Pathways and Water Mass Transformation Along and Across the Mohn-Knipovich Ridge in the Nordic Seas

Atlantic Water takes various pathways through the Nordic Seas, and its transformation to denser waters forms a crucial connection to the lower limb of the Atlantic Meridional Overturning Circulation. Circulation maps often schematize two distinct pathways of Atlantic Water: one following the Norwegian Atlantic Slope Current along the continental slope of Norway and one following the Norwegian Atlantic Front Current along the Mohn and Knipovich Ridges. In this paper, the connectivity between the northward ﬂ ow along these ridges is investigated. Analyzing trajectories of surface drifters and ARGO ﬂ oats, we ﬁ nd that only 8% of the ﬂ oats that travel near the mid ‐ ocean ridges take the frontal pathway to the north. Indeed, by tracing numerical particles in a realistic numerical simulation, part of the water mass traveling along the Mohn Ridge follows the 2,500 m isobath eastward and joins the slope current, instead of ﬂ owing north along the Knipovich Ridge. Furthermore, north of 74°N, frequent exchange between the slope current and the front current is observed. Therefore, the slope current and front current are less isolated than often schematized. Additionally, the observational data set reveals substantial cross ‐ ridge exchange; 31% of the ﬂ oats that travel within 60 km from the mid ‐ ocean ridges cross it. Results from numerical simulations indicate that the cross ‐ ridge exchange leads to cooling and freshening of the Atlantic Water along the front. Deployments of ﬂ oats near the mid ‐ ocean ridges are needed to investigate the pathway of Atlantic Water and its exchange across the ridge in more detail. Circulation (AMOC). A key region for maintaining this circulation is the Nordic Seas, where Atlantic Water is modi ﬁ ed due to large ocean heat losses to the atmosphere. The transformation of Atlantic Water is not uniform, as the water mass can take various pathways. Two main routes are often described: one along the continental slope of Norway and one along the topographic ridges that separate the warm eastern basin from the cold western basin. In this paper, we investigate the latter pathway using a Lagrangian approach. That is, we follow the ﬂ ow, instead of investigating properties at ﬁ xed positions. We ﬁ nd that the ﬂ ow along the mid ‐ ocean ridges is not continuous as often schematized; substantial exchange between the ﬂ ow along the ridge and the ﬂ ow along the continental slope occurs. Furthermore, we observe various trajectories that cross the mid ‐ ocean ridge. Results from numerical simulations indicate that this cross ‐ ridge exchange can lead to enhanced transformation of the Atlantic Water. Therefore, this process is important for setting the properties of water masses that enter the Arctic Ocean and of water masses that contribute to the AMOC.


Introduction
The Nordic Seas are characterized by a warm and saline Atlantic water mass in the east and a fresh and cold Arctic water mass in the west. Due to the anomalously warm and salty eastern basin, more than 60% of the total heat loss in the northern North Atlantic Ocean occurs in the Nordic Seas (Chafik & Rossby, 2019). As such, the transformed water masses leaving the Nordic Seas play a major role for the strength of the Meridional Overturning Circulation (MOC; e.g., Chafik & Rossby, 2019;Lozier et al., 2019). The upper limb of the MOC carries the warm Atlantic Water (AW) into the Nordic Seas via different branches (Figure 1a, Orvik & Niiler, 2002). The AW transforms to denser water masses along its path northward (e.g., Bosse et al., 2018), from where it flows toward the Arctic Ocean or returns south within the East Greenland Current (EGC). This gradual transformation of AW is seen as the main contributor to the variability of the dense water leaving the Nordic Seas via the overflows (i.e., Eldevik et al., 2009;Mauritzen, 1996).
However, the AW in the eastern basin of the Nordic Seas cannot be seen as one homogeneous layer that is modified in a uniform way; upon entering the Nordic Seas, the AW follows various branches that affect different parts of the Nordic Seas (Hansen & Østerhus, 2000). AW circulation in the Nordic Seas is typically schematized by a two-branch structure (Orvik & Niiler, 2002). One branch, the Norwegian Atlantic Slope Current (NwASC, hereinafter the slope current), follows the Norwegian continental slope as a topographically controlled barotropic current. The second branch, the Norwegian Atlantic Front Current (NwAFC, hereinafter the front current), is a baroclinic jet associated with the Arctic Front (AF, Orvik & Niiler, 2002). The AF is characterized by a strong thermocline and halocline that separate the Arctic Water in the west from the AW in the east (Figure 1b Bosse . This density front is aligned with the mid-ocean topographic ridges in the Nordic Seas: the Mohn Ridge and the Knipovich Ridge (see Figure 1a). Along the Knipovich Ridge, the front current is often referred to as the western branch of the West Spitsbergen Current (Walczowski, 2013).
Hydrographic measurements at various sections across the AF reveal strong horizontal gradients in temperature and salinity along the topographic slope and show a clear northward flow along the ridges (i.e., Rossby, Ozhigin, et al., 2009;van Aken et al., 1995;Walczowski, 2013). Recent estimates from glider transects by Bosse and Fer (2019) indicate that the AW transport along the front current is comparable to the transport along the slope current (∼3.2 Sv). It is therefore surprising that barely any surface drifters (Orvik & Niiler, 2002;Poulain et al., 1996) or RAFOS floats (Rossby, Prater, et al., 2009) follow this frontal pathway depicted in Figure 1a all the way through the Nordic Seas. The conclusion drawn by Orvik and Niiler (2002), that the NwAC maintains a two-branch structure throughout the Nordic Seas, was based on a composite of many partial pathways of surface drifters. Therefore, it can be questioned whether these pathways also apply to water parcels that come from south of 65°N and travel all the way to the Arctic. The fact that very few surface drifters follow the Mohn-Knipovich Ridge may be attributed to the nonuniform deployment and relatively large mortality of surface drifters in the Nordic Seas (Koszalka et al., 2011(Koszalka et al., , 2013. But still, it is as of yet unclear whether the northward flow along the Mohn Ridge is connected to the downstream northward flow along the Knipovich Ridge in a Lagrangian sense. Apart from AW modification in the along-front direction, across-front fluxes resulting from baroclinic instability of the front can impact water mass transformation processes. Evidence for such cross-frontal exchange is described by Budéus and Ronski (2009), who observed patches of AW in the Greenland Basin originating from the AF. In addition, analysis of satellite measurements and estimates from surface drifter data show enhanced eddy kinetic energy (EKE) along the Mohn Ridge (Koszalka et al., 2011;Trodahl & Isachsen, 2018). Analyses of hydrographic in situ measurements yield an eddy heat flux of 8-17.7 × 10 12 W across the AF ( van Aken et al., 1995;Walczowski, 2013). That is, ∼30% of the total heat loss of the eastern basins in the Nordic Seas is due to eddy heat exchange with the colder and fresher western basins (Segtnan et al., 2011). The warm anticyclonic eddies shed by the front from east to west can be important for the deep convection processes in the Greenland Basin, as they provide a salt source that can sustain deep convection (Brakstad et al., 2019). In turn, the eddy-induced exchanges across the ridge cool and freshen the AW (Saloranta & Svendsen, 2001), which affects the overturning in the Nordic Seas and the properties of the water masses entering the Arctic.
However, no significant cross-ridge exchange is captured by trajectories of surface drifters (Poulain et al., 1996), subsurface RAFOS floats (Rossby, Prater, et al., 2009), or ARGO floats (Latarius & Quadfasel, 2016;Voet et al., 2010). In fact, Rossby, Ozhigin, et al. (2009) suggest that cross-ridge exchange is unlikely, as the AF density gradient is strong and tightly locked to the sloping topography. Furthermore, analyses of a SF6 tracer, injected in the Greenland basin in 1996, do not show significant mixing between the Greenland Basin and the Lofoten Basin (Messias et al., 2008;Olsson et al., 2005). It is therefore unclear how important the cross-front transport is for water mass transformations in the Nordic Seas.
Successive analyses of the growing Lagrangian observational data set have shown to continuously improve understanding of the Nordic Seas surface and mid-depth circulation and variability Koszalka et al., 2011;Poulain et al., 1996;Rossby, Prater, et al., 2009;Orvik & Niiler, 2002;Voet et al., 2010). Therefore, in this paper we revisit the surface drifter and ARGO float trajectories with a special focus on trajectories along and across the Mohn and Knipovich Ridges. In order to investigate the connection between the flow along the Mohn and Knipovich Ridges and possible pathways across the ridges, results from the observed trajectories are compared to trajectories from a realistic, high-resolution numerical calculation (section 3). Furthermore, the importance of baroclinic cross-ridge exchange for water mass transformation along AW pathways is investigated in an idealized numerical configuration of the Lofoten and Greenland Basins as well (section 4). A discussion and conclusions are provided in section 5.

Methods
When investigating pathways and connectivity between different ocean branches, a Lagrangian approach is most suitable van Sebille et al., 2018). Therefore, we use all trajectories available from surface drifters and ARGO floats that at some moment during their lifetime passed through the Nordic Seas. Details on this data set and the selection procedure to analyze the flow along and across the Mohn and Knipovich Ridges are provided in section 2.1. The observed trajectories are limited not only in quantity but also in their capability to trace water masses. As ARGO floats and surface drifters float at a fixed depth, they are unable to represent any vertical displacement of water masses, which is likely to occur in the Nordic Seas. Therefore, we performed additional Lagrangian studies to evaluate pathways and water mass transformation of AW along the mid-ocean ridges using both a high-resolution, realistic ocean model (section 2.2) and an idealized numerical simulation (section 2.3).

Trajectories From Observations
The surface drifter data set is obtained from the Global Drifter Program (GDP) database (https://www.aoml. noaa.gov/phod/gdp/ updated through 30 June 2019 when downloaded on 3 December 2019). The surface drifters consist of a buoy at the surface and a subsurface drogue, and their trajectories represent the circulation at 15 m depth. Trajectories of surface drifters that have lost their drogue are not taken into account. A quality control and interpolation of each drifter position to 6 hr intervals were performed by the AOML/NOAA Drifter Data Assembly Center (Lumpkin & Pazos, 2007). In contrast to surface drifters, ARGO floats typically drift at 1,000 m depth. Every 10 days, they descend to 2,000 m depth and subsequently ascend to the surface profiling the water column. At the surface, they transmit their data and position to a satellite and return to their parking depth. Tracing the profiling location of ARGO floats visualizes the circulation at mid-depth (Voet et al., 2010). The ARGO float data are obtained from the International Argo Program (downloaded on 18 November 2019 from https://coriolis.eu.org), and only quality-controlled profiles are taken into account.
At time of download (3 December 2019), 564 surface drifters and 283 ARGO floats had been drifting through the Nordic Seas. To investigate the flow along and across the Mohn and Knipovich Ridges, trajectories that came in close vicinity to the mid-ocean ridges are selected. To this end, it is assumed that the front current flows along the 2,250 m isobath (Bosse & Fer, 2019), which is schematized by a straight line in Figure 2. The width of the front is estimated to be ∼40 km (Bosse & Fer, 2019), but the position of the front itself can be variable. Therefore, an area is defined by a 60 km radius from the 2,250 m isobath (red shading in Figure 2). Every float or drifter entering this area is used for analysis. This way, only 40 surface drifters and 50 ARGO floats remain, which allows for subsequent separation based on individual pathways by eye.

Trajectories From a Realistic Ocean Model
As the number of surface drifters and ARGO floats in the vicinity of the mid-ocean ridges is relatively low, results from the observed trajectories are compared to trajectories from a realistic ocean model: the Regional Ocean Modeling System (ROMS) applied to the eastern Nordic Seas (Figure 3a). The model has 800 m horizontal resolution and 60 vertical levels, ranging from 2-5 m at the surface to 60-70 m near the bottom. Details on the model configuration can be found in Dugstad, Koszalka, et al. (2019). Thanks to its high spatial resolution, the model resolves mesoscale and some of the submesoscale processes in the Nordic Seas and captures the variability and dynamics of the area well (Dugstad, Koszalka, et al., 2019;Trodahl & Isachsen, 2018).
We use the Lagrangian data set obtained by Dugstad, Koszalka, et al. (2019), who created this data set to investigate variability of the inflow to the Lofoten Basin. They used the 6-hourly output of the 3-D velocity fields from the realistic model, spanning years 1996 to 1999, to advect numerical particles. These particles were deployed in a rectangular grid of 40 × 40 drifters (white dots, Figure 3a) every week for 3 years at three different depth levels (15, 200, and 500 m). This spatially uniform deployment allows for a better representation of the flow, in contrast to the observed drifters and floats that are deployed along specific sections and in locations, which can lead to limited and biased statistics (Dagestad et al., 2018). The particles were advected for 1 year using the OpenDrift tool with a fourth-order Runga Kutta time stepping scheme (Dagestad et al., 2018, OpenDrift is an open source code that can be accessed at https://github.com/OpenDrift/opendrift). No explicit horizontal or vertical diffusion was added. Dugstad, Koszalka, et al. (2019) showed that turbulent transport processes are well captured by this data set.
Here, the data set is analyzed to study the pathways of AW along the mid-ocean ridge to the north. As we are only interested in AW, deployments northwest of the Mohn Ridge are not taken into account. Moreover, only particles that leave the domain within the advection time and were deployed at 200 m depth are considered. In this way, 115,515 trajectories are selected. The mean northward velocity is used to separate the trajectories further based on where the particles leave the domain (see solid red lines in Figure 3b). Particles that flow along the continental slope of Norway leave the domain between 9.5°E and 16.4°E; particles that follow the Knipovich Ridge leave the domain between 2.5°E and 9.5°E.

Trajectories From an Idealized Numerical Configuration
As the realistic ocean model simulates only part of the eastern Nordic Seas, an idealized configuration of the Massachusetts Institute of Technology general circulation model (MITgcm, Marshall et al., 1997) is used to shed light on the connection between the eastern (Lofoten) and western (Greenland) basin. Furthermore, the importance of the frontal instability for AW transformation is investigated using this model. Although  idealized, the model captures the dynamics of the Nordic Seas pertinent for this study (Ypma et al., 2020). The model has a 5 km horizontal resolution and 33 vertical levels ranging from 20 m thickness at the surface to 200 m thickness near the bottom. The warm eastern basin is separated from a colder western basin by a mid-ocean ridge at 1,200 m depth ( Figure 4). Details on the model configuration, as well as on the hydrodynamic properties of the circulation, can be found in Spall (2011) and Ypma et al. (2020).
The model domain consists of two parts separated by an island. The area north of the island represents the Lofoten and Greenland Basins. There, the ocean surface temperature is restored to an atmospheric temperature of 4°C with a timescale of 2 months to resemble a realistic surface heat flux (Ypma et al., 2020). Furthermore, temperature and velocity are restored over the full water column south of the island (see black box Figure 4a); no other surface forcing is applied. This induces a buoyant cyclonic boundary current, representative of the AW inflow. This inflow separates into a slope current and a front current, mimicking the NwASC and NwAFC, respectively (arrowheads in Figure 4a). As described in, for example, Spall (2010) and Isachsen (2015), the steepening of the continental slope near the Lofoten Islands is essential to create the observed enhanced eddy activity that induces the advection of AW throughout the eastern basins of the Nordic Seas. The idealized model captures the magnitude of EKE near the steep slope (shading in Figure 4a). Enhanced eddy variability is also found along the unstable front current. The total eddy heat flux across the thermal front in the idealized model is 10 × 10 12 W (Figure 4c), which is of similar magnitude as estimated from observations by van Aken et al. (1995) and Walczowski (2013).
To investigate the water mass transformation of the AW in this model, numerical particles are released every 15 km × 15 km east of the ridge at 200 m depth similar to the approach in the realistic numerical simulation (red dots in Figure 4a). The particles are advected forward in time along their 3-D trajectories using the Connectivity Modeling System (CMS, Paris et al., 2013) within the daily averaged three-dimensional velocity fields of the MITgcm simulation. A time step of 1 hr is used, and the new position of each particle is calculated using a fourth-order Runge Kutta stepping scheme in time and a tricubic interpolation in space. As for the realistic simulation, no horizontal or vertical diffusivity is added to the particles, so the particle motion is purely advective.
Particles are released daily for only 30 days and advected for 4 years. This is sufficient to capture the variability of the eastern basin since the idealized model simulation does not have a seasonal cycle. As we are interested in the dynamics of the front current, only those particles that have traveled within part of the front current are selected (red box Figure 4a and red lines Figure 4b). This results in 59,346 particles that are used for the analysis.

Along and Across Frontal Pathways
In this section, the connection between the frontal currents along the Mohn and Knipovich Ridges is investigated using both the trajectories from surface drifters and ARGO floats (as described in section 2.1) and the trajectories from the realistic numerical simulation (as described in section 2.2). Furthermore, pathways that reveal cross-ridge exchange between the Lofoten Basin and the Greenland Basin are discussed.
Most of the pathways taken by the surface drifters that come near the ridge (red shaded area in Figure 2) follow only part of the mid-ocean ridges as highlighted by the pathways in Figure 5a. From the 36 surface drifters that travel along the mid-ocean ridges, only five follow the path northward along both the Mohn Ridge and the Knipovich Ridge (green trajectories in Figure 5a). All other trajectories display a path that alternates between the slope current and the front current as indicated by the orange and brown trajectories in Figure 5a.  (c, d) The probability that a particle will leave the domain along (c) the slope or (d) the mid-ocean ridge is given by the number of particles that cross each 0.1°lon × 0.1°lat gridbox and leave through the specified transect divided by the total number of particles that cross that gridbox. The 33% probability contours are indicated by the orange (particles that leave along the slope) and green line (particles that leave along the front). Results shown are diagnosed from the realistic numerical simulation (section 2.2).
This exchange between the slope and front currents north of 72°N is also displayed by the trajectories of the ARGO floats (Figure 5b). In total, 10 of the 43 floats flow along the Knipovich Ridge and originate from the slope current. The mean flow at 1,000 m depth along the Mohn Ridge is southward (Voet et al., 2010). So most of the floats that follow the Mohn Ridge travel south (39 of 43 floats) and leave the Lofoten Basin along the Jan Mayen Ridge (27 of 43 floats). Furthermore, as most of the ARGO floats were originally deployed in the basin interior, the floats tend to follow the closed cyclonic circulation within the Lofoten Basin (Voet et al., 2010). It is therefore not surprising that only two ARGO floats show a connection between the northward flow along the Mohn Ridge and the Knipovich Ridge.
As the number of surface drifters and ARGO floats that approach the mid-ocean ridges is limited, trajectories in the realistic model configuration released at 200 m depth (section 2.2) are used to investigate the connection between the slope and front currents in more detail. The numerical particles that leave the model domain toward the north(west) are separated in two pathways: one along the continental slope (36% of the total number of particles; Figure 6a) and one along the Knipovich Ridge (10% of the total number of particles; Figure 6b). Most of the other particles leave the domain toward the southeast along the Jan Mayen Ridge or to the north within the Norwegian Coastal Current (54%).
The normalized particle density distribution in Figure 6 highlights which pathways the numerical particles take depending on where they leave the domain. Just as observed in the trajectories of the surface drifters, some of the trajectories along the Mohn Ridge follow the 2,500 m isobath at 72°N toward the southeast to join the slope current. Vice versa, some of the trajectories that follow the slope current flow along the same isobath toward the northwest to join the flow along the Knipovich Ridge (see example trajectories in Figures 6a and 6b). Therefore, the trajectories in the realistic numerical simulation, in line with the observed trajectories ( Figure 5), suggest substantial exchange between the front current and the slope current, before these branches continue north of 72°N.
The connection between the front and the slope currents is further investigated by computing the probability that a particle at any given location will leave the domain via the continental slope or via the Knipovich Ridge (Figures 6c and 6d). The probability is given by the ratio of the number of particles that cross a gridbox and leave the domain at a specified location to the number of particles that cross that gridbox in total. Therefore, darker shading in Figures 6c and 6d indicates a large probability that a particle in that region will leave via the continental slope (Figure 6c) or via the Knipovich Ridge (Figure 6d).
Along the entire length of the Mohn Ridge, particles display a similar probability (>33%) of leaving the domain via the slope current and leaving the domain along the Knipovich Ridge (striped area in Figures 6c and 6d). This again emphasizes the exchange between the front and slope currents and reveals that the flow along the Mohn Ridge does not form a one-to-one connection with the flow along the Knipovich Ridge.
An interesting aspect of the pathway highlighted by the particle density distribution in Figure 6b is that some particles shift from the eastern flank of the Mohn Ridge to the western flank of the Knipovich Ridge. Furthermore, ∼2% of the total outflow in the realistic numerical simulation consists of particles that cross the mid-ocean ridge from the Lofoten Basin into the Greenland Basin (not shown). The individual trajectories show strong variability along and northwest of the Mohn Ridge. As they originate from the AW layer in the east, they could represent AW from the frontal current detached by frontal instabilities traveling into the Greenland Basin as observed by Budéus and Ronski (2009).
The observed trajectories of surface drifters reveal the instability of the front current as well (Figure 7a). Especially along the Knipovich Ridge, multiple surface drifters cross the summit of the ridge and show meandering or circular (eddying) trajectories near the ridge. The trajectories of the 14 surface drifters shown in Figure 7a all indicate significant cross-ridge exchange between the Lofoten and the Greenland Basins.
Furthermore, 14 ARGO floats are found to cross the mid-ocean ridge from the Greenland Basin into the Lofoten Basin (or vice versa; Figure 7b). However, the cross-ridge exchange at the ARGO float depth is not necessarily related to the frontal instability. A zoom-in of four ARGO float trajectories near the ridge is shown in Figure 8. Three of these floats start in the cyclonic gyre of the Greenland Basin (Figures 8a-8c). When they approach the ridge, they cross, travel southward, and start to follow a

10.1029/2020JC016075
Journal of Geophysical Research: Oceans cyclonic circulation along the outer rim of the Lofoten Basin. The example trajectory in Figure 8d shows similar behavior, except that the float continues its path northward along the Knipovich Ridge.
It is beyond the scope of this study to assess the mechanisms behind the cross-ridge behavior of the ARGO floats. The position of the ARGO floats is only known at specific moments in time and the bathymetry near the mid-ocean ridges in the Nordic Seas is very complex. However, exchange between the western and eastern basins through gaps in the topographic ridges has been hypothesized by e.g., Spall (2010). Although Olsson et al. (2005) and Messias et al. (2008) showed that a fast pathway to exit the Greenland Basin exists via the Jan Mayen Current, the ARGO floats shown here that originate from the Greenland Basin do not seem to have a preferred location where they cross the mid-ocean ridge. Not surprising, they seem to be steered much stronger by topography than the surface drifters that showed cross-ridge exchange. Possibly, enhanced mean flow within canyons as discussed by Ruan and Callies (2019) can lead to the behavior seen by the ARGO floats.
In conclusion, results from surface drifter trajectories (Figure 5a), ARGO float trajectories at 1,000 m ( Figure 5b) and trajectories in a realistic numerical configuration at mid-depth ( Figure 6) all indicate that the connection between the flow along the Mohn Ridge and the flow along the Knipovich Ridge is not as strong as suggested by surface circulation schematics currently in use (e.g., Orvik & Niiler, 2002). Furthermore, several surface drifters and ARGO floats cross the mid-ocean ridge from one basin into the other. Therefore, in addition to the large eddy variability in the Lofoten Basin (Bosse et al., 2018), the enhanced exchange between the slope current and the front current north of 72°N and the excursions of water masses across the mid-ocean ridge lengthen the residence time of the AW within the eastern basin of the Nordic Seas. This potentially has implications for the water mass transformation of the AW in the eastern basin, which is discussed in the following section.

Importance of Frontal Instability for Water Mass Transformation
Next, we investigate the dependence of AW transformation on possible pathways related to the instability of the front current. The realistic numerical configuration is limited by its domain, and the low number of ARGO floats and surface drifters available near the front current is not sufficient to perform any statistical analysis on water mass transformation. Furthermore, since the surface drifters and ARGO floats drift at a specified pressure level, they are not suited to trace a particular water mass that might move in the vertical. Therefore, the fate of the AW is also traced in the idealized model configuration of the Lofoten and Greenland Basin introduced in section 2.3. In this section, the frontal pathways and AW transformation found in the idealized model configuration are compared to the results from the realistic numerical configuration.
To investigate the role of the frontal instability, only those particles are taken into account that leave the domain within the front current (realistic numerical configuration; see section 2.2) or traveled within the front current (idealized numerical configuration; see section 2.3). Furthermore, particles that follow the mean flow along the mid-ocean ridge are separated from the particles that at some point along their trajectory meander into the Greenland (western) Basin. In both model simulations, the mean flow along the mid-ocean ridge is mainly confined to a constant isobath (dashed red lines in Figures 3a, 3b, 4a, and 4b).
Particles that do not cross this isobath are assumed to follow the front current to the north (front-pathway); particles that do cross this isobath are assumed to be steered into the western basin by the instability of the front current (front_west-pathway).
By applying this selection procedure to the particles in the realistic model, the two exit pathways along the eastern and western flank of the Knipovich Ridge that were already identified in Figure 6b are now separated (Figures 9a and 9b). The front-pathway consists of particles that partly follow the Mohn Ridge but also originate from the slope current. In contrast, the front_west-pathway mainly consists of particles that travel along the Mohn Ridge. Most of these particles cross the mid-ocean ridge toward the northwest at 73°N where they continue their path along the western flank of the Knipovich Ridge.
The particles in the idealized numerical simulation display a similar behavior (Figures 9c and 9d). The frontparticles indicate exchange with the slope current mainly in the northern part of the domain, whereas the front_west-particles reside mainly at the western side of the front current. Therefore, the front_west-

10.1029/2020JC016075
Journal of Geophysical Research: Oceans particles are found farther offshore within the boundary current in the western basin compared to the frontparticles. Unlike the particles in the realistic simulation, there is no strong evidence for a preferred location where the front_west-particles cross the ridge into the western basin. This can be explained by the uniform shape of the mid-ocean ridge in the idealized numerical configuration.
Despite the similarities of the frontal pathways in the realistic and the idealized simulations, the number of particles that follow the front-pathway and front_west-pathway differ between the two simulations. In the realistic simulation, most particles (71%) cross the ridge and flow along the western flank of the Knipovich Ridge, whereas in the idealized simulation, most particles (81%) stay within the mean flow of the front current. Note that the percentages calculated are based on the total number of particles following each path, not on the total transport carried. Furthermore, although the idealized numerical configuration captures some of the key features of the Nordic Seas, a realistic bathymetry and surface forcing (e.g., there is no wind forcing in the idealized configuration) will certainly play a role on the strength and separation of various pathways within the Nordic Seas. However, as the frontal instability is mostly a result of the strong horizontal gradients in density across the ridge, the idealized configuration can still be used to investigate water mass transformation processes occurring along the front.

Journal of Geophysical Research: Oceans
To assess the impact of the frontal current instability on the water mass transformation in the idealized basin, for each pathway the temperature of the particles at the deployment location is compared to the temperature at the outflow (Figure 10a). The distribution of the particles' temperature at the outflow shows two distinct peaks at ∼8.2°C and ∼7.9°C (dashed lines Figure 10a). The colder signal at the outflow is related to the pathways that interact with the western basin interior (the front_west-particles; yellow lines in Figure 10a), whereas the warmer outflow waters correspond to the particles that travel within the front (the front-pathway; gray lines). Note that the temperature distribution of the front-particles and front_west-particles is indistinguishable at deployment (solid gray and yellow lines in Figure 10a). This indicates that the excursion into the western basin is essential to yield a stronger water mass transformation.
The temperature and salinity distribution of the particles released in the realistic simulation also show that the water mass transformation of AW depends on the pathway the water takes through the basin; particles that cross the Mohn Ridge toward the west are colder and fresher when they leave the model domain than the particles that stay within the front current (Figures 10b and 10c). The cooling and freshening of the water mass that crosses the mid-ocean ridge by frontal instabilities is likely due to mixing with the cold and fresh water masses within the Greenland Basin. Furthermore, the excursion across the ridge increases the transit time (not shown). Therefore, the cooling of the water mass continues as long as the water mass is exposed to the atmosphere.

Conclusions and Discussion
In this paper, we have investigated whether the flow along the Mohn Ridge and the flow along the Knipovich Ridge are connected in a Lagrangian sense. As the number of surface drifters and ARGO floats near the mid-ocean ridges is limited, we have additionally analyzed trajectories of numerical particles in a realistic and an idealized numerical simulation. These simulations allowed us not only to investigate the connection between the slope and front currents of the Norwegian Atlantic Current (NwAFC and NwASC, respectively) but also to trace water mass transformations along different pathways.
Based on this study, it is clear that care should be taken when using velocity and travel direction from surface drifters to derive circulation maps. These maps, as, for example, derived by Orvik and Niiler (2002), show the Eulerian mean velocity field but cannot provide any information on the connection between different branches in a Lagrangian sense. Based on the observed trajectories, we found only a weak connection between the northward flow along the Mohn Ridge and along the Knipovich Ridge, unlike how the NwAFC is normally schematized. Only 5 of the 40 surface drifters showed the traditional pathway north along both ridges. Instead, most drifters displayed exchange with the slope current ( Figure 5a).
Trajectories in the realistic numerical simulations indicated a stronger connection between the flow along the Mohn Ridge and the Knipovich Ridge than the observed trajectories. However, the model results showed a clear branching of the flow along the Mohn Ridge into different paths to the north ( Figure 6). Therefore, in the realistic numerical simulation, the front current along the Mohn Ridge did not form a one-to-one connection with the front current along the Knipovich Ridge either, corroborating the conclusions drawn from the observed trajectories.
Frequent exchange between the slope and front current was seen in the observed trajectories and the realistic model (Figures 2, 5 and 6). Many surface drifters and ARGO floats traveled along the 2,500 m isobath near 72°N toward the east (west) to join the slope (front) current on their path northward. This path was present in the realistic numerical simulation as well. Further north, floats crossed from the eastern branch of the West Spitsbergen Current to the western branch of the West Spitsbergen Current (or vice versa). The front current and slope current are therefore not isolated but clearly connected in a Lagrangian sense. Rossby, Prater, et al. (2009) observed a similar connection between the front current and the slope current just north of the Iceland-Scotland Ridge using RAFOS floats. Our results suggest that also north of 72°N significant exchange between the two branches occurs.
There are various mechanisms that can contribute to exchanges between the front and slope currents. The surface drifters in the Nordic Seas flow within the Ekman layer and are therefore subject to Ekman transport (Brambilla & Talley, 2006). Although the wind stress curl in the Nordic Seas is predominantly positive, wind reversals can occasionally enhance exchange between the boundary and the interior of the Lofoten Basin 10.1029/2020JC016075

Journal of Geophysical Research: Oceans
and therefore enable crossovers between the slope and the front current. In addition to wind effects, the strong eddy variability near the Lofoten Islands provides a means for AW to enter the Lofoten Basin interior as well (Fer et al., 2020). The latter mechanism seems to be responsible for AW transport from the slope current to the interior at depth, whereas wind processes seem to steer the surface waters across the Vøring Plateau entering the Lofoten Basin from the south .
Although wind effects are likely to play a role for the pathways of the surface drifters, the trajectories of the ARGO floats and the modeled trajectories from the realistic and idealized simulations are positioned below the Ekman layer. Below the surface, the mean circulation patterns are strongly dominated by topography (e.g., Nøst & Isachsen, 2003;Voet et al., 2010). Voet et al. (2010) observed the tight flow alignment with topography using so-called topostrophy derived from ARGO float trajectories, which is a measure of the flow alignment with isobaths. However, the region north of the Lofoten Basin is characterized by low values of topostrophy (his Figure 9), indicating that the ARGO float trajectories cross isobaths there. This is in agreement with our findings, where in addition to the ARGO floats, also, surface drifters showed similar behavior in this area. Such a crossover region could indicate that in this area the strength of eddies and meanders is of similar magnitude as along-flow mean velocities (Bower, 1991;Rossby, Prater, et al., 2009).
The numerous exchanges between the front current and the slope current lengthen the residence time of AW within the eastern basins of the Nordic Seas, which could enhance the total heat loss over the basin. Furthermore, the link between these two branches weakens the direct propagation of AW anomalies from the Greenland-Scotland Ridge to Fram Strait. It is therefore important to increase our knowledge not only on processes within the Lofoten Basin but also on the region north of 74°N and in particular on the local front current dynamics.
Analyses of trajectories in both the realistic and the idealized numerical configurations showed that AW transformation in the Nordic Seas depends on the path that the AW takes. Additionally, the water mass transformation itself could to some extent determine which path the AW will follow downstream. Our results suggest that the coldest and freshest water masses are formed along the front current, due to the frontal current instability and exchange with the western basin. Although not investigated in this study, the observed pathways are likely to display seasonal variations (e.g., Bosse & Fer, 2019). On longer time scales, it is known that the strength of the front current depends on the phase of the North Atlantic Oscillation (NAO); the front is weaker during the negative phase of the NAO (Raj et al., 2019). Also, under future scenarios, the front is expected to weaken, due to a reduction in the density gradient between the Lofoten Basin and the Greenland Basin (Ypma et al., 2020). If, due to the decrease in baroclinic instability, the exchange between the east and west reduces, the AW heat loss in the eastern basins could decrease and convection in the Greenland Basin could weaken due to a smaller supply of salty waters from the east.
Although not observed previously (Latarius & Quadfasel, 2016;Poulain et al., 1996;Rossby, Prater, et al., 2009), we found 28 ARGO floats and surface drifters that crossed the mid-ocean ridges. This might seem like a small number, but compared to the total number of observed trajectories that reached the mid-ocean ridges (90 floats), the observed cross-ridge exchange may be substantial. As in situ measurements suggest strong northward flow along the mid-ocean ridge (Bosse & Fer, 2019), it is surprising that out of the 564 surface drifters and 283 ARGO floats in the Nordic Seas, so few are found close to the Mohn and Knipovich Ridges. We would therefore encourage deployment of drifters within the strong northward flow along the mid-ocean ridge, to gain further insight in the pathways and residence time of AW in the eastern Nordic Seas.

Data Availability Statement
The ARGO data were collected and made freely available by the International Argo Program and the national programs that contribute to it (