Late summer particulate organic carbon export and twilight zone remineralisation in the Atlantic sector of the Southern Ocean

Introduction Conclusions References

. A high surface export production, which can account for up to 30-50 % of the Net Primary Production (NPP), has been proposed to be partly related to seasonal blooms of diatoms (Rutgers van der Loeff et al., 1997. The usually observed strong decrease of C export flux with depth indicates intense attenuation of downward fluxes of biogenic particles (Martin et al., 1987). This was also supported by studies of the particulate barium distribution (Cardinal et al., 2005;Dehairs et al., 1997;Jacquet et al., 2008) and in a lesser extent observations about excess 234 Th activity (Savoye et al., 2004a;Usbeck et al., 2002), which help documenting the remineralization length scale for organic matter (OM) in SO mesopelagic waters. Despite these substantial progresses, the SO remains an oceanic 10 region largely unresolved in terms of observations and experiments, and therefore, large discrepancies between model estimates persist (Gruber et al., 2009;McNeil et al., 2007). Better constraining the processes that favor long-term sequestration of carbon in the Austral Ocean still represents a major scientific issue. We report here new estimates of the POC export production from the Atlantic sector 15 of the Southern Ocean, along a transect from the Cape Basin to the Northern Weddell gyre as part of the BONUS GoodHope (BGH) expedition (R/V Marion Dufresne) during the IPY 2008 (February-March 2008). The observed trends are discussed as a function of the successive frontal systems crossed by the BGH section between the Cape Basin and the northern Weddell Gyre. 20 The POC export fluxes were inferred from measurements of the short-lived radionuclide 234 Th (t 1/2 = 24.1 d), which is now recognized as a robust proxy of the short-term dynamics of biogenic particles (Cochran and Masqué, 2003). Naturally occurring 234 Th is the decay product of 238 U, which is conservatively distributed in the open ocean, proportional to salinity (Chen et al., 1986;Pates and Muir, 2007). Unlike 238 U, 234 Th has a 25 strong affinity for particulate matter and its activity distribution through the water column offers a means for quantifying export flux and aggregation/disaggregation of particles on regional and seasonal scales (Buesseler et al., 1992 upper ocean and mesopelagic export of POC (Cochran and Masqué, 2003;Maiti et al., 2010;Savoye et al., 2004a). We also report mesopelagic carbon remineralisation fluxes estimated from excess Ba (Ba xs ; particulate Ba corrected for the lithogenic contribution). Ba xs profiles in the open ocean are characterized by maximum concentrations in the upper mesopelagic (∼150-500 m). This Ba xs is mostly present under the 5 form of micro-crystralline barite (BaSO 4 ) (Dehairs et al., 1980;Sternberg et al., 2008) and its formation is related to the decay of phytoplankton. Barite is precipitated in oversaturated micro-environments, mostly aggregates of organic material where bacterial activity is intense (Ganeshram et al., 2003). When micro-environments disintegrate and become remineralised in the mesopelagic zone, discrete barite crystals are re-10 leased and the Ba xs content can be related to carbon remineralisation activity (van Beek et al., 2009;Jacquet et al., 2008Jacquet et al., , 2011Sternberg et al., 2008). The time-scale involved in this process represents a few days to a few weeks (Ganeshram et al., 2003;Cardinal et al., 2005;Jacquet et al., 2008) whereby short term variabilities are smoothed out.
15 2 Materials and methods

Study area
The cruise track of R/V Marion Dufresne during the BONUS GoodHope (BGH) expedition (8 February-24 March 2008) is shown in Fig. 1 together with the position of major hydrographic fronts encountered along the north to south transect between Cape 20 Basin and northern Weddell Gyre. Starting from the subtropical domain, the cruise track crossed the southern subtropical front (S-STF), the sub-Antarctic front (SAF), the polar front (PF), the southern Antarctic circumpolar current front (SACCF), and finally the southern boundary of the ACC (Sbdy). Eleven stations were sampled for total 234 Th activity. Among these, five stations were also sampled for particulate 234 Th (S1 to S5) 25 in order to obtain POC/ 234 Th ratios of sinking particulate matter and derive POC export fluxes. Eight stations have been sampled for the Baxs proxy (S1 to S5 in addition to L3, L5 and L7).

Determination of total 234 Th activity
Total 234 Th activities were obtained from small volume (4 L) seawater samples collected from 12 L Niskin bottles. As detailed in Appendix 1, the samples for super stations (S1 Seawater samples were processed for total 234 Th activity measurement following the procedure developed by Pike et al. (2005). Samples were acidified to pH 2 using nitric acid and spiked with a known amount of 230 Th which serves as a yield monitor for 234 Th recovery. After 12 h equilibration the pH was raised to 8.5 using concentrated NH 4 OH 15 and Th was co-precipitated by adding 100 µL of MnCl 2 (2.0 g L −1 ) and 100 µL of KMnO 4 (7.5 g L −1 ). The samples were allowed to stand for 12 h before filtration on high-purity quartz microfiber filters (QMA, Sartorius; nominal pore size = 1 µm; Ø 25 mm). Filtered samples were dried overnight, mounted on nylon filter holders and covered with Mylar film and Al foil. On board each sample was counted twice using a low level beta counter 20 (RISØ, Denmark). Beta counting was allowed to proceed till counting uncertainty was below 2 % RSD. Residual beta activity was estimated for each sample after a delay of six 234 Th half-lifes (∼6 months) and was subtracted from the gross counts recorded on-board. For Th recovery, filters were dismounted and MnO 2 precipitates dissolved in 10 ml Introduction effect. Measurement uncertainty of 230 Th/ 229 Th ratios, in terms of Relative Standard Deviation (RSD), ranged from 0.1 to 1.6 % (n = 3 replicates) with dilution factors of 5 to 20. Estimated reproducibility of the method, evaluated with 9 standard solutions prepared separately and determined over different analytical sessions, was also particularly good and ranged from 0.5 to 1.3 %. The precision obtained with this simpli-15 fied procedure meets the requirements defined by Pike et al. (2005), who emphasize the need to achieve 229 Th/ 230 Th ratio errors of ≤2 % in order to reach accurate 234 Th activities. Th recoveries were estimated for every sample processed (n = 175) and measurement precision as obtained from triplicate analyses were all below 2 % RSD. Average Th recovery was 87 ± 2 % (n = 175). Uncertainties on total 234 Th activity are 20 reported in Appendix 1 and represent on average 0.10 dpm L −1 . The parent 238 U activity was estimated with salinity measurements using the relationship of Pates and Muir (2007):  (Savoye et al., 2006) and changes over time can be described using the following equation: Where λ is the 234 Th decay constant (0.0288 d −1 ); A U and A Th represent total 238 U and 234 Th activities, respectively; P is the net loss of 234 Th on sinking particles (i.e. vertical 234 Th flux) expressed in dpm m −2 d −1 , and V is the sum of the advective and diffusive fluxes.
If we neglect advective and diffuse fluxes (V = 0), the vertical flux of 234 Th (P) can 10 be estimated using the steady state (SS) approach ( d A Th d t = 0) In this study sites were not revisited over time (but see later) and therefore a first calculation was made using the SS assumption. In this case Eq. (1) can be simplified into: High resolution profiles of 234 Th activity in the upper 1000 m allow the export flux to be estimated from the upper ocean (surface export) as well as from the mesopelagic zone at 600 m depth (mesopelagic export). Integration of 234 Th activity along the water column was performed using a mid-point integration method. 234 Th surface export flux was also estimated using the non steady state (NSS) model NSS approach. NSS export flux at 100 m was estimated using the following equation (Savoye et al., 2006): where ∆t is the time interval between two successive occupations of a given site; A Th 1 and A Th 2 are the 234 Th activities for the first and second visit, respectively. This flux 5 approach assumes that: (i) the same water mass is sampled during both visits; (ii) 238 U activity is constant, and (iii) diffusive and advective fluxes are negligible (V=0).

Measurements of particulate 234 Th and POC
For particulate 234 Th and POC, suspended particulate matter was collected at five stations (S1, S2, S3, S4 and S5) via in situ large-volume filtration (150-2000 L) systems 10 (Challenger Oceanics and McLane WTS6-1-142LV pumps) equipped with 142 mm diameter filter holders. Two particle size classes (>53 µm and 53 1 µm) were collected via sequential filtration through a 53 µm mesh nylon screen (filter SEFAR-PETEX®; polyester) and a 1 µm pore size quartz fiber filter (QMA, Pall Life). Because suspended particles were also intended for other analyses by other participants, 14 C POC 15 and 210 Pb/ 210 Po, biomarkers, the filters were pre-conditioned prior to sampling. The PETEX screens were soaked in HCl 5 %, rinsed with Milli-Q grade water, dried at ambient temperature in a laminar flow hood and stored in clean plastic bags. QMA filters were precombusted at 450 • C during 4h and filters were stored in clean plastic bags before use. 20 After collection, filters were subsampled for the different end-users using sterile scalpels, a custom-build INOX steel support for 53 µm PETEX screens and a plexiglass punch of 25 mm diameter for QMA filters. For large size particles (>53 µm), particles on the PETEX screen parts dedicated to 234 Th were re-suspended in filtered seawater in a laminar flow hood, and collected on 25 mm diameter silver filters (1.0 µm porosity). Introduction Silver and QMA filters were dried overnight, and once mounted on nylon holders and covered with Mylar and Al foil were ready for beta counting. As for total 234 Th activity, particulate samples were counted twice on board until relative standard deviation was below 2 %. Residual beta activity was measured in the home-based laboratory after six 234 Th half-lifes (∼6 months).

5
Following beta counting, particulate samples were processed for POC measurement by Elemental Analyzer -Isotope Ratio Mass Spectrometer (EA-IRMS). Sizefractionated samples were dismounted from filters holders and fumed under HCl vapor during 4 h inside a glass desiccator, to remove the carbonate phase. After overnight drying at 50 • C, samples were packed in silver cups and analyzed with a Carlo Erba NA 10 2100 elemental analyzer configured for C analysis and coupled on-line via a Con-Flo III interface to a Thermo-Finnigan Delta V isotope ratio mass spectrometer. Results obtained for C isotopic composition of POC are not included in this work. Acetanilide standard was used for C concentration calibration. C blanks were 0.98 µmol and 0.54 µmol for QMA and silver filters, respectively. Results obtained for bulk POC and two size-15 segregated POC fractions (>53 µm and 53 1 µm) are reported in Table 1 along with  particulate 234 Th activity measured on the same samples.

Ba xs sampling and measurements
19-20 depths per station were sampled in the upper 1000 m using CTD rosette equipped with 12L Niskin bottles. were evaporated close to dryness and redissolved into ∼13 ml of HNO 3 2 %. The solutions were analysed by ICP-MS X Series 2 (Thermo) equipped with a Collision Cell Technology (CCT). Ba, Na and Al contents were analysed simultaneously (with CCT for Al and without for Ba and Na). To check whether internal standards ( 99 Ru, 115 In, 187 Re, 209 Bi) adequately corrected possible matrix effects, we analysed several cer-5 tified materials which also served to construct calibration curves. These standards solutions consisted of dilute acid-digested rocks (e.g. BHVO-1, GA, SGR-1), natural water (SLRS-4) and multi-element artificial solutions. Based on analyses of these standards, precision, accuracy and reproducibility are better than ± 5%. For more details on sample processing and analysis we refer to Cardinal et al. (2001). Detection limit in solution was calculated as three times the standard deviation of the on-board blanks and reaches 20 and 0.5 ppb for Al and Ba, respectively. BGH samples are largely exceeding this detection limit for Ba and on-board filtration blanks represented only 2 ± 0.8 % of the average sample Ba content. For Al, 23 over a total of 160 samples are below detection limit (DL), but concentrations are most of the time very close to DL. Indeed,

15
Al for on-board blanks represents 28 ± 14 % of average sample Al content. However, this did not significantly affect the Ba xs concen trations and the remineralisation fluxes, as discussed later.
Values of on-board prepared blanks were subtracted from sample values and excess Ba calculated by correcting total Ba for the lithogenic Ba contribution, using sample Al 20 content and a Ba:Al crustal molar ratio of 0.00135 (Taylor et McLennan, 1985). Na was also analysed to correct any sea-salt contribution to Ba xs . Remnant sea-salt was found to have but a negligible affect on Ba xs .

Carbon fluxes calculations from Ba xs depth profiles
Remineralisation carbon fluxes can be estimated using a relationship observed in ACC Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | from a 1-D advection diffusion model (Dehairs et al., 1997Shopova et al., 1995): where J O 2 is the O 2 consumption (µmol l −1 d −1 ), meso-Ba xs is the observed depthweighted average Ba xs value in the upper mesopelagic waters (125 to 600 m depth interval) and Ba residual is the residual Ba xs signal at zero oxygen consumption. For the 5 BaSO 4 saturated water column of the ACC this residual Ba xs was estimated to reach 180 pM (Monnin and Cividini, 2006;Monnin et al., 1999). Such value is expected also to prevail in deep waters (>600 m) were remineralisation is minimal compared to the upper mesopelagic. In the present study deep ocean Ba xs values (800-1000 m) are generally close to 200 pM. This also holds for the deep SAZ and the STZ waters 10 (stations S1, S2 and L3), known to be undersaturated for BaSO 4 (Monnin and Cividini, 2006;Monnin et al., 1999) and which therefore are expected to have smaller residual Ba xs contents. Since that is not observed here, we choose to apply a single value of 180 pM for Ba residual at all stations. Calculated J O 2 was then converted into carbon respired (C respired ) by: C respired is the organic carbon remineralisation rate (in mmolC m −2 d −1 ), Z is the thickness of the mesopelagic layer considered (i.e., 600-125 = 475 m), and RR is the Redfield C:O 2 molar ratio (125:175).

234 Th/ 238 U ratio profiles
The full data set, including activities of total 234 Th and 238 U, and corresponding 234 Th/ 238 U ratios can be found in Appendix 1. Figure 2  ratio decreases again between at 80 and 100 m depth (Fig. 2). A depletion of 234 Th indicates loss of 234 Th via particle scavenging and suggests a surface origin for this 30 m thick, less saline (salinity = 34.9), subsurface water tongue. This may be related to the particular hydrography in this zone, where cyclonic eddies can contribute to the subduction of surface waters of westward origin (Chever et al., 2010;Gladyshev et 20 al., 2008). Except for this particular feature at S1 it appears that deep total 234 Th activities in the subtropical zone (S1 and L2) are close to secular equilibrium with 238 U.
However, south of the STF substantial deep 234 Th excess is observed in the PFZ at 44.9 • S (station L3), at 46.0 • S (station L4) and at 47.5 • S (station S3), and also to some extent at 42.5 • S in the SAZ (station S2). 234 Th/ 238 U activity ratios largely >1.1 clearly 25 indicate substantial accumulation of 234 Th (i.e., excess 234 Th relative to 238 U) in the meso-pelagic zone, which can be attributed to particle remineralisation and/or disaggregation (Buesseler et al., 2008;Maiti et al., 2010;Savoye et al., 2004a returns to secular equilibrium. However, further south, in the northern part of the Weddell sea at 57.6 • S (station S5) we again observe layers of excess 234 Th activity (just below the ML and at 250 m depth; see Fig. 2).

Export flux of 234 Th
(a) Surface export flux 5 The fluxes of 234 Th out of the surface layer (upper 100 m and at MLD) were calculated using steady state (SS) and non steady state (NSS) models (Table 2  is about 30 % smaller than the flux from the ML. The SS export flux exhibits a latitudinal gradient which follows the structure of the upper ML (Fig. 4a). The smallest SS fluxes are observed in the STZ (stations S1 and L2), and also south of the ACC in the northern branch of the Weddell Sea Gyre (station S5) where the ML remains relatively shallow. 234 Th export is highest within the ACC, especially in its southern part between 20 the PFZ to the AZ (stations S3 to L7), where the upper ML extends quite deep (80 to 120 m) Although the steady state model is by far the most commonly used model in 234 Thexport studies, the assumption of constant 234 Th activity with time is only valid for uniform scavenging conditions (constant export production) (Savoye et al., 2006 in the mesopelagic zone (1.0 µmol dpm −1 at S1, 1.4 µmol dpm −1 at S2; 1.7 µmol dpm −1 at S3; 2.8 µmol dpm −1 at S4; 3.6 µmol dpm −1 at S5). For large (>53 µm) particles, POC/ 234 Th p ratios are much more variable in the mesopelagic zone (Fig. 6). A trend of decreasing POC/ 234 Th p ratios can be observed at S1 and S4 while ratios remain relatively unchanged, or even increase with depth at S3 and S2, respectively. At S3 the 15 very large values (14.1 to 16.3 µmol dpm −1 ) reported for the >53 µm particles between 400 and 500 m are likely due to zooplankton collected on the filter (revealed by visual inspection). For estimating POC export fluxes, the POC/ 234 Th p ratio of sinking particles at the export depth has to be determined (Buesseler et al., 1992). As recommended by 20 Buesseler et al. (2006), we consider the POC/ 234 Th p ratio of large (>53 µm) particles as representative of sinking material leaving the upper ML. In this study and as illustrated in Fig. 6, no clear relationship exists between POC/ 234 Th p ratios of >53 µm particles and depth making it difficult to calculate this ratio for the exact 100 m depth horizon using a fitting function, such as for instance a power law (Jacquet et al., 2011). In order 25 to take into account the observed POC/ 234 Th p variability below the export zone, we consider the average POC/ 234 Th p ratios between the basis of the ML and 300 m depth  (Fig. 7). The smallest ratios are observed in the STZ (S1; 1.9 ± 0.2 µmol dpm −1 , n = 3) and the SAZ (S2; 1.7 ± 0.2 µmol dpm −1 , n = 4). South of the SAF, the POC/ 234 Th p ratio of sinking particles gradually increases from 3.0 ± 0.2 µmol dpm −1 in the PFZ (n = 2; S3) to 5.6 ± 1.3 µmol dpm −1 in the southern ACC (n = 3; S4) and 4.1 ± 1.7 µmol dpm −1 in the WG (n = 4; S5). Such a southward increase of POC/ 234 Th p ratios of sinking 5 particles was also observed in the SE-Pacific, between the SAZ and the Ross Sea for particles sizing >70 µm (Buesseler et al., 2001).
To check the robustness of our averaging approach for evaluating POC/ 234 Th ratios of sinking particles we also fitted power law functions to the vertical profiles of POC and 234 Th p of the >53 µm particle size fraction ( between ML and 300 m (Fig. 7).

15
Though overall our POC/ 234 Th p ratios of sinking particles (Fig. 7) are of similar magnitude as those reported by others for the ACC, regional differences exist. For instance, POC/ 234 Th p ratios of sinking material in the PFZ (S3) and the WG (S5) are 1.5 to 1.8 times larger than those for the >70 µm size fraction sampled at the same locations 18 to 22 days earlier (Rutgers van der Loeff et al., 2011). Our range of POC/ 234 Th p ratio 20 for the >53 µm fraction (1.7-4.8 µmol dpm −1 ) is larger than the one reported by Coppola et al. (2005) for the Indian sector of the ACC (0.8-1.4 µmol dpm −1 ) but smaller than the ratios measured during bloom conditions in natural iron-fertilized settings close to Crozet (5.5-10.8 µmol dpm −1 ; Morris et al., 2007) and Kerguelen island ( at 100 m). For the PFZ, our POC/ 234 Th p ratio (3.01 ± 0.21 µmol dpm −1 ) is also smaller than the one observed in the PFZ during SAZ-Sense (5.13 ± 0.83 µmol dpm −1 at 100 m; Jacquet et al., 2011). At "L" stations we did not deploy large volume in-situ pumps. Therefore, for stations located close to a biogeochemical boundary (L2-STF, L3-SAF, L6-PF, and L7-Sbdy), 5 POC/ 234 Th p ratios were calculated by averaging the POC/ 234 Th p ratios measured in adjacent northern and southern zones (Table 3). For stations L4 and L5, located in the PFZ, we used the POC/ 234 Th p ratio obtained for S3 in the PFZ to calculate the POC flux.

15
Overall EP 100 ranges from 0.9 to 5.1 mmol m −2 d −1 and from 0.3 to 4.9 mmol m −2 d −1 based on the SS and NSS model, respectively (Fig. 8a; Table 3). The SS EP 100 increases progressively from north to south. It remains low in the northern part of the transect, from 0.9 to 1.9 mmol m −2 d −1 in the STZ (S1 and L2) and 1.7 mmol m −2 d EP 100 fluxes integrating the 15 to 22 day period preceding the BGH cruise also exhibit a latitudinal gradient but the variability is larger compared to the SS approach (Fig. 8a). The highest NSS EP 100 fluxes are observed in the AZ, south of the ACC (4.9 ± 3.2 mmol m −2 d −1 ) and in the WG (

Mesopelagic carbon export flux
In order to estimate the POC flux attenuation between 100-600 m depth we multi-20 plied the accumulation fluxes of excess 234 Th obtained in mesopelagic waters (based on SS model calculations) with the POC/ 234 Th p ratios of sinking particles (Fig. 8b).

15
A similar discrepancy between surface export and subsurface remineralisation is reported by Savoye et al. (2004a) for the AZ and in the Seasonal Ice Zone (SIZ) of the Australian sector. A possible explanation for this observed imbalance may be the decoupling of surface and mesopelagic processes, due for instance to lateral advection of surface waters. The strong eastward surface current in the central ACC may have 20 advected surface waters with lower 234 Th deficit and lower particle export relative to the signal captured at mesopelagic depth.

Ba xs profiles
Surface waters are depleted in Ba xs . Concentrations start to increase at the basis of the MLD where the density gradient gets steeper (Figs. 2 and 3b). The depth where the buildup of the meso-Ba xs starts is shallow (∼ 50 m) in the STZ and SAZ but then progressively increases in the ACC (∼100 m) to shoal again slightly southward. This follows quite well the latitudinal variation of the MLD (Fig. 4a and Table 2). This is consistent with previous observations and supports the view that aggregates formed at the basis of the mixed layer are loci where micro-barites precipitate (Cardinal et al., 2005). The Ba xs contents are usually maximal in the 200-400 m layer but high values exceeding 300 pM can extend down to 600-800 m (L3, SAF; L5, PFZ; S5, WG). The highest 5 value for the whole transect is reached at station L3 (SAF) at 250 m (>1000 pM). Such high values have already been reported on the SAF and SAZ (Jacquet et al., 2005). This Ba xs maximum is surrounded by values which remain high (>400 pM) over the 125-475 depth range. Ba xs contents are the lowest for the northernmost (STZ-SAZ) and the southernmost 10 (SACCF-AZ-WG) parts of the BGH section. This spatial variability is also clearly expressed in the depth weighted average mesopelagic Ba xs (meso-Ba xs ) contents (125-600 m; Table 4). Meso-Ba xs is minimal at S1 (STZ; 168 pM) and maximal at L3 (SAF; 497 pM). The two PFZ stations have meso-Ba xs contents exceeding 300 pM while all other stations have moderate meso-Ba xs contents (235-277 pM). Such a trend with 15 maximum meso-Ba xs values around the PFZ and lower values northward and southward has already been observed earlier (Cardinal et al., 2005). The variations of meso-Ba xs compare rather well with the excess Th flux integrated over the 100-600 m depth layer (Fig. 9a). Excluding one outlier (S3-PFZ station which exhibits by far the highest excess Th flux) there is a significant relationship between 20 these two geochemical parameters (R 2 = 0.73 p = 0.015, Fig. 8a).

Steady-state vs. non steady-state surface export production
The 234 Th-based approach reveals a latitudinal gradient of POC export production in late summer (Fig. 8a). The overall picture shows that SS EP 100 gradually increases from low values in the STZ (L2; 0.9 ± 0.2 mmol m 2 d −1 ) and the SAZ (S2;  (Chever et al., 2010;Klunder et al., 2011). Surface dis- 5 solved Fe (DFe) concentrations exhibit a rapid drawdown over 21 days, decreasing from 0.33 nM at 55.0 • S (station 119) and 0.34 nM at 56.0 • S (station 122)  to <0.1 nM at 55.2 • S (station L7) and 0.14 nM at 57.6 • S (station S5) (Chever et al., 2010). This fast assimilation of DFe (over 22 days) gives support for late summer phytoplanktonic production which may have triggered surface POC export.

POC/ 234 Th p ratio of sinking particles
The latitudinal trend of POC/ 234 Th p ratios (Fig. 7) suggests that sinking particles have variable properties, as depending on plankton community and/or particle dynamics. The increase of the POC/ 234 Th p ratio for the >53 µm fraction south of the PF would be consistent with larger particle volume to surface areas (V:SA), thus indicating an 15 increasing contribution of larger cells to export (Buesseler et al., 2006). Indeed, the occurrence of high POC/ 234 Th p ratios is well documented for SO high latitude systems dominated by diatoms (Buesseler et al., 2001(Buesseler et al., , 2003(Buesseler et al., , 2005Friedrich et Rutgers van der Loeff, 2002;Rutgers van der Loeff et al., 1997, 2002Savoye et al., 2008). This relationship between high POC/ 234 Th p ratio and diatom abundance seems to hold true small particles consistent with low POC/ 234 Th p ratios in sinking particles (Buesseler et al., 2006). South of the PF (50.4 • S), which marks the transition from oligotrophic to HNLC conditions with silicate-rich (>10 µm) surface waters (Bown et al., 2011;Fripiat et al., 2011), diatom abundance progressively increases with latitude and reach a maximum at 57.6 • S in the WG (S5). The size distribution of surface POC and 234 Th p 5 at this station contrasts with the ones at northern stations, with up to 26 % and 37 % of total POC and 234 Th p associated with the >53 µm size fraction, respectively. This suggests the influence of larger diatom cells to the surface water particle population and consequently could explain the higher POC/ 234 Th p ratio observed in sinking material. The surface water POC/ 234 Th p ratios of large and small particles differ between sites 10 (see Fig. 6). At S1, surface POC/ 234 Th p ratios of fine and large particles are similar (4.7 and 3.7 µmol dpm −1 , respectively). At S2, S3, and S5, POC/ 234 Th p ratios increase with decreasing particle size by a factor 1.5 (S2) to 2.0 (S3). However, at S4, the large particle POC/ 234 Th p ratio exceeds the ratio in small particles by 2.1 times. Buesseler et al. (2006) report an increase of POC/ 234 Th p ratios with particle size for different 15 oceanographic settings, including the SO. In our case only station S4 conforms to this pattern which fits the V:SA model, pointing to a dominant influence of increasing cell size on particle POC/ 234 Th p ratio. Decreasing or unchanging ratios with particle size, as observed at S1, S2, S3, and S5, must involve other controlling factors. Our results for the latter 4 stations are in good agreement with POC/ 234 Th p ratios obtained 20 for three particle size classes (1-10 µm, 10-50 µm, and >50 µm) during the ANTXXIV cruise conducted along the same section as BGH 3 weeks earlier (Rutgers van der Loeff et al., 2011). Similar POC/ 234 Th p ratios for both particle size classes as observed in the STZ (S1) may be consistent with rapid aggregation of small particles into larger sinking ones, possibly reflecting the impact of TEP-producing phytoplankton species 25 (Buesseler et al., 2006). On the other hand decreasing POC/ 234 Th p ratios with particle size, as observed for surface waters at S3 in the PFZ (2.4 µmol dpm −1 ; 1-53 µm fraction, 5.0 µmol dpm −1 ; >53 µm fraction) and to a lesser extent S2 (SAZ) and S5 (WG)  (Fig. 6), may reflect preferential C loss relative to 234 Th during large particle generation. This may include C degradation and recycling in the surface as well as variable C assimilation rates between trophic levels, including production of fecal material by zooplankton (Buesseler et al., 2006). 5 Chlorophyll a, POC and PON measurements reveal that phytoplankton abundance was highest in the STZ (L2) and in the SAZ (S2) in late summer 2008 (Joubert et al., 2011).

Surface export and biological production
To the south, algal biomass decreases progressively in the PFZ and reaches minimum values between the SACCF and the Sbdy. Particulate 234 Th determined at "Super" stations appears closely related to biomass distribution: the regression of surface POC 10 concentration (obtained from in-situ pumps sampling) and particulate 234 Th activity of total SPM yields a correlation coefficient (R 2 ) of 0.911 (n = 7). This relationship is preserved (R 2 : 0.808, n = 48) when considering full water column data obtained from in-situ pumps samples. Although particulate 234 Th appears to mirror plankton abundance, surface export production (EP100 and EPML) does not display any relationship 15 with algal biomass. In the STZ and in the SAZ, POC export fluxes are minimal, whereas in the low Chl a and POC area at S4 in the AZ, EP100 is highest (Fig. 8a). Insights into the processes controlling surface POC export can be given by nitrogen uptake measurements carried out using 15 N-labelled nitrate, ammonium and urea (Joubert et al., 2011). Results obtained in that study indicate that the late summer olig-20 otrophic conditions observed in the STZ support a phytoplanktonic community based on regenerated production with low f-ratio (0.2) and with nitrogen uptake being dominated by urea (70 % of total N uptake). This regenerated-based community appears dominated by small size phytoplankton, 51 % of Chl a is associated with picophytoplankton (<2 µm), what is consistent with the low POC flux deduced from 234 Th in 25 this zone (0.9-1.8 mmol m −2 d −1 ). To the south, the decrease of regenerated production documented by Joubert et al. (2011) and which is concurrent with a decreasing contribution of smaller sized phytoplanktonic, parallels the trend of increasing EP100 (Fig. 8a), indicating that enhanced nutrient recycling within the microbial loop appears to impact on POC export. Comparison between EP100 fluxes and urea uptake using a linear fitting function indicate a negative relationship (slope of -0.59) though the correlation is poor (R 2 : 0.195, n = 10), indicating other controlling factors are operating as 5 well. New production (NP) estimated from NO − 3 uptake rates and f-ratios offers a means to quantify the C export (Joubert et al., 2011). Although the approach relies on a number of underlying assumptions (steady state conditions; no nitrification in the euphotic layer; validity of Redfield stoichiometry to convert N uptake to C equivalents), new produc-10 tion represents potentially the amount of "exportable production" in the euphotic zone (Sambrotto and Mace, 2000). In Fig. 9, we compare NP deduced from nitrate uptake with SS EP100 (Fig. 9a) and with NSS EP100 (Fig. 9b) based on the 234 Th approach.
It can be observed that POC export fluxes deduced from 234 Th represent between 6 and 56 % of NP for SS EP100 and between 1 and 19 % of NP for NSS EP100. As 15 discussed by Henson et al. (2011) and Joubert et al. (2011) reasons for such discrepancy include a possible overestimation of f-ratio because of nitrification in the euphotic layer, the export of dissolved organic carbon and the fact that uptake of other reduced N species, such amino acids, is usually not considered in the f-ratio approach. Also, differences in time and space scales covered by the NP and the 234 Th approaches can 20 partly explain the observed discrepancies. Bearing in mind that NP represents the potential export of both dissolved and particulate material, lower POC export estimated using 234 Th approach tends to suggest that POC export efficiency is particularly low throughout the BGH transect. This is especially true for the SAZ (6 %), the PFZ (13 to 21 %), the AZ (7 %) and the N-WG (18 %) and to a lesser extent at the PF (29 %) 25 and at the SACCF (56 %). When considering the survey period defined by the NSS model (14 to 22 days before the BGH cruise), efficiency of POC export is even lower and represents only between 1 % (SAZ) to 19 % (SACCF) of "total exportable fraction" based on NP. Although the true export efficiency (Export/production ratio or ThE ratio,  Buesseler et al., 1998) has to be gauged against Net Primary Production (NPP) estimated from C uptake rates and which were not measured during BGH, the comparison between EP100 and NP may indicate that the biological C pump in the SE-Atlantic appears particularly inefficient (relative to nitrate uptake) in exporting C out of the euphotic zone during late summer.

Mesopelagic POC remineralisation
From the vertical distribution of excess 234 Th activities and Ba xs profiles in mesopelagic waters (Fig. 2), it appears that the height of the water column where particle remineralisation/disaggregation water is most intense, strongly varies along the BGH transect.
In the STZ (S1) and in the WG (S5) 234 Th accumulation is relatively shallow and peaks 10 in the subsurface between 120 and 250 m. Evidence of shallow remineralisation has also been reported for the NW Pacific (Maiti et al., 2010), the Sargasso Sea, as associated with mesoscale eddies (Buesseler et al., 2008), and in the SO during the SOFEX experiment (Buesseler et al., 2005). From the SAZ (S2) to the SACCF (S4) the water column layer of excess 234 Th and Ba xs is consistently broader. As shown in Fig. 2,   15 234 Th enrichments extend from below the upper mixed layer to 400 m in the SAZ (S2) and 1000 m in the SAF (L3). Such a thick layer of excess 234 Th is in line with previous studies carried out in different sectors of the SO, including the central Weddell Gyre (Usbeck et al., 2002), the Atlantic sector (Rutgers van der Loeff et al., 1997), and the Australian sector (Savoye et al., 2004a). It has also been reported for Ba xs (Jacquet et 20 al., 2008(Jacquet et 20 al., , 2011. Although there is a general positive correlation between remineralisation fluxes as calculated from Ba xs and 234 Th over the 100-600 m depth range, there also are differences. Indeed, two outliers corresponding to the maximum values of the Baxs-based (L3; SAF) and excess 234  bearing particles is not exactly overlapping with the release of Ba xs rich particles from the aggregates and micro-environments in which they originally formed. This likely reflects differences in carrier particle size and compositions. It should also be kept in mind that both proxies have limitations inherent to the conversion from Ba xs or 234 Th into carbon fluxes. It is quite striking to see that the correlation meso-Ba xs vs. 234 Th fluxes for the 100-600 m depth interval (Fig. 9a) is indeed better (has only one outlier) compared to the regression of the corresponding fluxes. It is likely that a significant part of the discrepancies between the two proxies comes from the assumptions made to calculate C fluxes. For Ba xs this is mainly based on the use of an empiric algorithm, for 234 Th, the main uncertainty probably resides in the choice of the POC/ 234 Th p ratio 15 of remineralised material. Despite such differences, remineralisation fluxes calculated using the Ba xs and excess 234 Th proxy approaches are of similar magnitude (Fig. 11). We also note that remineralisation fluxes of carbon (C respired ) are of the same order of magnitude as the fluxes of POC sinking from the surface, giving further support to the idea that a large 20 fraction of surface export production is strongly attenuated in the mesopelagic zone. However, some differences exist between the two proxies. Close to the SAF (44.9 • S), C respired from Ba xs data is ∼2 times higher than excess 234 Th-based estimates, suggesting that some C remineralisation took place earlier in the season and was not integrated by the excess 234 Th approach. By contrast, in the PFZ (47.5 • S), POC rem- 25 ineralisation deduced from excess 234 Th appears 5 times higher than C respired deduced from Ba xs . It is possible that in this case we see an effect due to physicochemical fragmentation of sinking aggregates as well particle input associated with by zooplankton Overall however, these results confirm that the PFZ (and SAF) is a zone of very efficient mesopelagic remineralisation. Remineralisation rates often exceed export rates 5 from the surface, probably reflecting the fact that the BGH cruise took place in late summer at a time where primary production is decreasing, i.e. when export was already relatively low. It is possible that mesopelagic remineralisation proceeds on particles which were formed earlier in the season and were associated with larger export fluxes. Such a trend toward higher remineralisation rates during the progress of the growth 10 season is a feature that has also been reported in the Southern Ocean (Cardinal et al., 2005;Jacquet et al., 2011).
The fate of the exported POC when transiting through the mesopelagic to bathypelagic zone (>1000 m depth) determines longer term sequestration of POC. In Fig. 12, POC sequestration efficiency to depths >600 m (here defined as the fraction of the ex-15 portable production reaching depths >600 m) is explored for the BGH area by considering the ratio of POC flux at 100 m depth over new production estimated from nitrate uptake (EP100/NP) versus the transfer efficiency through the mesopelagic, defined by the ratio of POC flux at 600 m (i.e. EP600 = EP100 -remineralisation flux) relative to the POC flux at 100 m (EP600/EP100) (Jacquet et al., 2011). Sequestration efficiency 20 appears negligible in the SAZ, the PFZ and the N-WG although the export production from the surface can represent up to ∼20 % of NP in the PFZ, <20 % in the N-WG, <15 % at the SAF and <5 % in the SAZ. The AZ and the STZ both have relatively low fractions of NP that are exported below 100 m depth (7 % for the AZ and 29 % for the STZ) but they significantly differ in the fraction of exported POC from the surface 25 (EP100) that reaches the bathypelagic zone below 600 m depth (20 % for the AZ and ∼50% for the STZ). Sequestration efficiency is higher in the STZ (15 %) than in the AZ where only ∼1 % of NP is exported below 600 m depth. Nevertheless, the highest sequestration efficiency is observed at the SACCF, there more than 25 % of NP is exported to the depth >600 m depth due to lower attenuation of the POC flux in the 100-600 m depth interval.

Conclusions
In this Southern Ocean study the distribution of short-lived 234 Th and biogenic particulate Ba (Ba xs ) are combined to document late summer export of POC from the surface 5 and its fate in the mesopelagic zone. Steady state modelling of 234 Th deficit predicts lowest export production in the STZ and the SAZ where highest levels of biomass are observed. To the south, across the PFZ into the Southern ACC, export production increases progressively, in line with substantial broadening of the surface mixed layer and increasing POC to 234 Th ratios of 10 sinking particles. South of the ACC, the AZ and the northern branch of the Weddell Gyre we observed a slight decrease in POC export, though values still exceed those for the STZ and the SAZ. This could partly result from a greater abundance of large diatoms in sinking material. Non steady state modelling of the 234 Th flux allowed to constrain export production over a period of 2 to 3 weeks prior to BGH expedition. For the area between SAZ and SACCF the non steady state model revealed significantly lower POC export compared to the steady state calculation, suggesting that late summer conditions with low silicate and iron levels, combined with predominance of regenerated production could be factors limiting export. In contrast, further south, in the AZ and the northern Weddell Gyre the two modelling approaches (non steady state and 20 steady state) yield similar values for POC export, indicating export production in this low productivity and high nutrient area remained relatively constant over the season. Although 234 Th-based export fluxes and new production estimates exhibit a similar increasing trend southward, 234 Th based fluxes are consistently lower. Considering that new production represents the "total potentially exportable fraction" of organic C, the discrepancy observed between the two proxies may indicate that surface POC export efficiency is particularly low in late summer. Below the export zone in the mesopelagic layer, excess 234 Th activities as well as accumulation of particulate biogenic Ba, provide strong evidence for significant though variable degrees of POC remineralisation. The attenuation of sinking particles appears particularly intense across the ACC, between the STF and the SACCF. While remineralisation in the SAZ, the AZ and the N-WG essentially occurs between subsurface 5 and 400 m, it extends much deeper for the region bounded by SAF and PF leading to highest attenuations of export being located there. Although some differences exist between the two independent proxies, excess 234 Th and meso-Ba xs yield similar estimates of POC remineralisation. When compared to export production we find that remineralisation of POC in the twilight zone is particularly efficient in the studied area 10 thereby controlling longer term bathypelagic POC sequestration.