Temperature and salinity were measured using a Seabird 911plus CTD at all biological sampling locations, including some additional profiles at the continental slopes of transects (Ingvaldsen et al., 2016—see their Figure 1 for CTD station locations). The CTD was equipped with an oxygen sensor (SBE 43) and a Seapoint Chlorophyll Fluorometer and a rosette system for collecting water samples. The conductivity, temperature, depth, and oxygen sensors are serviced and calibrated once a year by the manufacturer (Seabird). In situ water samples for salinity calibration (conductivity sensor) were taken at every station at maximum depth. The resulting accuracies of the pressure, temperature, and salinity measurements are estimated to 0.3 dbar, 0.001, and 0.002°C, respectively. Water samples for Winkler titration of oxygen were not obtained during the current investigations. However, the CTD's SBE43 oxygen sensor was calibrated on 13 February 2013 and data from this sensor was used to obtain a crude evaluation of ambient oxygen levels during the investigation, but also PANGAEA data by Rabe et al. (2014b) for a partly overlapping area (9 CTD stations east of longitude 2°E), in the eastern area of the Fram Strait in late June 2014, have been examined for comparison. The oxygen data of Rabe et al. (2014b) given in μmol/l were converted to ml/l using the ICES Unit conversion tools (http://www.ices.dk/marine-data/tools/Pages/Unit-conversions.aspx) (accessed 11 June 2017), and the relationship: 1 μmol O₂ = 0.022391 ml. As a proxy for phytoplankton biomass we use chlorophyll estimates based on old Seapoint factory calibrated fluorescence data. These should be considered relative values (“μg·l⁻¹, uncalibrated”) comparable between stations and does not imply an absolute measure of phytoplankton biomass (see also “Environmental Setting”).

Current velocities were measured with a RDI Sentinel 300 kHz lowered acoustic Doppler profiler (LADCP) mounted on the CTD carousel. The LADCP data were processed using methods common in the oceanographic community (LDEO-IX-8, Visbeck, 2002) and the barotropic tidal components were removed using the Arctic Ocean Tidal Inverse Model (AOTIM-5, Padman and Erofeeva, 2004). Sea ice concentration for the survey period was obtained from the National Snow and Ice Data Center (NSIDC) (Cavalieri et al., 1996, 1999).

In the current work we present the environmental situation from one transect on the west side of Spitsbergen (Transect 2—Fram Strait North), actually quite similar to the transect further south (Transect 6—Fram Strait South), and one transect north of Svalbard (Transect 4—Hinlopen). These two transects (cf. Figure 2) represent reasonably well the physical oceanography of key areas in the region during the investigations.

Acoustic data for estimation of the distribution and abundance of water column plankton and fish were collected with calibrated EK60 echo sounder split beam systems at the acoustic frequencies 18, 38, and 120 kHz at 1 ms pulse duration. The echo sounders were connected to transducers mounted on a protruding instrument keel with transducer faces ~3 m below the hull, usually ~8.5 m below the sea surface. The lower working threshold in terms of volume backscattering strength (S_v) in dB was set to −82 dB re 1 m⁻¹ (MacLennan et al., 2002). The vessel's EK60 systems are normally calibrated in January every year using standard methods and spheres (Foote et al., 1987; ICES, 2015a) and are known to be very stable over time (Knudsen, 2009). For the period 2010–2016 the vessel's 38 kHz EK60 system showed a <0.1 dB variation in S_v transducer gain.

Multi-frequency scrutinizing and target strength analysis were conducted with the Large Scale Survey System (LSSS) post processing system as described by Korneliussen et al. (2006, 2016), which also was used for exporting files for subsequent analysis by Matlab, Excel, or Systat. The processing involved selection of data to exclude and include, manual removal of noise (acoustic, electric, bubble, temporal noise from e.g., trawl sensors during trawl operations), correction of erroneous bottom detections, and surface originated noise. The allocation of Nautical Area Scattering Coefficient [NASC, s_A (m² nmi⁻²), MacLennan et al., 2002] values to various species or species groups and storage of these values in the database was done for 38 kHz frequency. In the upper ~200 m, where the signal/noise ratio on the 120 kHz echo sounder is above acceptable levels, all three frequencies were taken into account when inspecting the frequency response while below this depth, only 18 and 38 kHz were considered. Sequential thresholding was used to differentiate strong scatterers from weak scatterers. In the process the lower threshold (S_v) (LSSS–color scale, Korneliussen et al., 2016), was moved from the standard −82 dB upwards to a value where only the strongest scatterers remain visible on the echogram (e.g., −60 dB). The s_A corresponding to this S_v threshold was then allotted to the species or species group normally known to have a Target Strength (TS) above this threshold. Subsequently this s_A was subtracted from the total, and the remainder allotted to weak scatterers with TS below this threshold. In the Supplementary Material additional details are presented on the use of “sequential thresholding” and relative frequency response defined according to Korneliussen and Ona (2003) as r(f) ≡ s_v(f)/s_v (38 kHz), where s_v is the volume-backscattering coefficient, and the response at the acoustic frequency f is normalized to that at 38 kHz. Trawl data were used to corroborate the interpretation of the acoustic data. The acoustic backscattering data in the reports were in the form of s_A for 10-m depth intervals in units of (m² nmi⁻²).

The fairly low noise level enabled measurements down to about 800 m, while the main DSL concentrations were found not deeper than 600 m. Total backscatter was allotted using LSSS to the stronger scattering target categories (SC) 0-group fish, cod, capelin, redfish, and others (see ICES, 2015b; Korneliussen et al., 2016), then lumped to the category Strong_SC. The remaining backscatter including the micronekton krill, amphipods, and mesopelagic fish were lumped into the category Weak_SC. The two categories were summed to provide a third, “Total backscattering.” Micronekton as used herein is a combination of fish, krill, and a number of other animals in the 1–20 cm size range, nearly overlapping in size with what we normally term macroplankton (Cartes, 2009).

The above acoustic data for the three final categories were transformed to S_A [Nautical area scattering strength dB re 1 (m² nmi⁻²) by S_A = 10 log10 (s_A) and visualized on five of six transects along the ship's cruise trackline (Figure 2, Table 1) using “EasyKrig_V3.0.1-Matlab2012a,” a Matlab based tool written by Chu (2004, ftp://globec.whoi.edu/pub/software/kriging/easy_krig/; accessed 15 July 2013). The variogram model was the “general exponential-Bessel” and the Ordinary Kriging model was Point to Point with nugget set to 0, sill <1, length around 0.5, power > 1.5, and range ~0.5.

The scrutinized 38 kHz acoustic data were integrated vertically to provide the pattern of horizontal variability along each transect. Data were also averaged horizontally to provide vertical profiles of backscattering for particular subsections along each transect (Table 2). These subsections represent part of transects that were reasonable homogeneous with respect to bathymetry and acoustic backscatter in the DSL over the distance of the subsection, facilitating comparison of these between transects. In addition, acoustic data for selected subsections of transects were integrated to highlight the horizontal variability amongst the Weak_SC and Strong_SC categories along the transects. To compare the DSL in the different areas, the weighted mean depth of the backscattering (WMD) in the depth intervals of 250–600 m for each acoustic sub-section was computed using the following equation:

where z is the depth of interval j, s_A is the nautical area scattering coefficient value for that depth interval, and N is the number of depth intervals. The first transect along the shelf slope edge (Table 1), was not included in these analyses because of the variable bottom depth along the cruise track and the lack of ground-truth tows. All description of the patterns on transects and subsections are based on the 38 kHz data.

Samples of fish, micronekton, and zooplankton were collected with a variety of net and trawl systems (Figure 2). These included the Harstad trawl (Nedreaas and Smedstad, 1987; Godø et al., 1993; Dingsør, 2005), having a circumference of 320 m (Terje Hemnes, pers. comm., Åkrehamn Trålbøteri AS, http://www.tral.no/), although dimensions slightly change when being towed (Underwood et al., 2014), the Macroplankton trawl having a fixed mesh size of 4 × 4 mm from the mouth of the trawl to the cod-end, an approximately rectangular mouth opening of ~38 m² and a 92 m circumference (Melle et al., 2006; Wenneck et al., 2008; Krafft et al., 2010; Heino et al., 2011), the Åkra trawl (Valdemarsen and Misund, 1995), the current version with a trawl circumference of 538 m, the MIK-Ring Net (Munk, 1993; ICES, 2013—3.14 m²/1600 μm mesh size), the Multinet (Weikert and John, 1981—0.25 m²/180 μm mesh size), and the WP2 (0.25 m²)/Juday (0.1 m²) net (Juday, 1916; Working Party 2, 1968; both nets 180 μm mesh size). Trawl speed was ~2.5–3.5 knots, slightly depending on trawl being used and depth of trawling was monitored using a Scanmar depth sensor and trawl sonde. The Macrozooplankton trawl was additionally equipped with a combined Scanmar speed/symmetry sensor to allow the trawl speed through the water to be measured thereby allowing computation of the water volume filtered by the trawl.

The principal zooplankton sampling system was the combined WP2 and Juday net pair mounted on a single frame with two rings on which the net mouths were tied. This system was operated vertically, usually to within 10 m of the bottom, at most stations where the CTD was deployed. The WP2 sample was split and 50% was fixed in borax-buffered 4% formaldehyde for identification and enumeration purposes, while the other 50% was used for biomass estimation. This part was divided into three size fractions using sieves with mesh-sizes 2,000, 1,000, and 180 μm. Most animals retained on the 2,000 μm sieve were sorted, identified, and counted (Chaetognaths, the copepods Paraeuchaeta sp. and C. hyperboreus), while individual lengths of amphipods, fish, krill, and shrimps, were additionally measured after taxonomic identification and prior to rinsing in fresh water. The biomass retained on the 1,000 and 180 μm sieves as well as the identified animals belonging to the aforementioned groups above retained on the 2,000 μm sieve, were put on pre-weighed aluminum dishes and dried in an oven at 60°C overnight, after which they were packed and stored in a freezer at −20° awaiting new drying and weighing at the IMR onshore laboratory. After drying the summed dry biomass per group was measured.

The trawls were used to obtain a qualitative and semi-quantitative understanding of larger micronekton and fish that were present in the acoustic scattering structures observed. Hauls either targeted specific scattering structures (“targeted hauls”) with the aim to identify their constituents (Åkra and Harstad trawls) or they were standardized hauls to enable documentation of important acoustic scatterers in the water column. These included oblique hauls from near the bottom to surface using the Macroplankton trawl both in shallow and deep waters and standardized step-wise 0-group hauls (Dingsør, 2005), conducted in the uppermost 0–20–40 m using the Harstad trawl. With a trawl vertical opening of ~20 m, the sampling range was 0–60 m. For the Åkra trawl and Harstad trawl catches total numbers and wet weight were obtained for each taxonomic group or species being determined. Lengths were recorded for all specimens in the catch, and weight, age (by otoliths), maturity stage, and stomach content were determined for a subsample of the fish catch, while another subsample of the invertebrate part of the catch was worked up to species or genus if possible or to coarser groups like “Amphipoda” or “Euphausiacea” and their numbers and weights determined (±0.1 g). The Macroplankton trawl catches were worked up in a similar way after first determining total wet weight of catch (kg). Normally, all larger fish and jellyfish were determined to nearest possible taxon, counted, and wet weight measured. Because of the scarcity of fish in theses catches, individual lengths and wet weights were normally obtained. The remaining invertebrate catch was subsampled and worked up to nearest possible taxon and their numbers and wet weight determined (±0.1 g). In addition, individual lengths of amphipods, krill, and shrimps were measured (fresh length ± 1 mm). Due to methodological issues, such as depth of trawling, trawl variable mesh size, and differences in trawl mouth opening and mesh size between trawls, catches from the Harstad and Åkra trawl hauls were standardized to kg nmi⁻¹. The Macroplankton trawl catches were standardized to g m⁻³ using a computed haul volume filtered. The tabled numbers of Harstad and Åkra trawl species abundances must be considered indicative of their presence rather than actual abundances.

The Institute of Marine Research (IMR) fully adheres to Norwegian laws relevant to Ethics in Science as well as Animal Welfare. The legal and institutional framework within which IMR operate is detailed in OECD (2012, Part III, Chapter 21, p. 373–398). Under the current legal framework, there is no special permit requirements for at sea research and monitoring activities which do not involve experiments with live animals.

Atlantic Water (temperature >0°C and salinity >34.9, derived from Rudels et al., 2005) dominated from 1,000 m depth up to the surface layer in the Fram Strait North transect (Figures 3a,b). Over-riding the Atlantic Water was a fresher surface layer along most of the transect, although with strong lateral gradients (Figure 3a). In the western part of the transect, the presence of sea ice and melt water (with temperature below 0°C and low salinity) created a pronounced surface layer in the upper 30–40 m. Associated with this melt water layer was low fluorescence-chlorophyll values in the upper 20 m and a sub-surface maximum fluorescence-chlorophyll around 40 m, between the Atlantic Water and melt water. On the eastern side of the transect, the fluorescence-chlorophyll values indicated rather evenly distributed phytoplankton in the top 30 m.

Despite rather weak horizontal temperature gradients in the Atlantic Water, there were strong gradients in velocity (Figure 3b). While the middle of this transect was dominated by strong northward Atlantic Water flow (reaching velocities of 30 cm s⁻¹), both the eastern and western sides showed rather weak flow.

North of Svalbard, at Hinlopen, the northern most part of the transect was dominated by eastward flow of Atlantic Water between 10 and 1,000 m depth (Figures 3c,d). On top of this (in the upper ~10 m) melting sea ice made a fresh, cold surface layer. As opposed to the Fram Strait north, the subsurface chlorophyll maxima under the ice and along transect was much less prominent and did not reach the uppermost 10 m of the water column (Figure 3c). A westward current was evident at the shelf break. Crossing the shelf break onto the shelf, the currents were low and Atlantic Water dominated in most of the water column, except for the innermost (southernmost) part of Hinlopen. Phytoplankton fluorescence was higher on the outer shelf and beyond, than in the innermost parts.

The water column at all stations and to depths of ~2,600 m, seemed very well oxygenated with values in the range ~5.3–7.6 ml l⁻¹, while measurements >300 m were in the range 5.34–5.74 ml l⁻¹ with an average of 5.57 ml l⁻¹ (N = 5,408) based on the Helmer Hanssen Seabird SBE43 oxygen sensor data. However, the PANGAEA data calibrated and corrected using Winkler titration (Rabe et al., 2014b), from the same year and region (cf. section Materials and Methods), show that the oxygen values at depths greater than 300 m, were clearly higher than our values, within the range 6.78–7.28 ml l⁻¹ and with an average of 6.98 ml l⁻¹ (N = 18,665). Thus, the Rabe et al. (2014b), average value for all measurements below 300 m was 1.4 ml l⁻¹ higher than the average value based on our own measurements.

The vertically integrated water column backscattering, s_A, was highest along the western Svalbard shelf break (Transect 1) and across the southern portion of the Fram Strait (Transect 6—Table 3). The Strong_SC backscattering dominated over the Weak_SC except across the northern Svalbard shelf and into the Arctic Ocean (Transects 4 and 5), and only on the Hinlopen section (Transect 4) was the Weak_SC greater than the Strong_SC because there were few strong scatterers (Table 3).

For the Fram Strait South transect (Figure 2) a significant feature was the high backscattering at the surface (Figure 4). It extended from the coast and was most intense at the midway point and dominated by the Strong_SC category (Figure 5) that can also be seen from the total integrated values and for depths <100 m (Figure 6). The very strong backscattering in the upper 100 m (Strong_SC) resulted in the total backscattering for the water column being dominated by this component over the entire transect (Figures 6, 7 and Tables 3, 4). The Strong_SC contribution to the deep-scattering layer from 300 to 450 m was more important than the Weak_SC until midway along the transect to the west. Then the Strong_SC fraction declined and in the western portion of the subsection the Weak_SC accounted for the majority of the backscatter below 100 m depth (Figure 6).

Along the Fram Strait North transect (Figures 4, 5), high backscattering at the surface, also seen in the Fram Strait South transect, was evident, with a major contribution from the Strong_SC fraction, but there were scattered patches of high scattering in the Weak_SC fraction as well. A moderately strong scattering layer extended from the continental slope to the western end of the transect between 300 and 450 m. An important feature was that the DSL Strong_SC fraction was present from the continental slope to about two-thirds of the distance to the west and then became insignificant; the DSL Weak_SC contribution was moderate from the continental slope to the point where the Strong_SC scattering lost significance and then became stronger to the western end of the transect (Figures 4, 5). The abrupt change in the contributions of these two fractions occurred about where there was a cross-over from warm Atlantic water (>6°C; salinity >34.9) to colder melt water (~0.5–2.5°C; salinity ~31.8–33.0) at the sea surface (Figure 3). For the easternmost subsection (T2-1) closest to the continental shelf, Strong_SC backscattering was higher than Weak_SC in the upper 100 m (Figure 6) while the Weak_SC backscattering was higher than the Strong_SC below 100 m. Still the Strong_SC component was dominant in the water column as a whole (Figure 6, Table 4). Further offshore on Fram Strait North (T2-2), Strong_SC backscattering was lower and Weak_SC backscattering was higher throughout most of the water column (Figures 6, 7, Table 4). The pattern observed is nearly identical to the pattern seen for the Fram Strait South.

On the Along-Ice-Edge transect, from the western end of Fram Strait North to the northernmost station close to the start of the Hinlopen transect (Figure 2), the acoustic backscattering paralleled that observed on the Fram Strait transects. There was strong surface to 50 m backscattering until reaching about 140 nmi along the section. For the south-western subsection (T3-1) it was the Weak_SC that dominated the backscattering through most of the water column (Figure 6), including the DSL between 300 and 450 m. For the central deep shelf area subsection (T3-2, Figures 5–7), the Strong_SC component dominated in terms of total integrated backscatter, largely due to that component dominating in the uppermost 100 m of the water column. The variable backscattering further east (Figures 4, 5) is due to Strong_SC scattering that increased to moderate levels when the bottom shoaled to about 250 m (~nautical mile 110). Very low backscattering occurred at the shallowest portion of the transect (about 100 m) at nm 170. In the far northeast area (T3-3), there was again a DSL from 300 to 500 m dominated by high Weak_SC scattering.

The Hinlopen and Wijdefjorden transects across the Northern Svalbard Shelf region and into the deep Arctic slope water had quite similar backscattering patterns (Figure 4). The northern part of the Hinlopen transect had water depths up to 1,850 m, but due to a steep slope and lower quality acoustic data acquired during off-shelf station work, useful acoustic data were recorded to around 700 m bottom depth off the shelf and ended in Hinlopen Strait. Total backscattering was dominated by Weak_SC for the whole transect (Figure 5, Tables 3, 4), and is also reflected in the subsection vertical profiles (T4-1, Figures 5–7, Table 4). In this area, the marine mammal observers noted the presence of a number of whales. On the shelf, water column scattering was moderate and mostly in the upper 50 m with scattered patches of moderate to high Weak_SC backscattering also occurring in the 200–350 m depth zone (T4-2).

The more western Wijdefjorden transect was mostly over very shallow shelf waters. Moderate surface to 50-m backscattering was evident over the shelf and most due to the Weak_SC fraction (Tables 3, 4). Midway along the shelf, near a shallow bottom feature, there was strong scattering by the Strong_SC fraction (Figure 5). To the north beyond the shelf break (Figure 2), there was again strong surface scattering in which both Strong_SC and Weak_SC contributed significantly (T5-2, Figures 5–7) and as in the previous transects, there was strong backscattering centered around 400 m that was dominated by the Weak_SC fraction below 100 m depth (Figure 6).

The Weighted Mean Depth of the DSL backscattering values between 250 and 600 m for all transects averaged 407 m (Table 4). The three deepest WMD values (all > 440 m) were in the Arctic Basin: the Along Ice Edge transect (T3-3), Wijdefjorden (T5-2), and Hinlopen (T4-1). The shallowest (352 m) was in the first subsection (T3-1) along the Ice Edge transect. The integrated DSL values were mostly <50% of the total 600 m water column s_A except in the Arctic Basin where the DSL was about 80% of the total 600 m water column s_A (Table 4).

Irigoien et al. (2014) estimated the mesopelagic backscatter from depths of 200–1,000 m across the world's major ocean basins between 40°N and 40°S to have an average s_A of 1,864 ± 1,341 s.d. (m² nmi⁻²), while individual estimates ranged from 158 to 7,617 m² nmi⁻² at 38 kHz. They concluded that swimbladder “resonance” (the considerable amplification of backscattering that occurs when an object is insonified at a frequency that matches its own natural frequencies of vibration) was not a significant issue in most areas that were covered, which is interesting as a range of ecosystems and certainly quite diverse ensembles of mesopelagic species must have caused the backscattering recorded. Probably most of these communities were far more diverse in terms of number of species and types of swimbladder morphology than the simple mesopelagic community of the northeastern North Atlantic encountered during the current study. Swimbladder “resonance” is a well-known concern in the history of mesopelagic and myctophid acoustics (Sætersdal et al., 1999; Godø et al., 2009; Irigoien et al., 2014; Davison et al., 2015). It can complicate interpretation of s_A-values in time and space, and may result in overestimation of species abundance and/or biomass (see Sætersdal et al., 1999).

What could be interpreted as resonance during the current investigations was observed in the surface waters and most likely produced by 0-group fish as no mesopelagic fish where caught in the 0-60 m hauls at any time of the day (Table 5). The species that dominated our catches at depth (Table 5) and could be responsible for resonance, like B. glaciale, has a swimbladder either partly filled with gas and oil or no gas at all, and also less gas with increasing size (Bardarson, 2014; Scoulding et al., 2015). Modeled estimates based on a swimbladder volume of 1.16 mm³ in a 4.6 cm long B. glaciale gives resonance slightly above 18 kHz at 100 m depth while at 300 m the same sized swimbladder will give resonance slightly below 38 kHz (Bardarson, 2014). At 300 m there is a 15 dB difference in the target strength (TS₁₈ ≈ −75 dB, TS₃₈ ≈ −60 dB) for a B. glaciale of the above size (Bardarson, 2014). However, during the current study there was little evidence of resonance at depth as indicated by the frequency response at 18 kHz relative to 38 kHz, r(18 kHz), that was frequently close to 1.0 over a greater part of the investigated area (see Supplementary Material). Also, the White Barracudina (Arctozenus sp.) that were regularly caught at depth, does not have a swimbladder, implying that it would be a quite weak scatterer in the mesopelagic zone. Larger fish (i.e., cod and redfish) are known to have well developed gas filled swimbladders and a much higher Target Strength (TS) compared to the mesopelagic representatives above. They were, however, caught at depth, but seemed to be associated with the shelf and slope region, although occasionally their distribution was extended further off the continental shelf (Ingvaldsen et al., 2017).

The acoustic transects exhibit some key features (Figure 4). There was strong patchy scattering between the surface and about 50 m throughout most of the study area. In deep waters off the shelves there was strong scattering between 200 and 600 m dominated by larger fish (Strong_SC) close to the slope/shelf break and associated with the warm Atlantic Water moving north toward the Arctic Ocean. West of Svalbard the larger fish component diminished rapidly westwards, while the Weak_SC scatterers were increasingly dominant. There were no observed signals in temperature or salinity in the deeper portion of the water column that could be associated with the change in scattering structures across the deeper part of these western transects. However, the location of the Arctic front at the surface corresponded roughly to where the DSL changed from being dominated by larger fish to being dominated by Weak_SC at depth (Figures 4, 6). Also, in the western part of the transects, the Atlantic Water flow was weaker and sea ice and melt water dominated in the surface layer. In addition, west of the core Atlantic inflow along the Svalbard shelf there was almost no northwards flow (cf. Figure 3) over the entire water column. A similar pattern in acoustic backscattering at 38 kHz was observed by Dale et al. (1999) during an east-west transect from the Bear Island (74°30′N, 19°01′E) along latitudes 74 and 75°N to ~10°W in November 1995. The mesopelagic scatterers in the deeper part of the water column (>250 m) decreased substantially around 150 km west of Bjørnøya (~14°E) when approaching Arctic water masses.

In a 1993 survey of the Norwegian Sea (62–70°N) during June/July, the “deeper layer” (~200–500 m) DSL was 40% or more of the water column integrated backscatter and decreased in the total backscattering from south to north by approximately a factor of 2 (Figure 9; Melle et al., 1993). The integrated area backscattering coefficient (s_A–m² nmi⁻²), averaged over transects where bottom depth exceeded 509 m, ranged between 180 and 460 m² nmi⁻². Melle et al. (1993) found no clear-cut temporal change between April, May, and June/July in the DSL acoustic layers. On the Norwegian Sea slope (65°00′N, 04°50′E) in June 2000 in 844 m of water, Knutsen and Serigstad (2001) reported values for the DSL of 400–2,000 m² nmi⁻².

In the Irminger Sea, Magnússon (1996) observed a DSL between 400 and 800 m and latitudes ~60–68°N, although its vertical position varied somewhat depending on time of the year. Overall the integrated values over the water column ranged from 1,000 to 5,000 m² nmi⁻² and occasionally higher. In the same region, Anderson et al. (2005) showed integrated values for the 0–800 m water column during daytime in the same range (~600–2,000 m² nmi⁻²). The values of Magnússon (1996) and Anderson et al. (2005) are considerably above values for the region west and north of Svalbard and also generally higher than the average values for transects found in the south-eastern part of the Norwegian Sea around 62–64°N and 3–6°E (Melle et al., 1993; Knutsen and Serigstad, 2001). The Irminger Sea values are for the whole water column and are not entirely comparable to those of Melle et al. (1993). However, Anderson et al. (2005) reported daytime values in the 0–200 m layer that in most instances were <62 m² nmi⁻², while night-time values ranged from 42 to 235.8 m² nmi⁻², indicating that they are on the order of 10–100 times lower than the total water column (0–800 m) backscattering, thus emphasizing the importance and magnitude of the DSL itself.

The values recorded for the Irminger Sea (Magnússon, 1996; Sigurðsson et al., 2002; Anderson et al., 2005) and the Norwegian Sea region (Melle et al., 1993; Knutsen and Serigstad, 2001) fit the global picture. The values recorded for the Irminger Sea are considerably higher than those recorded for the eastern Norwegian Sea, while there is definite reduction toward the northern region as examined by Melle et al. (1993), approaching minimum values recorded globally by Irigoien et al. (2014). In a recent study, exploring changes in acoustic backscatter from off Norway to North of Iceland, Norheim et al. (2016) found that the integrated backscatter (s_A, m² nmi⁻²) considerably decreased from ~360 in the east to 15 m² nmi⁻² in the northwest. A similar South to North pattern was observed in our data combined with those of Melle et al. (1993), and Siegelman-Charbit and Planque (2016). Thus, moving from the south-eastern part of the Norwegian Sea north to Svalbard along the path of the Atlantic inflow, the mesopelagic s_A decreased substantially.

In the Svalbard region, the NASC integrated values for the DSL over 250–600 m, were mostly <60 m² nmi⁻² (Figure 9, Table 4) and compared to Melle et al. (1993) this is a factor of ~6–10 lower than the highest values found around 62°30′N, (Figure 9, t7). Only in the slope region of the northern Svalbard shelf did values range between ~100 and 150 m² nmi⁻². It is also interesting to note that the Weak_SC, the fraction that holds the weaker plankton and micronekton, was the main contributor to the scattering recorded in the DSL for the northern part of the three transects that stretches into the southern part of Sofiadjupet. Seasonal and year to year variability of the DSL in Svalbard waters might be greater than currently thought, but more data will be required to quantify such variability. Another important unknown, yet to be resolved, is the extension of the DSL northwards and westwards beyond the region of our study.

The DSL in the eastern part of the Norwegian Sea is associated with the Atlantic Water flowing north along the Norwegian coast (Melle et al., 1993; Torgersen et al., 1997; Dale et al., 1999; Knutsen and Serigstad, 2001; Norheim et al., 2016; Siegelman-Charbit and Planque, 2016). The magnitude of this inflow is diminished northwards due to flow of part of the Atlantic Water into the Barents Sea (Loeng, 1991) and retroflection of a portion of it westwards in Fram Strait (Nöthig et al., 2015; Nilsen et al., 2016). The remainder continues north along the Svalbard shelf into the Arctic Ocean proper. The deep and shallow biological constituents that passively follow the currents are reduced in numbers and biomass northwards toward the Arctic Ocean as portions of the Atlantic water is diverted (Figure 9).

In the southern Norwegian Sea, the DSL consisted of krill (M. norvegica), mesopelagic fish (Maurolicus muelleri, B. glaciale), shrimps (S. arcticus, Pasiphaea multidenta), and jellyfish (P. periphylla) (Melle et al., 1993; Torgersen et al., 1997; Knutsen and Serigstad, 2001), but also included blue whiting (Micromesistius poutassou). The composition of the DSL west and north of Svalbard was similar indicating that mesopelagic animals from the south contribute to the mesopelagic community further north along the western Svalbard archipelago and beyond. However, there are some important differences. Melle et al. (1993) stated that “the biomass of the layer increased from <10% of total biomass in the north to 20–50% in the south.” Both Melle et al. (1993) and Torgersen et al. (1997) report that the DSL was composed of two distinct layers in the south-eastern Norwegian Sea. Norheim et al. (2016) observed strong backscatter off Norway, spread over a relatively large depth range and in some areas the DSL had two layers. All authors suggest that the shallower layer involved the presence of the pearlside M. muelleri (Sternoptychidae). North of the region examined by Melle et al. (1993—Figure 1B, hatched area) at 69°N, 9.4°E, Kaartvedt et al. (1998) found M. muelleri in small dense schools in the upper layer during light summer nights, but there were few signs of the “typical” scattering structures found further south normally associated with M. muelleri (Melle et al., 1993; Torgersen et al., 1997, Norheim et al., 2016). However, total integrated values of Maurolicus were much lower here compared to the Storfjorden (Kaartvedt et al., 1998), and to the off-shelf registrations described in Melle et al. (1993) and Torgersen et al. (1997).

Only two specimens of M. muelleri and one specimen of blue whiting (M. poutassou) were collected in the 29 pelagic hauls in the Svalbard region. Hence, key components of the acoustic backscattering structures further south along the Norwegian coast, which are much more abundant here (Melle et al., 1993; Kaartvedt et al., 1998; Knutsen and Serigstad, 2001—~0.5–100 kg nmi⁻¹), diminish rapidly northwards. Since the environmental conditions in the core of the Atlantic current flowing northwards do not change dramatically, other factors including a changing light environment, loss of organisms through retroflection of Atlantic water toward colder regions in the west, predator–prey interaction, food availability as well as reproductive and recruitment constraints could all contribute to the observed patterns (cf. Kaartvedt et al., 1998; Kaartvedt, 2008). The mechanisms most important and responsible for the current observations are currently difficult to determine. Further ecosystem change should provide new insights as to which mechanisms are or have been at play.

On the Fram Strait South and North transects, the DSL was dominated by larger fish along the continental slope and the Weak_SC further offshore to the west. Although the shift in these DSL components coincided with changes in the surface water physical and biological structure (Figures 3, 4), it is more likely that the Strong_SC dominance was associated with fish such as redfish, and cod that normally reside in association with the shelf and slope, but here extended its distribution horizontally away from the shelf in search of prey (cf. Ingvaldsen et al., 2017). Moving northeastwards along the ice edge north of Svalbard and into colder surface waters, the large fish component was increasingly less important while the Weak_SC made up the majority of backscattering recorded.

Trawl samples taken on Hinlopen transect documented the shift in the acoustic patterns and structures found further west, as the key organisms on this transect were krill, particularly T. inermis and the amphipod T. libellula (Tables 5, 6). At the northernmost station (St20, 80°48′N, 15°40′E), where the DSL was fished, the krill M. norvegica and T. inermis dominated the catches, but typical mesopelagic organisms as the shrimp H. glacialis, the glacier lanternfish B. glaciale, and others were also present. The fauna still resembled the species composition further south in the Norwegian Sea (Melle et al., 1993; Torgersen et al., 1997; Knutsen and Serigstad, 2001).

The Weighted Mean Depth (WMD) of the DSL around Svalbard was somewhat deeper in the northernmost area compared to further south. The DSL appears to be following the Atlantic Water as it moves in and under the Arctic surface water and is displaced downward (Gjøsæter et al., 2017). This may explain the deeper depth distribution there and would be a further link between the biota and the water masses. On the other hand, several other factors, such as solar irradiance, in situ light conditions that will depend on various optical properties of the overlaying water masses (e.g., Lund-Hansen et al., 2015), or other environmental parameters (cf. Netburn and Koslow, 2015; Klevjer et al., 2016) might also affect the position of the DSL. However, the water column to the greatest depths seemed very well oxygenated although our own data might be slightly offset since the SBE43 measurements were not checked by Winkler titration of water samples. Oceanographic data published by Rabe et al. (2014b) for a partly overlapping area, show however, that the eastern area of the Fram Strait in June 2014 below 300 m depth, was very well oxygenated with high oxygen values in the range 6.78–7.28 ml l⁻¹, and it is thus believed that here were no constraints regarding oxygen concentration on the WMD of the organisms constituting the DSL in the area of these investigations. To further our understanding of the interplay between environmental variables and how they might impact the vertical extent of the DSL, in situ light data will be essential in future investigations.

In the epipelagial, the shallow scattering layer found close to the surface (~0–50 m) was a consistent feature of the region surveyed. An important constituent of this surface layer was the 0-group Sebastes spp., and this 30–50 mm fish was a major contributor to the acoustic backscattering observed in this layer, particularly on the west side of Svalbard. North of Svalbard, the abundance of the small 0-group Sebastes spp. was considerably reduced while the krill species M. norvegica and T. inermis and the hyperiid amphipod T. libellula were observed in much higher abundances than further to the south-west. These species compensated for the reduced abundance of Sebastes spp. in the acoustic backscattering observed.

On all transects the mesozooplankton biomass was very high, ranging from ~10 to 70 g DW m⁻² (Figure 8). An important constituent in the upper 50 m was the copepodite stages CI–CV of C. finmarchicus and smaller copepods like Oithona. These groups were abundant in the upper ~0–50 m as indicated by stratified Multinet tows (not shown) and likely were associated with a marked subsurface fluorescence maximum observed at ~25–30 m depth on the South and North Fram Strait transects.

The Strong_SC backscatter in the surface layer was reduced whilst moving north-east into the shelf and slope region north of Svalbard. The acoustics and biological sampling showed patches of smaller 0-group fish, but very few pelagic fish foraging on the large quantities of mesozooplankton. This food resource however was likely exploited by the 0-group fish community and it may also be important for mesopelagic fish and other predators residing in the DSL once the zooplankton enter diapause (Pepin, 2013). Baleen whales frequently sighted in deeper parts of the northern Svalbard shelf were also likely feeding on epipelagic concentrations of krill and hyperiid amphipod macroplankton that were frequently observed here.

The extent to which the mesozooplankton resource is exploited directly or indirectly by mesopelagic fish and associated deep water components is currently unknown, but needs to be addressed in future investigations. Acoustic monitoring of the water column including the mesopelagic domain along with ground-truthed species composition data could be an effective method to examine how changes in the epipelagic zone affect the deep-water communities of the Arctic Ocean. Such initiatives are strongly recommended (Rogers, 2015). The magnitude of DSL acoustic backscattering and the determination of its biological components both to the west of the Spitsbergen archipelago and northwards into the Arctic Ocean should be evaluated on a regular basis as an important additional method to properly assess ecosystem change, although methodological caveats and challenges still exist (Handegard et al., 2013; Davison et al., 2015).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.