The relatively mild climate of Norway is largely attributed to the northern extension of the North Atlantic Drift, the Norwegian Atlantic Current that transports warm and saline water masses toward the Arctic Ocean . These nutrient-rich warm waters contribute to supporting the entire food chain and sustain the productive waters around Norway (see, e.g., , for a discussion on recruitment of Atlantic cod stocks). The circulation pattern is organized in two main branches originating from the Iceland–Faroe and Faroe–Shetland gaps (Fig. a): the Norwegian Atlantic Slope Current (the slope current hereinafter) and the Norwegian Atlantic Front Current (the front current hereinafter). The diverging isobaths of the Lofoten Basin in the Norwegian Sea guide the two branches. The slope current follows the shelf break along the Norwegian continental slope northward and continues into the Barents Sea and Fram Strait. The front current follows the 2000 m isobath, veers west at the flanks of Vøring Plateau and continues poleward along the Mohn Ridge .

The front current, which is not addressed in this study, has not been measured in detail using current meter arrays, but geostrophic transport estimates are available from hydrography. At the Svinøy section (63∘ N, about 300 km downstream of the Faroe–Shetland Channel), a baroclinic geostrophic transport estimate of the front current was 3.4 Sv (1 Sv =106 m3 s-1) ; however, the total geostrophic transport from repeated Seaglider transects reached 6.8 Sv , implying a large barotropic contribution. Farther north, detailed glider observations of the front current over the Mohn Ridge confirm large transport rates, giving a 4.5 Sv annual average with an approximately 2 Sv barotropic contribution .

The University of Bergen, Norway, has monitored the slope current transport at the Svinøy section since 1995 at the location indicated by a star in the inset of Fig. b . The slope current there is about 40 km wide between the 200 and 900 m isobaths, with an annual mean speed of 0.3 m s-1. The average annual transport of this barotropic branch is 4.4 Sv . The slope current accelerates along steep topography off the Lofoten Escarpment near the Lofoten Islands . The Norwegian coastal current (blue arrow Fig. 1) carries relatively fresh water over the shelf and as the shelf gets narrow near the Lofoten Escarpment, there might be interactions with the slope current. Here, there are no published moored current meter records, but surface drifters indicate velocities reaching 1 m s-1 . The transport and variability of the slope current in this region are not known. It is hypothesized that the current becomes increasingly unstable near this topographic steepening. Using time-averaged fields of an eddy-resolving numerical ocean simulation, showed that the steep Lofoten Escarpment exhibits enhanced unstable baroclinic growth rates and large velocity variability, suggesting high lateral diffusion rates. The structure and transport of the slope current at the Lofoten Escarpment are the focus of this study.

The Lofoten Basin is affected by Atlantic Water (AW) transport and becomes a major heat reservoir that is exposed to large surface heat losses and substantial water mass transformations . AW enters the basin both as a broad slab in the upper layers between the two branches and by eddies detached from the unstable slope current . The eddy-induced lateral heat fluxes distribute heat in the basin . The region is energized, which is manifested in the map of average geostrophic eddy kinetic energy (Fig. a; see Sect. ) showing two maxima: one in the center, associated with a permanent, energetic eddy , and a secondary maximum closer to the slope, likely associated with the variability induced by the slope current. The energetics and the variability of the slope current remain to be constrained by observations.

The study was conducted as a part of the “Water mass transformation processes and vortex dynamics in the Lofoten Basin of the Norwegian Sea” (PROVOLO) project. The overall objective of PROVOLO was to describe and quantify the processes and pathways of energy transfer and mixing in the Lofoten Basin and their role in water mass transformation. Observations from multiple cruises, gliders and subsurface floats were analyzed and reported elsewhere with focus on AW transformation , the permanent Lofoten Basin Eddy and the frontal structure across the Mohn Ridge . The mooring component concentrated on the slope current. Here we report the first observations of the volume transport rates, energetics and their variability from weekly to seasonal timescales based on the mooring records.

The standard Gimsøy section was visited four times throughout the mooring period. An average section using the subset of stations taken in all four occupations is representative of the hydrography during the measurements (Fig. ; also compare to the section from climatology presented in Sect. ). The AW, identified by temperatures above 5 ∘C and Absolute Salinities SA>35.17, covered the 50–700 m layer from the shelf edge toward the basin, overlying the fresher and colder deeper water. The interface between these water masses meets the bottom slope at about 700 m. A relatively fresh layer on the shelf is associated with the Norwegian coastal current. The moorings MN and MW, marked in Fig. , show that the range of current measurements sufficiently covered the AW layer and the dynamical core at the slope identified by sloping isotherms.

A summary of the environmental forcing during the measurement period shows that the net surface flux was typical of the long-term average, with an event of strong heat loss exceeding the 1 standard deviation (σ) envelope from mid-February to early March 2017 (Fig. a). Wind speed showed seasonal variability, increasing from 5 m s-1 in summer to 12 m s-1 in winter (Fig. b). We averaged the EKEg from satellite altimetry in a 30 km radius at the basin mooring location (an EKEg maximum region; see Fig. a) and at MW and compare the evolution throughout the mooring deployment in Fig. c. The EKEg records confirm that MB is 2 to 5 times more energetic in general, except in summer when both locations were relatively quiescent.

Profiles temporally averaged in the winter (DJF), spring (MAM), summer (JJA) and fall (SON) months at MW and MN show strong vertical shear in u in the upper 600 m at both moorings (Fig. ). In contrast to the barotropic slope current at Svinøy, the current at the Lofoten Escarpment clearly has a strong baroclinic component. Background shear between 200 and 600 m depth was 0.05 to 0.10 m s-1 per 100 m in all seasons, with a maximum in the fall. The fall was characterized by maximum baroclinicity, whereas winter was characterized by maximum barotropic currents, consistent with increased winds. Increased baroclinicity in the fall could partly be due to seasonal freshening of the coastal current reinforcing the density gradients and partly due to increased Ekman transport toward the shore observed from September to March (see northward winds implying eastward Ekman transport in Fig. b). It is also likely that the slope current could interact with the fresh coastal current due to the narrow shelf off the Lofoten Escarpment.

Over the full record, the 200–600 m depth-averaged u was 0.15 m s-1. The strongest currents were observed in winter with an average of 0.20 m s-1 at MW and 0.25 m s-1 at MN (approximately twice the summer average) when the temperature was also the highest. The average winter temperature at MW was 7.3 ∘C compared to 5.8 ∘C (at MW) or 6 ∘C (at MN) in summer. The average temperature in the 200–600 m range was warmer by more than 1 ∘C in winter. This could partly be explained by vertical redistribution through winter vertical mixing of heat contained in the seasonal thermocline and partly by changes in AW properties flowing into the Nordic Seas. The cross-isobath component was weak (typically ±0.02 m  s-1) and increased in spring and winter, with the largest 200–600 m depth-averaged values in winter (0.05 m  s-1 at MW) and an increased variability with depth (Fig. b, d). In the upper part of the water column at MW, averaged cross-isobath velocities in fall exceeded the winter values. This is consistent with the increased EKEg at the MW location, calculated from satellite measurements in November 2016 (Fig. c). In deep layers (>900 m) at MW, barotropic currents were between 0.05 and 0.10 m s-1.

The summer profiles in Fig.  are averages over the summers of 2016 and 2017. When averaged separately (not shown), temperature profiles at MW are very similar, equal to within 0.5 ∘C in the upper 600 m and identical in deeper layers. At MW, u was 0.01 m s-1 larger below 300 m (a small barotropic increase) in summer 2017, and shear was stronger in the upper 300 m, increasing by 0.06 m s-1 to 200 m depth. At MN, summer average profiles of u in the bottom 250 m were identical in 2016 and 2017, but shear was stronger higher in the water column (above 400 m) in summer 2017, with u increasing by an additional 0.10 m s-1 to 200 m depth. This implies substantial interannual variability in the upper 300–400 m, which cannot be resolved by our limited times series.

The currents measured at moorings MW and MN were highly variable (Fig. ). The 14 d low-passed currents were strongest in the fall and winter (Fig. a–d). The annual cycle of the 200–300 m vertically averaged u at MN had an amplitude of 0.10 m s-1 and explains 20 % of the variance obtained using a sinusoidal fit to daily data (not shown). These figures are similar for MW for 300–400 m averaged currents (depth ranges are chosen to ensure continuous time series unaffected by mooring knockdowns). The cross-isobath components show a less pronounced seasonality with 0.01–0.02 m s-1 (5 %–15 % variance explained) at both moorings. The temperature record at MW also shows strong seasonality. The amplitude of the annual sinusoidal fit to the temperature time series increases from 0.6 ∘C at 200–300 m to 1 ∘C at 500–600 m, accounting for 60 %–70 % of the variance, and rapidly decays deeper.

The largest along-isobath currents reach 0.8 m s-1 at both moorings, last for 1 to 2 weeks and extend as deep as 600 m. In periods with strong u, the cross-isobath velocity is also energized. These energetic periods also correspond to the peaks in EKEg obtained from satellite altimetry at the MW location (Fig. c). Isotherms (available only at MW for the entire mooring record) show vertical displacements of the order of 100 m, consistent with mesoscale meandering of the slope current.

For comparison, in the Svinøy section showed that the fluctuations of the slope current are a combination of longer periodic forced oscillations, which are a direct response to the wind (periods in the 3–5 and 16–32 d bands), and free waves corresponding to the first and second topographic wave modes (dominant periods of 40–70 and 80–110 h).

We analyzed fluctuations in the low-passed fields relative to the annual cycle to assess dominant timescales and amplitudes of variability. The time series of fluctuations of u at MN averaged between 200 and 300 m shows 13 events with a peak-to-peak amplitude of 0.2–0.3 m s-1 and a mean duration of 8±2 d at an average interval (time separation between events) of 35±10 d. Similar numbers of events with a comparable timescale are detected for temperature oscillations exceeding 0.5 ∘C.

At shorter timescales, the 35 h to 14 d band-passed variability is shown in Fig. e–f for u. The structure of band-passed v is very similar (not shown), with approximately half the amplitude of u. The band-passed fields show highly energetic current variability reaching ±0.4 m s-1 (variability for v is ±0.2 m s-1). A similar event analysis of the fluctuations in the filtered band (averaged between 200 and 300 m at MN and between 300 and 400 m at MW) results in very similar properties for MN and MW. Typically, 40–50 events are detected in u with a peak-to-peak amplitude of 0.15–0.20 m s-1 and a mean duration of 2±1 d at an average interval of 10±7 d. The cross-isobath component shows about 40 events with a peak-to-peak amplitude of 0.10±0.03 m s-1 at similar durations and time intervals. The energetics and conversion rates are further discussed in Sect. .

We estimated a decorrelation timescale as the e-folding timescale from an exponential fit to the autocorrelation function from hourly velocity time series. At both moorings at the 650 m isobath (MN and MS), the 200–600 m depth-averaged along-isobath currents are correlated at timescales up to 6 d. The decorrelation timescale at MW is comparable (7.3 d). For reference, the advection time between the along-isobath separation of MS and MN is 2 d using the mean speed of 0.15 m s-1. Over the 26 km separation, u values at MN and MS are highly coherent, with a maximum correlation coefficient of r=0.6 at a 41 h lag (consistent with the 2 d advection timescale). The cross-isobath components are not significantly correlated. The lateral separation of 5 km between MN and MW is comparable to the Rossby deformation radius; here u is highly coherent (r=0.9 at 8 h lag), and the cross-isobath components are fairly correlated, with r=0.24 at a 2 d timescale with MN leading.

Transport calculations were made with daily averages of the 14 d low-passed current and temperature fields from moorings MW and MN using the along-isobath component of the current. First, transport densities (i.e., transport in a water column with 1 m width) were calculated by integrating vertically between 50 and 650 m depth, roughly corresponding to the AW layer. We extended the shallowest available measurement upward to 50 m, for which near-surface data are missing. The gaps in the velocity and temperature profiles vary between the moorings and are summarized in Sect. . A total transport was then estimated by assigning a constant width for each mooring (12.5 km for the outer mooring MW and 7.6 km for the inner mooring MN, justified below). Positive (northeastward, Qp) and negative (southwestward, Qn) transport as well as the net transport (Qp+Qn) in 1 ∘C temperature classes were computed. We estimated the transport of Atlantic Water (Q) as the net transport of water warmer than 5 ∘C. Average transports over the entire record, over summer months and winter months are listed in Table . Results are summarized in Fig. .

The moorings MN and MW are separated by 5.3 km (horizontal distance between the locations), and when projected onto the cross-isobath section the distance is about 5 km (the relative angle between the mooring line orientation and the cross-isobath direction is 20∘). We assume that the velocity measured at each mooring is representative for the half-width (2.5 km) to the next mooring. We further extend the width of MW 10 km off-slope (distance to the 2500 m isobath) and MN 10 km onshore (distance to the 250 m isobath). These choices are motivated by the coverage of the dynamic AW core at the Gimsøy section (see Fig. ). The outer edge corresponds to the location where the 5 ∘C isotherm is shallowest and covers the relatively steep lateral isopycnal gradient toward the slope. The width of the water column for the outer mooring MW used in the transport calculation is then 10+2.5=12.5 km. The width of the inner mooring is 2.5+5.1=7.6 km; 5.1 km is an effective width accounting for the shallowing bottom in the 10 km onshore of MN. The resulting cross-sectional area (600m×7.6km) is equivalent to the area between 50 and 650 m depth obtained by integrating the actual topography to 10 km onshore of the 650 m isobath.

The choice of total width for the transport calculation is consistent with the lateral structure of the depth-integrated geostrophic current inferred from the Gimsøy hydrographic section. From the four occupations of the Gimsøy section, we calculated the geostrophic transport relative to surface pressure. Depth-integrated geostrophic current peaks at an isobath between 500 and 750 m, suggesting that MN and MS are positioned near the maximum velocities of the slope current. The lateral structure of the depth-integrated relative geostrophic current was fairly symmetric and reduced to 20 % of its maximum over a total width of 25–30 km. This lateral structure is also consistent with the vertically integrated geostrophic shear from the annual mean climatology discussed later (red contours in Fig. a). As a result we find that the choice of cross-isobath width extending between the 2500 and 250 m isobaths for transport calculations is justified.

Two moorings closely spaced over the slope cannot resolve the full dynamics of the slope boundary current. However, the comparison with the Gimsøy section suggests that the dynamic core of the slope current can be captured by the mooring records. The individual occupations of the section show that the bulk of the AW is in the upper 650 m, which is resolved by our moorings. The relative geostrophic transports for AW calculated in the Gimsøy section between 50–650 and 50–1500 m were identical to within 0.1 Sv; hence, the limited vertical range of our transport calculation does not introduce additional errors in the baroclinic contribution.

Together with the temporal averages and 1 standard deviation, σ, we also report the standard error of the mean and a representative total transport error estimate. The standard error is calculated as se =σ/n using degrees of freedom (n) and taking into account the decorrelation timescale of 7 d (Sect. ). We calculate a representative transport error estimate for winter, summer and annual data points separately, accounting for the time variability in statistics. At each mooring, we assume root mean squared errors of about 20 % (4 km) in the effective width and 0.05 m s-1 in the depth-averaged current (corresponding to 30 m2 s-1 of transport density). A simple calculation using these figures, ignoring the statistics, would lead to an error of 0.12 Sv. Using the mean and σ of the observed transport density (for winter, summer and all data separately), we generate 100 random data points from a normal distribution and calculate the transport (without imposed error) using a 20.1 km width. The distribution of transport is approximately normal in each season, and this assumed distribution for error analysis is justified. We then generate 100 values for transport density and width from a random distribution with imposed errors and calculate the total transport (with error). The root mean squared value of the difference between transport values with and without error from this 100-point realization gives one error estimate. We draw 1000 bootstrap error estimates and average them to obtain the reported error. The transport error is 0.8 Sv for the annual average, 0.7 Sv for summer and 0.9 Sv for winter. This is typically less than the standard deviation and 3–4 times the standard error (Table ).

There is large variability in Q, with 1 to 4 Sv oscillations at a 2- to 4-week timescale (Fig. a). The transport variability can be due to the current meandering outside the moorings rather than a change in the along-isobath transport. Transport maxima were observed in winter. The transport approached zero at the trough of the oscillations, but the flow reversal was negligible. Total AW transport was typically northward. The monthly averaged transport of AW increased 3-fold in fall and winter, with a monthly average maximum of about 3.6 Sv in December, from about 1–2 Sv in summer (Fig. b). The transport in temperature classes is shown in Fig. . When averaged over summer and winter months separately, transport in high temperature classes (7–9 ∘C) was stronger in winter, whilst the low temperature classes (4–7 ∘C) were stronger in summer. This is because the maximum AW-layer-averaged temperatures occurred in winter (e.g., compare the winter and summer temperature profiles at MW; Fig. a) when the transport was also large (Fig. ). In winter, the vertical mixing of the warm surface layer resulted in a low stratified AW layer of 7–8 ∘C. The largest transport was in the 7–8 ∘C water for both seasons. We hypothesize that the largest warm water transport in winter is a consequence of the annual cycle of depth-averaged temperature coinciding with the time of the strongest barotropic currents in winter. Seasonal variability with transport and temperature maxima in winter and minima in the autumn was also observed in the Svinøy section, with an annual cycle amplitude in currents of about 0.10 m s-1 .

Statistics of the volume transport are summarized in Table . Overall, AW transport averaged over the entire record was 2.0±0.8 Sv (± total error; σ=1.3 Sv, se =0.2 Sv). The winter average (2.9±0.9 Sv) was larger than the summer average (1.6±0.7 Sv), which is significant when considering the se but not the total error. Averaged separately, transport was stronger in summer 2017 relative to summer 2016, increasing from 1.4±0.6 to 1.9±0.7 Sv in summer 2017. The difference was significant relative to the se.

The crude estimate of the width of the slope current must be treated with caution. The sensitivity to the choice of mooring width is approximately linear. Reducing the total effective width by a factor of 2, to 10 km, reduces the mean AW transport from 2.0 to 1.0 Sv. AW transports, on the other hand, are not sensitive to the definition of the AW temperature and vertical integration limits. Recalculating the transport using water with Θ≥3 (instead of Θ≥5) increases Q by less than 0.1 Sv. While the upper layers are characterized by lower-salinity water, the proportion of AW entrained into the upper 50 m should ideally be accounted for in the AW transport estimates. In the core of the slope current between MN and MW, salinity from the hydrographical atlas vertically averaged in the upper 50 m varies between 35.25 and 34.95 g kg-1 (not shown). Assuming shelf waters of salinity less than 34 g kg-1, the fraction of AW in the mixed water would exceed 65 % to 80 %. We limit our estimates at 50 m mainly because of a lack of reliable current measurements. Including the upper 50 m by extending the uppermost available current measurement to the surface and assuming a 100 % AW fraction increases the total mean transport by 0.3 Sv (from 2.0 Sv), well within the error estimates.

There is a substantial transformation of AW between the Svinøy (63∘ N) and Gimsøy (69∘ N) sections, discussed in detail by . Analyses in temperature–salinity space of isopycnal layers described how and where AW was progressively transformed to denser isopycnals. While the most important transformation occurred in the western part of the Lofoten Basin, lateral exchanges generated by instabilities of the slope current substantially modified the characteristics of the AW transported from the Svinøy to Gimsøy section. A climatological view of the hydrography shows the important poleward cooling and freshening of AW (Fig. ). As the AW is modified, isopycnals with a potential density anomaly σ0 less than 27.7 kg m-3 rise. At the core of the slope current, the displacement of the 27.5 isopycnal reaches 150 m, switching from being located below the AW core to above. This isopycnal layer is also where the largest spiciness injection – an indicator of water mass transformation by diapycnal mixing – was reported . Deeper isopycnals sink from Svinøy to Gimsøy, which could be related to the intermediate waters subducted along the Mohn Ridge front and AW transformations in the Lofoten Basin, decreasing the stratification in the AW pycnocline. As a result of winter mixing driven by intense air–sea fluxes, the AW pycnocline in the Lofoten Basin is more diffuse and deeper at around 800 m (vs. 500 m farther south). The cross-isobath temperature and salinity gradients across the slope current also exhibit a different structure, suggesting different contributions to geostrophic currents (via thermal wind balance) and a change with latitude in the baroclinicity of the slope current (Fig. ).

The cross-isobath gradients are relatively weaker at Gimsøy compared to the Svinøy section, and so are the temperature contribution (positive at the slope, negative on the shelf) and haline contribution (negative at the slope, positive on the shelf) to the geostrophic shear (Fig. ). Furthermore, the coastal current core – identified by the positive shear driven by salinity on the shelf – interacts more strongly with the slope current at Gimsøy. This can be explained by the steeper slope of the Lofoten Escarpment, which has a stronger control on the mean position of the slope current. Note that the broader region of isopycnal gradients at Svinøy does not necessarily imply a broader current but could result from a more variable position linked to a weaker steepness of the slope.

To further compare the baroclinicity of the slope current at these two locations, we vertically integrated the different contributions to the geostrophic shear with a level of no motion at the bottom (geostrophic velocity contours in Fig. ). The baroclinicity of the slope current indeed increases with latitude: poleward geostrophic currents exceed 0.6 m s-1 at Gimsøy compared to about 0.4 m s-1 at Svinøy, despite the stronger contribution from vertically integrated shear due to temperature (0.75 m s-1 at Svinøy vs. 0.56 m s-1 at Gimsøy). A strong negative shear due to salinity counterbalances the thermally driven geostrophic shear of the current at Svinøy (reaching -0.31 m s-1 integrated from the bottom to 150 m and -0.25 m s-1 to the surface). At Gimsøy, this value reaches only -0.12 m s-1 from the bottom to 250 m and becomes insignificant when integrated to the surface. This suggests that the cross-isobath salinity gradient is important for the baroclinicity of the slope current, even in a region where temperature accounts for most of the density variations. Changes in the baroclinicity of the slope current can thus be expected following AW salinity interannual variability (e.g., important freshening observed recently by ).

The kinetic energy content and variability of the slope current, as well as conversion rates associated with the barotropic and baroclinic instability of the current, are presently unconstrained by observations. Using our limited mooring records, we attempt to quantify the energetics of the slope current at the Lofoten Escarpment. For the following analysis, we obtained the fluctuations, denoted by primes, by band-pass filtering the hourly data with cutoff frequencies corresponding to 35 h and 14 d.

We start with the variability in depth-averaged along- and cross-isobath currents, the horizontal eddy kinetic energy density, EKE, and their relation to wind forcing. The EKE (J kg-1 or m2 s-2) is 1EKE=12u′2+v′2. The along-isobath current variability and the evolution of EKE were partly forced by the along-isobath wind modulating the geostrophic shear by cross-front Ekman transport (Fig. a–c). The annual average wind speed was 4.3±2.3 m s-1 (in this section the figures after “±” are 1σ). In winter, the average and maximum wind speed values were 6 and 11 m s-1, respectively. Depth-averaged (200–600 m) u at MW was 0.15±0.12 m s-1, with a maximum of 0.6 m s-1 in winter (winter average was 0.25±0.16 m s-1). The maximum correlation between depth-averaged u and the along-isobath component of the wind Wx was obtained at a 2 d lag (r=0.6). While no significant correlation was detected with the cross-isobath component, v increased in amplitude and variability in winter (from its annual average of 0.03±0.04 m s-1) to 0.05±0.07 m s-1, reaching a maximum of 0.26 m s-1.

From 30 d moving averages, EKE was (65±38)×10-4 m2 s-2, with a maximum of 185×10-4 m2 s-2. Daily average values were similar but with a 3 times larger standard deviation and a maximum of 790×10-4 m2 s-2. The maximum EKE was observed in winter, consistent with stronger and favorable downfront winds. When averaged over winter months EKE was (100±41)×10-4 m2 s-2.

An estimate of the baroclinic (BC) and barotropic (BT) conversion rates can be made by assuming no variability in the along-isobath direction and that the cross-isobath gradients dominate. Similar calculations were made in both idealized (channel) model studies e.g., and using mooring array data in the West Spitsbergen Current , in the East Greenland Current , and across the boundary current at the Beaufort shelf break and slope . A positive value of BC indicates conversion from mean potential energy into EKE by growing eddies extracting energy from the mean state. The conversion from mean kinetic energy into EKE is quantified by BT. In this case, the kinetic energy is extracted from the mean flow by eddies transporting along-isobath momentum down the mean velocity gradient e.g., . The baroclinic conversion rate can be approximated by 2BC=gv′ρ′‾∂z∂y, where the cross-isobath velocity fluctuation v′ and the density fluctuation ρ′ are obtained by 35 h–14 d band-pass filtering the hourly data; ρ0=1027 kg m-3 is a reference density, g is gravitational acceleration, and we applied a temporal averaging (overbar) using 30 d moving averaging. The mean isopycnal slope, ∂z/∂y, was calculated as (∂ρ‾/∂y)/(∂ρ‾/∂z). The barotropic conversion rate can be approximated by 3BT=-ρ0u′v′‾∂u‾∂y. As in the BC calculations, fluctuations are the 35 h–14 d band-passed hourly data, and time averaging is over 30 d. While the velocity data coverage is good in both moorings, density (through salinity measurement) measurements are limited. At MW, density measurements are available at target depths of 75, 380, 980 and 1476 m. At MN, the near-bottom sensor (648 m) recorded throughout, but the sensors at 165 and 455 m recorded only until September (the water column line was cut 3 months after the deployment). Note that motion-corrected mooring data were gridded and interpolated. Based on the density measurement coverage, we picked the 400 m level as a representative depth (in AW and in the wedge of the AW current with steep isopycnals; Fig. ) at which we can obtain vertical and lateral gradients but only for 3 months into the record. We calculated the vertical gradient at 400 m at MW using the records at 300 and 500 m and the lateral gradient from the records at 400 m. Whilst the baroclinic conversion rate time series is limited only to 3 months, the barotropic conversion rate can be calculated for the entire duration. We computed BT at 300, 400, 500 and 600 m depths. Results are summarized in Fig. d.

The average barotropic conversion rate (averaged over both moorings, over multiple levels and over 14 months) was (0.3±0.2)×10-4 W m-3. The maximum value reached 1.2×10-4 W m-3. The baroclinic conversion rate (only available in the summer for the first 3 months of the mooring period) was comparable at (0.4±0.6)×10-4, with a maximum of 1.7×10-4 W m-3. For reference, a conversion rate of 10-4 W m-3 for 1 d accounts for ρ0 EKE of O(10) J m-3 or EKE of O(100)×10-4 m2 s-2.

Observed EKE and the conversion rates at the Lofoten Escarpment can be compared to other relevant observations. In Fram Strait analyzed 12-year-long time series from moorings with a focus on the West Spitsbergen Current. EKE at 75 m depth was 50×10-4 m2 s-2 in summer and increased to 200×10-4 m2 s-2 in winter. At 250 m depth the magnitude was approximately reduced to half. These values, overall, are similar to the EKE at the Lofoten slope. In terms of baroclinic and barotropic conversion rates, the two sites are also comparable: in the West Spitsbergen Current BT was on the order of 0.1×10-4 W m-3, and BC at 75 m was 0.5×10-4 W m-3 in summer, increasing to 1.5×10-4 W m-3 in winter. Summer mean and maximum values are identical (within measurement uncertainties) to the corresponding values from our observations at 400 m in summer.

Using a mooring array analyzed the structure and variability of the shelf break East Greenland Current for the period September 2011 to August 2012. EKE at 100 m was up to 700×10-4 m2 s-2 in November 2011 when a reversal of shelf break current was observed; otherwise, typical values varied between 10×10-4 and 100×10-4 m2 s-2, similar to the values at the Lofoten Escarpment. Ignoring the energetic reversal event, BT at 100 m was on the order of 0.1×10-4 W m-3 and BC varied in the range of (1–5)×10-4 W m-3; both conversion rates are similar to those in the West Spitsbergen Current and the slope current.

The conversion rates calculated from our moorings may not be representative of the volume-averaged conversion rates over the slope region. In order to assess this, we compare the observations to high-resolution numerical model results in Sect. .

In order to better interpret the conversion rates obtained from moorings, we calculate volume-averaged conversion rates in the region using the outputs from high-resolution ROMS runs (Sect. .2). We first compute the baroclinic and barotropic conversion rates over a domain covering the slope region identified in Fig. . The conversion rates in a 3D, right-handed coordinate system are formulated in pp. 376–377. The baroclinic conversion rates are computed from 4BC=-ρ0u′b′‾⋅∇b‾/N2=g∂ρ‾∂z-1u′ρ′‾∂ρ‾∂x+v′ρ′‾∂ρ‾∂y. Here u=(u,v) is the horizontal velocity field (the formulation is valid for both the model grid and the along- and cross-isobath rotated coordinate system), b=-gρ/ρ0 is the buoyancy, N2=-gρ0∂ρ‾∂z is the buoyancy frequency, ρ is the potential density referenced to the surface, g is the gravitational acceleration and ρ0=1027 kg m-3 is a reference density. The primes denote deviations from an average state (overbar), averaged over multiple eddy timescales, e.g., for velocity u′=u-u‾. A positive value of BC indicates a transfer of potential energy from the mean flow to eddies.

We calculate the barotropic conversion rates from 5BT=-ρ0u′u′‾⋅∇u‾+v′u′‾⋅∇v‾=-ρ0u′u′‾∂u‾∂x+u′v′‾∂u‾∂y+∂v‾∂x+v′v′‾∂v‾∂y. A positive value of BT indicates a transfer of kinetic energy from the mean flow to eddies.

We compute BC and BT after interpolating the model fields to uniform z levels of 10 m vertical spacing. The time averaging is calculated over monthly windows to avoid any seasonal bias in the eddy fluxes. We arbitrarily chose the year 1999 from the model fields (available from 1996 to the end of 1999). Monthly conversion rates are then averaged vertically between 100 and 1000 m depth (i.e., we exclude the near-surface variability). A global annual average is then obtained by averaging over these 12 months. Results are shown in Fig. .

The baroclinic conversion rates are typically positive and largest along the slope, indicating that potential energy is extracted from the slope current to feed eddies that are generated there. The barotropic conversion rates, on the other hand, show larger spatial variability, with smaller amplitudes and abrupt sign changes along the slope. The baroclinic processes therefore appear to be the main contributor to the conversion of energy from the mean flow to eddies along the slope.

Monthly conversion rates over the slope, volume-averaged over the red box identified in Fig. a and between 100 and 1000 m depth, show that the baroclinic conversion rates dominate (Fig. a), implying that the baroclinic instability of the slope current extracts energy from the mean flow to eddies.

The motivation here is to assess whether the conversion rates obtained from a mooring array are representative of the volume-averaged values. To do this we define a segment across the slope (green in Fig. b) that stretches between the mooring positions of MW and MN and extend it further by 10 km at both sides. We then perform two types of calculations: (1) we compute BT and BC at the model grid resolution and average along the entire segment, and (2) we compute BT and BC using model data from the virtual mooring positions. To be consistent with the observations we apply Eqs. ()–(), rotate the coordinate system to along and across isobaths, calculate BC only at 400 m depth, and vertically average BT between 200 and 600 m. The motivation of performing the segment calculation is to better resolve the lateral shear (based on about 40 grid points compared to only two virtual moorings). The conversion rates are shown in Fig. b.

While there are differences between the segment and virtual mooring estimates (Fig. b), the conversion rates are comparable, with no systematic differences. Lateral shear and isopycnal slopes using only two moorings separated by about 5 km could thus be used in calculations of the conversion rates in one transect. We also note that BT is similar to the observations (blue line in Fig. d), with magnitudes of (0–1)×10-4 W m-3 and maximum values around 1×10-4 W m-3. Observed BC is available only in the summer months (red line in Fig. d) and compares fairly well with the BC from virtual moorings. However, a comparison with the volume-averaged conversion rates shows that calculations using virtual moorings alone overestimate BT, underestimate BC, introduce spurious changes in sign and are not representative of the overall conversion rates on the slope.

The discrepancy in BT is partly due to the different depth averaging (100 to 1000 m vs. 200 to 600 m; note that the latter range is constrained by available observations, whereas the former covers the depth range of interest on the slope region, excluding the upper surface processes) and partly because the volume-averaged calculations include the divergent terms (first and last term in Eq. ) in addition to the terms related to shear (second term). The highly variable spatial structure observed in BT cannot be resolved with a high-resolution single segment or a couple of moorings. Furthermore, volume averaging over BT with changing signs leads to a negligible average BT, which cannot be resolved with the moorings. The discrepancy in BC is mainly because the volume-averaged calculations are based on a depth average between 100 and 1000 m, whereas the mooring calculations are only taken at 400 m depth due to limited observations. BC cannot be captured by the calculations from a single level.

Based on the analysis of the model outputs, we conclude that the mooring-derived conversion rates must be interpreted with caution and may not be representative of the real conversion rates in the region. While we cannot confirm using the limited observations, the model results suggest that the average conversion rates on the Lofoten Escarpment are likely dominated by baroclinic instability of the slope current.

The Norwegian Atlantic Slope Current at the Lofoten Escarpment is described using 14-month-long mooring records in the period from June 2016 to September 2017. Despite the limited number of moorings, the observations resolve the core of the current from 200 to 650 m depth over the shelf break and the upper continental slope. The data set represents the first moored observations on a yearly timescale from this region and offers important constraints on the mean properties, transport rates, temporal variability and energy conversion rates of the slope current.

The 200–600 m averaged current shows an annual cycle with an amplitude of 0.1 m s-1, with the strongest currents in winter, and has a temporal average of 0.15 m s-1. The 14 d low-pass-filtered along-isobath currents reach 0.8 m s-1, lasting for 1 to 2 weeks, and extend as deep as 600 m. The variability in the along-isobath current is partly forced by the along-isobath wind stress, with a maximum correlation of 0.6 at a 2 d lag. In contrast to observations in Svinøy, the slope current is not barotropic and varies strongly with depth (shear of 0.05 to 0.1 m s-1 per 100 m in all seasons).

The average volume transport of Atlantic Water is 2.0±0.8 Sv, with summer and winter averages of 1.6±0.7 and 2.9±0.9 Sv, respectively. The largest transport is associated with warm water in all seasons, and the water temperatures are the highest in winter.

Calculations of the barotropic and baroclinic conversion rates from the moorings are supplemented by results from a high-resolution numerical model. While the conversion from mean kinetic energy into eddy energy (e.g., barotropic instability) is likely negligible over the Lofoten Escarpment, the baroclinic conversion from mean potential energy into eddy kinetic energy (e.g., baroclinic instability) can be substantial, with volume-averaged values on the order of 10-4 W m-3. Eddy kinetic energy and conversion rates in the slope current are comparable to the published results from the West Spitsbergen Current and the East Greenland Current.

Fishing activity in the region makes it highly challenging to maintain moorings; however, extended time series with better cross-isobath and vertical coverage are needed to study the dynamics and variability of the slope current. The attempts to calculate (observation-based) energy conversion rates remain inconclusive. The utilization of autonomous underwater vehicles, such as gliders, can help collect high-quality observations but will be difficult to operate in the strong boundary current. The slope current and its instability are important players in the energetics of the Lofoten Basin and merit further studies.