The “North Atlantic Stepping Stones: New England and Corner Rise Seamount Expedition” (EX-21-04) took place aboard the NOAA Ship Okeanos Explorer from 30 June to 29 July 2021. Twenty dives were conducted by the Deep Discoverer (D2) remotely operated vehicle (ROV), with nine dives made on eight CRS seamounts (Figure 1; Table 1). The dives spanned depths from 940 to 4,189 m, the ROV being deployed to the deepest point of the dive and ascended along non-linear transects. Detailed dive summaries for sites protected by NAFO have been provided by Waller et al. (2021).

The ROV was equipped with three high-definition Insite Pacific video cameras, including the primary Zeus Plus camera (tilt, 18× optical zoom), a Canon R3 still camera with an RF 24 mm macro lens for high-resolution imagery of organisms and habitat details, plus a dual-laser scaling system projecting two parallel beams 10 cm apart for size estimation. Lighting was provided by standard ROV-mounted LED arrays. Video footage was recorded continuously. D2 was also equipped with a SeaBird SBE-911 Plus CTD, with sensors recording salinity, temperature, pressure and dissolved oxygen, which generated data at a high sampling rate (24 Hz) (Cantwell et al, 2021). ROV positions were logged continuously during each dive using the onboard USBL system integrated with DVL and inertial navigation. The effective spatial accuracy of logged track points along the transects was on the order of 2–5 m, depending on ROV speed and maneuvering. ROV track distances were calculated as cumulative three-dimensional distances between smoothed spline-interpolated horizontal positions, plus vertical depth differences. Only continuous on-bottom transects were included in subsequent analysis, except during two dives (MacGregor Seamount, Dive 8; Yakutat Seamount Deep, Dive 10), technical issues required brief ROV recovery from the bottom and redeployment. Transect lengths given in Table 1 account for those interruptions.

Subsequent to the expedition, all video records from the nine dives were examined and annotated following methods presented by Beckmann et al. (2025). In brief: Visually different types of coral and sponge specimens or colonies were assigned unique morphotype codes and their numbers in each frame (to a maximum of 50). Encrusting taxa, small taxa of less than 50 mm size and specimens insufficiently resolved in the imagery were excluded. Annotations were performed using Ocean Networks Canada’s SeaTube platform (SeaTube Pro: Ocean Networks Canada v. 3.0). Each recorded morphotype was documented with still images and a written description following the CATAMI classification scheme (Althaus et al., 2015). Taxonomic assignments to the lowest possible level were made through reference to relevant literature (Best et al., 2010; Lapointe et al, 2020a; 2022; Kenchington et al., 2015; Preez et al., 2015; Quattrini et al., 2015; Ramiro-Sánchez et al., 2019; Sampaio et al., 2019; Watling et al., 2022; Watling and France, 2021; Wilborn et al., 2021; Wing and Barnard, 2004), and online sources. A full list of morphotypes and an identification guide are provided in Supplementary File 1.

For each annotation record of an individual, the associated depth, temperature, salinity and dissolved oxygen concentration data were extracted with SeaTube. Faulty salinity and oxygen measurements, caused by a sensor malfunction at Rockaway Seamount and identifiable as extreme outliers, were removed prior to analysis. Depth profiles of temperature, salinity and oxygen concentration through the entire water column were created from data collected during the ROV’s descent on the Rockaway Seamount (the deepest dive site; Table 1). Substrate was classified by visual examination of video segments corresponding to each annotation, following CATAMI substrate categories (Althaus et al., 2015). Representative images for each substrate class are provided in Supplementary File 2.

Sponge aggregations and coral gardens were identified by plotting annotation counts over time through each dive and verifying peak densities against video imagery. Start and end times of the ROV’s passage through each visible, continuous high-density patch were recorded. Within each patch, the two ROV track points furthest apart in 3D space were identified and the Euclidian distance between them calculated. Patches with a calculated length of at least 25 m were retained for analysis – that dimension being in approximate accord with the 25 m² minimum area threshold set by OSPAR for its identification of coral gardens (OSPAR Commission, 2010). The video records were further examined to determine whether the densities and species composition of corals and sponges within each analyzed patch could be considered as structure-forming VME habitats (sensu FAO, 2009).

Within each high-density patch, ten evenly spaced timestamps were generated between the determined start and end times. Video segments at these timestamps were reviewed and still frames extracted if they showed sharp focus, continuous patch coverage and ideally visible laser points. Between 7 and 10 frames from each patch were determined suitable for analysis.

Using Fiji (ImageJ 2.14.0), the scales of the extracted frames were determined from the 10 cm laser spacing; when lasers points were not visible, average frame widths from frames where they could be seen and taken with the camera at similar elevation and angle, were used. The still frame areas approximated 10 m² (± 10%). Within each frame, all visible individuals of each morphotype were marked and counted using the “Cell Counter” plugin of the software. Substrate class and seabed slope (classed as flat, slope, or wall) were recorded. The counts were rendered as estimated densities (individuals or colonies per m²) of each morphotype, though those are approximations as a result of the uncertainties in the seabed areas covered within each frame. The mean densities of each morphotype within each habitat patch were estimated by averaging the values for each frame. A slope-corrected seafloor area was calculated for each coral garden by creating a polygon from GPS points outlining the garden and computing its planar area in a local UTM projection. The average slope was estimated by fitting a plane to the depth values within the garden, and the planar area was divided by the cosine of this slope to obtain the slope-corrected surface area. These areas represent the approximate seafloor enclosed by the ROV tracks and may include some portions that were not directly observed.

To illustrate assemblage differences between identified coral or sponge patches and areas outside the habitat patches, the list of morphotypes was divided into those only recorded in the patches, those only recorded outside, and those present in both. For that purpose, the presence of morphotypes in the habitat patches was taken from the analysis of the selected frames, whereas records from outside were drawn from the video annotation.

All statistical analyses were conducted in R (version 4.4.1; R Core Team, 2023). Data manipulation and visualization was performed using tidyverse (Wickham et al., 2019) and ggplot2 (Wickham, 2016). Data exploration and code followed Beckmann et al. (2025). The datasets and R scripts used in this study are publicly available via GitHub (https://github.com/lara-maleen/Corner_Rise) and archived in Zenodo (Beckmann et al., 2026).

Morphotype densities were calculated for each dive by dividing the counts in the video annotations by the total transect length. The morphotype abundance matrix was Hellinger-transformed to reduce dominance effects while preserving ecological relationships. Pearson correlation coefficients were calculated between all environmental variables using pairwise complete observations. Variables with ∣R∣>0.7 were considered strongly correlated, while those with ∣R∣ between 0.3 and 0.7 were further evaluated using correlation plots. Environmental variables were standardized to z-scores for ordination and clustering analyses.

To examine depth-related habitat specialization of morphotypes, we calculated the depth range for each morphotype. Morphotypes with a broad depth range (>1,000 m) were classified as “generalists,” while those restricted to narrower depth ranges were considered “specialists.”

Species accumulation curves (SACs) were calculated for each dive and for the habitat patches. Abundances of morphotypes from annotated video and the analyzed frames were used as input. Curves were generated using the iNEXT package (Hsieh et al., 2025).

Pair-wise assemblage differences among dives were evaluated using Bray-Curtis dissimilarity computed from morphotype abundance data. Relationships between Bray-Curtis dissimilarity and both geographic distance and depth differences were assessed using Mantel tests and Pearson’s correlation coefficients (9,999 permutations). Beta diversity was further partitioned into turnover (β_SIM) and nestedness (β_SNE) components across sites using the betapart package (Baselga and Orme, 2012). Significance was assessed with permutation tests (999 permutations).

Unweighted pair group method with arithmetic mean (UPGMA) clustering was applied to the Hellinger-transformed morphotype abundance matrix using Chord distance. Clusters with approximately unbiased (AU) p-values ≥0.95 from multiscale bootstrap resampling (10,000 replicates) were considered statistically significant. The optimal number of clusters was determined using average silhouette width analysis with the cluster package (function silhouette) (Maechler et al., 2025). Environmental variables were clustered similarly using Euclidean distance on standardized data.

Non-metric multidimensional scaling (nMDS) of the Bray-Curtis dissimilarity matrices 1) between dives and 2) between habitat patches and areas outside of the patches, was performed with two dimensions (k=2, trymax=100). Distance-based redundancy analysis (dbRDA) of the Bray-Curtis dissimilarities was performed with the capscale function in the vegan package (Oksanen et al., 2025). Environmental collinearity was again assessed via variance inflation factors (VIF) using the vifstep function in the usdm package (threshold: VIF<5) (Naimi et al., 2014) which resulted in three final predictors (depth, latitude, longitude). The significance of the dbRDA model was tested using permutation ANOVA (999 unrestricted permutations). Site scores from the dbRDA ordination were clustered via k-means partitioning (k=3, 25 random starts) to enable comparison with the UPGMA clusters.

Morphotypes significantly associated with each cluster were identified with the indicspecies package (De Cáceres and Legendre, 2009) and designated as indicators. The Indicator Value (IndVal) statistic combining specificity and fidelity was calculated, with significance assessed by 999 permutations and p-value threshold of 0.05.

Bray-Curtis dissimilarities between clusters were examined with Similarity Percentage (SIMPER) analysis, using the vegan package. Species with significant contributions (p < 0.05, 999 permutations) were identified as key drivers of community differentiation between clusters.

Water-column temperatures decreased monotonically from the surface (Figure 3C), hence bottom temperature was highest, 5.16°C, at the Caloosahatchee dive site and lowest on Rockaway Seamount at 2.24°C There was a pronounced minimum in dissolved oxygen concentration at mid-depths. The lowest value recorded at the seabed was 7.79mg/L on Caloosahatchee Seamount, while the highest was 8.34 mg/L at the MacGregor site.

Bedrock was the dominant substrate at most sites, varying from 100% at Corner Rise 1, 98% at Hopscotch and 91% at Dumbbell, through 80% (and 17% gravel) at the shallow site on Yakutat Seamount, 77% (and 8% gravel) at Rockaway, to 50% bedrock (and 46% outcrop) at the deep site on Yakutat. Caloosahatchee Seamount showed a more even mix, with 48% bedrock and 41% gravel, while MacGregor Seamount exhibited the most heterogeneous terrain with 54% gravel, 27% bedrock and 19% outcrop. Substrate was excluded from multivariate analyses due to relatively limited variation and widespread bedrock dominance.

Bray-Curtis dissimilarity increased with both geographic distance (R = 0.42, p = 0.0187) and the depth difference between dive sites (R = 0.75, p = 0.0001).

Overall beta diversity across sites was high (β_JAC = 0.905). Most resulted from morphotype turnover (β_JTU = 0.883), indicative of a strong variation in species composition between sites, with a much smaller contribution from nestedness (β_JNE = 0.022). Turnover between sites was higher than within sites (mean β_SIM_between = 0.721 ± 0.1, mean β_SIM_within = 0.493 ± 0.092). Permutation tests confirmed that this difference was statistically significant (p = 0.001).

UPGMA clustering identified three primary assemblages: one associated with upper-mid bathyal sites (MacGregor, Caloosahatchee, Yakutat shallow and Yakutat deep); a second associated with lower bathyal sites (Castle Rock, Corner Rise 1, Dumbbell and Hopscotch); and a third corresponding to the abyssal site (Rockaway). The dendrogram and dissimilarity matrix were significantly related (r = 0.93), and silhouette analysis supported the optimal number of three clusters.

The nMDS ordination (k=2, stress = 0.041) displayed differences in community composition among the sites, which were grouped in ordination space as in their three clusters (Supplementary File 5).

The overall dbRDA model performed with depth, latitude and longitude as constrained variables (variables with no significant multicollinearity, VIF<5) was statistically significant (ANOVA: F = 1.67, p = 0.006), the first two axes explaining 24.65% and 15.66% of the variation, respectively (Figure 4). Depth was the only significant environmental driver (F = 2.12, p = 0.008) of the three. Additionally, Yakutat Seamount was sampled at two depths separated by approximately 400 m and were distinctly separated, with intra-seamount dissimilarity comparable to that observed between different seamounts.

Six morphotypes were significantly associated with the assemblage differences between the shallow and deeper groups identified by UPGMA clustering (p < 0.05; Figure 5). The lower bathyal cluster was characterized by CAO11 (Chrysogorgia sp. A, IndVal.g = 1.0), CAO21 (Keratoisididae, 1.0) and PD03 (undetermined sponge, 1.0) as indicator taxa. The upper-mid bathyal cluster featured CAO28 (Acanella sp., 1.0), CAO29 (Chrysogorgia sp. B, 0.998) and PD17 (Demospongiae, 1.0). The abyssal site hosted unique taxa including CAH23 (Abyssopathes sp.) and PH70 (Hyalonema (Corynema) sp.) and with that the exclusive deep dwelling families Cladopathidae and Hyalonematidae.

SIMPER analysis revealed clear contrasts among the clusters. Overall, Bray-Curtis dissimilarity between the upper-mid bathyal and lower bathyal cluster was high (0.93). While CAO11 (Chrysogorgia sp. A) contributed 0.7%, the top three species together accounted for only 12.4% of the total dissimilarity, indicating that the difference is spread across many species. The assemblage at the Rockaway site was more distinctive, with particular contributions to dissimilarity by CAO28 (9.5%, coral exclusive to the deeper sites) and PH13 (8.3%, sponge), while that between Rockaway and the upper-mid bathyal cluster was dominated by PD08 (13.1%, sponge exclusive to the shallower sites), with additional contributions from CAH04/CAH30 (Bathypathes spp.), PH07 (Sceptrulophora), and CAO11. Across all comparisons, Chrysogorgia sp. A (CAO11) emerged as a key taxon, distinguishing the groups.

Six high-density habitat patches were identified across five dive sites, with depths from 945 m to 2,495 m (Table 2, Figure 6). In each, either corals or sponges visibly outnumbered other megafauna (Figure 7), four patches being sponge-dominated, with Porifera making up 70.4% to 97.2% of the observations. The other two were coral gardens, with 5.2% of Porifera recorded at Caloosahatchee site and 25.9% at MacGregor (A). Total faunal densities within the patches were between 1.66 ± 0.75 ind/m² on MacGregor Seamount, and 4.90 ± 1.84 ind/m² on Caloosahatchee Seamount. The SACs for the VME habitats were not saturated, thus no morphotype richness analysis and comparison was conducted.

No morphotype occurred or dominated exclusively across all VME habitats. Despite high overlap in morphotype composition between the designated habitats and adjacent non-habitat areas, each habitat patch harbored unique taxa absent from surrounding habitats. An NMDS ordination (k =2, stress = 0.059) showed that the assemblages within and around identified habitat patches within a site were more similar to one another than to assemblages at other sites (Supplementary File 6).

Two high-density patches were identified at the Dumbbell Seamount: Habitat A (at mean depth of 2,403 m and ~55 m in length) and Habitat B (at 2,366 m and ~92 m in length). They exhibited similar assemblage compositions, dominated by sponge aggregations. The most abundant taxa included an unidentified demosponge (PD08) and fan-shaped or elongated glass sponges such as Farrea sp. A (PH24). Three sponge morphotypes, PH07 (Sceptrulophora sp.), PH58 (Farrea sp. B), and PH73 (Hexactinellida), were found exclusively within the patches and not in the surrounding area.

A single high-density patch was identified at a mean depth of 2,495 m on Corner Rise 1, spanning approximately 76 m, and dominated by the same ground-covering demosponge (PD08) seen on Dumbbell Seamount, as well as a fan-shaped bamboo coral (CAO21). A funnel-shaped euplectellid sponge (PH36) was observed exclusively within the habitat and was absent outside.

Two distinct patches of biogenic habitat, one extensive, were observed on MacGregor Seamount. Patch A, on a steep wall approximately 77 m in length, starting at 1,191 m depth, was classified as a coral garden and characterized by numerous undetermined stick-like corals (CA01), along with large, bushy Antipatharian colonies. Several taxa were exclusive to this high-density patch, including corals CA01, CAO02 (Lepdisis spp.), CAO30 (Thouarella cf. grasshoffi), CAO39 (Calyptrophora cf. clinata), and the sponge PH59 (Asconema sp.).

Patch B, on flat terrain at 945 m and spanning roughly 164 m, was dominated by bulky, undetermined sponges (P07) and bright yellow, ground-covering demosponges (PD17), both of which covered substantial areas. Interspersed among those were low densities of corals such as CAO22 (Keratoisidinae J3), CAH15 (Madrepora cf. oculata), CAH22 (Leiopathes sp.), and CAO34-38 (including Acanthogorgia sp., Paragorgia sp., and Corallidae). Aphrocallistes sp. and Rossellidae sponges were also present.

The two patches - wall and flat - shared several taxa, including CAH08 (Parantipathes sp.), CAH37, and CAO33 (Plexauridae). Notably, the MacGregor site exhibited the highest distinctiveness of all those surveyed, with 18 morphotypes recorded only there.

The Caloosahatchee site at 1,225 m hosted the second coral garden observed, extending continuously for approximately 387 m. It was dominated by primnoid corals, particularly Thouarella cf. grasshoffi (CAO30). Five coral taxa were exclusive to the high-density patch: CAH04 and CAH30 (Bathypathes sp.), CAH20 (Enallopsammia cf. rostrata), CAH35, and CAO42 (Iridogorgia cf. frontalis).

While the clustering of the sites into three distinct groups initially suggested a longitudinal biogeographic zonation, depth was the sole statistically significant driver, the appearance of an east–west trend perhaps arising from the distribution of the selected dive sites in relation to their surrounding water masses. Uniform sampling across the full CRS depth range would be needed to fully describe the zonation. Assemblage composition did not change steadily with depth, as might be expected if distance from the surface or water pressure was the primary control. Rather, there was a marked transition between the upper-mid bathyal and lower bathyal groups at about 2,000 m depth, suggestive of the influence of layered water masses.

The depths of the MacGregor, Caloosahatchee and Yakutat sites placed them in the North Atlantic’s complex Intermediate Layer (sensuLiu and Tanhua, 2021). During the fieldwork, that underlay the prominent oxygen minimum, centered around 800 m depth (Figure 3) within Eastern North Atlantic Central Water (sensuLiu and Tanhua, 2021). The shallowest dive sites reached depths where that water mass was mixed into the Intermediate Waters. Around the CRS, those are primarily derived from Subarctic Intermediate Water (itself formed by sinking of Warm Slope Water along the northern side of the North Atlantic Current: Liu and Tanhua, 2021), perhaps with an admixture of Mediterranean Water, which by some definitions penetrates even so far to the westward (Samborskaia et al., 2025).

In contrast, the deeper sites were bathed by the upper North Atlantic Deep Water – formed by mixing in the Irminger Sea of, primarily, Labrador Sea Water (LSW) and Iceland-Scotland Overflow Water that had passed the Mid-Atlantic Ridge through the Gibbs Fracture Zone (Liu and Tanhua, 2021; Henry et al., 2024). Still deeper, below about 3,000 m, the lower North Atlantic Deep Water contains a higher proportion of Denmark Strait Overflow Water and correspondingly less LSW, while there is an increasing admixture of Antarctic Bottom Water (termed Northeast Atlantic Bottom Water by Liu and Tanhua (2021)). At the depth of the Rockaway Seamount site, the contribution of the latter approaches 50% (Liu and Tanhua, 2021) – consistent with the observed temperature (~2.2°C) and salinity (~34.8 PSU).

Similar apparent associations between coral assemblages and water masses have recently been reported from around the Azores, in mid-Atlantic (Taranto et al., 2023), and from the North American continental margin, near the end of the New England Seamount Chain (Rakka et al., 2025). In the latter area, Rakka et al. (2025) noted that aragonite corals were found at depths bathed by higher-salinity North Atlantic Central Water, whereas calcite corals occurred at both lesser and greater depths, in lower-salinity waters of northern origins. In contrast, all assemblages observed on the CRS immersed in layers of similar salinity. Thus, horizontal connectivity, mediated by regional currents, and constraints on vertical dispersal of gametes and larvae, imposed by density differences between water masses, may dominate in structuring the observed zonation through recruitment (cf. Wang et al., 2020; Yearsley et al., 2020; Patova et al., 2025; Rakka et al., 2025; Taboada et al., 2025).

Beta diversity among coral and sponge communities was high, and most dissimilarity was driven by species turnover rather than nestedness. This indicates that species composition varies substantially across sites, with species replacement rather than simple loss of taxa driving community differences. Such high turnover is consistent with the observed strong environmental structuring. Similar patterns have been observed in other deep-sea seamount systems, where turnover dominates beta diversity and contributes to high regional diversity (e.g., Victorero et al., 2018; Taranto et al., 2023).

Nevertheless, sampling effort varied among seamounts, with most sites represented by a single ROV dive, which necessarily limits the ability to assess species uniqueness and completeness at individual features. Species accumulation curves did not reach an asymptote, indicating that additional sampling would likely reveal further taxa, including potentially rare or locally restricted species. As a result, we avoid overinterpreting differences in species richness among seamounts and focus instead on broader community structure, habitat-forming taxa, and depth-related patterns that are robust to sampling intensity. Despite these limitations, the observed densities of VME indicator taxa and the consistency of assemblage patterns across depth zones provide strong evidence for the ecological significance of the Corner Rise seamounts and support their consideration for precautionary protection.

We confirmed the presence of six high-density patches on the CRS, at depths from 945 m to 2,495 m, each extending over at least 25 m² - an extent matching the OSPAR guideline for identifying coral gardens (OSPAR Commission, 2010). The densities in those patches, 1.7 and 4.9 colonies per m² in coral gardens and 2.3 - 3.5 colonies per m² in sponge-dominated patches, were previously reported ranges. Coral gardens in UK waters average 1–9 colonies per m² (Henry and Roberts, 2014), while single-species gardens on the Mid-Atlantic Ridge reach maximum reported densities of 0.3 colonies per m² (Morato et al., 2021). The glass sponge Pheronema carpenteri has been found in aggregations of up to 1.53 m^-² on the Goban Spur (Hughes and Gage, 2004), but 15 individuals per m² have been observed in the geodiid sponge grounds on Sackville Spur (Beazley et al., 2015). In North Pacific seamount habitats, coral and sponge densities of up to 20 individuals per m² have been reported, though most identified patches had 2–3 individuals per m² (Beckmann et al., 2025).

Several of the taxa present are recognized VME indicators, including Thouarella cf. grasshoffi, Paragorgia spp. and Rossellidae sponges. FAO (2009) established criteria for identifying VMEs (Uniqueness or Rarity, Functional Significance of the Habitat, Fragility, Structural Complexity, and Life-History Traits, including slow growth, late maturity and low fecundity) which are met by many of our identified sponges and corals. On the CRS, we identified several VME coral taxa, such as Enallopsammia rostrata, Madrepora oculata, Leiopathes sp., Metallogorgia melanotrichos, Swiftia and Thouarella grasshoffi. Among the sponges, Hertwigia and Farrea build large complex structures, and Aphrocallistes vastus, a common species across the Atlantic, can form deep-water reefs (Austin et al., 2009).

Species compositions inside and around the CRS habitat patches did not differ significantly, hence the corals and sponges there do not form distinct high-density assemblages but simply dense aggregations of species already present across the broader seafloor. Thus, local availability of key resources, such as substrate and nutrients, appears to drive the formation of dense aggregations.

Our approach offers a means of reducing subjectivity in the identification of VMEs. Others have used kernel density estimation analyses to identify significant concentrations of VME Indicator taxa from research vessel trawl catches (Kenchington et al., 2014) and from underwater video (Meyer et al., 2019). The latter used kernel density estimates to visualize the spatial patterns of the most prominent megafauna, demersal fish, and skate egg cases on the summit of Schulz Bank (Arctic Mid-Ocean Ridge) to identify areas of dense aggregation within the sampled area. Rowden et al. (2020) and Beazley et al. (2015) each identified density thresholds where the abundance of a structure-forming taxon increased the diversity of associated species. Rowden et al. (2020) developed a quantitative approach to determine a density threshold of VME indicator taxa above which a VME was determined to be present, to be used with predictive distribution models on seamounts in the South Pacific. Beazley et al. (2015) used gradient forest analyses on imagery data collected from photographic transects in the northwest Atlantic to identify the structure-forming VME sponge density which elicited the largest turnover in megafaunal community composition indicative of an increase in biodiversity. Less quantitative approaches have been used by ICES, which endorses a somewhat circular and subjective definition of ‘VME habitats’ which are ‘records for which there is unequivocal evidence for a VME, e.g. ROV observations of a coral reef’ (ICES, 2016) leaving open the question of spatial scale. We suggest that the criteria used herein offer a quantitative basis for identification of VMEs from video observations that does not require full image analyses of associated species and are less computationally demanding than kernel density estimation.

Our study builds upon this framework by extending geographic and depth coverage across a larger number of CRS seamounts (n = 8), reaching depths of ~4,100 m - considerably deeper than the ~2,400 m maximum in the previous studies - and resolving a clearer depth-stratified east–west structure not previously reported for this region.

Comparisons of the taxa recorded on the two seamount chains suggest both parallels and regional divergences. The deeper sampling at CRS naturally revealed taxa absent from New England Seamounts (NES) surveys, including the black coral Abyssopathes and Umbellula sea pens at ~4,100 m on Rockaway Seamount. At depths around 1,800 m, where the two seamount chains may be connected by transport in the Subarctic Intermediate Water, the assemblages at CRS sites such as Yakutat and Castle Rock overlap with those reported from NES seamounts like Balanus and Retriever. Shared taxa include the corals Candidella imbricata, Metallogorgia melanotrichos, and Iridogorgia magnispiralis, along with sponges such as Hertwigia spp. and Farrea spp. Conversely, Lophelia pertusa, sea pens and the hexactinellid sponge Bolosoma, which were reported from NES, were not observed on CRS. Their absence may reflect dispersal limitations or other environmental differences between the seamount chains. Future studies combining particle-tracking models with population genomic data will be critical to test connectivity hypotheses (e.g., Taboada et al., 2023; Patova et al., 2025; Taboada et al., 2025), with candidate species including Metallogorgia melanotrichos, Enallopsammia rostrata, and potentially novel Hertwigia spp. from CRS and NES. To support these efforts, further taxonomic work is urgently needed. Several morphotypes recorded at CRS could not be confidently assigned to known taxa, as specimen collection was limited. Integrating morphological, molecular, and image-based approaches will be key to resolving cryptic diversity and generating reference-quality datasets. A strengthened taxonomic baseline will not only advance biodiversity science and biogeographic modeling but also improve the identification of VME indicator taxa - crucial for conservation management in under-surveyed regions such as the Corner Rise seamounts.

Comparisons of sponge diversity remain particularly constrained by differences in taxonomic resolution and methodology. While Lapointe et al. (2020b) identified numerous sponge taxa, many were reported under unique vernacular labels without accompanying image databases, limiting cross-study validation, and highlighting the need for image-supported reference collections.

The cross-jurisdictional position of the CRS and NES chains illustrates the need for cooperative conservation measures in ABNJ. The BBNJ agreement may provide the space for such discussions to take place. We also note that while bottom-fishing can cause significant adverse impacts to VMEs on seamounts (Waller et al., 2007; Lapointe et al., 2020a), other anthropogenic threats such as deep-sea mining (Washburn et al., 2023) must also be addressed. Fora such as the BBNJ would allow for such cross-sectoral discussions.

While NAFO had previously identified the presence of VME Indicators to 4,000 m (Waller et al., 2021), our study establishes the presence of VMEs to depths of 2,495 m on the CRS, while indicator taxa were found broadly along our transects, outside the high-density patches. Such scattered individuals and colonies may make significant contributions to connectivity, providing ‘stepping stone’ links mediated by dispersal (Wang et al., 2024). Thus, our results strongly support the continuation of the current NAFO protections beyond 2027. Further, although no review of VME within the WECAFC area has been scheduled, we recommend that the Commission maintains and strengthens its decisions concerning protection of the CSR. In particular, Caloosahatchee Seamount harbors high-density coral gardens, including VME-designated octocorals such as Iridigorgia cf. frontalis and Calyptrophora cf. clinata, while Dumbbell Seamount has dense sponge aggregations (mostly Farrea sp.), which meet VME criteria as reef-forming, disturbance-sensitive taxa, as well as several corals (e.g. Chrysogorgia cf. tricaulis).

Although we have recorded many taxa that meet FAO’s (2009) criteria for VME indicators on the CDS, none of them are listed as Indicator Species by NAFO (Annex I.E.). The Organization’s list is heavily biased towards species present on the existing fishing grounds within the NAFO regulatory area, where they may be encountered by commercial gears. We therefore recommend that NAFO generate a separate list of VME Indicator Taxa associated with the CRS and NES, so that future proposals for exploratory fishing can better anticipate potential VME encounters.

Advanced technologies have expanded our ability to explore the deep sea, and are generally targeted to scientific investigations that generate the foundational knowledge required for adaptive and evidence-based management. These efforts are essential not only for locating VMEs but also for understanding biodiversity patterns that underpin ecosystem function and resilience. While much of the CSR and NES remain unexplored, it will be critical for any exploratory fishing that might be permitted to conduct meaningful impact assessments.