Progress in
Oceanography
Progress in Oceanography 66 (2005) 211–230
www.elsevier.com/locate/pocean
Large scale flow separation and mesoscale eddy formation
in the Algerian Basin
P. Testor *, J.-C. Gascard
1
Laboratoire dÕOcéanographie Dynamique et de Climatologie, Univ. P. et M. Curie, 4 place Jussieu, 75252 Paris, France
Received 30 June 2002; received in revised form 14 April 2003; accepted 29 July 2004
Available online 2 June 2005
Abstract
During the ELISA/MATER experiment floats released at about 600 m depth in the Levantine Intermediate Water
layer south of Sardinia in July 1997 have revealed the existence of a coherent eddy, approximately 50 km in diameter
and lasting for several months. This anticyclonic eddy was first observed south-west of Sardinia in November 1997 and
drifted inside the Algerian Basin during the following months until April 1998. This eddy contained Levantine Intermediate Water at intermediate level and seemed to be related to 2 main large scale features: (a) a cyclonic gyre (250 km
in diameter and 3–4 months period) located in the Algerian Basin and (b) a boundary current located along the continental slope south and west of Sardinia and originating from the Sardinia–Tunisia channel. We will first describe the
‘‘Sardinian’’ eddy, from a kinematical point of view, and the Algerian Gyre and second, give some insights about the
eddy origin and its importance for LIW large scale spreading in the Western Mediterranean Sea.
2005 Elsevier Ltd. All rights reserved.
Keywords: Western Mediterranean Sea; Levantine intermediate water; Eddies
1. Introduction
The main information regarding the Algerian Basin concerns the circulation of the surface modified
Atlantic Water (AW) which has been intensively studied with satellite imagery, moored current meters (Millot, 1999; Puillat, Taupier-Letage, & Millot, 2002) and surface drifters (Font, Millot, Salas, Julia, & Chic,
*
Corresponding author. Present address: Leibniz-Institut für Meereswissenschaften (IFM-GEOMAR), Kiel, Germany. Tel.: +49
431 600 4154; fax: +49 431 600174151.
1
Present address: Laboratoire dÕOcénographie et de Climatologie: Expérimentation et Approache Numérique, ISPL, Univ. P. et M.
Curie, Paris, France.
E-mail addresses: [email protected], [email protected] (P. Testor).
0079-6611/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.pocean.2004.07.018
212
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
1998). The mean transport of AW associated with the Algerian current flowing along the north African continental slope, is 1.7 Sv (Benzhora & Millot, 1995). The current is unstable and generates large anticyclonic
Algerian Eddies (AEs) around 1–2E longitude. They are about 100 km in diameter and have a rotational
period of order of a week corresponding to a relative vorticity of f/10 and peak orbital surface velocities
of about 50 cm s1. AEs generally extend from the surface down to about 350 m depth which corresponds
to the maximum depth reached by AW, but they can episodically extend much deeper. They circulate cyclonically around the Algerian Basin at mean speed of few cm s1, detaching from the north African coast at
about 800 0 E. Usually, the average time for AEs to move around the Algerian Basin is about 1 year. Their
lifetime is usually greater than 1 year and can reach 3 years according to Puillat et al. (2002).
Deeper water masses circulation in the Algerian Basin and through the Channel of Sardinia–Tunisia are
less well-known. This channel is deep (1900 m) and wide (70 km at 500 m depth) and provides a unique
opportunity for intermediate and deep waters to communicate between the Algerian Basin and the Tyrrhenian Sea. At higher levels a main core of Levantine Intermediate Water (LIW), centered around 350 m
depth and characterized by high temperature and salinity maxima, flows westwards from the Tyrrhenian
Sea to the Algero-Provençal Basin along the northern side of the Channel of Sardinia–Tunisia. This flow
also concerns the Tyrrhenian Deep Waters (TDW) which are situated below between approximately 700
and 1500 m depth. These waters result from the LIW and the Western Mediterranean Deep Water
(WMDW) which would have mixed together in the Tyrrhenian Sea (at about a 2:1 rate) although these
mixing processes are not yet well documented (Hopkins, 1988, Rhein, Send, Klein, & Krahmann, 1999).
The WMDW flowing in the Channel of Sardinia–Tunisia and entering the Tyrrhenian Sea balances this
export of TDW (0.2 Sv according to Hopkins (1988)) in the Algerian Basin. Based on CFC measurements,
Rhein et al. (1999) observed a TDW outflow of 0.4 Sv between 600 and 1600–1900 m depth from the Tyrrhenian Sea through the Channel of Sardinia in the Algerian Basin.
On the southern side of the Channel of Sardinia–Tunisia and at intermediate levels, modified LIW follows the path of AW heading east (Benzhora & Millot, 1995) and enters the Tyrrhenian Sea where it mixes
with the LIW outflow coming from the Eastern Mediterranean Sea through the Channel of Sicily (Bouzinac, Font, & Millot, 1999; Sammari, Millot, Taupier-Letage, Stefani, & Brahim, 1999). Note, these deep
interbasin exchanges are well illustrated by CTD casts in the Channel of Sardinia–Tunisia (Figs. 2 and 3 in
Bouzinac et al. (1999)) showing very clearly the main core of LIW-TDW flowing westwards south of Sardinia and WMDW flowing eastwards north of Tunisia at deeper levels. Nevertheless, these deep water masses
exchanges have not been quantified since geostrophic calculations did not reveal any significant baroclinic
currents and the barotropic component is unknown.
South of Sardinia, the LIW vein flowing westwards is relatively narrow (50 km), reaching 800 m and
characterized by pronounced temperature–salinity maxima (13.9 C) at about 300 m depth. The LIW core
turns northwards when leaving the Channel of Sardinia–Tunisia following the continental slope west of
Sardinia (Katz, 1972; Millot, 1999; Perkins & Pistek, 1990; Wüst, 1961). West of Sardinia, the vein is wider
(120 km) and temperature maxima of 13.9 C are found locally between 250 and 350 m depth across a 10–
20 km narrow section (Millot, 1999). In the center of the Algerian Basin, LIW is generally significantly
modified with typical temperature maxima half a degree colder (around 13.35 C) due to increased mixing
with the surroundings. TDW evidenced by Rhein et al. (1999) and characterized by a minimum of CFC
concentration centered around 1000 m depth, fills the layer between LIW and WMDW throughout a large
part of the Algerian Basin. WMDW and TDW are the 2 main components of deep waters in the Algerian
Basin. CFC measurements allowed to identify TDW more efficiently than temperature-salinity characteristics alone, TDW being situated on the mixing line between WMDW and LIW on a h–S diagram. The lowest CFC concentrations (comparable with CFC values found in the Tyrrhenian Sea) characterizing TDW,
are found along the continental slope of Sardinia under the main core of LIW. According to Millot (1999)
and Rhein et al. (1999), TDW follows the general cyclonic boundary circulation around the Algero-Provençal Basin like LIW and mixes in the interior of the basin.
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
213
Patches of LIW have been observed in the interior of the Algerian Basin with temperature maxima
around 13.8 C (Fuda, Millot, Taupier-Letage, Send, & Bocognagno, 2000; Millot, 1999) giving evidence
to a direct transport of LIW from the LIW vein off Sardinian continental shelf and slope, towards the deep
basin. The mesoscale activity characterizing AW and evidenced with surface measurements, is able to slow
down and even to reverse the along slope circulation in the Channel of Sardinia–Tunisia according to Bouzinac et al. (1999). Patches of LIW observed in the Algerian Basin interior (Millot, 1999) may result also
from AEs entrapping patches of LIW from the LIW vein and advecting them towards the middle of the
basin. These patches have also been described as filaments swirling around AEs (Emelianov, Millot, Font,
& Taupier-Letage, 1999). Millot (1999) also considered that LIW vein could generate anticyclones (‘‘Leddies’’) resulting from (1) the vein instability (Afanasiev & Fillipov, 1996; Baey, Renouard, & Chabert dÕHières, 1995) (2) an increase in the vertical density gradient between the Tyrrhenian Sea and the Algerian Basin
that would reduce the LIW layer thickness and create negative vorticity in accordance with potential vorticity conservation and (3) the momentum conservation of a flow following a continental slope presenting
an angle to the right (Pichevin & Nof, 1995).
During MATER/LIWEX, Gascard et al. (1999) observed a deep cyclonic gyre circulation at 600 m depth
in the Algerian Basin (Fig. 1). This gyre type circulation is different from the well-known general cyclonic
alongslope circulation in the whole Algero-Provençal Basin since this cyclonic gyre also concerns waters in
the interior of the basin. This large scale steady circulation revealed by Lagrangian floats from July 1997 to
July 1998, has been defined as the Algerian Gyre. The Algerian Gyre extension is variable approximately
250 km in the East–West direction and 100–200 km in the South–North direction, occupying a large part of
the Algerian Basin between the Balearic Islands, Sardinia and North Africa. This gyre is characterized by
Fig. 1. Trajectories of floats #09, #11, #14, #39 and #88 from 31 September 1997 to 26 December 1997 color-coded according to in
situ potential temperature. Locations of moorings (crosses) equipped with currentmeters at 100, 350, 1000 and 1800 m depth.
214
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
horizontal velocities of 5–10 cm s1 as indicated by floats drifting at its periphery and completing a full loop
in 3–4 months. This is 3–4 times faster than the time needed by AEs (about 1 year) to circulate around the
Algerian Basin (Puillat et al., 2002). During the same experiment, an eddy characterized by a strong LIW
signature was also revealed by Lagrangian floats at the periphery of the Algerian Gyre and was first presented by Testor and Gascard (2003) as a LIW eddy. In this paper, we will mainly present a kinematical
analysis of this eddy, based on data obtained during MATER/LIWEX experiment and will propose new
insights about the origin of this eddy and its importance for LIW circulation in the Western Mediterranean
Sea. We conclude that this eddy, hereafter named ‘‘Sardinian Eddy’’, has a distinct origin from the Leddies
hypothesized by Millot (1999) and are also quite different from Meddies observed in the North Atlantic
(Richardson, Walsh, Armi, Schröder, & Price, 1989).
2. MATER experiment general presentation
During the MATER experiment, 15 subsurface Lagrangian floats have been located acoustically from
July 1997 until July 1998 in the Algerian Basin. These so-called RAFOS floats are isobaric like Swallow
floats (Swallow, 1955) and drift according to horizontal currents at a predetermined and quasi-constant
depth. They are localized by acoustic triangulation using an array of moored acoustic sources transmitting long range propagating signals. Floats receive signals transmitted by acoustic sources every 4 h. The
mean absolute accuracy for floats positioning is about 2 km but relative precision is better than 300 m.
They also measure in situ temperature and pressure every 4 h. At the end of their mission they ascend to
the surface and start transmitting data to satellites (NOAA/ARGOS). Twelve floats were recovered during LIWEX in July 1998 and deep CTD casts were taken at the locations where floats had been
recovered.
Floats were ballasted to drift in the intermediate layer to document the subsurface general circulation of
this basin and to complement other observations of ocean circulation carried out during the MATER/ELISA operation (MAST3 – EU final scientific report, 1999). During MATER/ELISA 1 cruise in July 1997, 10
floats had been launched in the center of the Algerian Basin and 5 floats had been launched south of Sardinia in the Levantine vein; the 15 drifting floats remained in the water at about 600 m depth for one full year
(until July 1998).
3. Description of large and mesoscale features revealed by the MATER Lagrangian float experiment
We will focus on the autumn 1997–spring 1998 period during which floats highlighted an eddy-like flow
field of marked LIW that we named Sardinian Eddy and also a gyre-like circulation that we named Algerian Gyre.
3.1. Deep cyclonic circulation in the Algerian Basin: the Algerian Gyre
Floats #09 and #11 launched in the center of the Algerian Basin in July 1997 clearly show the large cyclonic gyre deep circulation in the Algerian Basin (Fig. 1), the so-called Algerian Gyre (Gascard et al.,
1999). They drifted in the Gyre throughout the whole experiment and measured velocities of 5.6 ± 2.2
and 5.1 ± 1.8 cm s1, respectively. The Algerian Gyre was very energetic between November 1997 and January 1998 with velocities up to 8 cm s1 at the periphery (float #09, Fig. 2). Floats drifting in the gyre measured potential temperatures around 13.1 C at about 600 m depth in contrast to the high temperatures
recorded in the Sardinian Vein west of Sardinia or in the Sardinian Eddy at the periphery of the gyre by
other floats.
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
#14
#09
|U| (m.s )
0.1
0.05
0
13.6
dd/mm/yy
13.4
13.2
13
14/07/97 13/08/97 12/09/97 12/10/97 11/11/97 11/12/97 10/01/98 09/02/98 11/03/98
0
13.6
dd/mm/yy
13.4
13.2
13
12/10/97 11/11/97 11/12/97 10/01/98 09/02/98
dd/mm/yy
dd/mm/yy
#39
#11
0.1
|U| (m.s )
0.1
0.05
0
13.6
dd/mm/yy
0.05
0
13.6
dd/mm/yy
°
pot.temp. ( C)
|U| (m.s )
0.05
°
pot. temp. ( C)
pot. temp. (°C)
|U| (m.s )
0.1
pot.temp. (°C)
215
13.4
13.2
13
14/07/97 13/08/97 12/09/97 12/10/97 11/11/97 11/12/97 10/01/98 09/02/98 11/03/98
13.4
13.2
13
12/10/97 11/11/97 11/12/97 10/01/98 09/02/98
dd/mm/yy
dd/mm/yy
#88
|U| (m.s )
0.1
0.05
pot. temp.
0
13.6
dd/mm/yy
13.4
13.2
13
14/07/97 13/08/97 12/09/97 12/10/97 11/11/97 11/12/97 10/01/98
dd/mm/yy
Fig. 2. (Full line) Time series of velocities and potential temperatures measured by floats #09 and #11 drifting in the Algerian Gyre; by
float #88 drifting in the Sardinian Vein; by floats #39 and #14 drifting in the Sardinian Eddy. (Dotted line) Time series of float #39
drifting in another eddy (SCV-S1, Testor and Gascard, 2003).
The Algerian Gyre was revealed by other floats (Fig. 3) and has a strong seasonal variability: it is stronger in winter (U 10 cm s1) than in summer (U 5 cm s1). Fig. 3 also shows f/H contours which are
closed in the Algerian Basin and which coincide with the structure of the gyre as revealed by floats, both
in summer and winter. This configuration is in favor of a barotropic gyre circulation.
216
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
Fig. 3. Trajectories of all LIWEX floats color-coded according to potential temperature (top) between 11 November 1997 and 25
January 1998 ; (bottom) between 25 January 1998 and 24 June 1998. Isocontours f/H normalized by f0 at 3830 0 N and H0 = 2830m.
3.2. Flow separation south-west of Sardinia
CTD casts 048–057 undertaken during the ELISA experiment (Fig. 4) document the structure of the
fluid flowing northward along the western slope of Sardinia and in the Algerian Gyre. South-west of Sardinia geostrophic calculations referenced to 1000 m depth on the ELISA CTD casts 048–057 are presented in
Fig. 4. They do not indicate any current greater than 0.5 cm s1 below 300 m depth (AW layer). In contrast,
south-west of Sardinia, velocities at 600 m depth are larger (order of 5 cm s1) which implies that currents
are mainly constant along the vertical below the AW layer.
Velocities at float #88, reached a peak value of about 10 cm s1 (Fig. 2) when it was the closest to the
2000 m isobath. At this time, float #09 velocities (Fig. 2) drifting nearby float #88, are about the same as
float #88 velocities. There is a strong temperature gradient between these two floats (Fig. 4). Being less than
10 km away from each other, float #88 measured a temperature 0.4 C greater than float #09, indicating
that float #09 was in the Algerian Gyre whereas float #88 was in LIW that had recently come from the
Tyrrhenian Sea (Fig. 1). Around 3900 0 N, 730 0 E the two float paths separated; float #88 followed the
continental slope, and float #09 remained in the cyclonic Algerian Gyre, detaching from the continental
slope.
South-west of Sardinia, the LIW core is characterized by anticyclonic horizontal shear between the Algerian gyre and the continental slope (Fig. 5). This shear is evaluated to be about 0.9 to 0.4 · 105 s1.
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
048
049
050
051
052
053
18
055
.6
13
13.3
13.4
mLIW
054
056
057
0
908
909
910
21
14.9
MAW14.7
14.1
MAW
5
13.
.7
13
23 21
20 1622
.7 14.3
111445.9
.5
.1 1313.9
14
.6
.3
.7
.4
13.153 13 13
13.3
19
17
13.5
13.7
0
217
13.9
13.7
13.6
.7
13
LIW
LIW
13.6
13.2
13.1
13
Algerian Gyre
Sardinian
13.4
13.3
13.2
13.1
Vein
TDW
mTDW
TDW
depth (dbar)
depth (dbar)
13.3
12.
95
WMDW
0
20
40
km
60
80
0
10
20
km
30
Fig. 4. Sections of potential temperature undertaken in July 1997 (048–057) and in July 1998 (908–910) across the Sardinian Vein. See
Fig. 5 for the locations of ELISA and LIWEX CTD casts.
Note that the assumption of homogeneity for alongslope projection is justified as along slope velocities
measured by float #88 are about the same before and after the float detected the maximum velocity allowing us to define the new referential.
In this region, the LIW circulating along the continental slope south and west of Sardinia, is confined
between the energetic cyclonic Algerian Gyre and the continental slope (Fig. 4).
3.3. LIW Sardinian vein west of Sardinia
Float #88 at 600 m depth remained in marked LIW until 28 January 1998 indicating deep currents flowing along the slope of Sardinia. During all its journey, it measured potential temperatures of 13.5–13.6 C at
600 m depth (Fig. 2) and closely followed the 1000 m isobath (Fig. 1). Velocities measured by this float were
2.8 ± 2.4 cm s1.
The waters coming directly from the Tyrrhenian sea have their TS characteristics modified while flowing
northwards along the slope as indicated by CTD casts 908–910 (Fig. 4). This shows, as already suggested by
Millot (1999), that some dramatic change in the structure of the alongslope flow occurs south-west of
Sardinia.
Geostrophic calculations with reference at 1000 m depth on the LIWEX CTD casts 908–910 presented in
Fig. 4 do not indicate any current greater than 0.5 cm s1 below 300 m depth either. Accordingly, the Sardinian Vein is composed of waters coming directly from the Tyrrhenian Sea mainly confined between 1000
and 2000 m isobaths. It flows northwards west of Sardinia and seems to be a sluggish circulation characterized by no vertical shear below the surface layer of AW.
218
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
Fig. 5. Top: components of velocities measured by floats #09 and #14 in a new referential (X,Y) between 15 and 16 November 1997
(black) and between the 16 and 18 November 1997 (red). This is before and after float #88 measured a maximum in velocity which was
tangent to the slope. The position where this maximum has been measured is taken as the origin of the new axes. The new directions
(X,Y) are respectively orthogonal to the slope and along slope (the slope ‘‘seen’’ by floats is almost constant during that period). The
alongslope (full) and orthogonal to the slope (dotted) components of velocities are plotted along the X-axis in this new referential.
Bottom: trajectories of floats #09 and #14 between 15 and 18 November 1997. Along slope component plotted on the X-axis of the new
referential. The dashed-dotted box is oriented according to the new referential and shows approximately the region considered as
homogeneous along the Y-direction and during this period. It shows the shear in the boundary region between the continental slope
and the Algerian Gyre. Position of CTD casts shown in Fig. 4. As a comparison, see Fig. 8 in Rhein et al. (1999). The blue line shows
the positions of the section of their LADCP, CTD and CFC measurements.
3.4. Sardinian Eddy
In November 1997, two floats (#39 at 700 m depth and #14 at 600 m depth launched south of Sardinia in
July 1997) began to describe anticyclonic loops of 3–4 weeks period, indicating that they were both drifting
in the same mesoscale eddy. Both floats measured very high temperatures (13.6 C at 600 m depth and
13.4 C at 700 m depth) typical of LIW off the coast of Sardinia.
Float #39 left the Sardinian Eddy on 26 December 1997 and float #14 remained within the eddy until
April 1998, indicating an eddy life-time greater than 5 months. Since low-pass filters applied to floats fixes
allow to have an estimation of the vortex center locations, the relative positions of floats to the estimated
center of rotation of the Sardinian Eddy were computed as well as the corresponding orbital velocities considering this eddy was axisymmetric. Thus, a radial distribution of orbital velocities (Fig. 6) could be estimated. This gives a typical Sardinian Eddy radius length scale of about 25–30 km (defined as the distance of
maximum orbital velocity) confirmed by an upper bound of the eddy radius according to a float passing
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
219
nearby at the same depth (#09 at 40 km from the center, Fig. 1) and not influenced by the vortex rotations.
The eddy core was in quasi solid body rotation and had a relative vorticity f f/16 leading to a Rossby
number of 0.06.
Three phases can be identified from floats trajectories (Fig. 7):
Phase 1:First, floats #14 and #39 drifted in the LIW core south-west of Sardinia (Figs. 7 and 2). Both
measured high temperatures and low velocities typical of LIW off Sardinia. Temperatures decreased
slowly until the beginning of November, indicating they were drifting in a slightly modified LIW while
velocities increased. Floats #11 and #09 were embedded in the Algerian Gyre and measured temperatures and velocities (Fig. 2), in contrast to floats #14 and #39.
Phase 2:At the beginning of November, floats #39 (around 700 m depth) and #14 (around 600 m depth)
detached from the continental slope and began to describe anticyclonic loops of about 3–4 weeks period
while still drifting in warm waters typical of LIW. At this time, they were clearly trapped in the Sardinian
Eddy. The two floats drifted almost diametrically opposed and at about the same distance from the eddy
Sardinian eddy
16
orbital velocity (cm/s)
14
12
10
8
6
4
2
0
0
5
10
15
20
25
30
35
radial distance (km)
30
20
10
0
0
10
20
30
40
50
Fig. 6. (Bottom) Radial structure of the orbital velocities of the Sardinian Eddy inferred from an estimation of the eddy center of
rotation as a function of time. (Top) Relative positions of floats #39 (·) and #14 (+) to the center of the Sardinian Eddy.
220
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
x (km)
(a)
0
50
x (km)
100
150
200
250
(b)
0
50
100
150
200
250
0
40.5
#88
0
2000
00
38.5
#11
y (km)
#11
#88
#09
#39
#14
39
#39
#14
20
latitude N
2000
0
39.5
1000
1000
1000
40
#09
20
00
38
37.5
2000
1000
37
2000 1000
5
(c)
6
7
8
9
longitude E
40.5
0
#88
100
40
#14
39
100 km
38.5
#09
#39
y (km)
latitude N
2000
0
39.5
20
00
38
#11
37.5
2000
1000
37
5
6
7
8
9
longitude E
Fig. 7. Trajectories of floats #09, #11, #14, #39 and #88 between (a) 27 September 1997 and 8 November 1997, (b) 8 November 1997
and 26 December 1997, (c) 26 December 1997 and 1 April 1998. Full line is used for floats drifting in the LIW flow south-west of
Sardinia, in the Sardinian Vein or in the Sardinian Eddy, dashed line for floats drifting in more modified LIW.
center for 2 complete eddy revolutions. During this period, the Sardinian Eddy translated at a velocity of
about 3 cm s1 at the periphery of the Algerian Gyre as shown by time series of floats velocities indicating oscillations of 3–4 weeks periods and amplitudes about 5 cm s1, centered around 3 cm s1 (Fig. 2).
Temperatures decreased significantly when the eddy detached from the Sardinian Vein, indicating an
overall response to a colder environment.
Phase 3:At the end of December, the Sardinian Eddy remained on a steady position north of the Algerian Gyre for several weeks. At the same time, float #39 drifted away from the Sardinian Eddy center
while float #14 went closer to the center. Float #14 remained in the eddy until mid April 1998 while float
#39 moved away from it and began to drift without transition in a very distinct, much smaller and coherent vortex advecting newly formed WMDW (SCV-S1, in Testor & Gascard (2003)). At the end of
March, the Sardinian Eddy started to propagate southwards with advection speeds about 3 cm s1
and, finally float #14 left the Sardinian Eddy around mid-April.
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
221
4. Differential kinematic parameters
4.1. Estimation of differential kinematic parameters
A method based on the works of Molinari and Kirwan (1975) and Sanderson (1995) allows to analyse
float trajectories in terms of differential kinematic parameters (DKPs)-relative vorticity (vor), divergence
(div), shear (shr) and stretch (str)- for a water parcel defined by a cluster of floats, by computation of horizontal velocity gradients (Eq. (1)). For a water parcel with velocities ðuðx; yÞ; vðx; yÞÞ,
g11 ¼
o
u
;
ox
g12 ¼
o
u
;
oy
g21 ¼
ov
;
ox
g22 ¼
ov
;
oy
ð1Þ
and vor = g21 g12; div = g11 + g22; shr = g21 + g12; str = g11 g22.
A cluster of N floats drifting close to each other, is taken as a water parcel in which we assume that horizontal speed gradients are homogeneous. Then, these gradients can be estimated by inverting (using least
squares method) the system (Eq. (2)) in which we express the velocity of each float in a Taylor expansion
about the center of mass of the cluster.
ui ¼ U þ g11 ðxi X Þ þ g12 ðy i Y Þ þ u0i ;
ð2Þ
vi ¼ V þ g21 ðxi X Þ þ g22 ðy i Y Þ þ v0i ;
where (xi,yi) are the floats positions, (ui,vi) are the floats velocities, (X,Y) is the position of the mass center of
the cluster of floats (xi,yi), (U,V) is the velocity of the center (X,Y) (mean velocity of the parcel) and ðu0i ; v0i Þ
are residual velocities.
Assuming that velocity gradients
Plinked
P to the water column are constant during T time steps and defining a cluster centre as ðX ; Y Þ ¼ NT1 Ni¼1 Tk¼1 ðxi ðkÞ; y i ðkÞÞ, we can increase the number of equations used for
the linear trend (Eqs. (3) and (4)). We can rewrite all these equations using matrices:
U ¼ RA þ E;
with
0
ð3Þ
u1 ðt1 Þ U
B
..
B
B
.
B
B
B uN ðt1 Þ U
B
B
..
B
B
.
B
U ¼B
..
B
.
B
B
B u1 ðtT Þ U
B
B
B
..
B
.
@
uN ðtT Þ U
and
A¼
g11
g12
v1 ðt1 Þ V
1
C
C
C
C
C
vN ðt1 Þ V C
C
C
..
C
C
.
C;
C
..
C
.
C
C
v1 ðtT Þ V C
C
C
C
..
C
.
A
vN ðtT Þ V
!
g21
.
g22
..
.
0
x1 ðt1 Þ X
B
..
B
B
.
B
B
B xN ðt1 Þ X
B
B
..
B
B
.
B
R¼B
..
B
.
B
B
B x1 ðtT Þ X
B
B
B
..
B
.
@
xN ðtT Þ X
y 1 ðt1 Þ Y
1
C
C
C
C
C
y N ðt1 Þ Y C
C
C
..
C
C
.
C;
C
..
C
.
C
C
y 1 ðtT Þ Y C
C
C
C
..
C
.
A
y N ðtT Þ Y
..
.
0
u01 ðt1 Þ
v01 ðt1 Þ
1
C
B
.. C
B ..
B .
. C
C
B
C
B 0
B uN ðt1 Þ v0N ðt1 Þ C
C
B
B .
.. C
C
B .
B .
. C
C
B
E¼B
.. C
B ...
. C
C
B
C
B
B u0 ðtT Þ v0 ðtT Þ C
C
B 1
1
C
B
C
B .
.
.. C
B ..
A
@
0
0
uN ðtT Þ vN ðtT Þ
Thus, we can determine A and the residual matrix by calculating the linear trend every T time steps:
A ¼ ðR? RÞ1 R? U ;
E ¼ U RA.
ð4Þ
222
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
4.2. Okubo–Weiss criterion
A criterion addressing how to partition the fluid into regions with different dynamical properties has
been derived by Okubo (1970) and Weiss (1991). With this criterion and the knowledge of the differential
kinematical parameters of a water parcel, one can determine if this parcel is more typical of a vortex or a
filament:
< 0 ) vortex;
k ¼ ðshr2 þ str2 Þ vor2
> 0 ) filament.
To increase the number of data used for the linear regression presented in Section 4.1, as we have generally only 2–3 floats in the water parcel, 5 time steps (20 h) have been used for the computations, low compared to the time scale of the mesoscale phenomena which are studied hereafter. Every phenomena
characterized by smaller time scales are filtered out by this method. Computations were repeated 5 times
(shifting time from 0 to 4 time steps), and the scatter of these 5 different estimates indicate the accuracy
of the results.
Fig. 8 shows computations of differential kinematic parameters for the situation presented in Fig. 5,
involving floats #09 and #88 from 15 to 18 November 1997. Estimates of the shear are about
0.5 · 105 s1 and agree with previous estimates (Section 3.2). The Okubo–Weiss criterion also revealed
that floats were drifting in a region dominated by filamentation (shear and stretch, k > 0) during that
period.
This criterion was also applied to the water parcel represented by floats #39 and #14 (Fig. 9). One can
clearly see two different dynamical regimes. The two floats were first drifting in a region dominated by filamentation (k > 0) which corresponds to the LIW flow south-west of Sardinia and then passed to an eddy
flow field (k < 0) around 8 November 1997 when the two floats left the vicinity of the continental slope to
drift in the Sardinian Eddy. Note that from that date, estimations of the relative vorticity f gives f/f about
.0.05, consistent with previous estimations of relative vorticity associated to the Sardinian Eddy (Section
3.4).
The same computations were carried out using floats #39, #14 and #09 between 20 November 1997 and
12 December 1997 (Fig. 10). The definition of the water parcel in this case is more tricky, but it shows interesting results as calculations are almost unchanged for relative vorticity (about 0.5 · 105 s1) with or
without adding float #09. Peak values are due to numerical limitations, when the surface of the parcel nears
0, the three floats being on a line. Between these peaks, float #09 acted as if it belonged to the Sardinian
Eddy from a vorticity point of view. Velocities measured by floats #09, #14 and #39 during that period also
suggest that float #09, defining the periphery of the gyre, can also define the periphery of the Sardinian
Eddy (Fig. 11). Indeed, velocities at floats #14 and #39 were about the same as velocities at float #09, when
they were the closest to float #09.
5. Discussion
5.1. Algerian Eddies
Long-lived AEs were followed in the Algerian Basin during the whole experiment (July 1997–July 1998).
They formed along the African coast around 1–2E longitude by instabilities of the Algerian Current and
then propagated eastwards until they reached about 800 0 E longitude. Here they detached from the coast
and moved cyclonically around the basin. They have a clear sea surface temperature signature and Puillat
et al. (2002) described in great detail the paths of AEs during the ELISA experiment (July 1997–July 1998)
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
floats #09 and #88
x 10
) du/dx+dv/dy (s )
223
2
0
x 10
2
0
)
x 10
2
0
dv/dx+du/dy (s )
x 10
2
0
14/11/97
15/11/97
16/11/97
17/11/97
18/11/97
19/11/97
15/11/97
16/11/97
17/11/97
18/11/97
19/11/97
x 10
2
8
4
2
2
shear +stretch
2
6
0
14/11/97
time (jj/mm/yy)
Fig. 8. (Top) Differential kinematic parameters computed for the water parcel defined by floats #09 and #88 (Fig. 5): div (divergence;
ou
ov
ov
ov
þ oy
), vor (relative vorticity; ov
ou
), str (stretch; ou
oy
), shr (and shear; ox
þ ou
). (Bottom) Okubo–Weiss criterion. Full line is the
ox
ox
oy
ox
oy
mean of 5 computations, the shaded area represents the standard deviation.
based on AVHRR images. SST images allow the positioning of AEs only in clear sky conditions. We
checked that other data (TOPEX/POSEIDON tracks and ELISA current meters; non-published yet) can
also be used to localize AEs with a better time resolution than with SST images only. CTD casts
(geostrophic computations), current meters records at 100, 350, 1000 and 1800 m depth and floats trajectories (around 600 m depth) showed that their vertical extension varied a lot. It appeared that 600 m deep
floats did not generally show any influence of the large anticyclonic AEs up above. Current meters at
100 m always showed the signature of these eddies, but for deeper current meters, it was in agreement with
floats.
Between September 1997 and January 1998, two AEs (96-1 and 97-1) were identified and located along
the Algerian coast, quite far away from Sardinia (Puillat et al., 2002). 96-1 circulated from 6E to 8E during that period and 97-1 from 3E to 4E along the north African coast with advection speed of few cm s1.
Their diameter varied between 100 and 170 km and they were limited to the AW layer during that period.
They remained quite far away from the south-west corner of Sardinia and consequently they could
hardly have been involved in the formation of the Sardinian Eddy revealed by floats #14 and #39. On
the other hand, AEs seem to participate to some LIW transport across the Algerian Basin. Indeed, by
the time AEs reach the south-west corner of Sardinia, LIW filaments appeared swirling around AEs as described by Emelianov et al. (1999).
224
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
Fig. 9. (Top) Differential kinematic parameters computed for the water parcel defined by floats #14 and #39 (Fig. 1): div (divergence;
ou
ov
ov
þ oy
), vor (relative vorticity; ov
ou
), str (stretch; ou
oy
), shr (and shear ; ov
þ ou
). (Bottom) Okubo–Weiss criterion. Full line is the
ox
ox
oy
ox
ox
oy
mean of 5 computations, the shaded area represents the standard deviation.
5.2. Barotropic motions
5.2.1. Algerian gyre
The ELISA current meters deployed at 1800, 1000 and 350 m depth in the Algerian Basin (MAST3 – EU
final scientific report, 1999, see Fig. Sa1 for the position of the moorings) confirm that the cyclonic large
scale gyre circulation is barotropic. In the surface layer (ELISA currentmeters at 100 m depth), the gyre
is hidden by strong mesoscale currents due to AEs (orbital velocities 50 cm s1) which were generally limited to the surface AW layer during this experiment.
AEs migrate cyclonically around the Algerian Basin more slowly (1 year) than the floats below (3–
4 months) and this is largely due to AEs being blocked several weeks along the African slope. This indicates
that the cyclonic Algerian Gyre is strongly influencing the general movement of AEs around the basin.
CTD casts carried out during ELISA experiment were mainly limited to 0–1000 m depth so they could
not give a good representation of deep circulation. Nevertheless, no doming of isopycnals between 0 and
1000 m depth was observed, which is consistent with a barotropic gyre.
So, these arguments are all in favor of a barotropic Algerian Gyre strongly influenced by the bottom
topography and earth rotation (see Section 3.1).
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
225
floats #09, #14 and #39
x 10
div (s )
2
0
x 10
vor (s )
2
0
x 10
str (s )
2
0
x 10
shr (s )
2
0
23/11/97
28/11/97
03/12/97
08/12/97
23/11/97
28/11/97
03/12/97
08/12/97
x 10
2
shr +str
2
2
8
6
4
2
0
time (jj/mm/yy)
Fig. 10. (Top) Differential kinematic parameters computed for the water parcel defined by floats #09, #14 and #39 (Fig. 1): div
ov
ov
ov
(divergence; ou
þ oy
), vor (relative vorticity; ox
ou
), str (stretch; ou
oy
), shr (and shear; ov
þ ou
). (Bottom) Okubo–Weiss criterion. Full
ox
oy
ox
ox
oy
line is the mean of 5 computations, the shaded area represents the standard deviation.
5.2.2. Alongslope current south-west of Sardinia
It has to be pointed out that the velocity (LADCP measurements) and CFC measurements presented by
Rhein et al. (1999) were undertaken approximately at the same time as our observations (Figs. 4, 8 and
Plate 1 in their paper). Rhein et al. (1999) showed that TDW coming from the Tyrrhenian Sea extended
from 700 down to 1500 m depth just below the LIW layer in the whole basin. Our Fig. 5 deduced from
floats drifting around 600 m depth is quite similar to Fig. 8 in Rhein et al. (1999) representing the mean
velocity between 800 and 1500 m depth. It indicates that currents beneath the AW surface layer (0–
350 m) and down to 1800 m depth or to the bottom when it is shallower, are almost constant with depth,
south-west of Sardinia. Thus, the circulation south-west of Sardinia appears to be mainly barotropic as the
Algerian Gyre.
In this region, the Algerian Gyre and the circulation originating from the Tyrrhenian Sea and following
the continental slope south of Sardinia, merge and correspond to a maximum velocity of about 10 cm s1.
This maximum velocity extends over a large depth range in the LIW-TDW layer and is situated above the
2500 m isobath (compare Figs. 5 and 4 with Figs. 5 and 8 in Rhein et al. (1999)). The LIW and TDW cores
coming directly from the Tyrrhenian Sea through the Channel of Sardinia–Tunisia and characterized by
high temperature/salinity and low CFC values (<0.7 pmol kg1) respectively, are mainly confined to the
right-hand side of the velocity maximum in a region dominated by an horizontal anticyclonic shear.
226
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
0.1
U (m/s)
0.08
0.06
0.04
#09
#14
#93
0.02
0
120
23/11/97
28/11/97
03/12/97
08/12/97
23/11/97
28/11/97
03/12/97
08/12/97
L (km)
100
80
60
40
20
dd/mm/yy
160
10/12/97
#39
140
30/11/97
#14
10/12/97
120
30/11/97
20/11/97
y (km)
20/11/97
100
10/12/97
30/11/97
#09
80
60
20/11/97
40
0
20
40
60
80
100
120
140
x (km)
Fig. 11. (Top) Times series of velocities at float #09, #14, and #39 and (Bottom) their trajectories between 20/11/97 and 11/12/97.
Indirect estimates of Rhein et al. (1999) from tracers lead to an annual mean inflow of TDW from the
Tyrrhenian sea of 0.4 Sv. Fig. 5 (as well as with Fig. 8 in Rhein et al. (1999)) would indicate an instantaneous transport of about 0.8 Sv with mean velocities of 4 cm s1 between 600 and 1600 m depth across a
section 20 km wide crossing the region of pronounced LIW and TDW influence. In comparison, the Algerian Gyre involves a mean recirculation of 4 Sv which is one order of magnitude higher, considering a mean
orbital velocity of 4 cm s1 across a section 1000 m high (between 600 and 1600 m depth) and 100 km wide
(between the center and the periphery of the gyre). Also, in contrast to the Algerian Gyre, which appears
very energetic, the Sardinian Vein west of Sardinia appears as a more sluggish circulation closely following
the continental slope northwards.
5.3. Sardinian Eddy formation
Concerning the origin and formation of the Sardinian Eddy, several relevant aspects have to be examined. As mentioned in Section 1, Millot (1999) proposed 3 mechanisms for the formation of eddies south-
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
227
west of Sardinia. We propose another one due to the Algerian Gyre, which separates from the slope and
was never observed before.
5.3.1. Gyre flow separation
The Algerian Gyre (approximately 250 km diameter and 3–4 months period) appears as a dominant and
permanent feature in the Algerian Basin. It seems to extend down to the bottom having maximum intensity
at great depths (500–1800 m depth), but also influences the AEs in the surface layer as well. It is characterized
by a seasonal variability in horizontal extension and intensity. It reaches its maximal intensity of 8–10 cm s1
at its periphery in late autumn-winter. This gyre implies that a large amount of water separates from the continental slope at about 3900 0 N, whereas the Sardinian Vein flows along the slope west of Sardinia.
The analysis of differential kinematic parameters (Section 4.2) indicates that the LIW flow south-west of
Sardinia is a region dominated by filamentation (mainly anticyclonic shear). The Sardinian Eddy characteristics seem to be mainly due to the anticyclonic shear in this region, as the relative vorticity of the eddy
was about equivalent to the shear vorticity (0.5 · 105 s1) near the continental slope, (Sections 3.2 and
4.2). Part of the shear would have been transferred to curvature vorticity (about half) and the other part to
shear vorticity, as the eddy is in quasi-solid body rotation, once the water parcel had left the vicinity of the
Sardinian slope. This is illustrated by Okubo–Weiss criterion evidencing 2 regimes, before (filaments) and
after (vortex) November 8, 1997 (Figs. 8 and 9). Time series of velocities from floats and differential kinematic parameters indicate that parcels at the periphery of the Algerian Gyre could also characterize the
edge of the Sardinian Eddy from kinetic energy and relative vorticity points of view (Section 4.2). That suggests that the flow separation south-west of Sardinia is certainly involved in the formation of the Sardinian
Eddy.
Marshall and Tansley (2001) derived an implicit formula for the separation of a current from the coast.
Scaling analysis indicates that the condition for a current separating from a vertical wall is
1=2
U
r<
;
b
where r is the radius of curvature of the coast, U the speed of the boundary current and b the gradient of the
Coriolis parameter in the downstream direction. For b 1011 m1 s1, U 0.1 m s1 (which are values
corresponding to the gyre in winter when it is fully developed), it requires r < 100 km. The separation from
the coast of a jet or a gyre would require the generation of eddies to maintain the separation according to
Tansley (2001). Around the south-west corner of Sardinia (Fig. 7(c)) the continental shelf and slope form a
bulge characterized by a radius smaller than 100 km satisfying the Marshall and Tansley (2001) criterion for
flow separation and eddy formation.
We propose the Sardinian Eddy first observed on November 8, 1997 exactly at the separation location
and then migrating on the edge of the Algerian Gyre, is generated according to this process.
We have no precise indication about the vertical structure of this Sardinian Eddy since float #14 drifted
at about 600 m depth and float #39 at about 700 m depth. NOAA/AVHRR SST or SEAWIFS images
could not be used intensively due to a strong cloud cover during the autumn–winter season in this region.
But, the formation mechanism we propose suggests this would be a strongly barotropic eddy. However, no
surface signature of this eddy could be found neither on the few SST images, nor on TOPEX/POSEIDON
tracks due to sampling and resolution problem.
5.3.2. Other mechanisms
LIW and TDW flow south of Sardinia, join the Algerian Gyre south-west of Sardinia and may continue
to flow along the western Sardinian slope (float #88, Section 3.2). Could an internal dynamic stability of
this flow be involved in the generation of Sardinian Eddies ? The horizontal structure of isopycnals is very
228
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
weak in the LIW–TDW layer (300–1600 m) and geostrophic calculations do not show any significant vertical shear along the slope of Sardinia. So, the hypothesis of Millot (1999) assuming that LIW eddies are
generated by an unstable intermediate jet of LIW seems unprobable. If an instability process was involved
in the generation of LIW eddies, it would rather involve a horizontal shear as for a barotropic instability.
However, second-order derivatives of velocities are needed to estimate a criterion of barotropic instability.
Another hypothesis of Millot (1999) considers that an increase in the vertical density gradient would produce anticyclonic vorticity. A production of anticyclonic vorticity of f/16 requires a shrinking of 10 m for
a water parcel initially at rest and extending over 160 m according to the conservation of potential vorticity.
This had not been observed on CTD casts carried out along the slope, but the CTD array was certainly not
sufficient to detect such shrinking.
The effect of a sharp slope angle to the right, so-called ‘‘eddy-gun’’ (Pichevin & Nof, 1995), could also
generate an eddy as a result of momentum conservation and requires an agent responsible for the removal
of the eddy from the generation area – this is provided by b-effect or advection (or both).
So, while we cannot exclude these effects, the Sardinian Eddy formation mechanism we propose is different and involves the newly discovered Algerian Gyre circulation feature which is far more energetic than
the circulation associated with LIW and TDW south-west of Sardinia.
5.4. Comparison with Meddies
The situation met south-west of Sardinia is really different from the situation met south-west of Spain
and links between the generations of Meddies and Sardinian Eddies (as sometimes it has been suggested
for Meddies and Leddies) are not likely. The presence of an energetic cyclonic flow (which seems to contribute actively to the generation of LIW eddies) is also in contrast with the general circulation south of
Spain. Even if Meddies and Sardinian Eddies had about the same radius of about 25–30 km (Richardson
et al., 1989), their density structures and relative vorticities are very different. Indeed, Mediterranean Water
composing the core of Meddies has a strong density signature (almost constant potential density) due to
homogenization and entrainment while cascading from Gibraltar Strait. This is in contrast to LIW flowing
along the slope of Sardinia in a strongly barotropic flow, presenting typical temperature–salinity characteristics which compensate so that there is almost no density difference between the modified LIW in the Algerian Basin and the LIW coming recently from the Tyrrhenian Sea. Also, a Meddy has about f/2 relative
vorticity due to the density distribution (Richardson et al., 1989), which is a very strong anomaly compared
with the f/16 of the Sardinian Eddy. In summary, this clearly shows that these two kinds of eddies are very
different dynamical features: the Meddies belong to the Submesoscale Coherent Vortex category (McWilliams, 1985) in contrast to Sardinian Eddies.
6. Conclusion
The anticyclonic Sardinian Eddy revealed by the MATER (ELISA-LIWEX) experiment is characterized
by a radius of about 30 km, a rotational period of 3–4 weeks and if we approximate it with a quasi-solid
body rotation, a mean relative vorticity of about f/16 corresponding to a Rossby number R0 0.06. The
origin of this Sardinian Eddy seems to be strongly related to some interactions between the large scale cyclonic Algerian Gyre and the general boundary circulation entraining LIW along the continental slope of
Sardinia. The Sardinian Eddy appeared precisely at the separation location between these two large scale
features and evolved later on at the periphery of the Algerian Gyre. The Sardinian Eddy has a strong LIW
signature as the water at 600 m depth in the core of the eddy has the same characteristics as the LIW found
along the slope West of Sardinia. Its relative vorticity may be due to the anticyclonic shear between the
Algerian Gyre and the circulation along the continental slope south-west of Sardinia where the Algerian
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
229
Gyre and the Sardinian Vein merge and split. The formation mechanism seems mainly due to the energetic
cyclonic Algerian Gyre separating from the continental slope south-west of Sardinia and as a result of (1)
the geometry of the continental slope, (2) the speed of the flow intensified by the gyre at this location and (3)
the planetary vorticity gradient along the direction of the flow. The flow separating from the slope generates
eddies like Sardinian Eddies according to a process similar to the one described by Tansley (2001) and Marshall and Tansley (2001), dealing with the Gulf Stream separating from the east north American coast near
Cape Hatteras.
This observation of a Sardinian Eddy is a valuable information concerning the pending question of how
the LIW spreads in the Algerian Basin. Our findings contribute (1) to characterize Sardinian Eddies from
an eddy-structure point of view (kinematics and dynamics) (2) to shed light on the origin of such eddy and
(3) to estimate how important are Sardinian Eddies for the general circulation of LIW and TDW in the
Mediterranean Sea. The Sardinian Eddy structure described is noticeably distinct from Meddies and Leddies previously reported in the literature.
In addition, the MATER 1997/98 experiment indicated a Submesoscale Coherent Vortex composed of
newly formed WMDW (SCV-S1) interacting with the Sardinian Eddy (Testor & Gascard, 2003) indicating
that Sardinian Eddies can also have a significant role influencing the spreading of newly formed WMDW in
the Algero-Provençal Basin and vice versa.
The vertical extension of the Sardinian Eddy could not be precisely specified but it may have well concerned the whole water column. Each Sardinian Eddy would amount to about 3 · 1012 m3 of LIW and
TDW leaving the boundary circulation towards the interior of the Algero-Provençal Basin. This is equivalent to a volume of water passing through a section 30 km wide and 1 km high, with a mean velocity of
about 4 cm s1 during 1 month. In other words, a single Sardinian Eddy can account for as much as
1 month of the inflow of waters coming from the Tyrrhenian Sea, which is quite significant and should have
important consequences on the Sardinian Vein flowing northwards west of Sardinia. The Algerian Gyre is
characterized by a strong seasonal variability and Sardinian Eddies will mainly be formed in winter when
the gyre has its greatest intensity. Accordingly, 4–5 Sardinian Eddies may very well be formed each year
and can account for almost 50% of the waters entering into the Algero-Provençal Basin through the Channel of Sardinia.
Acknowledgments
The MATER experiment has been funded by the European Union in the framework of the MAST3 program under contract MAS3-CT96-0051. We acknowledge Antonio Lourenço and Catherine Rouault for
their expertise in operating floats and data processing. MATER lagrangian float experiment (deployment)
was accomplished in the context of ELISA on board French research vessel Le Suroıˆt (Ifremer). We
acknowledge Isabelle Taupier-Letage and Claude Millot for letting us have access to the SST, CTD and
currentmeters data collected during the ELISA experiment. The recovery experiment (LIWEX) was accomplished on board of French research vessel Thetys II (CNRS-CIRMED). P. Testor was supported by a
DGA grant.
References
Afanasiev, Y., & Fillipov, I. (1996). Generation of intermediate vortices in a rotating stratified fluid: laboratory models. Journal of
Geophysical Research, 101, 18167–18174.
Baey, J. M., Renouard, D., & Chabert dÕHières, G. (1995). Preliminary results about the stability of an intermediate water current.
Deep-Sea Research, 42(11–12), 2063–2073.
230
P. Testor, J.-C. Gascard / Progress in Oceanography 66 (2005) 211–230
Benzhora, M., & Millot, C. (1995). Characteristics and circulation of the surface and intermediate water masses off Algeria. Deep-Sea
Research, 42, 1803–1830.
Bouzinac, C., Font, J., & Millot, C. (1999). Hydrology and currents observed in the channel of Sardinia during the PRIMO-1
experiment from November 1993 to October 1994. Journal of Marine Systems, 20, 333–355.
Emelianov, M., Millot, C., Font, J., & Taupier-Letage, I. (1999). New data on Levantine Intermediate Water circulation in the western
Mediterranean Sea, In: A. Zatsepin (Ed.), Oceanic fronts and related phenomena, Konstantin Fedorov Memorial Symposium
Proceedings, UNESCO, Paris, IOC Workshop Report No. 159, pp. 132–137.
Font, J., Millot, C., Salas, J., Julia, A., & Chic, O. (1998). The drift of modified atlantic water from the Alboran sea to the eastern
Mediterranean. Sci. Mar., 62(3), 211–216.
Fuda, J. L., Millot, C., Taupier-Letage, I., Send, U., & Bocognagno, J. M. (2000). XBT monitoring of a meridian section across the
Western Mediterranean Sea. Deep-Sea Research I, 47, 2191–2218.
Gascard, J. C., Rouault, C., & P. Testor (1999). General ocean circulation and subsurface mesoscale eddies in the Algerian Basin. In
Fourth Mediterranean targeted project workshop, MATER, Perpignan, France, pp. 50–51.
Hopkins, T. S. (1988). Recent observations on the intermediate and deep water circulation in the Southern Tyrrhenian Sea.
Oceanologica Acta, special issue, 9, 41–50.
Katz, E. J. (1972). The levantine intermediate water between the strait of Sicily and the strait of Gibraltar. Deep-Sea Research, 19,
507–520.
Marshall, D., & Tansley, C. E. (2001). An implicit formula for boundary current separation. Journal of Physical Oceanography, 31,
1633–1638.
MAST3 – EU final scientific report, (1999). Contract MAS3-CT96-0051.
McWilliams, J. (1985). Submesoscale coherent vortices in the ocean. Reviews of Geophysics, 23(2), 165–182.
Millot, C. (1999). Circulation in the Western Mediterranean Sea. Journal of Marine Research, 20, 423–442.
Molinari, R., & Kirwan, A. D. (1975). Calculations of differential kinematic properties from lagrangian observations in the Western
Caribbean Sea. Journal of Physical Oceanography, 5, 483–491.
Okubo, A. (1970). Horizontal dispersion of floatble particles in the vicinity of velocity singularities such as convergences. Deep-Sea
Research, 17, 445–454.
Perkins, H., & Pistek, P. (1990). Circulation in the algerian basin during June 1986. Journal of Geophysical Research, 95, 1577–1585.
Pichevin, T., & Nof, D. (1995). The eddy cannon. Deep-Sea Research, 43(9), 1475–1507.
Puillat, I., Taupier-Letage, I., & Millot, C. (2002). Algerian Eddies Limetime can near 3 years. Journal of Marine Systems, 31(4),
245–259.
Rhein, M., Send, U., Klein, B., & Krahmann, G. (1999). Interbasin deep water exchange in the western Mediterranean. Journal of
Geophysical Research, 104, 23495–23508.
Richardson, P. L., Walsh, D., Armi, L., Schröder, M., & Price, J. F. (1989). Tracking three meddies with SOFAR floats. Journal of
Physical Oceanography, 19, 371–383.
Sammari, C., Millot, C., Taupier-Letage, I., Stefani, A., & Brahim, M. (1999). Hydrological characteristics in the Tunisia–Sardinia–
Sicily area during spring 1995. Deep-Sea Research, 46-I, 1671–1703.
Sanderson, B. G. (1995). Structure of an eddy measured dy drifters. Journal of Geophysical Research, 100, 6761–6776.
Swallow, J. C. (1955). A neutral-buoyancy float for measuring deep currents. Deep-Sea Research, 3, 74–81.
Tansley, C. E., & Marshall, D. (2001). Flow past a cylinder on a b plane with application to Gulf Strean separation and the Antartic
Circumpolar Current. Journal of Physical Oceanography, 31, 3274–3283.
Testor, P., & Gascard, J.-C. (2003). Large scale spreading of deep waters in the western Mediterranean Sea by submesoscale coherent
eddies. Journal of Physical Oceanography, 33, 75–87.
Weiss, J. (1991). The dynamics of enstrophy transfer in two-dimensional hydrodynamics. Physica D, 48, 273–294.
Wüst, G. (1961). On the vertical circulation of the Mediterranean Sea. Journal of Geophysical Research, 66, 3261–3271.