The study site is defined around the Kerguelen Plateau, from 64 to 75∘ E, and between 47 and 51∘ S (Figs. 1 and 2). This region includes various biogeochemical systems, from the typical High Nutrients Low Chlorophyll (HNLC) waters generally found south of the Polar Front (PF), to the highly productive waters over and downstream of the Plateau (Fig. 1). The most extensive survey dedicated to understanding the functioning of this iron fertilized environment was conducted during the KEOPS2/OISO20 cruise, when physicists, biogeochemists and biologists were brought together to investigate the mechanisms at play in different regions of the bloom. The previous KEOPS/OISO12 cruise, conducted in January–February 2005, involved as much expertise, but it was limited to the southern part of the bloom situated over the Plateau. Below we describe the sampling and measurement techniques followed during the KEOPS2/OISO20 cruise, which are very similar to those followed during the first KEOPS survey (Jouandet et al., 2008), and during all OISO cruises (Metzl et al., 1999; Jabaud-Jan et al., 2004; Metzl, 2009; Lourantou and Metzl, 2011).

The KEOPS2 survey was conducted onboard R.R.V. Marion Dufresne in 2011, from the 17 October, when the ship approached the Kerguelen Plateau on its western side, to the 21 November, when the ship headed north of Kerguelen Island (see cruise track in Figs. 1 and 2). Our work is mostly based on hydrological and biogeochemical properties measured underway in surface waters (∼5 m depth). Underway surface measurements include the fugacity of CO2 (fCO2), sea surface temperature (SST), sea surface salinity (SSS) and fluorescence. In addition, atmospheric CO2 was measured every 4 h, and surface water samples were colleted every 4 h for total CO2 (TCO2), total alkalinity (TA), chlorophyll a and nutrients concentrations and every 8 h for salinity.

The technique for fCO2 measurements was described in details by Poisson et al. (1993), Metzl et al. (1995, 1999) and Jabaud-Jan et al. (2004). In short, sea surface water is continuously equilibrated using a “thin film” type equilibrator thermostated with surface seawater. The CO2 in the dried gas is measured with a non-dispersive infrared analyser (Siemens). The analysis of standard gases used for calibration indicated an accuracy to ±0.7 µatm. All fCO2 values presented here are normalized to 1013 hPa. TCO2 and TA were measured onboard, using a potentiometric method with a closed cell, and Certified Referenced Materials (CRMs, batch 111, provided by A. Dickson, SIO, University of California). Accuracy estimated from the CRMs analysis was ±4 µmolkg-1 for both TA and TCO2. SST and SSS were obtained from two Sea-Bird thermosalinographs. SSS was checked against measurements performed onboard using a salinometer (Park et al., 2014). Accuracy for SSS was evaluated to ±0.005. Nutrients were also measured directly onboard using a Skalar auto-analyzer (Blain et al., 2014). Accuracy was determined by measuring certified standards (CERTIPUR, Merck): it was 3.5 % for nitrate and 2.2 % for silicic acid (Si(OH)4) for concentrations around 35 µmolkg-1. Underway chlorophyll a measurements were obtained by filtering between 1 and 2 L of seawater on 25 mm GF/F Whatman filters (pressure <200 mbar). Filters were stored at -80 ∘C until measurements were performed at LOCEAN using a Hitaschi F-4500 spectrofluorometer, after extraction of the filters in 90 % acetone, following the method described by Neveux and Lantoine (1993). The precision of chlorophyll a measurements is typically around 5 %. We used chlorophyll a measurements to calibrate underway fluorescence data, but due to the poor correlation obtained (r2=0.5), calibrated fluorescence data must be considered with care.

Water samples were also collected at depth using a rosette equipped with 22 Niskin bottles and a Seabird CTD (Conductivity-Temperature-Depth) sensors. We used the data obtained between 0 and 20 m to complement underway measurements, with the exception of chlorophyll a data as it was not obtained using the same technique (high performance liquid chromatography) as for underway surface data. CTD (Conductivity-Temperature-Depth) profiles were also obtained at all stations, allowing to evaluate the mixed layer depth. Most of the stations are situated along the north–south and east–west transects (along ∼72∘ E and ∼48.5∘ S), that cover the southern and eastern sides of the Plateau and the plume offshore (Fig. 1). Here we present measurements collected at the station A3 situated over the Plateau in the core of the southern bloom investigated during the first KEOPS survey, and which was revisited twice during the KEOPS2 survey and again during the following OISO cruises.

The KEOPS2 survey started in October, when surface chlorophyll a concentrations were still low in most of the studied area due to light limitation (Fig. 1a). As a consequence, surface fCO2 was close to equilibrium with atmospheric CO2 (388 µatm) or higher, with the exception of shallow coastal waters that acted as a large sink for atmospheric CO2, with surface fCO2 values as low as 300 µatm (Figs. 2a and 3a). South-east of Kerguelen Island, where the plateau deepens to more than 500 m, we observed a large region of CO2 outgassing (ΔfCO2 around +20 µatm) associated with high surface salinity (Fig. 2e). High surface fCO2 values are representative of winter conditions in the Southern Ocean when vertical mixing maintains high concentrations of TCO2 in surface waters, therefore acting against the effect of cooling (e.g. Metzl et al., 2006). A smaller area of CO2 outgassing was also observed on the warm side of the Polar Front (PF) zone before the onset of the bloom. On the contrary, surface fCO2 measured in October on the cold side of (within) the PF meander was close to equilibrium with atmospheric CO2.

Weekly satellite images of chlorophyll a show an early phytoplankton bloom in the warm waters of the PF zone east of 73∘ E at end of October. It was rapidly followed by an increase in chlorophyll a concentration in the cold waters over and downstream of the Plateau, such that early in November most of the study region acted as a sink for atmospheric CO2, with surface fCO2 values ranging from 275 to 390 µatm. The minimum and maximum fCO2 values were obtained in shallow waters (<500 m). Very low fCO2 values, down to 100 µatm below the atmospheric level, were also measured on the warm side of the PF zone (east of 73.5∘ E), whereas within the PF meander (71–73.5∘ E) the mean ΔfCO2 was around -30 µatm. The region over the Plateau south of Kerguelen Island, that acted as a source of CO2 in October, also turned into a relatively large sink in November, with ΔfCO2 values ranging from -80 to -30 µatm (Figs. 2a and 3a).

In order to identify the mechanisms responsible for the evolution of surface fCO2 observed at the onset of the phytoplankton bloom, we analyzed in more details the underway surface measurements of fCO2 and other relevant parameters collected in different regions of the bloom. Below, we first present observations collected offshore (bathymetry >1000 m) downstream of the Kerguelen Plateau, a region referred to as the Plume, and which includes the cold waters within the PF meander (71–73.5∘ E, hereafter the Middle zone), as well as the warmer waters found in the PF zone (north of 47.4∘ S and east of 73.5∘ E). Next we present observations collected over the Plateau (iron source) south of the PF. We conclude the section by discussing the mean air–sea CO2 fluxes calculated in the different regions in October and November.

The survey downstream of the Kerguelen Plateau started on the 21 October, when chlorophyll a concentrations were still low and surface fCO2 was close to equilibrium with atmospheric CO2 (Figs. 1a and 2a, Table 1). Surface and water column measurements were collected along the north–south transect at ∼72∘ E, starting over the Plateau (station A3), covering the western part of the Plume (north of 49.5∘ S), and crossing the Polar Front at his northernmost position (47.4∘ S, Fig. 2). Surface fCO2 measurements obtained in the Plume show fluctuations around the equilibrium value with a maximum amplitude of 45 µatm (Fig. 3a). Because of low chlorophyll a concentrations, surface fCO2 is expected to vary with temperature and mixing. It is clear from Fig. 3 that the high surface fCO2 (small outgassing) observed between 48 and 49∘ S is related to high SST. However, less than half of the spatial variability in surface fCO2 can be explained by the effect of temperature alone (∼15 µatm∘C-1 according to Takahashi et al., 1993). Consequently, more than half of the observed variability is due to the higher TCO2 concentrations measured in the area of CO2 outgassing compared to the surrounding surface waters, which must be explained by lateral and/or vertical mixing. The rapid variations in SST observed within the PF meander reflect the re-circulation of southern (cold) and northern (warm) surface waters (Park et al., 2014). However, since the contribution of northern waters is expected to reduce surface TCO2 concentrations in the Plume compared to the cold southern waters, we conclude that lateral advection has a minor impact on surface fCO2 (due to the opposite effects of temperature and TCO2). Instead, our observations point to vertical mixing as the main driver for the observed spatial variations in surface fCO2. Indeed, the CO2 outgassing observed between 48 and 49∘ S is related to relatively deep layers (80–110 m), while the small uptake of CO2 observed in the colder waters around 49∘ S (Fig. 3) can be attributed to the combined effect of reduced mixing (60–70 m) and low but significant chlorophyll a concentrations (Table 1).

The passage of the PF was marked by a rapid change in surface temperature from ∼2.5 to 3.9 ∘C between 47.5 and 47.4∘ S (Figs. 2 and 3). This resulted in a rapid increase in surface fCO2 from near-equilibrium values south of the PF (387±7 µatm) to a small source of CO2 north of the PF (406±12 µatm), which can be explained by the temperature effect alone (+22 µatm). In the warm waters of the PF zone, we recorded the maximum fCO2 value of 430 µatm at 47.3∘ S, followed by a sudden decrease down to near-equilibrium, and a second fCO2 maximum at 46.8∘ S (Fig. 3a). Such rapid changes in surface fCO2 reflected the impact of mesoscale features associated with the frontal structure (Park et al., 2014).

The second visit of the Middle zone occurred 10 days later (31 October to 2 November) when both surface and water column measurements were collected along the east–west transect along ∼48.5∘ S. At that time the whole region had turned into a small sink for atmospheric CO2 of the order of 10 to 20 µatm (Fig. 4a). This decrease in fCO2 resulted from the opposite effects of surface warming by ∼0.7 ∘C (+11 µatm) and a decrease in TCO2 concentrations by ∼10 µmolkg-1 (Table 1). The latter could partly be due to a small increase in chlorophyll a concentration observed over the 10 days period. Both the changes in TCO2 and chlorophyll a concentrations are likely related to the reduction in vertical mixing revealed by reduced mixed layer depths (40–90 m) and a decrease in surface salinity (Table 1). We observed a larger CO2 sink in the western (cold) part of the Plume (∼20 µatm) compared to its eastern (warm) part (∼10 µatm), which is explained by both temperature and TCO2 concentrations. The low surface fCO2 and TCO2 measured in cold waters were associated with high surface salinity and low Si(OH)4 concentrations, which could reflect an increased biological activity in the waters of southern origin (consistent with observations over the Plateau, see Sect. 3.3).

In the warm waters of the PF zone, the mixed layer was already relatively shallow at the end of October (<80 m, Table 2). Ten days later it was less than 40 m deep east of 73.5∘ E, and a large area of intense biological uptake of CO2 was observed, with surface fCO2 values as low as 290 µatm (Fig. 4a) and chlorophyll a concentrations between 2 and 6 µgL-1 (Table 1). The reason for the early onset of an intense bloom in this area is not yet fully understood. One mechanism likely involved is the fast delivery of iron from the Plateau north of Kerguelen Island thanks to the strong jets associated with the frontal structure in this region (e.g., Lourantou and Metzl, 2011 ; Park et al., 2014), since we can expect that the amount of iron lost during the transit (due to biological and physical processes) is reduced. In the Middle region, the first important drawdown of fCO2 (down to 350 µatm) was observed on the 3 November over a small area (72.8–73∘ E). Two days later, we measured a very similar CO2 sink at the same latitude, but covering a larger area (72.9–73.4∘ E). This area was characterized by warmer SST (∼3.5 ∘C) than the surrounding waters (∼3.0 ∘C, Fig. 4), which may reflect reduced vertical mixing and/or the incursion of warm waters from the PF zone.

By the 8 November most of the area investigated downstream of the Plateau acted as a moderate to large sink for atmospheric CO2. In the Middle zone, the mean CO2 sink was of the order of 30 µatm, and was associated with a doubling in chlorophyll a concentrations since the last visit three days before (Table 1). The effect of enhanced photosynthesis on surface fCO2 was partly compensated by a warming by ∼0.3 ∘C on the eastern part of the Middle zone (72.5–73.6∘ E). The concomitance of surface warming and increased biological uptake of CO2 is attributed to reduced vertical mixing (<80 m). In the eastern part of the Plume, surface fCO2 remained low until the end of the survey, whereas large and rapid spatial variations were observed in surface fCO2 measured in the western part (71–72.5∘ E, Fig. 4a). In the region between 72 and 72.3∘ E, where temperature remained relatively low all through November (around 3 ∘C), we observed an increase in surface fCO2 to near-equilibrium with atmospheric CO2 on the 10 November, followed by the re-establishment of the CO2 sink, and a second increase to near-equilibrium four days later. Similar changes in surface fCO2 were observed west of 72∘ E, but the driving mechanisms were somewhat different. Indeed, the increase to near-equilibrium observed on the 18 November was due to the combined effect of surface warming by ∼0.5 ∘C (delayed by ∼10 days compared to the eastern part of the Plume) and reduced chlorophyll a concentrations (Table 1).

Low fCO2 values (down to 320 µatm) were measured at the end of November where the PF reaches its northernmost position. These extreme values, however, were very localized. The mean fCO2 in the PF zone at the end of November was 360 µatm, with maximum values of 380 µatm.

The southern part of the bloom, observed to the south-east of Kerguelen Island, where the Plateau deepens to 500–1000 m, was investigated in details during the first KEOPS cruise in January–February 2005. These previous measurements, obtained during the declining phase of the bloom, showed a core of maximum chlorophyll a concentrations (>1 µgL-1) between 50 and 51∘ S at ∼72∘ E (station A3), associated with minimum surface fCO2 values (58±11 µatm lower than in the surrounding HNLC waters, Blain et al., 2007). This area was revisited three times during the KEOPS2 survey (Table 1, Fig. 3).

In October, we measured high surface fCO2 in the Plateau region (50–51∘ S, 70–72∘ E), about 25 µatm above atmospheric CO2, whereas in HNLC waters upstream of the Plateau, surface fCO2 was near-equilibrium (Fig. 2a, Table 1). The CO2 outgassing observed over the Plateau is clearly related to the deep mixing observed in this area (MLD >140 m, Fig. 2) compared to HNLC waters upstream (Table 2), leading to higher TCO2 and nutrients concentrations over the Plateau (Table 1). Surface temperature was slightly lower over the Plateau than in the waters upstream, and chlorophyll a concentration was slightly higher. This suggests that both temperature and primary production partly compensate for mixing with the CO2-rich subsurface waters.

The combined effect of vertical mixing, temperature and photosynthesis on surface fCO2 was also observed between the first visit over the Plateau on the 18 October and the second visit on the following day, when the mean surface fCO2 was reduced by 13 µatm and surface waters were cooler by ∼0.2 ∘C (Fig. 3, Table 1). Sea surface cooling likely resulted from the low air temperature, combined with reduced vertical mixing, as suggested by water column measurements collected on the 20 October that showed an increase in temperature from the surface to the base of the mixed layer (by ∼0.4 ∘C, Fig. 5). According to Takahashi et al. (1993), the temperature effect explains about 25 % of the sudden decrease in surface fCO2. The remaining 75 % is attributed to enhanced photosynthesis. Indeed, water column measurements revealed that TCO2 concentrations were about 15 µmolkg-1 lower at the surface than in the Winter Water layer found between 175 and 200 m (subsurface temperature minimum, Fig. 5). This signal supports the idea that the biological pump of CO2 was already active at the end of October, despite the deep mixed layer. This is further supported by a small increase in surface chlorophyll a concentrations observed between 18 and 21 October (Table 1), which may hide a larger increase in subsurface waters (chlorophyll maximum).

Surface measurements were collected three weeks later (8–9 November) in the northern part of the Plateau region (49.5–50∘ S, Fig. 3, Table 1). SST was higher than during the first visit (+0.5 ∘C) due to warming in the atmosphere, and SSS was slightly lower, both suggesting reduced vertical mixing. Surface fCO2 decreased by ∼90 µatm over the 3-weeks period, in response to an increase in chlorophyll a concentrations by a factor 10 (>2 µgL-1) and possibly reduced mixing with subsurface waters. This resulted in a strong CO2 sink (∼60 µatm) and minimum values in surface TCO2 and nutrients. At this time, when the bloom was growing, surface chlorophyll a concentrations were twice as large over the Plateau than in the Plume offshore, and the CO2 sink was twice as strong.

One week later (15–17 November), the last visit over the Plateau revealed an increase in surface fCO2 by ∼25 µatm and reduced spatial variability (Fig. 3a), associated with sea surface cooling and mixed layers as deep as during the first occupation in October (∼150 m, Table 2). The salinity profile was also very similar, while temperature increased by ∼0.5 ∘C in the mixed layer due to warming in the atmosphere (Fig. 5). The concentration in chlorophyll a was quite homogeneous in the mixed layer, reflecting an effective mixing at the time of observation. This is supported by strong winds measured on the 15 November (15–20 ms-1), and a rapid slow down on the following day (∼10 ms-1). Our observations suggest that this short event of strong winds induced an increase in surface fCO2 of the order of 30 µatm, followed by a decrease of about 10 µatm when winds are reduced (although SST increased by ∼0.2 ∘C). Despite this deep mixing event, the Plateau region acted as a substantial sink for atmospheric CO2, relatively steady at the end of the survey (39±6 µatm), whereas in the Plume downstream of the Plateau, we observed large and rapid spatial and temporal variations in surface fCO2 resulting in a moderate CO2 sink at the end of the survey (26±10 µatm).

Air–sea CO2 fluxes were generally small in October (Table 2). The largest flux was an outgassing of CO2 observed over the Plateau south of Kerguelen Island, which we attributed to deep mixing (140–180 m). Based on our observations, we can expect substantial variations in air–sea CO2 fluxes in this region due to rapid changes in winds and the subsequent stratification/destratification of the upper ocean. In the HNLC region upstream of the Plateau, we calculated a mean CO2 flux close to zero, associated with shallower mixed layer (80–150 m). In the Middle zone, where the mixed layer was even shallower (60–110 m), the mean CO2 flux was also close to zero before the onset of the bloom. This resulted from the combined effects of lateral and vertical mixing on SST and TCO2 leading to a small outgassing of CO2 in the warm waters, and low but significant biological activity that induced a small CO2 sink in the cold waters of southern origin. In the PF zone, north of the plume, we estimated a small outgassing of CO2 at the end of October, mainly explained by the temperature effect (warm waters). In shallow waters (<500 m), our observations indicate a mean air–sea CO2 flux close to zero in October, mostly due to small ΔfCO2 values and the combination of CO2 source (salty waters) and CO2 sink areas. In the few places where we measured low fCO2, notably inside the Morbihan Gulf (SSS <33.5, Fig. 2), the uptake of CO2 was of the order of 30 mmolCm-2day-1 for observed winds around 15 ms-1.

In November, the fertilized region over and downstream of the Plateau was globally a large sink for atmospheric CO2, as can be expected from chlorophyll a concentrations (Fig. 1). The largest uptake of atmospheric CO2 was observed in the PF zone (∼20 mmolCm-2day-1) and was associated with maximum chlorophyll a concentrations (Table 1). The mean CO2 uptake estimated over the Plateau was twice as low than in the PF zone, but slighlty larger than in the Middle zone (Table 2). However, the mean fCO2 drawdown observed from October to November was three times larger over the Plateau (60 µatm) than in the Middle zone (20 µatm). We do not have observation in the PF zone east of 73.5∘ E before the onset of the bloom, but we can expect that surface fCO2 was close to equilibrium with atmospheric CO2 (similarly as what we observed in the PF zone north-east of Kerguelen Island and in the eastern part of the Middle zone). This would give a mean fCO2 drawdown around 80 µatm, i.e., four times larger than in the Middle zone.

In shallow waters (bathymetry <500 m), air–sea CO2 fluxes were highly variable in November due to the large variability in surface fCO2. In coastal waters, the mean CO2 uptake was as large as in the PF zone (∼20 mmolCm-2day-1).