High variability of export fluxes along the North Atlantic 1 GEOTRACES section GA01: Particulate organic carbon export 2 deduced from the 234 Th method 3

. In this study, we report Particulate Organic Carbon (POC) export fluxes estimated using the 234 Th-based 16 approach in different biogeochemical basins of the North Atlantic, as part of the GEOTRACES GA01 expedition 17 (GEOVIDE, May-June 2014). Surface POC export fluxes were deduced by combining export fluxes of 234 Th with the 18 POC to 234 Th ratio of sinking particles at the depth of export. Particles were collected in two size classes (>53 µm and 19 1-53 µm) using in-situ pumps and the large size fraction was considered as representative of sinking material. Surface 20 POC export fluxes revealed latitudinal variations between provinces ranging from 1.4 mmol C m -2 d -1 in the Irminger 21 basin where the bloom was close to its maximum peak, to 12 mmol C m -2 d -1 near the Iberian Margin where the bloom 22 had already declined. In addition to the bloom staging, the variations of POC export fluxes were also related to the 23 phytoplankton community structure. In line with previous studies, the presence of coccolithophorids and diatoms 24 appeared to increase the POC export flux while stations dominated by pico-phytoplankton cells, such as cyanobacteria, 25 were characterized by lower fluxes. The surface POC export fluxes were then compared to in-situ and satellite primary 26 production (PP) in order to assess the export efficiency. This ratio strongly varied regionally and was generally low (≤ 27 14 %), except at two stations located near the Iberian margin (35%) and within the Labrador basin (38%), which were 28 characterized by unusual low in-situ PP. We thus conclude that the North Atlantic during this period was not as efficient 29 in exporting carbon from the surface, as described in recent studies. Finally, we estimated the flux of POC exported 30 100 m below the surface export depth in order to investigate the transfer efficiency along the section. This parameter 31 was also highly regional-dependent but the lowest attenuation of the POC flux was observed at stations where 32 coccolithophorids dominated.

). In addition, deep samples 127 (n=15; between 1000 and 3500 m) were taken for the calibration of low level beta counting (van der Loeff et al., 2006). overnight, mounted on nylon holders, covered with Mylar and aluminum foil and 234 Th activity counted using low 136 level beta counters (RISØ, Denmark). Beta activity counting was continued until a relative standard deviation (RSD) 137 ≤ 2% was reached. At the home-laboratory, residual beta activity was measured for each sample after a delay of six 138 234 Th half-lives (~6 months) and these residual counts were subtracted from the gross counts obtained on-board. All

146
which is within the range indicated by Pike et al. (2005). The total 234 Th recovery, involving all the steps described 147 above, was 91 ± 14 % (n=200). Uncertainty on total 234 Th activity was estimated using error propagation and accounts 148 between 0.04 and 0.10 dpm L -1 .

149
The 238 U activity was deduced from salinity using the

156
Two size fractions of particles were thus collected: the small size fraction (referred to as SSF hereafter, 1-53 µm) and 157 the large size fraction (referred to as LSF hereafter, >53 µm). Filters were cleaned prior to the cruise as follows: PETEX 158 screens were soaked in 0.6M HCl, (Normapur, Merck) rinsed with Milli-Q water, dried at ambient temperature in a 159 laminar flow hood and stored in clean plastic bags; QMA filters were pre-combusted at 450 °C during 4 h and stored 160 in aluminum foils until use. In-situ pumps were deployed on a stainless steel cable between 15 and 800 m and the 161 pumping time was approximatively 2-3 hours (Table S2).

162
After collection, filters were processed on board. The 142 mm PETEX screen was cut into quarters using a clean 163 scalpel and two quarters were processed in this study. Large particles collected on the screen were rinsed off using 164 0.45 µm filtered seawater and re-filtered under a laminar flow hood on a silver filter (SterliTech, porosity=0.45 µm, 165 diameter=25 mm) for the first quarter of the PETEX screen and on a GF/F filter (Whatman®, porosity=0.7 µm, 166 diameter=25 mm) for the second quarter.

167
The QMA filters were sub-sampled with a perspex punch of 25 mm diameter.

168
Then, silver, GF/F and QMA filters were dried overnight (50 °C) and prepared for beta counting (see Section 2.2).

169
After counting the residual beta activity (~ 6 months later), samples were prepared for POC, PN analyses along with where ATh is the activity of total 234 Th in dpm L -1 ; AU is the salinity-derived activity of 238 U in dpm L -1 , λ is the 234 Th 181 decay constant (0.0288 d -1 ); P is the net removal of 234 Th on sinking particles in dpm L -1 d -1 ; V is the sum of the 182 advective and diffusive fluxes in dpm L -1 d -1 .

183
Assuming steady state (constant total 234 Th activity with time) and neglecting the physical term V (Buesseler et al.,

184
1992), the net export flux of particulate 234 Th can be determined using the following equation: where Pz is the integrated flux of 234 Th from the surface to the depth z in dpm m -2 d -1 .
where AThp is the activity of particulate 234 Th (in dpm L -1 ); the scavenging flux J described above becomes here the

215
where Jz in dpm m -2 d -1 is the net integrated flux of dissolved 234 Th to the depth z. In our case, the calculation was 216 performed at the Eq depth for comparison with the 234 Th export flux.

217
In a similar way, Equation 7 is simplified to:

219
Under these conditions, the net flux of scavenging J (source term) is defined by two output terms, the export of     The complete dataset of total 234 Th, 238 U activities and the corresponding 234 Th/ 238 U ratios are presented in Table S1 248 and Figure 2 shows the depth profiles of total 234 Th and 238 U activities. A deficit of 234 Th relative to 238 U ratio indicates 249 a loss of 234 Th due to the export by particles (Buesseler et al., 1992;Cochran and Masqué, 2003 (Table 1). Similar fluxes were found at both integration depths with 273 differences smaller than 12%, except at Stations 32 and 51 where fluxes at Eq were 36 and 46% greater than those at 274 the base of the PPZ. Considering that there can be export (or remineralisation) below or above the PPZ depth, in the 275 following, only the export fluxes at the Eq depth are discussed as they represent the fully-integrated depletion of 234 Th 276 in the upper waters and thus the maximal export.

277
The highest 234 Th export fluxes at Eq using the SS model were observed in the west European and Icelandic basins,  Table 1).

280
Using the SS model, we also determined the 234 Th scavenging fluxes at the Eq depth (Equations 8 and 9), along the 281 transect. These fluxes ranged from 1495 to 3917 dpm m -2 d -1 at Stations 38 and 21, respectively (Table 1). In general,   Table S2.

289
LSF particles were collected on silver and GF/F filters (see Section 2.2), and POC concentrations and 234 Th activities          Table 2) in order to compare both estimations.

342
Except at Stations 1, 26 and 64, the POC fluxes were between 1.1 to 1.5 folds higher when using the SSF ratio.  (Table 3).

350
The POC export fluxes at Eq using the LSF ranged from 1.   the PP varied by a factor of 6, ranging from 27 ± 5 at Station 69 to 174 ± 19 mmol C m -2 d -1 at Station 26 (Table 3).

360
Low PP were determined in the Iberian basin, with one of the lowest values measured at Station 1 (33 mmol C m -2 d -

363
Similarly, the Station 32, within the Icelandic basin was highly productive with a PP reaching 105 ± 11 mmol C m -2 d -364 1 but Station 38 was characterized by a lower PP (68 ± 7 mmol C m -2 d -1 ). Within the subpolar area, the PP was high 365 in the Irminger basin, ranging from 137 ± 2 to 166 ± 32 mmol C m -2 d -1 at Stations 44 and 51, respectively, but the PP 366 was lower in the Labrador basin, ranging from 27 ± 5 to 80 ± 21 mmol C m -2 d -1 at Stations 69 and 77, respectively.

367
In addition to incubation data, we looked at the annual record of satellite-derived PP in order to document the recent 368 trend in the biological production before the cruise. 8-days averaged PP data for the year 2014 are shown in Figure 4.

379
Using the 8-day average data, PP was estimated for the preceding month (32 days) and the whole productive period 380 prior to the sampling date in order to account for different timescales in PP and to compare with export fluxes estimates 381 (Table 3 and

397
we explore quantitatively or qualitatively, the potential errors introduced in our calculation.

398
Lateral processes associated to high velocities currents and intense mesoscale activity are known to affect the 234 Th

421
Downwelling systems, such as the intense convection that occurred in the Labrador basin during the winter prior to 422 our sampling (Kieke and Yashayaev, 2015), are also prone to impact the 234 Th distribution. However, a strong vertical 423 advection would homogenize the 234 Th activities in the water column, which is not the case during our study (Fig. 2).

432
In conclusion, although considered to have limited or no impact on the measured 234 Th export fluxes, physical 433 processes should be kept in mind when interpreting these export fluxes.

444
In order to evaluate the potential error introduced by the SS approach, we have attempted to apply a NSS model.

513
where the PP remained rather low along the season (Fig. 4)

526
The west European basin

527
Relatively high POC export fluxes were observed at Stations 21 and 26, reaching respectively 4.8 ± 0.8 and 7.9 ± 5.0 528 mmol m -2 d -1 . The sampling was performed during the bloom, and the highest PP peak along the section was observed 529 at Station 21 ( Fig. 4 and 5) just before the sampling. Station 26 was also sampled after a first bloom peak, and these 530 prior-sampling and high PP peaks could have promoted these high exports. These stations were also characterized by 531 an important proportion of micro-phytoplankton communities which could also explain the high POC exports.

538
The Icelandic basin

539
One of the highest POC export flux along the transect was determined at Station 32, reaching 8.3 ± 0.5 mmol m -2 d -1 540 while the POC flux at Station 38 was 4.8 ± 0.4 mmol m -2 d -1 . Both stations were sampled during the productive period, 541 although the peak of the bloom was not yet reached (Fig. 4). Nevertheless, the PP at Station 38 remained rather low 542 along the season (Fig. 4)

567
Overall, POC exports varied strongly along the transect with a factor of 8.6 between the highest and the lowest POC 568 export flux. The magnitude of these fluxes seems to be dependent on the phytoplankton community structure and thus 569 on the particle composition and density (Buesseler, 1998;Francois et al., 2002;Honjo, 1996;Lam et al., 2011). The

578
In order to study the biological carbon pump in the North Atlantic, we used two parameters: the export efficiency 579 (ThE), which is calculated by dividing the POC export flux at Eq by the PP (Buesseler, 1998) and the transfer efficiency 580 (T100) which is calculated by the POC export flux at 100 m below Eq divided by the POC export flux at Eq (Fig. 8).

581
Note that the POC export flux at Eq+100 (Table 3)

605
This large range confirms that export efficiencies are highly variable with time and that the North Atlantic during the 606 period of our study seemed to behave like most of the highly productive areas of the world's ocean, with a low export 607 efficiency.

608
However, the ThE calculation is based on two parameters that are integrating processes over different time scales: 24 609 h for in-situ PP and several weeks for export. As a result of this temporal mismatch and due to the strong variability in 610 PP, ThE ratios were also estimated using the satellite-derived 8-day, 32-day and seasonal PP (

644
The carbon export fluxes varied by a factor ~ 9 along the transect highlighting an important spatial variation. In the 645 North Atlantic, some studies reported similar POC export estimates but some others determined much higher POC

659
iii) The phytoplankton community seems to impact the particle composition and density, which play a crucial

663
For most stations, the fraction of primary production that is exported from the surface zone (export efficiency) was 664 ≤14%, which is in agreement with the global ocean export efficiency (~10%; Buesseler, 1998). Export efficiency was 665 also inversely related to primary production, highlighting that the North Atlantic during our study seems to behave like 666 most of the highly productive areas of the world's ocean, with a low export efficiency. Finally, the fraction of POC