Surface circulation in the Eastern Mediterranean using drifters ( 2005 – 2007 )

Surface circulation in the Eastern Mediterranean using drifters (2005–2007) R. Gerin, P.-M. Poulain, I. Taupier-Letage, C. Millot, S. Ben Ismail, and C. Sammari Istituto Nazionale di Oceanografia e di Geofisica Sperimentale (OGS) – Borgo Grotta Gigante, 42/c-34010 Sgonico (Trieste), Italy Université de la Méditerranée, CNRS UMR 6535, Laboratoire d’Océanographie et de Biogéochimie (LOB), Antenne de Toulon, BP 330, 83507 La Seyne, France Institut National des Sciences et Technologie de la Mer (INSTM) – 28, rue du 2 mars 1934 2025 Salammbô, Tunisia Received: 2 February 2009 – Accepted: 15 February 2009 – Published: 6 March 2009 Correspondence to: R. Gerin (rgerin@inogs.it) Published by Copernicus Publications on behalf of the European Geosciences Union.


Introduction
The surface circulation in the eastern basin of the Mediterranean Sea was described for the first time at the beginning of the twentieth century by Nielsen (1912) who used hydrographic data from the Danish cruise onboard R/V Thor (1908)(1909)(1910) and hypothesised that a major role was played by the Coriolis effect.In Nielsen's representation, the circulation of the lighter Atlantic water (AW) is counterclockwise across the whole basin.Other descriptions of the surface circulation, inferred from geostrophic computations of more substantial hydrographic datasets, were proposed subsequently.Ovchinnikov (1966) suggested a circulation displaying several closed circuits in both the Southern Ionian and the area between 20 • E and 36 • E that, in the following, will be denoted as the Levantine sub-basin, and noted some seasonal variability in the circulation patterns.Lacombe and Tchernia (1972) described a flow of AW entering the Sicily Channel, crossing the Ionian North of two clockwise circuits and reaching Libya near 20 • E. From there, they represented the flow spreading toward the open sea, so that near 25 • E it extends from the Libyo-Egyptian coast as far as south of Crete.The AW counterclockwise circuit is depicted continuously at basin scale.Additionally, all these historical diagrams show a counterclockwise circulation in the Northern Ionian.
Then, studies performed in 1985-1992 during the POEM (Physical Oceanography of the Eastern Mediterranean) international program proposed another description of the circulation in the basin.Robinson et al. (1991) suggested a schema in which the main feature in the Levantine sub-basin is a current, named the Mid Mediterranean Jet (MMJ), crossing northeastward from ∼24 • E, to southeast of Cyprus.This MMJ bifurcates to delineate circulation features such as "Mersa Matruh" and "Shikmona".There is no evidence of AW flow along Africa.This schema was completed later (Robinson and Golnaraghi, 1993;Malanotte-Rizzoli et al., 1997)  to 40 • N before turning southward and joining the central branch offshore Libya near 20 • E, displaying a new circulation feature: a clockwise circuit in the northern part of the Ionian.This northern branch, also observed by Mauerhan (2000) using drifter data in the period 1986-1999 and by Lermusiaux and Robinson (2001) during the summer of 1996, can be attributed to interannual variability or decadal variability (Pinardi et al., 1997;Hamad et al., 2005;Millot and Taupier-Letage, 2005).Other studies (Pinardi and Navarra, 1993) reveal the importance of the wind stress which influences the circulation at seasonal time scale.
In recent years, the circulation of the basin was studied using long time series of satellite infrared images and reconsidering all available in situ observations.A new schema was proposed (Hamad et al., 2005(Hamad et al., , 2006;;Millot and Taupier-Letage, 2005).This schema depicts a general circulation counterclockwise all around the basin (as done by Nielsen) that is mainly constrained along the upper part of the continental slope, hence forming a basin-wide alongslope gyre.In the southern part, the flow, called the Libyo-Egyptian current, is unstable and generates anticyclonic eddies evolving in time and space.These eddies have diameters of 100-150 km, lifetimes of several months (up to more than one year), and propagate downstream along the slope at 1-3 km/d.They perturb the alongslope flow, which can result in dispatching AW toward the open sea."Mersa Matruh" and "Shikmona" are described as areas where eddies (generated upstream) tend to accumulate and/or merge.Finally, the anticyclonic eddies induced by the Northerly Etesian winds in the Ierapetra and Pelops areas have been shown to propagate, have lifetimes exceeding one year and therefore possibly co-exist with the newly-generated features.The cross-basin MMJ feature, described in the POEM diagram, is thus interpreted as the part of AW flow that is carried around the northern edges of the (successive) anticyclonic eddies generated either by the basinwide alongslope gyre in the southern part of the basin or by the wind in its northern part.
Recent works based on numerical simulations show the existence of an alongslope basin-scale circulation with no definitive evidence of the MMJ (Alhammoud et al., 2005)  or display the MMJ close to the African coast (Korres and Lascaratos, 2003), hence making it part of the above-described basin-wide alongslope gyre.The alongslope circulation is supported also by satellite altimetric data (Rio et al., 2007) and by in situ XBT data (Fusco et al., 2003).The lack of in situ observations in the southernmost parts of the basin (see e.g.Fig. 2 of Manca et al., 2004) was one of the reasons for the discrepancies of the AW circulation schemas in that area.et al., 2007;Beranger et al., 2007).The main aim of the EGITTO/EGYPT program was to study the surface circulation using surface drifters, with particular focus on the southern parts of the basin.
In the following, all the predominant circulation features with near-circular motion at scales up to 100-150 km are called eddies, whether they are generated by instability of the alongslope circulation, or forced by the wind or controlled by bathymetry.Following Robinson et al. (1991), the structures with scale of 50-150 km are called sub-basin eddies and the dynamical structures with size of 5-15 km are called mesoscale eddies.For instance, mesoscale structures correspond to instability features of the sub-basin scale eddies.This definition for the mesoscale in the Mediterranean is compatible with the values of the first internal Rossby deformation radius obtained by Grilli and Pinardi (1998).
In this paper, we present the results of the investigation of the spatial and temporal variability of the surface circulation in the basin obtained using 97 satellite-tracked drifters.The drifters were released in the Sicily Channel and in areas of specific interest in the Levantine sub-basin, from September 2005 to March 2007.The drifters Introduction

Conclusions References
Tables Figures

Back Close
Full and their deployment strategy are described in Sect.2, together with informations on the database and the methodology used to process the data.Pseudo-Eulerian maps of mean flow, eddy variability and the energy levels are presented in Sect.3. First the total dataset is considered and then it is divided in two extended seasons to explore the seasonal variability.The results are discussed and summarized in Sects.4 and 5, respectively.
2 Data and methods

Drifter characteristics
The drifters used in this study are the mini WOCE-SVP (model CLEARSat-15) manufactured by Clearwater Instrumentation.For details on the WOCE-SVP design, see Sybrandy and Niiler (1991) and Lumpkin and Pazos (2006).The surface buoy is tethered to a cylindrical drogue (with a diameter of about 60 cm and 5-m long), centred at a nominal depth of 15 m, that holds the drifter almost motionless with respect to the superficial layer.The buoy is equipped with a sea surface temperature sensor and a tension sensor that allows checking the presence of the drogue.All drifters were tracked with the Argos Data Location and Collection System carried by the NOAA and METOP polar orbiting satellites.This system provided about 9 locations/d with accuracy better than 1000 m (Argos User's Manual, 2008).

Strategy of deployments
A total of 97 drifters were deployed between September 2005 and March 2007 in the basin (Fig. 1 and Table 1).In the Sicily Channel and in the Ionian a seasonal variability was expected.As a result, the deployments were organized on a seasonal basis (approximately every 3 months).Five to six drifters were released across a transect at the entrance of the channel, both from research vessels (R/V Hannibal and R/V

Data and processing
Among the 97 drifters deployed, 5 stopped emitting immediately after deployment and were lost, while 8 were recovered and redeployed (hence being considered as new drifters).We thus obtained 100 individual trajectories that provide a database extending from 5 September 2005 to 31 October 2007 and covering the southern half of the basin (Fig. 2).
Drifter raw data were edited for outliers and spikes using automatic and manual techniques based on maximum distance, maximum speed and maximum angle between two consecutive points criteria, as described in Poulain et al. (2004).Edited positions were interpolated at 2-h intervals with a kriging optimal interpolation schema (Hansen and Poulain, 1996).The interpolated positions were then low-pass filtered using a Hamming filter with cut-off period at 36-h to eliminate high frequency current components (tidal and inertial currents) and were finally sub-sampled at 6-h intervals.Velocity components were then estimated from centred finite differences of the 6-h sub-sampled Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion longitude and latitude.About 40% of the observations were collected by drifters which had lost their drogues (see statistics in Table 2).In order to use the maximum number of observations for the computation, undrogued drifter data can be corrected using linear regression models of the differences between nearly collocated and co-temporal undrogued and drogued drifter velocities as a function of wind (Poulain et al., 1996(Poulain et al., , 2008;;Pazan and Niiler, 2001).In particular, in this paper, we performed such a correction using the formulation introduced by Poulain et al. (2008) deduced from the present EGITTO/EGYPT database.
Following the method detailed in Poulain (2001) and Emery and Thomson (2001), we computed the pseudo-Eulerian statistics to describe the near-surface circulation of the basin.Kriged sub-sampled data were grouped into 0.5 • ×0.5 • bins, organized on a grid with 0.25 • ×0.25 • mesh size to smooth spatially the results.Bins with less than 10 observations and containing data from less than two different drifters were rejected for the computation of the statistics.Trajectories outside the basin (i.e.north of Sicily) were not considered.Kinetic energy per unit of mass was computed and considered as the sum of two terms: the mean kinetic energy of the mean flow per unit of mass (MKE) and the mean kinetic energy of the fluctuations per unit of mass, also called eddy kinetic energy (EKE).The definitions of these statistics can be found in Poulain (2001).
Additionally, to investigate the seasonal variability, including the possible reversal of the mean circulation in some areas, the pseudo-Eulerian statistics were also split in two extended seasons: the winter from November to April and the summer from May to October.This separation captures most of the seasonal variability, as already shown by Poulain and Zambianchi (2007) in the Sicily Channel.A statistical analysis of temperature profiles (CTDs and XBTs) available in the NODC-OGS dataset (http: //nodc.ogs.trieste.it/)allowed checking that, in the basin, these extended winter and summer correspond to the most homogeneous and stratified conditions of the water column, respectively.Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion

Total dataset
The temporal distribution of the data (Fig. 3) reveals a classical sawtooth shape due to the limited operational lifetime of the drifters and to the successive deployments, with strong rises in correspondence to the major deployment episodes (e.g., the EGITTO-1 cruise in November 2005, with 19 drifters released).The maximum number of drifters operating simultaneously occurred on April 2006 with 37 units.The drifter lifetime in the basin is shorter than the nominal value (autonomy of the battery) because of the high probability of stranding or being caught by seafarers.The maximum lifetime is about one year and the mean half-life, that is the time after the deployment for which 50% of the drifters still provide useful data, is more than 100 days (see Table 2).The drifter trajectories were analysed during the first ten days after each deployment in the Sicily Channel (Fig. 4a-g), between 19 • E and 30 • E (Fig. 5) and in the easternmost part of the Levantine, south of Cyprus (Fig. 4h).Forty two drifters were released across the Sicily Channel during seven deployment episodes (Table 1 and Fig. 4ag).The majority of the drifters entered immediately the basin (34 over 42).However, north-westward motions to the Tyrrhenian were observed for 4 deployment episodes, especially in October 2006 (2 out of 6 drifters, Fig. 4e) and in January 2007 (3 out of 5 drifters, Fig. 4f).The initial trajectories of the drifters in the channel over ten days are highly variable.They can display loops that are either clockwise, mainly in its northern part or counterclockwise, mainly in its southern part.East of 20 • E, the striking features of the first 10-day trajectory segments (Fig. 5a-d eddies (south of EE on Fig. 5a, between 23 • E and 25 • E on Fig. 5d) and meandering structures.The trajectories of the drifters released in the southernmost region, when not in an eddy close to the coast, show an eastward current along the upper part of the slope, identified by the 200 m isobath (see at ∼22 • E and ∼28 • E on Fig. 5a, at ∼21 • E and between 25-26 • E on Fig. 5b, between 24-25 • E on Fig. 5d).Additionally, interactions between the eastward slope current and eddies are well evidenced by the trajectory of the southernmost drifter on ∼28 • 30 E in Fig. 5a, that of the drifter bending sharply offshore near ∼27 • E in Fig. 5b and that of the drifter veering eastward inshore at ∼24 • E in Fig. 5d.
Finally, the six drifters released south of Cyprus (Fig. 4h) evidence most of the anticyclonic eddy that sits over the Eratosthenes seamount (ESE).Three drifters converge at ∼33 • N and then move eastward due to the interaction between two circulation features located southeast of ESE as evidenced by means of satellite thermal images (not shown).
In order to delineate the areas characterized by strong currents we plotted the drifter trajectories corresponding to low-pass filtered velocities larger than 30 cm/s (Fig. 6).The drifter trajectory composite map (Fig. 2) and the distribution of the number of observations per bins (Fig. 7a) reveal that the dataset is widespread.In the Sicily Channel and in the Levantine sub-basin between 20 • E and 30 • E, the number of ob-Introduction

Conclusions References
Tables Figures

Back Close
Full • E and 19 • E, the current is southward and reaches the Libyan coast.The flow in the Southern Ionian is not well defined, especially in its southeasternmost part, but between 14 • E and 16 • E where there are more than 100 observations in bins (Fig. 7a).
There is no clear continuity of the flow from the Ionian to the area off Cyrenaica at 19-20  The variance ellipses superimposed on the EKE (Fig. 7c) show that along the continental slope, at basin-scale, the ellipses are generally flattened with the principal axis parallel to the slope.In the open sea, in the Levantine, the ellipses are more isotropic, indicating that the fluctuating currents occur in all the directions.At ∼25 • E, the principal axis of the ellipses is oriented N-S as far as Crete, as well, but to a lesser extent, along 31 • E. The EKE is maximum in correspondence to the Pelops eddy (PE), LE1, LE2, IE and EE eddies as well as along the slope between 20 • E and 30 • E.

Seasonal variability
The low-pass filtered trajectories (Figs.8a and 9a) divided by extended seasons reveal that the geographical coverage is broader in winter.However, the Sicily Channel, the central Ionian and the Levantine sub-basin between 20 and 30 • E are well covered in both seasons.The Northern Ionian is mainly covered in summer, while the easternmost Levantine is mainly depicted during winter.
The data densities, in the well-covered areas of Fig. 7a, are similar for both seasons (not shown), but it is important to note that the extended summer results are mainly due to the drifters operated during the summer 2006 (see the temporal distribution of the data, Fig. 3).For both extended seasons, the number of observations per bin is maximum (about 200 observations) in the Sicily Channel (due to the seasonal deployments), off Cyrenaica (Libya, between ∼20 • E and 23 • E) and in the main anticyclonic eddies in which drifters remained entrapped and re-circulated several times.The IE is well delineated in the low-pass filtered drifter tracks during the extended summer (Fig. 9a), since drifters were released inside it.Its signature is less intense during winter since drifters were entrained only around it (Fig. 8a).LE1 was tracked during all the year 2006 with thermal satellite images and it was checked that it moved westward, along the slope, between 20 • E and 24 • E at least (Taupier-Letage, 2008).Its evolution can be deduced from the drifter trajectories, especially in summer (Fig. 9a).EE is localized at the offshore end of the Herodotus trough and during summer at least two other eddies co-existed between 27 • E and 30 both seasons.During winter (Fig. 8a), one drifter delineates the PE anticyclone near 37 • N, 21 • E and another one identifies the counterclockwise circuit south of Turkey.
The mean flow during winter (Fig. 8b) shows that the flow entering the Sicily Channel is oriented toward SE.Three veins can be recognized.The northern one remains parallel to the island coast and enters the central Ionian between Sicily and Malta.The central vein meanders more southward until it joins the northern one (at ∼35 • N and ∼16 • E) and then reaches Libya at 19-20 • E. The third vein flows along the Tunisian shelf, but its description is limited to the entrance of the Sicily Channel.In the Southern Ionian the spatial coverage is too scarce to infer a general pattern.However, an eastward current is suggested alongslope and the cross-shore trajectories could be related to circulation clockwise loops, between 14 • E and 17 • E and between 16 • E and 20 • E. Farther to the East, the alongslope current appears clearly from ∼23 • E all the way to the Middle East regions.The signatures of all the eddies identified in Fig. 2 and another anticyclonic eddies (south of IE, centred at ∼25 • E) are clearly visible.In the Levantine sub-basin, the winter circulation computed is similar to the total average (Fig. 7b).
As previously said, in summer drifters do not cover the eastermost part of the basin: data are scarce east of 30 • E and only spotty.The mean flow pattern (Fig. 9b) in the Cretan Passage is quite similar to the one displayed during the winter (and to the total average).However, there are some local differences.There is a definite westward current alongslope between 20 • E and 23 • E, corresponding to the southern part of LE1.One must also note the southward flow south of IE and EE.The ESE and the LE can be identified as well as a cyclonic eddy southeast of ESE.In contrast, the flow is different in the Sicily Channel and in the Ionian.The main feature in the Sicily Channel is the northernmost vein that splits into two branches when entering the Ionian, south of Sicily.One branch goes to the North, reaching 39 • N, before veering to the South.
The main part of the other branch generates a large clockwise loop in the Southern Ionian (between 14 • E and 17 • E) extending to the Tunisian slope and the remaining part continues to the East and meets the former branch in the Central Ionian.The flow near 33 • N, between 20 • E-23 and seems to continue westward as far as 12 • E. The second clockwise circuit in the Southern Ionian is smaller than in winter and confined more to the slope.The westward propagation of LE1 elongates zonally its signature from ∼21 • E to ∼24 • E.
In both extended seasons, MKE (Figs. 8b and 9b) is maximum in correspondence to the circulation structures and areas already evidenced in Fig. 7b.The area around Sicily shows a marked increase of MKE in summer with values double than in winter.The variance ellipses and the EKE were calculated for the two extended seasons (Figs.8c and 9c).A seasonal variability is evident only off Sicily, elsewhere not much difference from the total pool is evidenced.

Discussion
Surface drifters were deployed in the eastern basin of the Mediterranean Sea from September 2005 to March 2007, in the framework of the EGITTO/EGYPT program, to describe its surface circulation patterns, and in particular to find out how the AW flows in the basin.The drifters, which were deployed in the Sicily Channel and in the Levantine sub-basin, provide a reasonably good coverage in the southern regions (Fig. 2) poorly sampled in the past.
The number of observations in the 0.5 • ×0.5 • bins can exceed 600 due to some drifters trapped inside or on the edges of eddies.The Southern Ionian is sampled mainly in an area oriented north-east and extending from Libya at 15 • E to the central Ionian.Especially during summer, some drifters moved to the Southern Ionian between 15 • E and 20 • E, but they are too few and the related statistics do not represent a reliable mean flow in this area.On the other hand, the Levantine sub-basin between 20 • E and 30 • E is well covered during all the period of study.
The main feature of the surface circulation in the Ionian is an eastward flow of AW entering the Sicily Channel and crossing the Ionian in its central part (Fig. 7a).It can be compared to the mean circulation already described by Malanotte-Rizzoli et al. (1997).In summer, this flow is stronger (compare Fig. 8b and 9b) and, east of

Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version

Interactive Discussion
Sicily, splits in two branches, one of which is northward as far as 39 • N and then turns clockwise and proceeds to the South.This clockwise elongated circuit seems to be related to the wind forcing that influences the circulation at seasonal time scales as explained by Pinardi and Navarra (1993).It might be also related to interannual or decadal variability (see introduction); however, the limited temporal extension of the database do not allow to infer any links in these respects.In the southern part of the Sicily Channel, the near-surface flow is southeastward, following approximately the 200-m isobaths delineating the Tunisian slope, and corresponds to the Atlantic Tunisia Current (Poulain and Zambianchi, 2007) or to the South Tunisia vein (Millot and Taupier-Letage, 2005).
The trajectories characterized by velocities larger than 30 cm/s (Fig. 6) show an alongslope flow in the Southern Ionian, which is comforted by the variance ellipses flattened along the slope (Fig. 7c).However, the flow pattern is complex since there is one eastward current from ∼12 to ∼17 • E and one westward from ∼17.5 to 18.5 • E. The computed mean flow might suggest two clockwise loops (although the eastern one is not closed, Fig. 7b) that would be offshore enough to allow an eastward current alongslope.Part of this complexity can be explained by the existence of sub-basin eddies that can develop and interact with the eastward alongslope current, as shown by Hamad et al. (2006).A seasonal variability is also evidenced.In winter, there is only little evidence of the alongslope flow at ∼15 • E, while in summer, the flow is predominantly westward along the Libyan slope.In the western part (from Tunisia to ∼16 • E), this reversal is confirmed by Poulain and Zambianchi (2007) and presumably related to the forcing by the prevailing southeastern winds.Additionally, especially during summertime, the mean flow pattern shows a southward current up to the Libyan coasts along 19 • E. Its fate is not known due to the drifter stranding (see e.g.Fig. 9a) and the poor coverage there.Thus, in the computed mean flow maps there is no clear continuity of the eastward flow off Cyrenaica.In such a complex and variable area, more data are required to achieve a reliable circulation pattern.East of ∼20 • E, the main feature is an alongslope counterclockwise flow off Africa Introduction

Conclusions References
Tables Figures

Back Close
Full and Middle East.In this respect, there is an overall agreement with the historical diagrams (Nielsen, 1912;Ovchinnikov, 1966;Lacombe and Tchernia, 1972) and a specifically good agreement with the most recent one (e.g.Millot and Taupier-Letage, 2005).A strong flow is evident mostly in winter, east of 21 • E, along the Libyo-Egyptian slope, while in summer it appears weaker.While there is no evidence of this current east of 30 • E due to the lack of data during summertime, in winter this flow is shown continuing alongslope northward off the Middle East regions and one drifter also moves westward in the area south of Turkey as the Cilician Current (Robinson et al., 1991) or the generic Northern Current (Millot and Taupier-Letage, 2005), completing the nearsurface counterclockwise circulation in the basin.
All the circulation structures of the Levantine sub-basin, pointed out in the low pass filtered trajectories (Fig. 2), can be found again in the mean flow map (Fig. 7b).
The elongated flow pattern off Cyrenaica, visible in the mean flow maps (Figs.7b, 8b, 9b), results both from the averaging of the trajectories that went around the two successive anticyclonic eddies LE2 and LE1 (see Figs. 1, 2 and 5a-c) and from the ∼1 year-long westward propagation of LE1 (Taupier-Letage, 2008).During summer 2006, LE1 was very close to the Libyan slope, so that the westward current induced on its southern edge was stronger than the general eastward flow and consequently the AW circulation was reversed (Fig. 9b).
The AW can also flow eastward in the open sea, after its deflection from the Libyo-Egyptian current by aticyclonic eddies, it is then transported by the successive eddies, as in a paddle-wheel effect.Therefore, on the mean flow maps (Figs.(Figs.8a and 9a).

Conclusions
The drifter observations presented in this paper provide an improved or novel description of the surface circulation in most areas of the eastern basin of the Mediterranean Sea for the period September 2005-October 2007.However, we have to keep in mind that the drifters, being Lagrangian in nature, covered the basin non-uniformly both in space and in time, resulting in possible bias errors in the statistical results calculated from the drifter trajectories.Nevertheless, significant results were obtained, such as the branching of the AW flow in the Ionian, the existence of the eastward alongslope flow at least from 23 • E, continuing its counterclockwise circuit up to the Middle East, the preponderance of sub-basin anticyclonic eddies in the southern parts of the Levantine sub-basin and their role in deflecting AW offshore, which does result in an open sea eastward transport of AW.In addition, seasonal reversal of the circulation in the Southern Ionian (westward in summer and eastward in winter) is observed.It is hoped that Lagrangian observations in the Eastern Mediterranean will be continued as part of scientific and operational oceanography programs in order improve further our understanding of the circulation in this basin.Introduction

Conclusions References
Tables Figures

Back Close
Full to detail the circulation in the Ionian: the flow of AW, named the Atlantic Ionian Stream (AIS) in the Sicily Channel, is depicted meandering and bifurcating, east of Sicily, in the central and Northern Ionian.The central branch crosses the Ionian, while the northern branch extends up ) are the clockwise loops (especially in November 2005 and April 2006), due to the strategy of seeding drifters along transects in the anticyclonic eddies (main eddies: LE1, LE2, IE and EE).From November 2005 to April 2006 a westward propagation of LE1 is observable, as well as that of LE2 between February and April 2006.Both eddies are also closer to the coast.EE was seeded in November 2005, while IE was seeded in April 2006.The curves and sharp bends in trajectories indicate the presence of other 533 Maximum velocities are mainly recorded in the central part of the Sicily Channel and southeast of Sicily, and along the Libyan slope, in the Southern Ionian.Further to the East, they are observed along the Libyo-Egyptian slope.Additionally, they underline the importance of the eddy activity and the positions of the main eddies (LE1, LE2, IE and EE).Velocities larger than 30 cm/s correspond to the slope current east of ∼29 • E up to the Middle East regions and the Latakia eddy (LE), as well as an area between EE and ESE.However, care must be taken because, during the period of study, features have varied in both space and time.Eddies have propagated, have vanished, appeared or have merged as in the case of the IE created in 2005 that co-existed and rapidly merged with the IE created in 2006 (seeTaupier-Letage, 2008).
5 • ×0.5 • bins are larger than 300.The mean surface flow computed is depicted in Fig.7bsuperimposed on the MKE.The overall circulation is quite complex but it shows a current entering the Sicily Channel and flowing eastward mainly on the Sicilian side.It passes between Sicily and Malta and splits into two branches when entering the Ionian.One branch goes to the North, reaching 39 • N, before veering to the South.The other one crosses the Ionian generating a large clockwise loop centred at ∼15 • E, 34 • N and joins the previous branch in the central Ionian, at ∼19 • E. Along 12 there is no definitive flow along the slope, while an eastward flow meanders in the open sea along ∼34 • N.Both observations are related to the sub-basin eddies LE2 and LE1.They modified the flow along the slope according to their position from the coast (closer: westward current; further: eastward current) and diverted the flow toward the open sea.Additionally, the reversal of the current, combined with the westward propagation of the eddies, decrease the mean current computed alongslope.East of ∼23 • E there is an eastward flow along the slope, that continues in a counterclockwise circuit off Middle East regions and south of Turkey.At ∼24 • E, the meandering open sea part of the flow splits in two different paths.The first part joins the alongslope eastward flow mentioned above, either near 23 • E or between 25 • E and 27 • E. The second part is sharply deflected towards Crete, around IE, and then continue eastward driven, successively, by the northern edges of EE and ESE, up to reaching the Middle East slope and the LE.Large values of MKE (Fig. 7b) are prevailing in the Ionian, in correspondence to the flow crossing the Ionian, the northward branch north-east of Sicily and the circuit centred at ∼15 • E. Large values also occur offshore Libya, between 20 • E and 24 • E, along the slope east of 23 7b, 8b and 9b), a part of the flow appears to meanders from the northern edges of LE1 successively on those of IE, EE and ESE, up to LE.This flow can be identified as the MMJ introduced byRobinson et al. (1991).However, the north-south orientation of the ellipses at ∼25 • E, ∼27 • E and ∼31 • E (Figs. 7c, 8c and 9c) indicates that, at these longitudes, the eastward open sea flow was not direct, but rather resulted, for example, from the succession of a southeastward flow on the eastern side of LE1 and a northeastward flow on the western side of IE, as also displayed by the individual drifter trajectories

Fig. 1
Fig. 1 Eastern Mediterranean area (30°-39°N, 10°-37°E) with geographical names and sites of drifter deployment (coded per year).The area between 20°E and 36°E is denoted as the Levantine sub-basin.The 200 m, 1000 m, 2000 m and 3000 m isobaths are represented in grey tones.

Fig. 3 .
Fig. 3. Temporal distribution of the drifter data in the basin.Number of drifters per day from 5 September 2005 to 31 October 2007.The maximum number of drifters operating simultaneously occurred on 24 April 2006 (37 units).
Fig. 7 Statistics computed with the entire dataset.a) Data density (saturated at 300 observations per bins).b) Mean flow superimposed on the MKE (saturated at 250 cm 2 /s 2 ).The mean flow arrows are centred at the centre of mass of the observations in each bin.c) Variance ellipses superimposed on the EKE (saturated at 500 cm 2 /s 2 ).In all the figures, data are grouped into 0.5° x 0.5° bins overlapped by 50% and results for bins with less than 10 observations and 2 different drifters are not plotted.The 200 m, 1000 m, 2000 m and 3000 m isobaths are represented in grey tones.

Fig. 7 .
Fig. 7. Statistics computed with the entire dataset.(a) Data density (saturated at 300 observations per bins).(b) Mean flow superimposed on the MKE (saturated at 250 cm 2 /s 2 ).The mean flow arrows are centred at the centre of mass of the observations in each bin.(c) Variance ellipses superimposed on the EKE (saturated at 500 cm 2 /s 2 ).In all the figures, data are grouped into 0.5 • ×0.5 • bins overlapped by 50% and results for bins with less than 10 observations and 2 different drifters are not plotted.The 200 m, 1000 m, 2000 m and 3000 m isobaths are represented in grey tones.
Fig. 8 Statistics for the extended winter.a) Low pass filtered drifter data.b) Mean flow superimposed on the MKE (saturated at 250 cm 2 /s 2 ).The mean flow arrows are centred at the centre of mass of the observations in each bin.c) Variance ellipses superimposed on the EKE (saturated at 500 cm 2 /s 2 ).In figures b) and c), data are grouped into 0.5° x 0.5° bins overlapped by 50% and results for bins with less than 10 observations and 2 different drifters are not plotted.The 200 m, 1000 m, 2000 m and 3000 m isobaths are represented in grey tones.

Fig. 8 .
Fig. 8. Statistics for the extended winter.(a) Low pass filtered drifter data.(b) Mean flow superimposed on the MKE (saturated at 250 cm 2 /s 2 ).The mean flow arrows are centred at the centre of mass of the observations in each bin.(c) Variance ellipses superimposed on the EKE (saturated at 500 cm 2 /s 2 ).In (b) and (c), data are grouped into 0.5 • ×0.5 • bins overlapped by 50% and results for bins with less than 10 observations and 2 different drifters are not plotted.The 200 m, 1000 m, 2000 m and 3000 m isobaths are represented in grey tones.