Early spring mesopelagic carbon remineralization and transfer efficiency in the naturally iron-fertilized Kerguelen area

. We report on the zonal variability of mesopelagic particulate organic carbon remineralization upper 150 m export, resulting in a zero transfer efﬁciency to depths > 800 m. In the polar front meander (time series), the capacity of the meander to transfer carbon to depth > 800 m was highly variable (0 to 73 %). The highest carbon transfer efﬁciencies in the meander are furthermore coupled to intense and complete deep ( > 800 m) remineralization, resulting again in a near-zero, deep ( > 2000 m) carbon sequestration efﬁciency there.

upper 150 m export, resulting in a zero transfer efficiency to depths > 800 m. In the polar front meander (time series), the capacity of the meander to transfer carbon to depth > 800 m was highly variable (0 to 73 %). The highest carbon transfer efficiencies in the meander are furthermore coupled to intense and complete deep (> 800 m) remineralization, resulting again in a near-zero, deep (> 2000 m) carbon sequestration efficiency there.

Introduction
While numerous artificial (Boyd et al., 2000(Boyd et al., , 2004Gervais et al., 2002;Buesseler et al., 2004Buesseler et al., , 2005de Baar et al., 2005;Hoffmann et al., 2006;Boyd et al., 2012;Smetacek et al., 2012) and natural (Blain et al., 2007;Pollard et al., 2009;Zhou et al., 2010Zhou et al., , 2013 ocean iron-fertilization experiments in the Southern Ocean have demonstrated the role of iron in enhancing the phytoplankton biomass and production in high-nutrient low-chlorophyll (HNLC) regions, determining to what extent fertilization could modify the transfer of particulate organic carbon (POC) to the deep ocean is far from being comprehensively achieved (Lampitt et al., 2008;Morris and Charette, 2013;Le Moigne et al., 2014;Robinson et al., 2014). This is partly due to the short term over which the observations were made, precluding extrapolation to longer timescales. Moreover, when assessing whether Fe-supply could induce vertical POC transfer, the magnitude of the export from the surface is not the only important parameter to take into account. Indeed, POC fate in the mesopelagic zone (defined as the 100-1000 m depth layer) is often largely overlooked although these depth layers are responsible for the remineralization of most of the POC exported from the surface layer (Martin et al., 1987;Longhurst, 1990;Lampitt and Antia, 1997;François et al., 2002;Buesseler et al., 2007b;Buesseler and Boyd, 2009). Only few studies have considered mesopelagic carbon (C) remineralization rates (Buesseler et al., 2007a;Jacquet et al., 2008aJacquet et al., , b, 2011aSalter et al., 2007) to estimate the response of deep POC export to fertilization. Assessing mesopelagic C remineralization is pivotal for evaluating remineralization length scale as well as the timescale of the C storage in the deep ocean. Indeed the typical depth of the main thermocline, 1000 m (IPCC, WG1, 2007, chapter 5) is often referred to as the horizon, clearly removed from the surface ocean and atmosphere (Passow and Carlson, 2012). Overall, assessing mesopelagic C remineralization will allow one to better quantify the ocean's biological carbon pump and its efficiency in the global C cycle which holds large uncertainty and is currently under debate (e.g. from 5 Gt yr −1 in Henson et al., 2011 to 21 Gt C yr −1 in Laws et al., 2000 and 13 Gt yr −1 in the IPCC WG1 report (2013, chapter 6)).
The present work aims at understanding the impact of a natural iron-induced bloom on the mesopelagic POC rem-ineralization and zonal variability in the Kerguelen area (Southern Ocean). Here, C remineralization was assessed from particulate biogenic Ba (hereafter called excess Ba or Ba xs ; mainly forms as barite BaSO 4 crystals) contents in the mesopelagic water column. The link between barite and C remineralization depends on the fact that this mineral precipitates in oversaturated micro-environments (biogenic aggregates) during the process of prokaryotic degradation of sinking POC (Dehairs et al., 1980(Dehairs et al., , 1992(Dehairs et al., , 1997Stroobants et al., 1991, Cardinal et al., 2001Jacquet et al., 2007Jacquet et al., , 2008bJacquet et al., , 2011aPlanchon et al., 2013;Sternberg et al., 2007Sternberg et al., , 2008aSternberg et al., , 2008b. Once the aggregates have been remineralized, barites are released and spread through the mesopelagic layer. Overall, earlier work highlights the fact that suspended barite in mesopelagic waters builds up over the growing season and reflects past remineralization activity integrated over several days to weeks (Dehairs et al., 1997;Cardinal et al., 2005;Jacquet et al., 2007Jacquet et al., , 2008b). An algorithm linking mesopelagic Ba xs contents to oxygen consumption (Shopova et al., 1995;Dehairs et al., 1997) allowed remineralization of POC fluxes to be estimated for the mesopelagic layer. Combined with surface C production and export estimates, mesopelagic Ba xs also highlights the deep carbon transfer efficiency of the system. From earlier studies, the efficiency of C transfer through the mesopelagic layer was reported to increase under artificially induced (EIFEX; Strass et al., 2005;Smetacek et al., 2012) and natural (KEOPS; Blain et al., 2007) Fe-replete conditions (Jacquet et al., 2008a, b;Savoye et al., 2008) compared to Fe-limited, non-bloom HNLC reference stations in the Southern Ocean. In contrast, C transfer efficiency through the mesopelagic layer was reported to be lower in natural Fe-replete locations during the SAZ-Sense (Sub-Antarctic Zone Sensitivity to Environmental Change) cruise off Tasmania (Jacquet et al., 2011a, b). Differences in plankton community structure and composition (e.g. diatoms vs. flagellates, type of diatoms) were suggested as possible causes of such discrepancies in C transfer efficiency through the mesopelagic layer (Jacquet et al., 2008a(Jacquet et al., , 2011a. Also, differences in integration timescales for the processes that control the carbon fluxes in artificially vs. naturally Fe fertilized systems may yield an incomplete picture of the C transfer potential and lead to misleading conclusions. Here, we examine changes in mesopelagic POC remineralization during the early spring (October-November 2011) KEOPS 2 expedition to the naturally iron-fertilized area east of Kerguelen Islands. The hydrographic structure of the Kerguelen area generates contrasted environments that are differently impacted by iron availability and mesoscale activity. The specific objectives of the present work are to assess the zonal variability of mesopelagic C remineralization and deep C transfer potential, and to identify possible causes of this variability. The same area was visited earlier in 2005 during summer at a late stage of the bloom (KEOPS 1; January-February 2005), offering a unique opportunity to estimate the main carbon fluxes over most of the growth sea- son. Mesopelagic C remineralization estimates are compared to particle and biological parameters as reported in other papers included in this issue Christaki et al., 2014;Dehairs et al., 2014;Laurenceau-Cornec et al., 2015;Planchon et al., 2014;van der Merve et al., 2015) and in Blain et al. (2007), Christaki et al. (2008), Jacquet et al. (2008a), Park et al. (2008) and Savoye et al. (2008).

Study area
The KEOPS 2 (Kerguelen Ocean and Plateau compared Study) cruise was conducted in austral spring at the onset of the bloom from 10 October to 20 November aboard the R/V Marion Dufresne (TAAF/IPEV). The KEOPS 2 expedition studied the Kerguelen Plateau area (Indian sector of the Southern Ocean) which is characterized by the passage of the polar front (PF), as illustrated in Fig. 1a. The Kerguelen Plateau is surrounded by the Antarctic Circumpolar Current (ACC) whose main branch circulates to the north of the plateau (Park et al., 2008). A second branch of the ACC circulates to the south of Kerguelen Islands to later join a branch of the Fawn Trough Current (FTC). The FTC has a main northeast direction, but a minor branch splits away northwestward to join the eastern side of the Kerguelen Plateau (Park et al., 2008;Fig. 1a). These particular hydrographic features generate a mosaic of recurrent massive bloom patterns in the northeastern part of the plateau, and the possible sources and mechanisms for fertilization were investigated during ANTARES 3 (1995; Blain et al., 2001) and the KEOPS 1 cruise (January-February 2005, late summer conditions; Blain et al., 2007Blain et al., , 2008. During KEOPS 2 the evolution of Chl a data based on multi-satellite imagery of the study area revealed the presence of different Chl a rich plumes   (Fig. 1a; e.g. Chl a map from 11 November 2011). Stations were sampled in distinct zones covering these different bloom patterns (Fig. 1a) (corresponding stations are reported in Fig. 1b): (a) on the shallow plateau (station A3; see 1 in Fig. 1a). Note that station A3 coincides with a site studied during the KEOPS 1 cruise, and that it was sampled twice over a 27-day period; (b) in a meander formed by a quasi-permanent retroflection of the polar front (PF) and topographically steered by the eastern escarpment (Gallieni Spur) of the Kerguelen Plateau (mainly stations E, sampled as a quasi-Lagrangian temporal series) (see 2 in Fig. 1a); (c) along a north-south transect (referred to as TNS stations; see 3, grey line in Fig. 1a) and a eastwest transect (referred to as TEW stations; see 4, grey line in Fig. 1a), both crossing the PF; and (d) in the polar front zone (PFZ) in the vicinity (east) of the PF (station F-L; see 5 in Fig. 1a). Furthermore we also sampled a reference HNLC, non-bloom, non-Fe-fertilized station southwest of the plateau (station R-2; see 6 in Fig. 1a). Station locations are given in Table 1.
Detailed descriptions of the complex physical structure of the area, circulation, water masses and fronts are given in Park et al. (2014). Briefly, the main hydrodynamic features observed during the cruise are the following (see θ − S diagram, Fig. 2a): (1) north of the PF, stations in the PFZ (TNS-1, TEW-8 and F-L) present Antarctic Surface Waters (AASW; θ 4 • C and density < 27); θ − S characteristics between 150 to 400 m at station F-L (and to a lesser extent at station TNS-1) reveal the presence of interleaving with waters from northern (subantarctic) origin, centred between the 27.  Based on the θ − S characteristics (Fig. 2a, b) and surface phytoplankton biomasses, we can schematically group the stations as follows. The R-2 HNLC reference station (white dot in Fig. 1b) is characterized by a very low biomass (with low iron contents; Quéroué et al., 2015). Stations TEW-3 and TNS-8 (black dots) are characterized by low to moderate biomass and Fe contents. Stations A3 and E-4W (red dots; south of the PF) as well as stations TNS-1, F-L and TEW-8 (blue dots; north of the PF) are characterized by high biomass and iron contents. Stations in the core of the PF meander (green dots; stations TNS-6, E-1, E-2, E-3, E-4E and E-5 considered as a temporal series) are characterized by moderate biomass and iron contents.

Sampling and analyses
Twenty-two CTD (conductivity, temperature, and depth) casts (surface to 500-2000 m) were sampled for particulate barium (Table 1) using a CTD rosette equipped with twenty-two 12 L Niskin bottles. Deep particulate Ba profiles (> 1000 m) were not systematically obtained from the same CTD cast, but from successive casts sampled closely in time and space and having similar θ − S data profiles. In the following, we use both the station and CTD numbers to refer to stations.
Four to 7 L of seawater were filtered onto 47 mm polycarbonate membranes (0.4 µm porosity) under slight overpressure supplied by filtered air (0.4 µm). The filters were rinsed with Milli-Q grade water (< 5 mL) to remove sea salt, dried (at 50 • C) and stored in Petri dishes for later analysis. In the laboratory, we performed a total digestion of samples using a tri-acid (0.5 mL HF / 1.5 mL HCl / 1 mL HNO 3 ; all Suprapur grade) mixture in closed Teflon beakers overnight at 90 • C in a clean pressurized room. After evaporating until nearly dry, samples were re-dissolved into around 13 mL of 2 % HNO 3 . The solutions were analysed for Ba and other major and minor elements using a ICP-QMS (inductively coupled plasma quadrupole mass spectrometer; X Series 2 Thermo Fisher) equipped with collision cell technology (CCT). To correct instrumental drift and matrix effects, internal standards and matrix-matched calibrations were used. We analysed several certified reference materials which consisted of dilute aciddigested rocks (BHVO-1, JB-3 and JGb-1), natural riverine water (SLRS-5) and multi-element artificial solutions for these external calibrations. Based on analyses of these external standards, accuracy and reproducibility are both better than ±5 %. For more details on sample processing and analysis we refer to Cardinal et al. (2001). Among all elements analysed, we were particularly interested in Ba and Al. The presence of sea salt was checked by analysing Na and the sea-salt particulate Ba contribution was found to be negligible. Average detection limits equal 0.6 nM for Al and 3 pM for Ba. Detection limits were calculated as 3 times the standard deviation of the blank measured on board and then normalized to an average dilution factor of 385, i.e. particles from around 5 L of Milli-Q water, dissolved in a final volume of 13 mL as for the samples. Biogenic barium (hereafter called excess Ba or Ba xs ) was calculated as the difference between total particulate Ba and lithogenic Ba using Al as the lithogenic reference element (Dymond et al., 1992;Taylor and McLennan, 1985). At most sites and depths, the biogenic Ba xs represented > 95 % of total particulate Ba. Lithogenic Ba reached up to 20 % of total particulate Ba at some depths in the upper 80-100 m mainly at station R-2 and stations north of the polar front (i.e. TEW-8, F-L and TNS-1). The standard uncertainty (Ellison et al., 2000) of Ba xs data ranges between 5 and 5.5 %. Ba xs and Al data are reported in Appendix A.

O 2 consumption and POC remineralization
The rate of oxygen consumption and particulate organic carbon remineralization rate in the mesopelagic layer (later referred to as MR) can be estimated using an algorithm relating mesopelagic Ba xs contents and oxygen consumption based on earlier observations in the Southern Ocean (Shopova et al., 1995;Dehairs et al., 1997Dehairs et al., , 2008. The detailed calculations are described in Jacquet et al. (2008aJacquet et al. ( , 2011a. Briefly, we use the following equations: where J O 2 is the O 2 consumption (µmol L −1 day −1 ) and C respired is the mineralization rate of organic carbon (in mmol C m −2 day −1 ; MR); Ba xs is the depth-weighted average (DWA) Ba xs value, i.e. the Ba xs inventory divided by the depth layer considered Z, Ba residual is the residual Ba xs signal (or Ba xs background) at zero oxygen consumption and RR is the Redfield C / O 2 molar ratio (127/175; Broecker et al., 1985). DWA Ba xs values were calculated for both layers at 150 to 400 m (plateau and deep stations) and 150 to 800 m (deep stations only; see details further below). The residual Ba xs is considered as "preformed" Ba xs , left over after partial dissolution and sedimentation of Ba xs produced during a previous phytoplankton growth event. In BaSO 4 , saturated waters, such as the ones filling the whole ACC water column (Monnin et al., 1999), this background Ba xs value was considered to reach 180 to 200 pM, which is rather characteristic for the deep ocean (> 1000 m) (see Dehairs et al., 1997;Jacquet et al., 2008aJacquet et al., , 2011. In the present study we used a Ba xs background of 180 pM. We take the opportunity here to also compare O 2 consumption rates for the KEOPS 1 expedition (D. Lefèvre, unpublished data) with KEOPS 1 Ba xs data published earlier (Jacquet et al., 2008a). No such O 2 consumption data are available for KEOPS 2. During KEOPS 1, dark community respiration (DCR) was estimated from changes in the dissolved oxygen concentration over 72 h incubations. Discrete samples were collected at three depths in the mesopelagic www.biogeosciences.net/12/1713/2015/ Biogeosciences, 12, 1713-1731, 2015 zone from Niskin bottles and placed into 125 cm 3 borosilicate glass bottles according to the WOCE (World Ocean Circulation Experiment) procedure, and oxygen concentration was determined by Winkler titrations using a photometric endpoint detector (Williams and Jenkinson, 1982). By integrating DCR data with the water column, we estimated the rate of oxygen consumption (referred to as JO 2 -W). We compared JO 2 -W obtained from incubated oxygen samples with the rate of oxygen consumption based on KEOPS 1 mesopelagic Ba xs contents (Eq. 1; later referred to as JO 2 -Ba). Dissolved oxygen was measured three times at station A3 (same location as during KEOPS2) over a 19-day period (A3 CTD no. 32, 74 and 119). Dissolved oxygen was also measured at station C11, located off-shelf in less productive HNLC waters (51.65 • S, 78.00 • E; not shown in Fig. 1) and was sampled two times over a 10-day period (C11 CTD no. 42 and 83). Figure 3 compares JO 2 -W and JO 2 -Ba for repeat stations A3 (no. 32, 74 and 119) and C11 (no. 42 and 83) (integration between 150-300 m). JO 2 -W rates range from 0.082 to 0.208 mmol m −2 day −1 at station A3 and from 0.292 to 0.528 mmol m −2 day −1 at station C11. Although JO 2 -Ba rates (from 0.846 to 1.555 mmol m −2 day −1 ) are slightly higher than JO 2 -W, JO 2 rates are of the same order of magnitude and present the same trend. We observed a significant positive correlation between the JO 2 rates (R 2 =0.90; p < 0.01), with a slope of 0.64. The difference in oxygen consumption rates can be explained by the integration time of both methods (a few hours for the incubations vs. few days to weeks for Ba xs ) and by the fact that KEOPS 1 occurred at the decline of the bloom (late summer; low organic substrates), which would explain the lower JO 2 rates as estimated by the incubation method.
Overall, these results highlight the need for further constraining spatial and temporal variability of deep ocean oxygen utilization via a combination of direct rate measurements and the Ba xs proxy. In the present work, O 2 consumption and POC remineralization were assessed from Ba xs inventories and Eqs. (1) and (2). C remineralization rates are given in Table 1. Relative standard uncertainties (Ellison et al., 2000) of C remineralization ranged between 4 and 20 %.

Particulate biogenic Ba xs profiles
Ba xs profiles in the upper 800 m are reported in Fig. 4. The complete whole water column data set is given in Appendix A. From previous studies we know that Ba xs in surface waters is distributed over different, mainly non-barite biogenic phases (see Stroobants et al., 1991;Jacquet et al., 2007;Cardinal et al., 2005;Sternberg et al., 2005). As such, these do not reflect POC remineralization processes, in contrast to mesopelagic waters where Ba xs is mainly composed of barite (Dehairs et al., 1980)  degradation of sinking POC (Martin et al., 1987;Sarmiento et al., 1993;Buesseler et al., 2007b). For KEOPS 2 we observed that Ba xs concentrations generally increase below 150 m (i.e. they increase above the background level set at 180 pM), but some sites have ocean surface Ba xs contents significantly larger than background (E-1, 896 pM at 21 m; E4-E, 563 pM at 93 m). Such values are not unusual, and very high surface values have been observed occasionally in earlier Southern Ocean studies. During KEOPS 1, surface Ba xs maxima at the three A3 repeat stations ranged from 1354 to 5930 pM at 50 m, likely associated with phytoplanktonderived particles (Jacquet et al., 2008a). The following part focuses on the mesopelagic zone where most of the remineralization of exported organic matter takes place. The Ba xs profile for station R-2 (CTD no. 17) displayed a characteristic mesopelagic Ba xs maximum, reaching up to 834 pM at 304 m, which is actually one of the highest values observed for the whole study (Fig. 4a). Ba xs profiles for stations A3 above the Kerguelen Plateau (A3-1 CTD no. 4 and A3-2 CTD no. 107; Fig. 4b) had lower mesopelagic Ba xs content, with values ranging from about 80 to 350 pM. For both A3 visits, Ba xs values increased close to the seafloor, reaching up to 1108 pM (A3-1, 474 m) and 1842 pM (A3-2, 513 m). In contrast, station E-4W (located further north along the margin in deeper waters, but with similar θ − S and Chl a characteristics as station A3) displayed a large mesopelagic Ba xs maximum reaching up to 627 pM at 252 m (Fig. 4c). Station TEW-3 (located on the Kerguelen Plateau, in waters with similar θ − S and Chl a characteristics as station TNS-8) had a profile similar to the one observed at station A3-2, but compared to plateau sites A3-1 and A3-2, no increased Ba xs contents were observed in bottom water (Fig. 4d). The other stations of the study area ( Fig. 4d- files similar to the one at station E-4W, showing the characteristic Ba xs maximum between 200 and 500 m. Note that for most of the stations, Ba xs concentrations in waters below the mesopelagic maximum did not systematically decrease to reach the Ba xs background level (180 pM; see above). In some cases Ba xs contents significantly higher than residual Ba xs were observed until below 1000 m (see Appendix A). This is particularly salient at stations TNS-6, E-1, E-2 and F-L where Ba xs values below 1000 m reach 410 pM at 1886 m (TNS-6) and 436 pM at 1498 m (E-1). These cases of high, deep Ba xs contents clearly contrasted with the values observed at station E4-E (Fig. 4h).

Depth-weighted average Ba xs content of mesopelagic waters
Since the base of the mixed layer was generally shallower than 150 m, this depth is taken as the upper boundary of the mesopelagic domain. The depth-weighted average (DWA) Ba xs contents, calculated for the 150-400 and 150-800 m depth intervals, are given in Table 1. For the profiles on the plateau (500 m water column), bottom waters with evidence of sediment resuspension were not taken into account when calculating DWA Ba xs values (≥ 400 m). Particle size spectra indicated that sediment resuspension occurred especially at stations A3 and TEW-3 (Jouandet et al., 2014;van der Merve et al., 2015;). Thus, at site A3 ( Fig. 4b), DWA Ba xs was calculated for the layer between 150 and 354 m for A3-1 (CTD no. 4) and between 150 and 405 m for A3-2 (CTD no. 107). For station TEW-3 (CTD no. 38), DWA Ba xs was calculated for the water layer between 150 and 400 m (Fig. 4d). For the deep sites, we considered both the 150-400 and the 150-800 m depth intervals when calculating the DWA Ba xs contents. Depth-weighted average Ba xs values were translated into carbon remineralization rates using Eqs. (1) and (2) given above. These rates ranged from 2 to 91 mg C m −2 day −1 (Table 1). DWA Ba xs values range from 199 to 572 pM (Table 1) and fit within the range reported for polar front areas during previous studies (Cardinal et al., 2001Jacquet et al., 2005Jacquet et al., , 2008aJacquet et al., , b, 2011Planchon et al., 2013). However, we note that the mesopelagic Ba xs maximum at R-2 occurs at shallower depths, around 300 m, and that there is no evidence for elevated values at 500 m where the previous authors reported higher trace element and silica concentrations. Also, as reported above (see Sect. 2.2 and Appendix A), the higher lithogenic Ba fractions at R-2 (up to 20 % of the total Ba) occur only in the upper 80 m. Moreover, we note that surface waters at R-2 has already experienced some nitrate consumption as compared to subsurface winter waters (T min waters). Indeed, surface waters had 10 % less nitrate than winter water (26 µM at 5 m vs. 29 µM at 200 m), and the isotopic enrichment of this surface nitrate confirmed a suggestion of uptake (see Dehairs et al., 2014). Also,  reported relatively low Si : C and Si : N ratios for surface ocean suspended matter), pointing to the development of a diatom assemblage just prior to the sampling, consistent with the high dissolution rates of biogenic silica (BSi) that Closset et al. (2014) reported for R-2 surface waters. It is therefore likely that the mesopelagic Ba xs content at R-2 indeed reflects remineralization of organic material that was fuelled by an important past early spring production and export event. Similarly, during late winter (November 1993) F. Dehairs (unpublished results) observed the presence of significant numbers of barite microcrystals in mesopelagic waters at the KERFIX time series station (50 • 40 S, 68 • 25 E) located east of R-2. Results would thus suggest the occurrence in this HNLC area of recurrent brief early spring diatom productive period pulses and subsequent export and remineralization activity in the underling layers. Chl a satellite images (Giovanni -Interactive Visualization and Analysis, NASA GES DISC) corroborate that the R-2 and KERFIX area is occasionally subject to enhanced biomass during early spring; b. The two successive visits (27 days apart) at site A3 yielded relatively low DWA Ba xs values of 267 and 316 pM, and a quite similar value was observed for the shallow station TEW-3 (324 pM), located further north on the plateau and north of the PF. Note that for comparison purposes, we recalculated the DWA Ba xs and MR values of KEOPS 1 by considering upper and lower mesopelagic layer boundaries of 150 and 400 m rather than 125 and 450 m, as in Jacquet et al. (2008a). Also, in the aforementioned study the high Ba xs contents observed near the seafloor were not excluded from the calculations, while they are here. These increased benthic boundary layer Ba xs contents (observed also during KEOPS 2) are due to sediment resuspension which extended up to 70 m above the seafloor during KEOPS 1 (Blain et al., 2008;Venchiarutti et al., 2008;Armand et al., 2008). Because of these slightly different depth intervals over which Ba xs values were integrated, the KEOPS 1 values discussed here will be slightly differ-ent from those reported in Jacquet et al. (2008a). At the other depths, the lithogenic Ba contribution at A3 (KEOPS 2) was only minor; c. The time series stations in the polar front meander had DWA Ba xs contents ranging from 258 to 427 pM (150-400 m), so reaching values exceeding those on the plateau. For these time series, stations' values decreased between day 0 (TNS-6) and 12 (E-3), and then increased again at days 22 (E-4E) and 27 (E-5). Stations E-4W and TNS-8, above the plateau but in deeper waters close to the Kerguelen margin, at the edge the high biomass plume (Fig. 1), had the highest DWA Ba xs values (468 and 473 pM, respectively; 150-400 m), not considering the R-2 reference station. The polar front F-L site, although located within the eastern part of the high biomass plume, had a smaller DWA Ba xs value of 345 pM (150-400 m) and the nearby station TEW-8 had the lowest DWA Ba xs value of the whole study area (199 pM; 150-400 m).

Mesopelagic Ba xs and bacterial production
Previous studies revealed that the shape of the columnintegrated bacterial production (BP) profile (i.e. the attenuation length scale) was important in setting the Ba xs signal in the mesopelagic zone Jacquet et al., 2008aJacquet et al., , 2011a. Mesopelagic Ba xs content is smaller when most of the column-integrated BP is restricted to the upper mixed layer (indicating an efficient, near-complete remineralization within the surface), compared to situations where a significant part of integrated BP was located deeper in the water column (reflecting significant deep bacterial activity and POC export). During KEOPS 2 the incorporation of 3 Hleucine was used to estimate bacterial production. BP data are described in Christaki et al. (2014). In Fig. 5 we compare column-integrated BP at 150 m over 400 m (BP150 / 400) and DWA Ba xs for the 150-400 m depth interval, along with the relationship obtained during KEOPS 1 (BP200 / 125 and 150-450 m DWA Ba xs ; Jacquet et al., 2008a;Christaki et al., 2008). Excluding stations A3, E-1, E-2 and E-3, KEOPS 2 data presented a significant correlation (R 2 =0.88; p < 0.01) and a similar trend to the one reported for KEOPS 1. A similar picture was obtained when integrating DWA Ba xs and BP up to 800 m (not shown). The time series "E" stations in the meander revealed a shift from stations E-1, E-2 and E-3 to stations E-4E and E-5, i.e. towards the trend reported above. A shift was also apparent at station A3 from KEOPS 2 (early spring) to KEOPS 1 (late summer). It is thus possible that results reflect the occurrence of different stages of bloom advancement. The large variability of the Ba xs and BP relationship during KEOPS 2, especially at A3 site and in the me -Biogeosciences, 12, 1713-1731, 2015 www.biogeosciences.net/12/1713/2015/ Table 2. Comparison of mesopelagic POC remineralization (MR) with primary production (PP) and export production (EP). All fluxes in mg C m −2 day −1 . r ratio is the ratio of Mr over EP. The C sequestration (or transfer) efficiency at 400 and 800 m (T400, T800) is the fraction of C export (EP) at 150 m exiting through the 400 m (T400) or the 800 m (T800) horizons. See text for further information on calculation.  Cavagna et al. (2015). d EP data from .  (Christaki et al., 2008;Jacquet et al., 2008a). ander, could reflect the temporal evolution and patchiness of the establishment of mesopelagic remineralization processes in this polar front area.

Fate of exported organic C in the mesopelagic zone and deep water column
An important question relates to the fate of the exported POC: how much of this POC is respired in the mesopelagic waters and how much escapes remineralization and is exported to deeper layers where longer-term sequestration is likely (see e.g. Passow and Carlson, 2012;Robinson et al., 2014;Schneider et al., 2008). To address these questions, we defined two ratios: (1) the mesopelagic C remineralization efficiency (r ratio in Table 2), which is the ratio of mesopelagic C remineralization (MR, based on the DWA Ba xs concentrations) over C export (EP) from the 150 m horizon (based on 234 Th, see Planchon et al., 2014), and (2) the C transfer efficiency at 400 and 800 m (i.e. T400, T800 in Table 2), which is the fraction of C export (EP) at 150 m passing through the 400 m (T400) or the 800 m (T800) horizons (e.g. T400 = EP400 / EP150 = 1-(MR / EP150), with MR / EP150 = r ratio; see above). This approach is similar to the one developed by Buesseler and Boyd (2009) stating that a conventional curve-fitting of particle flux data (i.e. power law or exponential) skews our interpretation of the mesopelagic processes. They recommended the use of combined metrics to capture and compare differences in flux attenuation. In the following, we compare MR fluxes for the different KEOPS 2 areas (reference site; plateau sites; polar front and polar front meander) and discuss remineralization and transfer efficiencies for those sites for which MR, primary production (PP) and/or EP data (Table 2) were available. PP data were estimated from uptake experiments including 24 h incubations at different PAR levels over the euphotic layer, i.e. up to the 0.01 % PAR level . EP data were estimated from 234 Th activities and 234 Th / POC ratios and are discussed in Planchon et al. (2014). The thorium method integrates POC export over a 1-month period ( 234 Th half live is 24.1 days). We remind the reader here that MR fluxes based on mesopelagic Ba xs reflect past remineralization activity integrated over several days to a few weeks (Dehairs et al., 1997;Jacquet et al., 2007Jacquet et al., , 2008b Lefèvre et al. (2008) and Mosseri et al. (2008), EP data are detailed in Savoye et al. (2008) and Ba xs data are described in Jacquet et al. (2008a).

Reference station R-2
Since station R-2 had the highest DWA Ba xs content, it yielded the highest MR flux of the whole study area (91 mg C m −2 day −1 ; 150-800 m; Table 2). In contrast, both PP and EP fluxes at R-2 were very low (132 and 10 mg C m −2 day −1 , respectively) and the calculated MR flux exceeded EP (Table 2). The resulting export efficiency (EP / PP) was high, and T400 and T800 values (the fraction of EP exported deeper than 400 and 800 m, as defined above) equal 0 (i.e. no export of POC beyond 400 and 800 m; note that > 100 % values, i.e. MR > EP, were set to zero in Fig. 7a and Table 2). The fact that MR exceeds EP therefore implies a non-steady state condition at the R-2 site. As reported above, R-2 probably experienced a brief early spring diatom production pulse days to a few weeks before the start of the KEOPS 2 cruise, followed by subsequent export and very important remineralization activity in the underling layers as depicted by MR data.

Station A3 on the plateau
The MR fluxes on the plateau varied little between the two visits 27 days apart (Table 1) and, moreover, as discussed below they were similar to summer values obtained during KEOPS 1 (see Jacquet et al., 2008a) when the same A3 site was sampled three times over a 19-day period. While during KEOPS 2 (spring) MR fluxes at A3 ranged from 11 to 14 mg C m −2 day −1 (with a standard uncertainty of around 5 %), they were slightly larger during KEOPS 1 (summer; 17 to 23 mg C m −2 day −1 ) (Fig. 5). We observed differences in the mesopelagic POC remineralization efficiency between the two seasons (r ratio, blue values in Fig. 6, Table 2). During KEOPS 1, r ratios (MR / EP) remained low, ranging from 7 to 9 % of EP at A3, while during KEOPS 2, r ratios were slightly higher but decreased from 29 % (A3-1; first visit) to 13 %, 27 days later (A3-2; second visit). This variation in r ratio during KEOPS 2 is mostly due to an increase of EP (from 47 to 85 mg C m −2 day −1 ; Planchon et al., 2014) over the same period while MR showed little change. Although at this early stage of the season (spring) PP at A3-2 had already reached 2172 mg C m −2 day −1 , EP remained relatively low (85 mg C m −2 day −1 ).
Here EP accounted for only about 4 % of PP (low export efficiency; see green data points in Fig. 5). These conditions suggested that phytoplankton biomass had accumulated in the surface waters without significant export at that point, or that C had been channelled to higher trophic levels as suggested by Christaki et al. (2014). Note that a negative relationship between primary productivity and surface carbon export efficiency has already been reported from previous studies in the Southern Ocean (Lam et al., 2007;Morris et al., 2007;Savoye et al., 2008;Jacquet et al., 2011a, b). Among possible explanations for the occurrence of highproductivity low export efficiency regimes in high-latitude systems Maiti et al. (2013) mentioned differences in trophic structure, grazing intensity, recycling efficiency, high bacterial activity or increase in DOC export, but the exact reason remain unclear. In contrast, during KEOPS 1 (summer), EP fluxes reached 250 mg C m −2 day −1 at 125 m (14-31 % of PP), while PP ranged from 865 to 1872 mg C m −2 day −1 , reflecting enhanced export efficiency (Jacquet et al., 2008a;Savoye et al., 2008). It is important to underline the fact that MR at station A3 was only slightly higher in summer than in spring, especially considering the large differences in export efficiency between seasons. According to results from sediment traps deployed over 1 year at the A3 site, Rembauville et al. (2014) reported that 60 % of the annual POC export at the base of the mixed layer occurred over a short periods of time representing < 4 % of the year and was composed of small highly silicified, fast-sinking, resting spores of diatoms that bypass grazing pressure. According to these authors, the pulses are linked to nutrient depletion dynamics inducing resting spore formation. During the rest of the year, the flux was composed of small diatoms (empty frustules) and small fecal pellets, with efficient C retention in the surface layer or transfer to trophic levels. If we consider that export conditions during KEOPS 2 are more similar to those prevailing most of the year, it is surprising that during KEOPS 1 (which would reflect an export event toward the end of the growth season) MR is not more important. This would indicate that fast-sinking, highly silicified and pulsed material was directly transferred to the bottom without major remineralization. Note for example that at the complex R-2 reference station, a small export event  held heavily silicified diatoms  and that the material was efficiently remineralized in the upper mesopelagic layer as witnessed by the high MR values we observed for that station. For the KEOPS 2 A3 site, Laurenceau-Cornec et al. (2015) reported that the sinking flux collected in the upper layer using gel-filled sediment traps was composed of phytodetrital aggregates that held slightly silicified diatoms . Even considering the shift from slightly to highly silicified material transfer between spring (KEOPS2) and summer (KEOPS 1), MR only slightly increases between both periods. Also, the mesozooplankton biomass at A3-2 was one of the highest of the KEOPS2 cruise, with a doubling from KEOPS 2 (early spring) to KEOPS 1 (late summer) . It is thus possible that at A3 the export event reported above, combined with a lasting grazing pressure, could have induced this rather low and enduring mesopelagic remineralization.   Since no PP data are available for that station, the EP / PP value has been set to 0. Isolines represent the modelled 1, 5, 10, 20 and 30 % of PP export to depths > at 400 or 800 m, and represent export efficiency.

Transfer efficiency
We also wonder if the shallow water column at A3 combined with lateral advection above the plateau plays a role in triggering the mesopelagic POC remineralization activity and in setting its efficiency. For KEOPS 1, Venchiarutti et al. (2008) reported that lateral advection over the plateau could significantly impact particle dynamics. During KEOPS 1, station B1 (CTD68), located on the plateau upstream from A3 according to the plateau circulation (Park et al., 2008), exhibited a very similar Ba xs distribution as station A3: low mesopelagic Ba xs and important bottom resuspension (not shown here; see Jacquet et al., 2008a). These strong similarities in Ba xs profile shapes would indicate that next to the pulsed nature of the events, the dynamics on the shallow plateau play an important role in limiting the extent of mesopelagic POC remineralization processes.
In Fig. 7a, the ratio of EP to PP (export efficiency) vs. the fraction of EP exported deeper than 400 m (i.e. T400; defined above) is shown for both KEOPS cruises. Note that for station A3-1 (KEOPS 2), there are no PP data. The A3 site shows increasing EP / PP ratios from spring (KEOPS 2) to late summer (KEOPS 1), and so do the T400 values (A3-1: 70 %; A3-2: 87 %; KEOPS 1 A3 site: 92 ± 1 %). Station E-4W is located in waters with similar θ − S and Chl a characteristics as the A3 plateau site but has a deeper water column (1384 m has PP and EP fluxes of the same order of magnitude (Table 2)). However, MR values (36 mg C m −2 day −1 ; 150-400 m) are larger at E-4W, resulting in a lower T400 value of around 33 %, compared to 87 % for A3-2 (Fig. 7a). When integrating between 150 and 800 m, T800 at E-4W equals 0 (i.e. no export of POC beyond 800 m; Fig. 7a and Table 2). Station F-L (in the vicinity of the PF; 74.7 • E) appears to function in a similar way as observed for E-4W (71.4 • E). PP at station F-L is relatively high (3380 mg C m −2 day −1 ), while EP is quite low (43 mg C m −2 day −1 ), reflecting the fact that the biomass was not yet exported from the surface waters or was transported to higher trophic levels. Since MR fluxes are slightly lower (21 mg C m −2 day −1 ; 150-400 m) at F-L than at E-4W, resulting T400 values are higher (52 %) there.
Overall, during KEOPS 2, it appears that biomass at stations A3, E-4W and F-L (sites of high productivity) had been accumulating in surface waters (e.g. transfer to higher trophic levels) and export had not yet started, considering the early stage of the season during KEOPS 2. Our observations allow us to conclude the following: 1. Both seasons (KEOPS 1 and KEOPS 2) showed a similar functioning of the mesopelagic ecosystem at A3. The rather low and enduring MR fluxes under high production and variable export regimes (high export efficiency during KEOPS 1 and low export efficiency during KEOPS 2) indicated that here mesopelagic remineralization does not represent a major resistance to organic matter transfer to the seafloor at A3. On average (considering both seasons, but excluding A3-1), the C transfer efficiency into the deep (> 400 m) as assessed by PP, EP and MR fluxes comparisons reached 91 ± 3 % at A3; 2. In contrast to A3, E-4W and F-L showed important mesopelagic remineralization rates, reducing the effi-ciency of C transfer beyond 400 m to 33 and 52 %, respectively, and to zero for both stations beyond 800 m. Bottom depth, lateral advection, zooplankton grazing pressure and the pulsed nature of the POC transfer at A3 were the particular conditions that could drive the differences in C transfer efficiency between A3 and E4-W and F-L and limit the extent of MR processes at A3.

Stations in the meander
Temporal short-term changes for the stations TNS-6, E-1, E-2, E-3, E-4E and E-5, located in the polar front meander, will be discussed in this section. Note that no PP or EP data exist for TNS-6. From Table 2 it appears that PP almost doubled between E-1 and E-5, but this increase was not paralleled by an increase of EP and MR, except for the 30 % EP increase from E-1 to E-3. In fact, overall EP shows a decreasing trend with time, while MR (150-400 m) stays rather constant, except for the decrease between E-1 and E-3 ( Table 2). As reported above such a mismatch may result from differences in timescales characterizing the different processes that were compared. The most likely explanation is that in this early stage of the growth season, phytoplankton biomass accumulated in the surface layer and export lagged behind. The ratio of EP to PP vs. T400 and T800 showed a large variability in transfer efficiency inside the meander (Fig. 7b). PP and EP fluxes increased by about 30 % from E-1 to E-3, but a concomitant decrease of mesopelagic MR yielded to an enhanced transfer efficiency, from 74 to 92 %, through the 400 m boundary, and from 52 to 73 % through the 800 m boundary. This suggests that significant remineralization should have occurred at greater depths (even > 1000 m), and it is also reflected by the presence of Ba xs maxima below 1000 m (see Fig. 4h and Appendix A). This was particularly salient when plotting Ba xs contents vs. depths over the 27day observation period (Fig. 8). The high, deep water Ba xs values in Fig. 8 were not taken into account when integrating TNS-6 and E-1 profiles between 150 and 400 or even 150 and 800 m (Fig. 5e). Considering that the seafloor in the meander area is at about 2000 m depth, it seems unlikely that these high Ba xs contents at depths > 1000 m were due to sediment resuspension. Also, particle spectra for these sites do not reveal any bottom resuspension (Jouandet et al., 2014;van der Merve et al., 2015). Therefore, the high, deep (> 1000 m) Ba xs contents at TNS-6 and E-1 most likely reflects the fact that here significant remineralization of POC material actually did occur in the bathypelagic domain and even down to the seafloor. Note that suspended particles in the depth range containing the deep Ba xs maxima were dominated by the < 2 µm size fraction (M. Zhou, personal communication, 2014). When integrating the Ba xs contents from 150 m to the seafloor at stations TNS-6 and E-1, MR fluxes increase to 156 and 184 mg C m −2 day −1 , respectively. Such C fluxes were similar to the EP values (maximum value of 130 mg C m −2 day −1 at E-3) and suggested that the exported Biogeosciences, 12, 1713-1731, 2015 www.biogeosciences.net/12/1713/2015/ POC was entirely remineralized in the water column leaving no C for transfer to the sediments. Overall, the temporal pattern of mesopelagic remineralization described above reflects two successive events of particle transfer: a first transfer from a previous bloom (occurred before visiting TNS-6 and enduring at E-1) and a second transfer from E-4E to E-5. The first transfer was evidenced by the downward (to the bottom) propagation of the mesopelagic Ba xs maximum signal, which mostly weakens at E-2. The second event was reflected by the occurrence again of important mesopelagic Ba xs build-up at E-4E and E-5. Overall, our results indicated the large capacity of the polar front meander to transfer POC material to depth, but in contrast to station A3 on the plateau, this transfer was coupled to intense and near-complete POC remineralization (as also observed at E-4W and F-L). Between-site changes in mesopelagic carbon remineralization due to unequal biomass productivity and iron fertilization over the Kerguelen Plateau were thus relatively complex. Furthermore, the conditions in the meander area seems to corroborate results obtained in the iron-replete Subantarctic Zone east of the Tasman Plateau (Australian sector of the Southern Ocean; SAZ-Sense cruise; Jacquet et al., 2011a, b), where the mesopelagic remineralization efficiency reported was relatively high (on average 91 %) and the deep (> 600 m) carbon transfer weak (< 10 %). Finally, the important Ba xs contents reported between 1000 and 2000 m during the first stages of the meander time-series support recent results indicating for the Southern Ocean that 1000 m is insufficient as an ocean-wide reference for carbon transfer and sequestration potential (Robinson et al., 2014).