The deepening of the Reykjanes Ridge southward from Iceland is associated with several fracture zones, including the BFZ at 57∘ N (Fig. 1). The BFZ axis extends quasi-zonally from the Iceland Basin to the Irminger Sea and intersects the rift valley of the Reykjanes Ridge at 56.75∘ N, 34.17∘ W. The BFZ entrance is composed of a narrow sill (8.7 km wide – following the 2000 m isobaths) at 56.73∘ N, 33.72∘ W along the eastern side of the rift valley. This eastern sill reaches a bottom depth of ∼ 2150 m (Fig. 1c). Immediately west of the eastern sill, the axis of the BFZ intersects the rift valley of the Reykjanes Ridge that is oriented northeast–southwest and reaches bottom depths larger than 2500 m. A second sill at the BFZ exit is located on the western side of the rift valley (56.75∘ N, 35.55∘ W). It reaches a bottom depth close to that observed for the eastern sill, although it is larger when considering the 2000 m isobaths (11.3 km). At the approach of the Irminger Sea, the western sill connects to two channels deeper than 2500 m and separated by a seamount with a summit that lies at about 1700 m.

This study is based on high-resolution measurements obtained during four cruises between 2015 and 2018 at key locations of the BFZ (Fig. 1 and Table 1). Two sections are localized within the BFZ. The East section was occupied at the eastern entrance of the BFZ in 2015, 2016 and 2017. Depending on the cruises, it was located either at or 5.5 km upstream of the eastern sill. The Middle section was occupied in the rift valley in 2015 and 2018. At the exit of the BFZ, the West section intersected the two channels west of the western sill in 2015. These three hydrographic sections were designed to study the deep circulation and associated evolution of ISOW that enters the BFZ from the Iceland Basin, circulates in the rift valley of the Reykjanes Ridge and exits the BFZ toward the Irminger Sea.

Twenty-one CTDO2 (conductivity, temperature, depth and oxygen) stations were first implemented along the three sections (stations 16–20 and 96–107) in June 2015 on the French N/O Thalassa during the RREX15 cruise (Branellec and Thierry, 2016). It consists of six stations at the East section, six stations at the Middle section and four stations at the West section. The nominal station spacing was less than 2 km along the East and Middle sections and less than 10 km along the West section. We also use station 122 carried out at 52.7∘ N, 35.1∘ W during the RREX15 cruise for a comparison with the hydrographic properties in the CGFZ.

A year later, the BOCATS16 cruise carried out five CTDO2 stations along the East section in July 2016 on the Spanish B/O Sarmiento de Gamboa (stations 105–109; Branellec and Lherminier, 2017).

In addition to these two cruises, the RREX17 cruise carried out six CTDO2 stations at the East section in August 2017 on the French N/O l'Atalante (stations 86–91; Branellec and Thierry, 2018). The stations were 5.5 km west of those carried out during RREX15 and BOCATS16 to intersect the eastern sill of the BFZ. For simplicity, we refer to this section as the East section as well.

Finally, the OVIDE18 cruise carried out six CTDO2 stations along the Middle section in July 2018 on the French N/O Thalassa (stations 103–108; Branellec et al., 2019).

During the four cruises, CTDO2 stations were also carried out along the so-called OVIDE line (Fig. 1a) that intersects the Reykjanes Ridge at 58.8∘ N, 31.3∘ W (Lherminier et al., 2007). Here, we consider the stations on the eastern side of the Reykjanes Ridge, from 27.2 to 30.7∘ W, to analyze the along-stream evolution of ISOW upstream of the BFZ.

The accuracies of the temperature, practical salinity, pressure and dissolved oxygen concentration are better than 0.002 ∘C, 0.002, 1 dbar and 1.5 µmol kg-1 for the four cruises, except for the dissolved oxygen concentration, for which the accuracy was estimated at 2 µmol kg-1 for OVIDE18. Figures A1 and A2 show the salinity and potential temperature of the East and Middle sections.

The CTD rosette was equipped with a 300 kHz downward-looking and a 150 kHz upward-looking L-ADCP (lowered acoustic Doppler current profiler, RD Instruments). The upper layer current velocity components were measured by two S-ADCPs (shipboard acoustic Doppler current profiler, RD Instruments) operating at 38 (OS38) and 150 kHz (OS150), with the exception of BOCATS16 for which S-ADCPs operated with a combination of 75 (OS75) and 150 kHz (OS150). The maximum depths reached by the S-ADCP signal were 1300–1400 m for the OS38, 700–800 m for the OS75 and 200–250 m for the OS150. The calibrations and processing of these measurements were identical for all four cruises. As described by Petit et al. (2018), the OS150 datasets were used to correct the calibration of the second S-ADCP, and the velocity profiles were averaged over 2 km segments along the sections.

These measurements were used to estimate absolutely referenced geostrophic velocities perpendicular to the sections and the associated gridded transports (Fig. 2 and Table 3). The geostrophic velocities were estimated from the CTDO2 measurements using the seawater toolbox and were referenced by velocity measurements from the OS38 and OS75 (Petit et al., 2018). The geostrophic velocities were then evaluated by comparison to the L-ADCP velocities at the East and Middle sections (Figs. A3 and A4). Note that L-ADCP measurements provide local measurements of the total velocity field at each hydrographic station while geostrophic velocities are averaged velocities between two successive stations (Lherminier et al., 2007), so we cannot expect a perfect agreement between the two datasets. Nevertheless, the comparison of the amplitude and spatial structure of the velocity field reveals the remarkable agreement between the geostrophic and L-ADCP velocities for each section. The largest difference in velocity is found at the East section in 2017 where the geostrophy-based velocity in the ISOW core is about -0.36 m s-1, while the L-ADCP-estimated velocity is about -0.14 m s-1.

Data from three Deep-Arvor floats are used to assess the pathway and hydrographic evolution of ISOW as it flows westward through the BFZ. The Deep-Arvor floats are Argo floats profiling down to 4000 m (Le Reste et al., 2016; Le Traon et al., 2020) and returning top-to-bottom profiles of temperature, salinity and dissolved oxygen concentration as a function of pressure every 10 d. One Deep-Arvor float (WMO 6901603) was deployed at the East section during RREX17 (station 89 at 56.73∘ N, 33.72∘ W), and two Deep-Arvor floats (WMO 6902881, 6902882) were deployed simultaneously at the Middle section during OVIDE18 (56.80∘ N, 34.17∘ W) (Table 2). The Deep-Arvor float 6901603 completed 12 cycles before it died on 26 November 2017. Floats 6902881 and 6902882 completed a few cycles in the BFZ before exiting the fracture zone and drifting into the Irminger Sea. The three floats drifted mostly in the range 1900–2100 dbar within the ISOW core layer. Float 6901603 grounded at cycles 2, 11 and 12 and drifted at shallower pressure during these cycles. Trajectories within the ISOW layer of the Deep-Arvor floats are considered from cycle 2 because the floats are programmed to rise to the surface immediately after reaching their parking depth at the end of the first descent.

Temperature, practical salinity and pressure were measured using a Seabird SBE41CP CTD sensor with a target accuracy of 0.002 ∘C, 0.004 and 7 dbar, respectively. The salinity data were corrected following the procedure specified for Deep-Argo floats (Cabanes et al., 2016; Wong et al., 2021). A compressibility term (CPcor) was applied to correct a pressure-dependent conductivity sensor bias. For the three floats, the nominal CPcor value provided by Seabird (CPcor = -9.57 × 10-8 dbar-1) is replaced by the Argo recommended value, CPcor = -13.5 × 10-8 dbar-1. No further salinity bias or drift was detected for 6901603 and 6902882, while 6902881 was corrected for a salinity bias of 0.002. This bias was determined by comparison with a ship-based calibrated cast acquired at float deployment.

The potential temperature–salinity (θ–S) properties of ISOW at the BFZ are compared to those observed in the East Reykjanes Ridge Current along the OVIDE section (Fig. 3). The comparison in the density range 27.8–27.87 kg m-3 reveals three geographical areas along the OVIDE section that are related to the three ISOW branches permanently observed at 30, 29 and 27∘ W along the eastern flank of the Reykjanes Ridge (Daniault et al., 2016; Xu et al., 2010; Petit et al., 2019). ISOW east of 29∘ W is fresher than that observed in the BFZ, while ISOW west of 30∘ W has θ–S properties close to those observed in the BFZ. The θ-S properties between 29–30∘ W are variable in time, being fresher than at the BFZ in 2015 and 2016 but matching the ISOW properties in the BFZ in 2017 and 2018. The BFZ is thus mainly fed by the western branch of ISOW flowing west of 30∘ W along the 2150 m isobath, which is the depth of the sill at the eastern entrance of the BFZ. The BFZ is also fed by the middle branch of ISOW flowing between 29–30∘ W in 2017 and 2018. However, while ISOW at the eastern entrance of the BFZ is lighter than 27.87 kg m-3, ISOW between 29–30∘ W reaches denser values. Then, only the lightest variety of ISOW in this branch enters the BFZ.

Once in the BFZ, the following analysis relies on a comparison of the velocity fields and hydrographic properties observed in the ISOW layer between the East and Middle sections and along the pathway of the Deep-Arvor float 6901603. The geostrophic velocity section acquired in 2016 is not discussed here because the sampling of the stations, limited to the two flanks of the channel, was not appropriate to capture the ISOW transport at depths higher than ∼ 1500 m within the sill.

A comparison of the velocity fields observed in 2015 and 2017 in the ISOW layer of the East section reveals differences in the amplitude and location of the ISOW through-flow (Fig. 2). In 2015, the westward flow is intensified at the center of the sill (56.74∘ N), while a weak eastward flow is observed along its northern wall (56.77∘ N). In 2017, the westward flow is observed all along the section and is intensified along the northern wall of the sill. Our dataset does not allow us to determine whether these differences are due to temporal variability of the inflow from the Iceland Basin or to differences in the local bathymetric constraints within the narrow channel of the BFZ entrance, as the East section in 2015 is localized slightly upstream in the channel compared to the East section in 2017 (Fig. 1b). The ISOW transport is estimated at -0.7 ± 0.1 Sv in 2015 and -0.5 ± 0.2 Sv in 2017, which represents 56 % and 49 % of the top-to-bottom transport, respectively (Table 3). The similar ratio of ISOW to top-to-bottom transports between 2015 and 2017 suggests that the BFZ through-flow is influenced by local barotropic circulation, as observed at the CGFZ (Bower and Furey, 2017; Bower and von Appen, 2008; Racapé et al., 2019).

At the Middle section, the velocity field in the ISOW layer is composed of a westward flow along the northern wall of the rift valley and an eastward flow along its southern wall in 2015 and 2018 (Fig. 3). This circulation at the Middle section, following the flanks of the section, suggests that ISOW flows along a cyclonic loop in the rift valley of the Reykjanes Ridge, possibly driven by the bathymetry. The associated top-to-bottom and ISOW transports are 5 times weaker in 2018 than in 2015 (Table 3), although the ISOW transport represents the same proportion of the top-to-bottom transport for these two years (43 % in 2015 and 44 % in 2018). The reduced ISOW transport in 2018 is due to an overall reduced westward flow combined with an enhanced eastward flow, which is composed of an additional core of ISOW at 56.78∘ N that is not observed in 2015.

The hypothesis of a cyclonic pathway for ISOW in the rift valley is supported by the trajectory of the Deep-Arvor 6901603 float in summer 2017 (Fig. 1d). The float drifted northward at 1200 dbar during its first two cycles from the eastern entrance of the BFZ (Table 2). Although this parking depth is not associated with ISOW density, we note that its pathway is parallel to the 2100 m isobath of the seamount north of the sill. North of the seamount, the float drifted within the ISOW layer at 2100 dbar and reached the northern part of the rift valley at cycle 5. Finally, the float drifted mainly southwestward and then westward as it crossed the western sill of the BFZ at its last cycle. Hence, the trajectory of the 6901603 Deep-Arvor float sketches a cyclonic pathway in the rift valley of the Reykjanes Ridge during its 12 cycles.

The transformation of ISOW in the BFZ is now analyzed by comparing the θ–S properties of ISOW between the East and Middle sections in 2015 (Fig. 4a–b). The θ–S range of variation on a given isopycnal does not exceed 0.01 ∘C and 0.005 between the East and Middle sections. This shows that ISOW does not exhibit a significant evolution of its hydrographic properties by isopycnal mixing between the two sills of the BFZ in 2015. This is also true when considering the Deep-Arvor float properties (not shown). A decrease of ∼ 0.015 kg m-3 in density of the near-bottom water is however observed at the Middle section. This can possibly be ascribed to the bathymetry of the sill at the eastern entrance of the BFZ, in case the densest water cannot overflow this topographic obstacle at 2150 m depth (black star in Fig. 4b). This is however not supported by the velocity section at the East section in 2015 (Fig. 2), which shows westward velocities down to the bottom, so that we rather interpret this property change as evidence of vertical mixing.

To better show this vertical mixing, we now consider the distribution of ISOW transport in temperature bins of 0.1 ∘C at these two sections (Fig. 4c–d). Their comparison reveals a net decrease in ISOW transport in the temperature bins 3.7 and 3.2 ∘C from the East to the Middle section. On the contrary, the ISOW transport significantly increases in the intermediate temperature bins 3.3–3.5 ∘C at the Middle section. The temperature range of the ISOW layer is thus reduced by 0.3 ∘C during its transit through the BFZ, with 77 % of ISOW localized within the 3.3–3.6 ∘C bins at the Middle section. These results highlight that ISOW is mainly homogenized as it flows westward in the BFZ, likely due to a vertical mixing within the ISOW layer.

To assess the evolution of ISOW as the overflow exits the BFZ through the two western channels, we use L-ADCP profiles at the West section in 2015 and two Deep-Arvor floats deployed at the Middle section in 2018, 6902881 and 6902882. The analysis is based on L-ADCP instead of geostrophic velocities because the distance between the stations at the West section is too large to resolve the ISOW flow in these narrow channels.

The data suggest that the main exit pathway for ISOW follows the northern walls of the western channels and that incursion of dense water from the Irminger Sea follows the southern walls, leading to a mixing between NADW and ISOW from the BFZ. Indeed, the RREX15 ship-based data show that each of the western channels is composed of a westward flow along the northern walls (stations 16 and 18) and an eastward flow along the southern walls (stations 17 and 19) in the ISOW layer (Fig. 5). Accordingly, the two Deep-Arvor floats followed the northern wall of the southernmost channel when exiting the BFZ until at least 35.5∘ W (Fig. 1d). The trajectory of float 6902881 became more chaotic further west.

Both the Deep-Arvor floats and the RREX15 ship-based data reveal that ISOW is colder and fresher in the western channels than in the rift valley (Figs. 4b and 6c–d). The RREX15 data additionally show that, for density lighter than σ0 < 27.855 kg m-3, the θ–S properties found along the northern walls of the channels are at intermediate positions between those in the BFZ and those observed along the southern walls of the channels. This suggests that ISOW exiting the BFZ along the northern walls of the channels is isopycnally mixed with colder and fresher NADW circulating in the Irminger Sea. The lower part of the layer (σ0 > 27.855 kg m-3) is fresher along the northern walls than in the upper part. This lower part of the layer cannot be renewed by BFZ through-flow, whose density is lower than 27.855 kg m-3, and must be mainly fed by denser NADW. The ISOW layer at the exit of the BFZ is thus a superposition of dense waters of different origins: the upper part results from the mixing of BFZ through-flow with fresher and colder NADW, while the lower part is mainly composed of dense NADW flowing in the Irminger Sea. Interestingly, the superposition of these dense waters creates a maximum in salinity at 27.83 kg m-3 (Figs. 4b and 6d), which was not observed at the upstream sills.

The Deep-Arvor floats then spread in the Irminger Sea and mix with surrounding water masses along their pathways. The two floats followed similar pathways in an interval of several weeks: heading toward the tip of Greenland and being abruptly diverted northeastward to join the Irminger Current (Fig. 6a–b). However, upon reaching ∼ 39∘ W at the exit of the BFZ, the float 6902882 first headed southward along the Reykjanes Ridge until 54∘ N and then turned back northward until 58∘ N. The southward spreading of this float along the western flank of the Reykjanes Ridge provides an opportunity to investigate the influence of the BFZ through-flow on the NADW properties flowing northward. More precisely, we now compare the θ–S properties of ISOW at the western exit of the BFZ with those in the vicinity of the CGFZ.

The RREX15 ship-based data show that the westward core of ISOW reaches a maximum in salinity at 27.88 kg m-3 at the CGFZ, while the maximum in salinity observed in the western channels of the BFZ is at 27.83 kg m-3 (Fig. 4b). The difference in density for the ISOW core between the two fracture zones is confirmed by the along-track properties of the float 6902882 (Fig. 6d–e). The core of ISOW is eroded toward lower densities as the float leaves the vicinity of the CGFZ to reach BFZ latitudes (Fig. 6e–f). North of the BFZ, the maximum in salinity is observed at lower densities (27.81 kg m-3), and the lower part of the layer (σ0 > 27.855 kg m-3) forms a mixing line between ISOW at 2.8 ∘C/34.94 and DSOW at 1.5 ∘C/34.9. The inflow of DSOW is possibly sourced by the recirculation evidenced by the two floats in the vicinity of Greenland. Hence, the limited northward erosion of the ISOW core reveals the key role of the BFZ through-flow in maintaining a maximum in salinity for the ISOW layer at these latitudes.