Mean Structure and Seasonality of the Norwegian Atlantic Front Current Along the Mohn Ridge From Repeated Glider Transects

The poleward flow of Atlantic Water in the Nordic Seas forms the upper limb of the meridional overturning circulation driving an important heat transport. The Norwegian Atlantic Front Current along the Mohn Ridge between the Greenland and Norwegian Seas is characterized for the first time, using repeated sections over 14 months from autonomous underwater gliders and two research cruises. The Norwegian Atlantic Front Current follows the 2,550‐m isobath with a width of 38 ± 2 km and absolute geostrophic velocities peaking at 0.56 ± 0.03 m s−1. The mean transport of Atlantic Water is 3.2 ± 0.2 Sv (equivalent to temperature transport of 71 ± 5 TW). Seasonal variability was observed with a magnitude of 0.8 Sv and maximum values in the fall. The deep currents at 1,000 m explained most of this seasonal variation and were anticorrelated with time‐integrated wind stress curl over the Lofoten Basin. Part of this flow might recirculate within the Lofoten Basin, while the rest continues toward the Arctic.


Introduction
Ocean circulation plays a key role in the climate system by redistributing heat from low to high latitudes (Ganachaud & Wunsch, 2000). The upper limb of the meridional overturning circulation in the Atlantic carries warm and salty Atlantic Water (AW) poleward across the north Atlantic toward the Arctic Ocean. The role of Nordic Seas (Greenland, Iceland, and Norwegian Seas) in the meridional overturning circulation, through heat exchanges with the atmosphere and production of North Atlantic deep water, is increasingly recognized (Chafik & Rossby, 2019;Lozier et al., 2019). The transport of AW affects the climate of western Europe (Årthun et al., 2017;Rhines et al., 2008), water mass transformation (Bosse et al., 2018;Mauritzen, 1996;Rossby et al., 2009), sea ice extent (Lind et al., 2018;Onarheim et al., 2014), as well as marine ecosystems (Årthun et al., 2018;Fossheim et al., 2015;Oziel et al., 2017). Characterizing the structure, magnitude and variability of this poleward transport is therefore highly important.
The ocean currents carrying warm AW in the Nordic Seas are organized in two poleward branches (Orvik & Niiler, 2002) (Figure 1a): the barotropic Norwegian Atlantic Slope Current (NwASC) and the baroclinic Norwegian Atlantic Front Current (NwAFC). The NwASC follows the shelf break northward and continues to the Barents Sea and Fram Strait. The NwAFC is the western branch and follows the 2,000-m isobath along the Vøring Plateau and then the Mohn and the Knipovich ridges farther north (Orvik & Niiler, 2002) (see Figure 1a and supporting information Figure S1 for a detailed bathymetric map of the Mohn Ridge region), before it partly joins the Fram Strait branch and recirculates into the Greenland Sea. Along their poleward pathway, the slope and frontal branches exchange waters (Rossby et al., 2009) and a slab of AW fills the Lofoten Basin (LB), forming the largest heat and salt reservoir of the Nordic Seas (Björk et al., 2001;Bosse et al., 2018;Mork et al., 2014; see Figure S2).
South of the Vøring Plateau at the Svinøy section (63 • N), continuous measurements since 1995 set the NwASC annual transport to 4.2-4.4 Sv (Orvik & Skagseth, 2003;Orvik et al., 2001) (Sv = 10 6 m 3 s −1 ). This barotropic branch is relatively well constrained by observations. Transport of the NwAFC in the same section, on the other hand, is highly uncertain. Based on limited hydrographic data and assuming a deep level of no motion, a baroclinic geostrophic transport estimate was 3.4 Sv (Orvik et al., 2001). Similar baroclinic transport calculations using limited surveys gave 2.4 Sv over the Mohn Ridge (Gascard et al., 2004), 2.6 Sv (Walczowski et al., 2005) and 3 Sv (van Aken et al., 1995) over the Knipovich Ridge. While these transports are roughly consistent, all exclude the barotropic contribution which was suggested to be important using theoretical arguments (Orvik, 2004), limited deep current profiling (Walczowski et al., 2005), and . Purple dots are CTD/L-ADCP stations taken during two PROVOLO cruises. The mean position and direction of the individual PF samplings define the mean PF axis. Cross-front distance and time distribution of (c) absolute salinity (0-25 m average) and (d) conservative temperature (300-400 m average) for individual dives (circles) and the objectively interpolated fields with correlation scales of 30 days and 30 km. absolute geostrophic current calculations referenced to shipboard current measurements (Cisewski et al., 2003). Using Seaglider sections in the NwAFC at the Svinøy section, Høydalsvik et al. (2013) obtained absolute geostrophic transport (referenced to glider depth-averaged currents [DAC]) of 6.8 Sv, that is, larger than the transport of NwASC. Currently, the magnitude and structure of the NwAFC transport are thus poorly constrained by observations. The LB is characterized by intense winter heat losses to the atmosphere (annual average of 80 W m −2 ) (Isachsen et al., 2007;Richards & Straneo, 2015), a deep cyclonic circulation (Voet et al., 2010) (dashed line in Figure 1a), and winter mixed layer depths exceeding 500 m (Bosse et al., 2018;Latarius & Quadfasel, 2016;Nilsen & Falck, 2006). Consequently, a sharp and dynamic density front is established over the Mohn Ridge that separates the warm AW from cooler and less saline Arctic waters (Mork & Blindheim, 2000). The NwAFC flows as a baroclinic jet associated with this "Polar Front" (PF; also called "Arctic Front"). The frontal system is hypothesized to contribute approximately half of the total northward AW transport (Gascard et al., 2004;Høydalsvik et al., 2013;Orvik et al., 2001;Walczowski et al., 2005) and to mediate exchanges between the Norwegian and Greenland Seas (Segtnan et al., 2011;van Aken et al., 1995;Walczowski, 2013). However, despite its impact on ocean circulation and biology (van Aken et al., 1991), observations at the PF are scant.

10.1029/2019GL084723
Here, we present detailed observations of the PF and NwAFC along the Mohn Ridge using sustained measurements over a period of 14 months. We report the first accurate estimates of absolute AW transport, as well as its components associated with the deep currents at 1,000 m and with the geostrophic shear above this level. The data are collected from two cruises and repeated transects using autonomous underwater gliders. The average structure is well resolved, and the seasonal variability is discussed within the limitation of the data set. The NwAFC at the Mohn Ridge is a key contributor to redistribution of the AW heat toward the Arctic and must be monitored.

Cruise Data
Measurements were made under the PROVOLO project ("Water mass transformation processes and vortex dynamics in the LB of the Norwegian Sea"). A 150-km section across the Mohn Ridge was repeated during two PROVOLO cruises with RV Håkon Mosby in June 2016 (14 stations) and RV Kristine Bonnevie in March 2017 (13 stations). Conductivity-Temperature-Depth profiles were acquired using a Sea-Bird Scientific SBE 911plus with pressure, temperature, and salinity data accurate to ±0.5 dbar, ± 0.002 • C, and ±0.003 g kg −1 , respectively. Conservative temperature , absolute salinity S A , and potential density anomaly were calculated using the TEOS-10 Gibbs Seawater Oceanographic toolbox (McDougall & Barker, 2011). Horizontal current profiles were acquired by a pair of 300-kHz lowered acoustic Doppler current profilers (L-ADCP) attached to the rosette, sampling in 8-m vertical cells. The L-ADCP data were processed using the velocity inversion method of Visbeck (2002), implemented in the LDEO software version IX-12 (Thurnherr, 2010), with typical horizontal velocity uncertainty of 2-3 cm s −1 . The data set and further details can be found in Fer et al. (2019).

Seaglider Missions
Autonomous underwater gliders collect data along a sawtooth trajectory between the surface and a typical maximum depth of 1,000 m (Testor et al., 2019). Between two consecutive dives, an estimate of the DAC is obtained by comparing the dead-reckoning position to accurate GPS fixes. Two Seagliders were deployed successively in June 2016 and January 2017 and repeatedly sampled the PF down to 1,000 m for 14 months with a 1.5-month gap in January 2017. In total, 25 sections were occupied from June 2016 to July 2017 along four main lines spread over about 150 km along the ridge's axis (Figure 1b). No difference in the PF structure could be identified along the ridge. Each section took on average 10 days to be completed, while the frontal zone was crossed in about 3 days. The Seagliders were equipped with unpumped SBE-41 C/T sensors. Average horizontal and vertical speeds were 15 km day −1 and 8 cm s −1 , respectively, giving a typical horizontal separation between profiles of 2.5 km (5 km between DAC estimates) and a vertical resolution of 0.8-1.6 m with the sampling rate of 10-20 s. Salinity was calculated after a thermal lag correction (Garau et al., 2011). Finally, glider temperature and salinity were corrected by an offset against the calibrated cruise Conductivity-Temperature-Depth profiles. The data set and further details can be found in Bosse and Fer (2019).

Temperature, Salinity, and Velocity Sections
For the 25 glider sections, along-stream absolute geostrophic velocities were computed in a cross-front axis defined by the direction of DAC, following Todd et al. (2016) and assuming DAC is aligned with the baroclinic surface jet. This hypothesis was validated by L-ADCP profiles in summer, while in winter, a cyclonic rotation of currents with increasing depth was noted. This can be due to -spiral and surface cooling effects (Schott & Stommel, 1978). The traditional calculation of geostrophic velocities perpendicular to the glider track leads to smaller velocity amplitude, larger apparent width, and similar transport (see Figures S3 and S4 and Table  S1 for methodological details). Scattered ∕S A were then optimally interpolated on a regular grid using a two-dimensional Gaussian correlation function: cov(X, Z) = e (X, Z) + (1 − e) exp(−X 2 ∕L 2 r − Z 2 ∕L 2 z ) with the Dirac function, e = 0.05 the relative error, L r = 20 km, typical half-width of the PF, and L z = 15 m, typical vertical scale of the seasonal thermocline. A 20-km scale corresponds to a sampling time scale of about 32 hr, larger than the dominant tidal (about 12 and 24 hr) and inertial (about 13 hr) periods, effectively filtering internal waves (Rudnick & Cole, 2011). Along-stream geostrophic velocities (U g ) were finally obtained by vertically integrating the geostrophic shear and constraining its depth average by the DAC. The PF position (y = 0) was defined as the location where 0-to 50-m averaged velocity was maximum.
For each glider section, absolute geostrophic velocities were decomposed into velocities at 1,000 m (U 1,000 g ) and the remainder due to geostrophic shear above this level (U shear g ) (i.e., equivalent to assuming a level of no motion at 1,000 m). Seasonal composites were then obtained by averaging sections from different months (winter: JFMA, summer: MJJA, and fall: SOND). These 4-month periods correspond to the short period of heat gain from the atmosphere from May to August and the long period of cooling from September to April (Yu et al., 2017). Finally, the mean PF section was calculated as the average of the three seasonal composites. For a consistent and homogeneous data set, shipboard measurements were not included in the composites but were used to describe the PF characteristics (seasonal and annual means summarized in Table S1). The water mass definitions follow the literature (Blindheim, 1990)-note that absolute salinity is larger by about 0.17 g kg −1 compared to practical salinity: AW, S A > 35.17 g kg −1 ; new AW (nAW), S A > 35.27 g kg −1 and > 6 • C; modified AW (mAW), 35.28 < S A < 35.34 g kg −1 and 27.71 < < 27.82 kg m −3 ; and Arctic Surface Water (ASW), S A < 35.12 g kg −1 for < 27.8 kg m −3 and S A < 35.07 g kg −1 otherwise. Mixed layer depths are estimated with a threshold on potential density increase of 0.03 kg m −3 (de Boyer Montégut et al., 2004).

Volume and Temperature Transport
Optimally interpolated sections of U g and were used to estimate the NwAFC volume and temperature transports. Lateral integration limits for each section were obtained by fitting a Gaussian jet (U m e −0.5x 2 ∕ 2 with U m the maximum velocity and the variance of the Gaussian function) to the depth-averaged geostrophic velocit, restricted to the values above half of the maximum value. Integration limits, also defining the width of the PF (L int in Table S1), were then delineated by regions of poleward velocity (i.e., positive) within the maximum interval of x = ±2 , accounting for about 95% of the total transport of the theoretical jet. This was motivated to exclude spurious transports away from the front axis, captured because of the slow glider speed (about 3 days to cross the PF). Along-stream velocities were then categorized in different water masses and integrated.
For transports from the cruise data, direct measurements by L-ADCP were used, and the jet axis was defined at the location of maximum in 0-to 200-m depth-averaged velocity. Temperature transport across the PF was computed using a reference temperature of −0.1 • C, the mean temperature of the outflow from the Arctic Ocean (Aagaard & Greisman, 1975). Because these estimates are based on single sections and mass is not conserved, they represent temperature transport rather than heat transport.
Error in DAC from gliders can be considered as a random error of 0.02 m s −1 (Rudnick et al., 2018). A statistical analysis on 100 members of each cross-front section by adding random noise to DAC, together with the optimal interpolation error, gave a mean statistical error, averaged over the 25 glider sections, of 0.14 and 0.05 Sv for total and AW transports, respectively. Note that the errors in individual transport estimates are small compared to the standard error of the mean arising due to transect-to-transect variability (Table S1). Cross-front sections were considered as independent if separated in space and time by scales larger than typical mesoscale. In the Nordic Seas, the deformation radius is about 10 km (Nurser & Bacon, 2014). Assuming decorrelation scales of 25 km and 5 days, Sections 5/6 and 18/19 were identified as dependent ( Figure 1c) and averaged before computing seasonal means.

Atmospheric Reanalysis
Atmospheric forcing was obtained from ECMWF's ERA-5 reanalysis for the period from June 2016 to July 2017 (Copernicus Climate Change Service, 2017). Hourly 10-m winds at 0.25 • resolution were retrieved and averaged into 24-hr intervals. The daily wind stress was then computed as w = a Cd a |v w |v w , where v w is the wind, a = 1.2 kg m −3 the air density, and Cd a = 0.0012 the surface drag coefficient for moderate winds (Large & Pond, 1981).

Mean Hydrographic and Current Structure
The AW layer extends to a depth of 700 m in the LB and shoals toward the Mohn Ridge (see Figure S2). As captured by the satellite image in May 2016, the surface temperature across the PF decreased from about 6 to 1 • C through a succession of fronts influenced by mesoscale features (Figure 1b), similar to the PF structure described along the Knipovitch Ridge by van Aken et al. (1995). Higher salinities (about 35.3 g kg −1 ) were observed on the warm side of the front (Figure 1c), with a freshening trend consistent with recent observations (Bosse et al., 2018;Mork et al., 2019). The salinity on the cold side was more variable and reached minimum values from October to March. This relatively fresh water, associated with ASW, can be linked to seasonal advection of ASW from the Greenland shelf by the Jan Mayen Current (Figure S5). At the depth of the AW core, the PF manifested as an abrupt decrease of temperature of the AW layer (about −4 • C/40 km at 300-to 400-m depth; Figure 1d), stronger than at the surface. Instabilities of the front generate important lateral heat fluxes. In particular, warm anticyclonic features detaching from the front could restratify the upper layers on the cold side (Figure 1b).
The composite PF section confirms the existence of a steep temperature front at 200-to 500-m depth and a salinity signature near the surface (Figure 2). The isopycnal slope peaks at the front origin, defining the baroclinic jet of the PF. The geostrophic currents had an average maximum speed of 0.56 ± 0.03 m s −1 (mean ± standard error, standard error = ∕ √ n with the standard deviation and n the number of independent cross-front sections) with values ranging from 0.34 to 0.83 m s −1 . The absolute currents at 1,000 m were significant (0.03 ± 0.01 m s −1 ). The width of the PF was 38 ± 2 km. The PF position over the Mohn Ridge varied by about 100 km across the ridge between the 2,250-and 2,840-m isobaths, with an average at the 2,550-m isobath (with respect to the smoothed topography). This is about half the distance from the bottom to the ridge crest. More variability in PF position was observed during the winter and summer when deep currents were also weaker ( Figure 2f). The jet was observed always on the same side of the ridge (i.e., retrograde jet in direction opposite to topographic Rossby waves) indicating stabilization of the barotropic flow (Li & McClimans, 2000;Poulin & Flierl, 2005). The topographic control of the barotropic jet along the Mohn and Knipovitch ridges indicates that parts of the AW could recirculate and flow to the southeast into the LB ( Figure S6).
A mean flow in the opposite direction of the poleward frontal current was observed in the LB at about 40 km from the PF. This opposing flow was more intense at depth, with a mean value of about 0.05 cm s −1 , as theoretically predicted by Orvik (2004) and described by Voet et al. (2010) from the drift of Argo floats at middepth. The deep currents directly measured during the cruises confirm the return flow on the warm side of the PF with even larger amplitude (Figure 2f), but they might be influenced by surrounding mesoscale features. The flow associated with the cyclonic circulation was stronger during the winter while almost nil during the summer and fall (Figure 2g), indicating either a decrease in intensity or a shift in position toward the center of the LB.

Seasonal Evolution
The vertical structure of the PF shows a seasonal evolution (Figure 3). Deep vertical mixing reaching 500 m cooled the AW to form a mode water of about 5.3 • C in winter 2017 (mAW, see Figure 3d). Cross-front exchanges of warm waters prevent deep vertical mixing on the cold side where the mixed layer depth was only about 200 m, in line with the shoaling of the AW pycnocline (i.e., = 27.9 kg m −3 ). This transition zone, characterized in the upper 200 m by an intrusion of relatively cold and fresh AW of about 3 • C, extended to at least 50 km from the PF and was observed in all seasons (Figures 3a-3f), in agreement with climatological observations across the Mohn Ridge ( Figure S2). The strongest cross-front temperature and salinity contrasts were detected in the fall. This was due to the presence of a 50-m-thick ASW layer, as well as the arrival of new warm and salty waters (nAW) in summer between 100 and 200 m, originating from the NwASC (Figures 3k and 3l).
Density gradients in the PF were typically dominated by temperature, but salinity difference reaching −0.2 g kg −1 in fall compensated for approximately one third of the cross-front density increase from temperature. The temperature difference across the PF at 270 m depth increased from 3 .7 C in winter to • 4.7 C in fall. While the width of PF remained constant within errorbars throughout the seasons, the absolute • currents at 1,000 m reached 0.06 ± 0.01 m s −1 in fall, compared to insignificant velocities in winter and 0.04 ± 0.02 m s −1 in summer. The return flow associated with the deep cyclonic gyre circulation of the LB was captured in winter with a DAC of 0.10 m s −1 , whereas it was absent in summer and fall.

Volume and Temperature Transport
The average AW transport was 3.2 ± 0.2 Sv, varying from 1.6 to 5.4 Sv with a seasonal evolution in volume and type of water masses (Figure 4a). The mAW was observed throughout the year but dominated the winter transport with 1.0 ± 0.9 Sv. On the contrary, the nAW was absent during the winter and peaked during the fall (0.8 ± 0.6 Sv). The ASW was occasionally transported with the PF, mostly during the summer and fall: In 4 out of 25 sections, ASW transport was larger than 0.1 Sv, with a maximum of 0.4 Sv observed in November when the surface salinity was the lowest. Note, however, the 2-month gap in the data due to the turnover of the gliders in January.
The AW transport was decomposed into a component associated with the deep currents measured at 1,000 m and the remainder associated with the geostrophic shear above this depth. While the total transport had a marked seasonal signature varying from 2.8 ± 0.4 Sv in winter to 3.6 ± 0.2 Sv in the fall, the transport due to geostrophic shear was fairly constant with 3.0 ± 0.2 Sv. This transport is in good agreement with previous estimates along the Mohn Ridge (Gascard et al., 2004;Orvik, 2004) and farther north (van Aken et al., 1995;Walczowski et al., 2005). The seasonal variation in AW transport was thus primarily due to variation in the transport associated with deep currents that increased from near zero values in winter to 0.8 ± 0.2 Sv in fall, accounting for 22% of the total AW transport. Sinusoidal fits to the observed seasonal transport yielded minimum transport in March and amplitudes of 0.1, 0.2, and 0.4 Sv for the geostrophic shear, deep currents, sand total transports, respectively (Figure 4c). We conclude that the AW transport at the PF is dominated by currents associated with the geostrophic shear above 1,000 m with a seasonal variability due to deep currents, representative of the barotropic circulation.
The poleward temperature transport at the PF was 71 ± 5 TW (TW = 10 12 W). As a comparison, the AW temperature transport (using the same temperature reference) is about 300 TW at the AW inflow into the Nordic Seas (Hansen & Østerhus, 2007) and 30-60 TW toward the Arctic Ocean in Fram Strait (Schauer et al., 2004;Tsubouchi et al., 2018).
The wind stress curl drives variations of the barotropic circulation (e.g., see Voet et al., 2010, for the LB). We averaged the wind stress curl over the central LB, inside the 2,800-m isobath guiding the cyclonic gyre circulation of the LB. On seasonal time scales, low-pressure systems traveling across the Nordic Seas typically result in positive wind stress curl during the winter. In summer, cyclonic activity ceases, and wind stress curl relaxes to negative or small values (Figure 4d). Interestingly, an anticorrelation can be noticed between the sign of wind stress curl and the AW transport variations. Positive wind stress curl indeed increases the cyclonic LB gyre circulation, although Voet et al. (2010) showed that winds cannot alone explain the gyre variability. Consequently, a decrease in the poleward transport at the PF resulted from a reduction of the barotropic component (induced by a strengthened opposing barotropic circulation).

Discussion and Concluding Remarks
We found an anticorrelation between the AW transport variations along the Mohn Ridge and the sign of the wind stress curl over the LB. Previous studies have established a positive correlation between the AW transport along the Svinøy slope and the local wind forcing and the large-scale North Atlantic Oscillation (Orvik & Skagseth, 2003a;Skagseth, 2004;Skagseth et al., 2004). The NwAFC and NwASC seem to react in opposite ways to the wind forcing in the LB. Numerical models could help understand the response of the slope and frontal Norwegian Atlantic Currents in the LB where observationsremain scarce.
Deep velocities of 0.1-0.2 cm s −1 have been directly measured by lowered profilers. The large abyssal velocities should have important implications for mixing over the rough topography of the Mohn Ridge ( Figure  S1). The interaction of the barotropic jet of the PF and tides with numerous seamounts of O(1-10 km) horizontal scales found along the ridge could generate internal waves and turbulent mixing (Nikurashin & Ferrari, 2011). Moreover, the Mohn Ridge is an active hydrothermal vent (Schander et al., 2010), for which small-scale currents and mixing are important (Vic et al., 2018). The mixing generated by the barotropic component of the frontal current and tides is still unknown and needs further investigations.
The PF position over the ridge slope was fairly stable in the fall, when the frontal jet was more intense and characterized by stronger deep currents. The horizontal shear of deep currents, however, remained quite comparable throughout the year. Topograpic slope plays an important role in the development of barotropic and baroclinic instabilities (Isachsen, 2015;Li & McClimans, 2000). The PF had shallower water on its cyclonic flank (i.e., in the opposite direction of topographic Rossby wave propagation), which is known to stabilize the flow (Poulin & Flierl, 2005). Variable winds associated with low-pressure systems can also excite upstream-propagating topographic Rossby waves (Li & McClimans, 1998). The strong variability in temperature and salinity across the PF captured by Seagliders and a satellite image suggest that the PF instability can be important and deserves further studies. A thorough understanding of the cross-front exchanges is indeed required to assess the role of lateral fluxes in setting the water column stratification and the local ecosystem dynamics, as well as in the propagation of heat and salt anomalies toward the Greenland Sea.
Average poleward volume and temperature transports of AW along the Mohn Ridge are 3.2 ± 0.2 Sv and 71 ± 5 TW. The transport associated with deep currents at 1,000 m accounts for about 10% of the total (up to 22% with 0.8 Sv in fall). This transport component, documented for the first time here, has often been neglected in previous studies mainly due to limitation of hydrographic surveys without direct currents measurements. The NwAFC is slightly weaker than the slope current, but comparable to the front current at the Svinøy section (63°N). Contrary to costly regular ship surveys, the glider technology can be effectively used to monitor topographically steered frontal currents such as the NwAFC. In particular, the fate of the flow of AW downstream of the Mohn Ridge remains unclear. Its connection with the Knipovitch Ridge and farther north to Fram Strait could be investigated by sustained sampling using autonomous underwater gliders.