Methodology from Tsounis, G., Edmunds, P.J., Bramanti, L. et al. Mar Biol (2018) 165: 29. https://doi.org/10.1007/s00227-018-3286-2

Surveys were conducted at East Cabritte and Europa Bay (Fig. 1) in July and August 2014, and March 2015. The sites represent contrasting exposure regimes, suggested by the exposure of East Cabritte to prevailing winds and swells, and the shelter of Europa Bay in the lee of Cabritte Horn. Measurements of physical environmental conditions (described below) were used to quantify these differences.

At each site, a 50 × 10 m study area was haphazardly established, within which aspects of the biota and the physical environment were measured. The long axes of the areas were parallel to the shore and ranged from 7.5- to 9.0-m depth at East Cabritte, and from 5.6- to 8.0-m depth at Europa Bay, and the short axes were perpendicular to the shore and crossed a depth gradient of 5.6–7.2 m at Europa Bay, and 7.5–9.0 m at East Cabritte. The depth ranges of the study sites differed, because the reef at Europa did not extend into deeper water, while the reef at East Cabritte markedly steepened above 7 m. Five, 10-m transects were equally spaced along, and perpendicular to, the long axis of each study area. Octocoral community structure was compared between sites using octocoral diversity, size, and density in both multivariate and univariate statistics frameworks.

Physical environmental condition
Environmental conditions at each site were characterized in the summer of 2014 and the winter of 2014–2015 through measurements of water motion, benthic rugosity, sedimentation, and light intensity. Water motion was characterized using two methods, first, using the wave climate recorded by an NOAA buoy moored 7.8 km from the study site (CariCOOS Data Buoy C at Mooring VI-105), and second, through direct measurements of integrated water motion using clod cards (Doty 1971).

Hourly wave direction (degrees relative to north) from March 15th 2011, 17:00 h to March 2nd 2015, 21:00 h was obtained from the NOAA buoy VI-105 (http://www.caricoos.org/drupal/virgin_islands), with measurements averaged by hour from a sampling frequency of 2 Hz in 17 min bursts. To obtain hourly averages, a varying number of records were averaged depending on the coincidence of the 17-min sampling bursts with the 60-min averaging period. Using hourly averages, the proportion of time (i.e., percentages based on number of hours) when waves directly impacted each site was calculated based on the direction from which the waves originated. The two sites were impacted by waves originating from dissimilar, but partially overlapping directions, because the sites differed in orientation and location along the shore relative to the southerly projection of Cabritte Horn (Fig. 1). Europa Bay is exposed to waves from 135° to 250°, and East Cabritte to waves from 60° to 135°. To capture these effects, the number of hours describing mean wave directions corresponding to each of these directional bins was quantified, without considering wave refraction around Cabritte Horn. Wave height was not evaluated using data from this buoy, as its distance from our study sites made estimates of wave height unreliable.

Integrated water motion was measured in situ using clod cards (Doty 1971) that were prepared in a single batch for each deployment, dried to a constant weight at 50 °C, and weighed prior to use. Clod cards had similar initial weights [128 ± 2 g (mean ± SE, n = 78)], and were deployed in July and August 2014, and March 2015, and assigned to each site in a paired design (two clods per site). Clods were secured for 24–48 h to posts ~ 30 cm above the benthos at 9-m depth adjacent to, but outside of, the octocoral canopy. Following deployment, clods were dried to a constant weight at 50 °C, and integrated water motion was evaluated from the dissolution of plaster in units of g day−1.

Sedimentation was measured with sediment traps in two deployments for 8 and 9 days in 2014 (to begin a new measurement when a storm at the end of the first deployment saturated the traps), and in a single deployment for 12 day in 2015. Both sites were monitored simultaneously. The traps consisted of PVC tubes (20 × 5 cm ID) that were deployed 60 cm above the benthos (Edmunds and Gray 2014). Traps were capped in situ, returned to the lab, and filtered through pre-weighed filters (Whatman #113). Filters and sediment were rinsed with freshwater to remove salt, dried to a constant weight at 50 °C, and weighed (± 1 mg). Sedimentation was normalized by catchment area of the traps, and time (mg cm−2 day−1).

In situ light intensity was measured using two integrating submersible light meters (JFE-Advantech Compact-LW) fitted with a cosine-corrected collector sensitive to photosynthetically active radiation (PAR, 400–700 nm) and a wiper blade that cleaned the collector prior to each measurement. The meters were deployed in a paired design at the two sites for 8 days in 2014 (August 10–15th and August 18–19th) and 8 days in 2015 (March 3–13th). Each meter was attached to a post at 9-m depth adjacent to the octocoral community, but ~ 5 m from the nearest octocoral colony to avoid shading. Light intensity was recorded at 0.033 Hz, and data were used to generate two dependent variables, one recording the maximum daily intensity (μmol m−2 s−1) and the other recording the intensity integrated over each 24-h period (units of mol m−2 day−1).

Benthic rugosity was determined along the five transects at each study plot using a light chain (10-mm links) which was laid along each transect to conform to the reef surface. Rugosity was calculated as the quotient of the linear distance and the conformed length of the chain (Luckhurst and Luckhurst 1978).

The hypothesis that the sites differed in environmental parameters was tested with univariate ANOVA using R (R Development Core Team 2008). Sediment traps and clods cards were not deployed at both sites in synchronous deployments due to logistical constraints, and these data were compared between sites and times using a two way, Model I ANOVA. Light intensity differs among days, and, therefore, was compared between sites using a within-subject design in the aov function in R, accounting for variation over time by considering deployment day as a blocking factor. Substratum rugosity was compared between sites using one-way ANOVA. In all cases, the ANOVA assumptions of normality and homoscedasticity were tested through graphical analyses of residuals.

Octocoral community structure

Species richness:
Octocoral species richness was compared between sites based on 50 quadrats (1 × 1 m) that were sequentially placed along the five, 10 m transects that crossed the short axis of the study plots, and censused for octocoral presence. Surveys began in July and August 2014, and were concluded in February and March 2015 (i.e., two field trips were required). Octocoral diversity was determined using Pielou’s Evenness Index (J´) (Pielou 1966), and the Shannon–Wiener Diversity index, H´ (Shannon 1948). This study considered adult octocorals, and excluded recruits (i.e., colonies ≤ 5 cm tall [HR Lasker, unpublished data]) from the surveys. However colonies ≤ 5 cm were censused if it was obvious that they had been larger adults that were reduced in size by predators. Octocorals were identified to the lowest taxonomic-level possible, as determined through voucher samples that were microscopically inspected for sclerites (after Bayer 1961). Preliminary sampling revealed 10 genera and 35 species at the two sites, but a small number (< 1.6%, n = 1290 colonies) could not be identified to species and were scored by genus (mostly Eunicea and Pseudoplexaura). Initial work indicated 39 nominal species (Edmunds and Lasker 2016), though subsequent analysis refined the species count to 35 (this study). We do, however, highlight the fact that the distinctions between Pseudoplexaura wagenaari and P. flagellosa, those between Plexaurella dichotoma and P. fusifera and and those between Eunicea laxispica, Eunicea mammosa and Eunicea succinea are difficult to make, especially in the field. For this study, we opted to distinguish between these species in our analyses, based on the best information available (spicule analysis), but acknowledge that this might not always be feasible in future studies, where pooling these pairs will facilitate consistent long-term data series analyses using multiple observers. Rarefaction curves (sensu Coleman et al. 1982) were used to evaluate the efficacy of the sampling regime (i.e., number of 1 m2 quadrats) in quantifying octocoral species abundance. At each site, the number of species as a function of sample size (number of quadrats) was analyzed using the specaccum option in the vegan package (version 2.3.2) for R [R Development Core Team 2008 (Oksanen et al. 2015)], and species abundance was evaluated by the asymptote of the curves against sample size.

Colony abundance:
To compare community structure of octocoral colonies between sites, we randomly subsampled 32 of the 50 quadrats (each 1 × 1 m) along the transects (described above) to remove the biases associated with uniform sampling (Sokal and Rohlf 1995). Densities (colonies m−2) by species were log(x) transformed and used to compute Bray–Curtis dissimilarity indices after applying a dummy value (+ 1) to account for paired observations of zero (Clarke et al. 2006). Dissimilarity indices were compared between sites using a one factor PERMANOVA with 999 permutations. Dissimilarity indices were produced using the vegdist function, and PERMANOVA was performed using the ADONIS function, both in the vegan package (version 2.3.2) for R [R Development Core Team 2008 (Oksanen et al. 2015)]. A similarity percentage analysis (SIMPER, Clarke 1993) was performed using the simper function in the vegan package (version 2.3.2) for R, and used to assess the contribution of individual species to the total dissimilarity between sites. Spatial variation in multivariate community structure was visualized using ordination plots generated by non-metric dimensional scaling (NMDS) that were based on Bray–Curtis dissimilarities (using the Vegan package in R).

Colony size:
The colony size–frequency distributions of the three most common octocorals that could be identified in the field (Antillogorgia americana, Eunicea flexuosa, and Gorgonia ventalina) were compared between sites. Colony heights were surveyed using 1-m-wide belt transects placed along the five transects dividing the study plots. Colonies were measured as encountered within these survey areas, with the objective of measuring 75–100 colonies of each species for each size class at each site. When too few colonies were found to meet the target sample size, additional non-overlapping belt transects were censused within the study plot to reach the target number of colonies. To test for differences in colony sizes for the three species between sites, one-way PERMANOVA with 999 permutations were performed (Anderson 2001) using the Adonis function in the vegan package (version 2.3.2) for the R software [R Development Core Team 2008 (Oksanen et al. 2015)]. Two-sample Kolmogorov–Smirnov tests using the R software were performed to compare the complete size–frequency distributions for each species between sites.

Community structure resolved by genus versus by species:
To evaluate the effect of taxonomic resolution on the differences in community structure detected between sites, multivariate analyses were conducted with genus- and species resolution, and the contribution of each genus or species (respectively) to total dissimilarity between sites was resolved using SIMPER.