Dataset: Mean percentage cover of corals and Porites astreoides at each site by year at St. John, VI from 1992 to 2019

Coral abundance was evaluated at multiple sites from 1992–2019, with different variables quantified at combinations of sites and years. The abundance of Porites astreoides was measured using photoquadrats recorded annually in May (1992, 1995, 1997), June (1993), or July/August (all other years). Photoquadrats (0.5 × 0.5 m) were recorded at six sites at 7–9 m depth between Cabritte Horn and White Point, and were placed along permanently marked transects (20 m from 1992–1999 and 40 m from 2000–2019) at positions that were randomized annually. The six sites were randomly chosen in 1992 to sample fringing reefs between White Point and Cabritte Horn. Images were recorded with cameras using 35-mm film (which allowed 18 photoquadrats to be recorded at each of two sites with one roll of film) from 1992–1999 (and subsequently digitized), and digital images thereafter, with a resolution of 3.34 MP from 2000–2006, 6.1 MP from 2006–2010, 12.1 MP in 2011, 16.2 MP from 2012–2015, and 36.3 MP from 2016–present. Cameras were mounted on a framer and fitted with two strobes, and the images were used to quantify the size and abundance of P. astreoides, and the percentage cover of coral (pooled among taxa and also for P. astreoides). There were ~ 18 photoquadrats site-1 from 1992–1999, and ~ 40 photoquadrats site-1 thereafter.

The analysis by colony focused on the yellow morph of Porites astreoides, and colonies were manually counted and measured using ImageJ software. The images were outsourced to students to measure colony sizes, and all results were screened for accuracy by the first author. Screening involved randomly selecting quadrats from each site and year combination, and repeating the analyses to evaluate precision. Where mismatches were detected, the set of photoquadrats was analyzed a second time by more highly trained research assistants. Colony sizes were determined from the mean of two planar diameters (± 0.1 cm), scaled from the size of the quadrat. Colonies were measured if they were enclosed by the quadrats, or if more than half of the roughly circular colonies was within the quadrat and their size could be estimated from a single diameter assuming they were round. The same photoquadrats were used to measure percentage coral cover using CoralNet software with 200 dots randomly located on each image, and dots manually annotated to quantify all stony coral, and Porites spp. separately. Porites spp. included both massive (P. astreoides) and branching forms (P. porites, P. furcata, and P. divaricata) because all corals were uncommon on these reefs, and because juvenile colonies of branching forms are challenging to discern from massive forms before a branching morphology is established.

To provide context to the analysis of colony density and size for P. astreoides, the demographic supply of colonies was evaluated using in situ surveys for juvenile colonies (≤ 4-cm diameter) at 5–6 adjacent sites at 5–9-m depth, and coral recruits using settlement tiles at 5-m depth at five other sites (Fig. S1); all sites were close to one another along ~ 4.4 km of shore. Surveys for juvenile colonies were completed along one 40-m transects at each site, with five sites surveyed in 1994 and 1996–1998, two sites in 1995, and six sites in all other year. Forty quadrats (0.5 × 0.5 m) were placed at random locations along each transect for the enumeration of juvenile colonies. Sites were treated as statistical replicates, and small colonies of P. astreoides were assumed to be juvenile based on the size at which sexual maturity is reached for this species. Recruitment of Porites spp. was measured using unglazed terracotta settlement tiles (15 × 15 × 1 cm) secured with their rough surface facing down in clusters of 15 at each of five sites. Sample sizes of tiles were reduced in 2018 as a result of Hurricanes Irma and Maria in 2017. Tiles were individually attached to the reef using a stainless steel stud through their center, which held them approximately horizontal with a 1 cm gap beneath that is favored for coral recruitment (Mundy 2000). Tiles were seasoned in seawater for ~ 6–12 months prior to installation to develop a biofilm, and the first tiles were installed in August 2006. For the first two years, tiles were replaced in January and July/August, but from 2008 they were exchanged annually in July/August.

Freshly collected tiles were soaked in bleach to remove organic material, rinsed in fresh water, dried, and then inspected for coral recruits using a dissecting microscope (40 x magnification). Recruits were identified to family (results for Poritidae are presented here) and densities were scored by tile (sum of recruits on the top, bottom, and sides). The two samplings per year in the first two years were summed to estimate annual recruitment by site, and in all years, mean annual recruitment (± SE) was estimated using sites as replicates (n = 5). Following scoring, tiles were cleaned in dilute HCl, rinsed, and stored in seawater beneath the lab dock for seasoning until use the following year.

Changes in the density and size of colonies of Porites astreoides among years were graphically displayed using scatter plots in which sites were the replicates. Changes over time in density and size were evaluated using repeated measures (RM) PERMANOVA in which time was the RM factor and mean values by site and year were replicates. Least squares linear regression was used to test for trends in variation in colony density and size over time, in both cases using mean values by year using sites as replicates. The same statistical procedures were applied to coral cover, density of juvenile colonies, and recruitment. The assumptions of the statistical procedures were tested through graphical analyses of residuals, and all statistics were conducted using Systat 13.0 software.