Dataset: Millepora features

This study was undertaken at Cabritte Horn on the south shore of St. John, US Virgin Islands across three decades. 

Photographic quadrat sampling
Photographic sampling of quadrats (0.5 meters x 0.5 meters) was performed annually from 1992 to 2021 using cameras mounted on a framer at a fixed height above the reef.  Color slide film was used from 1992–2000 (and digitized at 4000 dpi), with digital photography implemented in 2001. Photoquadrats were recorded at random positions along the transect, which was 20 meters long from 1992–1999 and 40 meters long from 2000 onward when the sample size was increased to ~40 photoquadrats per year.   

The photoquadrats were used to quantify Millepora abundance using ImageJ software (Abràmoff et al., 2004), and taxa were resolved to the M. alcicornis/M. complanata complex (hereafter described as M. alcicornis). As 99.8% of colonies were M. alcicornis, all colonies were considered to be a single species and described as Millepora sp. Areas of encrusting Millepora sp. were located in each photoquadrat, and outlined and measured for area with all pieces of encrusting sheets quantified separately. Because Millepora species’ sheet forms meander across the substratum and are prone to fission, it was not always clear where colonies began or ended within photoquadrats. Colonies therefore were defined as areas of autonomous tissue, so portions of colonies that were partially within the photoquadrats were scored as separate colonies (although such cases were relatively uncommon).

To access images, see Related Publications section below for Coral Image Downloads and Viewer link (Cabritte Horn images under "Random Sites")   

Morphology
Branches on each sheet were quantified as the number of proximal “roots” where they attached to the encrusting sheet, and as the number of distal growing points on each root. Branch fragments lying on the benthos were not counted. The size of Millepora sp. colonies was determined from the mean planar area of autonomous portions of sheets (cm²), and roots and growing points were normalized to the sheet (Roots 100 cm² and Growing Points 100 cm²) and colony (Roots per colony and growing points per colony); growing points were also expressed per root (Growing Points per root). The area of sheets were summed by quadrat and used to calculate the percentage cover of Millepora sp. (%).  Occasionally Millepora sp. was found encrusting octocoral colonies on which they appeared as long branches with miniscule sites of basal attachment; such colonies were excluded from analyses of roots and growing points.

The capacity of Millepora species to exploit a sheet-tree morphology was quantified through the quotient of roots and growing points to the area of the sheet (Roots 100 cm-2 and Growing Points 100 cm-2). Quotients quantified the exploitation of  “trees” relative to “sheets”, but they have the limitation of not being able to distinguish between effects caused by the growth of new roots or growing points versus changes in absolute area of the sheets. The relationships between these quotients and environmental conditions (rainfall, seawater temperature, hurricanes) were evaluated using generalized additive models (GAMs) that supported tests for complex non-linear relationships with multiple predictors. GAMs were prepared using the mgcv package (version 1.8-34) (Wood, 2011; Wood, 2017) in R (version 4.0.5), accessed through the XLSTAT (version 2021.2.1, Addinsoft, Paris) add-in to Excel16.54 (Microsoft). Models were run using Gaussian errors, cubic splines, and variance components estimated by REML. Restricted Maximum Likelihood Models were restricted to three quantitative effects to enhance interpretation (Fisher et al., 2018), and the best model was identified from the lowest corrected Akaike Information Criterion (AICc) (Burnham & Anderson, 2004). Summary of model selection using GAMs is detailed in Table S2 of Edmunds (2022).

Environmental parameters
Rainfall values from thirty years of monthly rainfall measurements (in centimeters) are presented in a supplemental document.  From 1991 to 2011 rainfall values were obtained from the Southeastern Regional Climate Center (www.sercc.com), which compiled data from a rain gauge in Cruz Bay, St. John (Station 671980). Where this record was incomplete, values were obtained from Catherinburg (Station 671348), East End (Station 672551), or through interpolation. From 2012 onward, rainfall was measured using a Standard Rain Gauge (NOAA, National Weather Service) deployed on the north shore of St. John (18.3558°N, -64.7660° W) at Station VI-SJ-3 (https://wys.cocorahs.org). 

Seawater temperature was measured with loggers (primarily Onset Computer Corp., Hobo U22-001, ± 0.2°C) sampling at 0.0011Hz and located ~ 900 m from Cabritte Horn at Yawzi Point (Edmunds & Gray, 2014). Temperature was averaged by day and month to characterize the mean of the hottest three months prior to each sampling (“Temperature1”, August–October, mean ± SE, n = 3), and between 31 July and the previous 1 August by study year (“Temperature2”, mean ± SE, n = 12 months).  Mean seawater temperatures can be accessed here: https://www.bco-dmo.org/dataset/875694, and in the Related Datasets section.

Hurricanes and intensity were evaluated as a qualitative effect. Ranks were assigned based on hurricane tracks (https://www.nhc.noaa.gov) in conjunction with local knowledge of wind speeds and impact.  Years of major hurricanes were assigned a rank of one (1), and all other years a rank of zero (0).  Major storms affecting the St.John study area include Hurricanes Marilyn (1995), Georges (1998), Lenny (1999), Earl (2010), Irma (2017) and Maria (2017) which occurred in the Fall (Sep-Nov). Since ecological sampling was performed in July/August, the impact of these Fall storms was first recorded in the photoquadrat images taken 8-10 months later (in the following calendar year). For example, Hurricanes Irma and Maria occurred in September 2017, but any effect on the Millepora at Cabritte Horn was not detected until  sampling took place in Summer of the subsequent year.  Please refer to the "Hurricane ranking" file in the Supplemental Files section for additional details.