Impact of open-ocean convection on particle fluxes and sediment dynamics in the deep margin of the Gulf of Lions

The deep outer margin of the Gulf of Lions and the adjacent basin, in the western Mediterranean Sea, are regularly impacted by open-ocean convection, a major hydrodynamic event responsible for the ventilation of the deep water in the western Mediterranean Basin. However, the impact of open-ocean convection on the flux and transport of particulate matter remains poorly understood. The variability of water mass properties (i.e., temperature and salinity), currents, and particle fluxes were monitored between September 2007 and April 2009 at five instrumented mooring lines deployed between 2050 and 2350-m depth in the deepest continental margin and adjacent basin. Four of the lines followed a NW–SE transect, while the fifth one was located on a sediment wave field to the west. The results of the main, central line SC2350 ("LION") located at 42°02.5′ N, 4°41′ E, at 2350-m depth, show that open-ocean convection reached mid-water depth (≈ 1000-m depth) during winter 2007–2008, and reached the seabed (≈ 2350-m depth) during winter 2008–2009. Horizontal currents were unusually strong with speeds up to 39 cm s −1 during winter 2008–2009. The measurements at all 5 different locations indicate that mid-depth and near-bottom currents and particle fluxes gave relatively consistent values of similar magnitude across the study area except during winter 2008–2009, when near-bottom fluxes abruptly increased by one to two orders of magnitude. Particulate organic carbon contents, which generally vary between 3 and 5%, were abnormally low (≤ 1%) during winter 2008–2009 and approached those observed in surface sediments (≈ 0.6%). Turbidity profiles made in the region demonstrated the existence of a bottom nepheloid layer, several hundred meters thick, and related to the resuspension of bottom sediments. These observations support the view that open-ocean deep convection events in the Gulf of Lions can cause significant remobilization of sediments in the deep outer margin and the basin, with a subsequent alteration of the seabed likely impacting the functioning of the deep-sea ecosystem.

waters to depth through dense shelf water cascading (DSWC) and open-ocean convection (OOC), have mainly been studied from a physical oceanography viewpoint (Marshall and Schott, 1999;Ivanov et al., 2004). The Mediterranean Sea constitutes a remarkable marine domain where exchanges with the Atlantic Ocean through the Strait of Gibraltar and deep-water formation drive thermohaline circulation in the different sub-basins and in the whole basin, in turn controlling the distribution of biogeochemical compounds and thus shaping the ecosystem (MER-MEX Group, 2011). Studies conducted in the Mediterranean Sea have shown how significant the role of dense deep-water formation and associated winter vertical mixing could be. Sinking dense water carries large amounts of organic matter, including particulate and dissolved carbon (POC and DOC, respectively), from the productive surface layer to depths below. Therefore, the removal of POC and DOC from surface waters and their injection to the deep Mediterranean Sea are strongly affected by deep-water formation and the vertical mixing that comes with it (Avril, 2002;Canals et al., 2006;Santinelli et al., 2010). Subsequently, DSWC and OOC lead to the formation of new, ventilated deep waters in the Mediterranean Basin. The occurrence of an efficient transfer of particles from the surface layer to depth linked to peak fluxes because of coastal and open-ocean dense water formation has been shown by several authors in the Gulf of Lions and in the nearby Ligurian Sea, both in the northwestern Mediterranean Sea, in winter 1999Sanchez-Vidal et al., 2009;Miquel et al., 2011). It has also been shown that high fluxes captured by deep-sea sediment traps correlate with the concentrations of large particles. The main export of such large particles occurs in winter during episodes of enhanced vertical mixing (Durrieu de Madron et al., 1999;Stemmann et al., 2002).
Some authors have also assessed the role of E-SE storms in transferring particulate matter to the deep Gulf of Lions Sanchez-Vidal et al., 2012). The former inferred that the high particle fluxes observed in the outer margin and basin during the two winter periods between November 2003 and April 2005 were related to lateral inputs from the adjacent shelf due to the occurrence of strong E-SE storms or DSWC, while the latter evidenced that most organic matter transferred to depths below came from the shelf following a large eastern storm in late December 2008 and that it was transferred to depths below associated with the finest fraction of particle fluxes. However, none of these two studies addressed the effects on particle fluxes of open-ocean deep convection that occurred at the same time. Open-ocean deep convection in the western Mediterranean Basin and the formation of western Mediterranean Deep Water (WMDW) have been investigated by numerous authors (see, for instance, Marshall and Schott, 1999 and references therein) following the MEDOC Group pioneer work in the Gulf of Lions (MEDOC Group, 1970). However, very few studies have focused to date on the impacts of OOC over matter and energy transfers, deep-water sediment dynamics and, ultimately, the deep ecosystem. Guidi-Guilvard (2002) showed that variations of the hyperbenthic communities in the Ligurian Sea resulted from both near-bottom currents and particle flux variability. Pusceddu et al. (2010) reported and compared the main impacts and ecosystemic effects produced by DSWC and OOC in the Gulf of Lions and the Aegean Sea, the latter in the northeastern Mediterranean Sea. These authors found that increased contents of particulate organic matter in deepsea sediments diminished benthic abundance and changes of benthic biodiversity. Martín et al. (2010) showed that sediment resuspension and abnormally high near-bottom particulate fluxes occurred in the Ligurian Sea during a severe OOC event in winter [2005][2006]. Puig et al. (2012) inferred, after an analysis of hydrological data of the last two decades that intense deep-water formation in winters 1999, 2005 and 2006 caused the formation of a thick bottom nepheloid layer that eventually extended over the entire western Mediterranean deep basin.
With the exception of the above-mentioned studies, it becomes quite clear that in the Gulf of Lions particulate fluxes and their sedimentary and ecosystemic impacts in the inner and mid-margin, including submarine canyons and adjacent open slopes, have received substantially more attention than in the deep outer margin adjacent to the Algero-Balearic Basin. We can also state that all studies emphasize the strong interannual variability of matter and energy transfers and impacts associated with Eastern storms and to DSWC and OOC reaching the seabed in the Gulf of Lions margin and adjacent deep basin.
Some modeling exercises on sediment transfer and vertical mixing in the Gulf of Lions have also been carried out. They have stressed, first, the impact of DSWC on the transfer of particulate matter to the deep margin and basin (Ulses et al., 2008) and, second, that both DSWC and OOC will likely experience a reduction in their frequency and intensity during the 21st century as a result of an increasing ocean stratification and the weakening of the thermohaline circulation because of the warming of ocean waters (Herrmann et al., 2008;Somot et al., 2008). The impact of global warming on the frequency and intensity of large storms is more uncertain (Ulbrich et al., 2009;Marcos et al., 2011;Young et al., 2011).
In this paper, we aim at characterizing the impact of OOC on particulate fluxes, and at inferring its potential imprint on sediment dynamics, after 20 months of continued observation of the water mass structure, hydrodynamics and mass fluxes in the Gulf of Lions deep outer margin and adjacent basin. By comparison with previous studies, we illustrate a number of scenarios on the effects of OOC over sediment transport, and discuss its potential effects on bathypelagic biological activity and benthic habitats.  .

General setting
The deep outer margin of the Gulf of Lions and the adjacent basin are bounded by Corsica and Sardinia to the east and the NE opening of the Valencia Trough and Balearic Islands to the west. To the north, a rather complex topography forms the continental slope of the gulf, which is incised by numerous submarine canyons (Fig. 1). The largest of these canyons extend towards the deep margin and basin at depths from 2000 to 2400 m. Particularly prominent are the N-S ori-ented Sète Canyon, towards which all canyons of the western Gulf of Lions converge, and the Petit-Rhône canyon-channel system feeding the Rhône deep-sea fan, with a neo-channel feeding the youngest lobe, or neo-fan (Droz and Bellaiche, 1985;Droz et al., 2001). Oceanic circulation in the northwestern Mediterranean Sea is characterized by a large-scale cyclonic pattern (see insert in Fig. 1). The northern branch of this cyclonic gyre, the Northern Current, flows westwards and south-westwards www.biogeosciences.net/10/1097/2013/ Biogeosciences, 10, 1097-1116, 2013 along the continental slope and continues southward through the Ibiza Channel. Part of it however flows eastward between the Balearic Islands and Corsica along the North Balearic Front (for a review see Millot and Taupier-Letage, 2005). Associated with this cyclonic gyre, a dome-like structure develops that reaches its highest point around 42 • N-5 • E. Dense water formation in the Gulf of Lions occurs owing to the effect of frequent and durable northwesterly (Tramontane) and northerly (Mistral) continental winds in winter. Dense deep waters form in two distinct locations: offshore, over the deep margin and adjacent basin, in a preconditioned area (Marshall and Schott, 1999 and references therein), and over the continental shelf (Durrieu de Madron et al., 2005;Ulses et al., 2008). OOC results from a combination of regional circulation (i.e., the cyclonic gyre involving the rising of intermediate and deep water masses) and meteorology (i.e., winddriven cooling and mixing of surface waters) that predisposes the water column to locally overturn (Marshall and Schott, 1999). Moreover, the N-S oriented positive relief formed by the Rhône deep-sea fan also contributes to trap the above waters following the Taylor column effect, thus prolonging the exposure of surface waters to the cooling action of the atmosphere (Madec and Crepon, 1991). Vertical convection mixes surface water with warmer but saltier intermediate water and deepens the mixed layer, which can eventually reach the seabed (> 2000-m depth) (Mertens and Schott, 1998). In the last 7 yr bottom-reaching convection over the outer margin and adjacent basin was observed in (López-Jurado et al., 2005Houpert et al., 2012). According to Marshall and Schott (1999), OOC is a three-stage process: (i) the "preconditioning phase" results from the wind-driven mixing of the surface layer and the erosion of the seasonal thermocline during autumn; (ii) the "mixing phase" in winter is characterized by a gradual deepening of the surface mixed layer and by the possible homogenization of the whole water column describing a large chimney-like structure; and (iii) the "sinking and spreading phase" takes place at the end of the winter when the restratification of the surface layer is due to reduced wind forcing and increased solar heating. Testor and Gascard (2006) studied the post-convection spreading phase in the northwestern Mediterranean Sea and showed that numerous eddies drifted away from the convection area and advected newly formed deep waters far away from the source region.
In situ current measurements on the deep slope and in the basin are scarce. Near-bottom currents in the deep basin are generally rather low (< 10 cm s −1 ), though substantially higher velocities (up to 50 cm s −1 ) have been occasionally measured during wintertime (Millot and Monaco, 1984;Palanques et al., 2009). Such currents are believed to be capable of reworking loose sediments, in an area where sedimentation rates are particularly low (i.e., 0.01-0.05 cm yr −1 after Miralles et al., 2005).

Mooring lines
Five mooring lines, deployed from September 2007 to April 2009, constituted the observational design in our study. They were contained within a circle of about 30 km in radius centred at 42 • 04 N, 4 • 38 E, which is the position of the Météo-France meteorological buoy "LION" (Fig. 1). The 5 lines were positioned on distinctive features of the Gulf of Lions' continental rise and basin. Lines SC2160, SC2240 and SC2350 were located at different water depths (last four digits) along the distal reach of Sète Canyon (SC). The westernmost line (SW2060) was located on a large sediment wave field west of Sète Canyon (SW) (Jallet and Giresse, 2005) atop of the Pyrenean Canyon Deep Sedimentary Body (Canals, 1985;Alonso et al., 1991), also named Pyreneo-Languedocian Sedimentary Ridge  to the south of the lower reach of Cap de Creus Canyon. The easternmost line, HC2300, is part of the HYDROCHANGE (HC) network (http://www.ciesm.org/marine/programs/hydrochanges.htm) and was deployed on the Rhône neo-fan lobe, south of the neo-channel (see Sect. 2: General setting).
The overall recording and sampling period lasted from September 2007 to April 2009. Relevant data on instruments for the different mooring lines are summarized in Table 1. SW2060 and SC2350 lines were operated by CEFREM/LOCEAN, SC2160 and SC2240 lines by IFREMER, and HC2300 line by MIO. All lines were equipped with current meters between 20 and 45 m above bottom (mab). With the only exception of the HC2300 eastern line, all lines were also equipped with a sequential sediment trap between 25 and 55 mab. The central line SC2350 ("LION") extended upward more than 2 km from the seafloor at 2350 m up to 150 m below sea surface. It included an additional sediment trap and current meter pair at mid-water depth (∼ 1000-m depth). SC2350 and HC2300 lines were equipped with additional CTD sensors (see Sect. 3.2: Hydrological time series and profiles).
The dynamics of the mooring lines were assessed using the Mooring Design and Dynamics software (Dewey, 1999) and validated by comparing the modeled deepening of the line to the pressure sensor records of current meters and/or CTD probes. Tilting and deepening of the mooring lines were insignificant during periods of weak current velocity (< 11 cm s −1 ) in spring and summer, and increased during periods of stronger currents (up to 39 cm s −1 ), mostly   In addition, the mooring line LDC1000 deployed since 1993 on the Lacaze-Duthiers Canyon slope at 1000-m depth was used as a slope reference to observe dense shelf water cascading events . Only potential temperature time series from September 2007 to April 2009 was used in this study.

Hydrological time series and profiles
Conductivity, temperature and depth time series were recorded at the SC2350 and HC2300 stations with SBE 37SMP instruments (Table 2). Four CTD sensors were clamped at 158, 690, 1507, and 2315-m depth on the SC2350 line, and one CTD sensor was clamped at 2287-m depth on the HC2300 line. RBR1000 temperature loggers were also clamped on the upper kilometer of the SC2350 line between SBE 37SMP CTD sensors to increase the vertical resolution (resulting in 8 sensors in the first deployment and up to 21 sensors in the third deployment). CTD data were corrected for offset and drift using pre-and post-deployment calibrations performed by Sea-Bird Electronics.
Additional comparisons between instruments were performed with respect to CTD casts carried out prior and after the turnarounds of mooring lines. CTD casts were done using a SBE 911 probe occasionally equipped with a Seapoint backscatter sensor at 880 nm to measure turbidity. The range of the turbidity sensor was 0-5 formazine turbidity units (FTU). ITS-90 was utilized for temperature. Temperature records from current meters were used at 500 and 1000 m depth on the LDC1000 line.

Current meter time series
Current speed and direction were recorded with mechanical or acoustic current meters (Table 1). The sampling interval was set to 30 min or 2 h. Due to a breaking at the base of SC2350 line during the second recovery and the loss of the current meter, no near-bottom current data were obtained at this site for the third deployment period (September 2008 to March 2009).

Sediment trap time series
Settling particles were collected with automated sediment traps. SC2160 and SC2240 lines held Technicap PPS5 coneshaped trap with a 1 m 2 mouth covered with a honeycomb baffle (with cells 10 cm deep and 1 cm in diameter). PPS5 traps were equipped with 24 collecting bottles. The sampling interval was set to 7 days during the first deployment (28 March 2008 to 29 August 2008) and to 9 days during the second deployment (1 September 2008 to 5 April 2009); see also in Table 1  velocity, the trap shape and aspect ratio (height/diameter), the tilt, and the settling velocity of particles (Baker et al., 1988;White, 1990;Buesseler, 1991 and reference therein). Despite hydrodynamic biases associated with intermittent strong currents and tilt, and different trap designs (conical and baffled PPS5 versus unbaffled cylindric PPS3), the similarities of temporal variability of near-bottom fluxes at the different sites suggests that data are robust enough to draw conclusions on the overall seasonal and interannual variability of downward particle fluxes in the study area.
For all traps, the collecting bottles were filled with a 3-5 % (v/v) formaldehyde solution in filtered seawater deployment. After recovery, large "swimmers" were removed by hand, and the remaining particles were filtered through glass-fiber filters for carbon analysis and 0.45 µm pore size membranes (Millipore or Nuclepore) for total mass flux determination, rinsed with Milli-Q purified water and dried at 40 • C. Particulate organic carbon concentrations in trap samples were measured with a LECO WR12 elemental analyzer (SC2160 and SC2230 samples) or LECO CN 2000 analyzer (SW2050 and SC2350 samples) after removing carbonates with a 2N HCl solution (Weliky, 1983).

Sediment cores
Sediment cores were sampled in April 2009 at three locations close to our experimental site with a multicorer using 10 cm in diameter, 30 cm long Plexiglas tubes. Each core was sliced in different layers of 0.5 cm down to 5 cm, and of 1 cm downcore (> 5 cm), stored in polyethylene plastic bags and preserved at 4 • C until analysis. Organic carbon and grain size were analysed on surface sediment samples (0-0.5 cm), whereas the entire core was used to estimate apparent sedimentation rate (Table 3).
POC content was measured with a LECO CN 2000: aliquots of freeze-dried, ground sediment were reacted with a 2N HCl solution to remove carbonates prior to analysis (Weliky, 1983). Grain size analyses were undertaken using a Malvern Mastersizer 2000 laser diffraction particle size analyzer; aliquots of fresh raw top sediment were prepared in 500 mL of water to obtain suspensions in a suitable dispersion to carry out the measurements. A small ultrasonic treatment was applied to break up loosely held agglomerates. For each sample 2 dispersions were prepared and the measurement was repeated 3 times until a stable dispersion had been achieved. Measurable sizes ranged from 50 nm to 1000 µm, and the results were then classified on the basis of textural features according to the Udden-Wentworth scale in three main categories: clay < 4 µm, 4 µm ≤ silt < 63 µm and 63 µm < sand < 1 mm.
Apparent sedimentation rates were determined by alphamass spectroscopy of freeze-dried, ground sediment. 210 Pb and 210 Po extraction was made by complete acid digestion (successively HNO 3 , HNO 3 -HClO 4 , HCl and HF), followed by spontaneous deposition of polonium on a silver disc (Flynn, 1968). The disc was placed between ZnS(Ag) phosphors and counted on a total alpha counter (Hallden and Harley, 1960). A subsequent deposition and counting, some 6 months later, recorded 210 Po ingrowth from 210 Pb in the sample. Activities at the time of sample collection were calculated from the two counts and the right decay corrections. Uncertainties were calculated by standard propagation of the ±1 sigma counting errors of samples and blanks. Apparent sedimentation rates were calculated using the CF-CS model (Constant Flux-Constant Sedimentation) of Anderson et al. (1987). In this model, the compaction effect is not considered, and apparent rates correspond to maximum values.

Satellite data of sea surface chlorophyll a concentration
Monthly  (Fig. 8a). Monthly rather than weekly or daily composites were chosen to limit noise or gaps due to cloud cover.

Integrated net heat fluxes
Components (sensible, latent and radiative) of atmospheric heat fluxes were collected from ECMWF (European Centre for Medium-Range Weather Forecasts). The model is based on a 0.25 • × 0.25 • lat/long grid and 3-h forecast intervals. We chose to extract data from the closest grid point (42 • N, 4.75 • E) to MF-LION location (Fig. 1).

Side-scan sonar
Side-scan sonar data was obtained onboard R/V Professor Logachev in August 2004 (Lastras et al., 2007), using a MAK-1M deep-towed platform. The MAK-1M consists of a 30 kHz sidescan sonar that yields a total swath range of up to 2 km from a mean altitude of 100 mab, with a variable resolution of about 7 to 1 m across track and along track. The system was operated along the canyon floor of the Sète canyon distalmost reach and was used to characterize seafloor morphology near the mooring sites. nutrients in the surface layer, (ii) an increase of chl a concentration during the fall planktonic bloom owing to erosion of the seasonal thermocline and, (iii) a minimum chl a concentration and production during winter due to the intense vertical mixing, and (iv) a large increase in late winter/early spring associated with the large planktonic bloom that takes place when the restratification of the surface layer occurs and the surface layer has been enriched in nutrients from deeper waters. As already shown by Morel and André (1991) and Santoleri et al. (2008), the location and extent of the OOC region in the Gulf of Lions was approximated for each winter from the patch of minimum chl a concentrations characteristics of deep-water formation. During winter 2007-2008, only a small patch of low (minimum 0.5 mg m −3 ) chl a concentration was visible in January, while during winter 2008-2009 a large patch of low chl a lasted from December to February with minimum chl a values of 0.2 mg m −3 . The five deep mooring lines were located inside the OOC region, while the shallower mooring line in the Lacaze-Duthiers Canyon was outside.

Temporal variability of integrated net heat fluxes and potential temperature
The surface cumulated net heat flux for the convection region (Fig. 4a)  The impact of the heat loss was clearly visible on the sea surface temperature (Fig. 4b), which decreased from 20-25 • C in September to a minimum of about 13 • C in late December (preconditioning phase). The effect on the water column (Fig. 4c) (Fig. 5b), followed by a gradual decrease linked to the continuation of the water column cooling by atmospheric forcing. A similar temperature increase was observed in mid-February at the other sites from the lower resolution current meter temperature sensor (data not shown). A rapid restratification of the upper water col- umn started in late-March for both years, marking the beginning of the sinking and spreading phase (Fig. 4c).
On the slope site (LDC1000), the temperature records at 500 and 1000-m depth showed a slight cooling of about 0.1 • C during wintertime, in particular at mid-depth (Fig. 5c). The absence of significant (> 1 • C) near-bottom temperature drop at 1000-m depth, characteristics of deep cascading of dense shelf water Palanques et al., 2012), indicated that water exported from the shelf was not dense enough to reach the rise and remained on the upper slope.

Variability of currents
The magnitude of near-bottom currents at the different sites and mid-depth currents showed a significant coherency (Fig. 6). Two distinct periods may be defined, i.e., the late spring-early autumn (May-October) period with low values (average and maximum speeds of about 3 and 11 cm s −1 , respectively) that contrasted with the late fall-early spring (November-April) period, with higher values (average and maximum speeds of about 6.7 and 40 cm s −1 , respectively). Important differences appeared for the two late fall-early spring periods (Fig. 6). During November 2007-April 2008, few significant current bursts were recorded close to the bottom at SC2350 and HC2300 and at mid-water depth at SC2350. During November 2008-April 2009, a first current burst took place in November 2008 and strong currents persisted in February and March 2009 at all sites. Average current velocity during that latter period was about 13 cm s −1 . Comparison of near-bottom current statistics at HC2300 (Table 4) for the February-April periods indicated that (i) average currents were stronger during winter 2009 (11.2 cm s −1 ) than during winter 2008 (5.9 cm s −1 ), and (ii) the occurrence www.biogeosciences.net/10/1097/2013/ Biogeosciences, 10, 1097-1116, 2013

Variability of chl a, mass flux, and particulate organic carbon content and flux in the convection region
The time series of monthly averaged surface chl a concentration in the convection region (Fig. 8a) showed a typical annual signal with summer minimum and spring maximum. In summer chl a concentration ranged between 0. The main statistics of total mass flux (TMF) and particulate organic carbon (POC) content for all traps are given in Table 5. Time-weighted average of near-bottom TMF for the entire sampling period varied within a narrow range (360-477 mg m −2 d −1 ), while the mid-depth time-weighted average TMF was much lower (90 mg m −2 d −1 ). Likewise, the flux and time-weighted average POC content in nearbottom traps ranged between 1.7 and 2.2 %, whereas the flux and time-weighted average POC content at mid-depth was more than twice as high (5.5 %).
Total mass fluxes and POC content at the different sites showed rather similar temporal variations (Fig. 8b-f). Minimum TMF (5-9 mg m −2 d −1 ) were recorded in summerfall and were associated with high POC contents. Maximum POC contents were observed in November 2008 with values of 5.8 and 7.7 % near the bottom and increasing with water depth, and 16.5 % at mid-depth. Maximum TMF were recorded near the bottom and at mid-depth in April-May 2008, and uniquely near the bottom in February-March 2009. In April-May 2008, high TMF were associated with high POC contents (4.9-5.8 %), resulting in high POC fluxes (3.2-4.5 mMol C m −2 d −1 ). In February-March 2009, near-bottom maximum TMF (2204-7180 mg m −2 d −1 ) were one order of magnitude larger than the average flux, but were associated with the minimum POC contents (0.8-1.2 %), producing nonetheless significant POC fluxes (1.3-4.9 mMol C m −2 d −1 ). At the mid-depth TMF (84-135 mg m −2 d −1 ) as well as POC content (1.9-3.3 %) and flux (0.16-0.24 mMol C m −2 d −1 ) were low. The continuation of the TMF pulse in late March-early April 2009, visible at the SC2160 and SC2240 sites (Fig. 7c-d), was associated with higher POC content (1.7-2.2 %), and POC flux at both sites was about 2-3 mMol C m −2 d −1 .
Comparison of near-bottom TMF at SC2350 for the winter period (December-March) indicated that the maximum TMF (7180 mg m −2 d −1 ) recorded in winter 2008-2009 was about two orders of magnitude higher than the maximum TMF (169 mg m −2 d −1 ) recorded in winter 2007-2008. At the same time the minimum POC content (0.8 %) in winter 2008-2009 was three times lower than the POC content (2.8 %) recorded in winter [2007][2008]. These results at the near-bottom trap contrast with those at the upper trap that do not exhibit such large fluctuations between the two winters.

Change of the bottom nepheloid layer structure
The CTD cast performed next to the SC2350 site in March 2008 ( Fig. 9a and c)  deep water masses, and the resulting distribution produced a thicker bottom turbid layer that extend more than 1000 mab.

Seabed characteristics
Surface sediment characteristics at different sites are listed in Table 3. Sediment accumulation rates ranged between 0.07 and 0.22 cm yr −1 , and increased together with the clay content (from 15.8 to 26.3 %). These values are in agreement with other studies previously conducted in the deep basin of the Gulf of Lions (Miralles et al., 2005 and references therein) and in the Ligurian Sea (Martín et al., 2009;Heimbürger et al., 2012). Despite differences in sediment grain size and sedimentation rates, POC content was approximately the same at all sites (around 0.6 %). Seabed morphology across the sediment wave field and along the distal reach of the Sète Canyon, where the mooring sites are located, is presented in Fig. 10. Furrows (Fig. 10b) are observed close to the SC2160 mooring line while crescent scours (Fig. 10c) and scours (Fig. 10d) are observed close to SC2240 and SC2350 mooring sites, respectively.

Discussion
Measurements obtained between September 2007 and April 2009 in the convection zone of the Gulf of Lions allow observing two contrasting winters. Winter 2007-2008 is characterized by a mild convection with a mixed layer depth limited to 1000-m depth, low current speeds and low particle fluxes in the water column and near the bottom. In contrast, winter 2008-2009 is characterized by a deep convection, with a mixed layer depth reaching to the bottom (2350m depth), stronger current speeds in the water column and near the bottom, and significant particle flux near the bottom. The extent of the convection zone, characterized by the minimum concentration of surface chl a, is smaller in winter 2007-2008 than in winter 2008-2009. The difference in the intensity of convection between the two winters primarily results from larger net heat losses during fall/winter 2008-2009 compared to fall/winter 2007-2008. At first, we consider the seasonal variability of currents and particle fluxes and interpret them in light of previous observations gathered in the northwestern Mediterranean Sea. Second, we will discuss the impact of the interannual variability of convection events and interactions with dense shelf water cascading on deep particle fluxes and the sustainment of a bottom turbid layer in the deep basin. Finally, we will discuss the effect of convection events on sedimentary substrates and deep-sea ecosystems.

Comparison with other Northwestern
Mediterranean oceanic sites al., 2012) the large TMF is rather originating from both deep DSWC and OOC events than from deep DSWC event only. Deep particle fluxes are also monitored at two other sites in the northwestern Mediterranean Sea, east of the Gulf of Lions: the ANTARES Site (42 • 50 N-6 • 10'E at 2400-m depth) and in the Ligurian Sea at the DYFAMED site (43 • 25 N-7 • 52 E at 2330-m depth). At the ANTARES site, TMF measured between July 1997 and January 1998 varied between 19 mg m −2 d −1 (end of August) and 352 mg m −2 d −1 (mid-October and mid-November) with a clear seasonal cutoff between the summer and fall periods (Amram et al., 2003). In the absence of any peculiar convection event and low current velocities (< 15 cm s −1 ) these TMF match with those observed during the same seasons in the Gulf of Lions (5-169 mg m −2 d −1 ). More recently, Al Ali et al. (2010) reported for the April to early-September 2005 period an average TMF of 77.9 mg m −2 d −1 and an average POC content of 5.5 % at 2235-m depth (150 mab). In early April 2005, TMF reached a maximum of 1331 mg m −2 d −1 , and POC content dropped to 1.5 %. Although no current meter data were presented, the authors suggested that the high flux and low POC content may be due to (1) sediment resuspension by deep water formation and/or (2) large particle export flux by intense DSWC event of winter 2004(Canals et al., 2006. Subsurface (200-m depth) and mid-depth (1000-m depth) particle fluxes are also monitored at DYFAMED site in the nearby Ligurian Sea (Miquel et al., 2011). Seasonal patterns of TMF in the Gulf of Lions, with highest export flux during winter-spring and lowest export flux during summer-fall, are similar to those observed at the DYFAMED site.

Variability of open-ocean convection and its impacts on deep particle fluxes
The different OOC intensities observed for both winters are typical for the interannual variability already observed in the region (Mertens and Schott, 1998;Bethoux et al., 2002;Herrmann et al., 2009). According to these authors, 18 yearly events of deep OOC, responsible for the ventilation of deep waters, occurred during the 1971-2008 period. Thus, our observations present a unique opportunity to compare two con-trasted winters and to assess the impact of deep OOC on deep particle fluxes. Compared to the mild convection in winter 2007-2008, the larger intensity of the convection during the winter 2008-2009 results in lower surface phytoplankton concentration in the water column in the convective area, a significant increase of near-bottom TMF, and an overall decrease in POC content (Figs. 2, 3, and 8a). Two potential mechanisms are considered to explain these differences: the variability of the exported particle fluxes from the surface layer, and resuspension of seabed.
Causal relationships between hydrological processes in, and around the convection region and the planktonic production have been shown based on modeling approaches (Lévy et al., 2000;Auger, 2011). These studies emphasize the major effect of vertical mixing and mesoscale activity on primary production and POC export. They showed that, for deep winter mixing biomass in the convective area is strongly diluted because of the limitation of photosynthesis by the reduced exposure time of phytoplankton to light, and mechanical decoupling of prey and predators. The majority of phytoplankton production is obtained at the rim of the convective area, Biogeosciences, 10, 1097-1116, 2013 www.biogeosciences.net/10/1097/2013/ where the mixed layer is shallowest. Auger (2011) further indicates that POC export by turbulence and advection culminate during the vertical mixing period, while POC export by sedimentation, which is directly related to the organic matter content, culminates during the spring phytoplankton bloom following the peak of vertical mixing. Hence, one might infer that the mild convection (reaching 700-m depth) in late March 2008 does not to affect mid-depth and near-bottom fluxes and POC content ( Fig. 8e and f). Conversely, the low biomass and primary production during the intense vertical mixing in late February 2009 likely contribute to the reduction of POC content observed at mid-depth and near the bottom (Fig. 8). After the mixing period, the increased TMF and POC content observed in April-May 2008, as well as the slight increase of POC content observed in late Marchearly April 2009 are probably caused by the settling and advection of organic matter produced in the surface layer during the late winter-early spring blooms. It is noteworthy that highest POC contents at 1000-m depth and near the bottom, which likely relate to rapid settling of organic matter produced during the fall bloom, are observed in both years (2008 and 2009) in October-November. The potential impact of deep OOC on sediment remobilization at 2330-m depth has been shown by  in the nearby Ligurian Sea. The authors showed that, during the winter 2005-2006 deep OOC event, nearbottom currents were markedly intensified (up to 39 cm s −1 ), and that near-bottom particle fluxes increased at the same time by two orders of magnitude. In the Gulf of Lions, TMF at mid-depth and near the bottom for the SC2350 site show very similar variations, except during the winter 2008-2009 when near-bottom TMF exceeded mid-depth fluxes by two orders of magnitude ( Fig. 8e and f). This abrupt increase of near-bottom TMF takes place at all sites and comes at a time when bottom currents significantly increase (average velocity ∼ 13 cm s −1 and peaks > 25 cm s −1 , Fig. 6). Such current velocity peaks, which exceeded the critical threshold for fine silts (< 16 µm) as calculated from the SEDTRANS model (Li and Amos, 2001), are probably large enough to resuspend the superficial fine sediments of the deep basin. The strong vertical mixing ( Fig. 5a and b) and significant mesoscale activity of horizontal currents (Fig. 7) also allow the dispersal of the suspended sediment several hundred meters above the seabed (Fig. 9b) and over a widespread area. Moreover, the contribution of sedimentary material poor in POC along with the dilution of the biomass by vertical mixing result in low POC content in near-bottom sediment trap samples. Thus, in the absence of noticeable lateral input from the nearby margin by DSWC that remains mild during the period of study (Fig. 5c) The authors relate the large increase of near-bottom flux, which peaked at 16270 mg m −2 d −1 , primarily to lateral advection of dense turbid water cascading from the western shelf of the Gulf of Lions. They considered that potential contribution of deep sediment resuspension by strong currents (between 20 and 47 cm s −1 ) could have been also possible, but less intense due to the position of the trap farther from the seabed (250 mab). Given the absence of turbid bottom layer in the years preceding this event  the availability of easily resuspendable sediment may have been reduced. However, other studies conducted in 2005-2006 on the western slope of the Gulf of Lions showed that the maximum near-bottom (30 mab) TMF collected at 1900-m depth during the intense 2006 DSWC event range between 3200 and 5700 mg m −2 d −1 and were about one order of magnitude lower than the near-bottom TMF at 1000-m depth on the upper slope and canyons that range between 34 100 and 90 080 mg m −2 d −1 (Pasqual et al., 2010). The latter work concluded to a significant decrease in the transport capacity of the dense shelf-water plume along its down-canyon and down-slope propagation. These observations suggest that the particle load transported during this deep DSWC event would have been even more diluted when reaching the deeper OOC area. Fluxes measured at 1900-m depth during the winter 2005-2006, however, are four times lower than those measured at 2400-m depth during the winter 2004-2005, while these two years were characterized by intense DSWC and OOC events. The large discrepancy suggests an additional source of particulate matter in the deep basin, such as sediment resuspension by open-ocean deep convection, must be considered to explain the significant increase in particulate fluxes during winter 2004-2005. We propose here three different scenarios to better describe the plausible origin of the near-bottom TMF in the deep basin of the Gulf of Lions during winter: (1) moderate TMF (< 1000 mg m −2 d −1 ) from a dominant biological source resulting from the surface production export during the winters of shallow OOC, as in 2008; (2) large TMF (up to 10 000 mg m −2 d −1 ) from a dominant sedimentary source due to the remobilization of surface sediment of the deep basin during winters with both deep OOC and shallow DSWC, as in 2009; and (3) extreme TMF (> 10 000 mg m −2 d −1 ) from a dominant sedimentary source due to the remobilization of surface sediments from the shelf and slope and mainly from the deep basin during winters with both deep OOC and deep DSWC, as in 2005.
In a compilation of numerous CTD casts conducted in the western Mediterranean from 1998 to 2011, Puig et al. (2012) show that the large bottom nepheloid layer observed during our study period (2007)(2008)(2009)  the replenishment of the bottom nepheloid layer to the remobilization of basin sediment by deep OOC.

Impacts of open-ocean deep convection events on seabed and benthic ecosystems
While deep OOC is very infrequent in the Ligurian Sea (one single event for the 1995-2007 period as observed by Marty and Chiavérini, 2010) and thus rarely affects the deep sediment (Martin et al., 2010). On the contrary, the recurrent deep OOC events in the Gulf of Lions continuously rework bottom sediments and a long-term alteration of the sediment might be expected. Various geophysical and acoustic imagery investigations performed on the slope and rise of the Gulf of Lions (Kenyon et al., 1995;Droz et al., 2001;Bonnel et al., 2005;Jallet and Giresse, 2005;Lastras et al., 2007) revealed erosional features generated by bedload transport processes (sediment waves, crescent scours, furrows, grooves, and mega-ripples). Examples of such bedforms close to the mooring sites are given in Fig. 10. The question of the origin of these erosional bedforms has been little investigated. On the slope and particularly in the Cap de Creus Canyon, Canals et al. (2006) and Puig et al. (2008) show that the rapid downward bottom currents and sand transport associated with DSWC events are prone to generate distinctive erosional bedforms, such as megascale longitudinal furrows, which extend also over the middle canyon down to 1400-m depth. On the rise, the flattening of the seabed and the decreasing of density contrast of the dense water plume with the ambient water slow the cascading currents and its erosive potential .
At the base of slope, the observed strong bottom currents in the present and previous studies (Millot and Monaco, 1984;Schott et al., 1996;Palanques et al., 2009) mostly relate to the deep OOC events. The periodical winnowing of surface sediments by these currents is thus thought to be responsible of the low sedimentation rates (Table 3). The observed giant erosional bedforms of the seabed (Fig. 10) form under unidirectional strong bottom currents that cannot be generated by deep OOC events, characterized by changing directions (Fig. 7). Depending on their orientation and shape they are interpreted (1) as formed by recent bottom current from the Cap de Creus and Sète Canyon, possibly triggered after DSWC events (Lastras et al. 2007), (2) as relict forms related to turbiditic current spillover from the adjacent Petit-Rhone turbiditic channel (Kenyon et al., 1995;Bonnel et al., 2005;Wynn et al., 2005). It has been pointed out that the floor of the giant scours is devoid of Holocene deposits, while a Holocene drape is well developed on the adjacent seabed ( Fig. 10 in Dennielou et al., 2009). This can be interpreted as a possible imprint of the interaction of the deep OOC with the seabed. Unfocused bottom currents could undergo local acceleration inside the scours, sufficient to prevent the deposition of sediment through the Holocene.
To date, the effect of dense water formation on deep benthic ecosystems in the Gulf of Lions has been only described for DSWC event (Pusceddu et al., 2010. The rapid sinking of dense water results in important immediate modifications that can be summarized in (i) increased organic matter content in the deep basin; (ii) lower benthic abundance; and (iii) modified benthic biodiversity. At longer time scales the analysis reveals that the Gulf of Lions deep benthic ecosystems is characterized by a very fast (months) recovery after such intrusions .
Our observations reveal the effect of OOC on POC load in the deep water layer. Inputs of POC exported from the surface layer following the winter 2007-2008 mild OOC or primarily released by resuspended sediments during the winter 2008-2009 deep OOC event may fuel the deep-sea pelagic microorganisms and trigger enhanced biological activity (Boutrif et al., 2012). The POC derived from sediment resuspension is generally much older (up to several thousand years old as described by Tesi et al. (2008Tesi et al. ( , 2010) and more refractory than the POC from the downward export of modern marine phytodetritus. The response of deep-sea ecosystems is thus likely dependant on the intensity of OOC that regulate the quantity and quality (i.e., labile vs. refractory) of the POC.
Besides, several studies conducted in other dynamic deep environments around the world indicated that sediment reworking can be ecologically important on the distribution, abundance, and structure of meiofauna associated with the sediment surface (Aller, 1989;Gage et al., 1995;Thistle et al., 1999). Erosional periods can be either beneficial (dispersal enhancement, benthic crowding relief, sediment properties improvement) and detrimental (damaging of animals by removal from the seabed, expatriation, exposure to water predators) to benthic organisms. It is thought that deep OOC events in the Gulf of Lions result in positive or negative effects on the benthos, by fuelling the deep sea floor with large amounts of bioavailable particles or by disrupting the benthic habitats.

Conclusions
The objective of this 1.5-yr study conducted in the deep basin of the Gulf of Lions (northwestern Mediterranean Sea) was to understand the impact of the open-ocean convection intensity that controls the seasonal variability of the hydrology, hydrodynamics and biogeochemistry of this open-ocean region, on the mid-depth and near-bottom particulate fluxes measured by sediment traps. The major outcomes of this study are 2. By comparison with previous studies, range and primary origin of near-bottom fluxes are believed to vary according to the deep-water formation intensity from (1) moderate TMF (< 1000 mg m −2 d −1 ) from a dominant biological source resulting from the surface production export during the winters of shallow OOC; (2) large TMF (up to 10 000 mg m −2 d −1 ) from a dominant sedimentary source due to the remobilization of surface sediment of the deep basin during winters with deep OOC and shallow DSWC; and (3) extreme TMF (> 10 000 mg m −2 d −1 ) from a dominant sedimentary source due to the remobilization of surface sediments from the shelf and slope and mainly from the deep basin during winters with deep OOC and deep DSWC.
3. The observations suggest that the recurrence of deep OOC in the area has a long-term effect on seabed morphology (i.e., by prevention of particle sedimentation) and thus should be considered as a major driving force for deep sedimentary dynamics.
4. Open-ocean deep convection has to be considered, together with dense shelf water cascading, as a major driving force for benthic ecosystems as it occasionally fuels them with labile POC from the surface layer and/or disrupts the benthic habitats by reworking superficial sediment. A better understanding of the composition of the particulate flux (i.e., biogenic and lithogenic contents), and its associated elements (including contaminants) generated by deep OOC events is now required to better assess the impact of such events on benthic ecosystems.