The 226 Ra–Ba relationship in the North Atlantic during GEOTRACES-GA01

. We report detailed sections of radium-226 ( 226 Ra, T 1 / 2 = 1602 years)

Abstract. We report detailed sections of radium-226 ( 226 Ra, T 1/2 = 1602 years) activities and barium (Ba) concentrations determined in the North Atlantic (Portugal-Greenland-Canada) in the framework of the international GEO-TRACES program (GA01 section -GEOVIDE project, May-July 2014). Dissolved 226 Ra and Ba are strongly correlated along the section, a pattern that may reflect their similar chemical behavior. Because 226 Ra and Ba have been widely used as tracers of water masses and ocean mixing, we investigated their behavior more thoroughly in this crucial region for thermohaline circulation, taking advantage of the contrasting biogeochemical patterns existing along the GA01 section. We used an optimum multiparameter (OMP) analysis to distinguish the relative importance of physical transport (water mass mixing) from nonconservative processes (sedimentary, river or hydrothermal inputs, uptake by particles and dissolved-particulate dynamics) on the 226 Ra and Ba distributions in the North Atlantic. Results show that the measured 226 Ra and Ba concentrations can be explained by conservative mixing for 58 and 65 % of the samples, respec-tively, notably at intermediate depth, away from the ocean interfaces. 226 Ra and Ba can thus be considered conservative tracers of water mass transport in the ocean interior on the space scales considered here, namely, on the order of a few thousand kilometers. However, regions in which 226 Ra and Ba displayed nonconservative behavior and in some cases decoupled behaviors were also identified, mostly at the ocean boundaries (seafloor, continental margins and surface waters). Elevated 226 Ra and Ba concentrations found in deepwater in the West European Basin suggest that lower Northeast Atlantic Deep Water (NEADWl) accumulates 226 Ra and Ba from sediment diffusion and/or particle dissolution during transport. In the upper 1500 m of the West European Basin, deficiencies in 226 Ra and Ba are likely explained by their incorporation in planktonic calcareous and siliceous shells, or in barite (BaSO 4 ) by substitution or adsorption mechanisms. Finally, because Ba and 226 Ra display different source terms (mostly deep-sea sediments for 226 Ra and rivers for Ba), strong decoupling between 226 Ra and Ba were observed at the land-ocean boundaries. This is especially true in the 1 Introduction The primary source of radium-226 ( 226 Ra, T 1/2 = 1602 years) to the ocean was found to be diffusion from deep-sea sediments following the decay of its parent isotope, 230 Th (Koczy, 1958;Kröll, 1953). This mode of introduction led Koczy to use radium data to derive vertical eddy diffusivities and velocities in the deep sea (Koczy, 1958). Since then, 226 Ra has been widely used to study the ocean circulation and mixing on a global scale (Chung and Craig, 1980;Ku et al., 1980). In the framework of the Geochemical Ocean Sections Study (GEOSECS) program, special attention was given to 226 Ra as its solubility and half-life made it an ideal candidate as a chronometer of the global thermohaline circulation. In particular, its 1602 years of half-life is more adapted than the longer half-life of carbon-14 ( 14 C, T 1/2 = 5730 years) that had also been used for that purpose. Therefore, the global oceanic distribution of 226 Ra in seawater was extensively documented throughout the major ocean basins and a unique database was generated during the GEOSECS program (Broecker et al., , 1967Chung et al., 1974;Ku et al., 1970;Ku and Lin, 1976). Barium (Ba) is an alkaline earth element like 226 Ra. Therefore they share a similar geochemical behavior in the ocean (Chan et al., 1976;Fanning et al., 1988;Mathieu and Wolgemuth, 1973). As such, Ba was proposed as a stable analog of 226 Ra in order to use the 226 Ra / Ba ratio as a clock in a similar manner as the 14 C / 12 C ratio. However, the recognition that 226 Ra and Ba participate in upper ocean biological cycles (Ku and Luo, 1994) introduced additional complications for the use of the 226 Ra / Ba ratio as a time tracer for deep water ventilation. Both 226 Ra and Ba indeed increase with increasing depth, thus reflecting uptake due to biological processes in surface waters, particle scavenging and subsequent release at depth through the dissolution of the settling particles (Broecker et al., 1967;Ku et al., 1970;Ku and Luo, 1994). 226 Ra and Ba are thus not only controlled by physical processes but appear to be incorporated in settling particles such as calcareous and siliceous shells, or in barite (BaSO 4 ), which precipitates in the mesopelagic zone (Bishop, 1988;Chan et al., 1976;Dehairs et al., 1980;Boyle, 1989, 1990). Hence, despite different principal sources to the ocean (rivers in the case of Ba and marine sediment diffusion for 226 Ra), their distributions are affected by similar processes in the water column. Barium displays a linear correlation with 226 Ra in the global ocean, resulting in a fairly constant 226 Ra / Ba ratio of 2.2 ± 0.2 dpm µmol −1 (dpm, disintegrations per minute) (Chan et al., 1976;Foster et al., 2004;Ku et al., 1980;Li et al., 1973;Östlund et al., 1987). Similarly, strong correlations were found between Ba-Si (silicate) and 226 Ra-Si, although no obvious process links 226 Ra or Ba with Si. The latter relationships appeared to be more surprising because Si is not a chemical analog of Ra and Ba. It was first proposed that diatom frustules exported from the upper water column could adsorb Ra and Ba, these elements being then released at depth following the dissolution of the siliceous tests (Bishop, 1988;Chung, 1980;Kumar and Li, 1996). More recent studies showed that the similar behaviors of Ba and Si (and alkalinity) reflect similar dissolved-particulate interactions McManus et al., 1999;Rubin et al., 2003). Indeed, Ba is not mechanistically coupled with alkalinity or silicate. Rather, the observed relationships may result from the spatial coherence of different carriers (barite, opal and carbonate) overprinted by hydrodynamics. The formation of biogenic silica, CaCO 3 and barite in surface water and their subsequent dissolution in the deeper water column may thus generate parallel oceanic distributions. While barite has been shown to be the main carrier that controls the Ba water column distribution, the relationship between Ba and Ra remains unclear.
While the global GEOSECS program provided valuable information on the coupling between biogeochemical cycles of 226 Ra and Ba in the ocean, several unknowns still remain. In this work, we take advantage of a new worldwide program, GEOTRACES, to provide new information on the distribution of 226 Ra and Ba and their relationship in the ocean. GEOTRACES program aims to characterize the distribution of trace elements and their isotopes (TEIs) (sources, sinks, internal cycling) in the ocean through a global survey consisting of ocean sections and regional process studies.
In the present study, we report dissolved 226 Ra activities and dissolved Ba concentrations in the North Atlantic Ocean and Labrador Sea (GEOVIDE project, GA01 section). The North Atlantic region hosts a variety of globally significant water masses with complex circulation patterns (García-Ibáñez et al., 2018;Lherminier et al., 2010). This area is crucial for the thermohaline circulation and thus for global climate, through its important role in the ventilation of the deep layer of the global ocean (Seager et al., 2002). As part of this process, the meridional overturning circulation (MOC) includes the northward transport of warm subtropical waters. These surface waters are then cooled and transformed into subpolar waters and may reach the Labrador and Irminger seas where deepwater formation and deep convection take place (Bennett et al., 1985;Pickart and Spall, 2007;Yashayaev et al., 2007). We propose to study the relationship between 226 Ra and Ba and to test the conservative behavior of these tracers in this specific region. We further document the Ra-Ba-Si relationship along the GA01 section, as it was done in previous sections conducted during the GEOSECS program and more recently along the GEOTRACES GA03 section.   Fig. 1). The water samples described here were collected on board the R/V Pourquoi Pas?. The section crossed different topographic features and regions with contrasting biogeochemical patterns. It complemented the sections GA03 (US-GEOTRACES) and GA02 (Dutch GEOTRACES), which were also conducted in the Atlantic Ocean in the framework of the GEOTRACES program. Seventy-eight stations were visited during the GEO-VIDE project.

Sample collection
At 15 of the 78 stations completed during the GA01 cruise, up to 22 discrete 10 L seawater samples were collected through the water column from Niskin bottles. Due to gravity, the seawater samples passed through 10 g of acrylic fibers impregnated with MnO 2 (called "Mn fibers"), which quantitatively adsorb radium isotopes (assumed to scavenge 100 % of Ra;van Beek et al., 2010;Moore and Reid, 1973). High-resolution vertical profiles of 226 Ra were thus built to provide a detailed 226 Ra section. The samples were unfiltered since particulate 226 Ra activities are typically 2 orders of magnitude lower than the dissolved 226 Ra activities (van Beek et al., 2007(van Beek et al., , 2009. From the same Niskin bottles, 15 mL was collected to determine the Ba concentration, so that Ba and 226 Ra analyses were conducted from the same initial sample, which allows us to investigate the 226 Ra / Ba ratio in the samples. The Ba samples were collected in pre-cleaned polypropylene bottles (rinsed three times with the same seawater sample), acidified with 15 µL of HCl (10 M, Merck, Suprapur) and kept at room temperature for later analysis.

Analysis of dissolved 226 Ra activities via 222 Rn emanation
Radium-226 was determined via its daughter, radon-222 ( 222 Rn; T 1/2 = 3.8 days) using a radon extraction system followed by alpha scintillation counting (Key et al., 1979). The Mn fiber samples were placed into gas-tight PVC cartridges (Peterson et al., 2009) that were flushed with helium (He) for 5 min at 250 mL min −1 . The cartridges were sealed and held for approximately 2 weeks (minimum of 5 days) to allow for 222 Rn ingrowth from 226 Ra decay. The 222 Rn was then flushed out from the cartridges using He and cryo-trapped in copper tubing using liquid nitrogen. The copper trap was heated to allow the 222 Rn to be transferred to evacuated "Lu-cas cells" via a stream of He. The Lucas cells are airtight chambers with inner walls coated with silver-activated zinc sulfide that emits photons when struck by alpha decay particles (Key et al., 1979;Lucas, 1957;Peterson et al., 2009). The cells were held for 3 h to reach the secular equilibrium of all 222 Rn decay chain daughters. After 3 h, the samples were counted overnight on a radon counting system (model AC/DC-DRC-MK 10-2). The counting uncertainties (1 SD, standard deviation) were within the range of 2-5 % for 10 L volume samples. All samples were appropriately ingrowth and decay corrected. The combined Lucas cell and detector background was ∼ 7 % of the typical total measured sample activity. The method was standardized using NIST (US National Institute of Standards and Technology) 226 Ra (20 dpm) sorbed onto MnO 2 fiber and analyzed in the same manner as the samples, with uncertainties (1 SD) of 5 % (Charette et al., 2015;Henderson et al., 2013). Vertical profiles of 226 Ra from the GEOTRACES GA01 (this study) and GA03 (Charette et al., 2015) sections that were located in close proximity off Portugal ( Fig. 1) were compared and showed a good agreement with increasing activities and increasing depth ( Fig. S1 in the Supplement).

Analysis of dissolved Ba concentrations
Barium concentrations were measured using an isotope dilution (ID) method (Freydier et al., 1995;Klinkhammer and Chan, 1990) by high-resolution inductively coupled plasma mass spectrometry (HR-ICP-MS). This method was adapted to a Thermo Finnigan Element XR instrument (MIO, Marseille). The Ba measurements presented here are the sum of dissolved Ba and a very small fraction of particulate Ba (generally < 1 % of total Ba, along GEOVIDE up to 1.3 % at the bottom of station 32 due to presence of a nepheloid layer; Lemaitre et al., 2018a) released from the samples as a result of the acidification step. Hence, while the measurements reported herein are total Ba, they are within analytical uncertainty representative of the dissolved Ba pool. The samples (0.5 mL) were spiked with 300 µL of a 135 Ba-enriched solution (93 % 135 Ba; 95 nmol kg −1 ) and diluted with 15 mL of acidified (2 % HNO 3 , 14 M, Optima grade) Milli-Q grade water (Millipore). The amounts of sample, spike and dilution water were assessed by weighing. The reproducibility of this method is about 1.5 % (1 RSD, relative standard deviation), as tested on repeated preparations of the reference solution SLRS-5 (NRC-CNRC river water reference material for trace metals). Average Ba values obtained for SLRS-5 were 13.48 ± 0.20 µg L −1 (1σ ) with RSD of 1.5 %, which is in good agreement with the certified values (SLRS-5 13.4 ± 0.6 µg L −1 ). The limit of detection calculated as 3 times the standard deviation of the procedural blank was 0.09 nmol L −1 .

Multiparameter mixing model
An optimum multiparameter (OMP) analysis was used to distinguish the relative importance of physical transport (i.e., water mass mixing) from nonconservative processes (input from the sediments, rivers or hydrothermal vents, dissolution of particles; uptake by particles and dissolved-particulate dynamics) on the 226 Ra and Ba distributions in the North Atlantic. We used the OMP analysis computed for the GA01 section by García-Ibañez et al. (2018) with 12 source-water types (SWTs). Based on historical data reported from the North Atlantic, we defined 226 Ra and Ba end-member concentrations associated with each SWT (Table S1 in the Supplement). The characteristics of SWTs (potential temperature, salinity and geographical location) reported by García-Ibañez et al. (2018) were used to determine the SWT endmembers for 226 Ra and Ba. In some cases, data from the GA01 section were used for the SWT end-member (Table S1). These 226 Ra and Ba SWT end-members were then used to calculate the 226 Ra and Ba concentrations that strictly result from mixing of the different water masses. In this way, we estimated the conservative components of 226 Ra and Ba, which can in turn be compared to the in situ concentrations used to generate the nonconservative component of 226 Ra and Ba along the GA01 section. The uncertainties for the SWT end-members were considered to be ±0.6 dpm 100 L −1 for 226 Ra and ±1.5 nmol L −1 for Ba, based on typical measurement uncertainties and spatial variability. A Monte Carlo method (250 random perturbations) was used to propagate the SWT end-member uncertainties to the 226 Ra and Ba results of the OMP analysis. This allowed us to estimate conservative component uncertainties of ±0.9 dpm 100 L −1 for 226 Ra and ±4.4 nmol L −1 for Ba. When the measured 226 Ra and Ba values were different from the conservative component values (taking into account the uncertainties on these values), 226 Ra and Ba were considered nonconservative. These nonconservative values can either be positive (representative of a net addition of 226 Ra and Ba) or negative (representative of a net removal of 226 Ra and Ba).
Note that the OMP analysis was not solved where nonconservative behavior of temperature and salinity is expected (that is, for waters above 100 m and for waters with salinities lower than 34.7). In the latter cases, changes in water mass properties may indeed be due to air-sea interaction or input of fresh water (e.g., near Greenland shelf; Daniault et al., 2011).

Hydrodynamic context
The OMP analysis was used to identify the different water masses (Table 1) crossing the GA01 section. The potential temperature-salinity diagram for all the GA01 stations along with the different SWT end-members used in the OMP analysis are represented in Fig. 2. The salinity section is shown in Fig. 3. The different water masses present along the GA01 section are described below.

Upper waters
Three main water masses were found in the upper waters (< 1000 m) of the investigated area (Fig. 3). First, the Central Waters occupied the upper eastern part of the GA01 section from the Iberian Peninsula to the Rockall Trough (stations 1 to 26). Their distribution was associated with the circulation of the North Atlantic Current (NAC). The NAC transports warm and saline waters northward, connecting the subtropical and the subpolar latitudes, and is part of the upper layer of the Atlantic meridional overturning circulation (AMOC) in the North Atlantic subpolar gyre. The NAC flows eastward from the Grand Banks of Newfoundland, splitting into four branches west of the Mid-Atlantic Ridge (MAR), while incorporating local water masses (Fig. 1). East of the MAR, the two northern branches of the NAC flow northward into the Icelandic Basin, the Rockall Plateau and the Rockall Trough, while the two southern branches flow southward into the West European Basin. The Central Waters can be identi-fied by the highest potential temperature of the entire GA01 section and are represented by two end-members called East North Atlantic Central Waters (ENACW 16 and ENACW 12 ). The ENACW 16 is warmer (16 • C) than the ENACW 12 , which can be identified with a potential temperature of 12.3 • C (Fig. 2).
Part of the Central Waters carried by the NAC recirculates toward the Iceland Basin and the Irminger Sea, leading to the formation of subpolar mode waters by mixing and cooling in the subpolar gyre (Lacan and Jeandel, 2004;Mc-Cartney, 1992). Iceland Subpolar Mode Water (IcSPMW) is formed in the Icelandic Basin and is located, along GA01, over the Reykjanes Ridge (stations 32 and 38) (Fig. 3). The IcSPMW is described by two end-members, the SPMW 7 and the SPMW 8 , which are distinguished by their potential temperatures of 7.0 and 8.0 • C (Fig. 2). Once formed, the Ic-SPMW follows the Irminger Current.

Intermediate waters
The Subarctic Intermediate Water (SAIW) originates in the Labrador Current (Read, 2000). The SAIW is associated with the advection of waters from the Labrador Sea within the NAC; it subducts below the Central Waters at approximately 600 m. Low salinities (34.8 and 34.7) and potential temperatures of 4.5 and 6 • C are representative of the two SAIWs, SAIW 4 and SAIW 6 ( Fig. 2).
Around the Rockall Plateau, the SAIW overlies the Mediterranean Water (MW). The MW enters the North Atlantic through the Gibraltar Strait and flows northward while extending westward. The MW can be identified in the West European Basin at approximately 1200 m (stations 1 and 13 in Fig. 3) with a salinity of 36.5 ( Fig. 2; García-Ibáñez et al., 2015).
The Labrador Sea Water (LSW) is found in multiple locations and at different water depths along the GA01 section (Fig. 3). The LSW is formed by progressive cooling and freshening in winter due to deep convection. The LSW can be characterized by its minimum salinity (34.87) and its minimum potential temperature (3 • C) (Fig. 2). The LSW contributes to the stratification of the interior of the North Atlantic and its boundary currents, and spreads at intermediate depths in three different basins intersected by the GA01 section (Fig. 1). The three independent pathways are (i) northward into the Irminger Sea (station 44); (ii) eastward across the MAR, through the Charlie-Gibbs fracture zone, then northward into the Iceland Basin (station 32) and eastward into the West European Basin (stations 21 and 26); and (iii) equatorward as a major component of the North Atlantic Deep Water in the Deep Western Boundary Current (DWBC), which constitutes the lower limb of the AMOC. Along these paths, the LSW mixes with both the overlying and underlying water masses and becomes warmer and saltier (Lazier, 1973).
The Polar Intermediate Water (PIW) is characterized by very low salinity (34.9) and potential temperature (less than 2 • C) ( Fig. 2) and is defined as a separate upper core on the Greenland slope. The PIW is episodically injected into the Irminger Sea and originates from either the Arctic Ocean or the Greenland shelf (Falina et al., 2012;Jenkins et al., 2015;Rudels et al., 2002).

Overflow waters and deepwater
The Iceland-Scotland Overflow Water (ISOW) originates at the Iceland-Scotland sill and entrains the overlying warm saline Atlantic waters (SPMW and LSW). ISOW identification features are a potential temperature of 2.6 • C and a salinity of 34.98 (Fig. 2;van Aken and Becker, 1996). ISOW was found at stations located on the eastern flank of the Reykjanes Ridge (stations 32 and 38) and near Greenland (stations 60 and 64) at great depth (2000-3500 m) (Fig. 3).
Overflow waters coming from the Denmark Strait mix with both the SPMW and the LSW during descent into the Irminger Sea to form the Denmark Strait Overflow Water (DSOW) ( Fig. 1) (Read, 2000;Yashayaev and Dickson, 2008). DSOW is located at the northern end of the Irminger Sea (station 44) and occupies the deepest part of the Greenland continental slope (stations 69 and 77) (Fig. 3). At bottom depth, DSOW is easily identified by a minimum potential temperature of 1.3 • C (Fig. 2).
In the Southern Ocean, the Antarctic bottom water (AABW) is formed by deep winter convection of surface waters. AABW flows to the north along the eastern side of the Atlantic and contributes to the formation of the lower Northeast Atlantic Deep Water (NEADWl) once this water penetrates the Iberian Abyssal Plain by crossing the Mid-Atlantic Ridge (Fig. 1). The NEADWl lies at the bottom of the West European Basin (stations 1 to 26 in Fig. 3) with a mean salinity of 34.895 and a potential temperature of 1.98 • C (Fig. 2). Then, the NEADWl recirculates into the Rockall Trough and Figure 4. Distribution of (a) dissolved 226 Ra activities (dpm 100 L −1 ), (b) dissolved Ba concentrations (nmol L −1 ) and 226 Ra / Ba ratio (dpm µmol −1 ) along the GA01 section. Station numbers are found above the panels. The sampling depths are shown for each vertical profile (black dots). meets ISOW in the Iceland Basin (van Aken, 2000;McCartney, 1992;Schmitz and McCartney, 1993).

Distribution of 226 Ra and Ba along the GA01 section
The 226 Ra distribution for the GA01 section is presented in Fig. 4 with Ba concentrations and 226 Ra / Ba ratios. The 226 Ra activities and Ba concentrations in the water column range from 7 to 25 dpm 100 L −1 and from 33.6 to 81.5 nmol L −1 , respectively. These data are in good agreement with Atlantic data from the GEOSECS program, which range from 6.8 to 23.4 dpm 100 L −1 for 226 Ra and from 35 to 105 nmol L −1 for Ba . Y=0.24x +7.3 ± 0.1 Figure 5. Relationships between 226 Ra and Ba (red dots) and between 226 Ra and Si(OH) 4 (blue dots) along the GA01 section in the North Atlantic. The best linear fit for the two plots is also reported (R = 0.93 for the two plots). The slopes of the relationships between 226 Ra and Ba and between 226 Ra and Si(OH) 4 are expressed in 10 −2 dpm nmol −1 and in 10 −2 dpm µmol −1 , respectively.
For both 226 Ra and Ba, the vertical gradient is stronger in the eastern part of the section (West European Basin) than on the western part of the section (from Reykjanes Ridge to Newfoundland). Both are particularly high near the seafloor in the West European Basin ( 226 Ra: 14-25 dpm 100 L −1 ; Ba: 63-82 nmol L −1 ) and are in agreement with data previously reported for this region Charette et al., 2015). At intermediate depths, Ba concentrations range from 40 to 50 nmol L −1 in the West European Basin (stations 1 and 21) and 226 Ra activities range from 9.5 to 10.9 dpm 100 L −1 , also in good agreement with literature data (Charette et al., 2015;Schmidt and Reyss, 1996). Low 226 Ra and Ba are found in the upper waters of the West European Basin and the Iceland Basin (8.1-8.9 dpm 100 L −1 and 35-43 nmol L −1 , respectively). Intermediate 226 Ra activities and Ba concentrations (9 dpm 100 L −1 and 42 nmol L −1 , respectively) are observed in bottom water in the Irminger and Labrador seas. Between the Reykjanes Ridge and Newfoundland, 226 Ra activities range between 7 and 10 dpm 100 L −1 in surface and intermediate waters. Similarly to 226 Ra, Ba concentrations are relatively low in this area, ranging from 39 to 50 nmol L −1 . The distributions in 226 Ra and Ba are explained to a first order by the different water masses present in the region, as discussed below.

226 Ra-Ba and 226 Ra-Ba-Si relationships
A linear correlation between 226 Ra and Ba is observed for all data collected along the GA01 section (Fig. 5). The slope of the 226 Ra-Ba linear regression obtained by this study in the North Atlantic is 2.5 ± 0.1 (2 SD) dpm µmol −1 , which agrees with the slope of the 226 Ra-Ba linear regression of 2.3 dpm µmol −1 reported during the GEOSECS program for all the oceanic basins (Chan et al., 1976). The intercept on the horizontal Ba axis is 4.4 nmol L −1 for the GA01 section, which is in agreement with the GEOSECS data (Chan et al., 1976;Li et al., 1973). This positive intercept may be the result of a larger riverine Ba input relative to 226 Ra (Ku and Luo, 1994). The slope of the 226 Ra-Ba linear regression reported from the GEOSECS program is similar from one oceanic basin to another. The 226 Ra / Ba ratio (slightly different from the slope) is also fairly constant throughout the global ocean (2.2 ± 0.2 dpm µmol −1 ; Östlund et al., 1987). This pattern indicates that 226 Ra and Ba may behave similarly in the ocean. Since 226 Ra and Ba are incorporated in settling particles such as calcareous and siliceous shells or barite (BaSO 4 ) by substitution or adsorption mechanisms (Bishop, 1988;Dehairs et al., 1980;Boyle, 1989, 1990) and are then released at depth following the dissolution of these particles, the constant 226 Ra / Ba ratio suggests that fractionation between 226 Ra and Ba during these processes is small.
Investigations conducted during the GEOSECS program further concluded that 226 Ra and Ba were tightly correlated with orthosilicic acid (Si(OH) 4 ) (Chan et al., 1976;Chung, 1980;Ku et al., 1970;Ku and Lin, 1976) despite the fact that 226 Ra, Ba and Si(OH) 4 exhibit different chemical behavior in the water column and different source terms. A Ra-Ba-Si relationship is also observed along the GA01 section (Fig. 5). Si(OH) 4 concentrations generally increase with increasing depth, with a steeper gradient in the West European Basin (Introduction Paper, 2017; This issue), as it was also the case for 226 Ra and Ba (Fig. S2). The link between 226 Ra, Ba and Si has been shown to reflect parallel dissolved-particulate interactions between barite and biogenic silica (Bishop, 1988;Chung, 1980;Jacquet et al., 2005Jacquet et al., , 2007Jeandel et al., 1996); the main carrier of 226 Ra in the ocean, however, remains an open question. The oceanic Ba-Si and 226 Ra-Si relationships may thus be the result of the interaction between ocean biogeochemistry and the water mass transport.
In contrast to the 226 Ra-Ba relationship, the slope of the 226 Ra-Si(OH) 4 relationship observed during GEOSECS exhibited significant spatial variability from one oceanic basin to another (Li et al., 1973). First, 226 Ra and Si are not chemical analogues, as it is the case for 226 Ra and Ba. Second, the variability observed in the 226 Ra-Si(OH) 4 relationship may be related to the large variability in the Si(OH) 4 distribution, which is mostly governed by the preformed nutrient concentrations of waters feeding into the main thermo-cline from surface waters of the Southern Ocean (Sarmiento et al., 2007). In the case of GA01, the 226 Ra-Si(OH) 4 linear regression slope is 2.4 ± 0.9 × 10 3 dpm mol −1 (2 SD). As a comparison, the 226 Ra-Si(OH) 4 slope reported for the GEOTRACES-GA03 section conducted south of the GA01 section in the Atlantic Ocean was 2.1 × 10 3 dpm mol −1 (Charette et al., 2015). The slope of the 226 Ra-Si(OH) 4 linear regression is 34.3 × 10 3 dpm mol −1 in the Pacific Ocean and 14.5 × 10 3 dpm mol −1 in the Antarctic Ocean. The 226 Ra-Si(OH) 4 relationship has an intercept with the vertical axis of 7.3 ± 0.1 dpm 100 L −1 , which represents the residual 226 Ra resulting from the total usage of Si in surface waters . According to (Shannon and Cherry, 1971), the removal of 226 Ra in the upper waters is limited by Si. For both the 226 Ra-Ba and 226 Ra-Si(OH) 4 relationships, several values are clearly outside of the linear regression trend (Fig. 5), a pattern that indicates deviation from the relationship usually observed between 226 Ra and Ba (or Si(OH) 4 ). Such deviations may result from nonconservative processes.

226 Ra and Ba distributions and their relationship with hydrography
A striking feature of the GA01 section is that the 226 Ra activities and Ba concentrations are particularly high in the West European Basin below 2000 m (Fig. 4), in the NEADWl. This pattern can also be observed in the GA03 section conducted south of the GA01 section (Charette et al., 2015), the two sections being separated by only ca. 500 km in that basin. The NEADWl is mainly formed from waters with a southern origin (Read, 2000). South of the Antarctic Convergence, the surface waters contain high 226 Ra activities from the upwelling of deepwater enriched in 226 Ra and associated with the circumpolar current (Ku and Lin, 1976). The convection of these surface waters leads to the formation of the 226 Rarich AABW, which circulates northward into the Atlantic Ocean. However, the high 226 Ra activities of the NEADWl cannot be solely explained by the high 226 Ra activities of these waters of southern origin. This will be discussed in Sect A slight increase in 226 Ra is observed between 1000 and 1600 m in the West European Basin (Fig. 4; Stations 1 and 13) where a salinity maximum is identified. This pattern is associated with the MW. This is corroborated by the slightly higher Ba concentrations and lower 226 Ra / Ba ratios, as ob-served in the Western Mediterranean Sea (van Beek et al., 2009), with these waters spreading westward into the Atlantic Ocean. At these stations, between 30 and 79 % of the water found at 1000-1600 m is of Mediterranean origin (MW), according to the OMP analysis (García-Ibañez et al., 2018;this issue).
Relatively uniform and low 226 Ra activities and Ba concentrations are found between the surface and 2500 m in the Labrador Sea, up to 2000 m in the Iceland Basin and deeper in the Irminger Basin (Fig. 4). These distributions can be related to the LSW, which is formed by winter convection in the Labrador Sea (Fröb et al., 2016;Pickart et al., 2003;Yashayaev and Loder, 2016). When formed, the LSW transports the characteristics of surface waters to the deep ocean (i.e., low 226 Ra activities and low Ba concentrations). The LSW then spreads into the Irminger and the Iceland Basin while conserving its low 226 Ra and Ba signatures. Relatively low 226 Ra activities and Ba concentrations are found at bottom depths in the Irminger and Labrador seas and may be associated with DSOW, which is also a recently ventilated water mass (Lazier, 1973).
Finally, according to the OMP analysis, ISOW is present at several stations along the GA01 section (García-Ibañez et al., 2018). First, on the eastern flank of the Reykjanes Ridge (station 32), 68 % of the water between 2700 and 3000 m is considered to be ISOW. Then, in the Labrador Sea (stations 69 and 77), an average of 58 % of the water between 2100 and 3000 m is identified as ISOW. The slight increase in 226 Ra activities and Ba concentrations observed at these locations may thus be related to the ISOW.

Conservative versus nonconservative behavior of 226 Ra and Ba
The use of an optimum multiparameter (OMP) analysis allowed us to distinguish the relative importance of physical transport (i.e., mixing of water masses) from nonconservative processes on the 226 Ra, Ba and 226 Ra / Ba ratios distributions in the North Atlantic (Fig. 6). The comparison between the vertical profiles of 226 Ra and Ba determined along the GA01 section, and those derived from OMP analysis (Fig. 7) clearly indicates deviations from the conservative behavior and reflects either an input of 226 Ra or Ba (positive anomalies highlighted in red; same color code as in Fig. 6) or a removal of 226 Ra or Ba (negative anomalies highlighted in blue; same color code as in Fig. 6). This comparison reveals that for 58 % of the samples 226 Ra can be considered conservative (activities due to mixing and transport) along the GA01 section (i.e., 58 % of the 226 Ra anomalies are within the [−0.9 and 0.9 dpm 100 L −1 ] interval), whereas for 65 % of the samples Ba can be considered conservative (i.e., 65 % of the Ba anomalies are within the [−4.4 and 4.4 nmol L −1 ] interval).
A major finding of this study is that 226 Ra and Ba are predominantly conservative at intermediate depths mostly between 500 and 2000 m, but slightly deeper in the center of   As a comparison, the conservative 226 Ra and Ba vertical profiles derived from the OMP analysis are also shown as solid grey lines. The discrepancy between the two vertical profiles indicates deviation from the conservative behavior and reflects either an input of 226 Ra or Ba (positive anomalies highlighted in red; same color code as Fig. 6) or a removal of 226 Ra or Ba (negative anomalies highlighted in blue; same color code as Fig. 6).
The OMP analysis has not been solved for the shallow coastal stations 53, 61 and 78. The 226 Ra / Ba ratios are also reported, together with the mean GEOSECS 226 Ra / Ba ratio (2.2 ± 0.2 dpm µmol −1 ; black dashed line) and its 1 standard deviation (grey shaded areas). Note that the scale may be different from one station to another and the vertical axis was cut to 1000 m. The seafloor is represented by the bottom axis. deep basins such as at stations 21, 44 and 69 (Fig. 6). These locations correspond to the depths at which the waters are far from the main sources and sinks of 226 Ra and Ba. The nonconservative 226 Ra (42 % of the 226 Ra) is mostly distributed close to interfaces such as surface/subsurface water and bottom water (both in the deep West European Basin and the Labrador Sea), near the main sources (seafloor or shallow sediments deposited onto the margins). The nonconservative Ba is mostly distributed in the upper 1500 m and in the deep West European Basin (Fig. 6). The 226 Ra / Ba ratios are also reported for all samples in Fig. 7. The mean ratio determined along the GA01 section is identical to the ratio determined during the GEOSECS program (2.2 ± 0.2 dpm µmol −1 ; Östlund et al., 1987). Of the 226 Ra / Ba ratios determined along the GA01 section 77 % are within the confidence interval [2.0-2.4 dpm µmol −1 ], indicating little deviation from the mean ratio, a pattern that is likely related to the similar chemical behavior in 226 Ra and Ba.

226 Ra inputs and their relationship with Ba
The deepwater of the West European Basin displays positive 226 Ra and Ba anomalies (stations 1 to 26; Fig. 7). The 226 Ra anomalies are initiated at shallower depths (ca. 300-2000 m) than the Ba anomalies (ca. 1000-2000 m) (Fig. 7). As mentioned above, the NEADWl -which is the main water mass of the deep West European Basin -is mainly formed from waters with a southern origin (mainly AABW) that are characterized by high 226 Ra and Ba concentrations. However, these southern waters experience a very specific history along their northward transport to the GA01 section that suggests that the high 226 Ra activities (and Ba) of the NEADWl cannot be solely explained by the high 226 Ra activities (and Ba) of these waters of southern origin. In order to explain the positive 226 Ra and Ba anomalies in the deepwater of the West European Basin, we thus need to investigate the fate of 226 Ra and Ba in the waters of southern origin that travel northward and reach section GA01. Figure 8 was computed by combining GEOSECS and TTO data ( 226 Ra, Si(OH) 4 , salinity and temperature) associated with the AABW (Spencer, 1972) that travels northward between 60 • S and 40 • N in the West Atlantic Basin. The same data ( 226 Ra, Si(OH) 4 , salinity and temperature) associated with the NEADWl in the East Atlantic Basin and along GA01 are also reported.
Between 60 • S and the equator, the high 226 Ra activities of the AABW are associated with relatively low salinity, low temperature and high Si(OH) 4 (Fig. 8). Then, while crossing the Mid-Atlantic Ridge at the equator and at 11 • N, the AABW undergoes several important transformations: 226 Ra activities and Si(OH) 4 concentrations decrease, while salinity and temperature tend to increase (Fig. 8). The 226 Ra and Ba end-members for the NEADWl were chosen at this specific location to coincide both geographically and with the characteristics (salinity, temperature and Si(OH) 4 ) of the NEADWl end-members used for the OMP analysis (Figs. 8,  S3). This end-member has been chosen far from the GA01 section in the OMP analysis (García-Ibáñez et al., 2018), because between 11 • N and the GA01 section (Fig. 8), salinity, temperature and Si(OH) 4 concentrations display relatively constant trends indicating no major modifications. In contrast, the 226 Ra activities display a significant spatial variability north of 11 • N and clearly increase towards the north (Fig. 8), a pattern that is especially clear when taking the GA01 data into account (Fig. 8). This 226 Ra increase is thus decoupled from salinity, temperature and Si(OH) 4 ), and likely explains the positive anomalies deduced from the OMP analysis in the deep West European Basin (Fig. 7). The specific history of these waters of southern origin (waters initially with a high 226 Ra activity; decrease in the 226 Ra activity at the equator and at 11 • N; new increase of 226 Ra activity north of 11 • N) suggest that the 226 Ra anomalies observed in the West European Basin are explained by inputs of 226 Ra along the northward transport of these waters. The positive anomalies result from the input of 226 Ra (and Ba) following either (i) dissolution/remineralization of settling particles that incorporated 226 Ra and Ba in the upper water column (this includes the dissolution of barite, since the waters of the Atlantic Ocean are undersaturated with respect to barite; Monnin et al., 1999) and/or (ii) diffusion of 226 Ra and Ba from deep-sea sediments (Cochran and Krishnaswami, 1980) (see Sect. 4.4). Of special note are stations in the West European Basin, which are especially deep (down to 5500 m). Deep sediments generally display elevated 230 Th activities due to scavenging of 230 Th from the entire water column (Bacon and Anderson, 1982;Nozaki, 1984). The highest dissolved 230 Th activities reported along the GA01 section were thus observed in the deepwater of the West European Basin (Deng et al., 2017(Deng et al., , 2018. Consequently, because 226 Ra is produced by the decay of 230 Th in the sediment, the 226 Ra diffusive flux in this area is expected to be especially high. The input of 226 Ra in the West European Margin is accompanied by Ba input since (i) positive Ba anomalies are also observed in the deepwater and (ii) the 226 Ra / Ba ratios do not significantly deviate from the mean GEOSECS 226 Ra / Ba ratio (Fig. 7a). One exception is found at station 21 in the West European Basin, which displays high 226 Ra / Ba at approximately 4000 m (up to 3.17 dpm µmol −1 ). At several stations (21, 26, 32, 38, 44, 60, 64 and 77), lower beam transmission values near the seafloor indicate presence of suspended sediments likely associated with a nepheloid layer. Nepheloid layers are turbid layers formed episodically by strong and intense abyssal currents that are transported along isopycnals away from the site of resuspension of bottom sediments (McCave, 1986). The presence of a benthic nepheloid layer is also indicated by high particulate iron concentrations at these stations. Such processes may thus contribute to releasing 226 Ra (and potentially Ba) to the deep water column, following desorption or dissolution of the particles. Similar 226 Ra maxima have been observed in the deepwater of the northeastern Pacific, suggesting that the 226 Ra flux is not uniform over the ocean bottom even on a regional scale (Chung, 1976). Suspended particle dissolution may also play a role here, notably for Ba. This will be discussed in more detailed in Sect. 4.4.
Positive 226 Ra anomalies are also found in deepwater at several other stations located in relatively deep basins (> 1200 m) along the GA01 section (e.g., stations 32, 38, 44, 60, 64, 69 and 77). Most of these anomalies are associated with 226 Ra / Ba ratios higher than 2.4 dpm µmol −1 . The 226 Ra positive anomalies observed at the stations mentioned above are thus best explained by the diffusion of 226 Ra from the sediment. However, these latter stations do not exhibit a positive Ba anomaly and Ba tends to be conservative. Consequently, the 226 Ra / Ba ratios in the deepwater of these stations is often significantly higher than the mean GEOSECS value (stations 21, 32, 38, 60, 64; Fig. 7). This pattern is different to that observed in the West European Basin, a discrepancy that may be explained by the different sediment composition in the two regions, the different residence times of deepwater in contact with deep-sea sediments (Chung, 1976) and/or the different roles played by suspended particle dissolution.
A strong 226 Ra positive anomaly is observed in the deepest sample collected at station 38 above the Reykjanes Ridge. It cannot be completely excluded that this positive anomaly is due to the hydrothermal vent since hydrothermal systems are known in the area (Fig. 1). High particulate iron and aluminum concentrations were also observed at these stations Menzel et al., 2017Menzel et al., , 2018. Enrichment in 226 Ra has indeed been observed in hydrothermal system plumes at mid-ocean ridge (Kadko, 1996;Kadko and Moore, 1988;Kipp et al., 2017;Rudnicki and Elderfield, 1992). Moreover, the 226 Ra enrichments are accompanied by slight Ba enrichments, which may support the hydrothermal origin hypothesis, since hydrothermal venting at mid-ocean Ridge constitutes the second major external source of Ba to the ocean (Edmond et al., 1979).
Finally, high 226 Ra / Ba ratios variations are also observed in shallow coastal waters (Fig. 7c). At stations 53 and 61, high 226 Ra / Ba ratios are found close to the bottom, in agreement with the input of 226 Ra from the sediment, whereas low 226 Ra / Ba ratios are found in the subsurface at stations 57, 61 and 78, in association with low salinities (Fig. S2). The low 226 Ra / Ba ratios are thus explained by the input of meteoritic water in coastal areas, since such waters are known to be the predominant source of Ba to the ocean (Martin and Meybeck, 1979;Wolgemuth and Broecker, 1970). At these shallow stations, the different source terms between 226 Ra and Ba therefore lead to important variations in the 226 Ra / Ba ratios (Fig. 7c.; stations 53, 57, 61 and 78). These observations clearly indicate that 226 Ra may sometimes be decoupled from Ba.

226 Ra removal and its relationship with Ba
Relatively few 226 Ra negative anomalies were found along the GA01 section. At the deep open-ocean stations, the neg-ative anomalies are mostly observed in the upper 1000 m ( Fig. 7; stations 13 , 21, 26, 32, 38, 44 and 77), but can also be found as deep as 2000 m (i.e., stations 64 and 69). In most cases, the negative 226 Ra anomalies are associated with significant negative Ba anomalies (stations 13,21, 26, 38, 44, 64 and 69). Such features are likely explained by biological mediated processes including the incorporation of 226 Ra and Ba in planktonic as calcareous and siliceous shells (Bishop, 1988), in Acantharian skeletons made of celestite (SrSO 4 ) or in barite (BaSO 4 ) crystals (van Beek et al., 2007;Chow and Goldberg, 1960;Shannon and Cherry, 1971;Szabo, 1967;Wolgemuth and Broecker, 1970).
Particularly low dissolved 226 Ra / Ba ratios (< 2 dpm µmol −1 ) are found in the upper 50 m at stations 21, 32, 64, 69 and 77, a pattern that was also observed in the upper 150 m of the Sargasso Sea, where van Beek et al. (2007) reported similarly low dissolved 226 Ra / Ba ratios that were accompanied by high 226 Ra / Ba ratios in suspended particles. This latter pattern was attributed to the incorporation of 226 Ra and Ba in Acantharian skeletons. The low dissolved 226 Ra / Ba ratios (e.g., 1.7 dpm µmol −1 , station 77) observed in the upper 200 m along the GA01 section may thus be attributed to Acantharians, which have skeletons that incorporate 226 Ra preferentially to Ba (van Beek et al., 2007Bernstein et al., 1998). The presence of Acantharians was not studied during GEOVIDE. However, previous studies reported the presence of Acantharians in this area, for example in the Iceland Basin and in the East Greenland Sea (Antia et al., 1993;Barnard et al., 2004;Martin et al., 2010).
Several phytoplankton blooms were observed along the GA01 section, as indicated by the chlorophyll a concentrations (Chl a) (Fig. S4). The largest bloom was observed in the Labrador Sea in May 2014. Diatoms were the dominant species in the Irminger and Labrador seas and on the Greenland and Newfoundland margins during GA01 (up to 55 % of the total Chl a concentration; Tonnard et al., 2018). In the West European Basin, Chl a was lower in May and June 2014 and coccolithophorids were the dominant species in that area (Tonnard et al., 2018). In these two regions, diatom frustules and coccolithophorids may thus contribute to the removal of 226 Ra and Ba (Bishop, 1988;Dehairs et al., 1980), from the water column in areas that were characterized by noticeable negative anomalies.
Additionally, because the Labrador Sea was sampled in June, during the decline of the bloom, barite that is presumably formed following the decay of settling organic matter may also contribute to the removal of Ba (and 226 Ra). High particulate excess Ba (Ba xs ) concentrations were indeed observed at stations displaying significant Ba negative anomalies: Ba xs concentrations reached a maximum of 400 m at station 13, 21 and 26) and between 400 and 800 m near Greenland, at stations 44, 64 and 69 (Lemaitre et al., 2017(Lemaitre et al., , 2018b. These Ba xs profiles can be related to the relatively high particulate organic carbon (POC) export flux de-termined at these stations (e.g., at station 69; Lemaitre et al., 2017Lemaitre et al., , 2018a. This POC flux would promote barite formation in the subsurface (Dehairs et al., 1980;Legeleux and Reyss, 1996) but also deeper in the water column (van Beek et al., 2007). Similarly, Jullion et al. (2017) -by using a parametric OMP analyses as applied in the Mediterranean Sea -also reported quantification of the nonconservative component of the Ba signal and suggested that the Ba negative anomalies potentially reflected Ba subtraction during barite formation occurring during POC remineralization. The winter deep convection in the Labrador Sea may also potentially explain these relatively deep Ba anomalies by transporting negative Ba and 226 Ra anomalies waters toward the deep sea, as was observed in the Mediterranean Sea (Jullion et al., 2017). With the exception of the acantharian skeletons that may impact the dissolved 226 Ra / Ba ratios in the upper 200 m, the removal of 226 Ra and Ba that takes place deeper in the water column or that involves other processes (e.g., barite precipitation) does not seem to significantly affect the dissolved 226 Ra / Ba ratios.
In the shallow coastal stations, lower 226 Ra / Ba ratios are observed (Fig. 7). These low ratios may also result from the removal of 226 Ra and Ba by planktonic shells and/or barite or from scavenging by lithogenic particles. However, because these stations are coastal stations, various processes are at play in a relatively shallow water column (i.e., diffusion of 226 Ra from the sediments; input of Ba from meteoric water; removal of Ba and 226 Ra by shells and barite), thus complicating the interpretation of the vertical profiles. We note that the low 226 Ra / Ba ratios observed in the surface of shallow stations near the coast of Greenland (stations 57 and 61) and Newfoundland (station 78) are associated with low salinities (Fig. 7c). This decoupling between 226 Ra and Ba may be explained by input of fresh water into the coastal zone.
Finally, at several stations, a decrease in the 226 Ra activities is observed near the seafloor (stations 13,21,44,60,64 and 77;Fig. 7). Similar decreasing trends near the seafloor have been reported in the southwest Atlantic and North Pacific for 230 Th (Deng et al., 2014;Okubo et al., 2012), a reactive element that strongly adsorbs onto suspended particles. This trend for 230 Th was explained by nuclide scavenging at the seafloor (Deng et al., 2014;Okubo et al., 2012). Radium-226 -although it is much less particle-reactive than 230 Th -and Ba may also be scavenged by resuspended particles near the seafloor and adsorb onto the surfaces of Mn oxides (Moore and Reid, 1973). High particulate trace element concentrations were also observed at stations 26, 38 and 69 and may be related to nepheloid layers that impact the deep water column up to 200-300 m above the seafloor Menzel et al., 2017Menzel et al., , 2018.

Estimation of 226 Ra and Ba input fluxes into the West European Basin
A strong 226 Ra positive anomaly is observed in the NEADWl between stations 1 and 21 and below 3500 m. On average, it is 3.3 dpm 100 L −1 over this vertical section. This anomaly reflects a concentration difference between the 226 Ra measured along GA01 and the 226 Ra due to water mass mixing. This concentration difference is associated with the northward transport of the NEADWl, estimated to be 0.9 ± 0.3 Sv (10 6 m 3 s −1 ) at 45 • N (GA01 section) (Daniault et al., 2016;McCartney, 1992). Therefore, the positive concentration anomaly can be converted to a 226 Ra flux that has to be added to this water mass, F Input-Ra , calculated as follows: where A is the mean positive anomaly of 226 Ra (in dpm m −3 ) determined by the OMP analysis; T NEADWl is the transport associated with the NEADWl (in m 3 s −1 ). This 226 Ra input may be associated with a sedimentary source and/or may result from the dissolution of suspended particles. First, we will study the hypothesis of the sedimentary source; the suspended particle source will be discussed second.
The NEADWl at 45 • N is made of up to 92 % of the 11 • N NEADWl end-member. Therefore, the sedimentary input along the northward transport of the NEADWl is calculated across a sediment area between 11 and 45 • N (Fig. S3). We consider a distance of 4209 km between 11 • N and the GA01 section and a distance of 1475 km between stations 1 and 21. This leads to a horizontal area of 6.21 × 10 6 km 2 (assuming a constant bathymetry), across which the sedimentary input is assumed to take place.
The 226 Ra flux diffusing out of bottom sediments, F Sed-Ra (in dpm cm −2 yr −1 ), can be calculated using Eq. (2), assuming that the anomaly is entirely explained by the sediment source: where S is the surface area described above (in cm 2 ) and F Input-Ra is 1.67 × 10 8 dpm s −1 . The calculated F Sed-Ra is 14.8 ± 3.1 × 10 −3 dpm cm −2 yr −1 , which is within the range of fluxes reported in the literature. For example, Cochran (1980) reported F Sed-Ra in the range of 1.5 × 10 −3 dpm cm −2 yr −1 for the Atlantic Ocean to 2.1 × 10 −1 dpm cm −2 yr −1 for the Pacific Ocean (Fig. 9). Li et al. (1973) estimated 226 Ra fluxes diffusing out of the sediment in the Southern Ocean and on the Antarctic shelf of 6.2 × 10 −4 dpm cm −2 yr −1 and 1.6 × 10 −3 dpm cm −2 yr −1 , respectively. The F Sed-Ra calculated here is thus slightly higher than the 226 Ra sedimentary fluxes reported in the Atlantic Ocean by Cochran (1980). Note, however, that the  Cochran (1980). The 226 Ra flux calculated in this study to explain the positive anomalies in the West European Basin is also reported (red dot). 226 Ra fluxes released from the sediments vary locally as a function of the sedimentary 230 Th activity, bioturbation rates, but also as a function of the sediment type and composition (Chung, 1976;Cochran, 1980). The 226 Ra fluxes reported in the Atlantic Ocean by Cochran (1980), which are the lowest of all basins, are only available for calcareous ooze type sediment (Cochran, 1980). The NEADWl may cross different types of sediments along its northward path in the Atlantic Ocean. This includes calcareous oozes, fine-grained calcareous sediments and clay (Dutkiewicz et al., 2015). In particular, 226 Ra diffusion is expected to be higher in these two latter sediment types (Cochran, 1980). As for Ba, the mean positive anomaly deduced from the OMP analysis is 7.0 nmol L −1 leading to a F Input-Ba of 69.5 mol s −1 . In the same way as 226 Ra, a Ba sedimentary flux F Sed-Ba of 3.19 ± 1.4 nmol cm −2 yr −1 would be required to explain the Ba anomalies observed in the West European Basin. This flux is on the low end of the Ba sedimentary fluxes reported in different ocean basins, which range from 1.0 to 30 nmol cm 2 yr −1 (Chan et al., 1977;Falkner et al., 1993;McManus et al., 1999;Paytan and Kastner, 1996).
Alternatively, the dissolution of settling particles could also contribute to the 226 Ra and Ba anomalies observed in the deepwater of the West European Basin. Assuming a steady state, we may undertake a mass balance calculation for particulate 226 Ra and Ba in the same box as described above (i.e., box defined between 11 • N and the GA01 section, between stations 1 and 21 -1475 km -and between 3500 m depth and the seafloor; Fig. S5). Particles enter the  Figure 10. 226 Ra and Ba flux estimations: F Part-x is the particulate flux entering the box, F Accumulation-x is the sediment accumulation flux and F Dissolution-x is the flux of particle dissolution assuming that all the settling particles dissolve. x is either 226 Ra or Ba. Both maximum and minimum values are shown for F Part-x and F Accumulation-x . Max F Dissolution-x represents a maximum value since it is calculated by subtracting the minimum value of F Accumulation-x from the maximum value of F Part-x . F Dissolution-x is thus between zero (if all F part-x accumulates in the sediment) and this latter value.
box from above as settling particles, but also horizontally, carried within the water masses at 11 • N that travel northward. Particles leave the box by different processes (accumulation in the sediment or northward transport by the water masses) or may dissolve while settling in the box. In the absence of precise information about the particulate 226 Ra and Ba fluxes entering and exiting the box horizontally (i.e., the particulate 226 Ra and Ba concentrations at 11 • N and at the GA01 section), we assume that they are of equal importance and therefore that they cancel each other out in the mass balance calculation.
The vertical particulate flux entering the box from above can be calculated as follows: where x is either 226 Ra or Ba, Cp 3500 is either the particulate 226 Ra activities or the particulate Ba concentrations at 3500 m, V s is the settling speed for suspended particles and S is the horizontal surface area described above (6.21 × 10 6 km 2 ). We use the value of 0.007 dpm 100 L −1 for the mean 226 Ra particulate activity at 3500 m, which was reported for the Atlantic Ocean, Sargasso Sea (van Beek et al., 2007) and the value of 0.087 nmol L −1 for the mean Ba particulate concentration at 3500 m, which was determined along the GA01 section (Lemaitre et al., this issue). We use the settling speeds (V s ) reported for suspended particles in the literature and that typically range from 100 to 1000 m yr −1 (Bacon and Anderson, 1982;Krishnaswami et al., 1976;Roy-Barman et al., 2002). The F Part thus obtained range from 1.4 × 10 6 to 13.8 × 10 6 dpm s −1 for 226 Ra, while the F Part range from 1.7 and 17.2 mol s −1 for Ba. Of this total F Part , a fraction may dissolve while settling, while the remainder will accumulate in the sediment. This dissolution flux is named F dissolution-x , where x is either 226 Ra or Ba. We use the sediment Ba ac-cumulation rates reported by Gingele and Dahmke (1994) in the Atlantic Ocean to calculate the particulate Ba flux that exits the box (F Accumulation-Ba : 2.0 to 13.4 mol s −1 ); hence, by difference the F dissolution-Ba is 0-15.2 mol s −1 (Fig. 10). This value is of the same order of magnitude of the F Input-Ba needed to explain the Ba anomalies (6.28 mol s −1 ). Therefore, in the case of Ba, the dissolution of settling particles may entirely explain the OMPA-derived anomalies. The sediment 226 Ra accumulation rates can be calculated from the Ba accumulation rates estimated above using the 226 Ra / Ba ratio determined in sinking particles collected in the Sargasso Sea near the seafloor (i.e., 1.5 dpm µmol −1 ; van Beek et al., 2007). The sediment 226 Ra accumulation flux thus calculated, F Accumulation-Ra , is 2.9 × 10 6 -19.6 × 10 6 dpm s −1 , leading to F dissolution-Ra of 0-10.9 10 6 dpm s −1 (Fig. 10). Therefore, F dissolution-Ra cannot account for more than 37 % of the required 226 Ra flux (F Input-Ra ). This implies that even if the settling speed is high (1000 m yr −1 ; high turnover of the particles), the particle dissolution cannot account for the entire F Input-Ra . The remaining part (minimum of 63 %) therefore has to be sustained by 226 Ra diffusion from the sediments.
While the above calculations have to be taken with caution given the numerous assumptions in the mass balance model, overall they suggest that the 226 Ra positive anomalies observed in the West European Basin may be explained entirely by the 226 Ra that diffuses out of the sediments. However, it cannot be excluded that the dissolution of settling particles also contributes to the 226 Ra enrichment. In contrast, the Ba positive anomalies may be explained either by the diffusion of Ba from sediment, by the dissolution of settling particles or by a combination of these two processes. These conclusions are in line with the current knowledge about 226 Ra and Ba sources in the deep ocean Chan et al., 1976Chan et al., , 1977Ku et al., 1980). We investigated the distribution of dissolved 226 Ra activities and Ba concentrations in the North Atlantic Ocean along the GA01 section. To a first order, the 226 Ra and Ba patterns appear to be correlated with the water masses (e.g., high 226 Ra and Ba in NEADWl in the West European Basin; low 226 Ra and Ba in Central Waters; slight increase of 226 Ra in the MW). Using a mixing model, we show that the measured 226 Ra and Ba concentrations can be explained by conservative mixing for 58 and 65 % of the samples, respectively, notably at intermediate depths (mostly between 1000 and 2000 m) and slightly deeper in the middle of deep basins, away from the ocean interfaces. These locations correspond to the depths at which the waters are away from the main sources of 226 Ra and Ba. 226 Ra and Ba can thus be considered to be conservative tracers of water mass transport in the ocean interior on the space scales considered here, namely on the order of a few thousand kilometers.
Our study also highlighted several regions in which significant input or loss of 226 Ra and Ba takes place. In the West European Basin, the deepwater (NEADWl) accumulates both 226 Ra and Ba while flowing from 11 • N to the GA01 section. Mass balance calculations suggest that these enrichments are predominantly explained by sediment diffusion for 226 Ra, with a possible contribution from suspended particle dissolution, while both the sediment and suspended particle dissolution could significantly contribute to the Ba enrichments. This pattern contrasts with that observed in the deepwater collected elsewhere along the section that does not display Ba enrichments associated with the 226 Ra enrichments. Bottom nepheloid layers may also contribute to the release of 226 Ra (and Ba) to the deep water column at several stations. Interestingly, nepheloid layer processes seem to also act as potential removal of 226 Ra (and Ba) at several other stations. Significant input of Ba likely associated with meteoric waters is found in the upper water column near Greenland. Finally, 226 Ra and Ba are removed from the upper water column, likely primarily due to biological mediated processes, such as incorporation of 226 Ra and Ba into barite (BaSO 4 ), which are presumably formed following the decay of settling organic matter and/or adsorption onto diatom frustules, a mechanism that would explain the 226 Ra-Ba relationship reported here. Similarly, strong correlations were also found between Ba-Si and 226 Ra-Si, although no obvious process links 226 Ra or Ba with Si, except maybe the adsorption of Ba and ( 226 Ra) onto diatom frustules. It cannot be excluded, however, that the observed Ba-Si and 226 Ra-Si relationships may result from the spatial coherence of different carriers overprinted by hydrodynamics.
Our study also provides evidence of significant decoupling between 226 Ra and Ba. In the upper 200 m, the 226 Ra / Ba ratios reach low values (< 2 dpm µmol −1 ), a pattern that has been observed in other regions, and was related to acantharian skeletons that incorporate 226 Ra preferentially to Ba (van Beek et al., 2007;Bernstein et al., 1998). Finally, deviations from the mean GEOSECS 226 Ra / Ba ratios were observed in the shallow coastal waters of Greenland and Newfoundland: the predominant input of Ba due to the input of meteoric water leads to lower 226 Ra / Ba ratios, whereas near the seafloor, the input of sedimentary 226 Ra leads to higher 226 Ra / Ba ratios.
The absence of a stable isotope for radium led geochemists to consider Ba a stable analog for 226 Ra because 226 Ra and Ba display similar chemical behavior, with the aim of using the 226 Ra / Ba ratio as a chronometer for the thermohaline circulation. This study confirms that 226 Ra and Ba behave similarly in the ocean interior away from external sources, both elements being predominantly conservative in the studied area over distances on the order of a few thousands of kilometers. However, this study also highlights regions in which 226 Ra and Ba deviate from a conservative behavior, an important consideration when considering the balance between the large-scale oceanic circulation and biological activity over long timescales. Decoupling between 226 Ra and Ba has been observed, in most cases at the ocean boundaries as the result of dissolved 226 Ra and Ba external sources. In addition, suspended particle dissolution may differently impact the dissolved 226 Ra and Ba content of intermediate and deepwater (as shown for the NEADWl); such process would therefore potentially modify their 226 Ra / Ba ratios and would complicate the use of this ratio as a chronometer. Inclusion of the different sources and sinks and particledissolved interactions in global ocean models should help to refine the use of the 226 Ra / Ba ratio as a clock to time the thermohaline circulation, as was proposed several decades ago during the GEOSECS program.
Occitanie-Pyrénées-Méditerranée (European Regional Development Fund). We are grateful to the captain and crew of the N/O Pourquoi Pas?. The GEOVIDE project is co-funded by the French national program LEFE/INSU (GEOVIDE), ANR Blanc (GEOVIDE, ANR-13-BS06-0014) and RPDOC (ANR-12-PDOC-0025-01), LabEX MER (ANR-10-LABX-19) and IFREMER. The GEOVIDE cruise would not have been achieved without the technical skills and commitment of Catherine Kermabon, Olivier Ménage, Stéphane Leizour, Michel Hamon, Philippe Le Bot, Emmanuel de Saint-Léger and Fabien Pérault. We are grateful to Manon Le Goff, Emilie Grosstefan, Morgane Gallinari and Paul Tréguer for Si(OH) 4 sampling and analysis. This work was also co-funded by the French national program LEFE/INSU "REPAP" (PI Stéphanie H. M Jacquet) and the US National Science Foundation (PI Matthew A. Charette, OCE-1458305;OCE-1232669). For this work Maribel I. García-Ibáñez and Fiz F. Pérez were supported by the Spanish Ministry of Economy and Competitiveness through the BOCATS (CTM2013-41048-P) project co-funded by the Fondo Europeo de Desarrollo Regional 2014-2020 (FEDER). Several figures were constructed using Ocean Data View (Schlitzer, 2003). Therefore, Reiner Schlitzer is warmly thanked. Satellite chlorophyll a visualizations used in this study were produced with the Giovanni online data system, developed and maintained by the NASA GES DISC.
Edited by: Gideon Henderson Reviewed by: two anonymous referees