Cold‐water coral assemblages on vertical walls from the Northeast Atlantic

In this study, we assess patterns of cold‐water coral assemblages observed on deep‐sea vertical walls. Similar to their shallow‐water counterparts, vertical and overhanging walls in the deep sea can host highly diverse communities, but because of their geometry, these habitats are generally overlooked and remain poorly known. These vertical habitats are however of particular interest, because they can protect vulnerable coral ecosystems from trawling activities. As such, it is important to understand their ecology and assess their global importance.


| INTRODUC TI ON
In order to address conservation needs in the deep sea, we need to understand better the spatial distribution of ecologically important habitats, and this can be facilitated by identifying the factors that play significant roles in shaping biological spatial patterns. The relative importance of these variables are expected to change across habitats and scale, with environmental factors most likely to explain species coexistence patterns at broader spatial scales and biotic processes at finer scales (Tamme, Hiiesalu, Laanisto, Szava-Kovats, & Pärtel, 2010).
From a conservation perspective, identifying the drivers behind biodiversity patterns can help shape approaches to marine spatial planning (Economo, 2011). For example, when environmental controls are most important, maximizing representation might be favoured, while when biotic controls are most relevant, protection of specific features may be optimal. Along the same line, niche theory suggests that each species exploits its environment differently (i.e., niche differentiation) and communities arise from heterogeneity in environmental conditions and limited resources (Hutchinson, 1957). This hypothesis implies that more complex environments, which provide increased niche differentiation, may act as biodiversity hotspots of particular conservation value. The main issue in the deep sea is that most habitats are poorly studied, with rarer ones still being discovered.
In shallow waters, the slope of the seabed has long been identified as an important structuring component of benthic communities (Witman & Dayton, 2001). When comparing vertical and horizontal substrata, predation (by fish and urchins, for example), sunlight intensity, sedimentation rates and wave action may all be structuring factors (Miller & Etter, 2008;Sebens, 1986). This can often lead to horizontal substrata being dominated by macroalgae and vertical sites being colonized by epifaunal suspension feeders (Miller & Etter, 2011). In tropical coral reefs, corals are often more abundant on vertical surfaces, where competition with algae and sedimentation rates may be reduced (Birkeland, 1977;Rogers, Fitz, Gilnack, Beets, & Hardin, 1984;Sheppard, 1982). Rich and abundant communities of suspension feeders on vertical walls have also been reported for deeper waters (Haedrich & Gagnon, 1991), but it is only in recent years that technological advances, particularly the increasing use of remotely operated vehicles (ROV), have allowed for more detailed descriptions of such environments (Bell, Alt, & Jones, 2016;Huvenne et al., 2011;Johnson et al., 2013) As a result, large vertical reefs of overhanging scleractinians (Brooke & Ross, 2014;Fabri et al., 2014;Huvenne et al., 2011;Van den Beld et al., 2017) and walls inhabited by Alcyonacea (Brooke et al., 2017;Edinger et al., 2011;Quattrini et al., 2015) or dominated by other community types such as bivalves (Johnson et al., 2013;Ludvigsen, Sortland, Johnsen, & Singh, 2007) and sponges (Bell et al., 2016;Brooke et al., 2017;Genin, Paull, & Dillon, 1992) are being discovered. Some of these steep walls have been reported as harbouring the highest abundances of corals or bivalves in the area (Gasbarro, Wan, & Tunnicliffe, 2018;Johnson et al., 2013;Morris, Tyler, Masson, Huvenne, & Rogers, 2013) with high numbers of other associated species also observed (Robert, Jones, Tyler, Rooij, & Huvenne, 2015).
Cold-water corals are currently at risk from both environmental changes and direct anthropogenic impacts (e.g., trawl fisheries) (Freiwald, Helge Fosså, Grehan, Koslow, & Roberts, 2004) and are the subject of targeted conservation strategies (Huvenne, Bett, Masson, Bas, & Wheeler, 2016;Ross & Howell, 2013). The general association between cold-water corals and seafloor slope is well established and has successfully been employed to inform management Rengstorf, Yesson, Brown, & Grehan, 2013), but vertical walls are of particular interest as they can provide natural protection against bottom trawling. Larval dispersal from these refuges may also help recolonize damaged habitats.
However, despite their likely significance, the diversity, abundance and uniqueness of such habitats along the continental slope are not known as the lower resolution bathymetric maps generally available for the deep-sea underestimate slope, while sampling techniques such as trawls and towed cameras are not suited to vertical habitats.
In this paper, we examine the spatial patterns of coral assemblages observed on deep-sea vertical walls at local and regional scales in the Northeast Atlantic and examine implications for management and conservation. The objective is to identify whether cold-water coral assemblages on verticals walls harbour different species assemblages than corals on flat terrain. We use video analysis of ROV transects carried out at five sites in the Northeast Atlantic to investigate species assemblage, diversity and niche differentiation.

| ME THODS
We investigated composition of vertical cold-water coral communities through case studies in the Northeast Atlantic. These case studies included a landslide escarpment and four sites in two submarine canyons (Table 1, Figure 1). During the "Slope Collapses on Rockall Bank and Escarpment Habitats" (SORBEH) cruise in July 2014 (RV Celtic Explorer-14011), the ROV Holland I was employed to survey a submarine landslide headwall scarp (Georgiopoulou, Shannon, Sacchetti, Haughton, & Benetti, 2013;Figure 1a). As part of the "Complex Deep-sea Environments: Mapping habitat heterogeneity As Proxy for biodiversity" (CODEMAP) project (cruise RRS James Cook-125), two branches of Whittard Canyon, with walls on both sides, were surveyed using the ROV Isis (Figure 1b). Another wall was surveyed in the nearby Explorer Canyon (Figure 1b,c). In addition, two ROV dives completed during a previous cruise (JC-036) in Whittard Canyon were also included in this analysis. For each of these dives, sections of videos recorded with the ROVs moving vertically from the base to the top of the walls were separated into transects for analysis (18 in total). To assess differences in species assemblages, we included an additional 10 video transects (from CE-14011, JC-125, JC-036 and another older Whittard Canyon expedition, JC-010) where cold-water corals on flat ground were observed at comparable depths. Dive locations and geological settings are summarized in Table 1.
Remotely operated vehicle positioning was obtained using an ultra-short baseline system (USBL) with an accuracy of 1% of TA B L E 1 List of cruises, vehicles and general environmental characteristics for each dive. Location names for Whittard Canyon based on Amaro et al. (2016) Cruise ROV Overlapping frames were extracted and imported to form a "stack" in the freely available software ImageJ (National Institutes of Health, https ://imagej.nih.gov/ij/). Each organism larger than 20 mm was identified, marked and its pixel position recorded to avoid risks of double counting using the "Cell Counter" plugin. A single observer made the species identifications using imagery catalogues available for the area (Howell & Davies, 2010;Howell, Davies, & Beld, 2017;Jones et al., 2009) or used morphospecies (also known as operational taxonomic unit) when species-level identification could not be achieved (ophiuroids, hydroids and brachiopods were not included as reliable counts could not be obtained). Use of morphospecies enables the differentiation of taxa below the lowest taxonomic level to which an organism can be identified based on imagery alone, using features such as colour, growth form, branching pattern, ecological information (e.g., depth), etc. (Howell et al., 2019). As this approach complicates comparison between research groups or reuse of data, a reference image for each morphospecies (as well as the species matrix) is provided as Supporting Information. Differences in species composition were investigated using non-metric dimensional scaling carried out on a Bray-Curtis resemblance matrix computed on the Hellinger-transformed (Legendre & Gallagher, 2001) species matrix.
Alpha (i.e., within-sample) diversity was evaluated using rarefaction curves. To investigate the composition of associated species, reef-building coral colonies were not included in these analyses.
To further establish how vertical walls may provide different habitats, we investigated the spatial distribution of three Alcyonacea spe- and Video S1). Structure-from-Motion techniques allow for scaled and georeferenced 3D reconstructions to be achieved from a single camera moving around a scene (Ullman, 1979). The commercial software files and, for each, terrain descriptors were assigned by computing the mean of the ten nearest points. In addition, at each point, the presence of dead or live coral framework was assessed. As our aim was to determine whether there was a difference in environmental conditions between the localities where a species occurs from those generally available in the area (e.g., background), rather than to model species-environment relationships, we opted for an ordination technique. Ordination techniques allow for direct comparisons of environmental space and are less likely to overestimate niche overlap (Broennimann et al., 2012). To determine whether the investigated species' niche differed from the average conditions available in the background, the approach created by Dolédec, Chessel, and Gimaret-Carpentier (2000) was applied, using the "niche" function of the "ade4" R package (Dray & Dufour, 2007). We first carried out a principal component analysis (PCA) based on a matrix composed of the environmental conditions at all locations where the three species were present as well as at 10,000 randomly drawn background points (Broennimann et al., 2012;Di Cola et al., 2017). For each species, a centre of gravity was calculated in PCA space based on the rows representing each species' presences. From this centre of gravity, the square distance to the PCA centre, called the outlying mean index (also termed marginality), represents the level of difference between the average habitat conditions used by that species and the average conditions available in the background.
To examine niche overlap, the environmental space, as represented by the first two axes of the PCA, was divided into a 100 × 100 grid and a kernel density function was applied to obtain the smoothed density of occurrence of the chosen species in each grid cell (Broennimann et al., 2012). Niche overlap between species x and y was calculated using the metric D as presented in Warren, Glor, and Turelli (2008), where p is the occupancy as obtained from the kernel density function for each cell i. D varies from 0, when there is no overlap, to 1, when there is complete overlap. Permutation tests (using 999 permutations) were employed to assess test significance.

| RE SULTS
Based on the video transects, a total of 38,720 individuals/colonies from 112 morphospecies (of which 26 were cold-water coral morphospecies) were encountered on the walls of the Rockall Bank Escarpment, Whittard and Explorer Canyons (Tables S1 and S2), with cold-water corals (Scleractinia, Alcyonacea and Antipatharia) forming the overall largest component ( Figure S1, Figure 3). However, sponges were more prevalent on the walls of the Rockall Escarpment while bivalves (the deep-sea oyster Neopycnodonte zibrowii and the limid clam Acesta sp.) were much more numerous on Whittard Canyon's western middle branch walls. The Rockall Escarpment showed the highest species richness, followed by the southwest The sites assessed showed slightly higher richness at flat locations, but differences in species assemblages occurred between most sites as well as between flat and vertical sites ( Figure 4)
The yellow and pink morphospecies of Alcyonacea were frequently found attached to both dead and living (though likely not attached to live tissue) coral framework of D. pertusum, associated with higher rugosity values, while the red morphospecies tended to be observed on ledges where sediment had accumulated.

| D ISCUSS I ON
Within the Northeast Atlantic region examined here, species assemblages differed between most of the walls investigated as well as between vertical and horizontal habitats, although many species co-occurred. Differences in habitat use between certain species were demonstrated at the scale of a single wall, with the fine-scale structural complexity provided by vertical habitats likely providing additional niche space exploited by certain species.

| Drivers of cold-water coral assemblages
Many studies have found a range of environmental variables to be useful in explaining cold-water coral taxa spatial patterns (Table 2), with clear trends apparent in the scale at which particular environmental variables are significant. However, this may also reflect the lack of information regarding the spatial variability of certain predictors at specific scales. For example, variables such as substrate type become significant at finer resolutions, but are not often available for global assessment. Similarly, other variables, such as current speed or productivity levels, are available only at broader scales, even if finer-scale variations are also likely to be important.
Our results suggested that the differences in species assemblages observed between sites were, at least in part, depth related. Increases in depth correlated with decreases in measured temperatures, with the wall on the Rockall Escarpment being the coldest (4.7°C), followed by Whittard Canyon's western branch (5.6°C), eastern branch (6.9°C), middle western branch (10.8°C) and Explorer Canyon (11.7°C). Although some cold-water coral species co-occurred across walls, one species was usually dominant.
For example, the deepest and coldest wall, Rockall Escarpment, was dominated by S. variabilis and various taxa of Antipatharians.

while Isididae and
Antipathidae also have deeper mean depths than other deep-sea coral families (Etnoyer & Morgan, 2005). In the eastern branch of Whittard Canyon, isopycnal displacement caused by internal tides could lead to daily changes of up to 1°C in temperature (Hall, Aslam, & Huvenne, 2017), and the wider temperature tolerance window or stronger physiological capacity for adjustment to temperature fluctuations of D. pertusum when compared to M. oculata (Naumann, Orejas, & Ferrier-Pagès, 2014) could be another reason for observed differences in abundance across branches.
Comparing walls of similar depth on opposite sides of Whittard Canyon's western middle branch, we found a very high similarity in species composition and diversity. However, in the case of the eastern branch, differences were clearly apparent and were almost as large as differences with the transects on flat terrain.
Differences in the geology of these two walls could in part explain this pattern. Walls in Whittard Canyon were generally composed of friable, less competent sedimentary units of varying thickness (Carter et al., 2018;Robert et al., 2017), but one notable exception is the southwest facing wall in Whittard Canyon's eastern branch.
This wall was composed of two lithologies, including a harder rock that seemed resistant to erosion. More competent, resistant rocks were also found along the Rockall Escarpment, and both walls appeared more hospitable to other colonizing organisms (mainly brachiopods, sessile holothurians and sponges), which led to higher diversity. Soft sediments dominated the surroundings of most coral patches at flat sites, and the inclusion of soft sediment associated fauna, occurring between coral patches, increased diversity.
Despite differences in species composition between horizontal and vertical sites, flat terrain transects were more similar to their same-site similar-depth vertical counterparts than to each other.
Within sites, the higher variability observed at flat sites is likely linked to the greater separation in transect locations. It may be that high dispersal rates increase the number of shared species, while environmental conditions play a role in regulating their relative abundance.
Differences in species dominance could also arise as a result of the first species to colonize, with the established species outcompeting the others for space, leading to possible alternate states (Sutherland, 1974). In shallower coral reefs, competition with faster growing organisms, such as barnacles, tunicates and bryozoans, can reduce coral recruitment (Birkeland, 1977). However, once established, large colonial organisms reduce exposed substrate and overgrow adjacent individuals, potentially limiting the occurrence of certain species (Jackson, 1977;Sebens, 1986), while the additional structural complexity may favour colonization by other taxa. Although the lower competition rates occurring in deeper waters may diminish the importance of such mechanisms, this could potentially explain the lower number of species observed on the northeast facing wall of the eastern branch, where D. pertusum completely covered the wall in certain areas. On the other hand, for species occurring on coral framework away from the wall (such as the yellow and pink morphospecies of Alcyonanceans), this apparently small change in location likely influenced the hydrodynamic regime encountered and the ability of an individual to capture food (Gori, Reynaud, Orejas, & Ferrier-Pagès, 2015;Orejas et al., 2016;Purser, Orejas, Moje, & Thomsen, 2014). One could also expect that on highly friable rock, coral colonies may be size/weight limited and less able to overgrow and limit the space available for colonization by other organisms, possibly explaining bivalve dominance in the western middle branch. On the northwest facing wall of the eastern branch, bed-scale variations in rock strength and friability also led to the formation of ledges through preferential erosion where sediment accumulated, contributing to fine-scale heterogeneity and niche separation.
Resource availability may further play a role in determining whether colonies of smaller fast-growing organisms establish first and outcompete the larvae of slower growing species or whether these slower growing organisms can develop enough to eventually become dominant (Birkeland, 1977;Kneitel & Chase, 2004;Lavorel & Garnier, 2002). For example, owing to greater filtering capacities, bivalves may be better able to handle less regular food supply than corals (Johnson et al., 2013). Modelling studies of internal tides within Whittard Canyon have found high energy levels, with some particularly high near-bottom velocities in the eastern and western branches (Amaro et al., 2016;Aslam, Hall, & Dye, 2018).
Depending on the orientation of individual walls to the oncoming current, the hydrodynamic regime created may trap food particles and increase food availability in certain areas while nearby areas in the lee side of the wall experience a different regime. Nepheloid layers have been recorded in proximity to the investigated walls in Whittard Canyon (Huvenne et al., 2011;Johnson et al., 2013;Wilson, Raine, Raine, Mohn, & White, 2015). Such benthic nepheloid layers were particularly rich in fresh and labile organic matter, but showed differences in content across branches (Huvenne et al., 2011;Wilson, 2016).

| Conservation Importance
Based on currently available global bathymetric maps, estimating the area potentially represented by vertical walls is very difficult. For example, based on the 15 arc second resolution of the SRTM15_PLUS global satellite bathymetry grid (Olson, Becker, & Sandwell, 2016), the highest slope value calculated for Whittard Canyon, an area known to harbour vertical walls, is 53° ( Figure 6).
As such, if areas with slopes >20° are taken as representing areas of potential very steep topography in specific deep-sea features with complex topographies such as canyon systems, seamounts, ridges and escarpments as compiled in the global seafloor geomorphic features catalogue by Harris, Macmillan-Lawler, Rupp, and Baker (2014) (Figure 1a), this could add up to 421,000 km 2 (with an additional 682,000 km 2 on escarpment, which can overlap with other features). Of course, not all walls would occur in broadly suitable cold-water coral habitats, but 6,000 km 2 (with an additional 12,000 km 2 on escarpment) of these potential walls occurred in regions predicted to be suitable for cold-water corals based on the modelling of Davies and Guinotte (2011) (Mayer et al., 2018). To put these estimates into perspective, a world-wide assessment of shallow tropical coral reefs, based on 500 m resolution data, estimated that they covered 212,340 km 2 (Burke, Reytar, Spalding, & Perry, 2011).
A similar assessment is not available for cold-water coral reefs, but our estimates indicate a surface area equivalent to 8% of the estimated surface area of shallow coral reefs is covered by currently overlooked vertical cold-water coral habitat. Although only 508 of the geomorphic features have at least one OBIS (Ocean Biogeographic Information System) record for any of the five previously listed species (Figure 7), this small number illustrates how little we know of the global spatial distribution of cold-water coral species, with many areas of the world remaining greatly under-sampled and poorly mapped. In particular, coral samples from vertical walls would be especially underrepresented in the OBIS records because of the associated sampling difficulties. Most records available would result from traditional sampling techniques (e.g., trawls and dredges), which would have purposefully avoided high profile structures. Collections on vertical walls would require ROV or manned-submersible, and even so, sampling while hovering remains problematic for most vehicles. Recent developments in the use of underwater hyperspectral imagery may help counteract such difficulties by reducing the need for samples for species identification and improving automated quantification of live coral cover for monitoring .
Until very recently (see Davies et al. (2017)), hierarchical classification schemes aimed at informing ecosystem-based TA B L E 2 Environmental variables found to be important drivers of cold-water coral spatial patterns from species distribution model studies habitats. Yet, such walls likely play an important role in supplying larvae to surrounding areas (Smith & Witman, 1999), especially considering that D. pertusum larvae could possibly survive for more than three weeks (Brooke & Ross, 2014;Larsson et al., 2014). In addition, the rough topography can also protect these vulnerable cold-water coral ecosystems from threats such as commercial bottom trawling. However, such activities also lead to other effects, such as increased suspended particulate matter and change in organic content (Puig et al., 2012;Wilson, Kiriakoulakis, et al., 2015). Although many CWC are tolerant of moderate to high sedimentation rates (Brooke, Holmes, & Young, 2009), it has been suggested that reduced sediment accumulation on vertical walls could be beneficial for feeding ability and larvae survival (Brooke & Ross, 2014;Brooke et al., 2017), and anthropogenic changes to such processes could have unknown effects on vertical wall assemblages. As the temporal variability of most environmental factors, and their influence on spatial patterns, remains mostly unknown, cold-water corals on rocky walls will continue to be vulnerable to the current changes in ocean conditions, while acute disturbances are likely to arise following increasing anthropogenic activities in the deep sea.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated during the production of the current study are available from the corresponding author upon reasonable request.