Journal of Geophysical Research: Oceans

Circulation over a submarine canyon in the NW Mediterranean

Authors


Abstract

[1] The circulation in Blanes canyon, an interruption in the NW Mediterranean continental shelf north of Barcelona, was investigated. The study employs data from oceanographic surveys carried out in the summer and fall of 2003. Velocity data show that in the vicinity of the shelf break the flow is deflected along the canyon walls. A cyclonic mean flow can be seen over the canyon mouth owing to vortex stretching of fluid parcels advected across the shelf break. Field observations are in qualitative agreement with fundamental fluid dynamic considerations based on potential vorticity conservation and friction effects at lateral boundaries. Evidence is given that upwelling is found near the shelf break inside the canyon in the two field experiments. This upwelling extends vertically from the seasonal thermocline (at about 100 m) to the shelf-slope front (at about 200 m). There is no evidence that upwelled water can reach the continental shelf.

1. Introduction

[2] The continental margins along oceans and seas are often interrupted by submarine canyons. These topographic features are preferred pathways for shelf-slope exchanges of material and energy, and play important roles in distributing suspended particulate matter and exporting sediments to the open ocean [Durrieu, 1994]. From a biological point of view, submarine canyons are preferential recruitment habitats [Sardà et al., 1994] and are also seen as biodiversity hot spots [Gili et al., 2000]. Higher than average abundance of fish eggs and larvae [Palomera, 1992; Sabatés and Olivar, 1996] has been observed in the vicinity of several NW Mediterranean canyons, including Blanes canyon.

[3] Submarine canyons can deflect the incoming flow and modify the local circulation. Numerical models developed to predict the local circulation show that the width of the canyon is an important parameter [Klinck, 1988]. Narrow canyons (defined as those canyons whose width is smaller than half of the internal Rossby radius of deformation) have a strong effect on circulation, and are normally characterized by a general cyclonic circulation within the canyon because of the stretching of water parcels passing over the shelf break [Hickey, 1997]. Wide canyons only modify the circulation along the topography [Klinck, 1988]. The direction of the incoming flow is another key factor in the local dynamics [Klinck, 1996]. When the incoming current has the coast on its left (a left bounded current), the current-canyon interaction causes local upwelling over the entire canyon. Freeland and Denman [1982] proposed that the up-canyon flow could be driven by the onshore unbalanced pressure gradient. When the flow has the coast on its right (a right bounded flow), the current-canyon interaction causes asymmetry in the vertical velocity field; downwelling is forced over the upstream wall, whereas upwelling is forced over the downstream wall. The total net vertical flux generally indicates downwelling [Klinck, 1996]. However, recent laboratory experiments performed on the Coriolis turntable demonstrated that right bounded flows can force local upwelling in canyons because of strong vortex stretching of the flow passing across the shelf break [Boyer et al., 2006].

[4] The importance of submarine canyons in modifying the synoptic and local scale circulation in the NW Mediterranean has been pointed out in several papers. In Palamos canyon, Álvarez et al. [1996] conducted a number of case studies on a range of different currents impinging on the canyon. These produced different shelf-slope exchanges that in turn modified the biomass pattern in the area. In Blanes canyon, Rojas et al. [1995] found an anomalous situation, in which the flow above and just below the thermocline was reversed because of the presence of cold, fresh waters coming from the Gulf of Lions continental shelf. These field studies in the NW Mediterranean region provide good data sets for the motion field in submarine canyons. These data can then be compared with laboratory and numerical model results in order to better understand canyon flows.

[5] The purposes of the present communication are (1) to present the data obtained in two oceanographic cruises undertaken in a NW Mediterranean canyon (Blanes canyon) during 2003 as part of a multidisciplinary project for understanding the circulation in Blanes canyon and the geological and biological processes associated with this circulation (RECS project, REN2002-04556-C02-01/MAR) and (2) to delineate field observations that have clear counterparts in the laboratory and numerical models considered here.

2. Strategy and Methods

2.1. Flow Characteristics

[6] The general circulation in the NW Mediterranean is governed by the Northern Current, a baroclinic current formed by water of Atlantic origin (Atlantic Water, or AW) flowing over water of Levantine origin (Levantine Intermediate Water, or LIW). The path of the Northern Current describes a general cyclonic flow along the NW Mediterranean continental slope [Millot, 1999]. This current arrives at the Catalan Sea from the Gulf of Lions, following the continental shelf and slope in a SW direction. The surface signature of the Northern Current is enhanced by a density front, which deepens toward the coast, intersecting the continental slope at the shelf break depth [Font et al., 1988, 1995]. In addition to this baroclinic flow, the current has a barotropic component that involves the entire water column, including LIW and the underlying Western Mediterranean Deep Water (WMDW) [Conan and Millot, 1995]. The Northern Current is 30 to 50 km wide, and is characterized by a vertical velocity profile of about 30 to 50 cm s−1 at the surface, decreasing approximately linearly with depth to speeds of a few centimeters per second at several hundred meters depth [e.g., Lapouyade and Durrieu, 2001]. This velocity is maintained to approximately the bottom [e.g., Flexas et al., 2002]. The Northern Current is subjected to meanders and anticyclonic eddies [Font et al., 1995] because of mixed baroclinic-barotropic instabilities [Albérola et al., 1995a; Flexas et al., 2004, 2005].

2.2. Setting

[7] Blanes canyon is situated off the Catalan coast, in the NW Mediterranean (Figure 1). We define the canyon mouth as the line joining the extreme offshore points of the shelf break (150 m isobath) on each side of the canyon. According to this definition, the mouth of the canyon is along latitude 41.52°N, joining the 150 m isobath, and is 16 km wide (Wsb in Figure 1). The canyon mouth, however, has a particular structure and in only 2 km onshore it decreases to about a half its width (W130 in Figure 1). In this case, the maximum depth over the canyon walls is 130 m deep. This definition of the canyon mouth is based on the consideration that water parcels passing over the canyon rim are more influential in modifying the local current structure than those passing the canyon beyond its mouth. In an oceanographic framework, the region offshore the canyon mouth is also of interest, mainly for biological and geological purposes. In the present study, the region defined between the head of the canyon (the most coastal part of the canyon, where both canyon walls join) and the mouth of the canyon (Figure 1), is named the “upper canyon,” whereas the offshore continuation of the canyon is referred to as the “lower canyon.”

Figure 1.

(top) Plan view of Blanes canyon bathymetry showing the shelf break depth (i.e., the 150 m isobath) with a thick bold line, the width at the shelf break Wsb, the length L, the width W, and the width at the 130 m isobath W130. Location of Malgrat de Mar meteorological station is marked with a white circle. (bottom) ADCP/CTD stations are marked with circles. Interpolated grid points are marked with dots. The rectangle indicates the box used for control volume calculations (the upper canyon). Sections used in Figures 5, 7, 10, and 13 are shown with labels used in Figures 5, 7, 10 and 13.

[8] Blanes upper canyon is 16 km (L) long and 8 km (W) wide, where the width is taken at the midupper canyon (Figure 1). The axial bottom depth at its mouth (H) is 1100 m, the depth of the canyon from its rim to the bottom (hm) is 950 m, and the half width (l) is 4 km. Thus, the fractional height (hm/H) is 0.86 and the aspect ratio (H/l) is 0.3. The internal Rossby radius of deformation is given by Rd = NH/f, where N is the buoyancy frequency, and f is the Coriolis parameter. For the two periods studied here (fall and summer), N (averaged from the surface to 1100 m depth) varies from 1.87 × 10−3 to 2.03 × 10−3 s−1 for fall and summer respectively. These values lead to a range of Rd ≃ 20–22 km. The stratification parameter (S = NH/fl) varies from 5.1 to 5.6, and the vertical stratification scale (Tr = fl/N) ranges from about 195 m to 210 m. In winter, however, the water column homogenizes, N is about 1.05 × 10−3 s−1, leading to S = 2.9 and a Tr of 380 m. Considering that the Northern Current varies from 30 to 50 cm s−1 at the surface, linearly decreasing to 3 to 5 cm s−1 at about 500 m depth, the incident flow over the canyon (U) at the shelf break depth ranges from 8 to 14 cm s−1, and the Rossby number (Ro = U/fl) varies from 0.2 to 0.35.

[9] Given that the internal Rossby radius of deformation is about twice the width of the canyon, Blanes canyon is considered a narrow canyon [Klinck, 1988]. Blanes canyon is also deep, that is, the depth of the canyon below its rim is roughly twice the depth of the incident flow (950 m versus 500 m). The shape of the upper canyon is fairly symmetrical except at its mouth, where the upstream wall is wider than the downstream flank. The canyon axis is fairly perpendicular to the local isobaths, and shifts eastward at the head of the canyon.

2.3. Data

[10] The data presented here were taken during two oceanographic cruises performed at the beginning of summer 2003 (22–25 June) and fall 2003 (2–5 November). The velocity and hydrographic sampling consisted of a rectangular grid of 59 stations distributed in 9 N–S transects of 7/6 stations each (Figure 1). At each grid point conductivity-temperature-depth profiler (CTD) and Vessel Mounted acoustic Doppler current profiler (ADCP) data were measured. The separation between stations was about 4 km in longitude inside the canyon and about 8 km outside the canyon. In latitude, the separation between stations was about 8 km.

[11] These surveys represent the first high-resolution 4 km sampling performed over Blanes canyon. The experiment was designed to assure enough grid points within the study region to resolve canyon-scale structures over the upper and lower canyon, according to the finest resolution bathymetry available at the time [Canals et al., 1982]. However, a new finer-scale bathymetric resolution [Canals et al., 2004] recently became available, and revealed the narrowness of the upper canyon. This new resolution of the bathymetry also revealed that only four casts were actually performed inside the upper canyon: two at its head, and two over the midcanyon. The coarse resolution of the data within the upper canyon is a limiting factor of the present data set. As such the vorticity field cannot be calculated in the upper canyon.

[12] The velocity data were obtained from the surface to 360 m at 8 m intervals using a vessel mounted ADCP (VMADCP). The data were averaged every 5 minutes to obtain several vertical profiles at each station. In order to calculate absolute velocities, ADCP data were referred to the bottom when correctly detected; otherwise the data were referred to the ship velocity measured by the 3DF ASTHECH GPS three-dimensional positioning system or by the differential GPS (GPSD). The ADCP data were preprocessed according to (1) a maximum threshold for the absolute velocities, (2) a maximum threshold for the velocity variations between two correlative measurements, and (3) the intensity of the echoes received by the ADCP. After visual checking of every profile, the velocity profiles were averaged to obtain a single velocity profile at each station.

[13] The hydrographic profiles were obtained with a MK-III CTD deployed down to the bottom. Water samples obtained during the cruise were used to calibrate the CTD conductivity cell. The accuracy in salinity was 0.01 units. The hydrographic profiles were preprocessed using the Sea-Bird Electronics Data Processing software.

[14] Wind data were obtained from two inland stations close to Blanes, namely Malgrat and Castell d'Aro, and from the 18 km resolution meteorological model MASS. Both meteorological stations belong to the XAC network of the Catalan Meteorological Office (SMC), whereas the MASS model is used by SMC to perform daily meteorological predictions (see www.meteocat.com for more information). The three sources of information were consistent for the periods studied here. Previous to and during the summer cruise, only sea breezes were observed, characterized by intensities of 2–3 m s−1 (Figure 2a). During the fall cruise winds were also mild; NE winds of 5 m s−1 were registered only once, several days previous to the cruise (Figure 2a).

Figure 2a.

Wind data (in m s−1) at 10 m above sea level for the (top) summer cruise and (bottom) fall cruise. Cruise periods are marked with boxes. Data obtained from Malgrat de Mar meteorological station (see Figure 1).

[15] Tides in the Mediterranean Sea are small mainly because the Mediterranean is a semi-enclosed basin with a very narrow outlet/inlet into the Atlantic (the Strait of Gibraltar). Except near the Strait of Gibraltar, the northernmost part of the Adriatic Sea and Gabes Gulf (off Tunisia), the maximum tidal amplitudes in the Mediterranean Sea are less than 10 cm [e.g., Tsimplis et al., 1995]. Tide characteristics obtained from Puertos del Estado Tide Gauge Network show that near Blanes canyon (the so-called Barcelona site) the largest diurnal and semidiurnal tide constituents O1, K1, M2, and S2 are on the order of centimeters, and thus can be safely neglected (data obtained from Puertos del Estado web site: www.puertos.es).

[16] In order to account for the lack of synopticity of the ADCP and hydrographic data presented here, current meter data from several moorings located in the upper and lower canyon, contemporary to the summer and fall cruises, were examined. During the fall cruise, a sensor located at the upper canyon mouth, at a nominal depth of 375 m, showed quasi-steady flow speed and direction during the 3 day survey period (Figure 2b). During the summer cruise no instruments were deployed in the upper canyon, but two moorings deployed over the lower canyon showed no significant changes in the velocity magnitude or direction during the 3 day survey period (Figure 2b). Thus the data presented here are considered synoptic.

Figure 2b.

ADCP horizontal velocity fields at 250 m and 350 m and 6-hour averaged current meter data from several moorings located at 250 m and 375 m at the upper and lower canyon during (right) fall cruise and (left) summer cruise. ADCP arrows nearest to recording current meter locations are highlighted in red. Data during surveys from moored sensors at indicated depths are also shown.

2.4. Methods

[17] To compute dynamic variables, observations were first interpolated onto a 30 × 30 regular grid by means of an Optimum Interpolation technique [Bretherton et al., 1976]. All fields were spatially smoothed with a cutoff wavelength of 10 km, a compromise which allowed us to (1) maintain the structures observed when using smaller wavelengths and (2) obtain reliable vorticity values. According to Gomis and Pedder [2005], in order to obtain reliable vorticity values it is necessary to smooth the sampling grid with a cutoff wavelength of about 3 times the separation between grid points. For a separation of 4 km, this is a minimum of a 12 km cutoff wavelength. If cutoff wavelengths larger than 10 km are used, however, some features inside and outside the canyon are elongated or merged. In such cases smoothing clearly modifies features of the canyon's scale. Therefore, 10 km was considered the optimal cutoff wavelength, because this allowed the canyon-scale structures to be maintained, while being close to the recommended cutoff wavelength for reliable vorticity estimates. Although vorticity fields are qualitatively reliable, vorticity gradients can only be considered approximate. Comparing the results obtained with a 10 km and a 12 km cutoff wavelength, the horizontal vorticity gradients decrease by a half in the latter case.

[18] In order to estimate vertical velocities, the total vertical volume flux per unit area through areal elements at 10 m separation distances in the vertical is obtained using a control volume approach employing conservation of volume principals. The selected control volume is a box defined by the mouth of the upper canyon at the 130 m depth isobath (W130 at latitude 41.536°N) and the respective perpendicular lines toward the head of the canyon at longitudes 2.786°E and 2.939°E (Figure 1). The upper boundary condition is set at the surface, where the vertical velocity is zero. The vertical volume flux for the first layer at a depth of 10 m is obtained by balancing the vertical volume flux with the net horizontal flux through the vertical sides of the control volume. The vertical volume flux through the next level at 20 m is then obtained in the same way as for the 10 m level, although now the flux through the upper surface of the control volume (i.e., at 10 m) is also accounted for. With this method we estimate the mean vertical volume flux at each 10 m level from the surface to 360 m, which was the maximum VMADCP depth range in these cruises. Finally, dividing the total vertical flux at each depth by the horizontal area corresponding to this level (the horizontal area of the canyon every 10 m is taken from the bathymetry obtained by Canals et al. [2004]) we obtain an estimate of the mean vertical velocity as a function of depth. The procedure to obtain the total vertical flux per unit area, equation image, can be expressed as:

equation image

where XY is the horizontal area at each z-level, equation image are the horizontal velocity vectors, equation image is the vector normal to the horizontal area XY, and dA is the horizontal unit area.

[19] In order to investigate the accuracy of the vertical velocity estimates we performed several sensitivity tests in which we varied the box limits by plus or minus one grid point in longitude and latitude at each side of the box used as the control volume, therefore allowing the inclusion/exclusion of nearby CTD profiles (see Figure 1). Error bars were computed from standard deviation values obtained from these sensitivity tests. It is important to keep in mind that scales within submarine canyons are small, no more than the half width of the canyon, and thus estimates of fluxes into and out of boxes will have associated errors if flow features are missed between sample points. With the low spatial resolution of the present data set inside the upper canyon we cannot assure that all features are resolved, and therefore this will definitively impact our flux estimates.

3. Field Observations

[20] Two vertical sections taken upstream of the canyon show the incoming flow characteristics observed during each cruise (Figure 3). In fall, the core of the Northern Current is narrow and intense. The NC lies over the shelf break (at 150 m) with velocities of about 20 cm s−1. In summer, the NC is wider and less intense. At the shelf break depth maximum velocities of about 10 cm s−1 are found over the open slope. Hydrographic and dynamical observations obtained from the summer and fall cruises from the surface to 300 m depth are presented below.

Figure 3.

Vertical sections of velocity in E–W direction (in cm s−1) for the fall and summer cruises taken upstream of the canyon (Section I in Figure 1: longitude 3.02°E). Negative velocities (indicating westward direction) are in solid lines, whereas positive velocities (indicating eastward direction) are in broken lines. Approximate ADCP profiles are marked.

3.1. Fall

3.1.1. Hydrography

[21] During the fall cruise, the upper water column over the canyon head (i.e., from about 50 to 100 m depth) is occupied by relatively cold saline dense water (Figures 4a and 4b). This causes an inversion of the horizontal density gradient over the canyon head and a relative minimum of density over the lower canyon at these depths (Figures 4a and 4b). At 120 m (Figure 4c), the density front is located over the midcontinental slope. It is characterized by isopycnals ranging from 28.70 to 28.90, and forms a horizontal density gradient of 0.2/0.1° in latitude. The front deepens toward the coast (Figures 4c and 4d). At an upstream location (Figure 5a), isopycnals ranging from 28.70 to 28.90 intersect the slope from 140 m to 215 m depth. The bottom of the thermocline, characterized by the 28.50 isopycnal, is located at about 100 m, tilting down toward the coast and intersecting the slope at 120 m (Figure 5a). By comparing the across-slope vertical section taken upstream of the canyon with a section taken along the canyon axis (Figure 5a), we see that the isopycnals corresponding to the density front and above (e.g., from 28.60 to 28.90) stretch when entering the canyon.

Figure 4.

November cruise. Horizontal density field at (a) 50 m, (b) 80 m, (c) 120 m, and (d) 150 m depths. CTD positions are marked.

Figure 5a.

November cruise. Density sections across the shelf break obtained upstream before entering the upper canyon (Section I: longitude 3.02°E) and along the canyon axis (Section II: longitude 2.85°E). Approximate CTD profiles are marked at the surface. The density front is highlighted in bold. Sections are indicated in Figure 1.

[22] We show two vertical sections taken along hydrographic sampling lines (Figure 5b): one over the midupper canyon (Figure 1b) and one taken 2.5 km out of the upper canyon mouth (Figure 1d). A third section interpolated over the upper canyon mouth is also shown (Figure 1c). The vertical section taken at the midupper canyon (Figure 5b, Section B) shows isopycnal uplift at the upstream canyon wall and downward tilting over the downstream canyon wall. A vertical section taken at the canyon mouth (Figure 5b, Section C) shows different tilting of the isopycnals above and below the shelf break. Above the shelf break, downward tilting is observed over the upstream canyon wall and isopycnal uplift is observed at the downstream canyon wall. Below the shelf break, the tilting reverses: isopycnals between 28.70 and 28.80 (from 150 to 200 m depth) show uplift toward the upstream canyon wall and downward tilting at the downstream canyon wall, similar to that observed at the midupper canyon. A vertical section taken off the upper canyon mouth (Figure 5b, Section D) shows uplift of the isopycnals <28.65 over the downstream canyon wall.

Figure 5b.

November cruise. Density sections taken across the midcanyon (Section B), canyon mouth (Section C), and 2.5 km from the canyon mouth (Section D). Approximate CTD profiles are marked at the surface. The density front is highlighted in bold. Sections are indicated in Figure 1.

3.1.2. Dynamics

[23] Velocity fields obtained at different depths are shown in Figure 6. At 20 m (Figure 6a) the current's path is rather complicated: over the lower canyon the Northern Current flows in a SW direction, whereas close to the shelf break a branch of the current flows onshore after passing the upper canyon (at 41.55°N, 2.70°E). A cyclonic tendency is observed near the coast over the canyon head (at 41.65°N, 2.85°E), whereas upstream of the upper canyon, over the continental shelf, there is an anticyclonic tendency (at 41.60°N, 3.05°E).

Figure 6.

Velocity vectors at (a) 20 m, (b) 100 m, (c) 150 m, and (d) 200 m depths corresponding to November 2003.

[24] At 100 m and below (Figures 6b, 6c, and 6d), a different flow pattern is observed near the canyon mouth: velocity vectors close to the bathymetric contours follow the isobaths entering the upper canyon along the upstream wall and leave the canyon over the downstream wall. Major current-bathymetry interaction at the canyon mouth is observed at the shelf break depth, i.e., at 150 m (Figure 6c). The Northern Current over the lower canyon continues flowing in a SW direction, leaving the canyon along the downstream canyon wall.

[25] The vorticity field at the shelf break depth (Figure 7a) is positive near the canyon mouth. Over the lower canyon, away from the shelf break, there is a negative vorticity band close to a positive vorticity band, both showing three anticyclones/cyclones respectively, of about 16 km in diameter. The vertical structure of the cyclonic vorticity patch near the canyon mouth (Figure 7b) shows maximum vorticity values of 0.3f at 150 m, i.e., at the shelf break depth (here f ∼ 10−4 s−1 is the Coriolis parameter).

Figure 7a.

Relative vorticity at 150 m corresponding to November 2003. Vorticity units are in 10−5 s−1. Positive vorticity is shaded.

Figure 7b.

Vertical sections of vorticity taken 2.5 km off the canyon mouth (Section D) and over the lower canyon (Section E) corresponding to November 2003. Approximate ADCP profiles are marked at the surface. Sections are indicated in Figure 1. Vorticity units are in 10−5 s−1. Positive vorticity is shaded.

[26] The vertical velocities obtained using control volume calculations (see Section 2.4) are negative above the thermocline (located at about 100 m depth during this cruise), with maximum negative vertical velocities of −60 m d−1 at 50 m (Figure 8). From the bottom of the thermocline down to 200 m, vertical velocities are positive, with a maximum of 27 m d−1 at 125 m. Below 200 m some inversions are observed.

Figure 8.

Vertical velocities inferred from the Total Vertical Flux (per unit of horizontal area) obtained from November 2003 using the control volume detailed in the text (bold line). Positive values indicate upwelling. Error bars are standard deviation values computed from sensitivity tests considering plus or minus one grid point at each side of the box used as control volume (see Figure 1).

3.2. Summer

3.2.1. Hydrography

[27] In summer the density front is once again located over the midcontinental slope, and at 100 m is characterized by isopycnals ranging from 28.75 to 28.90, which form a horizontal density gradient of 0.15/0.1° in latitude (Figure 9a). The front deepens toward the coast (Figure 9), and isopycnals ranging from 28.75 to 28.90 intersect the slope from 150 m to 230 m depth (Figure 10a). The bottom of the thermocline, characterized by the 28.55 isopycnal, is located at about 60 m depth. By comparing a cross-slope vertical section taken upstream from the canyon (Figure 10a) and a section taken along the canyon axis (Figure 10b), it can be seen that the isopycnals stretch when entering the canyon.

Figure 9.

Horizontal density fields corresponding to summer 2003 obtained at (a) 100 m, (b) 130 m, (c) 190 m, and (d) 220 m depths. CTD stations are marked.

Figure 10a.

Density sections across the shelf break corresponding to June 2005 taken (left) upstream before entering the upper canyon (Section I: longitude 3.02°E) and (right) along the canyon axis (Section II: longitude 2.85°E). Approximate CTD profiles are marked at the surface. Sections are indicated in Figure 1.

Figure 10b.

Density sections taken across the canyon head (Section A), midcanyon (Section B), canyon mouth (Section C), and 2.5 km from the canyon mouth (Section D). Approximate CTD profiles are marked at the surface. The density front is highlighted in bold. Sections are indicated in Figure 1.

[28] Vertical sections taken at the upper canyon (Figure 10b, Sections A and B) and at the canyon mouth (Figure 10b, Section C) show tilting of isopycnals above and below the shelf break. In these three sections, above the shelf break the distance between isopycnals 28.70 and 28.75 is large over the upstream canyon wall, decreasing westward, toward the downstream canyon wall. This suggests that water just above and at the shelf break depth coming from the adjacent upstream shelf, is stretched when crossing over the upper canyon, and then compressed when it reaches the downstream wall. Below the shelf break depth (from 170 to 250 m depth) the pattern is reversed, and isopycnal uplift is observed at the upstream canyon wall (see isopycnals ranging between 28.80 and 28.95) whereas downward tilting occurs at the downstream canyon wall.

[29] Horizontal sections of the density field show similar features. From 90 m to 170 m depth the upper canyon is characterized by denser water over the downstream wall with respect to the upstream wall (Figures 9a and 9b). Below 180 m depth, the pattern inverses, and from 200 m to 300 m depth, the upstream canyon wall shows higher density than the downstream wall (Figures 9c and 9d).

[30] In order to compare the density structure of the front in summer and fall, we consider the stratification parameter S = NH/fL, where N2 is the Brunt-Väisälä frequency, H is the water column width under consideration, f is the Coriolis parameter and L is half the canyon width. We consider an average of N obtained at two representative stations (Figure 11) at the depths that correspond to the density front, i.e., from 130 to 250 m, and avoid the influence of the seasonal thermocline. The stratification parameter for the density front in summer is Ssummer = 2.67 × 10−3s−1 × 120 m/10−4 × 4000 m = 0.80 and Sfall = 3.83 × 10−3s−1 × 120 m/10−4 × 4000 m = 1.15 for fall. This shows that the stratification is about 1.45 times larger in fall than in summer. Varying the upper (from 130 m to 150 m) and lower (from 230 m to 250 m) limits of the water column width considered in this computation, the difference in stratification ranges between 1.41 and 1.51 times larger for the fall cruise than for the summer cruise.

Figure 11.

Brunt-Väisälä frequencies obtained from a typical CTD profile in the summer and fall 2003 cruises.

3.2.2. Dynamics

[31] During the summer cruise the Northern Current flows over the canyon (Figure 12a). The current is less intense and wider than in fall. The near-surface flow is coherent with the flow below the thermocline (Figure 12b). At the shelf break depth (150 m), the core of the current is located over the lower canyon (Figure 12b), intersecting the slope at 180 m. Significant flow modification due to bathymetry is observed from 180 m to 300 m. At these depths, the flow tends to enter the canyon mouth along the upstream wall and exits along the downstream wall (Figures 12c and 12d).

Figure 12.

Velocity vectors at (a) 50 m, (b) 150 m, (c) 200 m, and (d) 300 m depths corresponding to June 2003.

[32] At 200 m, the vorticity field near the canyon mouth is mainly cyclonic (Figure 13a). A negative vorticity patch is observed on the slope upstream of the canyon. Offshore, the vorticity shows two pairs of vortices, which are similar to those observed in the fall cruise (Figure 7a). A vertical section taken near the canyon mouth shows the vertical structure of the cyclonic vorticity patch with maximum vorticity values of about ∼0.1f (Figure 13b).

Figure 13a.

Relative vorticity at 200 m corresponding to June 2003. Vorticity units are in 10−5 s−1. Positive vorticity is shaded.

Figure 13b.

Vertical sections of vorticity taken 2.5 km off the canyon mouth (Section D) and over the lower canyon (Section E) corresponding to June 2003. Approximate ADCP profiles are marked at the surface. Sections are indicated in Figure 1. Vorticity units are in 10−5 s−1. Positive vorticity is shaded.

[33] The net vertical velocities obtained from the control volume calculations (see Section 2.4) are negative over the thermocline, showing a maximum of −28 m d−1 at 30 m (Figure 14). Net vertical velocities are positive below the thermocline. Maximum positive values of about 40 m d−1 are found between 90 and 180 m. At 230 m, the vertical velocities show a relative minimum of 10 m d−1; this value continues down to 360 m.

Figure 14.

Vertical velocities inferred from the Total Vertical Flux (per unit of horizontal area) obtained for June 2003 using the control volume detailed in the text (bold line). Positive values indicate upwelling. Error bars are standard deviation values computed from sensitivity tests considering plus or minus one grid point at each side of the box used as control volume (see Figure 1).

4. Discussion

[34] In the fall cruise, at 20 m the circulation over the upper canyon is highly variable in direction. The current close to the shelf break flows onshore after passing over the canyon (Figure 6a). Closer to the coast, there is a flow reversal over the canyon head and an anticyclonic tendency farther upstream (latitude 41.6°N, longitude 3.1°E) (Figure 6a) which are also observed at 100 m. These features might be due to the geostrophic adjustment of local hydrographic features reported above. In one hand, the inverse density gradient observed over the canyon head (Figures 4a and 4b) would force an anticlockwise circulation. On the other hand, the density gradient established between the relative minimum of density over the lower canyon and the saltier open sea water (Figure 4a) would cause, because of geostrophic adjustment, an onshore flow over the canyon (Figure 6a). In previous studies in the area, hydrographic features have been reported to modify the local circulation. Rojas et al. [1995] observed an anticyclonic eddy over Blanes canyon associated with a relatively cold fresh water patch located above the seasonal thermocline. More recently, Rubio et al. [2005] observed an anticyclonic structure, which was 45 km in diameter, reaching 100 m depth, and well below the seasonal thermocline, which modified the local flow.

[35] According to the observations from the two cruises, the Northern Current is modified by Blanes canyon. Flow modification is shown as a deflection of the current along the canyon walls. Major modification is observed at the shelf break (at 150 m) in fall (Figure 6) and below the shelf break (at 200 m) in summer (Figure 12).

[36] Over the lower canyon, the Northern Current flows in a SW direction, and leaves the canyon along the downstream canyon wall (Figure 6). Alternating negative/positive vortices (Figure 7) of about 16 km in diameter (close to the Rossby radius of deformation) are possibly related to horizontal velocity intensification of the Northern Current (Figure 6). As this occurs away from any topographic boundary it is not expected to be related to a topographic effect.

[37] The volume flux estimates obtained from ADCP data show that there is upwelling inside Blanes canyon (Figures 8 and 14). The upwelling is at its maximum near the shelf break depth, and extends from the bottom of the thermocline (at about 100 m) down to the density front (at about 200 m) (Figures 5a, 5b, 10a, and 10b). Independent data provided by CTD hydrographic casts show tilting of isopycnals along the canyon mouth and within the canyon (Figures 5b (Sections C and D) and 10b (Section D)). Since this tilting and the associated isopycnal stretching is relevant for vorticity dynamics when they are the result of the interaction between the large-scale flow and topography, it is important to eliminate other possible origins, particularly the possibility that isopycnal perturbation is related to internal baroclinic tides. Baroclinic tides might become important in stratified flows in places of abrupt bathymetric changes under the forcing of relatively large barotropic tides [Baines, 1982]. They can be further amplified within submarine canyons [Baines, 1983; García-Lafuente et al., 1999], which is a possibility that cannot be rejected or confirmed in the canyon under study because of the lack of suitable observations. There are, however, reasons suggesting that baroclinic tides are not the source of the observed isopycnal distribution within the canyon. Far from straits and narrow passages, barotropic tidal currents in the interior of the western Mediterranean Sea are negligible [Albérola et al., 1995b]; therefore the barotropic forcing needed is not expected to occur nearby Blanes canyon. The harmonic analysis of the velocity series measured at 375 m depth by the instrument placed at the upper canyon mouth (Figure 2b) gave a major semiaxis less than 0.3 cm s−1 for M2 and 0.4 cm s−1 for K1 (just about the signal-to-noise ratio), which is clearly insufficient to force baroclinic tides. Moreover, the similarity of the isopycnal spatial pattern found in both cruises does not point to a fast time-varying force, such as tides, as the underlying mechanism. Furthermore, the complex spatial pattern of this isopycnal tilting with isopycnals sloping upward (looking downstream) in the upper portion of the canyon and downward beneath (Figures 10b (Sections C and D) and 5b (Sections C and D) to a lesser extent) would require the existence of higher and energetic baroclinic modes. Therefore, even when the definitive rejection of internal tides as a possible contributor to the observed isopycnal distribution needs more robust confirmation, the aforementioned arguments have prompted us to discard this mechanism and to assume that the cause of the distribution is the flow-topography interaction as the large-scale flow passes over the Blanes canyon.

[38] In the two cruises, cyclonic vorticity is observed at the upper canyon mouth (Figures 7 and 13). In both cruises, water from the adjacent shelf stretches when flowing over the canyon (Figures 5a, 5b, 9a, 9b, 10a, and 10b). According to these observations the main source of positive vorticity is the stretching of water flowing from the adjacent shelf: owing to conservation of potential vorticity these fluid parcels turn cyclonically when entering the canyon. In the summer cruise, water coming from the adjacent shelf flows over the canyon (Figures 9a and 9b). From 90 m to 170 m the upper canyon is occupied by lighter water over the upstream wall and denser water over the downstream wall (Figures 9a and 9b). Density sections across the canyon mouth show that isopycnals from 28.70 to 28.75 are stretched over the upstream wall, and are compressed westward, over the downstream wall (Figure 10b (Sections C and D)). These observations suggest that local downwelling/upwelling occurs over the upstream/downstream upper canyon walls respectively. At these depths (i.e., roughly from 100 m to 200 m), according to the net vertical velocity estimates there is maximum net upwelling of about 40 m d−1 (Figure 14). Below 200 m depth the tilting inverses, which suggests localized upwelling/downwelling over the upstream/downstream canyon walls respectively. A similar discussion applies for the fall cruise, although in this case the bottom limit of the net upwelling is not well defined, showing a significant decrease in vertical velocity below the shelf break (i.e., below 150 m) and inversions below 200 m (Figure 8).

[39] In spin-up laboratory experiments performed on the large Coriolis turntable [Boyer et al., 2006] upwelling observed in canyons with right-bounded flows is explained as follows. The flow behavior in the vicinity of the canyon in a narrow region above and below the shelf break is governed by two basic processes. First, fluid parcels advected over the canyon are subjected to stretching of vortex lines passing over the shelf break and into the canyon topography [Hickey, 1997]. Owing to conservation of potential vorticity these fluid parcels turn cyclonically when entering the canyon. The second mechanism occurs near the canyon mouth and involves fluid particles roughly following the canyon depth contours. For right bounded flows, as in the present case, horizontal shear layer instability would lead to anticyclonic vorticity generated near the canyon mouth [Boyer et al., 2006]; therefore, these two processes compete near the canyon mouth. Far from the canyon mouth, i.e., over the midupper canyon, the effect of flow-topography interaction near the canyon mouth is very small. Just below the rim, the strong cyclonic vorticity acts like a pump, taking water away over the downstream canyon wall. Because of mass conservation, and without lateral inflows due to the presence of the canyon walls, some flow coming from below is needed to compensate this loss of mass; this consequently forces an upwelling inside the canyon which might even reach the adjacent shelf [Boyer et al., 2006]. By performing a numerical model experiment that calculates all the terms involved in the vertical component of vorticity it may be possible to resolve the complex dynamics in Blanes canyon. In particular, the cause of upwelling should be further investigated using numerical methods.

4.1. Differences Between the Fall and Summer Studies

[40] The interpretation of the results obtained from the summer cruise follow similar lines to those discussed for the fall cruise. However, some differences have been observed between the cruises. The Northern Current flow is larger in the fall cruise than in the summer cruise; the positive vorticity values observed at the canyon mouth are larger in the fall cruise than in the summer cruise; and the flow modification in fall occurs at shallower depths than in summer. Differences in stratification and in the location of the current incident on the canyon, the Northern Current, with respect to the shelf break explain some major differences observed.

[41] Because of geostrophic balance, the stronger density gradients that characterize the density front in the fall cruise (Figure 11) explain the larger flow velocities observed in the fall cruise compared to the summer cruise. The stronger flow associated with the Northern Current in fall with respect to summer forces a larger amount of water into the canyon, therefore increasing the positive stretching vorticity compared to summer.

[42] The location of the Northern Current with respect to the upper canyon explains the variation in depth of the maximum current-bathymetry interaction: the current in fall is deflected at the shelf break depth, whereas in summer the flow modification due to the canyon is observed below the shelf break. In both cases the Northern Current flows over the entire canyon, although in fall the core of the current is located over the shelf break, whereas in summer it is located slightly offshore. Consequently, maximum current topography interaction occurs at shallower depths in fall than in summer.

4.2. Comparison With Other Field, Numerical, and Laboratory Data

[43] Some similarities occur between Blanes canyon and US West coast canyons [Hickey, 1997]. In particular, cyclonic vorticity dominates within the canyon, both in upwelling and downwelling wind conditions. Hickey explains that the cyclonic vorticity is caused by the water layer being stretched in two different ways: it may fall into the canyon, or stretch gradually as it flows over the canyon.

[44] Right bounded flows simulated in general numerical and laboratory models [Klinck, 1996; Boyer et al., 2006], as well as in numerical models of NW Mediterranean canyons [Skliris et al., 2001; Jordi et al., 2005] are also in good agreement with the data presented here. Results show similar fields of velocity, flow deflection at the shelf break depth and positive vorticity inside the canyon. Quantitative agreement with results near the canyon mouth is found with the vorticity distribution obtained by Jordi et al. [2005]. On the other hand, in the laboratory model by Pérenne et al. [2001], for right bounded flow experiments, the authors obtained negative relative vorticity within the canyon. This is due to the dominance of negative shear vorticity generated at the mouth of the canyon, which was advected toward the head of the canyon.

[45] In previous studies, upwelling and downwelling over submarine canyons have been explained as the consequence of the unbalanced pressure gradient existing inside canyons [Freeland and Denman, 1982]. For left bounded flows, the unbalanced pressure gradient causes an in-canyon flow and therefore upwelling [Hickey, 1997; She and Klinck, 2000]. For right bounded flows (as in our case) the unbalanced pressure gradient would cause downwelling, contrary to our observations. Our data shows different flow directions on both walls of the upper canyon mouth, which suggests a complex pattern of the pressure gradient distribution, similar to that observed in Astoria canyon, except during active upwelling events [Hickey, 1997].

[46] Recent large-scale laboratory experiments give evidence of right bounded flows causing local upwelling in submarine canyons [Boyer et al., 2006]. Time mean flows, shown by Boyer et al. [2006, Figure 7], show that canyons are dominated by cyclonic vorticity, with a small vertical extent. At the shelf break level and below, the flow follows the canyon depth contours, entering the canyon along the upstream wall and leaving the canyon along the downstream wall. Upwelling occurs at these depths: maximum upwelling is observed at the shelf break depth inside the canyon, decreasing in intensity with depth. These observations are in agreement with the field data analyzed in the present work.

[47] In the paper by Boyer et al. [2006] upwelled water is able to reach the adjacent continental shelf [see Boyer et al., 2006, Figure 11]. This feature is not observed in the density fields obtained from the fall and summer cruises: the upwelling inside the canyon occurs below the shelf break without reaching the nearby continental shelf. These observations are in agreement with the study done by Klinck [1996], in which canyons with right bounded flows showed weak shelf-slope exchange.

[48] The control volume calculations used in this study only give the net upwelling/downwelling tendency, and do not show the preferred regions where upwelling occurs. However, the density distribution obtained during the summer cruise suggests that downwelling/upwelling preferred locations are situated over the upstream/downstream canyon walls, respectively. This would be in agreement with Klinck [1996] and Jordi et al. [2005], in which numerically simulated canyons with right bounded flows have a downwelling region on the upstream wall and an upwelling region at the downstream wall. In these numerical studies, the transport values obtained in the upstream and downstream wall regions are quite similar. However, Klinck [1996] always reports a net downwelling (unfortunately, Jordi et al. [2005] do not give net values for the steady circulation). The only explanation for the difference between the net downwelling reported by Klinck [1996] and the net upwelling values obtained from the field data is the particular topography of Blanes canyon, although this hypothesis needs to be tested numerically. To the authors' knowledge, Ardhuin et al. [1999] performed the only numerical study in which a realistic topography of Blanes canyon was used. Ardhuin et al. [1999] explored the response of Blanes canyon to different 1 day wind bursts of 10 m s−1, and found a highly ageostrophic response, which resulted in an upwelling/downwelling depending on the wind direction.

[49] Upwelling in submarine canyons has important biological implications, as reported in several studies on canyons along the western Canada coast under upwelling favorable winds [e.g., Allen et al., 2001]. However, the bio-physical coupling over abrupt topography is complex (see the review by Genin [2004]). According to the author, to enhance population growth, the upwelled water should reach the photic layer and remain for enough time to elevate the phytoplanktonic biomass. When considering the influence of the upwelling reported in this contribution to the local biology patterns some peculiarities should be taken into account: the uppermost vertical limit of the upwelling only extends to the seasonal thermocline (at about 100 m), and the upwelling is limited to the upper canyon entrance (i.e., it does not reach the lower canyon).

5. Summary and Conclusions

[50] A study of the flow in Blanes canyon using data from two oceanographic cruises has been conducted. Cyclonic flow deflections were observed in the vicinity of the shelf break near the canyon entrance in both the fall and summer cruises. These deflections are due to columns of shelf water stretching when they pass over the shelf break. General numerical models and laboratory experiments, field studies off the US West coast, and numerical models from other Mediterranean canyons are in good agreement with these observations.

[51] There are some differences between the two cruises studied. In the fall study the current is stronger and flows closer to the head of the canyon. The flow is modified at the shelf break depth in fall, whereas in the summer cruise this occurs slightly deeper. Vorticity values are stronger in the fall cruise than in the summer cruise. The differences observed between the two cruises are explained by the differences in stratification and in the location of the current with respect to the shelf break.

[52] Upwelling has been reported in a canyon with right bounded flow. Evidence is given that in Blanes canyon upwelling occurs inside the canyon at the shelf break depth, which extends in the vertical from the seasonal thermocline (at about 100 m) down to the shelf-slope front (at about 200 m). There is no evidence that upwelled water reaches the continental shelf. The results obtained in the laboratory model employed by Boyer et al. [2006] in the large-scale Coriolis turntable are in general agreement with these observations, except that in the laboratory model the upwelled water reaches the shelf.

[53] While this work contributes to shed some light on the circulation and dynamics of Blanes canyon, limitations of the sampling strategy over the upper canyon require additional efforts in order to solve the complex dynamics involved.

Acknowledgments

[54] The authors sincerely acknowledge B. Hickey, one anonymous reviewer, and editor R. Weisberg for their useful comments which largely improved the original manuscript. The authors thank the professional collaboration of the R/V Garcia del Cid crew members and UTM technicians. The authors acknowledge M. Canals for kindly providing the high-resolution bathymetry of Blanes canyon; the Catalan Meteorological Office (SMC) and G. Jordà are acknowledged for providing the meteorological data sets. I. A. Catalan and J. Garcia Lafuente are acknowledged for kindly discussing the possible biological implications of this work on pelagic species and for fruitful discussions on the role of tides in the Mediterranean Sea, respectively. The authors are grateful to the Environmental Fluid Dynamics program of ASU and his director, H. J. S. Fernando, for partially supporting the visit of M.M.F. at ASU. M.M.F. acknowledges a Juan de la Cierva postdoctoral fellowship at IMEDEA (CSIC-UIB). This work has been performed in the framework of RECS project (REN2002-04556-C02-01/MAR) funded by the Spanish Ministry of Science, and it was partially funded by the U.S. National Science Foundation under grant OCE-0527940. The authors wish to thank PRECARIOS, a postgraduate's organization whose dedication and effort contribute to improve young scientists working conditions in Spain (www.precarios.org).

Ancillary