Hydrographic sections were conducted south of Madagascar during the Agulhas Current Sources EXperiment (ACSEX-2) cruise survey in March 2001. The East Madagascar Current (EMC), presumably one of the sources of the Agulhas Current, was crossed four times over the Madagascar ridge. The upwelling cell, at the southeastern tip of Madagascar (inshore of the EMC), was hydrographically highlighted for the very first time during ACSEX-2. The behavior of the EMC south of Madagascar is studied using both hydrographic data and satellite imagery. At the southeastern end of Madagascar, the EMC turns westward and the dramatic change in the shape of the shelf favours the development of a cyclonic eddy that is embedded between the EMC core and the coast. The upwelling is then associated to the presence of this eddy and to wind favorable conditions.
 The East Madagascar Current (EMC) plays a potential controlling role on the major western boundary current (Agulhas Current) of the South West Indian Ocean. The EMC is narrow and intense and presents all the characteristics of a western boundary current [Lutjeharms et al., 1981]; [Shallow et al., 1988]. The behavior of the EMC south of Madagascar Island raises some controversy about its contribution to the Agulhas Current. Analysis of cruise data from a hydrographic section directly south of Madagascar, by [Lutjeharms et al., 1981] found no evidence of the EMC rounding the southern tip of Madagascar and continuing westward to supply the Agulhas Current. They noted however an eastward water movement at the surface, which appeared to be a continuation of a southward flow down the West Coast of Madagascar.
 Evidence of an eastward retroflection of the EMC has been seen in both infrared [Lutjeharms, 1988] and ocean color [Lutjeharms and Machu, 2000] images. [Gründlingh et al., 1991] suggest that eddies, formed in the vicinity of the EMC, may be advected south by this current. When such an anticyclonic eddy lies south of Madagascar, the rotational sense would lend credence to the concept of a retroflection. Blocking of the retroflection would eject anticyclonic eddies westward to the African continent, similar to rings shed from the Agulhas current south of South Africa. Recently, [Lutjeharms et al., 2000] advanced a possible seasonal variability in the speed of the EMC in phase agreement with both prevailing winds in the region and the flow of the South Equatorial Current. Using altimetry data, [Heywood and Somayajulu, 1997] showed an apparent seasonality with the EMC flowing directly towards the African coast in winter, but looping northward into the southern Mozambique Channel in summer.
 The area south of Madagascar was visited in March 2001 within the Agulhas Current Source Experiment (ACSEX) program. The aim of the ACSEX program is to study the sources of the Agulhas Current, which possibly control the shedding of rings at the retroflection of this current. These rings enter into the Atlantic Ocean and through this interocean exchange, participate in the global thermohaline circulation.
 During the ACSEX-2 cruise, conducted on board the R/V Pelagia, three hydrographic sections crossed the upwelling cell occurring at the southeastern end of Madagascar (Figure 1). The data from part of these sections show the first hydrographic evidence of this upwelling cell, which had previously been identified on satellite imagery [Lutjeharms and Machu, 2000] and [DiMarco et al., 2000]. [DiMarco et al., 2000] believe the upwelling off Madagascar is forced through a combination of upwelling favorable wind stress and frictional interaction between the southward flowing EMC and the continental shelf slope. In a later section, hydrographic data are used to put forward the mechanisms responsible for this remotely observed upwelling. A greater understanding of this inshore upwelling cell might be of great importance in comprehending the downstream behavior of the EMC.
2. Hydrographic Data
 On board the R/V Pelagia, the ACSEX-2 cruise surveyed the southwest Subtropical Indian Ocean from Cape Town (March 4th) to Reunion Island (March 26th, 2001). During this cruise, 35 CTD stations (stations 62 to 97) were conducted in a series of four sections which crossed the major features south and southeast of Madagascar (Figure 1). The paper will focus on data from stations nearest to the coast along transects C, D and E around the southeastern tip of Madagascar.
 Conductivity, temperature, pressure and fluorescence were recorded down to a maximum of 2400 meters depth on individual CTD casts; temperature was also measured continuously at the surface by an AQUAFLOW thermosalinograph. A 24-position rosette sampler was used, fitted with 5 and 10 liters NOEX sampler bottles. Water currents were monitored by an LADCP mounted on the CTD frame and by a vessel mounted ADCP. Samples were analyzed onboard for salinity, dissolved oxygen and nutrient concentrations (nitrate, nitrite, phosphate and silicate) by the Netherlands Institute for Sea Research (NIOZ) scientists [Van Aken, 2001].
3. Results and Discussion
 Part of temperature sections C, D and E (stations marked with stars on Figure 1) are shown on Figure 2. The rise of isotherms on each section makes the occurrence of an upwelling of cold water along the shelf evident and shows its extension. The further upstream we go, the less the offshore extension of the upwelling along the shelf. This upwelling along the slope is also visible on the vertical nutrient distribution. Table 1 shows nitrogen concentrations in the upper layer decreasing from the continental plateau (stations 69, 84 and 83, 85) towards the deep ocean (stations 71, 82, 86 and 87 with respect to transect considered). Station 70 behaves differently and will be discussed later. Lowest oxygen concentrations were encountered at station 69, 70 and 83 at 100 m depth (Table 1). At these locations, surface chlorphyll-a concentrations were higher and therefore the lower oxygen concentrations reflect the consumption induced by the remineralization of organic particles falling down through the water column.
Table 1. Nitrogen and Oxygen Concentrations for Section C, D and E. Stations Closest to the Coast are Presented
O2 (mmol/kg) d
Station 69 (section C)
Station 84 (section D)
Station 85 (section E)
 The upwelling cell off southeastern Madagascar is also well monitored by ocean color (Figure 3a).
 The satellite image of the 18th April 2001, simultaneous to our ship coverage of the area, displays some complementary information to describe the behavior of the EMC south of Madagascar. In Figure 2, the dome shape of isotherms 9° to 17°C at station 70 indicates the presence of a cyclonic eddy. Nutrients measured at station 70 (Table 1) agree with the occurrence of such an eddy feature since concentrations levels are lifted by 200 m compared to station 71.
 Surface currents represented in Figure 3a show the location of the core of the EMC while the velocity components of station 82 (Figure 3b) give the depth of the current that follows the steep narrow shelf along the east coast of Madagascar. From the southeastern point of Madagascar (47°E), isobaths follow the coast and have a concave shape until the southern end (45.5°E). As the current, which has a signal down to 1000 m, follows the shelf, this concavity is favorable to the development of a cyclonic eddy. SeaWiFS chlorophyll-a on Figure 3a and altimetry (Topex cyle 18705; not shown) confirm this hypothesis and exhibit an eddy centered at 25.5°S–46.5°E. This eddy is confined between the core of the EMC and the coast. Section C crosses its southwestern edge while section D seems to stand further north. This concavity in the shelf bathymetry seems to have the same effect on the EMC as the Agulhas Bight has on the Agulhas Current, where the existence of a permanent cyclonic eddy (between 22°E and 25°E) has recently been suggested [Roberts et al., 2002]. This eddy would result from the interaction of the Agulhas Current and the topography of the Agulhas Bight. Authors assume that the counter-current observed by [Beal and Bryden, 1997] under the Agulhas Current could find its source in this cyclonic eddy. Despite the presence of this cyclonic eddy, station 69 surface current on Figure 3b gives a seaward current while an opposite current would be expected. This reflects the Ekman transport relative to eastern winds that were dominating during the period of the cruise. It also explains the masking of the temperature signature of the eddy at the surface along section C (Figure 2c).
 Southeast of Madagascar, meridional and zonal current components of stations 82 and 87 shows a reverse of the current from a south-westward direction at the surface to a north-eastward flow at depth (Figure 3b). This reverse is about 200 m deeper for station 82 than for station 87. It appears as if the return current of the cyclonic feature embedded against the coast meets the EMC at the southeastern end of Madagascar. A single deep northward counter-current path would then be allowed to “leave” this concavity. As in the Agulhas Bight, this eddy would then be the source of a deep poleward current beneath the EMC.
 The Madagascar upwelling cell which was hydrographically measured for the first time during ACSEX-2 cruise, was found to have a spatial distribution which agrees well with remote color measurements. The combination of satellite imagery and in situ observations allowed us to outline the functioning of the Madagascar upwelling cell. We found that the change from a bathymetry characterized by a narrow continental shelf and a steep slope to one characterized by a wider and a less steep slope seems to be responsible for the development of an eddy feature whereas upwelling favorable wind conditions create an Ekman surface current responsible for the masking of the eddy surface signature. While [DiMacro et al., 2000] showed that the upwelling is increased in area and magnitude when wind-stress is increased, it should be examined whether the intensity and extent of the upwelling cell, found inshore of the EMC, influences the EMC behavior off the south coast of Madagascar. ACSEX-2 cruise current data along section B shows an EMC being steered through the bathymetric channel situated at 45.5°E–26.5°S (not shown; de Bruin, pers. com. but clearly suggested on Figure 3b chlorophyll scene). A modeling study might be necessary to further investigate the different upwelling scenarios and their impact on the role of the EMC in supplying the Agulhas Current.
 We would like to thank the NIOZ Institute for inviting us on board the R/V Pelagia. We acknowledge the dedicated effort of the officers and crew of the R/V Pelagia. E.M. was supported by a grant from the Centre National de la Recherche Scientifique to LEGOS. We thank R. Schlitzer since Figure 2 has been processed using Ocean Data View http://www.awi-bremerhaven.de/GEO/ODV) Ocean color data used in this study were produced by the SeaWIFS project at Goddard Space Flight Center. We finally thank J. Sudre for the final processing of some SeaWIFS images and V. Garçonfor a critical reading of the first version of the manuscript.