Southwest Approaches are incised by extensive submarine canyons including the King Arthur Canyon and multi-branched Whittard Canyon, which dominates this part of the margin. Whittard Canyon is a dendritic V-shaped canyon affecting over 150 km of the margin. This canyon displays a semi-circular shaped head with a concave profile (Cunningham et al., 2005). It is composed of four main branches (Western branch, Western middle branch, Eastern middle branch and Eastern branch), and is still active with mainly sediment transport from the head through turbidity currents (Amaro et al., 2016; Carter et al., 2018). Retrogressive erosion widens the canyon branches by wall failures (Amaro et al., 2016), and erosional processes such as slumping or density currents appear to mold the Whittard Canyon’s present-day bathymetry (Robert et al., 2017; Carter et al., 2018). The canyon acts as a preferential pathway for sediment to transit from the margin to the abyssal plain (Zaragosi et al., 2000). The extensive vertical flanks of the canyon form an important habitat for cold water coral, with assemblages different from those of flat sites (Robert et al., 2020). In addition, gullies and terraces of the middle eastern branch, imaged by sonar scan, appear swept clean of sediments and exhibit bedrock outcrops (Bourillet et al., 2006).

The Goban Spur is located south of the Porcupine Seabight and characterized by a steep (up to 40°) escarpment (beginning at −1,800 m) with an N120 direction leading to the Porcupine Abyssal Plain (<−5,000 m) (de Graciansky et al., 1985; Masson et al., 1989; Naylor et al., 2002; Dorschel et al., 2010). The Goban Spur was affected by the Variscan orogeny and subsequent rifting phases that overprinted the peri-Gondwanan basement (Kimbell et al., 2010; Yang et al., 2020). It is defined by complex structures affected by faults that appear geomorphologically as highs, lows, and escarpments.

The Porcupine Bank separates the Porcupine Seabight from the Rockall Trough and forms a continental block with steep lateral slopes (>7–4° north to up to > 20° south) extending for more than 800 km (Naylor and Shannon, 2005; Yang et al., 2020; Whiting et al., 2021). At present-day bathymetry, the Porcupine Bank is incised on its southwestern slope by multiple submarine canyons and gullies including the over >100-km Porcupine Bank Canyon formed by upslope–retrogressive slope (Elliott et al., 2006; Dorschel et al., 2010).

The Rockall Trough, a ~2,500-m deep basin, separates the shallow Rockall Bank and Fangorn High from the Porcupine Bank. The Rockall Bank slope is affected by along-slope and across-slope processes and features like erosional channels, iceberg scours, and slope failures (O’Reilly et al., 2001; Haughton et al., 2005; Shannon et al., 2007; Dorschel et al., 2010; Sacchetti et al., 2011). Cold water coral mounds have been described on the flanks of the eastern Rockall Bank associated with geologic and oceanographic control (Mienis et al., 2006; Shannon et al., 2007).

New seismic studies on the western Porcupine Bank show details of Atlantic rifting and the oceanic to continental crust transition, highlighting that inherited Caledonian and Variscan crustal structures that influence the geometries of the crustal domains (Whiting et al., 2021; Yang and Welford, 2021). It also shows north-westward increasing volcanism supporting the transition from magma-poor to magma-rich rifting along the Porcupine Atlantic margin, with exhumed mantle domains southwest of Porcupine Bank to Goban Spur (Whiting et al., 2021; Yang and Welford, 2021). More recent surveying in the Porcupine Bank (Strachan, 2021) and in Whittard Canyon (Cunningham et al., 2005; Carter et al., 2018) has indicated extensive areas of bedrock exposure. Bedrock outcrops are mainly observed in the head of the Porcupine Bank Canyon, on escarpment creating bedrock cliffs, along the steeply sloping center margin, or along moderate to steeply sloping seafloors (Strachan, 2021). In the Whittard Canyon, a side-scan sonar survey reveals areas of bedrock exposure located in the eastern branch (Cunningham et al., 2005).

Multibeam echosounder (MBES) data used in this study were provided by the Irish National Seabed Mapping Programme INFOMAR/INSS [INFOMAR, Geological Survey Ireland and Marine Institute (GOTECH, 2002)]. This work contains Irish Public Sector Data (Geological Survey Ireland and Marine Institute) licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license. The different concatenate bathymetries were obtained using the Kongsberg Simrad EM120, EM1002 multibeam echo sounder on board RSV Bligh, RSV Siren, and S/V Ocean Surveyor (GOTECH, 2002). The multibeam echo sounder data were processed using CARIS HIPS and SIPS multibeam processing software and the xyz files were gridded in Fledermaus and then the DTM was exported to ESRI asc. Export interpolated Surface as ArcGIS raster (GOTECH, 2002). The multibeam bathymetry and backscatter data were projected to UTM Zone 28N and gridded at 25-m resolution in Esri ArcGIS 10.8.1. The grid used as background bathymetry is from GEBCO Compilation Group (2022).

The benthic video assessment for bedrock occurrence was made using videos and datasets collected during SeaRover campaigns (Picton et al., 2021) using the Marine Institute’s Remotely Operated Vehicle (ROV) Holland 1 on board ILV Granuaile during SeaRover 2017 (O’Sullivan et al., 2017) and SeaRover 2018 (O’Sullivan et al., 2018), and onboard the RV Celtic Explorer during SeaRover 2019 (O’Sullivan et al., 2019). Data from videos from Autosub6000 AUV (autonomous underwater vehicle) onboard RRS James Cook during CODEMAP2015 cruise (Huvenne et al., 2016) were also used. Bedrock locations identified from past cruises CYMOR 2 (Pastouret et al., 1984), CYAPORC (Auffret et al., 1987) and GEOMANCHE 76/2 (Pautot, 1976) were also used ( Figure 1 ). These videos allow direct recognition of bedrock exposure in various regions, either morphologically or spatially, on the ICM ( Figure 1 ; Supplementary Figure 1 ). The videos were watched and the seabed substratum, where observed, was classified into classes: mud, pebbles, bedrock, and hardground ( Appendices A–C ). We were cautious about the apparent geometry and size to determine if the outcrop was in place and to limit the identification of large dropstones. This classification allowed us to indicate the presence or absence of bedrock/outcrops. The USBL from the ROV was retrieved and compared to time code to extract the location of bedrock outcrop transects.

To reduce potential oversampling due to differences in resolutions between MBES data and ROV observation points, the ROV video data were resampled using the Generate Points Along Lines tool in ESRI ArcMap 10.8.1 with 5-m intervals. Some transects were cleaned if the ROV was running in circles or remaining stationary for a long time in one position, resulting in point clustering and oversampling, to simplify the generation of clear ROV navigation line and points.

The study area has been divided into five zones to look more precisely at the bedrock suitability occurrence. These zones are the Southwest Approaches (including Whittard Canyon and King Arthur Canyon), the Goban Spur, the northern and southern Porcupine Bank, and the Rockall Bank (including Edoras High and Hatton Basin). Each zone depicts a large diversity of locations suitable for bedrock occurrence ( Figure 5 ). It can be noted that the raw BSI displays very high resolution over the overall study area with a precision of the order of decametric sedimentary structure ( Figure 6 ).

The Bedrock Suitability Index has improved bedrock mapping considering the scientific literature, by linking and aggregating multiple variables extracted from direct video observations into a single index. Accordingly, the BSI mapping allows, at a regional scale, us to indicate potential hard substratum occurrence along the ICM ( Figure 6 ). Here, according to the PCA test ( Figure 2 ), the parameters that seem to be descriptors of bedrock, for this model, in deep submarine settings are as follows: roughness, rugosity, slope, curvature, and BPI. The general high roughness and rugosity suggests that outcrops can be observed on the heterogeneous terrain. It seems to agree with previous mesoscale studies about hard substrate mapping using these parameters (Gratwicke and Speight, 2005; Wilson M. F. J. et al., 2007; Wilson S. K. et al., 2007; Purkis et al., 2008; Dunn and Halpin, 2009). The high BPI indicates that this terrain heterogeneity is often associated with topographic highs and lows and is also a good indicator for hard rocky terrain, conforming to studies using these parameters in species and habitat distribution modeling (Howell et al., 2016; Hu et al., 2020). Moreover, medium to high slope and curvature describing these outcrops also highlights the terrain heterogeneity favorable for flows and erosional processes that instigate the outcrops. The high reflectivity shown by the backscatter values also illustrates induration characteristics of potential bedrock occurrence.

The bedrock suitability index indicates areas sharing the same geophysical parameters as bedrock outcrops. Here, it also seems to emphasize features affected by erosion processes leading to outcropping. The pilot study site in the upper Porcupine Bank Canyon (Strachan, 2021) has shown good results for the applicability of the Bedrock Suitability index in highlighting bedrock occurrence on mainly cliffs, steep slopes, and rising from gentle slope outcrops. In this wider study, the main morphological features underlining potential bedrock, being classified with high bedrock suitability, are erosion zones associated with (1) downslope and along-slope transport associated with failure scarps, escarpments, canyons heads, talweg, or terraces; or (2) topographic features inherited from deep structural components of the margin such as positive and negative reliefs associated with fault walls.

In the eastern branch of the Whittard Canyon, retrogressive erosion linked to subsequent flows widens the canyons and leads to the instability of the canyon walls (Cunningham et al., 2005; Daly et al., 2018). Canyon development incising the Porcupine Bank has been associated with slope failure (O’Reilly, 2022). Along the Rockall Trough margin, bedrock exposed by multiple slope failure events linked to seismicity have been described (Sacchetti et al., 2011; O’Sullivan et al., 2020; O’Reilly, 2022). These instabilities and erosive processes may lead to an unveiling of the bedrock as the younger sedimentary and more unstable and unconsolidated succession covering is swept away.

In addition, the relict canyon talwegs shown here as potential bedrock location agree with Cunningham et al. (2005)’s observations on the relict canyons implying sub-crop or coarse material with along-slope processes transporting finer sediments to the eastern canyon branches and accumulating coarser material ( Figure 10 ). Moreover, previous studies showed, in the Southwest Approaches canyons, that turbidity currents, mainly in the upper part of the canyons, entail erosion and incisions of the canyon floors (Zaragosi et al., 2000; Cunningham et al., 2005; Bourillet et al., 2006). These active downslope processes infer active erosion within the upper canyons. Here, these erosion processes are highlighted by high suitability for bedrock to occur in the canyon’s head, either in Whittard Canyon or on the northern and southern Porcupine Bank canyons ( Figures 7 – 9 ).

Large-scale sediment slumps and block failure (Leynaud et al., 2009; Carter et al., 2018) appear to be direct erosion processes affecting the Whittard Canyon walls as well as the Rockall Bank and the northern Porcupine Bank (Faugeres et al., 1981; Elliott et al., 2006; Georgiopoulou et al., 2014). These processes are highlighted here by areas of high BSI overprinting on headwall scar locations and parts of canyon walls ( Figures 5A, B , 9 ).

On the Fangorn High ( Figure 12 ), the round-shaped features underlining potential outcrops might be related to mound constructions or volcanism (Barton and White, 1997; Stoker et al., 2012, CE21010 cruise). The topographic features highlighted by clusters of high suitability for bedrock occurrence in Goban Spur ( Figure 11 ) seem to be related to the deep fault network affecting the southern part of the ICM.

Underlying faults modeling the ICM play a crucial part in the general morphology of this deep offshore territory ( Figure 13 ). The Goban Spur and the Southwest Approaches, which are particularly affected by a deep fault network, display morphological features shaped by the fault scarps like escarpments associated with topographic high and lows. This connection between the deep structure of the margin and its geomorphology can also be observed on the bedrock probability occurrence map ( Figures 13A, B ).

On the northern part of Goban Spur, two conjugated sets of faults with 140° and 50° trend deeply affect the area ( Figure 13A ). The faults with 140° trend seem to mark escarpment slope toes. Indeed, the layout of the faults determines a limit between high BSI and very high BSI areas ( Figure 13B ). This can be observed on most fault scarps on the southern Goban Spur, creating a pattern of semi-horst and graben with preferential erosive and outcropping surfaces. The fault layout seems to correspond to the limit of high BSI clusters ( Figure 13B ). In Figure 13B , clusters of high suitability are located preferentially at the edge and toe of the slope of steep escarpments, as well as on the topographically disturbed seabed. This reinforces the confidence in BSI areas that overlay fault locations. Moreover, as hard surfaces are preferential substrates for important organisms such as cold-water corals, the erosion features and outcrops created by fault scarps create preferential environments to develop deep sea habitats.

Bedrock or rock boulders can provide excellent habitat for cold-water coral reefs or act as a hard substrate for them to grow over and encompass other species’ important assemblages (Gratwicke and Speight, 2005; Wilson SK et al., 2007; Purkis et al., 2008; Dunn and Halpin, 2009). Indeed, hard substratum seems to be a crucial parameter in species distribution and habitat mapping studies in warm provinces (Caddy, 2007; Casoli et al., 2019; Hu et al., 2020; Keenan et al., 2022), in Arctic to Subarctic regions (e.g., coastal northern Norway: Jonsson et al., 2004; Dunlop et al., 2020), in shallower settings like circa-littoral (e.g., Basque continental shelf: Galparsoro et al., 2015), or in deeper settings like continental slopes (Wilson M. F. J. et al., 2007). In particular, Wilson SK et al. (2007) and Dunn and Halpin (2009) showed with observations that cold-water coral reef habitat and other species assemblages were correlated with rugosity. Cliffs and vertical walls on complex geomorphologic features are favorable to cold-water reef development (Flach, 2003; Robert et al., 2017; Robert et al., 2020; Appah et al., 2021; O’Reilly, 2022).

Here, different areas seem to be favorable for rocky substrate outcropping, especially where along-slope processes are active (presumably bedrock is exposed through erosion at canyon heads, escarpments, etc.) and high BSI occurrence can be observed. Up to 20% of potential occurrences of bedrock can be concentrated in these areas, compared to 5% on the relatively flat seafloor across the margin ( Table 2 ). Eight percent can be found in canyon areas ( Table 2 ). It can be noted that 20% correspond to tectonically induced topographic relief (e.g., in the Goban Spur area, Table 2 , Figure 14 ), located in deep to very deep domains (<−2000 mbsl). It could coincide with deep habitat and rich species diversity, such as benthic communities, crustaceans, echinoderms, polychaetes, as observed on the Goban Spur area at <−2000 mbsl (Flach et al., 1998; Flach and de Bruin, 1999), and particularly sponges, cnidarians, and gorgonian observed on rocky substrates (Tyler and Zibrowius, 1992). Consequently, deep rocky areas located on the Goban Spur region could encompass a large variety of species and, thus, biodiversity hotspots could also be linked to the deep structure of the ICM.

Accordingly, per Table 2 and BSI mapping, potentially more than 5,000 km² (e.g., canyon area) could be favorable areas for habitat development and species like cold-water coral reefs on the ICM. As a result, bedrock mapping could identify suitable environments for reefs through morphometric properties and BSI establishment all along the ICM. Spatial distribution of hard substrate appears to be crucial to increase precision in species mapping and could help in habitat protection management.

The high potential for bedrock to occur in canyons, escarpments, or positive topographical features highlights both a preferential substratum for benthic species and locations with enhanced bottom currents. Future research could investigate whether BSI can be applied to map potential areas for benthic and sessile organisms that require both hard substratum and enhanced hydrodynamism.

This model could help identify probable bedrock locations at the regional scale (≥ 400,000 km²) up to 25-m resolution and cover huge areas with less precision but good indication. This model is GIS-based and time-efficient, requiring less computer resource usage despite the huge dataset (25-m-resolution grids on more than 400,000 km²).

Nevertheless, the model has its limitations. The intrinsic restrictions of the ROV limit the acquisition of videos in settings deeper than −3,000 m. In addition, video acquisition data on deep vertical and steep walls are scarce. This is reflected on the BSI as a huge part of vertical walls are illustrated with low BSI and seem underestimated as bedrock outcrop potential locations ( Figures 5 , 6 ), and because vertical walls can host a variety of habitats, there is no clear delineation between the bedrock occurrence of the vertical wall and sediment zonation. This could be resolved by a new video dataset or ground-truthing points, dedicated on vertical wall stratigraphy mapping. Likewise, there are constraints due to the quality of bathymetric and backscatter data with depth (more noise and reduced beam accuracy in progressively deeper water) ( Figures 5 , 6 ); imperfect integration of either position, orientation, or sound speed information can generate false seafloor roughness elements that overprint true geomorphology as demonstrated in Clarke (2018). In addition, there are some caveats when using uncalibrated backscatter values, especially considering different surveys and sources. These concerns were addressed using the Focal statistic tool to reduce the influence of noise in the dataset; standard deviation was multiplied with the original BSI map to give less weight to surfaces with high variability (displaying noisy bathymetry in the dataset) and more weight to low-variability areas (less noise in the dataset).

Morphologically significant rugosity may appear and disappear within a terrain, reflecting the changing imaging geometry rather than the real spatial distribution of natural morphology (Clarke, 2018). The general influence on the BSI leads to areas with more noise like the Feni drift/abyssal plain area, ≤3000-m-depth areas, and less noise in areas with shallower distinct morphological structures like the Whittard Canyon, Fangorn High, or north Porcupine Bank. Another approach would be to apply this model to other regional areas that were covered by large ROV campaigns or to lessen the bathymetric noise with further cleaning.