Inﬂuence of intense scavenging on Pa-Th fractionation in the wake of Kerguelen Island (Southern Ocean)

. Dissolved and particulate excess 230 Th and 231 Pa concentrations (noted 230 Th xs and 231 Pa xs respectively) and 231 Pa xs / 230 Th xs activity ratios were investigated on and out of the Kerguelen plateau (Southern Ocean) in the framework of the Kerguelen Ocean and Plateau compared Study project in order to better understand the inﬂuence of particle ﬂux and particle chemistry and advection on the scavenging of 231 Pa. In the wake of Kerguelen, particulate 231 Pa xs is relatively


Introduction
The Kerguelen plateau (South Indian Ocean) is an ideal laboratory to study the mechanisms of natural iron fertilization in the Ocean. Better defining these mechanisms was the main aim of the KErguelen Ocean and Plateau compared Study project (KEOPS, Blain et al., 2007). For this purpose, a good understanding of the particle dynamics and advection processes in this area was required. 231 Pa and 230 Th are natural radionuclides, uniformly produced in seawater by the decay of the homogeneously distributed uranium isotopes ( 235 U and 234 U respectively). Consequently, they are both produced at a fixed known rate in the ocean with a production activity ratio 231 Pa/ 230 Th of 0.093. Both radionuclides are particle reactive and therefore rapidly adsorbed onto settling particles and removed (scavenged) from the water column to the sediment. However, their affinity for particles differs, 230 Th adsorption coefficients being generally 10 times higher than 231 Pa coefficients (Anderson et al., 1983a;Moran et al., 2002;Chase et al., 2003). Consequently, 231 Pa has a residence time in the water column of 50-200 yr, longer than the 10-40 yr residence time of 230 Th (Anderson et al, 1983a;Nozaki et al., 1985; Published by Copernicus Publications on behalf of the European Geosciences Union. 3188 C. Venchiarutti et al.: Influence of intense scavenging on Pa-Th fractionation Walter et al., 1997), so that these two radionuclides are fractionated in seawater compared to the production ratio. 231 Pa is transported to areas of high particle fluxes (e.g. margins) prior to being scavenged. In contrast, 230 Th is mainly scavenged on its production site (Walter et al., 1997;Yu et al., 2001). This enhanced scavenging at ocean margins is named "boundary scavenging" (Spencer et al., 1981;Anderson et al., 1990). Boundary scavenging is the result of the combination of (i) an increasing particle flux (particle flux effect) from the open ocean to the margin and possibly from a change in particulate matter composition (particle composition effect) and (ii) a sufficiently long scavenging residence time of the element (e.g. 210 Pb,231 Pa. . . ) in the open ocean so that it can be transported from the open ocean where it is produced, to the ocean margins where it is efficiently scavenged.
As for 230 Th, the oceanic 231 Pa chemical behaviour is considered to be governed by reversible scavenging on settling particles (Bacon and Anderson, 1982). Therefore, in absence of lateral transport of these radionuclides by currents, reversible scavenging yields linear increases with depth of total (dissolved + particulate) 231 Pa and 230 Th concentrations below the euphotic layer and at least down to 1000 m depth (Nozaki et al., 1985;Roy-Barman et al., 1996;Choi et al., 2001;Marchal et al., 2007). Deeper in the water column, deviations from this linearity are often observed. This depletion can be due to either (i) depleted surface waters that are sinking and mixing with deep waters (Rutgers van der Loeff and Berger, 1993;Vogler et al., 1998;Moran et al., 2002;Coppola et al., 2006) or (ii) advection of waters depleted in radionuclides coming from the ocean margins, leading to low radionuclide concentrations in the water column (Venchiarutti et al., 2008;Roy-Barman, 2009).
Both radionuclides are therefore not only used as tracers for particle scavenging (Nozaki et al., 1981;Bacon and Anderson, 1982;Luo and Ku, 2004) but also for the deep ocean circulation or ventilation (Rutgers van der Loeff and Berger, 1993;Scholten et al., 1995;Moran et al., 1997). However, data on the distributions of 231 Pa and 230 Th in seawater and particles are still scarce, precluding a full understanding on these radionuclides oceanic behaviour.
For instance, the role of particle composition on the radionuclides scavenging remains unclear. Indeed, 231 Pa appears to be more reactive with inorganic particles (Fe 2 O 3 , MnO 2 ) or biogenic silica (opal) in carbonate-depleted areas, like in the Southern Ocean (Guo et al., 2002;Chase et al., 2002 andScholten et al., 2005). Consequently, in presence of particles rich in Fe 2 O 3 , MnO 2 or biogenic silica, both 230 Th and 231 Pa would be adsorbed onto the particles with almost the same efficiency, thereby yielding almost no fractionation between 231 Pa and 230 Th (Walter et al., 1999;Chase et al., 2002). However, questions still remain concerning the exact affinity of these radionuclides according to particle type, grain size and whether they are adsorbed or rather incorporated in the particles.
Moreover, recent debates question the relevance of 231 Pa/ 230 Th ratio as a circulation tracer. Indeed, it appears that the 231 Pa/ 230 Th could reflect the influence of the surrounding biological productivity (in particular diatoms) rather than the circulation imprint (Bradtmiller et al., 2006(Bradtmiller et al., , 2009Hayes et al., 2011), and questions still remain concerning the effect of ocean margins on the 231 Pa/ 230 Th.
Here, we present the 231 Pa distributions in both solid and dissolved phases as well as the 231 Pa/ 230 Th fractionation in the Kerguelen area during the KEOPS program. A salient feature of this area is the biogeochemical contrast between the Kerguelen plateau and the open-ocean (phytoplankton bloom versus High Nutrient Low Chlorophyll -HNLC-area). This study complements the work on 230 Th during KEOPS (Venchiarutti et al., 2008) which pointed out that: -Using the "classical" 1-D scavenging model does not allow obtaining realistic 230 Th scavenging rates over the Kerguelen plateau. Realistic scavenging rates were obtained by using a new advection-scavenging model that takes into account both 230 Th scavenging and advection of HNLC open-ocean waters over the productive plateau.
-Using this model, Venchiarutti et al. (2008) demonstrated that scavenging is strongly enhanced on the plateau and along its eastern flank (mean settling speed of small particles one order of magnitude larger than in the open ocean) and that, despite this strong increase of scavenging compared to the open ocean, the 230 Th concentration does not drop down to zero because the continuous inflow of open ocean water brings new 230 Th on to the plateau. It suggests that advection of waters from the open ocean could bring other trace elements as well and impact on the budget of these elements over the plateau.
-Enhanced scavenging of 230 Th on the plateau and along its eastern flank was tentatively attributed to the occurrence of particle re-suspension by nepheloid layers. Venchiarutti et al. (2008) also showed that the Kerguelen plateau represents an ideal "simple" case to study boundary scavenging because there is a high gradient of particle flux over a limited distance linked to a well constrained circulation over the plateau. As mentioned earlier, 231 Pa is more prone to boundary scavenging than 230 Th. Consequently, a significant boundary scavenging of 231 Pa is expected over the Kerguelen plateau. Analysing 231 Pa in this area should also bring constraints on Pa/Th fractionation during scavenging processes. In particular, precise budgeting of the advection and scavenging flux using the Kerguelen boundary scavenging model should allow elucidation of the respective influence of the particle flux versus chemistry effects, a question still in debate. Thus, the first aim of the present study is to determine if the intense scavenging highlighted by 230 Th on the Kerguelen plateau and along the eastern slope (Venchiarutti et al., 2008) also affects 231 Pa distribution. If the model proposed for 230 Th does not account correctly for 231 Pa, it would indicate the occurrence of processes allowing the decoupling of the two elements. The other aim is to determine the role of particle composition vs. particle flux on the sedimentary Pa/Th ratios. Hence, we propose to study the fractionation of these two radionuclides over and out of the Kerguelen plateau, based on 231 Pa partition coefficients (K Pa ) and the 231 Pa/ 230 Th fractionation factor (F Th/Pa ).

Regional settings: hydrography
The Southern Ocean is the largest HNLC area in the world Ocean (Sarmiento et al., 1998;Tréguer and Pondaven, 2002;Marinov et al., 2006). However, most of the islands of this ocean experience intense summer phytoplankton blooms (Pollard, 2009;Frew et al., 2006). In the case of the Kerguelen Island and plateau zone, a large bloom extends yearly South of the Polar Front (PF, Fig. 1) over the Kerguelen plateau, to which it is restrained by the bathymetry and surrounding HNLC waters (Blain et al., 2007;Mongin et al., 2008).
The Kerguelen plateau divides the Antarctic Circumpolar Current (ACC) into a northern flow, north of the Kerguelen Island and a southern flow, through the Fawn Trough, delimiting the so called "central Kerguelen plateau" (Park et al., 2008;Mongin et al., 2008). This latter extends therefore from the Kerguelen Island (north of the PF) to the Heard/McDonald Islands, its southern limit (Fig. 1). It is shallower than ∼560 m, with some shallow seamounts and is delimited by deeper troughs and ridges (Park et al., 2008;Mongin et al., 2008): at north, the PF Trough (∼650 m) and south, the Fawn Trough (∼2650 m) and along the southeastern part of the plateau, the Northwest-Southeast Trough (∼600 m).
LADCP measurements (Park et al., 2008) highlighted the contrast between the weak geostrophic circulation of the shallow plateau, with a general northward flow ≤5 cm s −1 and the circulation along the eastern flank of the plateau, where a strong north-westward branch (∼18 cm s −1 ) of the Fawn Trough Current (FTC) brings cold Antarctic waters, likely originating from south of the Elan Bank and the eastern Enderby Basin (Fig. 1). These waters include Winter Water (WW), Upper and Lower Circumpolar Deep Waters (UCDW and LCDW respectively). When exiting the FTC, these waters are encountering cold Antarctic waters within the northernmost extent of the Deep Western Boundary Current (DWBC) flowing along the eastern escarpment plateau from the Australian-Antarctic Basin (McCartney and Donohue, 2007;Roquet et al., 2009).
During the KEOPS cruise in the Austral summer 2005, three transects were covered on and out of the central Kerguelen plateau, stations numbered from 1 to 7 and 8 to 11 respectively ( Fig. 1). Station A3 was located in the core of the large bloom on the central plateau and station C1 was at the southern extent of this central plateau, whereas the offplateau stations Kerfix, A11, B11 and C11 were considered as open-ocean stations in HNLC waters (Fig. 1). The bloom was mostly dominated by large diatoms in surface waters, and sediments on the plateau were dominated by siliceous ooze . The export of particulate carbon on the Kerguelen plateau was found to be almost double that in the surrounding HNLC waters . Bacterial activity in the upper 125 m depth of the plateau was higher than in the HNLC area whereas the inverse was observed for the mesopelagic layer Jacquet et al., 2008). This difference was attributed to the occurrence of larger diatoms cell sizes on the plateau than in the HNLC waters.
Finally, Zhang et al. (2008), van Beek et al. (2008) and Chever et al. (2010 demonstrated that waters originating from further south of the Kerguelen plateau and spreading northward over the plateau, were likely enriched in trace elements due to shelf weathering in the vicinity of Heard Island. These advected waters may be an important source of trace elements to the Kerguelen plateau and contribute significantly to the natural fertilisation.

Sampling
The sampling and preparation procedures -common for both radionuclides -are fully detailed in Venchiarutti et al. (2008). Below, we will summarize only briefly some specific treatments related to the Pa analysis.
Seawater samples were collected during KEOPS cruise on board the R/V Marion Dufresne (19/01/05-13/02/05, Fig. 1) using General Oceanics 12 L-Niskin bottles and a Seabird SBE19 + CTD. The analysis of the dissolved 231 Pa was performed at 4 out of the 9 stations sampled during KEOPS: on (A3 and C1) and off (B11 and C11) the Kerguelen plateau. The core of the bloom (station A3) was sampled twice, 12 days apart, for the radionuclide analysis (referred as A3-33 for the first visit and A3-77 for the second).
Briefly, the 30 l-seawater samples were first filtered through 47 mm diameter Millipore filters (0.65 µm Durapore or 0.8 µm Versapore), spiked with the corresponding yield tracers, 50 pg of 229 Th and ca. 140 fg of 233 Pa (produced by neutron activation of 232 Th). After 12h equilibration, both radionuclides were co-precipitated, with KMnO 4 and MnCl 2 solutions, by addition of NH 4 OH, based on the protocol of Rutgers van der Loeff and Moore (1999). After 24 h homogenisation, the samples were filtered through 142 mm diameter Millipore membranes and the filters stored in Petridishes.
Particulate 231 Pa was collected using in situ pumps (McLane and Challenger Oceanic systems) and analyzed at 4 stations: A3-77, A11, C11 and Kerfix. At each station, 83-2564 l were filtered in situ through 0.65 µm Durapore or 0.8 µm Versapore filters (diameters of 142 or 293 mm) and the filters were folded and stored in Petri-dishes to be analysed once back at the home laboratory.

Analytical procedures
All the analytical procedures were conducted using acidcleaned containers and materials, double-distilled reagents, and the analyses were performed in clean-rooms (Class Iso 5-6) at the home laboratory LEGOS.

Dissolved samples
Back at the lab, the Mn-filter coprecipitates were processed for the analysis of dissolved 231 Pa. However, due to technical problems, the co-precipitation yield could not be determined except for the samples collected at A3-33, for which non-leached filters were counted for gamma spectrometry at the LSM (Laboratoire Souterrain de Modane, French Alps). The co-precipitation recovery was 103 ± 11 % for nine A3-33 samples (Table 1). This good recovery was encouraging and we assumed that the co-precipitation step was as efficient as at A3-33 for all the KEOPS stations.
The Mn-filter precipitates were leached in 9 M HCL baths. However, because of a 4-month mass spectrometer Table 1. Chemical and co-precipitation yields for the KEOPS 231 Pa samples.

Sample types
Chemical yields a Co-precipitation yields b Dissolved 231 Pa 14-97 % (57 ± 4 %) 103 ± 11 % (N = 23) (N = 9) Particulate 231 Pa 5-51 % (24 ± 5 %) X (N = 13) a Chemical yield ranges (left column) obtained by Isotope Dilution take into account all the chemical steps carried out after spiking the samples with the corresponding yield tracers (i.e after the co-precipitation step for dissolved samples and after leaching for the particles). Average yield values are reported into brackets (see details in Sect. 2.4.3). b Pa co-precipitation yields (right column) were estimated with gamma-spectrometry at station A3-33 (see details in Section 2.3.1). N represents the number of samples analysed and therefore used for the average values.
breakdown, the analysis of the dissolved 231 Pa concentrations at four stations (A3-77, B11, C11 and C1) was postponed for almost one year and consequently no 233 Pa spike remained at all in these samples. Therefore, the Mn-filter leachates at these four stations were split into two aliquots (each equivalent to 15 l seawater). One aliquot was dedicated in the first instance to the Th analysis (Th results are reported in Venchiarutti et al., 2008). The other aliquot was stored in 9 M HCl and HF to enable later determination of the 231 Pa concentrations by isotope dilution, after re-spiking with a new 233 Pa spike (prepared by milking a solution of 237 Np at the Alfred Wegener Institute, AWI).
After evaporation of the leaching solution to almost dryness, the samples were spiked with 236 U (∼40 pg) to trace 233 U bleeding in the Pa fraction (Choi et al., 2001). The Pa samples were then purified (from Th and U) using an anion exchange resin (AG1x8, 100-200 mesh). The elution steps are further documented in Venchiarutti et al. (2008) and Jeandel et al. (2011a).

Particle samples
The 231 Pa and Th particulate samples were leached in a mixture of concentrated HCl and HNO 3 followed by the addition of Suprapur HF, based on the protocols given in Tachikawa et al. (1997) and Venchiarutti et al. (2008).
As for the dissolved samples, machine breaks led us to split the leachates in order to allow the particle analysis of Th isotopes, performed on ∼20 % aliquot of the leachate (Venchiarutti et al., 2008). Concentrations of multiple elements were later determined with ICP-MS (Agilent) on 2 % of the remaining ∼80 % of the initial leachates. Ultimately, after spiking with the corresponding yield tracers ∼100 fg 233 Pa (AWI batch), ∼20 pg 229 Th and 15 pg of 236 U, the particulate Th and Pa of the remaining leachates were then separated and purified through the same anion exchange resin described in Venchiarutti et al., 2008. Both sets of particulate Th data, resulting from the analysis of the ∼20 % and ∼80 % leachate samples gave consistent results (Venchiarutti et al., 2008).

Spectrometric analysis
After the chromatographic separation, both particulate and dissolved 231 Pa purified fractions were measured within 24 h (so that 233 Pa decay to 233 U is minimal, see Sect. 2.4.2) in a 2 % HNO 3 solution on a Finnigan Neptune MC-ICP-MS (Observatoire Midi-Pyrénées, Toulouse) and their concentrations calculated by isotope dilution from the measured 233 Pa/ 231 Pa ratio.
Th isotope measurement is further detailed in Roy-Barman et al. (2005) and Venchiarutti et al. (2008). Pa analysis was performed according to the protocols derived from Choi et al. (2001) and Regelous et al. (2004). Pa samples were introduced into the plasma through a Cetac Aridus system, equipped with a PFA microflow nebulizer (Elemental Scientific, 50 µl min −1 ) and a membrane desolvation unit, resulting in a passive sample uptake (without peristaltic pump) of ∼60 µl min −1 . This system was optimized using N 2 (flow rate of 5 ml min −1 ) and Ar (sweep gas, flow rate of ∼10 ml min −1 ) gases.
Before each sequence for Pa measurements, the instrument was tuned and peak shapes were improved, for both Faraday cups and SEM, using a standard solution of natural U (SRM 4321 C), with a certified 238 U/ 235 U ratio of 139.67 ± 0.016. This U solution was also run every eight samples in order to bracket the samples and check the instrument short and longterm variability and mass bias (linear correction) during the analysis.
To prevent from any cross-contamination (carry-over) between samples, the system was thoroughly cleaned with a 2 % HNO 3 v/v solution after each sample run and acid blanks were run prior to each measurement.

Corrections for 232 Th tailing and isobaric interferences
The importance of 232 Th tailing correction for masses 230 and 231 have been previously underlined by Choi et al. (2001); Pichat et al. (2004) and Thomas et al. (2006). However, tailing effect might be attenuated for seawater samples because the 232 Th concentration is generally low. Nevertheless, the abundance sensitivity or tailing correction was checked for each of our samples and the counts on mass 231 were corrected using a linear correction based on the systematic measurements of the masses 230.5 and 231.5 for 231 Pa (Choi et al., 2001). The abundance sensitivity of the MC-ICP-MS at the time of the Pa analyses was ∼2 ppm for masses 1 amu apart, i.e. on 231 Pa peak, with the Retarding Potential Quadrupole lens-RPQ settings at this time.
For KEOPS samples, most of the 232 Th concentrations in seawater (Venchiarutti et al., 2008) were typically lower than 100 pg kg −1 (except for a few samples, notably at C1), so that the highest measurable 232 Th and 233 Pa quantities (for a 15 l sample) were typically lower than 1.6 ng (0.8 ng g −1 in 2 ml) and 140 fg respectively. Moreover, the efficient column separation of Th and Pa allows us to limit the percentage of 232 Th "bleeding" (<2.5 %) in the Pa fraction, thereby inducing a 233 Pa/ 232 Th ratio <3.73 × 10 −3 in the samples. Consequently, taking the highest 232 Th 1 H/ 232 Th ratio = 1 × 10 −5 yields to a hydride correction typically <0.3 %.

Corrections for uranium bleeding and in-growth
Adding 236 U to the seawater and particulate samples prior to the anion exchange column allowed us to monitor the efficiency of the column separation of U and Pa, and thus to ensure that the Pa fraction is free of any 233 U. Indeed, the presence of any 233 U in the Pa fraction or so called " 233 U bleeding" would interfere the counting on mass 233 (Choi et al., 2001;Pichat et al., 2004).
In spite of an efficient chromatographic separation that leaves only an insignificant residual U amount in the Pa fraction, the contribution of 233 U bleeding on the 233 peak in the Pa fraction was monitored for each sample and estimated following Eq. (1): (1) Where n is the number of counts on the masses 233 and 236, and the sub-scripts sple and spike refer to sample (the Pa fraction) and 236 U spike respectively. The estimate of bleeding is based here on the counts of uranium found in the Pa fraction and not on the measurements of the U ratios in the U fractions, preventing us to compare to the literature values (Choi et al., 2001;Pichat et al., 2004;Thomas et al., 2006). Most of the Pa analyses were not affected by U bleeding. Indeed, the 233 U bleeding exceeded 0.5 % of the total U in the Pa fraction for 10 samples only (reaching 11 % for A3-77, 400 m depth) but less than 0.01 % in the remaining samples. In case of significant bleeding (above 0.5 %), the aforementioned correction (Eq. 1) was applied.
All the dissolved 231 Pa samples were corrected for the ingrowth from the uranium, which was co-precipitated with MnO 2 , assuming a maximum of 5 % of co-precipitated U in the samples (Roy-Barman, personal communication). The mean U-decay corrections vary from 0.22 % (B11 measured 767 days after sampling) to 0.95 % (A3-77, measured 827 days after sampling).

Uncertainties and blanks
Uncertainties for 231 Pa concentrations, estimated by isotope dilution, were propagated at the 2σ level and include the instrument statistical error, carry-over correction, mass bias, spike contributions and overall procedural blanks. Corrections for the 232 Th tailing and 233 U bleeding were also taken into account.
For the dissolved samples, the procedural blanks (Table 2) were estimated by analysing (as for a sample, although not spiked) the co-precipitated filters (of two types: Durapore and Versapore) achieved on-board, using 10 l ultra-pure water and the same reactants and reagents as for the samples. Distinguishing contamination due to the use of one type of filters with respect to the other was not statistically significant. The blanks for the particles (Table 2) were achieved using Versapore filters. They were generally close to the background noise. The detection limit was estimated at 1.2 fg of 231 Pa and corresponds to three times the standard deviation based on two blanks for particles (N = 2).
Although we could not establish a significant statistic on our blank values, they are in the range of those gathered from the literature (Choi et al., 2001;Moran et al., 2002;Edmonds et al., 2004) whatever the extraction method (co-precipitation with Fe or Mn-Ox, filtration or siphoning-off).
The overall yields obtained for 231 Pa are reported in Table 1 and for Th isotopes in Venchiarutti et al. (2008) and are in the range previously reported in the literature (Choi et al., 2001;Thomas et al., 2006).

Results
In seawater, most of the 230 Th and 231 Pa are produced by the radioactive decay of the soluble and homogeneously distributed 234 U and 235 U respectively, providing the authigenic fraction of these radionuclides in any given sample. However, an accurate determination of this fraction requires the correction of the measured value from the detrital contribution which is supported by the U already contained in the particles. This latter is the lithogenic fraction of these radionuclides in the same sample. The lithogenic contribution can be critical in areas receiving strong lithogenic inputs.
Here, we estimated the 230 Th and 231 Pa scavenged from seawater and termed as "unsupported" (or in excess) 230  The mean natural activity ratio 235 U/ 238 U = 0.04605 (Scholten et al., 1995Moran et al., 2005) is applied in Eq.
(3), 238 U/ 232 Th crustal activity ratio has been measured in different locations and varies from 0.8 ± 0.2 (Anderson et al., 1990) in the Pacific, 0.7 ± 0.2 in the Atlantic (Scholten et al., 2008) and 0.4 ± 0.1 south of the Antarctic Polar Front (Walter et al., 1997). We used here a value of 238 U/ 232 Th = 0.8 ± 0.2 (Anderson et al., 1990) to estimate 231 Pa xs in the wake of Kerguelen. Dissolved and particulate 231 Pa xs concentrations on and off the Kerguelen plateau are reported in Tables 3 and 4, and in Figs. 2, 5 and 6 and expressed in dpm m −3 . All the 231 Pa xs / 230 Th xs ratios reported in Tables 3 and 4 and represented in Figs. 3 and 4 are activity ratios.
Most of the dissolved 231 Pa xs concentrations and 231 Pa xs / 230 Th xs ratios are in the range previously observed in the South-West Indian Basin (Thomas et al., 2006), in the Equatorial Atlantic (Choi et al., 2001;Moran et al., 2002) or in the South Atlantic (Rutgers van der Loeff and Berger, 1993;Walter et al., 2001).
Radionuclide concentrations in particles (Tables 3 and 4) are consistent with values found in the Labrador Sea (Moran et al., 2002) but somewhat lower than values obtained in the South Atlantic, south of the Polar Front by Rutgers van der Loeff and Berger (1993). Particulate 231 Pa xs / 230 Th xs ratios (Fig. 4) are in the range of the particulate ratios observed in the Southern Ocean (Walter et al., 2001).
On average, lithogenic contributions represent 2 % and 1 % of the dissolved and particulate 231 Pa xs concentrations respectively. The highest lithogenic contributions were observed at C1 (50 m depth, 10 % contribution) for dissolved 231 Pa xs and at A3-77 (350 m depth, 4 % contribution) for 231 Pa xs in particles.

On-plateau stations (0-560 m depth)
Dissolved 231 Pa xs was analyzed at 2 of the 6 onplateau stations (A3-77 and C1) that were studied for 230 Th xs (Tables 3-4 respectively, Fig. 2a, and for 230 Th xs , cf. Venchiarutti et al., 2008). At these two stations, dissolved 231 Pa xs distribution is constant within the uncertainties all along the water column (Fig. 2a). However, a surprising maximum of dissolved 231 Pa xs (0.079 dpm m −3 ) is observed at 150 m depth at the plateau station A3-77 (Fig. 2a). Most of the dissolved 231 Pa xs / 230 Th xs ratios are above their production ratio of 0.093 (Fig. 3a) throughout the water column at stations A3-77 and C1. The average 231 Pa xs concentration in particles (Table 3), measured only at station A3-77, is 0.017 dpm m −3 representing about 43 % of the total 231 Pa xs concentration at this station. Particulate 231 Pa xs / 230 Th xs ≥0.093 (Fig. 4) is observed at this latter station, with a decrease of the ratio with increasing depth.

Off-plateau stations (0-3275 m depth)
Dissolved 231 Pa xs was measured at 2 out of the 4 off-plateau stations (C11 and B11 stations, Fig. 2b) that were analysed for 230 Th xs . These stations are considered as representative of "open ocean" or HNLC stations.   Dissolved 231 Pa xs / 230 Th xs ratios are relatively constant with depth and above the production activity ratio of 0.093 (Fig. 3b). They display a few maxima in the surface waters, reaching 3.03 at station B11 (Table 4). Particulate 231 Pa xs / 230 Th xs ratios (Fig. 4) show a clear dependence with depth, displaying some maxima in surface waters (with a   mean value of ∼2.76 between 80 m and 200 m at station C11, Table 3) and then decreasing values with increasing depth.

Discussion
Here, we aimed at determining the mechanisms involved in Pa and Th scavenging, fractionation and their resulting distributions in seawater and particles in the wake of Kerguelen.

Fractionation of 231 Pa and 230 Th in the wake of Kerguelen
Dissolved 231 Pa xs / 230 Th xs activity ratio does not change significantly with depth on and off the Kerguelen plateau (Fig. 3). Contrastingly, particulate 231 Pa xs / 230 Th xs activity ratio decreases by a factor of ∼2 between shallow and deep waters whatever the station (Fig. 4). These variations could reflect changes in the rates of exchange processes (adsorption/desorption, aggregation/disaggregation) between particles and seawater and/or in the particles chemistry (composition or particle alteration). The fractionation factor (noted F Th/Pa ) is a good tool to investigate the influence of particle composition on the scavenging and fractionation of 231 Pa and 230 Th, assuming chemical equilibrium between dissolved and particulate phases. It provides information on the element reactivity according to the particle composition (Walter et al., 1999;Guo et al., 2002;Luo and Ku, 2004;Chase et al., 2002 and and is defined as follows: Where K Th and K Pa are the partition coefficients for 230 Th and 231 Pa respectively, defined as the activity ratio between particle and seawater distributions. Note that both activities are here expressed as dpm of Th and Pa per m 3 of seawater (Scholten et al., 1995;Venchiarutti et al., 2008;Roy-Barman, 2009) and not related to the mass of particles per litre of seawater (Chase et al., 2002), since no "extensive" study of the particulate composition or particle size/surface area was carried out during KEOPS cruise. Consequently, the partition coefficients determined here are dependent on the particles concentration and will change if particle mass changes. In most of the open ocean, Th is preferentially scavenged over Pa yielding a fractionation factor F Th/Pa of roughly 10 (Anderson et al., 1983a). In the Kerguelen area (Table 3), F Th/Pa values are very low, ranging from 0.06 ± 0.01 (C11, 200 m) to 1.6 ± 0.2 (C11, 700 m). F Th/Pa close to 1 is typically observed when opal abundance is high (Walter et al., 1997;Chase et al., 2002;Scholten et al., 2005Scholten et al., , 2008 reflecting the high affinity of Pa for opal. The low F Th/Pa values reflect the high affinity of 231 Pa for particulate matter. Indeed, in the Kerguelen area, K Pa values range from 0.05 ± 0.01 (C11, 700 m) to 3.1 ± 2.2 (C11, 80 m), with an average value of 0.86 (Table 3). This is almost two orders of magnitude larger than the K Pa values observed in areas where biogenic silica (BSi) is not prevalent in the particulate matter, like in the Equatorial Pacific and South-East Atlantic (∼0.01-0.04; Anderson et al., 1983a;Scholten et al., 2008). This high affinity of 231 Pa for particulate matter in the Kerguelen area is consistent with the high BSi values recorded during KEOPS cruise, with about 5 µmol L −1 over the Kerguelen plateau and ∼2 µmol L −1 at the off-plateau stations Fripiat et al., 2011), compared with the low BSi concentrations <0.5 µmol L −1 found in the Equatorial Pacific (Leynaert et al., 2001) and in the South-East Atlantic (Bishop et al., 1978).
Moreover, the observed F Th/Pa < 1 of the Kerguelen area suggests that high opal abundance favours adsorption of 231 Pa onto particles with more efficiency than 230 Th, thereby producing high particulate 231 Pa xs / 230 Th xs ratios in the euphotic layer. It confirms the enhanced affinity of 231 Pa for opal when opal represents more than 60 % of the particulate matter .
Using parameter such as F Th/Pa , K Th or K Pa to establish a link between the particulate matter composition and its effect on Pa and Th fractionation, we implicitly assume that Pa and Th have reached a chemical equilibrium between the dissolved and the particulate phases. However, such equilibrium may not be reached at all the water column depths, especially in the surface waters. Hence, it may not be possible to integrate this fractionation over the whole water column as suggested by Thomas et al. (2006) and Scholten et al. (2008). Therefore, we must remain cautious in our interpretation.
Nevertheless, the low F Th/Pa values and relationship between K Pa and BSi confirm that 231 Pa and 230 Th fractionation in the Kerguelen area appears driven by the biogenic opal content of the particles. This is consistent with the high abundance of large diatoms in the euphotic layer of the Kerguelen area Carlotti et al., 2008).

Evidence of boundary scavenging along the eastern slope of the Kerguelen plateau
Along the eastern escarpment of the Kerguelen plateau, cold Antarctic subsurface and deep waters are entrained by the northward branch of the Fawn Trough Current (Fig. 1) from the eastern station C11 toward B11 (Park et al., 2008 andMongin et al., 2008). In the deep waters (σ 0 = 27.63-27.84 kg m −3 , i.e. from 700 m depth to the bottom), 231 Pa xs concentration profiles differ (Figs. 2b and 5), while the deep water masses of these stations have similar θ-S characteristics (Park et al., 2008;Venchiarutti et al., 2008). The 231 Pa xs depletion is more pronounced at C11 than at B11. These features are qualitatively similar to those observed for the 230 Th xs profiles at the same stations (Venchiarutti et al., 2008). As for 230 Th, we attribute the 231 Pa xs depletion to an intense boundary scavenging in the water flowing along the eastern escarpment of the plateau, possibly due to particle re-suspension and/or nepheloid layers in the deep and bottom waters (Venchiarutti et al., 2008). Indeed, it has been shown (notably for 210 Pb) that particles supplied from the seafloor with nepheloid layers can have a strong impact on the radionuclide scavenging in the deep ocean (Nozaki et al., 1997;Okubo et al., 2007;Turnewitsch et al., 2008). However, the exact processes yielding this bottom scavenging remain to be determined. Indeed, if reversible equilibrium only is determining the dissolved and particulate Pa distribution when particles fall through the water column, re-suspended particles should be at equilibrium with bottom waters and hence particle re-suspension should not produce further Pa scavenging.
Although C11 seems located upstream of B11 (Fig. 1), some of the C11 deep waters have already interacted with the waters flowing at the contact of the slope, and likely with the sediments deposited on it (Park et al., 2008). This interaction of the deep waters with the eastern slope of the Kerguelen plateau is also imprinted in their Nd isotopic signature (Jeandel et al., 2011b). This is not the case for B11 deep waters, located in the core of the Fawn Trough and east off the Kerguelen plateau (Park et al., 2008). This explains the lower radionuclide concentrations at C11 compared to B11 (Figs. 2b and 5).
Comparing the mean dissolved 231 Pa xs and 230 Th xs concentrations between 700 m and 2800 m depth at B11 and C11, taking B11 as the reference station from the open-ocean (not affected by scavenging along the Kerguelen plateau slope), we estimated that this margin effect leads to a scavenging of 231 Pa xs = 37 ± 4 % and 230 Th xs = 10 ± 0.3 % between these two stations. Assuming that these depletions are only due to scavenging on particles, the average F Th/Pa along the escarpment can be estimated as ( 230 Th xs / 231 Pa xs )/((1-230 Th xs )/(1-231 Pa xs )) = 0.20. Taking the uncertainties on 230 Th xs and 231 Pa xs into account, we obtain a total range of F Th/Pa = 0.16-0.24. This range is consistent with the F Th/Pa < 2 estimated at C11 with only one value above 1, at 700 m depth at C11.
These scavenging estimates and low fractionation factor for 231 Pa xs and 230 Th xs confirm that in an environment dominated by BSi, 231 Pa removal is at least as efficient and possibly more efficient than 230 Th removal as we inferred in the previous section from the K Pa and F Th/Pa data.
Finally, in the case of the Kerguelen area, the high abundance of opal appears to enhance the "boundary scavenging" effect already generated by the difference in the radionuclide residence times (Anderson et al., 1983b).

Scavenging of 231 Pa on the Kerguelen plateau
We now apply the Kerguelen plateau boundary scavenging model developed for 230 Th (Venchiarutti et al., 2008) to 231 Pa, in order to determine the 231 Pa scavenging rate over the plateau, the influence of open ocean water advection and thereby the resulting Pa flux down to the sediment.
We assume that between 0 and 500 m, the 231 Pa concentration in the open ocean essentially increases linearly with depth and take the off-plateau station B11 (Fig. 2b) as reference for the open ocean with (d 231 Pa xs−t /dz) = 3.4 × 10 −4 dpm m −4 . In order to estimate d 231 Pa xs−t concentration at B11, particulate data missing at this station, we consider a value of 0.2 for K Pa as representative of the affinity of 231 Pa for particulate matter in the upper 500 m depth of the water column, in an area of the open ocean (i.e. here, out of the plateau) dominated by opal (Rutgers van der Loeff and Berger, 1993).
The vertical dissolved Pa profile over the plateau, derived from Eq. (13) in Venchiarutti et al. (2008), is then given by: Where A is a constant defined as A = (τ (1 + K Pa )/SK Pa ), τ is the transit time of the water at station A3 defined as τ = L/u = 0.33 y, taking an horizontal speed u of 5 cm s −1 (mean current velocity measured with moorings, for the stations of the transect A, cf. Park et al., 2008) and a distance over the whole plateau L = 520 km (Venchiarutti et al., 2008). Thus, we estimate the water residence time on the plateau of about 4 months, which may be considered as the lowest value for the water residence time, since this estimate is based on the highest mean current velocity on the whole plateau. We assume that Th and Pa are transported on the plateau at the same particle settling velocity and hence onto the same class of settling particles (S = 3000 m yr −1 , Venchiarutti et al., 2008). Over the plateau, we take the mean value K Pa = 0.56 ± 0.21 at station A3-77. The production rate of 231 Pa is P = 2.44 × 10 −3 dpm m −3 yr −1 .
Despite the few available data, their general agreement with the modelled profile at station A3 confirms the importance of advection on the distribution of 231 Pa over the plateau (Fig. 6). The model allows us to estimate that the vertical particulate 231 Pa flux settling at 500 m over the plateau is S×( 231 Pa xs−p ) 500m = 84.9 dpm m −2 yr −1 and represents almost 70 times the local in situ production integrated over a 500 m depth water column. This high 231 Pa flux is possible because the 231 Pa scavenging residence time (h/(2 SK Pa ) ∼ 0.15 yr) in the water column is shorter than the residence time of the water on the Kerguelen plateau (∼ 0.33 yr), thereby allowing an efficient scavenging and because the advected open ocean water has a high 231 Pa content (as in the Surface Mixed Layer and Winter Waters from the off-plateau station B11, Fig. 5).
Hence, the drawdown of the dissolved 231 Pa concentration over the plateau remains limited because advection of open ocean water brings continuously new 231 Pa over the plateau.  If similar scavenging conditions occurred without any advection of open-ocean water, the dissolved 231 Pa concentration should be much lower than observed (Fig. 6, curve "plateau profile without advection", F = 0). The dissolved 231 Pa concentrations at C1, nearby Heard Island, is consistent with the open-ocean profile (Fig. 6, solid bold line), suggesting that waters originating from this area may also be a source of 231 Pa-rich waters for the central plateau (A3 station), in agreement with Zhang et al. (2008) and Chever et al. (2010).
In the Panama and Guatemala Basins, the lack of Pa-Th fractionation associated with boundary scavenging of 230 Th and 231 Pa was explained by a moderately intensified scavenging compared to the open ocean with F Th/Pa << 10 rather than by a quantitative stripping of 231 Pa and 230 Th with F Th/Pa = 10 (Anderson et al., 1983b). In the Kerguelen plateau case, there is both F Th/Pa << 10 (Table 3) and a strongly intensified scavenging due to the high biological productivity, but not completely reflected in the dissolved 230 Th and 231 Pa concentrations due to the effect of advection.
In summary, the 231 Pa data confirm that there is an enhanced scavenging on the plateau, as already observed for 230 Th (Venchiarutti et al., 2008). Moreover, 231 Pa concentrations showed that this intensified scavenging on the Kerguelen plateau is due to both particle flux and particle composition effects, leading to a low Pa/Th fractionation. However, the "boundary scavenging" model proposed for the Kerguelen plateau in Venchiarutti et al. (2008), is a particular case of boundary scavenging model where the open ocean is infinitely large compared to the ocean margin (Roy- Barman, 2009). Compared to the boundary scavenging created by a large continental margin in the Pacific Ocean as described by Bacon et al. (1988), the strong 230 Th and 231 Pa scavenging at Kerguelen (a small spot in the Southern Ocean) should not create a significant depletion of these radionuclides throughout the whole Southern Ocean.
However, these results stress that, even for particulate reactive metals with short residence time in the water column, the effect of lateral advection over the Kerguelen plateau cannot be neglected. Indeed, subsequently to the publication of the boundary scavenging model for the Kerguelen plateau, Chever et al. (2010) clearly showed that lateral advection of dissolved iron towards the Kerguelen plateau should be also considered as a predominant source of total dissolved iron (total apparent particulate iron and dissolved iron) above the plateau.
Thus, even if the "advection-scavenging model" applied to 230 Th and 231 Pa, cannot be strictly applied to other elements (like iron), it nevertheless shows that it is necessary to take this advection into account to establish the budget of other particle reactive elements, which have been treated in a one-dimensional way over the Kerguelen plateau (Blain et al., 2007).

Conclusions
In the wake of the Kerguelen plateau, 231 Pa xs and 230 Th xs distributions in seawater and particles and their fractionation are in the range of those previously observed in the Southern Ocean. This study shows that in the Kerguelen wake dominated by biogenic silica, 231 Pa removal is at least as efficient and possibly more efficient than 230 Th removal, thereby setting the Pa/Th ratios and fractionation factor (F Th/Pa ≤ 1) in the water column, and consequently in the sediments of this area of the Southern Ocean.
We confirm that along the eastern plateau escarpment, an intense scavenging occurs in the deep water as depicted by the decrease of both dissolved 230 Th xs and 231 Pa xs concentrations. The lack of fractionation of 231 Pa xs and 230 Th xs (F Th/Pa ≈1) associated with this intense scavenging was attributed to nepheloid layers inducing re-suspension of BSi-rich particles stripping the deep water column of both radionuclides. To our knowledge, this work is the first clear observation of boundary scavenging directly from the decrease of both dissolved 231 Pa and 230 Th concentrations when a water mass flows along an "ocean boundary".
In addition, it sheds a new light on the boundary scavenging processes. Indeed, boundary scavenging, depicted in this study by changes in the fractionation of Pa and Th, occurs in the wake of Kerguelen due to the combination of both particle effect with a gradient in the particle fluxes (e.g. higher particle flux at the margin than in the open ocean) and the effect of particle composition (dominated here by opal).
On the Kerguelen plateau, using an "advection-scavenging model", we show that the advection of open-ocean water, rich in dissolved 230 Th and 231 Pa, plays a critical role in the radionuclides budget by balancing the drawdown of both nuclides concentrations due to the strong scavenging. Consequently, we strongly recommend that, so as to establish the distributions of other elements of interest in the water column in high scavenging areas (e.g. the Kerguelen plateau in this study), the advection of water from a lower scavenging area should be taken into account.
Ultimately, the future publications on Th and Pa distributions and their fractionation in contrasted areas of the ocean will provide more realistic representations of particle composition and particle dynamics, necessary to develop a more sophisticated modelling approach of the scavenging processes of particle-reactive elements in Global Circulation Models (Dutay et al., 2009).