Photoquadrats at Yawzi Point and Tektite were recorded using cameras attached to a framer that held them perpendicular to the reef. A Nikonos V (35-mm format) was used from 1987-1999, and digital cameras thereafter, with 3.3 MP resolution from 2000–2006, 6.1 MP from 2006–2010, 12.1 MP in 2011, 16.2 from 2012–2015, and 36.3 MP from 2016–present. Cameras were fitted with a strobe (Nikonos SB 105) and the images resolved objects ≥ 1-cm diameter.

The analyses at Yawzi Point and Tektite were augmented in 1992 with six additional sites that were selected using random coordinates constrained to hard substrata. This sampling focused on a habitat where boulders and cliffs of igneous rock are common, mean coral cover has remained < 5%, and Orbicella annularis has not been common since at least 1992. Five sites are at 9-m depth, with one at 7-m depth (RS9), and they have been recorded annually. These sites serve as replicates of reefs between Cabritte Horn and White Point, and are analyzed as the pooled random sites (PRS). Results from 1995 (May) and 1996 (May) are used to evaluate the effects of Hurricanes Luis and Marilyn, and from 2017 (July and November) to evaluate the effects of Hurricanes Irma and Maria. Each site consists of a permanently marked transect at a constant depth that was 20-m long from 1992–1999 (n ~ 18 photoquadrats site-1), but was extended to 40 m in 2000 when digital photography was implemented (n ~ 40 photoquadrats site-1). Photoquadrats (0.5 × 0.5 m) were recorded at random positions along each transect (and re-randomized annually) using cameras (as described above) attached to a framer that held them perpendicular to the reef. Cameras were attached to two strobes (Nikonos SB 105), and resolved objects to at least 5-mm diameter.

Photoquadrats were analyzed by overlaying them with a grid of 200 randomly-located dots and identifying the substratum beneath each dot. Images were analyzed manually prior to 2005, from 2005–2011 using CPCe software, and from 2012 to present, using CoralNet software with manual annotations. With this approach, the abundance of each substratum type is defined by the total number of dots that occur on top of it in each image, and when expressed as a percentage of the dot population on each image, provides a measure of percentage cover (hereafter “cover”) (Menge 1976). Two resolutions were applied to the analyses, first to resolve three functional groups (FG), coral (combined cover of scleractinians), macroalgae (algae ≥ 1-cm high, mostly Halimeda, Lobophora, Padina, and Dictyota), and a combined category of crustose coralline algae, algal turf, and bare space (CTB). Second, scleractinians were resolved to the lowest taxonomic level possible, which was genera at Yawzi Point and Tektite, and a combination of species and genera at the PRS.

In addition to the hurricanes described herein, St. John also was impacted by Hurricane Lenny on 17 November 1999 (Table S1). However, underwater damage attributed to this storm was minor, probably due to the modest local wind speeds (150 km h-1), and propagation of damaging waves east and south that reduced their impacts on the southern shore of St. John. The effects of Hurricane Lenny are not considered in the present analysis. Given the vagaries of fieldwork extending over 31 Years, it was not possible to standardize the timing of sampling that took place before and after each storm episode. Sampling took place 6 weeks after Hurricane Hugo, 8 months after Hurricane Luis, and 9 weeks after Hurricane Maria. Sampling after the two most recent storms was comparable to sampling after Hurricane Hugo with regards to the delay following the storms and, therefore, probably quantified mostly coral mortality directly attributable to physical damage, and blooms of macroalgae commensurate with the growth that is possible in two autumn months. The longer delay in sampling after Hurricanes Marilyn and Luis probably resulted in measurements of coral mortality that was caused both by direct physical damage and delayed-onset disease, as well as blooms of macroalgae that can grow over 8 months extending from autumn to spring.