On an ROV dive on 16^th April 2012, at 48.761°N, 10.461°W (Figure 1) we encountered a vertical wall extending from 631 m depth to 780 m depth. On this wall we identified the presence of large numbers of Acesta excavata on the vertical faces.

Acesta excavata and Neopycnodonte zibrowii were found together, mostly between depths of 633 m and 762 m, with particularly high densities in the vicinity of small overhangs (Figure 2). Nine taxa apart from the bivalves were identifiable in more than 10% of photographs. The two most abundant species were the scleractinian corals Desmophyllum dianthus and Madrepora oculata (Table 1). Comatulid feather stars were also relatively abundant, although the resolution of the photographs was sufficient neither to identify the feather stars to species level nor to conclude that they comprised a single species. In order of decreasing abundance was a species of small pink fish of the family Bythitidae living in the cryptic habitat created by the bivalves and corals, a pink tube anemone of the family Cerianthidae, a trochid gastropod Margarites sp., a portunid crab Bathynectes longispina, and the pencil urchin Cidaris cidaris. A species of Epizoanthus was also very abundant, being present on 24% of oyster shells.

Other species were present in less than 10% of the photographs. Some of these species were large and conspicuous despite occuring infrequently, for example, the bright orange flytrap anemone Actinoscypha saginata (Figure 3, bottom right), the carrier crab Paramola cuvieri, Lepidion eques, a characteristic fish species of deep waters in the region [18], and a conger eel, Conger conger, which was encountered at a depth of 681 m on the wall. Other crustaceans present in the photographs were squat lobsters Munidae sp., unidentified caridean shrimps and the euphausiid Meganyctiphanes norvegica. Apart from the crinoids and pencil urchin mentioned above, echinoderms in the photographs included four species of starfish. These comprised two poraniids, one almost certainly Poranius pulvillus, the other unidentified, a species of Ceramaster (Figure 3, upper left), and a stichastrellid, probably Stichastrella rosea. A second urchin species Echinus sp. (Figure 3, upper right) was also present. Other cnidarian species identified were the antipatharian coral Stichopathes sp. of which just two specimens were seen, a second species of zoanthid, Parazoanthus anguicomis, which was identified on a very small number of oyster shells, and the athecate hydrozoan Tubularia indivisa, which was more prevalent in shallower depths. Polychaetes are likely abundant in this cryptic habitat but are rarely visible in photographs. Two photographs capture Sabellidae sp. with their feeding tentacles displayed. This habitat is highly suitable for sabellids but it is likely that the turbulence caused by the ROV thrusters causes them to withdraw their characteristic tentacles making them hard to locate on photographs. Bryozoa were encountered relatively frequently but they were usually small and it was mostly not possible to identify them nor to elucidate the number of species present. However, a reticulate species of Reteporella, possibly R. incognita, was present, as was a cyclostome species that was probably Tervia irregularis. Sheet encrusting cheilostomes were occasionally present on oyster shells but could not be identified further. Distinctive sponges included the hadromerid Weberella bursa, the poecilosclerid Mycale lingua and an unidentified species of Tetractinellida. Two further sponge species were also tentatively identified from the upper part of the wall but the difficulty of sampling in overhang areas means that no voucher specimens were collected for confirmation. Nevertheless the blue sponge (Figure 3, upper left) could be Hymedesmia curvichela sensu [19] (not Hymedesmia curvichela Lundbeck, 1910) which we encountered (and collected) on a previous cruise to a canyon system north of the Porcupine Seabight. The bright yellow crust (also Figure 3, upper left) is likely Hexadella dedritifera Topsent, 1913, although this has recently been shown to be a species complex [20].

Size and density measurements in photographs were related to depth (Figure 4). The average height of A. excavata shells tended to increase with depth (F_1,20 = 5.41, p<0.05, r² 21%). The overall size distribution of measured shells was generally symmetrical around the mean of 10.5 cm (SE 0.37), with no obvious additional size classes. A linear regression of A. excavata density with depth was not significant; however, the highest densities in 20 m depth bands seemed to increase with depth. This observation was supported by quantile regression: coefficients for the 75% quantile were significant, indicating that upper density limits for A. excavata increased with depth. The maximum A. excavata density recorded was 25.2 m⁻². Neopycnodonte zibrowii also occurred at higher densities with increasing depth (F_1,20 = 4.6, p<0.05, r² 19%). The maximum oyster density observed was 16.1 m⁻². The patterns of increasing density with depth ended abruptly, with the deepest individuals found at low densities (values not directly estimated due to reflections of the laser sights off the substrate).

Correlations between species in the assemblage suggest common responses to environmental gradients and/or a biological interaction. Acesta excavata and N. zibrowii densities were positively correlated (r = 0.662, p<0.01). Of the nine identifiable taxa frequently observed in photographs, six were positively correlated with A. excavata and/or oyster densities: Desmophyllum dianthus, Cerianthidae sp, Comatulida sp(p)., Bythitidae sp., Bathynectes longispina and Margarites sp. (Table 1). The densities of two conspicuous species, the coral Madrepora oculata and the urchin C. cidaris, had no associations with A. excavata or N. zibrowii. Because of the nature of zoanthid colonies and the difficulties in identifying discrete colonies, it was not possible to test whether these were positively correlated with bivalve densities.

Other aspects of the habitat varied with depth. Most strikingly, the deepest part of the cliff had few A. excavata and the rest of the rock surface was relatively bare. This contrasts with shallower areas, where A. excavata was frequent, as were mobile species and colonies of corals, bryozoans, hydroids and sponges (Figure 3).

The water masses at the Whittard Canyon region are dominated by Eastern North Atlantic Water (ENAW) in the upper 500 m (Figure 5A) with ranges of 8< T <18°C, 35.2< S <36.7, and density, σ_t = 27–27.2 kg m⁻³. Below this, lies Mediterranean Outflow Water (MOW, 2.6< T <11°C, 35< S <36.2, with a core centred at σ_t = 27.5 kg m⁻³), found principally off slope, although the influence of MOW can be seen extending into the canyon branches. Labrador Sea Water (LSW) is found at intermediate depths with a core between 1900–2000 m. Vertical profiles of temperature, salinity and density, σ_t (Figure 5B–D) indicated a weakly stratified surface layer down to 400 m; the spring timing of the survey had not permitted full seasonal stratification to be developed. Between 600–1000 m evidence of the permanent thermocline can be seen, with strong stratification of the water column in both profiles. The ROV dive site (water depth 750 m), and the depths of the wall where high densities of suspension feeders were found (640–740 m), are located within this permanent thermocline.

Vertical profiles of beam attenuation (m⁻¹), used as a proxy for suspended material concentration, revealed a strong benthic nepheloid layer (BNL) present and numerous intermediate nepheloid layers (INL) within the canyon (Figure 6). Overall beam attenuation (m⁻¹) was highest in BNLs at 750 and 1000 m water depths. At 1000 m water depth, the BNL extended >100 m above the seabed, possibly associated with impinging MOW core found at this depth. A 200 m thick INL centred at 1100 m in the adjacent, deeper, vertical profile (water depth 1320 m) was likely associated with the detachment of the BNL. At the ROV dive site, the vertical profile of beam attenuation (m⁻¹) indicated the highest values of suspended material extending from the seabed at 750 m to ∼500 m depth. INLs observed between 650–750 m water depth in the 1000 m and 1320 m profiles were also likely to be associated with the BNL found at the ROV site.

The rough topographic nature of canyons is likely to result in dynamics of a complex nature where energy will be extracted from barotropic tides to baroclinic internal wave motion. A 6–7 hour repeat cross-canyon channel CTD transect ∼3 km downcanyon from the dive site suggested the presence of internal waves in the depth range 400–700 m (Figure 7). Internal waves generated at the barotropic semi-diurnal tidal period will propagate as a beam through the water column, periodically stretching and squashing the isopycnal surfaces at the depths where the internal wave energy is concentrated. This is highlighted in Figure 7 which shows the displacement of isopycnals based on the repeat CTD profiles at six locations across the canyon. At this time the upper 200 m of the water column was well mixed due to a severe storm a few days previously and is not shown. Maximum isopyncal displacement occurred close to the seabed at the dive site side of the canyon at ∼500 m and a secondary maximum was found at about 300 m on the opposite (northern) side. Relatively high isopycnal excursion occurred as a layer between these two maxima, as well as in a layer between 200–300 m which one might tentatively suggest emanated from the northern maximum. The band of high isopycnal excursion across the canyon between 300–500 m represented an angle (β) of ∼3 degrees from the horizontal. The buoyancy frequency (N), determined from vertical density profiles, together with an internal wave frequency (σ) appropriate for the semi-diurnal tide and Coriolis parameter (f), would suggest that internal waves would indeed propagate in beams 3 degrees from the horizontal (e.g. sin(β) = [(σ²−f²)/(N²−σ²) ]^1/2, [21]). This perhaps suggests that internal waves may have been generated at the top of the north canyon wall, with subsequent cross canyon propagation to a location immediately above ROV dive site wall. In addition there is high displacement at the foot of the canyon which may be due to tidal modulation of any bottom density flow or of any MOW water located between 800–1000 m. This would cause a change in density from horizontal advection rather than vertical displacements.

The resolution of the CTD transect was such that repeat profiles were made at water depths of 500 and 1000 m respectively, spanning the depth range of the vertical wall. The high calculated displacements above 300 m may also be the result of an internal wave beam originating from the same source, propagating up and reflecting from the surface mixed layer, where there are high displacements immediately below the bottom of the surface mixed layer. Again the propagation angles are consistent although the interpretation is somewhat speculative. Overall the repeat CTD transect suggests that a significant amount of baroclinic energy exists within the upper/mid canyon region. A BNL was generated at the depth of high displacement on the southern canyon wall (500 m) and an INL at that depth was present within the canyon (Figure 6). It is likely therefore that sediment mobilisation at depths above the wall and its associated fauna is a persistent process within the canyon.

Observations were made using the deep-water ROV Holland I during a cruise on the RV Celtic Explorer to the Whittard Canyon system (cruise CE12006, Figure 1). ROV Holland I is a Quasar work class ROV rated to 3000 m. It is equipped with several video camera systems, an OE14-208 digital stills camera and has two robotic arms, a slurp sampler and storage boxes for collecting fauna. Material retained in the slurp chamber was sieved with a 0.5 mm sieve and used to verify identifications from photographs. ROV depth and position were established via a Sonardyne Ranger USBL beacon system. Cruise CE12006 was targeting vertical walls in canyon systems as potential sites of high biomass. Potential wall areas were identified from the available INFOMAR bathymetry by targeting regions with high slope. The INFOMAR (Integrated Mapping for the Sustainable Development of Ireland’s Marine Resource) project is a joint venture between the Geological Survey of Ireland and the Marine Institute and provides high quality bathymetry to 25 m resolution, interpolated from 100 m spaced point data (http://www.infomar.ie/data/). Suitable sites were dived with the ROV. On 16^th April 2012, a dive at 48.761°N, 10.461°W (cruise CE12006, Event number 12) encountered a vertical wall extending from 631 m depth to 780 m depth. On this wall we identified the presence of large numbers of A. excavata on the vertical faces. On reaching the top of the wall, the ROV was flown to near the bottom of the wall to repeat a vertical pass while taking still photographs. Poor weather prevented any further dives on or near this site during the cruise.

Individual, non-overlapping, photographs were treated as individual quadrats to collect information on size distributions and densities of A. excavata. Size and area estimates were made for each photograph using paired laser guidelines orientated at 90° to the camera’s focal plane and separated by 10 cm. Measurements, calibrated to that photograph’s laser guideline separation, were made in imageJ, an open source Java-based image processing progamme (http://rsb.info.nih.gov/ij/). Counts of A. excavata and N. zibrowii were converted to densities m⁻². The attachment point of A. excavata shells to the substratum was taken as a basis for shell height measurements (the species has a straight dorsal margin). To minimize error, these measurements were only taken when the attachment point was clear and the shells were seen with a view of one valve or were seen in side view. Neopycnodonte zibrowii shells were similar in size to A. excavata, but the uneven nature of the oyster shells and the tendency for the edges of shells to overlap meant that shell sizes could not be confidently estimated. Quantile regressions were carried out using the quantreg package in R [29]. Overlapping photographs were mosaiced using check points added manually in the Hugin package (http://hugin.sourceforge.net/) to provide larger images of the assemblage. Organisms associated with the habitat were identified, where possible, from the images.

Sixty-six CTD casts were performed in two branches of the Whittard Canyon providing detailed hydrographic data. Vertical profiles of temperature, salinity, fluorescence and transmission were taken down to depths of 3150 m. Beam attenuation measurements were made using a 0.25 m path-length transmissometer (C-Star, WET Labs’) operating at 650 nm. Data were processed using Seabird data-processing software and MATLAB (Matworks, R2007a). Beam attenuation was calculated usingwhere x is the pathlength of the transmissometer and Tr is the transmittance output fromthe instrument expressed as

Voltage_(signal) is the output signal, Voltage_{(dark value)} is the dark offset for the instrument obtained by blocking the light path and Voltage_{(clean water calibration)} is the manufacturer’s supplied value for output in clean water.

Isopycnal displacements were calculated from a 6–7 hour repeat survey across the canyon close to the ROV dive site. To achieve this, the individual profiles were averaged over 20 m vertical bins to obtain individual vertical profiles of density and the density difference between the profile pair (ρ’). To calculate the isopycnal displacement (Z) at any depth, each pair of repeat profiles was averaged to form a mean vertical density profile and ρ’ was divided by the vertical density gradient of a range +/−40 m about each depth (dρ/dz),