The Salas y Gómez and Nazca ridges are two adjacent seamount chains off the west coast of South America that collectively contain more than 110 seamounts. The ridges support an exceptionally rich diversity of benthic and pelagic communities, with the highest level of endemism found in any marine environment. Despite some historical fishing in the region, the seamounts are relatively pristine and represent an excellent conservation opportunity to protect a global biodiversity hotspot before it is degraded. One obstacle to effective spatial management of the ridges is the scarcity of direct observations in deeper waters throughout the region and an accompanying understanding of the distribution of key taxa. Species distribution models are increasingly used tools to quantify the distributions of species in data-poor environments. Here, we focused on modeling the distribution of demosponges, glass sponges, and stony corals, three foundation taxa that support large assemblages of associated fauna through their creation of complex habitat structures. Models were constructed at a 1 km² resolution using presence and pseudoabsence data, dissolved oxygen, nitrate, phosphate, silicate, aragonite saturation state, and several measures of seafloor topography. Highly suitable habitat for each taxa was predicted to occur throughout the Salas y Gómez and Nazca ridges, with the most suitable habitat occurring in small patches on large terrain features such as seamounts, guyots, ridges, and escarpments. Determining the spatial distribution of these three taxa is a critical first step towards supporting the improved spatial management of the region. While the total area of highly suitable habitat was small, our results showed that nearly all of the seamounts in this region provide suitable habitats for deep-water corals and sponges and should therefore be protected from exploitation using the best available conservation measures.

Geo-referenced coral and sponge records were obtained from the Ocean Biodiversity Information System (OBIS 2020) the NOAA Deep-Sea Coral and Sponge Database (NOAA 2020), and supplemented with records from recent expeditions to the area (J López and E Easton, 2020, unpublished data). All records were obtained as presence-only records, with duplicate records removed prior to analysis. The bulk of records were focused on the Salas y Gómez and Nazca ridges, with another cluster of records in the neighboring Juan Fernández Islands region. We chose to focus on three higher taxonomic groupings that are often key foundation species on seamounts: stony corals (Order Scleractinia, n=233), demosponges (Order Demospongiae, n=275), and glass sponges (Order Hexactinellida, n=134).

Within the study area, a suite of 44 environmental variables known to influence the distribution of corals and sponges were constructed for use in models (Table 1). Bathymetry for the study area were obtained from the SRTM30+ dataset (Becker et al. 2009; Sandwell et al. 2014) at a native resolution of 0.0083° (approximately 1 km) and used in the creation of several additional layers.

A number of terrain metrics were derived from this bathymetry layer to define the shape of the seafloor. Slope, roughness, aspect, general curvature, cross-sectional curvature, and longitudinal curvature were calculated using the ArcGIS (v10.8, ESRI) toolkit ‘DEM Surface Tools’ (v2; Jenness 2004, Jenness 2013a). Slope was measured in degrees and calculated using the 4-cell method (Jones 1998). Aspect represents the compass direction of the steepest slope and was converted to an index of eastness using a sine transformation and an index of northness using a sine transformation. Curvature metrics assess the likely flow of water across a feature, with positive values generally indicating convex features that cause water to accelerate and diverge, in contrast to concave features where water would be expected to decelerate and converge. Roughness is a measure of topographical complexity, calculated here as the ratio of surface area to planimetric area, with more positive values indicating more complex terrain. The Topographic Position Index (TPI) was calculated using the toolkit Land Facet Corridor Designer (v1.2; Jenness et al. 2013b). TPI assesses the relative height of features compared to the surrounding seafloor, with positive areas indicating locally elevated features and negative values indicating depressions. TPI is scale dependent, and was calculated at scales of 1,000 m, 5,000 m, 10,000 m, 20,000 m, 30,000 m, 40,000 m, and 50,000 m. Finally, the Vector Ruggedness Measure (Hobson 1972; Sappington et al. 2007), which calculates terrain heterogeneity, was calculated with a neighborhood size of 3, 5, 7, 9, 15, 17, and 21 using the Benthic Terrain Modeler (v3.0; Walbridge et al. 2018).  

To complement the suite of terrain metrics, large-scale geomorphological features expected to provide suitable habitat for corals and sponges were obtained from Harris et al. (2014), including seamounts, guyots, canyons, ridges, spreading ridges, plateaus and escarpments. We applied an algorithmic approach to delineate seafloor features based on a modified SRTM+ bathymetry dataset, generating a total 29 geomorphic categories. See Supplemental Figure 21 for a map of geomorphological features in the study area.

Data describing benthic conditions at the seafloor were obtained from the World Ocean Atlas (v2; 2013), including temperature, dissolved oxygen, salinity, nitrates, and phosphates. Carbonate data including dissolved inorganic carbon (DIC), total alkalinity, and the saturation states of calcite and aragonite, were obtained from Steinacher et al. (2009). Current data describing regional horizontal and vertical current velocities were obtained from the Simple Ocean Data Assimilation model (SODA v3.4.1; Carton et al. 2008). Particulate organic carbon (POC) flux to the seafloor was obtained from Lutz et al. (2007). Raw benthic data layers were transformed to match the extent and resolution of the other environmental variables using an upscaling approach that incorporates bathymetry data to approximate conditions at the seafloor (sensu Davies and Guinotte 2011). This upscaling technique has previously been demonstrated to work effectively on both global and regional scales for a variety of data (Yesson et al. 2012; Georgian et al. 2019).

Surface conditions were assessed as chlorophyll a and mean sea surface temperature data obtained from the Aqua MODIS program (Aqua MODIS 2018). Both layers were calculated as the mean value from 2002-2016 at a native resolution of 4 km, and resampled to match the extent and resolution of the other environmental layers with no additional interpolation. 

Taxonomic data are presented as XX (longitude), YY (latitude), and either 0 (psuedoabsence) or 1 (occurrence) for each taxon (stony coral, demosponge, or glass sponge)

Environmental data are ascii rasters in the geographic coordinate system WGS84 with a grid size of 0.008333333333.