Comparison of horizontal and downward particle fluxes across canyons of the Gulf of Lions (NW Mediterranean): Meteorological and hydrodynamical forcing

ARTICLE IN PRESS

Continental Shelf Research 28 (2008) 1957– 1970

Contents lists available at ScienceDirect

Continental Shelf Research

0278-43

doi:10.1

� Corr

E-m1 Pr

Arendal

journal homepage: www.elsevier.com/locate/csr

Comparison of horizontal and downward particle fluxes acrosscanyons of the Gulf of Lions (NW Mediterranean):Meteorological and hydrodynamical forcing

J. Bonnin a,�, S. Heussner a, A. Calafat b, J. Fabres b,1, A. Palanques c, X. Durrieu de Madron a,M. Canals b, P. Puig c, J. Avril a, N. Delsaut a

a CEFREM, UMR 5110 CNRS, University of Perpignan, 52 Avenue Paul Alduy, 66860 Perpignan Cedex, Franceb GRC Geociencies Marines, Department of Estratigrafia, P. i Geociencies Marines, University of Barcelona, C. Marti i Franques s/n, 08028 Barcelona, Spainc Institut de Ciencies del Mar CSIC, Passeig Maritim de la Barceloneta 37-49, 08003 Barcelona, Spain

a r t i c l e i n f o

Article history:

Received 14 June 2007

Received in revised form

5 May 2008

Accepted 6 June 2008Available online 19 June 2008

Keywords:

Particle flux

Sediment transport

Advection

Temporal variability

Event-dominated

Cascading

Continental margin

NW Mediterranean

43/$ - see front matter & 2008 Elsevier Ltd. A

016/j.csr.2008.06.004

esponding author. Tel.: +33 540008857.

ail address: [email protected] (J.

esent address: UNEP Shelf Programme, GRID-

, Norway.

a b s t r a c t

In the framework of the European project EUROSTRATAFORM, an array of six near-bottom mooring lines was

deployed in the heads of the major submarine canyons incising the continental slope of the Gulf of Lions

(NW Mediterranean). All moorings were equipped with sediment traps, current meters and turbidity

Optical Backscatters Sensors (OBS) situated at few meters above the bottom. Particulate mass fluxes

were recorded weekly by the sediment traps between November 2003 and May 2004 and compared

with horizontal export fluxes obtained from the turbidity and current meters.

Results show that near-bottom vertical particulate mass fluxes generally increase along slope

between the northeastern-most and the southwestern region of the Gulf of Lions. Horizontal export

fluxes also show some increase southwestward with flux in the southwestern-most canyon up to two-

orders of magnitude higher than in northeast canyons. Furthermore, most of the export flux is driven by

energetic and short events such as marine storms and dense shelf water cascading. Although the

consequence of storms and cascading was obvious from traps and OBS measurements, substantial

differences between vertical and horizontal fluxes appeared. While sediment traps recorded the largest

vertical flux during the autumn storm (4 December) with contributions from river discharge and

resuspended material, horizontal flux as obtained from OBS data was highest in winter (21–24

February). We believe that such discrepancies are partly due to measurement artifacts originating from

the particle size effects and/or the different height above the bottom occupied by the traps and OBS

which complicate comparisons between vertical and horizontal fluxes. However, different water-

column conditions (stratified versus unstratified) together with the hydrodynamic regime during the

two periods appear as a likely explanation for the flux differences between fall and winter storms.

& 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Continental margins are complex regions where diversephysical, chemical, biological and sedimentological interactionswith the continent, the open ocean and the atmosphere occur.These interactions and the related processes play a fundamentalrole in the global biogeochemical cycles as continental marginsact on the one hand as a sink for particulate matter supplied bycontinental inputs and biological production, and on the otherhand as a source for the adjacent open ocean. Therefore, assessing

ll rights reserved.

Bonnin).

Arendal, Postboks 183, 4802

the functioning of shelves and slopes in the transfer of particulatematter to the deep ocean is necessary to estimate their influencein the global carbon cycle (Liu et al., 2000; Muller-Karger et al.,2005).

Large multidisciplinary research programs have been con-ducted in the recent past to determine the exchange fluxes fromthe productive shelf to the deep ocean, e.g., SEEP I (Walsh et al.,1988) and SEEP II (Biscaye et al., 1994) in the Middle AtlanticBight, ECOMARGE in the Gulf of Lions (Monaco et al., 1990), OMEXI at the Goban Spur on the NW European margin (van Weering etal., 1998), STRATAFORM on the Eel Margin (Walsh and Nittrouer,1999) and MATER in the Alboran Sea (Fabres et al., 2002). All theseprograms have indicated that cross-slope advection contributessubstantially to the flux of biogenic debris to the sediments on theslope. Such advection occurs via layers of enhanced particleconcentration called nepheloid layers that can be found in the

ARTICLE IN PRESS

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–19701958

vicinity of the bottom or at intermediate depths where theydetach from the shelf break and slope and spread along isopycnalsurfaces (McCave, 1986; Puig and Palanques, 1998; McCave et al.,2001). The main source of particles forming those nepheloidlayers is resuspended sediment by the action of waves, bottomcurrents or internal waves (Biscaye and Anderson, 1994; vanWeering et al., 1998; Monaco et al., 1999; Durrieu de Madron et al.,1999a; Bonnin et al., 2002). Several studies based on sedimenttraps and seabed investigations have shown that sediments canquickly reach the slope and abyssal regions. Walsh and Nittrouer(1999) have shown that variability of sediment-trap flux wascontrolled by a combination of shelf sediment resuspension, riverdischarge and margin circulation. Furthermore, a recent studyconducted on the Louisiana Shelf and the adjacent Mississipicanyon showed that labile sedimentary organic matter originatingfrom the Mississippi river and recently deposited on the shelf wastransported rapidly down the canyon after the passage ofhurricane Ivan in October 2004 (Bianchi et al., 2007), highlightingthe importance of canyons in river-dominated margins for theglobal carbon budget.

This study was part of the EUROSTRATAFORM project carried out indifferent European continental margins, in order to (1) quantifythe export flux of particulate matter from the shelf to the slope,(2) identify the principal mechanisms controlling the shelf-slopesediment exchange, and (3) examine how modern sediment-transport processes may influence the formation of the strati-graphic records on continental margins. The study we present inthis paper has been carried out on the margin of the Gulf of Lions(GoL), which is indented by several canyons running from theshelf break to the continental rise. Previous studies conducted inthe region have highlighted the spatial and temporal variability ofparticulate fluxes but were not able to resolve in detail suchvariability. We present here vertical fluxes from sediment trap

Têt

Aude

Hér

ault

Orb

-

--Tech

ylgA

Mosson

Lez

Vidourle

3° 4°

42°

43°

Cap de Creus

Cap Creus

Lacaze-Duthiers

Aude

Herault

02

40608090

200

1000

1500

200

600

Canyon head (300 m) short mooring

Mid-canyon (1000 m) long mooring

100

Fig. 1. Map of the study area with the position of the mooring lines in the head of the su

are schematic representations of the mooring lines: the short ones in the canyon head

data and horizontal fluxes from near-bed moorings measuredbetween fall and early spring in the major canyons of the GoL anddiscuss the mechanisms responsible of the flux variability. Theresults provide a synoptic view of particle export fluxes and allowa more robust assessment of the mechanisms, such as marinestorms and shelf dense water cascading responsible of most of theparticulate matter export in this region. In particular, the resultssuggest that differences between vertical and horizontal fluxes arerelated to the stratification of the water column.

2. Background

The GoL is a river-dominated continental margin. The shelf iscrescent-shaped, and the upper slope indented by numeroussubmarine canyons (Fig. 1). The Northern Current flows along theslope as a part of the cyclonic circulation of the westernMediterranean basin (Millot, 1999). A density front is formedbetween the low-salinity shelf water and more saline Mediterra-nean seawater, and this front influences shelf-slope exchanges(Durrieu de Madron et al., 1999). In winter, cold and drycontinental winds as well as a large air–sea temperaturedifference induce an important heat loss and evaporation fromthe sea. The sustained cooling and mixing of the surface layerleads to a mixing of the water column and the formation of densecold waters on the shelf that cascade down the slope (Durrieu deMadron and Panouse, 1996; Tusseau-Vuillemin et al., 1998; Dufauet al., 2004). The distribution of the suspended particulate matterconcentration shows a strong seaward decreasing gradient(Durrieu de Madron et al., 1990; Durrieu de Madron and Panouse,1996; Heussner et al., 2006) as a result of the large river inputsand the general circulation in the GoL. Furthermore, Durrieu deMadron et al. (1990) and Lapouyade and Durrieu de Madron

ArlesGd Rhône

Pt R

hône

Petit-Rhone

Grand-Rhone

Planier

0

AtlanticOcean

GoL

MediterraneanSea

TrapPPS3 560mabRCM 7/8558 mab

Floats

Floats

Floats

Trap PPS330 mabRCM 7/828 mab

Release

BallastLong line

(mid-canyon)Short line

(canyon head)

Trap PPS330 mab

RCM 7/828 mab

Floats

Release

RCM9/115 mab

Floats

5°E

bmarine canyons (stars) and at mid canyon depth (squares). On the right of the map

s and the long one deeper down.

ARTICLE IN PRESS

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–1970 1959

(2001) estimated that the horizontal transport of suspendedparticulate matter advected by the permanent along-slope currentincreases significantly downstream between the updrift (north-east, Planier) and downdrift (southwest, Cap de Creus) end of theGoL.

Particulate matter reaching the continental margin comes fromvarious sources. Primary sources are continental riverine dis-charge (dominated by the Rhone River), atmospheric (mainlySaharan dust deposition) and biological production. Resuspensionof sediment on the shelf and slope during energetic events alsocontributes to the particle budget on the shelf and slope. Due tothe predominant river inputs, suspended particle concentrationsgenerally decrease seaward. The distinct surface nepheloid layer,which extends over the shelf edge and slope, is primarilycomposed of mineral and biological material (Durrieu de Madronet al., 1990; Monaco et al., 1999). The bottom nepheloid layer(BNL), which extends to the shelf-edge and onto the upper slope(hence connecting the inner shelf to the submarine canyon heads),is mostly lithogenic (Monaco et al., 1999; Durrieu de Madron et al.,1999).

On a seasonal basis, total mass fluxes (TMF) are much largerduring winter and seem to covary with major sources ofparticulate matter (river discharge) and physical forcing (winterconvection, along-slope current intensity and meso-scale varia-bility) (Monaco et al., 1990, 1999; Durrieu de Madron et al., 1999).Annual mass fluxes are 3–10 times higher at the southwesternend of the GoL than at its northeastern end (Heussner et al., 2006).The limited variation of the chemical composition (majorconstituents) of the material collected in sediment traps suggeststhat the particle reservoir feeding shelf-slope exchanges is fairlyhomogenous at the scale of the GoL. Although particles advectedto the slope environment are predominantly lithogenic, theproportion of ‘‘fresh’’ organic matter remains high, particularlyat the southwest end of the GoL (Heussner et al., 2006).

During floods and major southeastern (marine) storms, riverplumes and shelf waters are mixed and dominantly advectedcyclonically along the shelf and exported at the southwestern endof the GoL (Ulses et al., 2008b). Dense water cascading has thepotential to transport massive amounts of particulate materialfrom the shelf to the deep sea (Canals et al., 2006; Heussner et al.,2006). This latter process appears to have large interannualvariability (Bethoux et al., 2002). Additional observed mechan-isms involved in sediment transport across the shelf break of theGoL include marine storm-generated downwelling (Palanques etal., 2006; Ulses et al., 2008b), shallow flows occasionallypenetrating over the shelf under the effect of wind (Millot,1990; Estournel et al., 2003) and current instabilities (Durrieu deMadron et al., 1999; Flexas et al., 2002; Petrenko et al., 2005).

Table 1Summary of moorings information

Mooring site Latitude (1N) Longitude

(1E)

Water

depth

(m)

Current

meters

depths (m)

Tr

de

(m

Cap Creus 42123.513 3119.239 302 297 27

Lacaze-Duthiers 42135.762 3123.841 309 304 27

Aude 42145.977 3140.361 309 304 27

Herault 42153.198 3159.059 302 297 27

Petit Rhone 42159.309 4128.566 302 297 27

Grand Rhone 42158.800 4147.974 301 296 27

Planier 43107.150 5112.870 307 302 27

Lacaze-Duthiers 42125.700 3132.650 1000 972 97

– – – 1000 502 50

Planier 43102.320 5111.270 1000 972 97

– – – 1000 502 50

3. Material and methods

3.1. Instrumental setup

Our observations were made during two-consecutivedeployments of six mooring lines, equipped with near-bottomsediment traps and current meters, in the heads of the majorsubmarine canyons of the GoL (Fig. 1; Table 1), from northeastto southwest: Planier (PLC), Petit Rhone (PRC), Herault (HEC),Aude (AUC), Lacaze-Duthiers (LDC) and Cap Creus (CCC) (Fig. 1).One mooring was deployed in the Grand Rhone Canyon, butunfortunately an instrument failure prevented the recoveryof data. The mooring lines were deployed at �300 m waterdepth and worked synchronously between November 2003 andMay 2004 with a gap of 4–9 days between the two deployments.All moorings were equipped with one trap positioned at28 m above bottom (mab). Traps were programmed to collectmaterial over periods of 7 days. All traps (Technicap PS3/3)were cylindro-conical (0.4 m diameter, 2.5 aspect ratio, 12receiving cups; Heussner et al., 1990), and coupled with AanderaaRCM9/11 current meters equipped with a turbidimeter (OpticalBackscatter Sensor, hereafter OBS) positioned at 5 mab. Thesampling interval of the current meters and OBS was set to20 min. Turbidity data recorded in formalin turbidity unit (FTU)were converted into suspended sediment concentrations usingthe ‘‘best-fit’’ calibration curves of Guillen et al. (2000) usingsurface sediment from the study area, and used to calculateintegrated horizontal fluxes (Fig. 4). However, one should beaware that such calibration can only provide an estimation of thehorizontal flux as conditions during this study (e.g., grain size) canbe different than those prevailing during the calibration work ofGuillen et al. (2000).

The Technicap PPS3 sediment traps are used for studies ofparticulate matter fluxes on continental margins (Monaco et al.,1999; Heussner et al., 1999; Bonnin et al., 2002, 2006). Trapsduring a few deployments did not function properly and did notcollect settling material: CCC (first deployment), AUC (seconddeployment) and PLC (second deployment). All current metersand turbidity sensors functioned properly except the HECturbidity sensor during the first deployment.

In addition, two long-term lines deployed at greater depths(�1000 m) in the PLC and LDC, provided TMF and current–velocitydata from 1 November 2003 to 31 May 2004. These lines are partof a long-term monitoring (since 1993). They are equipped withone additional sediment trap and current meter pair positioned at500 mab (Fig. 1). The sample collection interval was set between29 and 31 days for the traps, and 1 h for the current meters. Thelist of all moorings is presented in Table 1.

ap

pth

)

Start first interval

(dd-mm-yy)

End last interval

(dd-mm-yy)

Duration of

intervals (d)

No. of

complete

intervals

2 10 November 2003 5 May 2004 7 12

9 10 November 2003 5 May 2004 7 20

9 10 November 2003 5 May 2004 7 12

2 10 November 2003 5 May 2004 7 24

2 10 November 2003 5 May 2004 7 24

1 10 November 2003 5 May 2004 7 0

7 10 November 2003 5 May 2004 7 12

0 1 November 2003 15 May 2004 29–31 3

0 1 November 2003 15 May 2004 29–31 6

0 1 November 2003 15 May 2004 29–31 5

0 1 November 2003 15 May 2004 29–31 6

ARTICLE IN PRESS

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–19701960

3.2. River inputs

In the GoL, the freshwater and particulate matter inputs aredominated by the Rhone River, in the north-eastern part of thestudied area, which delivers 480% of the total (sediment andwater) discharge (Aloısi et al., 1977). Other rivers flowing into theGoL have typical Mediterranean behavior with low averagesediment input. However, during intense and abrupt floodepisodes, those rivers can reach particulate matter dischargecomparable to the Rhone River’s base flow. Variations of thesediment discharge of the Rhone River during the experimentalperiod were used as a reference for sediment input to the system(Bourrin et al., 2006). Data on water discharge and SSC weremeasured at Arles (50 km upstream of the Rhone River mouth) bythe Compagnie Nationale du Rhone. In addition, Tet Riversediment discharge was also calculated using the rating curveestablished by Serrat et al. (2001) to provide an idea of the inputfrom the smaller rivers.

3.3. Total mass fluxes: sediment trap samples processing

Heussner et al. (1990) gave a complete description of the PPS3sediment trap and the processing of the trap samples. Thepolypropylene receiving cups of the traps were filled up beforedeployment with a buffered 5% (v/v) formaldehyde solution in0.45mm filtered seawater. This poisoning solution limits degrada-tion of trapped particles and prevents the mechanical disruptionof swimming organisms that occasionally enter the traps duringsampling. After recovery, the cups were stored in the dark at2–4 1C until they were processed in the laboratory, within amaximum delay of a few months. Swimmers, which were not veryabundant, were removed in a two-step procedure. After decanta-tion of the supernatant, particles were wet-sieved through a 1 mmnylon mesh in order to retain the largest organisms. Swimmerssmaller than 1 mm were manually removed under a dissectingmicroscope using fine tweezers. The original samples were thenprecisely divided into aliquots for subsequent analyses using arotary splitting method (Heussner et al., 1990). Sample dry weightwas determined on four aliquots filtered onto 0.45mm Milliporefilters, rinsed with distilled water to remove salts, and dried at40 1C for 24 h. Overall precision for total mass flux derived from aprevious experiment is around 7% (Heussner et al., 1999).

3.4. Calculation of horizontal fluxes

For each mooring site, instantaneous suspended sedimentfluxes in the canyon heads were calculated by multiplying thecurrent–velocity in the canyon axis by the SSC. Instantaneous andcumulative suspended sediment fluxes over the whole experi-mental period were presented in Palanques et al. (2006). Here, forease of comparison with TMF, the calculated horizontal fluxeshave been integrated over the sediment trap sampling intervals(7 days).

3.5. Determination of siliciclastic and carbonate fraction

Fabres et al. (2008) describes in detail the methodology used todetermine the organic carbon (Corg), total carbon (Ctot) andbiogenic silica (BSiO2) concentrations in the samples collected bythe PPS sediment traps. Briefly, replicate subsamples fromsediment traps were filtered onto Whatman GF/F filters, rinsedwith distilled water, and oven-dried at 40 1C. Total carbon wasanalyzed by combustion in a LECO CN 2000 analyzer. Organiccarbon was analyzed in the same way after progressive, controlledacidification with 2 N HCl to remove carbonate. Organic matter

content was estimated as twice the organic carbon content.Carbonate content was calculated assuming all inorganic carbon(Ctot–Corg) was CaCO3 and using the molecular mass ratio 100/12(Monaco et al., 1990). After filtration onto 0.45mm Milliporefilters, biogenic silica content was measured following the methodof Mortlock and Froelich (1989). Finally, siliciclastic fractionabundance was calculated by difference (Eq. (1)), assuming thatno other major constituent besides organic matter, calciumcarbonate and biogenic silica was present in the samples. Itincludes quartz, feldspars, heavy minerals and aluminosilicate,and is generally referred to as the lithogenic fraction.

% siliciclastic fraction

¼ 100%� ð% organic matterþ % CaCO3

þ % BSiO2Þ. (1)

As the temporal variability of flux must be taken into account inthe computation of the means, a flux-weighted mean content wascomputed for the siliciclastic and carbonate fractions and eachtrap by weighting individual sample concentrations by theircorresponding mass fluxes: Cfw ¼

PCiFi/

PFi, where Cfw is the flux-

weighted content for a given element, and Ci and Fi are themeasured content and mass flux for sample i. This way ofcomputing a mean yields a ‘‘true’’ average value equivalent towhat would have been obtained if settling particles had beencollected as a single sample during the averaging period.

4. Results

4.1. River discharge of suspended sediments

The average Rhone River discharge for the whole observationperiod was 1.3�103 m3 s�1 (average SSC ¼ 0.45 g l�1) and reached9.8�103 m3 s�1 during the 4 December flood (peak SSC duringthat event: 2.82 g l�1). Sediment supplied by the Rhone during theDecember storm was 4.7�106 t. For comparison, the Tet Riverwater discharge was usually lower than 4�102 g s�1 during thestudy period although a flood occurred on 4 December with anaverage daily discharge of 4�104 g s�1 (Fig. 2). The SSC rangedbetween 0.02 and 0.8 g l�1. Guillen et al. (2006) estimated theamount of sediment supplied by the Tet to the sea during theflood to be about 2�104 t and the total sediment supplied overthe study period (December 2003 to mid March 2004) under non-flood conditions to be about 3�103 t.

4.2. Total mass fluxes

The range of TMF measured at the different mooring sitesduring the study is large, spanning over three-orders of magni-tude: from 0.6 to 109.5 g m�2 d�1 for individual sampling intervals(Fig. 3). Higher values are in the upper range of typical TMF valuesgenerally measured on continental margins. Biscaye and Anderson(1994) measured TMF (15 days average) at 400 m (8 mab) in theMiddle Atlantic Bight between 0.1 and 18 g m�2 d�1. Walsh andNittrouer (1999) measured TMF (6–10 days average) on the Eelcontinental margin at 450 m (15 mab) between 4 and 24 g m�2 d�1

and Bonnin et al. (2005) measured TMF (11 days average) on thecontinental slope of the Faeroe-Shetland Channel of up to90 g m�2 d�1 at 800 m (2 mab). For the NW Mediterranean, therange of values observed here are typical of TMF values foundduring previous studies (Monaco et al., 1990, 1999; Puig andPalanques, 1998; Heussner et al., 2006), and they clearly are muchhigher than those measured in deep sea such as those in theLigurian Sea (e.g., Miquel et al., 1994).

ARTICLE IN PRESS

Fig. 2. Daily sediment discharge (solid line) for the Rhone and Tet Rivers. The

dashed line shows the sediment discharge integrated over trap intervals. Note the

large difference between the two rivers. The sediment discharge was calculated

using the daily water discharge measured at Arles (50 km upstream of the river

mouth) by the Compagnie Nationale du Rhone and the suspended sediment

concentration (SSC)—water discharge regression derived by Sempere et al. (2000).

Fig. 3. Total mass fluxes (TMF) for the six traps deployed in the head of the major

canyons of the Gulf of Lions. The scale has been kept constant to highlight

differences between PLC and LDC and CCC).

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–1970 1961

4.2.1. Spatial distribution

In this study, TMF showed significant spatial variability (Fig. 3).Lowest averaged TMF were found in the northeast (3.1 g m�2 d�1

in the head of the PLC), and highest TMF are observed in thesouthwest (13.4 g m�2 d�1 in the LDC). This is in good agreementwith previous studies conducted in this area (Monaco et al., 1990,1999; Durrieu de Madron et al., 1999; Heussner et al., 2006).However, the picture that emerges from our trap data is slightlymore complicated. When looking at the overall average (Table 2),HEC and PRC have higher TMF than CCC. If we consider the twodeployments separately, averaged TMF for the different canyonsduring the first deployment shows a gradient increasing from PLCto LDC (unfortunately, no TMF was available for CCC) but with arelatively low average TMF for the AUC. For the second deploy-ment, TMF is on average much lower with the highest TMFmeasured in the CCC. Lowest TMF was observed in the HEC for thesecond deployment (5.6 g m�2 d�1), although it should be notedthat no fluxes could be measured in the PLC during the seconddeployment, where we presume that even lower TMF would havebeen found. To give an idea, time-averaged TMF measured furtherdown the PLC, for the same period, by the traps deployed at 1000(30 mab) and 500 m (mid-water column) indicate relatively lowTMF of 1.9 and 1.6 g m�2 d�1, respectively.

4.2.2. Temporal variability

TMF was highly variable throughout the whole experiment.Although periods of high flux were recorded during bothdeployments, higher TMF were generally observed in fall. Thebase flux was lower during the second deployment with TMFvalues as low as o1 g m�2 d�1 during the second week of April inboth CCC and LDC. For canyons with data from both deployments,the first deployment was characterized by higher mean TMF. Thefirst deployment was marked by one high flux event that occurredat all sites in early December. During this event, TMF reached109.5 g m�2 d�1 in the LDC and was as high as 67.1 g m�2 d�1 in theHEC (Fig. 3). During the early winter, TMF was fairly low anddevoid of major peak until 22 February when high TMF wasobserved in the CCC, LDC and to a lesser extent in the PRC. Duringthis event, TMF reached 32 g m�2 d�1 in the CC and current speedwas as high as 82.5 cm s�1. A third TMF peak was recorded in midApril (second deployment). Except in the HEC where it reached22.5 g m�2 d�1, its intensity was lower than in February despitefairly vigorous bottom currents (Fig. 4).

4.3. Composition of trapped particles

Material collected by sediment traps during this project wasdominantly composed of greenish fine-grained particles. Fluffyaggregates were observed in nearly all samples in variableamounts. Fecal pellets were found in all samples but representedonly a small fraction of the total mass. Other constituents wereforams, shell fragments and lithogenic grains (quartz, feldspar,calcite, clay) that were observed systematically during high fluxintervals.

Bulk composition of trapped particles over the whole experi-ment highlights the predominance of lithogenic (siliciclastic)matter in all traps with an overall flux-weighted mean of 63.6%.Temporal and spatial variability was fairly low with individualinterval values ranging from 57.2% to 70.6%. Content of CaCO3 didnot vary much in time but showed a fairly clear decrease fromnortheast (PLC, 35.1%) to southwest (CCC, 29.4%) highlighting theinput from the Rhone River (Table 2). For canyons where trapsfunctioned during both deployments, averaged lithogenic andCaCO3 contents were higher in fall/winter than in winter/spring.For the biogenic components (Corg, BSiO2, 13C, chloropigmentsand amino acids) of the trap samples, the reader is referred toFabres et al. (2008).

ARTICLE IN PRESS

Table 2Average total mass flux (TMF, in g m�2 d�1) and flux-weighted average content (%)

of major constituents in sediment traps deployed in the canyons heads

Fall/winter

deployment

Winter/

spring

deployment

Overall

Cap Creus TMF – 7.5 7.5

Corg – 1.8 1.8

BSiO2 – 2.2 2.2

CaCO3 – 29.4 29.4

Lithogenic – 64.8 64.8

Lacaze-

Duthiers

TMF 21.5 5.9 11.7

Corg 1.3 1.9 1.4

BSiO2 0.4 3.2 1.1

CaCO3 31.0 29.1 30.5

Lithogenic 66.1 63.9 65.5

Aude TMF 7.8 – 7.8

Corg 1.3 – 1.3

BSiO2 0.6 – 0.6

CaCO3 31.4 – 31.4

Lithogenic 65.3 – 65.3

Herault TMF 14.3 5.6 9.6

Corg 1.0 1.6 1.2

BSiO2 0.3 1.8 0.7

CaCO3 28.7 30.4 29.2

Lithogenic 60.0 64.6 61.5

Petit-Rhone TMF 11.3 5.9 8.6

Corg 1.4 1.9 1.6

BSiO2 0.4 2.8 1.3

CaCO3 33.1 30.5 33.4

Lithogenic 63.8 62.9 65.8

Planier TMF 3.1 – 3.1

Corg 2.5 – 2.5

BSiO2 0.9 – 0.9

CaCO3 35.1 – 35.1

Lithogenic 58.9 – 58.9

Fig. 4. Current–velocities (cm s�1) and SSC (mg l�1) near the bottom (at 5 mab) for

all six canyons. Note the strong current (that lasted less than 12 h) on the 4

December in nearly all canyons except the PLC. Note also the occurrence of several

relatively intense flows (associated with storms) during the winter period

transport.

Table 3Statistics on current speeds (minimum, maximum and mean in cm s�1) and

frequency of observation (%) by velocity class in canyons heads between November

2003 and May 2004

Speed

(cm s�1)

Cap

Creus

Lacaze-

Duthiers

Aude Herault Petit-

Rhone

Planier

o5 54.8 53.3 77.6 78.1 63.9 82.4

5–10 19.8 21.3 11.6 14.5 20.2 10.3

10–15 7.3 9.1 7 3.5 6.8 2.9

415 18.1 16.3 3.8 3.9 9.1 4.4

Minimum 0.3 0.3 0.3 0.3 0.3 0.3

Maximum 82.4 60.7 36.7 48.7 57.2 51.9

Mean 10.7 9.4 4.7 4.3 6.1 4.2

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–19701962

4.4. Current– velocity and horizontal fluxes

Time series of near-bottom current–velocities for all canyonsare shown in Fig. 4, and statistics of current speed are presented inTable 3. Current magnitude shows great differences at the varioussites: averaged speed over the whole experimental period is4.2 cm s�1 in the PLC while it reaches 9.5 and 10.8 cm s�1 in theLDC and CCC, respectively. Peak velocities (reaching 82.4 cm s�1 inthe CCC in late February) occurred on several occasions. Majorpeaks were recorded in early December (lasting 12 h only), lateFebruary, late March and mid April and late April. Time seriesfrom CCC and LDC appear similar with concomitant peaks andcomparable magnitudes, although current intensity was alwayshigher in the CC canyon.

Horizontal flux data are presented in Fig. 5. The magnitude ofthe calculated horizontal component (reaching 400,000 g m�2 d�1

in the CCC) is systematically higher than TMF measured bysediment traps (sometimes by more than four-orders of magni-tude!). Generally, no correlation emerges between ‘‘vertical’’ trapflux and calculated horizontal flux. This is in good agreement withobservations of Gardner et al. (1997) who found that fluxesmeasured by traps are not a proportional measure of thehorizontal flux of particle past the traps. From our results, weobserve though that when the trap flux is markedly high, such asduring the early December interval in LDC or the late Februaryinterval in CCC, the horizontal flux is also maximum. Unfortu-

nately, it was not possible to do grain size measurement on alltrap samples to help with the interpretation of variations invertical flux. Grain size of trap material collected in the HE canyonaround the 24 April storm showed that sand size particlesrepresented nearly 40% of the TMF for this interval (F. Bourrin,unpublished data). This may explain why this interval horizontaland vertical fluxes co-vary. We can also speculate that laterallyadvected coarse and heavy particles (such as sand) were alsointercepted in the traps at times of the other major flux eventsdescribed earlier.

ARTICLE IN PRESS

Fig. 5. Horizontal flux (dashed line) integrated over the same period (7 days) as

sediment trap fluxes (solid line). Positive horizontal fluxes are orientated down

canyon while negative fluxes are up-canyon. Note the difference in magnitude

between horizontal (up to 440,000 g m�2 d�1) and downward particle fluxes (up to

120 g m�2 d�1).

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–1970 1963

5. Discussion

This study was designed to assess the role of submarinecanyons in the transfer of particulate matter from a river-dominated shelf to the deep sea. For this, understanding thehydrodynamic processes that control such transfer is essential.Studies carried out on various continental margins (SEEP, OMEX,ECOMARGE) during the last two decades have shown that most ofthe particulate matter transfer from the shelf to the deep oceantakes place within the bottom boundary rather than by settling ofparticles from the surface layer (Biscaye and Anderson, 1994; vanWeering et al., 2002; Inthorn et al., 2006).

Considering this, sediment traps were deployed as close aspossible to the bottom to intercept the particles advected withinthe bottom boundary layer. With the traps positioned so close tothe bottom, it is clear that the measured flux is composed of twocomponents: the primary settling flux and the horizontallyadvected flux (including resuspended and rebound particles;Walsh et al., 1988). The latter likely overwhelms the primary fluxwhich probably represents a minor contribution only of the fluxduring the highest flux events. Deciphering the different compo-nents of this mixture is particularly complex, and for this reason itis difficult to distinguish export of very recent and older material(McCave et al., 2001), to identify zones of origin for primaryparticles (Siegel and Deuser, 1997) or to identify the resuspensioncomponent of the flux (Walsh et al., 1988; Bonnin et al., 2002).

Consequently, we consider in this study the bulk of the export flux(primary+horizontally advected).

5.1. Quality check of the trap fluxes

The mooring lines were maintained taut by floats mounted atthe top and along the mooring lines. Examination of the currentmeter pressure sensors showed that the mooring line was close tovertical even during strong current episodes (up to 82 cm s�1).Concerning the tilt of the traps, no direct measurement wasavailable. Modeling of the tilt in response to the local hydro-dynamics was carried out for the traps (Technicap PPS 3) operatedduring this study using the MDD (mooring design and dynamics)program (Manno, 2007). The results have shown that for currentspeed as high as 80 cm s�1, tilt of the trap was around 151 whichwe believe according to Gardner (1985) would not affectsignificantly the collection efficiency.

Three dimensionless parameters are important in the collec-tion efficiency of cylindrical traps (Butman et al., 1986): (i) thetrap Reynolds number, Rt ¼ u�D/n, where u is the flow velocity atthe height of the trap mouth, D is the diameter of the trap mouthand n is the kinematic viscosity (ratio of fluid viscosity to fluiddensity), (ii) the trap aspect ratio, and (iii) the ratio of flow speedto particle settling velocity. Only the effect of Rt is generallychecked since the trap aspect ratio is constant and both flowspeed and settling velocity are variable and unknown. Accordingto Butman (1986) and US-GOFS (1989), the collection efficiency ofthe traps in this study (with aspect ratio 41) is expected todecrease over some range of increasing Rt for fixed values of theother two parameters. To test this, Reynolds numbers werecomputed using averaged speed at each trap for each individualsampling period, assuming a constant temperature and salinity(T ¼ 13 1C, S ¼ 38.5, nE1.310�6 m�2 s�1) at each depth. Therefore,variation of Rt is only dependant on variations of the currentspeed. If a strong effect of current speed on collection efficiencieswas to exist, one might expect some covariation of total mass fluxwith that of current speed. Individual correlation coefficients foreach trap were not statistically significant (Fig. 6). Since we werenot able to detect unambiguously any effect of current speed, webelieve that mass fluxes measured during this study were notsignificantly biased by the water flow.

5.2. Discrepancies between TMF and horizontal fluxes

Vertical fluxes measured by the sediment traps and thehorizontal fluxes calculated from turbidity and current metersdata are not well correlated (Fig. 5). Several studies have reportedcomparable differences between horizontal (based on opticalsensors and current meters) and vertical (trap-based) fluxes(Garnder, 1989; Gardner and Walsh, 1990; Walsh and Gardner,1992; Gardner et al., 1997; Walsh and Nittrouer, 1999). Gardneret al. (1997) concluded from such comparisons that sedimenttraps were very inefficient at collecting horizontally advectedmaterial.

Furthermore, it should be noticed that the sediment traps andOBS/current meters used for this study were not positioned at thesame height above bottom. Traps collected material at 28 mabwhile OBS sensors measured turbidity at 5 mab. The position ofthe OBS and current meters closer to the bottom obviously createsdifferences between OBS-based and trap-based fluxes as concen-tration of suspended sediment generally gets higher near thebottom. It is thus likely that most of the sediment transportoccurred below 28 m.

With respect to vertical fluxes, it is important to rememberthat these particles are also moving horizontally, though the

ARTICLE IN PRESS

Fig. 6. Current speed versus total mass flux for all trap sample intervals. Note the absence of strong correlation.

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–19701964

largest vertical fluxes may well occur when the currents weakenafter an event large enough to produce high values of SSC.Elevated current speed during storms could alter the collectionefficiency of the sediment trap (mainly CCC and LDC) and hence ofthe downward particle flux (e.g., Gardner, 1985). Maximum aswell as mean current speeds were higher in LDC and CCC than inother canyons, and currents exceeding 15 cm s�1 were also foundmore often in those two canyons (16.3% and 18.1%, respectively;Table 3). This latter value is generally considered as a thresholdabove which trapping efficiency decreases (e.g., Baker et al., 1988)which would suggest even higher TMF during those strongcurrent intervals. This could explain, at least partly, differencesbetween TMF and horizontal flux in February (horizontalfluxcTMF in LDC and CCC). Observed differences between trap-based flux and transmissometer-based flux have been attributedto the size of particles collected by sediment traps which arelarger (d40.5 mm; Walsh and Gardner, 1992) than particlesmeasured by optical devices which are more sensitive to particleswith diameter o20mm (e.g., Bunt et al., 1999). In addition,Hatcher et al. (2001) argued that OBS sensors depend less on

particle size (which is dependent on mass) than on particleprojected-area concentration that, depending on fractal dimen-sion, need not to be proportional to mass. This implies that astrong OBS signal (high turbidity) may not coincide with a hightotal mass flux as measured by sediment traps. Detailed data onsuspended sediment particle sizes, not available in our study,would be required to obtain more accurate estimates of horizontalflux from OBS data.

TMF and SSC were plotted together with average andmaximum current speeds to try to further identify reasons forthe differences observed in this study. Fig. 7 shows that co-variation between TMF and SSC is relatively good for somecanyons (AUC, HEC), but is extremely poor for the CCC (negativecorrelation) and LDC. Those differences were particularly sub-stantial during high flux and current regime events that occurredin December and February. In general, correlation between TMFand SSC was better during the fall/winter deployment than thewinter/spring one.

In this study, the greatest TMF measured by traps was oftenassociated with the presence of coarse, rapidly sinking particles like

ARTICLE IN PRESS

Fig. 7. For each canyon, averaged SSC (open circle) versus TMF (dot) are shown on the lower panel and averaged (black line) and maximum current speed (dashed line) at

2 m below traps (upper panel). Co-variation is relatively good when current speed is low as this is the case in the AUC, HEC and PLC but appears extremely poor when speed

is high (LDC, CCC).

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–1970 1965

sands that may be missed by the OBS. Larger proportions of sandgrains have been observed through stereomicroscope for sampleswith highest TMF. Trap samples from the HEC were also measuredusing a laser particle sizer for the period around the April storm (F.Bourrin, personal communication) when TMF is higher in the HECthan in the LDC and CCC. The results show that up to 40% sand werepresent in trap samples from the HEC at that time (14–20 April).Uncertainties in estimates of SSC from OBS data for coarsersediment may also contribute to the lack of relationship betweenSSC and TMF in the CCC during the February storm.

5.3. Spatial distribution of the flux: the influence of the general

circulation

As mentioned earlier, the Rhone River represents the majorsediment input into the GoL. However, highest fluxes were notfound in front of the Rhone River mouth but much further to thesouth. TMF measured in the canyons heads of the GoL displayedcomparable spatial distribution patterns as previous observationsmade in that region (Monaco et al., 1990, 1999; Heussner et al.,2006): TMF is minimal in the far northeast (PLC), maximum at thesouthwest (LDC and CCC) and intermediate in the central sector(Table 2). As already reported (Durrieu de Madron et al., 1990,1999), this distribution follows the general circulation of theNorthern Current that flows from northeast to southwest and thatultimately transports particulate matter towards the LDC and CCCcanyons (Fig. 8). In comparison, canyons located in the center of

the gulf (PRC, HEC and AUC) have relatively low TMF compared totheir southwestern counterparts except during the extreme stormthat occurred in early December 2003 (Fig. 3). During thisparticular event and in this central sector, TMF in the HEC(67.1 g m�2 d�1) was higher than in the PRC and AUC canyons (32.3and 24.3 g m�2 d�1, respectively). In late April, TMF in the HECreached 22 g m�2 d�1 while it was only 5.3, 7.4 and 12.9 g m�2 d�1

in the PRC, LDC and CCC, respectively (Fig. 3). This also suggests aslightly more complex circulation in the center part of the gulfsuch as eddies or meanders of the slope current that could havesome consequences on the circulation and shelf-slope export(Durrieu de Madron, et al., 1999). Since TMF is fairly weak in PLCand PRC, the HEC appears as a probable natural conduit for exportof particles delivered by the Rhone River. Progressive vectordiagrams of current–velocity at the head of the canyons (Fig. 9)show that current has a northeast–southwest direction for PLC,PRC and HEC but its direction changes for AUC suggesting that thecirculation is somehow perturbed between HEC and AUC. Oneshould keep in mind though that the currents measured in thehead of the canyons are very much constrained by the orientationof the axis of the canyon and may therefore not reflect the patternof the general circulation on the shelf. One more point concernsthe sediment discharge shown in Fig. 2. It is marked by a strongincrease in the Tet River (that we assume is representative of mostof the small rivers of the GoL). At the same time, no significantincrease was observed in the Rhone River suggesting that smallrivers may have, during short and intense flood events, someimpact on sediment transport to the deep sea.

ARTICLE IN PRESS

Fig. 8. Satellite (MERIS) image of the Gulf of Lions just after the storm that occurred on 4 December. Note the turbid plume stretching along the coast and transporting

material at the south west of the study area above the CCC. The red arrow depicts the general circulation pattern. This picture has been modified after figure from Guillen et

al. (2008) in order to highlight the turbid plume (shaded area).

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–19701966

5.4. Temporal variability of the flux: the impact of marine storms and

cascading

Particulate matter flux time series (downward and horizontal)reveal that the transfer of particulate matter to the slope and deepbasin is clearly event-dominated. This has been observed in othermarine environments such as the Eel continental margin (Walshand Nittrouer, 1999; Guerra et al., 2006) the Adriatic Sea (Fainet al., 2007) or the Portuguese margin (Vitorino et al., 2002) wherethe variability of sediment fluxes was controlled by a combinationof shelf sediment resuspension, river discharge, wind and margincirculation that was particularly strong during winter months.

As we just discussed, the impact of the December storm onTMF is visible in every canyon albeit with different strength. Theimpact of this storm though appears much less clearly on the cal-culated horizontal fluxes. In contrast, the impact of the Februarystorm on TMF is almost exclusively restricted to the southernmostcanyons while its impact on the horizontal fluxes can easily beobserved in most of the canyons (Fig. 5).

The two major storms observed during this study weregenerated by intense southeast winds that initiated a down-welling flow of shelf water that enhanced cross-shelf transport.Furthermore, during those two events, near-bottom current speedwas fairly comparable. SSC exceeded 4 mg l�1 in the LDC canyonhead for both events. What may cause the TMF differencesbetween the December and February storms? First, the stormlasted 3 days in February and only 12 h in December. Thus, stormduration may partly explain why horizontal fluxes are higher inwinter. From the analyses of the cumulative horizontal fluxes inthe canyons (Palanques et al., 2006) and from modeling of thewater masses (Ulses et al., 2008a), it also appears that thedifferences in the intensity of the shelf-slope exchanges betweenthose two events (FebruarycDecember) were the consequence ofthe cascading of dense shelf water that occurred in February in the

southwestern-most canyons (and to a lesser extent through thecanyons of the central sector of the GoL). Why then do our resultsindicate higher TMF in December than February? In December,large amounts of suspended sediment are delivered by all riversalong the GoL as the consequence of heavy rainfall thataccompanied the storm while rainfall was minimal during theFebruary storm (Guillen et al., 2006). From this, we assume thatthe large TMF measured in December reflect the input ofsedimentary material resulting from the massive river discharge.We reviewed earlier the potential instrumental bias that may alterthe signals from traps and optical devices. However, in the light ofthe data that we present here and that have been publishedelsewhere (Palanques et al., 2006; Ulses et al., 2008a), we believethat the stratification of the water column may have also played arole in the high TMF but relatively low horizontal flux in fall andthe reverse in winter. We propose here (Fig. 10) conceptualsynthetic diagrams depicting the different water conditions andsuspended particle distribution before, during and after theautumn and winter storm events, respectively. The contour linesof density are based on in situ observations and modelingpublished by Ulses et al. (2008a) while the suspended particlecontour line are extrapolated from TMF and SSC data presented inthis study.

In fall, the water column was still fully stratified. The down-welling generated by the storm promoted a down-slope bendingof the isopycnal surfaces. The plume of suspended and riverdischarged particles was likely advected towards the canyonheads where moorings were deployed (Fig. 10a and b). However,the storm lasted only a few hours and the isopycnals relaxed backto their initial (pre-storm) stratified position rapidly after thestorm hence preventing particles to be transported much deeperdown. The particulate material may have therefore been con-strained on the very upper slope as a result of the strongstratification (Fig. 10c) and was available for vertical settling after

ARTICLE IN PRESS

-800

-600

-400

-200

0

200

-800

Planier Canyon

N-S

Dis

tanc

e (k

m)

E-W distance (km)

Petit Rhone Canyon Herault Canyon

Aude Canyon Lacaze-Duthiers Canyon Cap Creus Canyon

2000-200-400-600-800

-600

-400

-200

0

200

-800

N-S

Dis

tanc

e (k

m)

E-W distance (km)2000-200-400-600

-800

-600

-400

-200

0

200

-800

N-S

Dis

tanc

e (k

m)

E-W distance (km)2000-200-400-600

-800

-600

-400

-200

0

200

-800

N-S

Dis

tanc

e (k

m)

E-W distance (km)2000-200-400-600

-800

-600

-400

-200

0

200

-800

N-S

Dis

tanc

e (k

m)

E-W distance (km)2000-200-400-600

-800

-600

-400

-200

0

200

-200

N-S

Dis

tanc

e (k

m)

E-W distance (km)800 10006004002000

Fig. 9. Progressive vector diagrams of the current–velocity for PLC, PRC, HEC, AUD, LDC and CCC. The overall current direction is very much associated to the direction of

each canyon axis at the location were the mooring were deployed.

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–1970 1967

the storm had passed. This is in good agreement with the resultsof Heussner et al. (2006) of long-term TMF time series who didnot find any correlation with TMF and storms at mid canyon depth(1000 m). TMF measured from sediment traps deployed in the LDCand PLC deeper down the slope (1000 m, long mooring; Fig. 1) donot indicate any increase of the particle transfer during theDecember event. The near-bottom trap deployed in the LDC didnot function during the fall/winter period which limits theinterpretation we can make from these data. We can, however,speculate from these trap data that sediment particles that wereintercepted in the traps at the canyons head did not reach depthsof 1000 m or more.

In winter, the water column was completely mixed due tostrong northerly winds that prevailed since late autumn: thewhole shelf water column became homogeneous and denser thanslope waters (Ulses et al., 2008a). The newly formed dense waterlikely cascaded down the canyon (mostly through the south-western-most canyons) and reached its equilibrium depth around400 (Ulses et al., 2008a; Fig. 10d). The severe storm that occurredon the 22 February had several effects on this preexistingsituation: (1) it facilitated resuspension of sediment on the shelfby the action of high waves (Palanques et al., 2006) and (2) itinduced downwelling of shelf turbid water in response to wateraccumulation near the coast (as during the December storm). Thislikely reinforced the down-canyon transport of dense shelf water.During the February storm (Fig. 10e and f), dense shelf water andassociated resuspended particles reached the canyon head,plunged rapidly within the bottom boundary layer to greaterdepth where it appears to reach its equilibrium at about 650 m(Ulses et al., 2008a), hence favoring sediment deposition aroundthis depth. During this storm, the suspended particles were

probably rapidly advected down the canyons within the bottomboundary layer explaining the low TMF recorded in the trapspositioned at 28 mab. The February TMF peak that was observedat 300 m in the CCC and LDC was not observed at 1000 m in theLDC (Fig. 11) during the same interval but a TMF increase wasobserved at 1000 m in the following trap interval (15 March–15April). This increase was not observed in the trap situated at mid-water depth (500 m) which suggests the near-bottom advection ofparticles as a result of dense water cascading. We should note forthis study, however, that the fall/winter 2003–2004 was not aparticularly cold winter and therefore the dense water cascadingwas not particularly strong. As a comparison, winter 1999 andmore recently 2005 (Canals et al., 2006) was much colder andmassive vertical fluxes were recorded at 1000 m (Fig. 11). Theperiod covered by this study was, however, exceptional in terms ofthe recurrence and the intensity of the storms that certainlyfacilitated the resuspension of shelf sediments.

6. Conclusions

Sediment trap data collected in the head of the majorsubmarine canyons of the GoL showed that particulate matterfluxes are generally larger in the southwestern canyons than intheir northern counterparts but the differences are not as high asthose observed in previous studies based on horizontal fluxes.Hence, the general circulation that flows southwest-ward anddecreases off-shelf certainly plays a role but TMF in the canyonsheads is also very much related to local hydrology and atmo-spheric events. On the one hand, a fall storm induced large riverdischarge and transport through the canyons as a result of

ARTICLE IN PRESS

Fig. 10. Schematic and conceptual diagrams showing the water-column conditions during autumn and winter storms in the LDC. Panels a–c illustrate the effect of the

December event on the transfer of particles. At that time of the year, the water column is still stratified which limits the impact in terms of export depth (�300 m). Hence,

despite the large amount of particles delivered by the flood that accompanied the storm (wet storm) and resuspended by the waves generated by the storm itself, the export

down the canyon is limited. Panels d–f illustrate the role of the cascading of shelf dense water in a fully homogenized water column during winter to resuspend and export

massively particle deeper down (�700 m). The solid black lines show isopycnals obtained from CTD surveys and modelling of Ulses et al. (2008a), color shaded areas depict

turbidity concentration, indicative scale based on SSC data measured for turbidity is shown dashed line and star represent the mooring line with the position of the

sediment trap. The hatched and shadowed areas represent zones of erosion and deposition, respectively.

Fig. 11. Time series of TMF measured at 1000 m water depth in the LDC between

1993 and 2005. The dashed line is a data extrapolation when traps did not

function. The grey area represents the period during which sediment traps were

deployed in the canyons heads, i.e., the period used in this study. Finally, note that

the year 2003–2004 as not exceptional in terms of particulate matter export to the

deeper part of the basin as compared to 1999 and 2005 when particularly intense

dense shelf water cascading occurred (Canals et al., 2006).

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–19701968

downwelling generated by the accumulation of water against thecoast. However, the material at that occasion was not transportedmuch deeper than the canyon’s head region owing to water-column stratification. On the other hand, the severe winter storm,that was not accompanied by heavy rainfall, facilitated the

resuspension of previously deposited material on the shelf andexport of particulate matter further down the slope within thebottom boundary layer as the consequence of favorable denseshelf water cascading conditions (unstratified water column)prevailing at that time.

This paper also addresses the differences of particulate matterfluxes measured either from sediment traps or using current–velocity and suspended matter concentration for the two majorexport events. The substantial differences between the two typesof fluxes (up to four-orders of magnitude) are, in our opinion,linked to (1) differences in the response of optical sensors versustraps to differences in particle size and composition, (2) thedifferent height above the bottom of the different instruments,and (3) the different water-column stratification conditions in falland winter that resulted in different transport mechanisms(settling versus lateral advection).

Acknowledgments

This study was supported by the EUROSTRATAFORM FP-5programme. We wish to thank the technicians and the crew of theR.V Tethys for their precious field assistance during the Eurostrata-form-Canyon cruises. Appreciation goes to Franc-ois Bourrin forkindly providing us the Meris Satellite image that we use in ourFig. 8. Comments by P. Wiberg, J.P. Walsh and one anonymousreviewer considerably improved the manuscript.

ARTICLE IN PRESS

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–1970 1969

References

Aloısi, J.C., Auffret, G.A., Auffret, J.P., Barusseau, J.P., Hommeril, P., Larsonneur, C.,Monaco, A., 1977. Essai de modelisation de la sedimentation actuelle sur lesplateaux continentaux franc-ais. Bulletin de la Societe Geologique de France 7,183–195.

Baker, E.T., Milburn, H.B., Tennant, D.A., 1988. Field assessment of sediment trapefficiency under varying flow conditions. Journal of Marine Research 46,573–592.

Bethoux, J.P., Durrieu de Madron, X., Nyffeler, F., Taillez, D., 2002. Deep water in thewestern Mediterranean: peculiar 1999 and 2000 characteristics, shelf forma-tion hypothesis, variability since 1970 and geochemical inferences. Journal ofMarine System 33–34, 117–131.

Bianchi, T.S., Allison, M.A., Canuel, E.A., Corbett, D.R., McKee, B.A., Sampere, T.P.,Wakeham, S.G., Waterson, W., 2007. Rapid export of organic matter to theMississippi Canyon. EOS 87 (50), 565–572.

Biscaye, P.E., Anderson, R.F., 1994. Fluxes of particulate matter on the slope of thesouthern Middle Atlantic Bight: SEEP-II. Deep-Sea Research 41 (2–3), 459–510.

Biscaye, P.E., Flagg, C.N., Falkowski, P.G., 1994. The shelf edge exchange processesexperiment, SEEP-II: an introduction to hypotheses, results and discussion.Deep-Sea Research II 41 (2–3), 231–252.

Bonnin, J., van Raaphorst, W., Brummer, G.-J., van Haren, H., Malschaert, H., 2002.Intense mid-slope resuspension of particulate matter in the Faeroe-ShetlandChannel: short-term deployment of near-bottom sediment traps. Deep-SeaResearch 49, 1485–1505.

Bonnin, J., Koning, E., Epping, H.G., Brummer, G.-J., Grutters, M., 2005. Geochemicalcharacterization of resuspended sediment on the southeast slope of theFaeroe-Shetland Channel. Marine Geology 214, 215–233.

Bonnin, J., van Haren, H., Hosegood, P., Bummer, G.-J., 2006. Burst resuspension ofseabed material at the foot of the continental slope of the Rockall Channel.Marine Geology 226, 167–184.

Bourrin, F., Durrieu de Madron, X., Ludwig, W., 2006. Contribution to the study ofcoastal rivers and associated prodeltas to sediment supply in the Gulf of Lions.Life and Environment 56 (4), 307–314.

Bunt, J.A.C., Larcombe, P., Jago, C.F., 1999. Quantifying the response of opticalbackscatter devices and transmissometers to variations in suspended particu-late matter. Continental Shelf Research 19, 1199–1220.

Butman, C.A., 1986. Sediment trap biases in turbulent flows: Results from alaboratory flume study. Journal of Marine Research 44 (3), 645–693.

Butman, C.A., Grant, W.D., Stolzenbach, K.D., 1986. Predictions of sediment trapbiases in turbulent flows: a theoretical analysis based on observation from theliterature. Journal of Marine Research 44 (3), 601–644.

Canals, M., Puig, P., Durrieu de Madron, X., Heussner, S., Palanques, A., Fabres, J.,2006. Flushing submarine canyons. Nature 444 (7117), 354–357.

Dufau, C., Marsaleix, P., Petrenko, A., Dekeyser, Y., 2004. 3D modelling of the Gulf ofLions’ hydrodynamics during January 1999 (MOOGLI 3 experiment) and latewinter 1999. WIW formation and cascading over the shelfbreak. Journal ofGeophysical Research 109 (C11).

Durrieu de Madron, X., Panouse, M., 1996. Transport de matiere en suspension surle plateau continental du Golfe du Lion—situation estivale et hivernale.Comptes Rendus de l’Academie des Sciences—Series IIa, Paris 322, 1061–1070.

Durrieu de Madron, X., Nyffeler, F., Godet, C.H., 1990. Hydrographic structure andnepheloid spatial distribution in the Gulf of Lions continental margin.Continental Shelf Research 10, 915–929.

Durrieu de Madron, X., Radakovitch, O., Heussner, S., Loye-Pilot, M.D., Monaco, A.,1999. Role of the climatological and current variability on shelf-slopeexchanges of particulate matter: Evidence from the Rhone continental margin(NW Mediterranean). Deep-Sea Research I 46 (9), 1513–1538.

Estournel, C., Durrieu de Madron, X., Marsaleix, P., Auclair, F., Julliand, C., Vehil, R.,2003. Observations and modelisation of the winter coastal oceanic circulation inthe Gulf of Lion under wind conditions influenced by the continental orography(FETCH experiment). Journal of Geophysical Research 108 (C3), 8059.

Fabres, J., Calafat, A., Sanchez-Vidal, A., Canals, M., Heussner, S., 2002. Compositionand spatio-temporal variability of particle fluxes in the Western Alboran Gyre,Mediterranean Sea. Journal of Marine System 33–34, 431–456.

Fabres, J., Tesi, T., Velez, J., Batista, F., Lee, C., Calafat, A., Heussner, S., Palanques, A.,Miserocchi, S., 2008. Seasonal and event-controlled export of organic matterfrom the shelf towards the Gulf of Lions continental slope. Continental ShelfResearch, this issue, doi:10.1016/j.csr.2008.04.010.

Fain, A.M.V., Ogston, A.S., Sternberg, R.W., 2007. Sediment transport event analysison the western Adriatic continental shelf. Continental Shelf Research 27,431–451.

Flexas, M., Durrieu de Madron, X., Garcia, M.A., Canals, M., Arnau, P., 2002. Flowvariability in the Gulf of Lions during the MATER HFF experiment (March–May1997). Journal of Marine System 33–34, 197–214.

Gardner, W., 1985. The effect of tilt on sediment trap efficiency. Deep-Sea Research32, 349–361.

Garnder, W.D., 1989. Baltimore canyon as a modern conduit of sediment to thedeep sea. Deep-Sea Research 36, 323–358.

Gardner, W.D., Walsh, I.D., 1990. Distribution of macroaggregates and fine-grainedparticles across a continental margin and their potential role in fluxes. Deep-Sea Research 37, 401–411.

Gardner, W.D., Biscaye, P.E., Richardson, M.J., 1997. A sediment trap experiment inthe Vema Channel to evaluate the effect of horizontal particle fluxes onmeasured vertical fluxes. Journal of Marine Research 55, 995–1028.

Guerra, J.V., Ogston, A.S., Sternberg, R.W., 2006. Winter variability of physicalprocesses and sediment-transport events on the Eel River shelf, northernCalifornia. Continental Shelf Research, doi:10.1016/j.csr.2006.07.002.

Guillen, J., Palanques, A., Puig, P., Durrieu de Madron, X., Nyffeler, F., 2000. Fieldcalibration of opticakl sensors for measuring suspended sediment concentra-tion in the Western Mediteranean. Scientia Marina 64, 427–435.

Guillen, J., Bourrin, F., Palanques, A., Durrieu de Madron, X., Puig, P., Buscail, R.,2006. Sediment dynamics during wet and dry storm events on the Tet innershelf (SW Gulf of Lions). Marine Geology 234, 129–142.

Hatcher, A., Hill, P., Grant, J., 2001. Optical backscatter of marine flocs. Journal ofSea Research 46, 1–12.

Heussner, S., Ratti, C., Carbonne, J., 1990. The PPS 3 time series sediment traps andthe trap sample processing techniques used during the ECOMARGE experi-ment. Continental Shelf Research 10, 943–958.

Heussner, S., Durrieu De Madron, X., Radakovitch, O., Beaufort, L., Biscaye, P.E.,Carbonne, J., Delsaut, N., Etcheber, H., Monaco, A., 1999. Spatial and temporalpatterns of downward particle fluxes on the continental slope of the Bay ofBiscay (Northeastern Atlantic). Deep-Sea Research Part II 46 (10), 2101–2146.

Heussner, S., Canals, M., Durrieu de Madron, X., Palanques, A., Carbonne, J., Delsaut,N., Guarracino, M., Saragoni, G., 2006. Interannual and spatial variability ofshelf-slope particulate fluxes and exchange processes in the Gulf of Lions(NW Mediterranean). Marine Geology 234, 63–92.

Inthorn, M., Wagner, T., Scheeder, G., Zabel, M., 2006. Lateral transport controlsdistribution, quality, and burial of organic matter along continental slopes inhigh-productivity areas. Geology 34 (3), 205–208.

Lapouyade, A., Durrieu de Madron, X., 2001. Seasonal variability of the advectivetransport of suspended particulate matter and organic carbon in the Gulf ofLions (NW Mediterranean). Oceanologica Acta 24, 295–312.

Liu, K.K., Atkinson, L., Chen, C.T.A., Gao, S., Hall, J., MacDonald, R.W., TalaueMcManus, L., Quinones, R., 2000. Exploring continental margin carbon fluxeson a global scale; EOS transactions. American Geophysical Union 81 (52),641–644.

Manno, S. 2007. Validation et applications d’un modele de comportementde lignede mouillage. Master Thesis. 30p.

McCave, I.N., 1986. Local and global aspects of the bottom nepheloid layers in theworld ocean. Netherlands Journal of Sea Research 20, 167–181.

McCave, I.N., Hall, I.R., Antia, A.N., Chou, L., Dehairs, F., Lampitt, R.S., Thomsen, L.,van Weering, T.C.E., Wollast, R., 2001. Distribution, composition and flux ofparticulate material over the European margin at 47–501N. Deep-Sea ResearchPart II 48, 3107–3139.

Millot, C., 1990. The Gulf of Lions’s hydrodynamics. Continental Shelf Research 10(9–11), 885–894.

Millot, C., 1999. Circulation in the western Mediterranean Sea. Journal of MarineSystem 20, 423–442.

Miquel, J.C., Fowler, S.W., La Rosa, J., Buat-Menard, P., 1994. Dynamics of thedownward flux of particles and carbon in the open northwestern Sea. Deep-SeaResearch Part II 41, 243–261.

Monaco, A., Courp, T., Heussner, S., Carbonne, J., Fowler, S.W., Deniaux, B., 1990.Seasonality and composition of particulate fluxes during ECOMARGE-I,western Gulf of Lions. Continental Shelf Research 10, 959–988.

Monaco, A., Durrieu de Madron, X., Radakovitch, O., Heussner, S., Carbonne, J.,1999. Origin and variability of downward biogeochemical fluxes on the Rhonecontinental margin (NW Mediterranean). Deep-Sea Research Part I 46,1483–1511.

Mortlock, R.A., Froelich, P.N., 1989. A simple method for the rapid determination ofbiogenic opal in pelagic marine sediments. Deep-Sea Research 36, 1415–1426.

Muller-Karger, F.E., Varela, R., Thunell, R., Luerssen, R., Hu, C., Walsh, J.J., 2005. Theimportance of continental margins in the global carbon cycle. GeophysicalResearch Letter 32, L01602.

Palanques, A., Durrieu de Madron, X., Puig, P., Fabres, J., Guillen, J., Calafat, A.,Canals, M., Heussner, S., Bonnin, J., 2006. Suspended sediment fluxes andtransport processes in the Gulf of Lions submarine canyons. The role of stormsand dense water cascading. Marine Geology 234, 43–61.

Petrenko, A., Leredde, Y., Marsaleix, P., 2005. Circulation in a stratified and wind-forced Gulf of Lions, NW Mediterranean Sea: in situ and modeling data.Continental Shelf Research 25, 7–27.

Puig, P., Palanques, A., 1998. Temporal variability and composition of settlingparticle fluxes on the Barcelona continental margin (Northwestern Mediterra-nean). Journal of Marine Research 56, 639–654.

Sempere, R., Charriere, B., Van Wambeke, F., Cauwet, G., 2000. Carbon inputs of theRhone River to the Mediterranean Sea: biogeochemical implications. GlobalBiogeochemical Cycles 14, 669–681.

Serrat, P., Ludwig, W., Navarro, B., Blazi, J.-L., 2001. Variabilite spatio-temporelledes flux de matieres en suspension d’un fleuve cotier mediterraneen: laTet (France). Comptes Rendus de l’Academie des Sciences, Paris 333,389–397.

Siegel, D.A., Deuser, W.G., 1997. Trajectories of sinking particles in the Sargasso Sea:modeling of statistical funnels above deep-ocean sediment traps. Deep-SeaResearch Part I 44, 1519–1541.

Tusseau-Vuillemin, M.H., Mortier, L., Herbaut, C., 1998. Modelling nitrate fluxes inan open coastal environment (Gulf of Lions): transport versus biogeochemicalprocesses. Journal of Geophysical Research 103 (C4), 7693–7798.

Ulses, C., Estournel, C, Bonnin, J., Durrieu de Madron, X., Marsaleix, P., 2008a.Impact of storms and dense water cascading on shelf-slope exchange in theGulf of Lions (NW Mediterranean). Journal of Geophysical Research 113,doi:10.1029/2006JC003795.

ARTICLE IN PRESS

J. Bonnin et al. / Continental Shelf Research 28 (2008) 1957–19701970

Ulses, C., Estournel, C., Durrieu de Madron, X., Palanques, A., 2008b. Suspendedsediment transport in the Gulf of Lions (NW Mediterranean): impact ofextreme storms and floods. Continental Shelf Research, this issue, doi:10.1016/j.csr.2008.01.015.

US GOFS, 1989. Sediment trap technology and sampling. In: Knauer, G., Asper, V.(Eds.), Report of the US GOFS Working Group on Sediment Trap Technologyand Sampling. Planning Report, vol. 10. US GOFS Planning Office, Woods Hole,MA, 94pp.

van Weering, T.C.E., McCave, I.N., De Stigter, H.C., Hall, I., Thomsen, L., 1998.Recent sediments, sediment accumulation and carbon burial at GobanSpur, N.W. European continental margin. Progress in Oceanography 42,5–35.

van Weering, T.C.E., de Stigter, H.C., Boer, W., de Haas, H., 2002. Recent sedimenttransport and accumulation on the NW Iberian margin. Progress inOceanography 52, 349–371.

Vitorino, J., Oliveira, A., Jouanneau, J.M., Drago, T., 2002. Winter dynamics on thenorthern Portuguese shelf. Part 1: Physical processes. Progress in Oceano-graphy 52, 129–153.

Walsh, I.D., Gardner, W.D., 1992. A comparison of aggregate profiles with sedimenttrap fluxes. Deep-Sea Research 39, 1817–1834.

Walsh, J.P., Nittrouer, C.A., 1999. Observations of sediment flux to the Eelcontinental slope, northern California. Marine Geology 154, 55–68.

Walsh, I.D., Fischer, K., Murray, D., Dymond, J., 1988. Evidence for resuspension ofrebound particles from near-bottom sediment traps. Deep-Sea Research 35, 59–70.