The São Paulo Ridge (SPR) is a 350 km-long linear geological feature located in the Continental Margin off Brazil, between 28–29°S and 40–45°W (Figure 1A). It is a component of the east-west trending Rio Grande Fracture Zone alignment that delineates the southern boundary of the São Paulo Plateau (Bassetto et al., 2000). As other aseismic ridges, the SPR is asymmetric with (a) a flat 2,000 m-deep northern flank, buried by sediments of the São Paulo Plateau, and (b) a steep southern flank (known as the São Paulo Escarpment), diving from 2,500 m-deep crests to a 4,200 m-deep foot (Gamboa and Rabinowitz, 1981; Alberoni et al., 2019). It is believed that SPR current morphology is largely derived from irregular submarine vulcanism that started in the Aptian period (∼120 m.a.), during the early South Atlantic Ocean expansion, when the SPR acted as a barrier obstructing the marine circulation toward the north and promoted the formation of a shallow environment where the deposition of a thick layer of evaporites resulted in the formation of the São Paulo Plateau (Gamboa and Rabinowitz, 1981; Bassetto et al., 2000; Ichisima et al., 2017). The southern flank interposes the northern flow of the Antarctic Bottom Water (AABW) generating countercurrents and eddies that mobilize sediments and form a trough delineating the deep contour of the ridge, also known as the São Paulo Channel (Figure 1B; Gamboa and Kumar, 1977; Alberoni et al., 2019). The NADW flows southward over the São Paulo Plateau, across the SPR crests and down over the escarpment where it overlays the AABW.

Data on SPR seafloor benthic habitats and megafauna were acquired during a research cruise conducted by the RV “Yokosuka” in 2013, under the “Iata – Piuná” consortium established between the Japan Agency for Marine-Earth Science and Technology (JAMSTEC), the Oceanographic Institute of University of São Paulo (IOUSP) and the Geological Survey of Brazil (CPRM) (Kitazato et al., 2017). The SPR area was explored between April 23 and 27 and included swath bathymetry and deep-sea dives of the manned submersible Shinkai 6500.

Bathymetric data were acquired during along-ridge transects with a 12 kHz hull-mounted Multibeam Echosounder System (MBES), integrated Differential Global Positioning System (DGPS) and compensated by the Inertial Measurements System (IMU). The data were filtered to remove spurious depth records and submitted to methods of linear interpolation to produce a Digital Bathymetric Model (DBM) with a cell size of 123 m. Video transects were produced during four deep-sea dives along a complete SPR southern flank depth profile, from 4,219 to 2,644 m depths (Figure 1 and Table 1). Dives 6K1333 and 6K1336 explored the abyssal region at the foot of the ridge. Part of these dives were dedicated at the study of a whale-fall carcass environment, encountered at 4,204 m depth, previously described by Sumida et al. (2016) and excluded from this study. Dives 6K1334 and 6K1335 described the seafloor along the depth gradient of the São Paulo Escarpment. All dives involved approximately 4 h of activities near the seafloor, including photo/video recording, geological and biological sampling. Video data was acquired by two HD-TV color video cameras, both positioned at the bow, 1.7 m above the vehicle’s bottom (Nakajima et al., 2014). Camera 1 angled obliquely 40° toward the seafloor and recorded continuously a fixed area ahead the bow of the submersible (horizontal acceptance = 90°, vertical acceptance = 57°). Camera 2 was mobile (pan – tilt) and was used for detailed observations of habitat features and megafauna species. Continuous information of date/time, depth (and altitude in meters) and the vehicle’s heading (in degrees) were overlaid in the videos. Horizontal position (latitude, longitude) was estimated by SSBL (Super Short Base Line) method which required a transponder mounted on the submersible and an array of transducers on the hull of the RV Yokosuka. In this method the position is estimated by (a) phase lag measured from angles of received sound waves, and (b) distance calculated from their traveling period. Records of latitude and longitude were transformed into easting and northing (UTM) to estimate the linear distance covered by the submersible. The area covered by video transects was calculated by multiplying this linear distance by the mean lower width (MLW) of images taken every 6 s, following the equation proposed by Jones et al. (2009) for oblique videos with cameras facing the imaging path:

where α is the altitude of the camera, θ and ω are the camera’s horizontal and vertical acceptance angles, respectively, and δ is the angle of the camera from vertical (Nakajima et al., 2014). Measurable uncertainty in area estimations derived from variability in α along video transects. Bootstrap 95% confidence intervals of MLW were calculated for video transects or segments of it, using the quantile method.

Megabenthos specimens were collected by the submersible’s manipulators and slurp-gun, photographed on board and stored in ethanol (75%) and formalin for posterior identification by specialists in the National Museum (Federal University of Rio de Janeiro) where they were cataloged.

Videos, sample records and oceanographic data and metadata are available in “DARWIN – Data and Sample Research System for Whole Cruise Information in JAMSTEC².”

The DBM was analyzed for seafloor morphology and segmentation (Brown et al., 2011). This process included the transformation of bathymetry data into secondary-derived layers (terrain variables) using the algorithm package Benthic Terrain Modeler (BTM) (Walbridge et al., 2018) contained in ArcGIS Desktop^® 10.2.2. Variables with the greatest potential for the description of benthic habitats at the corresponding spatial scales were chosen, including surface gradients (slope, aspect), relative depth (Bathymetric Position Index – BPI) and surface rugosity (Vector Roughness Measurement – VRM) (Wilson et al., 2007; Walbridge et al., 2018). BPI is a neighborhood analysis function of the mean depth around each cell in the DBM. Positive values correspond to features and regions that are higher than the surrounding area which characterize ridges; negatives values would represent depressions on the seafloor. Values equal or near zero are either flat or constant slope area (Walbridge et al., 2018).

The procedures for seafloor segmentation (Figure 2) started with overlaying 23 bathymetric profiles on the DBM, transversal to the bathymetric gradient of the São Paulo Escarpment (Supplementary Figure 1) for visual interpretation. Along each profile, the variables BPI-broad (annulus = 300 pixels ∼ 36,900 m), BPI-fine (annulus = 60 pixels ∼ 7,380 m) and slope were extracted and analyzed individually in order to identify Zonal and Structural classes (sensu Erdey-Heydorn, 2008). Zonal classification described broad scale surficial characteristic of the seafloor. Classes representing the main bathymetric components were visually defined, and had their boundaries delimited by depth and values of BPI-broad and slope (i.e., averages across the 23 transversal bathymetric profiles). Within each Zonal class, surficial characteristics of the seafloor were then described to a greater detail (Structural classification) using terrain variables BPI-fine and slope. Structural classes were again visually defined and delimited by maximum and minimum values of depth, BPI-fine and slope values (i.e., averages across the 23 bathymetric profiles). A classification table was then built, summarizing lower and upper limits of depth, slope, BPI-broad and BPI-fine for all visually defined Zonal and Structural classes (Supplementary Table 1). The classification table was then used as input data for semi-automated classification of new adjusted classes by BTM’s Classify Benthic Terrain tool. Such classification procedure (followed by spatial representation of the resulting classes) was performed several times until the spatial segmentation was considered satisfactory (Figure 2). Each model run was conducted after manual adjustments of the terrain variable values delimiting structural classes, and/or the creation of new intermediate classes.

Each video produced by camera 1 was initially observed to record changes in depth, altitude, submersible activities and image visibility. Also, seafloor features relevant for habitat classification (following Greene et al., 1999) were selected, namely: substrate texture, particle sizes, relief and others. These features were organized in a scale of “substrate types,” represented by capital letters, where: R = rocky ridge with rugged surface; F = flat rocky pavement with plain or granulated surface; B = loose boulders (>25.5 cm); C = cobbles (>6.5 cm and <25.5 cm); P = pebbles (>2 cm and <6.5 cm); G = gravel (>4 mm and <2 cm); U = unconsolidated substrate varying from fine mud to coarse sand.

A second analysis of these videos included only segments when the submersible was moving ahead. During these segments, substrate types were attributed to 1-min video intervals using a two capital letters system; the first referring to the substrate type covering more than 50% of the visible seafloor, and the second to the substrate covering between 30 and 50%. For example, a segment where seafloor was mostly unconsolidated (e.g., covered by biogenic sediment) with variable amounts of scattered pebbles was codded as UP. Also, when only one type of substrate was visible during the observed interval the correspondent letter was duplicated (e.g., UU = seafloor completely covered by sediments). These combinations of two substrate types were defined as “bottom types” (Greene et al., 2007; Tissot et al., 2007; Tissot, 2008).

Seascapes units were mainly defined by one or more dominant bottom types regularly combined along a continuous segment of the seafloor. Adjacent seascapes were delimited by an abrupt change in the dominant bottom type (or bottom type combination) observed for more than 10 s along the video track, indicating the end of a seascape unit and the beginning of a new one. Bottom types that contrasted with the dominant ones but did not persist long enough along the video track (e.g., a patch), were not considered a new seascape unit, but as part of the current seascape substrate variability. Descriptions of seascape units were complemented by “modifying” elements (e.g., currents, bioturbation), relief and slope (Greene et al., 1999), as well as by the influence of AABW and NADW on the seafloor. This influence was estimated by temperature and salinity data recorded continuously during the dive track by the submersible’s CTD. Mixing percentages of these water masses during the dives were calculated according to Mamayev (1975).

Videos produced by cameras 1 and 2 were analyzed for visible megafauna. Life forms smaller than approximately 5 cm and/or visible in images taken above 2-m altitude were not included in the analysis. Otherwise, videos were stopped at each sighting and recorded the organism type (“morphotype”) along with associated information, including date, time, depth (m), altitude (m), and heading (in degrees). Morphotypes were classified in higher taxa (Phylum, Class, Orders) and their consistency was double checked by repeating the analyses of videos produced by both camera 1 and 2. When no morphotypes could be confidently assigned to a given observed organism after these repeated analyses, it accounted for higher taxa quantification only. Some morphotype identifications to family, genus and species level were possible through collaboration with deep-sea fish and invertebrate taxonomists, and with the aid of deep-sea fauna image databases (e.g., OER’s Benthic Deepwater Animal Identification Guide³ and others).

Megafauna spatial distribution and abundance was analyzed by representing density of total recorded megafauna and/or taxonomic groups as a function of the depth strata and seascape units. Density was calculated by dividing organism numbers by the mean estimated areas covered by video transects (expressed as individuals. m^–2), and their 95%CI limits.

The SPR is depicted as the southern margin of the São Paulo Plateau, which extends to the north as 3,000 m-deep plane areas (Figure 1). Along the border of the plateau, prominent ridge crests rise to 2,200 m depths, over 1,000 m above the plateau level. These crests are separated by plane areas and channels that extend to the edge of the plateau, possibly characterizing paths of deep-water flow. The margin of the plateau is continuously bordered by the São Paulo Escarpment that forms a south – southeast facing wall, which is (a) higher/steeper in sectors adjacent to the plateau crests (Supplementary Figure 2, e.g., profiles 3, 11, and 19) and (b) lower/gentler in sectors adjacent to plane areas in between the ridge crests (Supplementary Figure 1, e.g., profiles 8, 15, and 22). The deep edge of the escarpment is connected to the 4,100–4,300 m-deep São Paulo Channel, a trough extending continuously along the SPR. To the south of this channel, the seafloor rises into a 4,000 m-deep plain area that is part of the Santa Catarina Plateau (Figure 1).

The spatial representation of BPI-broad delineated the areas comprised by broad bathymetric components, including the São Paulo Plateau, plateau crests, São Paulo Escarpment, São Paulo Channel and Santa Catarina Plateau (Figure 3). Secondary seafloor structures were evidenced from the spatial analysis of BPI-fine and included valleys and channels on the São Paulo Plateau, as well as discontinuities of the São Paulo Channel. Slope, aspect (easterness and northerness) and roughness (VRM) provided refined spatial representations of (a) channels and valleys bordering the plateau crests, and crossing the margin of the São Paulo Plateau, and (b) the escarpment, depicted as a steep, southern-faced roughed terrain (Figure 3). These were seafloor topography subject to classification by the morphometric analysis (see below).

The seafloor classification procedure resulted in 12 Structural Classes (Figure 4) enclosed within four Zonal Classes. Descriptors (upper and lower limits of depth, terrain variables and slope) and names attributed to each on of them are presented in Supplementary Table 1. The plateau (Zonal Class II) comprises plains (Class II.5), valleys (Class II.6), and gentle slopes (Class II.7) which were shallower than 3,700 m and with less than 5° slope. The SPR crest (Class I) is composed of seamount-like summits of high (Class I.2), moderate (Class I.3), and low (Class I.4) altitudes, some of the former topped by 2,240 m-deep peaks (Class I.1). Bordering the edge of the plateau, the São Paulo Escarpment (Class III) is formed by an upper terrace (Class III.8) followed by a deeper scarp (Class III.9). At the ridge foot (Class IV), the São Paulo Channel (Class IV.10) runs continuously along the entire SPR extension, sometimes bordered by local depressions next to the escarpment (Class IV.11), and connected to the south to the deep plains of the Santa Catarina Plateau (Class IV.12) (Figure 4).

The deep-sea dives crossed four zonal classes: ridge foot, ridge escarpment, plateau and ridge crest (Figure 4C). Seafloor texture was dominated by biogenic sediments usually mixed with scattered rocky particles (UP, UC), outcrops and crests (UF, UR) (Supplementary Table 2). These bottom types were recorded for over 56% of the observation time, followed by bedrock outcrops (RU + RC = 33.5%). Areas entirely covered by rocky substrata (e.g., RC, RR, RP, FP, PC) were recorded in 7.1% of the observed time. Most of the area observed (>90%) corresponded to gentle slopes (5–30°). Steep slopes (>30°) occurred in 2% of the observed area and only in the ridge escarpment.

Texture classification, slope, water masses and modifying elements allowed the differentiation of seven seascape units (Table 2). The “Abyssal Mud Field” (AMF), was characterized by an undulated sediment surface covering the São Paulo Channel seafloor, with isolated patches of pebbles and bedrock outcrops (UP, UR). This seascape dominated images produced by dive 6K1333 that explored a linear track along the foot of the ridge, in eastward direction over the 4,000–4,100 m isobaths. The effect of AABW flux was noticeable in this seascape, usually producing regular sand waves (Table 2). Adjacent do AMF, the “Debris Field” (DF) extended along the deep border of the escarpment generally characterized by a mixture of cobbles, pebbles, and boulders, interspersed with biogenic sediments (UC and UP). These bottom types, recorded in 64.5% of the total observation time of this seascape (Supplementary Table 2), were often associated with landslide signs and coated by a thin sediment cover (Figure 5A).

Moving into the lower section of the ridge escarpment (above 4,100 m depths) the seascape was dominated by a steep and roughed bedrock surface, the “Bedrock Cliff” (BC) (Table 2). Rocky ridges sometimes covered by piles of loose cobbles (RU and RC) dominated the seafloor (64.5% of the observation time, Supplementary Table 2) (Figure 5B). Small amount of sediments accumulated in cracks and crevices, and thinly covered bedrock particles. This seascape unit was interrupted at approximately 4,070–3,998 m depths by a narrow terrace formation, the “Granular Flat” (GF) (Table 2 and Figure 5C), where flat bedrock pavements were usually covered by pebbles and variable amounts of sediments (FP + FU = 71.4% of the observation time, Supplementary Table 2). A positive (convex) relief characterized the upper section of the escarpment (above 3,366 m depths), where the seascape “Bedrock Crest” (BCR) was defined by a mixed substrate formed by bedrock outcrops and sediment ponds (RU) (Table 2 and Figure 5D). This seascape unit was also observed in the ridge crest, between 2,578 and 2,657 m depths (Figure 5).

At the ridge plateau and the adjacent ridge crest, flat sedimented areas (UU), occasionally with sparse patches of pebbles and cobbles (UP, UC), formed a seascape unit called “Mud Terrain” (MT) (Table 2). Sediment surface was generally smooth, but often modified by animal tracks (Figure 5E). The ridge plateau also included a slightly convex “Granular Crest” (GCR) (Table 2 and Figure 5F) covered by pebbles, cobbles and soft sediments (RU + UP = 82.2% of observation time Supplementary Table 2).

Temperature of the water mass overlaying the seafloor ranged between 0.45 and 3.11°C, and salinity between 34.59 and 34.94 psu (Supplementary Figure 2), well within the limits of the NADW and AABW (according with Morozov et al., 2010). The AABW-NADW interface, delimited by the isotherms 1.8–2.0°C (Zhao and Thrunherr, 2017) at 3,409–3,433 m depths in the ridge escarpment (Supplementary Figure 3), coincided with the boundary between seascapes BC and BCR (Figure 5). Density decreased sharply in this depth zone characterizing mixing processes, i.e., warming of AABW (Zhao and Thrunherr, 2017) (Supplementary Figure 2). AABW influence predominated on seascape units AMF, DF and GF (>82% mixing) and NADW influence predominated on seascapes BCR, MT, and GCR (>85% mixing) (Table 5). Mixing percentages of these water masses varied widely on BC (AABW 20–93%, NADW 7–80%). Highest temperatures (2.75–3.10°C) were recorded at the plateau – ridge crest (3082–2676 m) and lowest (0.4–0.6°C) at the ridge foot (4,208–4,014 m).

A total of 914 organisms of the epibenthic and benthopelagic megafauna were recorded during 700 min of SPR seafloor observation. Over 70% of the records were sessile suspension feeders, including sponges (62.8%) and anthozoans (11.6%). These were followed by shrimp-like crustaceans (11.3%) and fish (Actinoperygii, 8.7%) (Table 3). A list of identified taxa is presented in Table 4.

Megafauna records were usually sparse (<0.01 individuals.m^–2) along the SPR depth gradient, but two patches of moderate concentrations (0.07–0.30 individuals.m^–2) mostly of suspension feeders, occurred at 3,800–3,300 m, and 2,900–2,700 m depth intervals (Figure 6). The former patch occurred in the ridge escarpment and the BC and BCR seascapes. The latter patch occurred at the ridge crest and within the reach of the GCR, MT and BCR seascapes. Both depth zones were covered by hard substrata (e.g., outcrops or loose particles), but also included a variable coverage of sediments.

Most taxonomic groups were observed above 2,900 m and were associated with mixed substratum in the BCR seascape (Table 3). Nearly 60% of all records of Porifera occurred in this seascape being largely represented by two species: a pedunculated sponge, Family Dendoricellidae (cf Pyloderma sp., Demospongiae), and Poliopogon amadou (Hexactinellida) (Figure 7). Corals (Anthipataria and Alcyonacea) also tended to occur deeper, between 3,300 and 3,800 m, on the ridge escarpment (Seascape BC). 74.3 and 74.0% of crustaceans and echinoderms, respectively, were recorded below 2,900 m (Table 3). The latter was dominated by Holothuroidea, Order Elasipodida, including Psychropotes semperiana, Benthodites sp., and Enypniastes sp. These were observed in BCR usually occurring on sediment ponds. Nearly 44% of fish records concentrated in the 2,700–2,900 m depth stratum (Table 3). One species, Acanthonus armatus (Ophidiiformes, Ophidiidae) was particularly abundant in association with the GCR seascape (Figure 7D).

Epibenthic and benthopelagic megafauna was recorded within a total area estimated in 35609.4 m² (34076.6–37236.8 m² 95% CI). Mean density was 0.026 individuals.m^–2 (0.024–0.027 individuals.m² 95% CI) decreasing progressively with depth (0.130–0.005 individuals.m^–2, Table 5). The sectors above 3,300 m, including GCR, MT and, BCR seascapes, exhibited more elevated concentrations of megafauna, all of them under a prevailing influence of NADW (usually > 80%, Table 5). Upper sectors of BCR (<2,900 m) exhibited the highest fauna concentrations (0.328 individuals.m^–2 – 0.300–0.360 95%CI) largely dominated by sponges cf Pyloderma sp. and P. amadou (Table 5 and Figure 7C). Estimated densities of taxonomic groups are presented in Supplementary Table 3.