Sognefjord is located on the western Norwegian coast, extends to about 205 km and has a maximum water depth of 1308 m (Poremba and Jeskulke, 1995; Storesund et al., 2017). The fjord has a 3-layer water column structure (Svendsen, 2006; Storesund et al., 2017): the top layer is brackish water (salinity ≤ 33 psu), formed by the mixture of freshwater runoff and seawater, and moves out of the fjord; the intermediate layer (salinity between 33 and 35 psu) is well-oxygenated, owing to exchanges with the Norwegian Coastal Current above the sill depth, and hosts compensatory flows that may be in-fjord or out-fjord; and the bottom, or basin, water below the sill depth (salinity > 35 psu after deep-water renewal) originates from Atlantic water and becomes gradually less dense between renewal events because of diffusion and mixing with the layer above, driven by tidal forcing. Whilst the shallow sill (170 m) at the mouth of the fjord allows for some exchange of water between the fjord system and the adjacent coastal water, the mixing is fairly reduced and strong stratification occurs because of the influx of terrestrial runoff (Storesund et al., 2017). Renewal of the basin water occurs approximately every 8 years (Buhl-Mortensen et al., 2020).

Conditions within the fjord basin below sill depth are fairly stable and homogeneous. Water temperature is consistently around 7.4°C and the salinity is generally greater than 35.0 psu (Witte et al., 2003; Storesund et al., 2017). Oxygen concentrations have been found to decrease with increasing distance from the sill and with depth, where Storesund et al. (2017) found the inner fjord and lower basin water to have oxygen concentrations below 4.5 mL L^–1 in May and November 2012.

Two video transects were performed with the work-class ROV Ægir 6000 in Sognefjord in July 2017 during a SponGES cruise with the R.V. G.O. Sars (Figure 1). Dive 1 (D1) was conducted within the central fjord (1155 – 85 m; 61° 6′ N, 6° 39′ E) and Dive 2 (D2) was located near the sill (1230 – 55 m; 61° 3′ N, 5° 22′ E), respectively. D1 and D2 had approximate transect lengths of 2.43 km and 1.39 km, and were located approximately 79 km and 17 km away from the sill, respectively. The approximate fjord side slopes for D1 and D2 are 43° and 44° inclined to the horizontal plane. Physical samples were collected by the ROV during the transects to help confirm identifications of fauna.

One ship-based conductivity–temperature–depth (CTD) sensor package cast was conducted for each transect to profile temperature (°C), salinity (psu), dissolved oxygen (mL L^–1) and chlorophyll a concentration (μg L^–1) throughout the water column. The bottom depths of the CTD casts corresponding to D1 and D2 were 1017 m and 1223 m, respectively. In addition, water samples were collected (using a rosette water sampler, on which the CTD package was mounted) from the different water masses for nutrient analysis. From the video footage, the percent cover of exposed hard substrate and soft sediment was estimated for each image analyzed.

Seawater samples for the analysis of inorganic dissolved nutrients (Si, PO₄, NH₄, NO₃, and NO₂) were collected with the CTD-rosette from selected depths. Sample depths were selected based on the profile of the CTD downcast, whereby samples were collected from five different depths. Subsamples were collected in 50 mL syringes, which were rinsed three times with water from the niskin bottles of the CTD rosette before being filled. After sampling on deck, samples were filtered through 0.2 μm filters and instantly sub-sampled into two vials, one of which was used for samples of ortho-phosphate (PO₄), ammonium (NH₄), nitrate (NO₃), and nitrite (NO₂), stored at −20°C and the other for silicate (Si), stored at 4°C. Nutrients were analyzed at NIOZ with a QuAAtro Gas Segmented Continuous Flow Analyzer. Measurements were made simultaneously on four channels for PO₄ (Murphy and Riley, 1962), NH₄ (Helder and De Vries, 1979), NO₂, and NO₃ (Grasshoff et al., 1983). Si was measured during a separate run (Strickland and Parsons, 1968). All measurements were calibrated with standards diluted in low nutrient seawater. For a detailed description of the sampling procedure we refer to Roberts et al. (2018).

Still images were extracted from the videos approximately every 30 s to ensure there was no spatial overlap between images during analysis. Due to fluctuations in ROV altitude and changes in turbidity or topography, some areas of the transects were not suitable for analysis. Images were excluded if they contained any of the following characteristics: (1) image area obscured by part of the ROV or suspended material, (2) ROV was collecting physical specimens, (3) ROV was too far from the substrate (>10 m), (4) camera angle was not seabed-facing, (5) image contained poor light visibility, (6) image was blurred, and (7) overlapping image area. Videos were rescanned during image annotation to help identify individuals that were difficult to decipher in the stills alone, and in some cases, new stills were extracted if a more suitable image was present ±5 s from the original still. This was mostly the case for images that were blurred or contained poor light visibility and more suitable images of the same area were available within 5 s of the original image extracted. Parallel lasers of known separation (which project spots onto the seabed in order to determine image scale) were only active for the first hour of D1. It was therefore not possible to determine area (m²) and density (individuals m^–2) for the survey and megafauna were enumerated based on individuals per image.

ImageJ (version 1.52) was used to annotate the extracted imagery. A virtual grid with 496 cells was overlaid on each image in ImageJ. The grid size was selected because the cells completely overlapped the images. Percent cover of substrate type was estimated by counting the number of grid cells that contained that particular substrate type, then calculating the percentage of the total number of grid cells represented by that value. The grid resolution was selected such that the cells were small enough to minimize the number containing multiple substrate types as a proportion of total number of cells (and thereby increase the precision in the percent cover estimates) and large enough that the gridlines were not so dense as to obscure the fauna present in the image. In cases where cells contained multiple substrate types, the cell was characterized as the substrate that covered more than half of the cell. All epibenthic megafauna individuals that were easily visible within the imagery were counted and identified to the lowest taxonomic level. Due to fluctuations in altitude during video transects, some fauna had to be identified based on gross morphology (e.g., white sponge 1, yellow sponge 2, etc.). Taxa were classified as rare if represented by three or less individuals.

In general, the water was slightly warmer, more saline, and more oxygenated in the water column below the sill at D2 (near-sill), relative to the same depths in D1 (central). The top layer of water was made up of warm brackish water (Figure 2), and a sub-surface chlorophyll maximum occurred at the halocline between the brackish surface water and intermediate water (approximately at 20 to 30 m). At approximately 80 to 100 m, dissolved oxygen decreased. Dissolved oxygen levels recovered in the intermediate water just below the sill depth. There was a gradual transition between the intermediate water mass and basin water until about 300 m, where a drop in temperature and dissolved oxygen levels occurred. The basin water was fairly homogeneous and there was not much difference in the water properties between the dives, where temperature was around 7.5°C, salinity at 35.06 psu, and dissolved oxygen around 4.2 mL L^–1. D1 was frequently covered in soft sediment with exposed hard substrate patches throughout the transect, whereas D2 had distinct regions of solely soft sediment (e.g., bottom of fjord basin) and exposed hard substrate (e.g., along slopes and cliffs). Within the Intermediate Zone, small boulders and rocks became more frequent, and the sediment type became more coarse.

Inorganic nutrient concentrations increased with depth (Table 1), showing highest concentrations below 450 m water depth, while lowest concentrations were measured in surface waters at both sites. Nutrient concentrations did not show large differences between the two sites.

In D1, from the suspended particulates observed in the video footage, predominately vertical settling appeared to occur. In these regions, sessile fauna and vertical rock walls were covered with a fine layer of particulate matter. However, D2 had higher observed turbidity throughout the dive compared to D1, with the settling of particulate matter apparent and some evidence of horizontal flow based on the position of the feeding apparatus of filter feeders and particulate direction changes.

In the PCA ordination, the first two principal components (PC) represented approximately 76% of the environmental variability within the two dives combined (47.1% and 28.6% for PC1 and PC2, respectively) (Figure 3). The PCA ordination showed that the images within the Basin Zones of both D1 and D2 were clearly distinguishable from the Intermediate and Above Sill Zones along the axes of PC1 and PC2. Dissolved oxygen (Eigenvector = 0.503), salinity (Eigenvector = −0.498), and depth (Eigenvector = −0.490) had the strongest influence on PC1. Percent cover of soft sediment (Eigenvector = −0.623) and exposed hard substrate (Eigenvector = 0.613) most strongly influenced PC2.

In total, D1 had 79 taxa with a total of 11557 individuals and D2 had 89 taxa with a total of 10615 individuals (Table 2). After rare taxa were excluded from the dataset, there was a total of 22105 individuals and 72 taxa identified within 511 images, where D1 had 11528 individuals and 57 taxa and D2 had 10577 individuals and 63 taxa. Porifera made up the majority of the taxa (24), followed by Echinodermata (16), and Cnidaria (10). The Above Sill Zone for both D1 and D2 had the lowest total abundances and number of species compared to the other zones.

The top 10 most abundant taxa for D1 were White Encrusting Sponge 1, Psolus squamatus (O.F. Müller, 1776), Hymedesmia sp., Brachiopoda 1, Munida tenuimana Sars, 1872, Gracilechinus acutus (Lamarck, 1816), Polymastia nivea (Hansen, 1885), Gracilechinus elegans (Düben & Koren, 1844), Yellow Encrusting Sponge 1, and Stylocordyla borealis (Lovén, 1868). For D2, the top 10 most abundant taxa were the White Encrusting Sponge 1, Echinoidea 1, P. squamatus, Hymedesmia sp., G. acutus, M. tenuimana, Yellow Encrusting Sponge 1, G. elegans, Acesta excavata (Fabricius, 1779), and Phakellia spp.

Within the Basin Zone, exposed hard substrate was often covered with polychaete tubes, the bivalve A. excavata, cnidarians, and encrusting sponges (Figure 4). Acesta excavata was present in high densities when observed on vertical rocky walls, often with juveniles and dense accumulations of encrusting polychaetes nearby. The octocoral Anthomastus grandiflorus (Verril, 1878) (Figure 4b) and large glass sponges Asconema aff. foliatum (Fristedt, 1887) (Figure 4e) occurred only on vertical rock walls. In D1, A. aff. foliatum was covered by a fine layer of suspended particulate matter (Figure 4e).

Soft bottom areas were characterized by the enigmatic asteroid Hymenodiscus coronata (Sars, 1871), M. tenuimana, Bathyplotes natans (Sars, 1868), Mesothuria intestinalis (Ascanius, 1805), Psilaster andromeda (Müller & Troschel, 1842), small patches of Kophobelemnon stelliferum (Müller, 1776) and carnivorous sponges, and in rare cases, Virgularia mirabilis (Müller, 1776). Signs of lebensspuren such as burrows containing M. tenuimana within or nearby were observed throughout both dives (Figures 4n,p).

In regions with mixed substrate types (e.g., exposed hard substrate and soft sediment), Psolus squamatus was often observed concentrated at breaks in the slope or protruding surfaces. Phakellia spp., Phakellia ventilabrum (Linnaeus, 1767), and Axinella infundibuliformis (Linnaeus, 1759) were commonly positioned along slopes, aligned with the direction of observed horizontal particle flow (Figure 4m). Large anemones like Bolocera tuediae (Johnston, 1832) were observed residing on exposed hard substrate walls with soft sediment surrounding the substrate.

In the Intermediate and Above Sill Zone of D2, it should also be noted that there was a sudden occurrence of Echinoidea 1 in large quantities (max. 859 individuals per image) from 240 m to 60 m water depth (Figure 4u). For D2, the Above Sill Zone had a higher proportion of echinoderms present (Figures 4v,y). Numerous fish species, including Chimaera monstrosa (Linnaeus, 1758), Sebastes viviparus (Krøyer, 1845) (Figure 4v), and Coryphaenoides rupestris (Gunnerus, 1765), were observed within the upper regions of D1 and D2. Anthropogenic waste was found in the D1 Above Sill Zone (Figure 4w).

The community composition of the non-rare taxa showed significant differences between the depth zones and dives (ANOSIM Global R: 0.261, p = 0.001) (Figure 5). For D1, the nMDS plots and pairwise ANOSIM test indicated that the Basin and Intermediate Zones shared a similar community composition (p > 0.05). All other zones showed significant differences in community composition.

The SIMPER analysis revealed that for both dives, all zones except for D2-Above Sill had P. squamatus, Hymedesmia sp., and White Encrusting Sponge 1 within the top five contributing taxa (Table 3). Taxa from the Echinodermata were more common in D2 than D1. For both dives, the Basin and Above Sill Zones had the largest difference in community composition (SIMPER, D1: Average dissimilarity = 63.03%; D2: Average dissimilarity = 78.87%). Munida tenuimana, Parastichopus tremulus (Gunnerus, 1767), Brachiopoda 1, G. elegans, and Pteraster militaris (O.F. Müller, 1776) contributed most to the differences between the Basin and Above Sill communities for D1. For D2, M. tenuimana, P. squamatus, G. acutus, Hymedesmia sp., and White Encrusting Sponge 1 contributed most to the differences between the two zones.

The diversity indices for the D2 (near-sill) zones were consistently lower than those of the corresponding zones in D1 (central) (Figure 6). For both dives, the diversity indices for the Above Sill Zones were lower than those of the deeper zones. The diversity indices all passed the Levene’s test of homogeneity (p > 0.05), and the one-way ANOVA indicated that there were significant differences (p < 0.001) in the indices for D1 and D2 and their respective depth zones. The species richness and diversity were statistically significantly different between the D1 and D2 respective Basin Zones (Tukey HSD: p < 0.01). There was a significant difference in species richness between the Basin and Above Sill Zones for D1 (Tukey HSD: p = 0.03) and trend toward significant difference between the Basin and Above Sill Zones for D2 (Tukey HSD: p = 0.069). For the Shannon-Wiener diversity index, there was a significant difference between the Basin and Above Sill Zones for D2 (Tukey HSD: p = 0.005), and a significant difference between the Intermediate and Above Sill Zones (Tukey HSD: p = 0.047).

Salinity and dissolved oxygen were the most influential variables on the diversity indices when considered separately, as revealed by GLMs (Table 4 and Supplementary Figure S3), and depth had little influence alone. For species richness, the combination of depth and dissolved oxygen explained 10.85% of the deviance within the dataset. For diversity, the combination of depth, salinity, and dissolved oxygen explained 19.01% of the deviance in the dataset.

In general, much of the same taxa composition was observed in both dives for depth zones below the sill depth. The largest difference in megabenthic community composition was found between the deepest and shallowest zones for both dives, and similar trends have been observed in other surveys (Starmans et al., 1999; Sswat et al., 2015; Molina et al., 2019). In the present study, the Basin Zone was characterized more by sessile fauna (e.g., P. squamatus, Acesta excavata, Hymenodiscus coronata, and sponges) and M. tenuimana, whereas the Above Sill Zone was more dominated by echinoderms and anemones.

Contrary to numerous fjord and shelf-based studies (Buhl-Mortensen and Høisaeter, 1993; Holte et al., 2004; Webb et al., 2009; Włodarska-Kowalczuk et al., 2012; Sswat et al., 2015; Gasbarro et al., 2018; Molina et al., 2019), we find that communities at the shallowest depths (Above Sill Zone) and in closer proximity to the sill (D2) have the lowest number of species and diversity. Buhl-Mortensen et al. (2017, 2020) observed a similar trend in relation to the proximity to sill, where the outer region in Sognefjord had lower species richness compared to the middle region (Buhl-Mortensen et al., 2017, 2020). However, Buhl-Mortensen et al. (2020) observed a trend of decreasing species richness with increasing depth, which was not observed in the present study. The trends observed in the present study are more consistent with shelf community patterns observed by Starmans et al. (1999), where shallower regions contained a lower number of highly abundant taxa than the deeper stations and diversity increased with increasing water depth. This reduction in species richness and diversity in D2 (near-sill) and the areas above the sill could be driven by changes in water mass characteristics or increased physical stress on the benthic communities as the environmental conditions become less stable (Starmans et al., 1999; Jones et al., 2007).

The basin waters of both dives were fairly homogeneous (Storesund et al., 2017) and likely contributed to the homogeneity observed in the species composition in the deeper regions. Water mass properties (e.g., temperature, salinity, dissolved oxygen) play a significant role in megabenthic community composition (Buhl-Mortensen and Høisaeter, 1993; Williams et al., 2010: Meyer et al., 2015), which appears to be the case for Sognefjord as well. The changes in species composition appear to gradually occur around the transition between the basin and intermediate water masses, which is at approximately 300 m (Storesund et al., 2017), and more clearly near the sill depth. Studies have shown that sills affect water mass dynamics in ways that are critical to the structuring of benthic communities (Strømgren, 1970; Rüggeberg et al., 2011).

As Buhl-Mortensen and Høisaeter (1993) stated, the environment in fjord basins is influenced by the sill depth. With shallow sills, organic matter becomes trapped within the inner fjord below the sill depth and is not flushed out readily by the adjacent coastal water (Klitgaard-Kristensen and Buhl-Mortensen, 1999). As such, it is possible that organic input from renewal events and terrestrial sources (e.g., rivers, runoff, snowmelt, etc.) accumulates and has longer residence times in fjord basins (relative to shallower waters), providing food and nutrients to the benthic fauna below the sill depth. In a recent study of the Sognefjord by Buhl-Mortensen et al. (2020), the authors observed continuous detritus cover on sloping bedrocks at depths greater than 400 m and terrestrial organic material mixed in with the basin’s soft sediment. The observed higher species richness and presence of deposit-feeding holothurians (Bathyplotes natans and Mesothuria intestinalis) and suspension-feeding Hymenodiscus coronata in the Basin Zone’s soft bottom regions indicate availability of organic matter to the basin floor (Roberts and Moore, 1997; Flach et al., 1998; Amaro et al., 2015).

The vertical-falling particulate matter along the rocky walls observed in D1 is likely an important food source for many of the filter- and suspension-feeders (e.g., encrusting sponges, Asconema aff. foliatum, encrusting polychaetes, and Acesta exacta) residing on the vertical rock walls or under overhangs. Areas of flow acceleration owing to irregular topography (e.g., ridges, peaks, and other elevated substrate) experience increased particle fluxes and are also likely important for suspension feeders (e.g., Psolus squamatus, Phakellia spp., Phakellia ventilabrum, and Axinella infundibuliformis) (Flach et al., 1998; Buhl-Mortensen et al., 2020). As is common in fjord environments, it is likely that the quality of food is lower in the basin and inner fjord compared to regions nearer to the sill (Klitgaard-Kristensen and Buhl-Mortensen, 1999). However, the higher species richness and presence of suspension- and deposit-feeders within the basin suggests the fauna may be adapted to the low quality of food, or that this is compensated by the stability of the basin environment. It is clear that a more rigorous study should be conducted to quantify and assess the quality of the organic matter supplied to the basin communities.

Depth, salinity, and dissolved oxygen were highlighted as important variables for the diversity indices. Depth acts as a proxy for other factors and it is likely that parameters which were not accounted for in the present study (e.g., food availability, particulate organic matter, localized hydrodynamics, pollution) are also influencing the patterns observed (Jones et al., 2007; Webb et al., 2009; Williams et al., 2010). Dissolved oxygen and percentage cover of substrate type varied most between the two dives, both of which are known to be critical for many benthic habitats (Holte et al., 2005; Williams et al., 2010; Sswat et al., 2015). The availability of hard substrate is important for sessile invertebrates (Williams et al., 2010; Buhl-Mortensen et al., 2012), and in this study, regions with exposed hard substrate were often covered with sponges, serpulid worm tubes, bivalves, and holothurians, similar to observations made by Gasbarro et al. (2018). However, for the Sognefjord megafauna community, there was not much difference in diversity and species richness between percent cover of hard substrate, soft bottom or mixed substrates. Therefore, it is possible that other factors like environmental dynamics or food availability is driving the patterns observed.

The megafauna communities at the mouth of the fjord (D2) showed lower diversity and species richness compared to central fjord (D1) communities. The central fjord environment is more stable than that of the fjord mouth, which is subjected to greater temporal variability (Storesund et al., 2017) due to exchange with the coastal ocean. Differences between the two dives are likely to be largely a result of horizontal environmental gradients along the fjord set up by biogeochemical and physical processes. For example, dissolved oxygen concentrations at the interface between the basin and intermediate water and at the sill depth differed slightly between dives (Supplementary Figure S1), the water being more oxygenated toward the sill (D2) because of the influence of coastal water. Dissolved oxygen concentrations at these depths are likely diminished with distance up-fjord by diffusion to and entrainment of vertically adjacent, less oxygenated waters and by the cumulative effects of (bacterial) respiration with distance from the sill (Storesund et al., 2017).

Areas with high environmental disturbance are characterized by an increase in mobile species, decrease in sessile fauna, and overall lower diversity (Włodarska-Kowalczuk et al., 2005; Jones et al., 2006; Webb et al., 2009; Hughes et al., 2010; Włodarska-Kowalczuk et al., 2012), as was observed in D2 and regions above the sill depth. The higher turbidity observed in D2 may have impacted the fjord benthic communities. Sessile suspension-feeding invertebrates are at a risk of smothering in areas with high quantities of suspended material in the water column (Jones et al., 2006; Kutti et al., 2015; Meyer et al., 2015). Fauna that are not limited by such conditions can persist (Rygg, 1985; Włodarska-Kowalczuk et al., 2005, 2012; Gasbarro et al., 2018), sometimes in high abundances, which could contribute to the increased abundance of echinoderms and reduced occurrences of sponges and sessile holothurians in the shallower regions of the fjord. Buhl-Mortensen et al. (2020) also noted that the shallower and silled regions of the fjord have relatively strong currents, whereas, the bottom currents in the basin were weak. This supports the general picture of gradients in environmental variability and stress within the fjord.

The environmental conditions in Sognefjord are affected by anthropogenic influences, such as cruise ships, fish farms, hydroelectric stations, and pollution (Manzetti and Stenersen, 2010). There is limited information concerning how such influences impact the Sognefjord community, though fish stocks have seen a considerable reduction (see Manzetti and Stenersen, 2010) and the shellfish community showed increased diarrhetic shellfish poisoning toxins with increased distance into the fjord (Ramstad et al., 2001).

A study by Rygg (1985) found that fjords with high pollutant concentrations were characterized by opportunistic species. Additional organic input from anthropogenic sources like fish farming or nutrient runoff may lead to hypoxic conditions in the fjord basin (Levin et al., 2009; Johansen et al., 2018). Johansen et al. (2018) predicted that increased organic matter within Norwegian fjord basins will lead to a dominance of deposit feeders, while the presence of suspension feeders will decline. Similar to the findings of Rygg (1985), Johansen et al. (2018) also found a shift in the community structure toward opportunistic species as a result of oxygen depletion and increased temperatures. Coastal water temperatures have been rising (Aure, 2016), which has led to increased temperatures within fjord basins (Johansen et al., 2018), resulting in changes in stratification and reduced oxygen supply. Long-time series temperature and organic input data are not readily available for Sognefjord, but if its environmental conditions follow the trajectories of other Norwegian fjords it is possible to predict a similar shift toward more opportunistic species. The present study does not include any temporal replication and the impact and future implications of anthropogenic-derived environmental change on the system is largely unknown and requires more research.