Submarine canyons are complex geomorphological features that incise continental margins to form pathways between the shelf and deep sea. (Huvenne and Davies, 2014; Amaro et al., 2016). The movement of water masses, sediments and organic matter over varying temporal scales through the canyon generates environmental gradients of physico-chemical properties that occur both horizontally, i.e. along or across the canyon axis, and vertically (Obelcz et al., 2014; Fernandez-Arcaya et al., 2017; Hall et al., 2017; Ismail et al., 2018). As a result, environmental conditions can vary over short spatial scales, such that different branches within a single canyon, or even opposing walls of the same branch may have different seafloor characteristics, and experience different hydrodynamic and sedimentary regimes (McClain and Barry, 2010; Aslam et al., 2018; Bargain et al., 2018; Ismail et al., 2018; Pearman et al., 2020). The high spatial and temporal heterogeneity in environmental conditions often results in enhanced regional and local productivity, biodiversity, and faunal abundance (De Leo et al., 2010; Vetter et al., 2010; De Leo et al., 2014).

Submarine canyons are listed by the FAO (2009) as topographic features that may support vulnerable marine ecosystems (VMEs). Vertical walls situated within submarine canyons are features of high conservation value, providing natural areas of protection for VMEs under increasing anthropogenic pressure from deep-sea trawling (Huvenne et al., 2011; Johnson et al., 2013). Vertical walls support a range of faunal assemblages (which make up VMEs) that exhibit high diversity (Robert et al., 2015; Robert et al., 2017; Pearman et al., 2020). Examples are walls supporting dense aggregations of reef forming scleractinian corals, Lophelia pertusa (recently synonymised to Desmophyllum pertusum (Addamo et al., 2016)) (Huvenne et al., 2011; Brooke and Ross, 2014; Fabri et al., 2014) and Madrepora oculata (Fabri et al., 2014), the stony coral Desmophyllum dianthus, the octocorals Paragorgia arborea and Duva florida (Brooke et al., 2017), the deep-sea oyster, Neopycnodonte zibrowii (Van Rooij et al., 2010; Fabri et al., 2014), and the fire clam, Acesta excavata (Johnson et al., 2013). On the other hand, other sections of vertical walls can be devoid of life (Robert et al., 2015; Pearman et al., 2020). Consequently, vertical walls contribute to a canyon’s habitat diversity in various ways.

Desmophyllum pertusum reefs and coral gardens are listed as ‘threatened or declining’ under Annex V of the Oslo-Paris convention agreement (OSPAR, 2008), under Annex 1 of the Habitats Directive (92/43/EEC, 1992) and as VMEs (FAO, 2008), requiring protection. Effective spatial management and conservation of vertical wall assemblages requires a deeper understanding of the processes that generate the observed faunal distribution patterns (Huvenne and Davies, 2014). However, despite the likely importance of vertical walls in supporting and protecting diversity hotspots and protected habitats, few ecological studies of wall fauna have been conducted (Robert et al., 2017; Robert et al., 2020) and our understanding of the processes that generate spatial patterns along them is limited.

Our limited understanding is, in part, attributed to the challenge of sampling deep-sea vertical walls and measuring the local environmental characteristics. As a result, vertical walls stayed largely unsampled prior to recent advancements in remote technologies (e.g. Remotely Operated Vehicles (ROVs)) (Huvenne and Davies, 2014). Additionally, the limitations in the resolution of ship-borne bathymetry prevents accurate delineation of vertical walls (Huvenne et al., 2011; Robert et al., 2017). Consequently, despite their likely importance, vertical walls remain under-represented and under-sampled environments of canyons, limiting our knowledge of canyon ecology. This is further confounded by the predominance of canyon studies which only model the probability of epibenthic species presence-absence (Robert et al., 2015; Bargain et al., 2018; Lo Iacono et al., 2018) or univariate faunal responses that condense faunal information into a single diversity index (Robert et al., 2015; Ismail et al., 2018), rather than representing wider, multivariate species assemblage data.

In general, the responses of canyon fauna are regulated by a complex interplay of multiple factors acting at different scales. Environmental factors (water mass properties, seafloor characteristics and food supply) are most likely to explain species patterns at broader spatial scales (McClain and Barry, 2010; Robert et al., 2015; Ismail et al., 2018) while biotic processes (e.g. competition) more often act at finer spatial scales (Robert et al., 2020). Stochastic events (disturbance) act at multiple scales (Pierdomenico et al., 2016). The interaction of these processes across different spatial and temporal scales makes identifying key factors that drive faunal patterns within heterogeneous canyon landscapes challenging.

Canyons are recognised as sites of intensified hydrodynamics, including energetic internal waves and internal tides (Liu et al., 2010; Hall et al., 2017). Internal (baroclinic) waves occur when there is a perturbation to the interface between layers of the water column with different temperatures, salinities, and thus densities. The perturbation is restored by local buoyancy, forming oscillations (waves) that propagate along the interface. In a continuously stratified water column, such as the open ocean, the waves propagate vertically as well as horizontally. Internal waves generated by tidal motions, and thus oscillating at tidal frequencies (e.g. semidiurnal), are termed internal tides (Wunsch, 1975). In canyons, internal tides are generated when surface (barotropic) tidal currents flow across steep canyon topography (Allen and Durrieu De Madron, 2009; Vlasenko et al., 2016; Hall et al., 2017). Internal wave–topographic interactions (e.g. generation, reflection and breaking) are determined by multiple factors, including wave frequency, buoyancy frequency (i.e. stratification) and latitude. These determine the slope of internal wave propagation (s_wave) which can be compared to the local bathymetric slope (s_bathy) to predict wave behaviour. Canyon walls are typically steep compared to the slope of semidiurnal internal tides (α = s_bathy/s_wave >1), a state known as supercritical, so these internal waves approaching a wall are reflected back into deep water and towards the canyon floor. Conversely, the floors of canyons often have a gentler slope than semidiurnal internal tides (α<1), a state known as subcritical, so these internal waves approaching from offshore are reflected up the canyon towards its head. These processes combine to focus internal tide energy towards the canyon boundaries (its walls and floor), intensifying near-bed tidal currents (Hall and Carter, 2011; Hall et al., 2014). Where the local bathymetric slope is equal or near-equal to the internal wave slope (α ≃ 1), the wave is trapped near the boundary, often leading to breaking – similar to surface waves breaking on a beach – which increases the turbulent mixing of heat, salt, nutrients, and particulate matter between the layers of the water column.

Internal tides are increasingly advocated as key environmental factors influencing species patterns in the deep sea (Huvenne et al., 2011; Johnson et al., 2013; Van Haren et al., 2017; Davison et al., 2019; Pearman et al., 2020). For example, research focussing on scleractinian cold-water coral (CWC) assemblages has highlighted the importance of local hydrodynamics (including internal tides) in supplying nutrients and food to sustain CWC populations and preventing sedimentation on the hard substratum that the corals colonise (Frederiksen and Westerberg, 1992; Thiem et al., 2006; Davies et al., 2009; Mienis et al., 2009; White and Dorschel, 2010). Through interactions with sloping topography, internal tides occurring within canyons may enhance near-bed currents and turbulent mixing, forming efficient food supply mechanisms to benthic communities (Johnson et al., 2013). For example, the aggregation of organic matter by internal tide driven resuspension and mixing is postulated to play an important role in supporting high densities of M. oculata on the southern wall of Cap de Creus Canyon (Orejas et al., 2009).Internal tides also influence the resuspension and advection of suspended material in nepheloid layers (White et al., 2005; Liu et al., 2010; Puig et al., 2014; Wilson et al., 2015), where enhanced amounts of suspended matter (including particulate organic matter) are observed, representing an important food source for deep-sea fauna (Demopoulos et al., 2017). Internal tide modulation of nepheloid layers can result in replenishment of food to the benthos over the tidal cycle (Davies et al., 2009) and has been linked to spatial distributions of antipatharians and gorgonians in canyons of the Bay of Biscay (Van Den Beld et al., 2017).

The vertical displacement of the water column strata, associated with internal tides, also results in temporal variability of water mass properties (including temperature, salinity, dissolved oxygen concentration) along the canyon walls. Fauna respond to such changes in water mass properties and hydrodynamics (Levin et al., 2001; Howell et al., 2002; Dullo et al., 2008; Fabri et al., 2017). In addition, spatial and temporal hydrodynamic variability has been linked to species richness and assemblage patterns on the Hebrides Terrace Seamount (Henry et al., 2014). Hydrodynamic variability of internal tides generates environmental heterogeneity in near-bed sheer stress and nutrient and sediment fluxes (Frederiksen and Westerberg, 1992), which are proposed to influence CWC coral mound formation in the North-East Atlantic (White and Dorschel, 2010). On the other hand, modelling indicates internal tide hydrodynamic variability is an important factor influencing larval dispersal on the Rosemary Bank Seamount (Stashchuk and Vlasenko, 2021). However, to date no studies investigating faunal responses to internal tide induced environmental heterogeneity have been conducted in submarine canyons.

Here for the first time we investigate if spatial and temporal gradients in hydrodynamic properties, induced by the internal tide, can explain variation in spatial patterns of faunal diversity and assemblage composition on deep-sea canyon walls. We utilise biological, hydrodynamic and bathymetry-derived datasets in an integrated approach to undertake a high-resolution analysis of wall assemblages within Whittard Canyon, North-East Atlantic. We ask the following questions: (1) Does epibenthic megafaunal assemblage composition change across hydrodynamic and substratum gradients on vertical walls and (2) which environmental variables exert the strongest influence on epibenthic megafaunal diversity and assemblage structure?

A number of studies have examined environmental drivers of faunal patterns in submarine canyons (Robert et al., 2015; Sigler et al., 2015; Pierdomenico et al., 2016; Domke et al., 2017; Bianchelli and Danovaro, 2019; Pierdomenico et al., 2019; Pearman et al., 2020) but until now, no study of canyon wall assemblages integrating both spatial and temporal oceanographic variability induced by the internal tide has been conducted. Using a multi-disciplinary approach, we have been able to further quantify spatial patterns in environmental variables and wall faunal assemblages in canyon settings.

Canyons are highly heterogeneous environments and the influence of spatial patterns in the environmental variables, coupled with the sample design of locations at two very different depths, makes it difficult to pull apart the role of environment vs the role of location (as illustrated by the strong significance of the RAC in the GLM and the overlap in variance partitioning in the RDA). Still, the nMDS and RDA results illustrate that faunal assemblages are not simply determined by ROV dive or sampling location since several assemblage clusters were observed from multiple dives ( Figures 5 , 7 and Table 5 ). Furthermore, the RDA analyses identified depth, slope, substratum and proxies of internal tide dynamics as important factors driving faunal patterns on canyon walls ( Figure 7 and Table 6 ). The GLMs for species richness and 1/D also identified slope, depth and substrate characteristics as influencing faunal diversity, but not any of the proxies of internal tide dynamics ( Table 4 ). Diversity metrics condense multivariate information (faunal composition and/or abundance) into a single measure that is not representative of species composition. In our data, areas of high diversity (species richness and 1/D) supported different faunal compositions, demonstrating that the sole use of diversity metrics to represent faunal variability may miss key aspects of species – environment relationships in canyons and so limit our understanding of processes driving faunal distributions. As such these results indicate that other environmental factors (in this case proxies of internal tide dynamics), in addition to those traditionally highlighted by studies modelling diversity (i.e. depth, slope and substrate) (Robert et al., 2014; Ismail et al., 2018), are likely acting to determine assemblage composition in canyons. Additional sampling of other vertical walls in this part of Whittard Canyon would certainly help to obtain a clearer insight into the role of the environmental conditions in influencing faunal distributions.

Our spatial analysis revealed that the environmental variables investigated were spatially organised in relation to depth which was identified as an important factor influencing faunal patterns ( Figure 8 and Tables 4 , 6 ). Within Whittard Canyon we found several oceanographic gradients (temperature, salinity and dissolved oxygen) that were correlated with depth and varied in intensity along the canyon ( Figures 2 , 3 and Supplementary Figures 2, 3, 6 ). Broad- scale environmental gradients of physico-chemical properties act to determine faunal patterns (Levin et al., 2001; Kenchington et al., 2014; Mcclain and Lundsten, 2015; Robert et al., 2015; Du Preez et al., 2016; Ismail et al., 2018) and are likely driving the difference in assemblages observed between walls of the upper and mid canyon ( Figure 7 and Table 5 ). However, the observation from our study of different assemblages from similar depth ranges ( Figures 5 , 7 , and Table 5 ) suggests that other processes, such as internal tides and substratum are working in concert at smaller spatial scales to drive spatial patterns in species assemblages on canyon walls.

Canyons are sites of intensified hydrodynamics including internal tides, which our study has shown generate spatial and temporal heterogeneity in water properties (temperature, salinity, density and dissolved oxygen concentration) ( Figure 3 ) and near-bed current speeds ( Supplementary Figure 1 ). In Whittard Canyon, short term temporal variability induced by the vertical isopycnal displacement of the M ₂ internal tide (represented by the variable M ₂ amplitude) was found to be a significant factor driving faunal assemblages on canyon walls ( Figure 7 and Table 6 ). Short-term variability in temperature, salinity and oxygen also drove differences in faunal assemblages on the Hebrides Terrace Seamount (Henry et al., 2014). Short-term internal tide induced variability in water properties may influence faunal distributions via species physiological constraints that limit their bathymetric distributions (Hutchins, 1947; Rowe and Menzies, 1969; Tietjen, 1971; Menzies and George, 1972; Van Den Hoek, 1982; Jeffree and Jeffree, 1994; Southward et al., 1995). The comparatively low variance explained by the vertical isopycnal displacement of the M ₂ internal tide in our study may reflect the restricted environmental range that was sampled. For example, despite local amplitudes of up to 140 m calculated for the M ₂ tide resulting in maximum tidal temperature variations of 1.55°C, these areas of high temporal variability did not coincide with data collected from vertical walls. Consequently, the temporal variability in oceanographic variables experienced by wall fauna was relatively consistent between dives, even if the absolute values differed ( Table 5 ).

The M₂ vertical isopycnal elevation amplitudes diagnosed here may also reflect spatial and temporal variation in internal tide kinetic energy and associated turbulent mixing (Van Haren et al., 2022). In Whittard Canyon, peaks in turbulent kinetic energy dissipation have been linked with resuspension of material, nepheloid formation and sediment movement (Van Haren et al., 2022) that indirectly influences fauna by resuspending and concentrating POM (Dell’anno et al., 2013; Demopoulos et al., 2017; Pearman et al., 2020). Additionally, internal tide kinetic energy influences fauna directly by elevating near-bed current speeds and associated physical stress (Weinbauer and Velimirov, 1996; Johnson et al., 2013; Orejas et al., 2016) and variation in internal tide kinetic energy has been correlated with fish larvae richness, abundance and assemblage composition in the Midrift Archipelago Region of the Tiburon Basin, Gulf of California (Ruvalcaba-Aroche 2019). Future observational campaigns, including moored Acoustic Doppler Current Profilers (ADCPs) and microstructure profiler surveys, will allow full diagnosis on internal tide energetics and associated turbulent mixing rates to quantitatively assess their influence on canyon fauna.

Internal tide–topographic interactions (indicated by bathymetric slope criticality to the dominant semidiurnal internal tide) are also linked to spatial heterogeneity in turbulent mixing (Wilson et al., 2015; Hall et al., 2017; Van Haren et al., 2017; Aslam et al., 2018) and near-bed current speeds both of which are linked to resuspension of POM (Thomsen and Gust, 2000; Wilson et al., 2015; Hall et al., 2017; Aslam et al., 2018). Deep-sea fauna predominantly rely upon the consumption of surface derived POM and internal tides interacting with supercritical slopes have been shown to form efficient food supply mechanisms capable of delivering high quality POM from surface waters to benthic assemblages at depth (Johnson et al., 2013; Mohn et al., 2014). Conversely, internal tides interacting with critical slopes may result in wave breaking and resuspension, and the mobilisation of older material from the seafloor that is often degraded and reworked material of lower quality POM. In Whittard Canyon, slope criticality was found to be a significant factor driving faunal assemblages on canyon walls, and it was mainly linked to assemblages that correlated with coral reef substrata ( Figure 7 and Table 5 ). Dives 116 and 263, conducted at similar depths, differed in slope criticality and assemblages observed ( Figure 6 ). On dive 263, brachiopods, large sponges and arborescent gorgonians were observed on walls where the slope was near critical ( Table 5 and Figures 4 , 6 and 7 ). In contrast, on dive 116, an assemblage characterised by D. pertusum was observed at similar depth, from a wall that was supercritical ( Table 5 and Figures 4 , 6 and 7 ). Isotopic analysis shows that D. pertusum has a broad trophic niche (Mueller et al., 2014; Demopoulos et al., 2017) having been known to feed on POM, zooplankton (Duineveld et al., 2007; Duineveld et al., 2012), bacteria and dissolved organic matter (Mueller et al., 2014) with a preference for high quality POM. On the other hand, isotopic signatures indicative of lower quality POM have been documented from brachiopods (Valls, 2017). Could the different spatial distributions of these assemblages be driven by the trophic niches of the characterising taxa that are able to capitalize on heterogeneity in POM influenced by the internal tide interacting with complex topography? Variability in the quality and amount of food supply is known to influence canyon faunal distributions (De Leo et al., 2010; McClain and Barry, 2010; Cunha et al., 2011; Chauvet et al., 2018). Furthermore, hydrodynamic and geomorphological processes have previously been proposed as factors influencing the supply and resuspension of particulate organic carbon to canyon environments and thus driving trophic structure, faunal assemblage composition and diversity (Dell’anno et al., 2013; Demopoulos et al., 2017). However, to confirm the role of the internal tide in generating spatial heterogeneity in food availability to which fauna respond, further trophic analysis of nepheloid layers in proximity to faunal assemblages in relation to internal tide dynamics would be required.

Near-bed current speed is also an important factor influencing food supply (Thomsen and Gust, 2000). Although R.M.S near-bottom baroclinic, barotropic and total current speed was removed from statistical analysis (due to collinearity with depth), data exploration showed that assemblages were distributed along a gradient of baroclinic (internal) current speed ( Supplementary Figure 8 ). Separation of species along a gradient of current speed could reflect feeding and morphological adaptations. Species vary in their feeding strategies and efficiency under different hydrodynamic regimes (Järnegren and Altin, 2006; Van Oevelen et al., 2016). Species may exploit exposed areas to increase food encounter rates (Davies et al., 2009; Howell et al., 2011; Rengstorf et al., 2013; Mohn et al., 2014; Van Oevelen et al., 2016; Bargain et al., 2018), or conversely avoid areas with high current speeds that may exceed food capture rates, damage feeding apparatus (Johnson et al., 2013; Orejas et al., 2016) or topple large arborescent species (Weinbauera and Velimirov, 1996). Current speed is a primary driver of coral distributions (De Clippele et al., 2018) and in our study the D. pertusum assemblage occurred in an area exposed to the highest speeds, which is consistent with published observations (Davies et al., 2009; Rengstorf et al., 2013; Mohn et al., 2014) including those from vertical walls (Brooke and Ross, 2014). On the other hand, larger gorgonians and sponges were observed in areas exposed to lower current speeds. Intensified currents are also linked to resuspension and increased turbidity, which both brachiopods and D. pertusum are noted to tolerate (James et al., 1992; Brooke et al., 2009) and may enable them to exploit these conditions. Corals also benefit from strong currents that reduce sediment settlement on corals, which in turn reduces cost expenditure associated with cleaning polyps (Brooke et al., 2009; Larsson and Purser, 2011).

The ability to exploit substrata may also influence faunal distributions in canyons. Canyon walls not only vary in their slope criticality but in their geological formation and fine scale structural complexity provided by ledges and organisms themselves. CWC species including arborescent gorgonians and scleractinians are considered ecosystem engineers capable of forming complex structures, which act to promote increased species richness by providing substratum for settlement, refuge, modification of local sedimentation and subsequent food availability (Buhl-Mortensen et al., 2010; Guihen et al., 2013), resulting in increased fine-scale environmental heterogeneity (Buhl-Mortensen et al., 2010) and diversity of associated species (Frederiksen and Westerberg, 1992; Henry and Roberts, 2007; Lessard-Pilon et al., 2010). In our study, species richness was highest where coral substrata co-occurred with mud on vertical walls with ‘step-wise’ substrata, whereby accumulations of mud supported additional soft sediment species, further increasing diversity ( Figure 4E ). The overall negative relationship between species richness and coral reef framework modelled by the GLM as opposed to the positive relationship between coral reef framework and 1/D likely represents the increased evenness among species on vertical walls that are dominated by D. pertusum and A. excavata. Desmophylum pertusum reefs will promote species richness to a point after which D. pertusum dominates, so that fewer species occur but those that do are relatively evenly represented resulting in increased 1/D (Henry and Roberts, 2007). Small-scale geomorphological features, such as ledges also appear to influence faunal distributions on walls ( Figures 4A, E ). In our study, certain species were observed aggregating in association with ledges ( Figures 4A, E ). Similar observations have been made from other vertical wall environments where the increased fine-scale structural complexity provided by ledges is proposed to contribute to fine-scale environmental heterogeneity and so promote niche differentiation (Robert et al., 2020). The fragile nature of the ledges has also been proposed as a limiting factor on maximum colony size of corals observed (Brooke et al., 2017; Robert et al., 2020). This postulation could explain the occurrence of the D. pertusum assemblage on the wall with wider stronger ‘steps’, observed from dive 116 ( Figure 4E ) that are capable of supporting greater weight and higher coral densities, compared to the thinner ledges observed elsewhere ( Figure 4A ). However, the existence of different communities associated with ledges in ours (dive 262, 250 and 116) and other studies of Whittard Canyon (Johnson et al., 2013; Robert et al., 2020), suggests that these features act to influence species patterns at fine spatial scales whilst other factors beyond substratum availability (e.g. depth, food supply and internal tide dynamics) influence assemblage patterns across walls at the canyon scale.

The findings of our research can be applied to other settings where internal tides interact with complex topography to generate spatial-temporal gradients in environmental conditions (i.e. seamounts, coral mounds and ridges) (Frederiksen and Westerberg, 1992; White and Dorschel, 2010). The importance of internal tides in inducing temporal variability in water properties and/or influencing food availability to sustain benthic assemblages in otherwise hostile conditions (i.e. oxygen minimum zones) have been reported from steep shelf (Hanz et al., 2019) and seamount environments (Van Haren et al., 2017). Internal tides are also considered important phenomena influencing cold-water coral mound development (White and Dorschel, 2010). However, few studies have explicitly incorporated the influence of internal tides (Van Haren et al., 2017). In light of our findings research in other complex settings should endeavour to incorporate internal tide data into their analysis.

Our results show that faunal patterns on vertical walls in submarine canyons are driven by broad-scale environmental gradients that co-vary with depth, but also has highlighted the role of the internal tide in generating environmental heterogeneity at a finer scale (via vertical isopycnal displacement of the M ₂ internal tide and associated short term temporal variability in water mass properties, topography tide interactions and current speed) and how this might influence faunal distributions within the context of the larger depth related environmental gradients. As sites of intensified hydrodynamics, where internal tides generate spatial-temporal gradients in environmental variables, incorporating internal tide data is necessary to fully understand the processes that influence faunal patterns in canyons (including vertical walls).

We demonstrate that multivariate analysis of species data provides greater sensitivity than univariate indices, providing further insight into how the environmental factors interact at different scales to generate variability in environmental conditions that control species abundances and ultimately which species become characteristic of assemblages. Specifically, we highlight the likely link between internal tides and their associated vertical displacement in generating both spatial and temporal gradients in water mass properties that in turn influence faunal patterns on canyon walls.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Conceptualization of paper, TP and VH, Methodology: CTD data provision FM, image annotation TP and KR, oceanographic data processing and analysis TP and RH, statistical analysis TP, Writing- original draft preparation TP, writing, reviewing and editing TP, AC, KR, RH, FM and VH. All authors contributed to the article and approved the submitted version.