Whether the three-dimensional (3D) structure of habitats influences and partition recruitment niches of corals is unknown. We developed a new method that combined Species Distribution Modeling and Structure from Motion to characterize and map the three-dimensional recruitment niches of two ecosystem engineers on Caribbean coral reefs, scleractinian corals and octocorals. In this repository, we include 48 3D models of small areas of the reef (i.e., within ~ 0.25 m2 quadrats) reconstructed with Structure-from-Motion, as well as the geospatial data used to characterize and map the realized recruitment niche for scleractinian corals and octocorals on Caribbean coral reefs. We conducted the study at two shallow, fringing reefs off the south shore of St. John, US Virgin Islands, named Grootpan and Europa Bays (18° 18.360’N, 64° 43.140’W, and 18° 19.016’N, 64° 43.798’W, respectively). Within each 0.25 m2 quadrat, we counted and marked all recruits (octocorals ≤ 5 cm height, and scleractinians ≤ 4 cm wide). DATASET DESCRIPTIONS: "Quadname_data.zip": In each of this folders we included all the data calculated within a quadrat: ASCII files (.txt). The annotated dense point cloud (.las) for each quadrat. The quadrat 3D model texture (.jpg). The quadrat 3D polygon mesh (.ply). The quadrat 2.5D Digital Elevation Model (i.e., DEM; .tif). Shape files with recruits local coordinates within each quadrat (.dbf, .prj, .shp, .shx). "datawide.rds": This is the file needed to run the analyses performed in Martínez-Quintana et al., 2023. This file is obtained after processing all the raw data calculated within each quadrat. All code associated with the workflow used to obtain the datawide.rds file and run the analyses performed in Martínez-Quintana et al., 2023 is available at github.com/AdamWilsonLab/meshSDM. IMPORTANT NOTES: Quadrat names starting with the letters “eu” indicate the data were collected at Europa Bay, whereas those starting with the letters “ec” indicate that data were collected at Grootpan Bay (commonly named East Cabritte). Each ASCII file (quadname_ASCII_subsampled_X.txt) contains the slope and roughness of the quadrat calculated on the point cloud at 5, 10, 20, and 100 mm scales, and the smooth point cloud used to calculate the topographic exposure index (TEI) described in Martínez-Quintana et al., 2023. Calculations were performed and ASCII files were created with CloudCompare. Each dense point cloud, mesh, texture, and DEM were calculated with Agisoft Metashape. Agisoft Metashape allows the user to classify and annotate groups of points in the dense point cloud. However, the list of classes provided by the software corresponds to the standard list used for terrestrial LiDAR data; these classes cannot be renamed within the software. Thus, for the present study, we coded the automatic semantic classifications available in Metashape as follows: Ground = Calcareous rock. Building = Igneous rock. High noise = Sand. Low vegetation = Adult Scleractinian corals. Medium vegetation = Adult Octocoral base. High vegetation = Sponge. Water = Octocoral recruit (named also ocr). Road Surface = Scleractinian recruit (named also scr). Unclassified = created points but never classified (excluded from the analyses). Low Point = noise (unreliable points). Transmission tower and Rail = Points outside the quadrat and excluded from the analysis.

Support for this work was provided by The National Science Foundation grants: OCE-1350146, OCE-1756381, and OCE-1801475