Samples were collected in the North-West Pacific in December 2002 on board RV Mirai, as part of the JPAC program (PIs: Makio Honda and Susumu Honjo) at station K1 (51°N-165°E; 5,130 m) in the Western Subarctic Gyre and at station K3 (39°N-160°E; 5,470 m) usually located north of the Kuroshio Front ( Figure 1 ). We also report data on samples collected at old station Hale Aloha (22°28’N; 158°8’ W) located north of Hawaii in June 2002 on board RV Ka’imikai-o-Kanaloa (KoK) and in August 2002 on board RV Wecoma (PI: Mike Bacon). This station is located south of station ALOHA that is the site of the Hawaii Ocean Time-series program (HOT, 22.75°N; 158.00°W). These sites off Hawaii are oligotrophic (low macronutrient and therefore low chlorophyll concentration; Karl and Lukas, 1996) with warm waters, whereas the waters of the Western Subarctic Gyre are cold and mesotrophic (Honda, 2003; Buesseler et al., 2008). Coccolithophorids (CaCO₃) are common off Hawaii (Cortes et al., 2001), whereas the waters of the Western Subarctic Gyre (e.g. stations K2 and K1; Figure 1 ) are dominated by diatoms (Honda et al., 2002; Honda et al., 2006). Station K3 also located in the northwestern North Pacific subarctic region is expected to be dominated by diatoms, based on sediment trap data from a nearby station (40°N/165°E) indicating that biogenic opal is predominant (Honda et al., 2002). The export of carbon is also higher in the Subarctic Gyre compared with the oligotrophic region off Hawaii (Buesseler et al., 2008).

The two investigated areas were studied later in the framework of the VERtical Transport In the Global Ocean, VERTIGO project, in years 2004 and 2005 (i.e., station Aloha, off Hawaii and station K2 in the Western Subarctic Gyre) (Buesseler et al., 2008). In particular, the Ba dataset (dissolved Ba and particulate Ba concentrations) obtained in the JPAC project on the entire water column can be compared to the Ba dataset collected during the VERTIGO project in the “twilight zone” (0-1000 m) (Dehairs et al., 2008). In addition, the vertical distribution of the ²²⁶Ra activities reported here complement the study of Kawakami and Kusakabe (2008) who investigated the horizontal distribution of ²²⁶Ra (and ²²⁸Ra) in surface waters of the northwestern North Pacific Ocean. Time-series sediment trap moorings have been deployed in the northwestern North Pacific Ocean and provide information on sinking particle fluxes in the region (Honda, 2003).

The vertical profiles of potential temperature, salinity and potential density are reported in Figure 2 of the Supplementary Material for both stations K1 and K3. Station K3 is usually located north of the Kuroshio Front. However, Argo floats measurements conducted in the Northwest Pacific (Oka et al., 2015) indicate that the Kuroshio may sometimes extent toward station K3 ( Figure 3 of Supplementary Material ). This was the case in 2002, when station K3 was investigated in the present study, as indicated by the higher potential temperature and higher salinity in the upper 400 m at station K3 ( Figure 2 of Supplementary Material ) and by geochemical tracers (see Ra Isotopic Ratios (²²⁸Ra/²²⁶Ra)).

Seawater samples were collected using 12 liter-Niskin bottles. Between 35 and 56 kg of seawater were collected at old station Hale Aloha and between 74 and 130 kg per depth were collected at stations K1 and K3 to a depth of 5200 m. Such relatively large seawater volumes were collected in order to allow the determination of ²²⁸Ra - in addition to ²²⁶Ra - which displays low concentrations in seawater. A separate aliquot (60 ml seawater) was collected for the determination of dissolved Ba concentrations. The large volume seawater samples (between 35 and 130 kg) were passed through a cartridge filled with acrylic fiber impregnated with MnO₂ (so called “Mn fibers”), following the technique described by Moore and Reid (1973). We used a flow-rate of 200 ml min^-1 to ensure the quantitative separation of ²²⁶Ra from seawater (Moore et al., 1985; van Beek et al., 2010).

Suspended particles were collected using McLane large volume pumps (WTS, McLane Labs, Falmouth, Ma, USA). Large volumes of seawater were filtered in situ through 142-mm diameter Versapor filters (acrylic copolymer on a nylon substrate; Pall Corporation), with a pore size of 0.8 µm. These large volume samples were required for gamma spectrometry and ICP-MS analyses. Additionally, seawater samples (9-11 L) were collected at station K3 using Niskin bottles and were passed through 47-mm polycarbonate filters with a pore size of 0.8 µm. To investigate the different Ba carriers within the suspended particles, the filters were sequentially leached using i, MilliQ water; ii, 10% (v/v) HNO₃; iii, 3% HCl; iv, hot 50% (v/v) HNO₃, following Ganeshram et al. (2003), slightly modified. Steps 1 to 3 were conducted on the ship immediately after collection using the clean room facility on RV Mirai. Step 4 was conducted back in the laboratory at the Woods Hole Oceanographic Institution (WHOI). Step 1 (MilliQ water) rinses dissolved Ba associated with residual seawater, weakly sorbed Ba and Ba released by cell lysis. 10% (v/v) nitric acid removes readily exchangeable and easily hydrolysable organically bound Ba. Celestite skeletons of acantharians, metal hydroxides and potentially also extracellular polymeric substances (EPS) are expected to be dissolved during steps 1 and 2. Step 3 - slightly modified from Ganeshram et al. (2003) following Weast et al. (1966), as done in van Beek and Reyss (2001) - dissolves BaSO₄. Finally, step 4 removes Ba associated with refractory organic matter.

For dissolved ²²⁶Ra, the Mn fibers were ashed in a furnace for 16 hours at 820°C (Charette et al., 2001). The ash was then transferred into counting vials, sealed to prevent any Rn loss and analyzed for their radium activity using the low-background gamma spectrometers at the underground laboratory of Modane (Laboratoire Souterrain de Modane, LSM, French Alps) and at the underground laboratory of Ferrières (LAboratoire de mesure des FAibles RAdioactivités, LAFARA, French Pyrénées). High-efficiency, low-background, well-type germanium detectors (430 cm³ and 950 cm³ at LSM, Reyss et al., 1995; 280 cm³ at LAFARA, van Beek et al., 2010; van Beek et al., 2013) were used. ²²⁶Ra activities were determined using the ²¹⁴Pb (295 keV and 352 keV) and ²¹⁴Bi (609 keV) peaks. ²²⁸Ra activities were determined using the 338, 911 and 969 keV peaks of ²²⁸Ac. Counting times for each sample ranged from 2 to 5 days, in order to allow the detection of the low ²²⁸Ra activities in addition to ²²⁶Ra activities. Uncertainties for ²²⁶Ra activities are errors due to counting statistics. The low counting background afforded by the counting facilities, the large sample volumes and the long counting times resulted in uncertainties for ²²⁶Ra (ie. ± 1 to ± 3%) smaller than reported from most of the measurements performed during the GEOSECS program (± 3% to ± 10%).

For dissolved Ba, aliquots of seawater were analysed by isotope dilution using ¹³⁵Ba. The determination of dissolved Ba concentrations at old station Hale Aloha was conducted using the ICP-MS facility (Element, Finnigan) at WHOI. For samples collected in the North West Pacific (K1 and K3 stations), analyses were made using the ICP-MS facility at Scottish Universities Research and Reactor Centre, East Kilbride, Scotland (VG PlasmaQuad II). Uncertainty on the Ba concentrations is estimated at ± 2%.

For particulate ²²⁶Ra and Ba, the filters were first placed in a counting tube and sealed prior to analysis to prevent any Rn loss, and counted for Ra isotopes using the low background gamma spectrometers at the underground laboratory of Modane (Laboratoire Souterrain de Modane, LSM, French Alps) using the same method as for dissolved ²²⁶Ra. The Versapor filters were subsequently dissolved in a Teflon beaker with a mixture of HNO₃ (ultrapure acid), HF (ultrapure acid), and HClO₄ (pure acid) on a hot plate to measure Ba, ²³⁸U and ²³²Th using the ICP-MS facility at Observatoire Midi Pyrénées, Toulouse (Elan 6000 Perkin-Elmer), as described in van Beek et al. (2009). Excess Ba (Ba_ex), excess ²²⁶Ra activities (²²⁶Ra_ex) and excess ²²⁸Ra activities (²²⁸Ra_ex) in suspended particles were determined by subtracting the Ba, ²²⁶Ra, and ²²⁸Ra associated with the lithogenic component, as described in van Beek et al. (2007). The solutions obtained during sequential leaching were analyzed using the ICP-MS facility (Element, Finnigan) at WHOI.

The barite saturation index is defined as the ratio of the aqueous barium sulfate ionic product to the barite solubility product at a given temperature and pressure. For a given seawater sample, the concentrations of Na, K, Ca, Mg, Cl and SO₄ are calculated from the measured salinity and from the ionic composition of standard seawater (Millero, 2006). The measured Ba and Sr concentrations are then used along with the major element concentrations to calculate ionic activities in the Na-K-Ca-Mg-Ba-Sr-Cl-SO₄-H₂O system based on Pitzer’s ion interaction model (Monnin and Galinier, 1988; Monnin, 1999), for the temperature and pressure of the sample under consideration. Sr substitution in barite has a limited influence on the barite saturation index (Monnin and Cividini, 2006).

The ²²⁶Ra activities and Ba concentrations determined in seawater at old station Hale Aloha, station K1, and station K3 are reported in Tables 1 - 3 of the Supplementary Material , and the vertical profiles are shown on Figure 2 . Both increase with depth, reflecting uptake during particle formation in shallow waters and subsequent release at depth. Similar profiles were reported in the North Pacific during the GEOSECS program (Chung, 1976; Chung and Craig, 1980). The ²²⁶Ra and Ba profiles are similar at the three stations, but above 3000 m, Ba concentrations (and ²²⁶Ra) are higher in the Subarctic Gyre (station K1) than at station K3 ( Figure 2 ), while below 3000 m, there is a clear decrease in Ba concentrations at K1. The Ba concentrations at old station Hale Aloha are similar in June (lower sampling resolution) and August 2002 ( Table 1 of the Supplementary Material ). In the upper water column, the ²²⁶Ra and Ba concentrations at station K1 are higher than those at stations located further south (K3 and old Hale Aloha), a pattern that was also observed during previous projects (Dehairs et al., 2008; Kawakami and Kusakabe, 2008). The higher concentrations in the upper waters of the northwestern North Pacific subarctic region are likely explained by the impact of the continents and continental margins in these regions (eg., Amakawa et al., 2004; Hawakami and Kusakabe, 2008; Lam and Bishop, 2008).

The barite saturation index was calculated at stations K1 and K3 considering either pure BaSO₄ or (Ba,Sr)SO₄ ( Figure 3 ). The surface waters at K1 and K3 are undersaturated with respect to barite. Equilibrium is reached at approximately 500 m depth at station K1 and at approximately 1000 m depth at station K3. Below these depths, the waters are either in equilibrium or supersaturated with respect to barite down to approximately 3500 m at station K1 and down to approximately 5000 m at station K3. This pattern is characteristic of Pacific waters and agrees with that reported by previous studies (Monnin et al., 1999; Monnin and Cividini, 2006). This index can be used to provide information on the fate of barite within the water column (preservation versus dissolution), which may in turn impact the distributions of ²²⁶Ra and Ba in the water column. Our data agree with those obtained at station K2 in the Western Subarctic Gyre and station Aloha off Hawaii by Dehairs et al. (2008) during the VERTIGO project.

The Ba and ²²⁶Ra concentrations are broadly consistent with the GEOSECS data set (Broecker et al., 1967; Wolgemuth and Broecker, 1970; Broecker et al., 1976; Chan et al., 1976; Chan et al., 1977; Moore et al., 1985). The ²²⁶Ra/Ba ratios below 500 m agree with the GEOSECS value of 2.3 dpm µmol^-1. In contrast to many GEOSECS profiles, however, the data reported here display significantly lower ²²⁶Ra/Ba ratios in the upper 500 m at stations K1, K3 and old station Hale Aloha ( Figure 4 ). Near the seafloor, a decrease in the ²²⁶Ra/Ba ratio is also observed at stations K1 and K3.

Some studies have reported Ra activities determined either in suspended particles (van Beek et al., 2007; van Beek et al., 2009; Bourquin et al., 2011; van Beek et al., 2013) or in sinking particles collected with sediment traps (Moore and Dymond, 1991; Legeleux and Reyss, 1996; van Beek et al., 2007; van Beek et al., 2009) in the ocean. The vertical profiles of excess Ba concentrations (Ba_ex) and excess ²²⁶Ra activities (²²⁶Ra_ex) in suspended particles at stations K1 and K3 display maximum values between 200 and 500 m ( Figure 5 ), similar to other oceanic basins (Dehairs et al., 1980; Dehairs et al., 1990; Jacquet et al., 2007; van Beek et al., 2007; Dehairs et al., 2008; Jacquet et al., 2008; van Beek et al., 2009). The Ba_ex and ²²⁶Ra_ex vertical profiles at station K1 are remarkably similar, with both profiles displaying a double peak in the upper 400 m. For old station Hale Aloha, we report total ²²⁶Ra activities that have not been corrected for lithogenic ²²⁶Ra, though this correction is generally quite small ( Table 4 of the Supplementary Material ). The ²²⁶Ra activities that have been determined at that station in June and August 2002 agree well with each other and display a maximum value at 250 m, a pattern that is also observed at station K1 (maximum ²²⁶Ra activity at 230 m). The ²²⁶Ra activities in suspended particles are lower at old station Hale Aloha than at stations K1 and K3 ( Figure 5 ). The Ba_ex data that we report here for stations K1 and K3 agree with the values reported by Dehairs et al. (2008) at station K2 in the Western Subarctic Gyre during the VERTIGO project, with maximum values being approximately 500 pmol kg^-1 in the upper 500 m. Dehairs et al. (2008) report Ba_ex concentrations that were lower at station Aloha than at station K2, a pattern that is similar for the ²²⁶Ra_ex activities reported here. We also report ²²⁸Ra_ex activities determined at stations K1 and K3 ( Figure 5 ). The ²²⁸Ra_ex activities are higher in the upper 500 m of the water column, a pattern that was also observed in other oceanic basins (van Beek et al., 2007; van Beek et al., 2009). We will later compare the ²²⁸Ra/²²⁶Ra ratios of the suspended particles to the dissolved ²²⁸Ra/²²⁶Ra ratios (see Ra Isotopic Ratios (²²⁸Ra/²²⁶Ra)).

The ²²⁶Ra/Ba ratios in suspended particles are higher in the upper water column and decrease with increasing water depth ( Figure 6 ). Two samples collected at station K3 in the upper 500 m display ²²⁶Ra/Ba ratios that are higher than the 2.3 dpm µmol^-1 (i.e. 3.2±0.3 dpm µmol^-1 at 250 m and 2.9±0.3 dpm µmol^-1 at 400 m). At station K1, two samples collected in waters above 150 m display high ratios (3.0 dpm µmol^-1 at 30 m and 110 m), but ratios close to 2.3 dpm µmol^-1 were also found at 80 m, 150 m and 230 m, indicating spatial heterogeneity in the composition/origin of the suspended particles.

Below 400-500 m, the ²²⁶Ra/Ba ratios in suspended particles are equal to or lower than the 2.3 dpm µmol^-1 value. Similar patterns were found in other oceanic basins (van Beek et al., 2007; van Beek et al., 2009). The higher ²²⁶Ra/Ba ratios found in suspended particles compared to the ratio in seawater at the same depth indicates that fractionation between ²²⁶Ra and Ba takes place during formation of suspended particles in the water column, contrary to earlier assumptions (Wolgemuth and Broecker, 1970; Ku and Luo, 1994).

The sequential leaching method of Ganeshram et al. (2003) was applied to an entire vertical profile of suspended particles collected with Niskin bottles (station K3) in order to provide additional information on the different Ba carriers within the water column ( Table 1 ). In surface waters down to 110 m, Ba is mostly found in the fractions released by the first two leaching steps (MilliQ water and 10% HNO₃), which suggests that the Ba pool in surface waters down to 110 m is relatively labile ( Figure 7 ). The labile pool of Ba at these depths likely consists in i) labile organic matter, including potentially EPS, ii) acantharians that are generally concentrated in surface waters (Michaels, 1988; Michaels et al., 1995) and iii) oxy-hydroxides. This labile pool of Ba then decreases with increasing water depth such that below 200 m, barite appears to be the dominant Ba carrier. This pattern thus agrees with the view that the decrease in the labile pool of Ba - following the decay of organic matter or dissolution of acantharian celestite - provides Ba in the microenvironment for barite precipitation in subsurface water (Chow and Goldberg, 1960; Dehairs et al., 1980; Legeleux and Reyss, 1996; Ganeshram et al., 2003; Sternberg et al., 2005; Martinez-Ruiz et al., 2020). The Ba_ex concentrations (non lithogenic Ba) are often considered to reflect barite, although Ba_ex may also include Ba associated with the different fractions invoked above; these latter Ba carriers can be considered as relatively minor contributors below 110 m ( Table 1 ). The Ba_ex concentrations determined in suspended particles collected using in situ pumps below the surface down to 2000 m (250-550 pmol kg^-1; Figure 5 ) are in relatively good agreement with the Ba concentrations associated with BaSO₄ in fraction #3 (200-630 pmol kg^-1; Figure 7 ) over the same depth interval, with the exception of a sharp peak at 250 m found in the sequential leaching experiment (1680 pmol kg^-1). This discrepancy may reflect spatial variability that leads to slightly different samples collected by the two methods (Niskin bottles versus in situ pumps). Note that such high particulate Ba concentrations (1680 pmol kg^-1) have already been reported by previous studies in other oceanic regions (Jacquet et al., 2007; van Beek et al., 2009) and correspond to a depth where the rate of barite precipitation is believed to be highest (e.g., Legeleux and Reyss, 1996). Maximal Ba_ex and ²²⁶Ra_ex concentrations were also found at this depth ( Figure 5 ). The Ba concentration associated with BaSO₄ remains relatively high down to 2000 m (629 pmol kg^-1; Figure 7 ), a pattern that may be related to the saturation state of intermediate waters with respect to barite in the North Pacific that preserves barite or limits its dissolution while settling ( Figure 3 ). At 5000 m, the Ba concentration associated with BaSO₄ is significantly lower than in the upper 500 m (166 pmol kg^-1), suggesting dissolution in water undersaturated with respect to barite. The pool of Ba associated with refractory organic matter decreases from the surface to the bottom. While the sharp decrease in the labile Ba pool in the upper 500 m may be the main contributor to barite precipitation in the upper water column, the release of the Ba associated with refractory organic matter at depth may also contribute to barite precipitation in waters below 500 m (van Beek et al., 2007).

The Sr concentration and the Sr/Ba ratio was also investigated in fraction #3 that is associated to BaSO₄ ( Table 1 ). High Sr concentrations and high Sr/Ba ratios (1.0-1.1 mol/mol) were found in samples collected in surface waters (40 m and 110 m). Since acantharian skeletons are expected to dissolve in the first two leaching steps, the high Sr concentrations found in fraction #3 are unlikely from acantharians, but must be included in BaSO₄. The high Sr content in these barite samples may indicate that barite found at these depths precipitated in microenvironments where significant dissolution of celestite took place and/or following a prior step of Sr-Ba bioaccumulation by EPS (since EPS was shown to concentrate Sr in addition to Ba; Martinez-Ruiz et al., 2020). Below 250 m, the Sr/Ba ratios are lower (0.019-0.085 mol/mol). These latter values are in good agreement with the ratios reported in barite separated from sediments collected in the Pacific Ocean (0.03-0.05 mol/mol; van Beek et al., 2003). This observation further supports the association of fraction #3 with BaSO₄. The different Sr/Ba ratios in BaSO₄ found either above or below 110 m may indicate a larger influence of the dissolution of acantharian skeletons - that releases Ba, Sr and Ra into the microenvironments - in barite formation in the upper 110 m or an evolution of the ratio following dissolution of barite during settling. Like the Sr/Ba ratio in fraction #3, the ²²⁶Ra_ex/Ba_ex ratio in suspended particles is also higher in the upper water column and decreases with increasing depth ( Figure 6 ). This pattern is also observed in the Sr concentrations and Sr/Ba ratios determined in suspended particles collected with in situ pumps ( Table 4 of the Supplementary Material ). The latter Sr/Ba ratios being higher than the Sr/Ba ratio determined from the sequential leaching in fraction #3 may be explained by the different Sr and Ba contributors in these samples (i.e. residual seawater, carbonates, barite, celestite), which would serve to mask the Sr/Ba ratio in barite. Similar relationships were found in the upper 150 m between Ra, Sr and Ba at the OFP station in the Sargasso Sea and at the DYFAMED station in the western Mediterranean Sea: high ²²⁶Ra/Ba ratios were found in both suspended and sinking particles with ratios being much higher than the ambient seawater ²²⁶Ra/Ba ratios. Values up to 3.8 dpm µmol^-1 were observed at 120 m in the Sargasso Sea (van Beek et al., 2007) and up to 8.4 dpm µmol^-1 at 140 m in the western Mediterranean Sea (van Beek et al., 2009), at depths where acantharians are found. In these latter studies, the relationship with acantharians was further suggested by the high particulate Sr content determined in the suspended particles in the upper 150 m of the two regions (van Beek et al., 2007; van Beek et al., 2009) and by the observation of acantharian skeletons on filters collected in the western Mediterranean Sea (van Beek et al., 2009).

The vertical gradient of the dissolved ²²⁸Ra/²²⁶Ra ratio in the water column allows us to investigate the depth of particulate Ba formation in the water column (Legeleux and Reyss, 1996; van Beek et al., 2007; Figure 8 ). The much higher dissolved ²²⁸Ra/²²⁶Ra ratios in the upper 500 m at station K3 compared with station K1 are due to higher dissolved ²²⁸Ra activities ( Figure 9 ) resulting from the northward extension of the Kuroshio to station K3 in 2002 (Oka et al., 2015; Figures 2 and 3 of Supplementary Material ). These waters acquired ²²⁸Ra when interacting with the sediments of the continental shelf when they were flowing northeastward along Honshu (Kaufman et al., 1973; Kawakami and Kusakabe, 2008; Charette et al., 2013). Consequently, the ²²⁸Ra_ex signature of the suspended particles is also higher at station K3 in comparison to station K1 ( Figure 5 ). At station K1, the ²²⁸Ra/²²⁶Ra ratios in the suspended particles collected in the upper 500 m generally match the ²²⁸Ra/²²⁶Ra ratios in the dissolved phase ( Figure 8 ): the ²²⁸Ra/²²⁶Ra ratios in the particulate phase in the upper 500 m follow the profile of the dissolved phase. This suggests that the particulate Ba collected at each of these depths was formed at these water depths and thus incorporated the ²²⁸Ra/²²⁶Ra signature of the dissolved phase. This observation agrees with previous studies conducted in other oceanic basins (van Beek et al., 2007; van Beek et al., 2009). At station K3, the ²²⁸Ra/²²⁶Ra ratios in the particulate phase in the upper 500 m are slightly higher than the dissolved phase, but remain close to the seawater ratios. Like at station K1, this may suggest that the particles incorporated the ²²⁸Ra/²²⁶Ra signature of the dissolved phase at these depths.

Below 500 m, the ²²⁸Ra/²²⁶Ra ratios in the particulate phase are higher than the seawater ratios at the corresponding depth at both stations K1 and K3 suggesting that particulate Ba originates from shallower depths, where it acquired a higher ²²⁸Ra/²²⁶Ra ratio. At these depths, the particulate Ba is mostly associated with barite ( Figure 7 ) that are found at depths where the waters are in equilibrium or supersaturated with respect to barite ( Figure 3 ). This indicates that the particulate Ba at intermediate depths (500-1500 m) mostly consists of settling barite that was formed in the upper 500 m of the water column. Such high particulate ²²⁸Ra/²²⁶Ra ratios indicating an export of barite formed in shallower waters were previously observed in other oceanic basins but only in sinking particles collected with sediment traps (Legeleux and Reyss, 1996; van Beek et al., 2007; van Beek et al., 2009) and not in suspended particles (e.g. van Beek et al., 2007). The sharp decrease in the Ba concentration associated with barite (fraction #3) below the Ba peak at 250 m ( Figure 7 ) - that indicates a maximum rate of barite formation at these depths - suggests that a large fraction of barite then dissolves in the upper 500 m of the water column (that were shown to be undersaturated with respect to barite) and are not exported to greater depth.

The general relationship between Ba and ²²⁶Ra ( Figure 1 of Supplementary Material ) led Chan et al. (1976) to propose that there was a constant ratio between Ba and ²²⁶Ra in the ocean above 2500 m (4.6 nmol ²²⁶Ra/mol Ba, equivalent to 2.3 dpm µmol^-1). The data reported here indicate that ²²⁶Ra/Ba ratios significantly lower than 2.3 dpm µmol^-1 can be found in the upper 500 m of the water column. A similar trend was observed in the Sargasso Sea (Ocean Flux Program Time-Series; van Beek et al., 2007; Figure 10 ). van Beek et al. (2009) also reported low ²²⁶Ra/Ba ratios (1.3-1.8 dpm µmol^-1) at the DYFAMED Time-Series in the Mediterranean Sea, but these low ratios were found throughout the water column (van Beek et al., 2009; Figure 10 ). Data from the Mediterranean Sea thus clearly fall below the regression obtained with the GEOSECS data ( Figure 10B ), a pattern that may - at least partly - be explained by the anti-estuarine circulation in the Mediterranean Sea that fills the entire basin with water originating from the upper water column of the Atlantic Ocean (expected to display a low ²²⁶Ra/Ba ratio, following van Beek et al., 2007). The data from the Mediterranean Sea also clearly highlight the temporal variability of the ²²⁶Ra/Ba ratio at the DYFAMED station.

Many of the vertical profiles obtained during GEOSECS do not show lower ²²⁶Ra/Ba ratios in the upper water column. This could be attributed to the lower sampling resolution and precision of the GEOSECS Ba and ²²⁶Ra measurements (Ku and Luo, 1994). We note however, that when all the ²²⁶Ra/Ba profiles obtained during the GEOSECS program for each ocean basins are plotted versus depth, the lowest ²²⁶Ra/Ba ratios are invariably found in the upper 500 m of the water column (left of the vertical line on Figure 11 ). Likewise, as noted before (Wolgemuth and Broecker, 1970; Chan et al., 1976; Ku and Luo, 1994), the linear regression of ²²⁶Ra activity versus Ba concentration has a significant, positive intercept on the Ba axis ( Figure 1 of the Supplementary Material ). The pattern reported here (i.e., low ²²⁶Ra/Ba ratios in the upper 500 m) does not appear universal, however. In particular, data from Foster et al. (2004) and Staubwasser et al. (2004) indicate that surface waters in the Ross Sea have ²²⁶Ra/Ba ratios similar to deep waters. This pattern is consistent with the deep convective mixing that dominates this region, which may mask fractionation observed in other regions. This is further confirmed by the lack of an intercept when regressing ²²⁶Ra activity and Ba concentration measured in samples from the Southern Ocean (Li et al., 1973).

The intercept on the regression of ²²⁶Ra activity versus Ba concentration has been attributed to the differences in the source of the two elements (Wolgemuth and Broecker, 1970; Chan et al., 1976; Ku and Luo, 1994). While Ba is mainly added to the upper ocean by rivers and potentially also by SGD, ²²⁶Ra is mostly added to the ocean by diffusion from deep-sea sediments. Simple two box models show that such differences in input can generate a lower ²²⁶Ra/Ba in the upper water column (Chan et al., 1976). However, the difference in the ratio produced by these simple models is smaller than those observed in the measurements reported here, suggesting that other mechanisms contribute to the observed vertical changes in Ra/Ba.

Barite, celestite, EPS and hydroxides (e.g. Fe, Mn hydroxides) are known to incorporate or adsorb Ba and Ra. Barite and celestite are believed to be the main Ba carriers in the water column, and to explain the vertical profiles of particulate Ba_ex and ²²⁶Ra_ex, which have higher concentrations in the upper ca. 500 m. The two main Ba (and Ra) carriers are not found at exactly the same water depth: the presence of celestite is mostly restricted to the upper 150 m, whereas barite is usually found below that depth (maximum concentration often observed at ca. 200-250 m). The high ²²⁶Ra/Ba ratios in suspended particles reported in this study were found both at shallow depths dominated by acantharians, as deduced from elevated particulate Sr concentrations (30 m and 110 m, station K1) and barite (250 m and 400 m, station K3). Because celestite is expected to be enriched in radium compared to barium (Bernstein et al., 1998), the high ²²⁶Ra/Ba ratios found in suspended particles in the upper ca. 150 m may be attributed to acantharian skeletons; such a fractionation process may thus contribute to decrease the dissolved ²²⁶Ra/Ba ratios in the upper water column. As deduced from the leaching experiments, barite (fraction #3) formed in the upper 110 m displays high Sr/Ba ratios, suggesting a potential link between acantharian skeletons and barite formation at these depths. The high ²²⁶Ra/Ba ratios found in the suspended particles at shallow depths may thus be related to celestite and to barite precipitating in microenvironments in which celestite - with a high ²²⁶Ra/Ba ratio - rapidly dissolves (Bernstein et al., 1987; Bernstein et al., 1992). Deeper in the water column, the leaching experiments showed that the decrease in the labile pool of Ba down to ca. 250 m (e.g. decay of labile organic matter) was accompanied by increased barite formation. Fractionation between Ba and Ra may thus take place during the bioaccumulation phase of Ba-Ra prior to barite formation within the microenvironments (e.g. adsorption onto bacterial biofilms; Martinez-Ruiz et al., 2020) or during barite precipitation itself. Below 500 m, the ²²⁶Ra/Ba ratios in suspended particles are lower than the ambient seawater ²²⁶Ra/Ba ratios in the three studies that have investigated both particulate and dissolved Ba and Ra (van Beek et al., 2007; van Beek et al., 2009; this study). The ²²⁶Ra/Ba ratios may thus be impacted during dissolution of barite that settles in undersaturated waters.