MUSIC algorithm (Schmidt 1986)which is routinely used by CODAR radars (Lipa et al2006) was preferred providing an azimuthal resolutionof 5 This angular resolution corresponds to 17 km at

range 20 km and 52 km at range 60 km The nominalrange resolution which is performed by a Fourier trans-formation of the chirps was 3 km However a smooth-ing average was performed along the range directionin order to increase the signal-to-noise ratio (SN) Thetemporal acquisition rate was 30 min A better SN ratiowas also obtained by performing a smoothing averagein time leading to an effective temporal resolution of1 h estimated to be sufficient to capture the surfacecurrent variability in the GoL

The radial current component maps obtained by thetwo radars were merged to calculate the current vectorsat the nodes of a predefined grid The mapping gridwas chosen rectangular orientated along westndasheast (x)and southndashnorth (y) directions Grid points are equallyspaced by 5 km along these directions

The precision of the radial speed depends primarilyon the limited integration duration The resulting valueis δVr = 21 cms which leads to an error of 3ndash35 cmsin both meridional and zonal current components overthe eddy region due to the geometrical dilution ofprecision

Figure 2 depicts the current vector coverage iethe ratio between the number of vector grid pointswhere current estimates are available and the totalnumber of processed grid points Some discontinuitiescan be noticed which correspond to external factorsas technical problems radiofrequency interferences orinterruption of the experiment during summertime forsafety reasons A loss of radar coverage was also no-ticed during strong winds especially during the Mistralwind regime This is consistent with the fact that prop-agation losses in HF ground wave propagation modeincrease with sea state leading to a decrease of theradar range as shown both theoretically (Barrick 1971)and experimentally (Forget et al 1982)

Fig 2 Time evolution fromJune 2005 to March 2007 ofmapping coveragecorresponding to the ratiobetween the number of vectorgrid points where currentestimates are available andthe total number of processedgrid points Time occurrencesof anticyclonic eddiesobserved by radars aresuperimposed in red with linelength proportional tolifetime in days


3 Eddy events observations and wind conditions

31 Eddy characteristics

As expected radar observations display the sea surfacemotion and mainly reveal in this micro-tidal environ-ment the surface drift due to the wind stress Specificharmonic spectral analysis also confirms the presenceof both inertial and diurnal oscillations as an oceanicresponse to wind gusts or breezes (Forget et al 2008)During periods of calm weather the general surfacecirculation is captured by the radars A northern branchof the NC flowing along the slope and sometimes in-truding the inner shelf is regularly observed The plumeof the Rhocircne river is often detected in the northern partof the radar coverage

In the present study we focus on a specific mesoscaleeddy structure which has been identified on radar cur-rent maps during the first year of the campaign (Allouet al 2010) The surface observation of such eddy wasconfirmed by the analysis of deep currents measured byfour current meter moorings on the shelf break duringwinter 2005ndash2006 and spring 2006 The eddy identifica-tion was done by visual inspection of HF radar surfacecurrents and vorticity maps This method was preferredto mathematical methods (Doglioli et al 2007 Nencioliet al 2010) since the radar observations were oftentoo noisy for an automatic multi-criteria detection andsome verification was necessary

From June 2005 to January 2007 27 anticycloniceddies have been recorded as those shown in Fig 3During the same period only two cyclonic eddies wereobserved Eddy occurrences are indicated by red verti-cal bars added to the radar coverage (Fig 2) the lengthis proportional to the eddy lifetime which varies from4 h to almost 25 days Gaps in the HF radar data com-plicate the exact estimation of the eddy lifetime forwhich often only a lower bound can be given Thusseveral eddy events could correspond to a unique struc-ture sporadically observed by radars at the surfaceConsidering the duration of the radar experiment(17 months of intermittent coverage) and the lifetime ofthe eddies these latter constitute a significative featureof the circulation in the eastern part of the GoL

The eddies were observed at irregular dates mainlyin June and in winter Among all the observations fiveeddy events (Fig 3) in December 2006 were particu-larly spectacular in terms of intensity and persistenceMoreover the good quality of radar current maps al-lowed a detailed description of the eddies characteris-tics and behavior (Fig 3) We therefore selected theseevents for the purpose of the present paper

These 5 December eddies diameters range from 20to 40 km Their current velocity values reach 04 msand are stronger over the shelf break in the south-ern part of the structure probably reinforced by thenorthern branch of the NC The Rhocircne plume appearsclearly on the second event (December 17th) and gen-

Fig 3 Snapshots of HF radarsurface currents 130-misobath is drawn


erates a convergent front It should be noticed that thecomplete eddy is not entirely detected by the radarsespecially near the coast Nevertheless the center ofthe structures remains obvious and can be numericallydetermined as the position of the velocity field mini-mum In December 2006 as for the whole data set eddycenters were located on the shelf or at the top of theslope near the 130-m isobath

When a continuous observation is available a west-ward drift of the structure is observed The eddiesfollow the shelf break direction with a drift velocityof the order of 025 ms The different eddy centertracks for December 2006 events are plotted in Fig 4Most of the eddies seem to appear within the radarcoverage As an example the time evolution of thefirst observed eddy is shown in Fig 5 at 2-h frequencyfrom its formation till the end The anticyclonic eddy isformed between the coastline and the 130-m isobath onthe 3rd of December remains distinguishable until the5th of December and eventually gets concealed by astrong northwestward flow A careful examination ofthe whole data set (200 snapshots) suggests that theeddies have not come from the east into the coveragearea Therefore a local generation is assumed

Past in situ data collected in the same area evidencedthe deep extension of similar anticyclonic structurePetrenko (2003) concluded to the presence of similar

Fig 4 Observed eddy westward trajectory in December 2006Each event corresponds to a different color specified in thelegend by the date of the first observation of the eddy togetherwith the lifetime in hours Circles correspond to the eddy centereach hour 130-m isobath is drawn

anticyclonic circulation from ADCP measurements inJune 1998 (Moogli 2 cruise) This observation wasconfirmed by SeaWiFs maps of surface chlorophyll-aconcentration Similarly a quick examination of thehull mounted ADCP database of the French NOTeacutethys II (httpsaveddtinsucnrsfr) confirms the ex-istence of anticyclonic structures in the water column atdifferent dates (eg 24 April 2000 1st November 2003)Gatti et al (2006) presented ADCP thermosalino-graph and remote sensing data collected in December2003 revealing the presence of an unusual eastwardbarotropic jet associated with the Rhocircne River Amongthe three suggested processes one corresponds to amovement of these freshwaters driven by an anticy-clonic eddy

32 Wind analysis

The surface layer dynamics are strongly correlated toatmospheric forcings which are remarkably intense inthe Gulf of Lions In order to investigate the effectof the wind forcing on the eddy generation we per-formed an analysis of wind conditions during the ob-servation period coincident to the strong eddy events inDecember 2006

This analysis uses wind estimated by the mesoscalemodel MM5 (Grell et al 1997) from the National Cen-ter for Atmospheric Research embedded in the Na-tional Centers for Environmental Prediction weatherforecast model The time and spatial resolutions arerespectively 3 h and 9 km (successively interpolated toa 3-km grid) To validate the model wind speed anddirection at 10 m height were compared to in situ windmeasurements provided by the Meteo France buoylocated at 47 E 421 N (position shown in Fig 1)The buoy sensors have been measuring hourly wind at-mospheric pressure temperature humidity and wavessince 2001 and are installed at 36 m above sea levelA logarithmic wind profile law is applied to the windintensity in order to get a comparable wind data set at10 m height

The wind rose at the buoy platform over December2006 is plotted in Fig 6a and shows the predominanceof the northerly winds in terms of occurrence (20ndash25) and magnitude (up to 18 ms) Figure 6b showsthe stick diagrams of the wind in December 2006 (ieusing the oceanographic convention indicating the di-rection of the velocity vector) obtained from the buoyplatform (top panel) and from the model at the buoylocation (middle panel) To get some insight into theeffect of the local wind on the eddy structure an ad-ditional stick diagram from MM5 at the eddy location


Fig 5 Sequence of surface currents measured by HF radars corresponding to the first eddy event from the 3rd to the 5th of December2006 The time incrementation for each figure is 2 h 130-m isobath is drawn


a

b

Fig 6 a December 2006 wind rose at buoy station the locationand extension of the fan sector indicate the wind origin and itspercentage and the colors refer to the wind intensity (ms) bwind sticks in December 2006 at comparable time resolution

(3 h) offshore buoy data (upper) MM5 model at buoy location(middle) and MM5 model at eddy location (lower) Gray boxescorrespond to radar observation of eddies


a

c

b

Fig 7 EOF analysis from MM5 wind data in December 2006 a Mean (meters per second) b EOF 1 (50 of the variance) c temporalcoefficient for EOF 1 (meters per second)

is plotted (bottom panel) The distance between thebuoy platform and the eddy core as determined on theradar current maps is about 100 km which is a largedistance compared to the size of the eddy structureunder investigation but no closer oceanic buoy windmeasurement was available

The wind appears to be highly variable in both in-tensity and direction Besides the predominant featurecharacterized by a continental northerly wind regime(either Mistral or Tramontane) the wind can experi-ence strong pulses (days 7 and 18) relaxations (days 26and 29) and reversals (days 4 6 8 16 and 26)

At the buoyrsquos location (top and middle plots) thedifferent wind fields display good global agreementsin direction with intense northwesterly and southsoutheasterly wind episodes Nevertheless careful in-vestigation exhibits important discrepancies at high fre-quency such as delays at sudden wind rotations (forinstance the southerly wind set up around December29th) The comparison of the wind model data at twodifferent locations (buoy platform in the middle panel

and eddy core in the lower panel) shows very littledifference in terms of magnitude and direction whichis confirmed by wind vector maps (not shown)

To highlight the eddy episodes gray boxes corre-sponding to the events observed by the radar system aresuperimposed to the stick diagrams in Fig 6 They wereobserved for different meteorological conditions Thefirst event (days 3ndash5) is characterized by an initial (andantecedent) southerly wind turning to a northwesterlywind The second event (days 16ndash17) is characterizedby prior southeasterly wind turning northerly Two suc-cessive eddy events were observed on days 22ndash24 and26ndash27 They correspond to periods that mainly exhibita northerly wind with either a decrease in intensity ora weak rotation The last episode is characterized bya strong southerly wind during the eddy observation(days 29ndash31)

An empirical orthogonal eigenfunction (EOF)method was applied to the MM5 data set of December2006 in order to highlight the predominant wind modesand their variability The EOF method (Lorentz 1956)


decomposes the data set into representative modes de-termined by empirical functions based on eigenmodesthat best describe the information in terms of varianceTo take into account the 2D nature of the problemboth wind components are analyzed in parallel andlinked in a cross-correlation function (Kaihatu et al1998)

Figure 7 shows the mean field the first spatial mode(EOF1 which accounts for 50 of the total variance)and the associated amplitude While the mean fieldis characterized by a weak northerly wind (5 ms)the first component presents oscillations of the windfrom south to north with very strong peaks (17 mson December 6 for Mistral 11 ms on December 8 forthe southerly wind at the buoy location) As indicatedby the principal component analysis of the wind andconfirmed by the local observations (the GoL is a well-known windy region) the predominant feature is theMistral This land wind occurs mainly in winter can lastfor few days reach very high speed (30 ms) and canblow homogeneously over large regions (Guenard et al2006) The temporal coefficient of the first EOF showsa regular sign inversion which means that southerlywind is also a preponderant wind sector

4 Analysis of eddy generating mechanisms

The goal here is to generate an anticyclonic eddy simi-lar to the observed one through specific and idealizedwind-driven simulations allowing a dynamical analysisto identify the dominant mechanism

41 Setup of the numerical model

The model used in this investigation is 3D hydrodynam-ical model for application at regional scale (MARS3D)documented by Lazure and Dumas (2008) It is a freesurface sigma-coordinate model resolving primitiveequations under Boussinesq and hydrostatic approxi-mations A barotropicndashbaroclinic mode splitting is used(Blumberg and Mellor 1987) with the same time stepthanks to the alternating direction implicit scheme forthe external mode The model uses a staggered C-gridaccording to Arakawa and Lamb (1977) The horizon-tal spatial resolution is 1200 m while the vertical direc-tion is discretized with 30 levels refined at the surfaceand at the bottom to better resolve the boundary layers

The turbulent scheme for vertical diffusion is theRichardson number dependent Pakanowski and Phi-lander formulation (Pacanowski and Philander 1981)while horizontal turbulent viscosity coefficients are

obtained using Smagorinsky formula (Smagorinsky1963)

νH = α xy

radic(partupartx

)2

+(

partv

party

)2

+ 1

2

(partuparty

+ partv

partx

)2

(1)with α = 02 and a minimum value set to 20 m2s in thedomain enlarged in boundary sponge layers to avoidnumerical instabilities

Previous studies of the northwestern Mediterraneansea circulation were conducted with MARS3D via nestedconfigurations investigating surface dynamics (Andreet al 2005) the variability of the NC (Andre et al2009 Rubio et al 2009b) wave- and current-inducedbottom shear stress (Dufois et al 2008) or anchovy re-cruitment through Lagrangian transport (Nicolle et al2009) Most of them used MM5 atmospheric forcing forsurface boundary conditions

Here the model is implemented in an idealizedconfiguration and not derived from a nesting chainto highlight specific physical processes The only back-ground circulation taken into account is an idealizedNC flowing westward along the continental slopeThus open boundary conditions were adapted to per-mit its entrance at the eastern side of the domain and itsoutflow at the west To generate the NC a realistic seasurface slope and density fields have been analyticallybuilt in agreement with the geostrophic balance Atopen boundaries temperature salinity and sea surfaceelevation were prescribed whereas the velocities had ano gradient condition To avoid spurious effects nearthe boundaries due to differences between the externalanalytical prescribed solution and the internal numeri-cal one a sponge layer was applied as shown in Fig 8Besides temperature and salinity were relaxed towardthe external forcing in a band of ten grid cells nearthe boundary with a characteristic time of few hoursYet unexpected oscillations of the sea surface levelappeared over the continental slope To overcome thisnumerical flaw we chose to simply prescribe a linearcombination of external analytical sea surface layer andinternal computed one αζinternal + (1 minus α)ζexternal

This above formulation is equivalent to a juxta-position of a Sommerfeld radiation condition and arelaxation condition (Blayo and Debreu 2005) tun-ning the relative importance of both effects throughalpha coefficient After few numerical experiments thevalue of alpha was set to 05 at the southern boundaryand 005 at the eastern one leading in both case toa strong relaxation toward the prescribed sea surfaceelevation The effective relaxation time is then 600 s atthe southern boundary and 315 s at the eastern one fora time step of 300 s On the West a single Sommer-


Fig 8 Model domain andbathymetry for realistic gulf(B1) Contour intervals are50 m until iso-200 m 200 mfrom iso-200 m to iso-3000 mThe boundary sponge layer issuperimposed in gray

feld radiation condition was considered to allow freeoutflow of the current This configuration satisfactorilyreproduced the NC without any perturbation inside thedomain of interest

Several numerical tests have been conducted whichdiffer according to the configuration and the windforcing The different options concerning bathymetrydesign thermohaline distribution background circula-tion river runoff and wind forcing are summarized inTable 1 and described hereafter

Three different bathymetries have been retained tostudy the impact of the coastline design and bottom

topography on the eddy generation The first one isbased on the best available estimate of the bathymetryfrom the coastline to the 2000-m isobath thus includingcanyons but smoothed at the boundaries to erase thetopographic constrain of the islands in the south of thedomain (Corsica and Balearic Islands) and to facilitatethe prescription of an idealized NC (Fig 8 run B1 inTable 1) The second one represents a circular gulf witha uniform continental slope corresponding to a filteredand smoothed version of the real bathymetry especiallyconcerning the coastline (B0) A third bathymetry isimplemented similar to the first one except a flat 200-

Table 1 Idealized runcharacteristics according tothe configuration and forcingoptions

Run_B1D1NC1R1W1 willcorrespond to realisticbathymetry vertical densitystructure with a NC signatureRhocircne river runoff andnortherly wind forcing Whennot specified wind intensity is16 ms and Rhocircne riverrunoff is 1200 m3s

Configuration 0 1 2options

Bathymetry Smoothed Realistic Flat 200-m shelfcoastline

Density Homogeneous Vertical profile(T = 13C

S = 385 psu)

Northern current No Yes

Rhocircne river runoff No Yesfresh and cold(T = 6C) water(debit 800 m3sor 1200 m3 s)

Wind forcing No Northerly Southerly(intensity 10 12 14 or 16 ms) (intensity 10

or 16 ms)(duration 2 3 or 4 days) (duration 1 2

or 4 days)


m-deep shelf to study the effect of the shelf topogra-phy (B2)

Regarding the initial thermohaline distribution twoidealized density fields are considered horizontally andvertically homogeneous over the whole domain (D0)or horizontally homogeneous but with a vertical profilecorresponding to a typical winter stratified water (D1)over the abyssal plain In the first case (D0) the NC(if any) is only balanced by a sea surface slope InDecember the seasonal stratification vanishes on theshelf due to repeated Mistral and Tramontane gustsTherefore the dynamics may be considered as fullybarotropic in the GoL except in the Rhocircne plumeand no interactions between eddies and stratificationare expected Figure 9a represents the temperature andsalinity profiles considered in case D1 Associated with

the vertical density profile a temperature and salin-ity boundary condition is imposed to mark the NCwith a thermohaline signature Figure 9b representsthe surface circulation and temperature after 1 monthof spin-up when the NC has been generated by a seasurface gradient (NC1) The coastal current follows thecontinental shelf with no intrusion into the GoL andcarries warm water Figure 9c represents a meridionalsection at 6 E across the NC showing its density signa-ture as imposed on the eastern boundary compared tothe surrounding initial thermohaline distribution (D1at latitude 424 N) The NC is well-defined downto 400 m depth 30 km wide with maximum veloci-ties around 035 ms This configuration using B1 forbathymetry D1 for density and NC1 for the currentis kept as reference as it is consistent with the liter-

a b

c

Fig 9 a Model initial stratification salinity and temperatureprofile Run_B1D1NC1R0 after 1-month spin-up b surface tem-perature (degree Celsius) and current vectors (meters per sec-

ond) The boundary sponge layer is superimposed in gray cdensity section and zonal velocity contours at longitude 6 EContour interval is 005 ms The transect is represented on b


ature for winter conditions (Conan and Millot 1995Mounier et al 2005) Conversely the simulation of abarotropic current (D0 NC1) shows less realistic char-acteristics (800 m deep 35 km wide maximum velocity02 ms) Some simulations are also conducted withoutany coastal current (NC0) by removing surface ele-vation and horizontal density gradients at the easternboundary

For all simulations the numerical spin-up phase is setto 1 month with no surface atmospheric forcing to allowa stable NC set-up and equilibrium over the wholedomain Indeed the domain-integrated kinetic energyapproaches stationary value after 15 days maximum

The influence of river runoff is tested in some simula-tions but limited to the Rhocircne river discharges (R1) asit provides 90 of the gulfrsquos freshwater input (Bourrinand Durrieu de Madron 2006) To simulate the Rhocircnerunoff freshwater debouches into the sea from a chan-nel added in the land mask with realistic width anddepth to take into account the input of momentumat the Rhocircne mouth The reference outflow rate isset to a constant value of 1200 m3s which representsthe average discharge of daily outflows measured inDecember 2006 while some specific tests on the dis-charge value are run with lower outflow of 800 m3sAs regard to the estimated residence time of freshwaterfrom the Rhocircne in the GoL a spin-up of 1 month is alsoconsistent with a realistic buoyancy forcing on the shelf(Durrieu de Madron et al 2003)

Once a stable background dynamic is generateddifferent meteorological forcings are applied They arereduced to wind stress thus no atmospheric heat ormass fluxes are taken into account The stress calcula-tion is done with a bulk type formulation and constantdrag coefficient (Cd = 1 210minus3) The EOF analysis jus-tifies the use of northerly and southerly idealized windsas our principal forcings in the numerical simulations(W1 and W2 in Table 1) The wind forcing starts witha linear ramp during 24 h and is then kept to a highuniform constant value during a few days

These wind forcings were tested separately on thereference configuration (run_B1D1NC1) correspond-ing to a coastal baroclinic current bordering a realisticshelf and on additional combinations of bathymetry de-sign and density structure according to Table 1 leadingto more than 20 runs The impact of wind and riverrunoff was also tested by changing their magnitudewithin a realistic range

The reproduction of wind-driven shelf circulationspreviously described by Estournel et al (2003) andPetrenko et al (2008) for different idealized wind forc-ings validates our reference configuration However inthis paper we describe modeling experiments resulting

from curl-free wind stress as no clear anticyclonic eddygeneration could be imputed on wind curl in the areaof interest according to several modeling tests Con-cerning the Rhocircne plume position which is also stronglydependent on wind forcing model results were coher-ent with previous studies for typical wind directions(Estournel et al 1997 Ulses et al 2005)

42 Northerly wind-driven circulation

Northerly wind forcing was tested on the referenceconfiguration corresponding to a realistic coastlinedensity front and NC (run_B1D1NC1R0W1 Table 1)In that case an anticyclonic eddy is generated under thesurface layer (0ndash20 m) Figure 10 represents a snapshotafter 4 days of wind forcing 16 ms intense to whichthe initial state without wind has been substractedThe difference highlights the anomalies in terms ofcirculation and surface elevation of the gulf dynamicsinduced directly by the wind

Wind stress acts dynamically on sea surface elevationover the whole gulf Figure 10 exhibits a high and lowpressure area on the western and eastern coasts respec-tively due to a surface westward Ekman transport Theresulting onshore pressure gradient at the eastern coastinduces an alongshore southeastward coastal upwellingjet which is in approximate geostrophic balance and fol-lows the coastline orientation The coastline irregulari-ties at 52 E behave like a wall constraining the currentto flow southward generating negative vorticity Thisanticyclonic eddy apparent off Marseille area on thevertically integrated velocity vectors exhibits speeds ofabout 02 ms Its position and size (30 km diameter)estimated by visual inspection of surface current andvorticity maps are in agreement with the character-istics of the eddies observed by radars In terms oflocation and size this eddy also corresponds to thoseobserved at different dates with the Tethys II ADCPdatabase and measurements by Petrenko (2003) duringthe Moogli 2 cruise in June 1998 with strong northerlywind conditions

The modeled structure extends from the bottomto under the surface layer where the circulation fol-lows the Ekman drift Figure 11 displays Hovmollerdiagrams of the zonal (a) (at longitude 51 E) andmeridional (b) (at latitude 432 N) integrated velocityalong sections crossing the eddy One can see the eddyformation during the wind stress entirely formed after3 days Then it remains stable as long as the wind keepsblowing

Once the deep anticyclonic eddy is formed a windrelaxation reduces the Ekman surface dynamics allow-ing its detection by radar measurements at the surface


Fig 10 Sea surface height(meters) and verticalintegrated velocities (metersper second) differencesCirculation from referencerun without wind(run_B1D1NC1R0W0) hasbeen substracted from thenortherly wind-drivencirculation(run_B1D1NC1R0W1 atday 5) to highlight anomaliesTransects for Fig 11a and bare also represented

The previous numerical experiment has been contin-ued after the formation of the deep eddy enabling aninvestigation of the behavior of the system when thewind stops Sequential vertical slices across the eddy(same transects than for Fig 11) are depicted in Fig 12to show the evolution of the vertical eddy structureWhile the strong and constant northerly wind is blowing(upper panels) the eddy is formed in the whole watercolumn under the surface layer In the surface Ekmanlayer no eddy is detectable and the southwestward flowis almost spatially homogeneous The surfacing of theeddy structure is made possible by the wind stop asshown on lower panels of Fig 12 corresponding to2 days after the wind relaxation

Figure 13 depicts the OkubondashWeiss parameter distri-bution at the surface at the same date This parameteris defined by the difference between the strain and thevorticity (W = s2

n + s2s minus w2 where sn ss and w are the

normal and shear components of strain and vorticityrespectively) W is a measure of the stirring and mix-ing in the eddy field and helps to identify eddy cores(Basdevant 1994) and barriers for transport applica-tions (Pasquero et al 2001 Testor and Gascard 2005Isern-Fontanet et al 2004) The strong negative coreat the eddy location is well-defined and representsnegative vorticity according to the current vectors Thecomputed eddy is slightly smaller (20 km diameter)than the observed ones (Fig 3) nearly at the samelocation especially for the 4th and 17th of Decemberevents and its lifetime is shorter (about 10 h) due toa strong superimposed inertial signal at surface Theassociated velocities are reduced to about 015 ms

The impacts of the bathymetric constraint the Rhocircneriver discharge and the NC are analyzed by running dif-ferent configurations Figure 14 depicts the circulationfor some of the configurations in terms of integratedvelocity (upper panels) and surface velocity (lowerpanels) Panels a and e are the resulting referencerun circulation showing the eddy clearly apparent offMarseille area on the depth integrated velocity vectorswith speed of about 02 ms (same as Fig 10) and thesuccessive surface detection after 2 days of wind relax-ation (same as Fig 13) According to several runs con-sistent with bathymetric options the barotropic eddycannot develop with an idealized smoothed coastline(B1 Fig 14b) or a constant 200-m depth shelf (B2Fig 14c) neither can the surface eddy after relax-ation This demonstrates the importance of the complexcoastline and the details of bathymetry for the genera-tion of confined eddies

Figure 14d h corresponds to run_B1D1NC0R1W1ie without any bordering current but including theRhocircne river output Neither the NC nor the freshwaterdischarge influences the deep eddy formation (paneld) Nevertheless after the wind relaxation the surfaceeddy intensity is slightly strengthen by the supply ofriver discharge Without any bordering current theeddy is drifted southward suggesting that the NC tendsto constrain the eddy close to the coast

To estimate the impact of the intensity of the windon the eddy generation and its temporal evolutionwe made additional tests by making vary the intensityof the wind in a realistic range according to Fig 6afrom 10 to 16 ms (Table 1) In the framework of our


Fig 11 Hovmoller diagram(run_B1D1NC1R0W1) alongeddy crossing sections azonal (meters per second) atlongitude 51 E and bmeridional (meters persecond) at latitude 432 Ndepth-integrated velocitysuperimposed to currentvectors The transects arerepresented in Fig 10

a

b

idealized model configuration only very strong windshigher than 14 ms led to the eddy evidence This lowerbound could be reduced when using more realisticconfigurations including an energetic initial field In-deed a 12-ms northerly wind was sufficient to generatea similar eddy feature when initializing with a morerealistic density field including horizontal gradients

Hence the generation of this eddy could be ex-plained by the relaxation of a strong northerly windletting the deep eddy reach the surface This hypothesisis corroborated by another simulation using the samewind relaxation but only after 2 days of constant windwhich is not sufficient for the generation of the deep

eddy In that case no significant negative vorticity isdetected at the surface

Similar wind scenarios could be identified in Fig 6and related to the eddy observation Strong northerlyevents as observed in all data set from the 18th to the22nd of December (Fig 6) lead to surface wind-drivensouthwestward circulation according to radar maps (notshown) as expected by Ekman dynamics theory Inthis usual configuration no surface eddy could be ob-served but a deep one could exist according to ouridealized simulations A surface eddy was measured afew days later (days 22ndash24) during a local wind relax-ation (Fig 6 lower panel) A faster wind sequence was


a b

c d

Fig 12 Vertical sections of meridional and zonal velocities (meters per second) crossing the eddy (run_B1D1NC1R0W1) after 4 daysof northerly wind stress (upper panels) and then 2 days of wind relaxation The transects are represented in Fig 10

Fig 13 Surface OkubondashWeiss parameter and currentvectors for run_B1D1NC1R0W1 at day 7Four days of continuousnortherly wind was followedby 2 days of relaxation


Fig 14 Idealized simulation for northern wind a e REF(run_B1D1NC1R0W1) b f idealized bathymetry and coastline(run_B0D0NC1R0W1) c g idealized topography (flat 200-mshelf run_B2D1NC1R0W1) d h REF with Rhocircne river dis-

charge no NC (run_B1D1NC0R1W1) andashd represent the verti-cally integrated currents after 4 days of wind endashh represent thesurface currents 2 days after the wind relaxation

present for the fourth eddy event in December 2006(days 26ndash27) with strong Mistral immediately followedby a relaxation during eddy observation

43 Southerly wind-driven circulation

The complementary feature appearing in the first windEOF (Fig 7) is a southerly wind with specific eventsat days 1ndash3 (before the generation of the first eddyobserved by radars) or days 29ndash31 during the last eddymeasurements (Fig 6) Our second idealized wind forc-ing (W2) specifically concerns this wind regime

According to run_B1D1NC1R1W2 a southerly windhas no effect on a barotropic anticyclonic circulationand we do not expect any deep eddy structure Thewind only affects the position of the Rhocircne river plumewhich is partly constrained to the northeastern coastof the GoL Ulses et al (2005) observed and modeledsimilar features in the Gulf of Fos the small semi-enclosed bay close to the Rhocircne mouth

After the relaxation of a 4-day strong homogeneoussoutherly wind a surface anticyclonic eddy is gener-ated initially positioned close to the coast then grow-ing and extending southward Figure 15 shows theOkubondashWeiss parameter corresponding to this surfaceeddy after 5 days without wind It shows the eddy corewith high vorticity (negative values) and the surroundedcirculation cell with high rates of strain (positive val-ues) with stronger extrema than for the simulationswith Mistral (Fig 13) The total eddy diameter is be-tween 20 and 30 km with velocities of 03 ms Thiseddy is generated approximately 15 days after the windrelaxation and grows for a few days

Figure 16 exhibits the temporal evolution of surfacedensity and current vectors at longitude 51 E bringinginto light the correlation between the southward exten-sion of the eddy with the lighter front Another inter-esting feature appears in the intermediate part of thediagram corresponding to a section 43ndash432 N and tothe 130-m isobath The arrows indicating the directionand speed of the current reveal an anticyclonic rotation


Fig 15 SurfaceOkubondashWeiss parameter andcurrent vectors forrun_B1D1NC1R1W2 atday 10 Four days ofcontinuous southerly windwas followed by 5 days ofrelaxation Transect forFig 16 is also represented

with a 175-h period for almost five successive periodsThis corresponds to inertial motion confirmed by thethermocline (not shown) signature both decaying intime due to frictional effects and being irregular due tothe presence of additional waves

Figure 17 illustrates the surface circulation atdifferent times (2 days in the upper panel and 5 daysin lower panels) after wind relaxation according todifferent run configurations Panels a and e correspondto the reference configuration B1D1NC1R1W2 (same

Fig 16 Hovmoller diagramof surface densitysuperimposed to currentvectors at longitude 51 E(starting at relaxation after4 days of southerly windrun_B1D1NC1R1W2) Thetransect is represented inFig 15


Fig 17 Idealized simulations for southern wind a e REF(run_B1D1NC1R1W2) with surface salinity (PSU) b f REFwithout Rhocircne river discharge (run_B1D1NC1R0W2) c g ide-alized topography (flat 200-m shelf run_B2D1NC1R1W2) d

h REF without NC (run_B1D1NC0R1W2) andashd represent thesurface currents 2 days after the wind relaxation endashf 5 days afterthe wind relaxation

as Fig 15) where the circulation is superimposed tothe surface salinity signature The eddy is marked bya freshwater patch from Rhocircne river freshwater thatwas advected eastward by the wind forcing and trappedin the eddy This transport effect could have importantlocal consequences regarding pollution and materialdispersion

A necessary condition for this generation mechanismis the freshwater output of the Rhocircne river as indicatedby the simulation without river discharge (panels band f) Additional simulations have been carried outwith the same wind forcing but different configurationcombinations to evaluate their influence on this mech-anism like the NC or the bathymetry Results withoutany current bordering the shelf (NC0 Fig 17d h) orwith a constant gulf depth set at 200 m (B3 Fig 17cg) demonstrate that neither the NC nor the gulf slop-ing topography are necessary for this eddy generationmechanism Yet both contribute to reinforce it in sizeand intensity and the NC clearly constrains the eddyclose to the coast Additionally the effects of windduration and intensity and Rhocircne river runoff havebeen investigated (Table 1) Two days of wind stress

is sufficient for the generation mechanism while lowervalues of wind intensity or freshwater runoff lead to asmaller and weaker eddy

Only southerly winds can constrain a part of theplume in the Marseille bay a necessary condition forthe buoyancy gradients formation and therefore theeddy generation Usually the plume is either flowingwestward of the river mouth for weak winds (due toCoriolis effects) or easterly winds either extendingsouthwestward for northwesterly winds (Ulses et al2005 Estournel et al 1997 2001 Broche et al 1998Marsaleix et al 1998) hence not really affecting theeddy area Yet this kind of structure is known in the lit-erature as a freshwater bulge and was recently observedclose to other estuaries as in the Hudsonrsquos river plumeby ocean color imagery and surface current radars(Chant et al 2008) and in the Columbia river plume byADCP and conductivityndashtemperaturendashdepth measure-ments (Horner-Devine 2009) The generation mecha-nisms have been previously studied theoretically orvia idealized studies (Yankovsky and Chapman 1997Avicola and Huq 2003) or (Choi and Wilkin 2007)pointing out to the importance of the high river dis-


charge as well as an irregular coastline featuring largeangle between the coast and the outflow Our studiedarea characterized by a complex coastline is consistentwith these characteristics

5 Conclusions

This research investigates the dynamics of anticycloniceddies which were observed off the coast of Marseillein the Gulf of Lions several times during a 17-monthHF radar campaign in 2005ndash2007 The eddy featuresa diameter reaching 40 km a variable lifetime (a fewhours up to 25 days) and an irregular occurrence Thispaper specifically focus on five of them which wereobserved with remarkable consistency and durationin December 2006 The main issue of this study wasto identify physical mechanisms that could lead to itsgeneration in particular the effect of wind forcingswhich are very specific and intense in the area Otherdistinctive features of the region were investigated toothrough different idealized configurations permittingto evaluate the influence of the bathymetry Rhocircneriver outflow buoyancy and general circulation (NC)

This process-oriented study suggested two possiblemechanisms radically differing one from each otherThe first one goes through two steps Due to the gulfrsquosgeometry and shallowness northerly winds create adepression at the eastern side of the GOL leadingto a southeastward barotropic geostrophic jet Thenoff Marseille the curving coastline constrains the up-welling jet southward generating negative vorticityThe anticyclonic structure is generated with an exten-sion corresponding to the shelfrsquos expansion similar tothe one observed at different depths by ADCP cam-paigns This eddy forms by intense wind conditionsthus only under the surface layer subject to Ekmancirculation According to numerical simulations theeddy is apparent at the surface after the wind relaxationand accelerated by both the Rhocircne water discharge andthe bordering northern current In this mechanism thecoastline and the bathymetry play a fundamental roleby confinement effect

The second possible generation mechanism is drivenby buoyancy forcing resulting from the Rhocircne riveroutflow Strong southerly winds constrain part of theplume in the Marseille bay During wind relaxation afreshwater surface lens or bulge detaches from the coastwith high buoyancy gradient leading to a well-definedanticyclonic eddy remaining stable for a few days

These generation mechanisms are both physicallyrealistic and coherent with the regionrsquos characteristics

The first one is assumed to be more common as clima-tological statistics on wind conditions in the GOL showhigher frequency of Mistral (45 of winds exceeding10 ms) compared to southerly winds (5 of windsectors)

The high wind intensity required to numerically re-produce the eddy in our idealized simulations can bereduced in the case of a more realistic and energeticdensity field However when a realistic configurationis used the precise process identification becomesdifficult since all forces interact simultaneously Thusa neutral initial state was chosen to analyze the eddydriving mechanism with adapted forcings that may bemodulated in realistic conditions

Additional tests on wind structure have been con-ducted from further EOF modes different typical windcombinations and spatial structures toward the realis-tic MM5 wind data Concerning the simulations forcedby additional idealized wind only a northwesterly wind(Tramontane) could generate a deep vortex with sim-ilar dynamics than with northerly wind Easterly oneswhich constitute the second EOF as well as spatial ortemporal combination of typical winds do not lead tothe generation of the vortex or to a clear identificationof a mechanism Eventually to complete the analysison the effect of wind forcing on the eddy generationthe MM5 atmospheric data set during December 2006was used to force the oceanic model both in an op-erational nesting way and in an academic way Noneof these runs managed to well reproduce a surfacevortex similar to the one observed by HF radars atthe correct dates Several causes can be at the originof the missing feature Intense wind events may beunderestimated by atmospheric models and the localrelaxation we have shown to be crucial for the eddysurfacing may be slightly displaced in space Besidesno significant negative wind stress curl are observed inthe area on the model outputs while it could favor theanticyclonic motion This reinforces the general ideathat reproducing with model such small-scale oceanicstructures as the eddy investigated in this paper remainsa major challenge

Very high resolution is needed for hydrodynamicconfigurations as well as for bathymetric precision andatmospheric forcings The use of recently developedmeteorological models with higher spatial and tem-poral resolution is expected to strongly influence andhopefully improve the prediction of mesoscale and sub-mesoscale coastal structures Furthermore the correctparametrization of momentum and heat fluxes at theairndashsea interface remains a challenge as this couldstrongly impact the oceanic circulation regardless ofthe resolution The assimilation of surface currents as


measured by coastal radars may also improve the nu-merical simulations

For further process-oriented investigations com-bined in situ observations would be useful permittingto monitor simultaneously surface dynamics throughHF radars and deeper circulation though ADCP tran-sects and density profiles for instance Additional pro-cesses evidenced by radar measurements would alsodeserve specific analysis like inertial motion coastalcurrent instabilities and surface circulation seasonalvariability

Acknowledgements The radar campaign was founded by theFrench ECOLO-PNEC (Programme National EnvironnementCocirctier) project We are thankful to Yves Barbin and JoelGaggelli who collected and post-treated the data The researchwas supported by IFREMER and METEO FRANCE in theframework of a PhD grant and by GIRAC Pocircle Mer Finallythe authors would like to thank ACRI ST company for the MM5configuration managing Most of the simulations have been runusing IFREMER calculation facilities

References

Alberola C Millot C Font J (1995) On the seasonal andmesoscale variabilities of the northern current during thePRIMO-O experiment in the western Mediterranean SeaOceanol Acta 18(2)163ndash192

Allou A Forget P Devenon JL (2010) Submesoscale vortexstructures at the entrance of the Gulf of Lions in the north-western Mediterranean Sea Cont Shelf Res 30(7)724ndash732

Andre G Garreau P Fraunie P (2009) Mesoscale slope currentvariability in the Gulf of Lions Interpretation of in situ mea-surements using a three dimensional model Cont Shelf Res29(2)407ndash423

Andre G Garreau P Garnier V Fraunie P (2005) Modelledvariability of the sea surface circulation in the north westernMediterranean Sea and in the Gulf of Lions Ocean Dyn55294ndash308

Arakawa A Lamb V (1977) Computational design of the ba-sic dynamical processes of the UCLA general circulationmodel Methods Comput Phys 17173ndash265

Astraldi M Gasparini G (1992) The seasonal characteristics ofthe circulation in the north Mediterranean basin and their re-lationship with the atmosphericndashclimatic conditions J Geo-phys Res-Oceans 97(C6)9531ndash9540

Avicola G Huq P (2003) The characteristics of the recirculat-ing bulge region in coastal buoyant outflows J Mar Res61(4)435ndash463

Barrick D (1971) Theory of HF and VHF propagation across therough sea 2 Application to HF and VHF propagation abovethe sea Radio Sci 6(3)527ndash533

Barth A Alvera-Azcarate A Weisberg RH (2008) Assimila-tion of high-frequency radar currents in a nested modelof the West Florida Shelf J Geophys Res 113(C08033)doi1010292007JC004585

Basdevant C Philipovitch T (1994) On the validity of the OkubondashWeiss criterion in two-dimensional turbulence Physica D11317ndash30

Bassin C Washburn L Brzezinski M McPhee-Shaw E (2005)Sub-mesoscale coastal eddies observed by high frequency

radar a new mechanism for delivering nutrients to kelpforests in the Southern California Bight Geophys Res Lett32(12)L12604

Blayo E Debreu L (2005) Revisiting open boundary conditionsfrom the point of view of characteristic variables OceanModel 9(3)231ndash252

Blumberg A Mellor G (1987) A description of a three dimen-sional coastal ocean circulation model Three-dimensionalcoastal ocean models Coast Estuar Sci 41ndash16

Bourrin F Durrieu de Madron X (2006) Contribution to thestudy of coastal rivers and associated prodeltas to sedimentsupply in the Gulf of Lions (NW Mediterranean Sea) Vie etmilieumdashLife Environ 56(4)307ndash314

Broche P Devenon J Forget P de Maistre J Naudin J CauwetG (1998) Experimental study of the Rhone plume Part Iphysics and dynamics Oceanol Acta 21(6)725ndash738

Chant RJ Glenn SM Hunter E Kohut J Chen RF HoughtonRW Bosch J Schofield O (2008) Bulge formation of a buoy-ant river outflow J Geophys Res-Oceans 113(C1)C01017

Choi BJ Wilkin JL (2007) The effect of wind on the dispersal ofthe Hudson River plume J Phys Oceanogr 37(7)1878ndash1897

Conan P Millot C (1995) Variability of the northern current offMarseilles western Mediterranean Sea from February toJune 1992 Oceanol Acta 18(2)193ndash205

Coulliette C Lekien F Paduan J Haller G Marsden J (2007) Op-timal pollution mitigation in Monterey Bay based on coastalradar data and nonlinear dynamics Environ Sci Technol41(18)6562ndash6572

Doglioli AM Blanke B Speich S Lapeyre G (2007) Trackingcoherent structures in a regional ocean model with waveletanalysis application to Cape Basin eddies J Geophys Res112(C05043) doi1010292006JC003952

Dufau-Julliand C Marsaleix P Petrenko A Dekeyser I (2004)Three-dimensional modeling of the Gulf of Lionrsquos hydro-dynamics (northwest Mediterranean) during January 1999(MOOGLI3 experiment) and late winter 1999 westernMediterranean intermediate waterrsquos (WIWrsquos) formation andits cascading over the shelf break J Geophys Res-Oceans109(C11)C11002

Dufois F Garreau P Le Hir P Forget P (2008) Wave- andcurrent-induced bottom shear stress distribution in the Gulfof Lions Cont Shelf Res 281920ndash1934

Durrieu de Madron X Denis L Diaz F Garcia N Guieu CGrenz C Loye-Pilot MD Ludwig W Moutin T RaimbaultP Ridame C (2003) Nutrients and carbon budgets for theGulf of Lions during the Moogli cruises Oceanol Acta26421ndash433

Echevin V Crepon M Mortier L (2002) Interaction of a coastalcurrent with a gulf application to the shelf circulation of theGulf of Lions in the Mediterranean Sea J Phys Oceanogr33188ndash206

Echevin V Crepon M Mortier L (2003) Simulations and analysisof the mesoscale circulation in the northwestern Mediter-ranean Sea Ann Geophys 21281ndash297

Essen HH Gurgel KW Schlick T (2000) On the accuracy of cur-rent measurements by means of HF radar IEEE J OceanicEng 25472ndash480

Estournel C Broche P Marsaleix P Devenon J Auclai F VehilR (2001) The Rhone river plume in unsteady conditionsnumerical and experimental results Estuar Coast Shelf Sci53(1)25ndash38

Estournel C Durrieu de Madron X Marsaleix P Auclair FJulliand C Vehil R (2003) Observation and modelisation ofthe winter coastal oceanic circulation in the Gulf of Lions un-der wind conditions influenced by the continental orography(FETCH experiment) J Geophys Res 108(C3)8059


Estournel C Kondrachoff V Marsaleix P Vehil R (1997) Theplume of the Rhone numerical simulation and remote sens-ing Cont Shelf Res 17(8)899ndash924

Flexas M Durrieu de Madron X Garcia M Canals M ArnauP (2002) Flow variability in the Gulf of Lions during theMATER HFF experiment (MarchndashMay 1997) J Mar Syst33197ndash214

Flexas M van Heust G Treling R (2005) The behavior of jetcurrents over a continental slope topography with a possibleapplication to the northern current J Phys Oceanogr 35790ndash810

Forget P Barbin Y Andre G (2008) Monitoring of surface oceancirculation in the Gulf of Lions (north-west MediterraneanSea) using WERA HF radars In Proceedings IGARSSBoston USA

Forget P Broche P Demaistre J (1982) Attenuation with dis-tance and wind-speed of HF surface-waves over the oceanRadio Sci 17(3)599ndash610

Gatti J Petrenko A Devenon J Leredde Y Ulses C (2006)The Rhone river dilution zone present in the northeasternshelf of the gulf of lion in December 2003 Cont Shelf Res261794ndash1805

Grell G Dudia J Stauffer D (1994) A description of the fifth-generation Penn- StateNCAR Mesoscale Model (MM5)NCAR technical report note TN-398 National Center forAtmospheric Research Boulder

Griffa A Lumpkin R Veneziani M (2008) Cyclonic and an-ticyclonic motion in the upper ocean Geophys Res Lett35L01608

Guenard V Drobinsky P Caccia J Tedeschi G Currier P (2006)Dynamics of the MAP IOP 15 Mistral event observationsand high-resolution numerical simulations QJR MeteorolSoc 132757ndash777

Gurgel K Antonischski G Essen H Schlick T (1999) WellenRadar (WERA) a new ground-wave HF radar for oceanremote sensing Coast Eng 37(3ndash4)219ndash234

Gurgel K Essen H Schlick T (2002) The role of HF radar withinoperational forecasting systems of the ocean In Geoscienceand remote sensing symposium IGARSS IEEE Interna-tional 1 pp 512ndash514

Henson SA Thomas AC (2008) A census of oceanic anticycloniceddies in the Gulf of Alaska Deep-sea Res Part 1 OceanogrRes Pap 55(2)163ndash176

Horner-Devine AR (2009) The bulge circulation in the ColumbiaRiver plume Cont Shelf Res 29(1 Sp Iss SI)234ndash251

Hu ZY Doglioli AM Petrenko AA Marsaleix P Dekeyser I(2009) Numerical simulations of eddies in the Gulf of LionOcean Model 28(4)203ndash208

Hua B Thomasset F (1983) A numerical study of the effects ofcoastline geometry on wind-induced upwelling in the Gulf ofLions J Phys Oceanogr 13(4)678ndash694

Isern-Fontanet J Font J Garcia-Ladona E Emelianov MMillot C Taupier-Letage I (2004) Spatial structure of anti-cyclonic eddies in the Algerian basin (Mediterranean Sea)analyzed using the OkubondashWeiss parameter Deep-sea ResII 513009ndash3028

Kaihatu J Handler R Marmorino G Shay L (1998) Empiricalorthogonal function analysis of ocean surface currents usingcomplex and real vector methods J Atmos Ocean Technol15927

Kaplan D Largier J Botsford L (2005) HF radar observationsof surface circulation off Bodega Bay (northern CaliforniaUSA) J Phys Oceanogr 110C10020

Lavrova OY Bocharova TY (2006) Satellite SAR observationsof atmospheric and oceanic vortex structures in the BlackSea coastal zone In Shea MA Gupta RK Menenti M

Lopez RA (eds) Remote sensing of oceanographic processesand land surfaces space science education and outreach (ad-vances in space research-series) vol 38 Elsevier Amster-dam pp 2162ndash2168

Lazure P Dumas F (2008) An externalndashinternal mode couplingfor a 3D hydrodynamical model for applications at regionalscale (MARS) Adv Water Resour 31(2)233ndash250

Lipa B Nyden B Ullman DS Terrill E (2006) Seasonde ra-dial velocities derivation and internal consistency In IEEEjournal of oceanic engineering vol 31(4) 4th radiowaveoceanography workshop N Queensland Australia 2004 pp850ndash861

Lorentz E (1956) Empirical orthogonal function and statisti-cal weather prediction Science report 1 Statistical Fore-cast Project Department of Meteorology MIT (NTIS AD110268)

Marsaleix P Estournel C Kondrachoff V Vehil R (1998) Anumerical study of the formation of the Rhone River plumeJ Mar Syst 14(1ndash2)99ndash115

McWilliams J (1985) Submesoscale coherent vortices in theocean Rev Geophys 23165ndash182

Millot C (1979) Wind induced upwellings in the Gulf of LionsOceanol Acta 2(3)261ndash274

Millot C (1990) The Gulf of Lionsrsquo hydrodynamics Cont ShelfRes 10(9ndash11)885ndash894

Millot C Wald L (1980) The effect of Mistral wind on theLigurian current near Provence Oceanol Acta 3(4)399ndash402

Molcard A Poulain P Forget P Griffa A Barbin Y GaggelliJ Maistre JD Rixen M (2009) Comparison between VHFradar observations and data from drifter clusters in theGulf of La Spezia (Mediterranean Sea) J Mar Syst 78S79ndashS89

Mounier F Echevin V Mortier L Crepon M (2005) Analysis ofthe mesoscale circulation in the occidental MediterraneanSea during winter 1999ndash2000 given by a regional circulationmodel Prog Oceanogr 66251ndash269

Nencioli F Dong C Dickey T Washburn L McWilliams JC(2010) A vector geometry-based eddy detection algorithmand its application to a high-resolution numerical modelproduct and high-frequency radar surface velocities in theSouthern California Bight J Atmos Ocean Technol 27564ndash579

Nicolle A Garreau P Liorzou B (2009) Modelling for anchovyrecruitment studies in the Gulf of Lions (western Mediter-ranean Sea) Ocean Dyn 59953ndash968

Orlic M Kuzmic M Pasaric Z (1994) Response of the Adri-atic Sea to the Bora and Sirocco forcings Cont Shelf Res14(1)91ndash116

Pacanowski R Philander S (1981) Parametrization of verticalmixing in numerical-model of tropical oceans J Phy Ocean111443ndash1451

Paduan J Rosenfeld LK (1996) Remotely sensed surface currentsin Monterey Bay from shore based HF radar (Coastal OceanDynamics Application Radar) J Geophys Res 101(C9)20669ndash20 686

Paduan J Shulman I (2004) HF radar data assimilation in theMonterey Bay area J Geophys Res 109(C07S09) doihttp1010292003JC001949

Parks AB Shay LK Johns WE Martinez-Pedraja J Gurgel KW(2009) HF radar observations of small-scale surface currentvariability in the Straits of Florida J Geophys Res-Oceans114C08002

Pasquero C Provenzale A Babiano A (2001) Parametrizationof dispersion in two-dimensional turbulence J Fluid Mech439279ndash303


Petrenko A (2003) Variability of circulation features in the Gulfof Lions NW Mediterranean Sea importance of inertial cur-rent Oceanol Acta 26323ndash338

Petrenko A Leredde Y Marsaleix P (2005) Circulation in astratified and wind-forced Gulf of Lions NW MediterraneanSea in situ and modelling data Cont Shelf Res 257ndash27

Petrenko A Dufau C Estournel C (2008) Barotropic eastwardcurrents in the western Gulf of Lion north-western Mediter-ranean Sea during stratified conditions J Mar Syst 74406ndash428

Reffray G Fraunie P Marsaleix P (2004) Secondary flows in-duced by wind forcing in the Rhone region of freshwaterinfluence Ocean Dyn 54179ndash196

Robinson AR (1983) Eddies in marine science Springer NewYork

Rubio A Arnau P Espino M Flexas M Jorda G Salat JPuigdefabregas J Arcilla A (2005) A field study of thebehaviour of an anticyclonic eddy on the Catalan conti-nental shelf (NW Mediterranean) Prog Oceanogr 66(2ndash4)142ndash156

Rubio A Barnier B Jorda G Espino M Marsaleix P (2009a)Origin and dynamics of mesoscale eddies in the Catalan Sea(NW Mediterranean) insight from a numerical model studyJ Geophys Res-Oceans 114C06009

Rubio A Taillandier V Garreau P (2009b) Reconstruction ofthe Mediterranean northern current variability and associ-ated cross-shelf transport in the Gulf of Lions from satellite-tracked drifters and model outputs J Mar Syst 78S63ndashS78

Sammari S Millot C Prieur L (1995) Aspects of the seasonal andmesoscale variability of the northern current in the westernMediterranean Sea inferred from PROLIG-2 and PROS-6experiments Deep-Sea Res 42893ndash917

Schmidt R (1986) Multiple emitter location and signal parameter-estimation IEEE Trans Antennas Propag 34(3)276ndash280

Shadden S Lekien F Paduan J Chavez F Marsden J (2009) Thecorrelation between surface drifters and coherent structuresbased on high frequency radar data in Monterey Bay Deep-Sea Res 56161ndash172

Shay L Cook T An P (2003) Submesoscale coastal ocean flowsdetected by very high frequency radar and autonomous un-derwater vehicles J Atmos Ocean Technol 201583ndash1600

Shay L Lee T Williams E Graber H Rooth C (1998) Effectsof low frequency current variability on submesoscale near-inertial vortices J Geophys Res 10318691ndash18714

Smagorinsky J (1963) General circulation experiments with theprimitive equation I The basic experiment Mon WeatherRev 11199ndash165

Stewart R Joy J (1974) HF radio measurements of surface cur-rents Deep-Sea Res 211039ndash1049

Testor P Gascard JC (2005) Large-scale flow separation andmesoscale eddy formation in Algerian basin Prog Oceanogr66211ndash230

Ullman D O Donnell J Kohut J Fake TAA (2006) Trajec-tory prediction using HF radar surface currents MonteCarlo simulations of prediction uncertainties J Geophys Res111C12005

Ulses C Estournel C Bonnin J Durrieu de Madron X MarsaleixP (2008) Impact of storms and dense water cascading onshelf-slope exchanges in the Gulf of Lion (NW Mediter-ranean) J Geophys Res-Oceans 113(C2)C02010

Ulses C Grenz C Marsaleix P Schaaff E Estournel C MeuleS Pinazo C (2005) Circulation in a semi-enclosed bay underinfluence of strong freshwater input J Mar Syst 56(1ndash2)113ndash132

Yankovsky A Chapman D (1997) A simple theory for the fateof buoyant coastal discharges J Phys Oceanogr 27(7)1386ndash1401

Zamudio L Hogan P Metzger EJ (2008) Summer generation ofthe Southern Gulf of California eddy train J Geophys Res-Oceans 113(C6)C06020

Generation mechanisms for mesoscale eddies in the Gulf of Lions radar observation and modeling

Abstract

Introduction

Setup of the HF radar system

Eddy events observations and wind conditions

Eddy characteristics

Wind analysis

Analysis of eddy generating mechanisms

Setup of the numerical model

Northerly wind-driven circulation

Southerly wind-driven circulation

Conclusions

References


2007 Molcard et al 2009) support for safe navigationand search and rescue operations (Gurgel et al 2002Ullman et al 2006) as well as description of coastal cir-culation (Shay et al 1998 2003 Paduan and Rosenfeld1996) The assimilation of surface current data (Paduanand Shulman 2004 Barth et al 2008) into numericalocean model may also considerably improve the fore-casts Quantitative information on eddy characteristicscan be captured like their diameter elongation ve-locities vorticity lifetime or trajectories (Bassin et al2005 Parks et al 2009)

A few physical processes are known that can gener-ate such mesoscale structures McWilliams (1985) andRobinson (1983) detailed some of them for mesoscalecoherent vortices barotropic or baroclinic instabilitiesof the current leading to eddy formation with verticaland horizontal dimensions comparable to the currentwidth and the characteristic Rossby radius topographicboundary effects (frictional drag when flowing aroundtopography) or topography variation under a flow(stretching or compression of the water mass creatingvorticity according to the conservation of potential vor-ticity) mixing and adjustments in an unstable frontalregion and the β-effect based on the latitudinal vari-ation of the Coriolis parameter Furthermore Orlicet al (1994) and Estournel et al (2003) described theinfluence of the wind curl on eddy formation

The Gulf of Lions (GoL) is a particularly relevantarea to study mesoscale structures as it is a continen-tal shelf characterized by a microtidal regime suitablefor the generation and evidence of eddy structuresIndeed this quite shallow gulf (90 m mean depth) is

bounded by a complex coastline and subject to majorforcing mechanisms (Fig 1) The GoL circulation isstrongly influenced by the along-shelf northern current(NC) or Liguro-Provenccedilal current which is the north-ern branch of the general surface cyclonic circulationin the northwestern Mediterranean Sea (Astraldi andGasparini 1992) The NC is known as a density currentsubject to geostrophic equilibrium and often exhibitsa noticeable thermal signature in winter as it carrieswarmer water from the Ligurian basin Its high seasonalvariability was evidenced by Millot (1990) Conan andMillot (1995) and Alberola et al (1995) From spring toautumn the NC flows relatively far from the coast andremains shallow (150 m deep) and wide (40ndash60 km)while in winter it gets deeper (more than 200 m) andnarrower (30 km) appears to be closer to the coast(Petrenko 2003) and to be more intense with a trans-port ranging from 15 to 2 Sv

Far from being a stable current the NC exhibitsan important mesoscale activity (Alberola et al 1995Sammari et al 1995 Rubio et al 2009b) Meanders aregenerated from baroclinic instabilities related to pertur-bations of the stratification (with periods of 75 daysand wavelengths of 60 km according to Flexas et al(2002)) or from barotropic instabilities due to bot-tom topography gradients with shorter periods (around35 days according to Flexas et al (2005)) Althoughthe NC generally adjusts itself above the 1000ndash2000-misobaths (Flexas et al 2005) it sometimes intrudes intothe shelf on the eastern part of the GoL under specificcondition of wind and stratification (Millot and Wald1980 Echevin et al 2002 Petrenko 2003) In contrast

Fig 1 Schematic circulationin the Gulf of Lions includingthe Northern Current themain freshwater output(Rhocircne river) predominantwind regimes and bathymetry(isobaths 50 100 200 5001000 2000 m) Radarstations are located at points1 and 2 with the coveragerepresented by dashed lines




































32 Wind analysis







a

b




a

c

b

















νH = α xy

radic(partupartx

)2

+(

partv

party

)2

+ 1

2

(partuparty

+ partv

partx

)2
















S = 385 psu)





or 4 days)





a b

c



























a

b






a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References




































32 Wind analysis







a

b




a

c

b

















νH = α xy

radic(partupartx

)2

+(

partv

party

)2

+ 1

2

(partuparty

+ partv

partx

)2
















S = 385 psu)





or 4 days)





a b

c



























a

b






a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References




a

b




a

c

b

















νH = α xy

radic(partupartx

)2

+(

partv

party

)2

+ 1

2

(partuparty

+ partv

partx

)2
















S = 385 psu)





or 4 days)





a b

c



























a

b






a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References


a

c

b

















νH = α xy

radic(partupartx

)2

+(

partv

party

)2

+ 1

2

(partuparty

+ partv

partx

)2
















S = 385 psu)





or 4 days)





a b

c



























a

b






a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References










νH = α xy

radic(partupartx

)2

+(

partv

party

)2

+ 1

2

(partuparty

+ partv

partx

)2
















S = 385 psu)





or 4 days)





a b

c



























a

b






a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References












S = 385 psu)





or 4 days)





a b

c



























a

b






a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References





a b

c



























a

b






a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References

























a

b






a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References


a b

c d


























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References
























5 Conclusions













References




























































































Abstract

Introduction




Wind analysis





Conclusions

References





References




























































































Abstract

Introduction




Wind analysis





Conclusions

References































































Abstract

Introduction




Wind analysis





Conclusions

References

generation mechanisms for mesoscale eddies in the gulf of...

Documents