Carbonate Budgets, Structure-from-Motion Products, and Topographic Complexity Measurements From Restored and Non-Restored Areas of Coral Reefs in the Lower Florida Keys

Study locations: Surveys were conducted at eight offshore reefs and three patch-reef sites in the Lower Florida Keys where Mote Marine Laboratory researchers have been actively restoring coral populations in recent years. The reefs at Looe Key, Summerland Ledges, American Shoal, and Mote Site C were surveyed in the summer of 2022 (July 11-18) and the sites at Sand Key, Rock Key, Eastern Dry Rocks, Marker 32, Dog's Leg, Cat's Paw, and Cook Island were surveyed in the summer of 2023 (July 11-18). Restored and non-restored areas of the reef were surveyed at each site. For restored transects, SCUBA (self-contained underwater breathing apparatus) divers identified areas of the reef with a high density of outplanted corals and placed transects within those areas. Non-restored transects were placed on nearby areas of the reef (adjacent spurs or ledges for offshore reef sites) where no outplants were present.

Photographic surveys for Structure-from-Motion (SfM): At each site, except for Summerland Ledges, 10–12 (approximately 10 x 2 meters [m] in size) photographic transects were surveyed by SCUBA divers using a downward facing Canon Powershot S120 camera in an underwater housing set to collect RAW image format in continuous shoot mode. Summerland Ledges is an experimental restoration site consisting of two 10 x 10 m restored area and one 10 x 10 m control plot. The two restored plots at Summerland Ledges were surveyed 2m from the reef base using a dual Nikon D7000 DSLR camera system and the control plot was surveyed using a Canon EOS R at the surface. The number of transects varied per site depending on the availability of restored or control reef. All images were collected 1-2m from the reef base using a double-lawnmower swim pattern ensuring 70-80% forward and lateral overlap between images. Prior to image acquisition, the diver used a metric field tape to delineate the sampling area then placed 3-4 coded 25-centimeter (cm) scalebar targets evenly throughout the transect areas to provide accurate scale for models. For more information about the photographic surveys and access to the associated imagery, refer to Johnson and others (2025).

Structure-from-motion data processing: SfM data products (point clouds, digital surface models, and orthomosaics) were generated using Agisoft Metasahpe Pro v2.0 or later generally following the protocols of Bayley and Mogg (2020). For a thorough description of the workflow and settings, refer to Toth and others (2025) supplemental information section. Briefly, the images from each transect were checked for proper orientation then uploaded into Metashape. Three-dimensional (3D) Models were then generated using the following basic steps: 1) align images into a sparse point cloud, 2) scale model and run initial bundle adjustment for lens calibration, 3) reduce number of low quality points in the model to improve accuracy, 4) orient and position the model in local coordinates, 5) generate dense 3D point cloud, 6) segment point cloud into classes: noise, reef base, canopy, and outplants, 7) filter point cloud based on segmentation classes to generate digital surface models (DSMs), 8) render orthomosaics and 9) export all SfM data products. The open-source R Multiscale DTM package (Ilich and others, 2023) was used to calculate structural complexity metrics from 1-centimeter (cm) resolution exports of the DSMs. The package calculates 15 metrics encompassing terrain attributes from five common terrain groups: slope, aspect, curvature, relative position, and roughness. For this study, four included Multiscale DTM roughness metrics were extracted with a 5x5 neighborhood window size: adjusted standard deviation, roughness index, and vector ruggedness. These metrics were found to be highly correlated so the analysis focused on vector ruggedness (VRM). The impact of restoration on average reef elevation (relative to the lowest point of each transect model) and overall surface area-to-planar-area ratio (rugosity) of the transects was also evaluated. SfM data are published in individual zipped folders, per study site (the eight offshore reefs and three patch-reef sites) and organized into sub-folders per data product: DEMs (Tagged Image File format, .tif), orthomosaics (.tif), and point clouds (LASer file, .las).

Quantifying percent cover of reef benthos: Percent cover of calcifying reef taxa (corals and crustose coralline algae) and other benthos were quantified by conducting point-count analysis of a 10 x 1 meter (m) belt transect within the two-dimensional (2D) orthomosaics described above using the online software, CoralNet (https://coralnet.ucsd.edu/). Briefly, 10, non-overlapping 1 x 1 m images were extracted from each orthomosaic and the benthos beneath each of 150 points were identified. All coral taxa were identified to species, except for Orbicella, Pseudodiploria, and Millepora spp. which were later pooled by genera. Other benthos, including crustose coralline algae, macroalgae, sponges, gorgonians, zoanthids, consolidated bare substrate (which was generally covered in turf algae), and unconsolidated substrate (sand and rubble) were identified categorically.

Bioerosion census surveys: USGS researchers conducted census surveys of bioeroding parrotfish, sponges, and urchins following the ReefBudget v2 protocol (Perry and Lange, 2019) in the same reef areas where the structure-from-motion surveys were conducted. Divers recorded the number and size of bioeroding sponges (Cliona aprica, C. caribbaea, C. tenuis, C. varians, C. delitrix, and Siphonodictyon coralliphagum) and bioeroding urchins (Diadema antillarum, Echinometra lucunter, Ec. viridis, and Eucidaris tribuloides) along the same 10-m transects surveyed for structure-from-motion. Sponge surveys were conducted within a 1-m belts around the transects and urchin surveys were conducted within 2-m belts. Researchers also recorded the species, size, and life phase of bioeroding parrotfish (Sparisoma viride, Sp. aurofrenatum, Sp. rubripinne, Sp. chrysopterum, Scarus vetula, Sc. taeniopterus, Sc. iseri, Sc. guacamaia, Sc. coeruleus, and Sc. coelestinus) within 8–10 larger (25 x 4 m) belt-transects to estimate site-level parrotfish bioerosion. Note that because of the larger size of the parrotfish transects and their highly mobile nature, parrotfish surveys could not be conducted separately in restored and non-restored areas of the reefs.

Calculating coral reef carbonate budgets: To calculate gross carbonate production (in kilograms per meter squared per year [kg m-2 y-1]), the percent cover of each reef calcifier (from the point count analysis) was multiplied by area-normalized taxon-specific calcification rates (Courtney and others, 2024) and summed for each transect. Microbioerosion along each transect was estimated by multiplying the total area of consolidated, non-calcifying reef substrate (from the point-count analysis) by the western Atlantic mean microbioerosion rate of 0.24 kg m-2 y-1 (Perry and Lange, 2019). Bioerosion by parrotfish, sponges, and urchins were estimated by multiplying by the species-, size-, and, for parrotfish, life-phase-specific bioerosion rates suggested in ReefBudget v2 (Perry and Lange, 2019). Those data were summed with microbioerosion to estimate total bioerosion. Total bioerosion was subtracted from gross carbonate production to estimate net carbonate production. Reef-accretion potential was estimated according to the following equation (Kinsey, 1985): Reef-accretion potential=(Net carbonate production)/(p(1-porosity)), where p is the density of calcium carbonate (2.9 grams per cubic centimeter; Kinsey, 1985) and porosity was estimated from cores of geologic reef framework in the Florida Keys built by branching or massive corals (for offshore and patch-reef sites, respectively; Toth and others, 2018).