The following methodology was extracted from Edumunds et al. (2020), figure and table references below correspond to that publication. This methodology section applies to this dataset in addition to other datasets published in Edmunds et al. (2020).

Sponge abundance was determined using photoquadrats recorded annually since 1992, but here they were sub-sampled by years scattered uniformly across the 26 year study (described below). These photoquadrats are part of a time-series that was designed to quantify the coral community (Edmunds 2013, 2019). The project consists of six sites on hard substrata at 7–9-m depth that were randomly selected in 1992, permanently marked, and recorded using photoquadrats (0.5 × 0.5 m) positioned randomly along a single transect at each site. Photoquadrats were recorded in May 1992, June 1993, August 1994, May 1995-1997, and July or August thereafter. Prior to 2000, ~ 18 photoquadrats site-1 were recorded annually using 35-mm film (Kodachrome 64), but from 2000, ~ 40 photoquadrats site-1 have been recorded using digital cameras ranging in resolution from 3.3 MP to 36.3 MP. Cameras were fitted with two strobes (Nikonos, SB 105), and color slides were scanned (4000 dpi) for analysis.

Definitive identification of Caribbean sponges requires analysis of spicules and other skeletal characteristics (Rützler 1978). As this was impossible with photoquadrats, a consensus set of 23 sponges was identified using expert opinion. This set resolved 19 species, 3 sets of congeners that could not be distinguished to species, and an unknown category (Table S1). Most sponges were identified to species (66%, n = 9,608 pooled over the whole study), 4% were identified to genus, and 31% were assigned to the “unknown” category. Surveys on the adjacent island of St. Thomas suggested the total sponge diversity includes over 100 species (Gochfeld et al. 2020). Between 100 and 300 sponge species is typical for a localized area in which multiple sites are combined (Wulff 2016), with individual Caribbean sites being characterized by 51–67 species (Villamizar et al. 2013, Wulff 2006b, 2013).

To evaluate sponge density, photoquadrats were opened in Adobe Photoshop CS5.1 software, and the number of sponges counted. Individuals were defined by their contiguous areas of sponge biomass, which were separated from other sponges of the same taxon. This method potentially overestimates sponge density where algae, other taxa, or topographically complex surfaces obscured connections among pieces of a contiguous sponge. The limitations of planar photographs in quantifying benthic communities have been inherent in the method since it was first used in modern ecology (Loya 1972). These issues are unavoidable with planar images, but their magnitude depends on the quantity of macroalgae and the rugosity of the benthic surfaces. With regard to rugosity, the study sites were located on carbonate pavement (RS11, 5), or igneous substrata (RS15, 2), or a combination of the two (RS9, 6) (Fig. 1), which provided relatively smooth surfaces with limited ability to obscure sponges (Fig. 2). Analyses of rugosity at Europa Bay (RS11) and East Cabritte (~ 500 m from RS2) in 2014 revealed that the mean (± SE) topographic complexity at ~ 8-m depth was 1.16 ± 0.02 and 1.22 ± 0.02, respectively (Tsounis et al. 2018). The values for topographic complexity from St. John describe relatively flat communities (Fig. 2) that correspond to 40 year minimal values across the Caribbean, where rugosity declined from ~ 2.5 in 1969 to 1.2 in 2008 (over a depth range from the surface to > 20 m) (Alvarez-Filip et al. 2009).

Evaluating the ecological importance of sponge density (i.e., sponges per area) requires measurements of sponge biomass (Wulff 2001, 2016), which are unobtainable from planar photographs. However, sponge biomass can be estimated from sponge volume calculated from linear dimensions and the volumetric formulae of geometric shapes matching those of the sponges (after Wulff 2001). We sought approximations of sponge volumes using photoquadrats, assuming that the shape of each sponge could be inferred from a planar image and volumetrically approximated by geometric shapes (i.e., rods, disks, and spheres). The volumes of these shapes were estimated using linear dimensions obtained from the photographs using ImageJ software (Abramoff et al. 2004). In the few cases where sponges were conical (e.g., Ircinia campana), it was not possible to accurately estimate their volume and they were excluded from the analysis. As I. campana accounted for only 1% of all sponges, the bias attributed to this affect was trivial. The volume of encrusting sponges was estimated assuming they were disks 1.7-mm thick (based on mean thickness of 1.7 ± 0.2 mm (± SE, n = 23) of the encrusting sponges Chondrilla caribensis forma hermatypica, Spirastrella coccinea, S. hartmani, Clathria venosa, Placospongia cf. intermedia, Acarnus nicoleae, and Cliona caribbaea selected at random for measurement in shallow water on the reefs of Belize during May 2019). Sponge volumes were summed by genus within each quadrat to provide a single replicate measure.

To measure sponge density and volume in the ~ 5,000 photoquadrats from 1992 to 2017, the 26 year record was analyzed in 2–3 year intervals. All six sites were analyzed for sponge density, but three sites (White Point, Cabritte Horn, and Europa Bay) were selected to estimate sponge volume. The sites for volume estimates were selected to span the range of sponge abundances observed along this shore, and to provide a tractable task commensurate with the limited capacity to estimate organism volume from planar images.

Physical environmental conditions

As part of the ecological monitoring in this location (Edmunds 2013; Edmunds and Lasker 2016), seawater temperature and rainfall have been recorded since 1989. The records from 1990 to 2017 were used to explore their capacity to account for variation in multivariate sponge assemblage structure, as well as the multivariate community structure defined by sponges, scleractinians, octocorals, macroalgae, and CTB. Seawater temperature was recorded at ~ 9–14 m depth at Yawzi Point using Hobo loggers (± 0.2°C [U22-001, Onset Computer Co., MA]) that sampled every 10–15 minutes. Records were collapsed by day and used to calculate annual summaries of mean temperature, minimum temperature, maximum temperature, the number of hot days (i.e., > 29.3°C), and the number of cold days (i.e., ≤ 26.0°C). From 1992 to 2011, rainfall (cm y-1) was obtained from the Southeastern Regional Climate Center (http://www.sercc.com), which compiled data from a rain gauge in Cruz Bay (Station 671980). Where this record was incomplete, values were obtained from Catherinburg (Station 671348), East End (Station 672551), or a mean value for the missing months calculated from all other values for that same month (in Edmunds and Gray 2014). From 2012, rainfall was measured using a Standard Rain Gauge (NOAA, National Weather Service) deployed on the north shore of St. John (18° 21´ 20.95N, 64° 45´ 57.53W). Temporal variation in the regional-scale physical environment was evaluated through a de-trended index reflecting the effect of the Atlantic Multidecadal Oscillation (AMO) as reported in Kajtar et al. (2019) and provided courtesy of the first author. Hurricane effects were evaluated on a categorical scale assigning impact values to storms of 0 (no storm), 0.5 (minor impacts), and 1 (major impacts), and summing impacts within each year (after Gross and Edmunds 2015).

Blank values in this dataset indicate "not applicable."

edmunds2020_fig3.csv