Coral skeletal core collection: We collected 101 skeletal cores from massive Porites coral colonies at ten reef sites representing two major reef environments, barrier reef and lagoon, the latter including fringing reefs around the uplifted karst Rock Islands. The two environments are broadly distinguishable in both physical (flow, temperature, and light regimes) and chemical (carbon system parameters, salinity) characteristics with generally higher flow, light, pH, and salinity and lower SST on the barrier reefs (Shamberger et al. 2014; Barkley et al. 2015).
Skeletal cores (20-40 cm in length) were collected in April 2011, September 2011, April 2012, August 2014, and January 2015 vertically from live coral colonies at 1-6 m depth using pneumatic drills with 3.8 cm diameter diamond drill bits. Core holes were filled with cement plugs hammered flush with the colony surface and sealed with underwater epoxy. Visual inspections of colonies 6-12 months after coring revealed significant overgrowth of plugs and no long-term impacts to the corals. Coral cores were oven-dried and scanned with a Siemens Volume Zoom Helical Computerized Tomography (CT) Scanner at Woods Hole Oceanographic Institution. 3-D CT scans of coral cores were analyzed using OsiriX freeware to visualize the 3-D image (Cantin et al. 2010; Crook et al. 2013) and an automated MATLAB code to quantify skeletal growth parameters and stress banding (DeCarlo et al. 2015).
Stress bands: Coral cores that included growth records prior to 1998 were assessed for the presence of high-density stress bands associated with elevated temperatures in 1998 (n = 86), and all cores were examined for stress bands in 2010. A stress band was defined as a region of the coral core > 1 mm in height and extending the entire width of the core where density values exceeded two standard deviations of the whole-core density mean. We defined a minimum band thickness in order to filter out smaller-scale density variability and high-density noise. A value of 1 mm for this thickness threshold was selected based on the average linear extension rates of Palau Porites corals (0.88 cm yr-1, interquartile range = 0.35 cm yr-1), where 1 mm represents, on average, approximately 10% of overall annual linear extension. High-density anomalies of this width therefore represent significant perturbations in growth. Density thresholds were set based on standard deviations from mean values in order to account for significant differences in density means and variability between individuals. Density values were normally distributed within coral cores, and values greater than two standard deviations were defined as the threshold for a stress band. This threshold was selected to aid in the identification of only the most anomalously high-density areas (i.e., areas with densities greater than approximately 95% of all values) while also minimizing the probability of type II errors in coral cores where stress bands exist but high-density values are slightly less extreme. Stress bands were identified as occurring within a particular year (specifically, 1998 and 2010) based on annual patterns of density banding, in which successive low density bands were counted down from the top of the core and subsequently dated based on the known date of collection. Although a small number of coral skeletal cores had occasional high-density regions in additional years, we did not consistently detect stress bands corresponding to years other than 1998 and 2010.
The percentage of Porites corals with stress bands was compared with community bleaching data for each reef site collected during the 1998 and 2010 high-temperature events. Bleaching data from 1998 were collected at nine reef sites in November 1998 using a point-intercept technique with three replicate 20-m transect surveys per site conducted at 3-5 m depth (Bruno et al. 2001). A subset of six of these nine sites was used to compare bleaching data to stress band records based on proximity to our core collection sites. Data from 2010 were collected at 80 randomly assigned reef sites in July and August 2010 with three replicate 30-m transect surveys conducted at 2-5 m depth (van Woesik et al. 2012). A subset of 31 of these sites was included in this study. Because in situ bleaching data were collected at randomized locations, the spatial matches between sites with bleaching data and sites with coral cores were not always exact. Therefore, for each coral core collection site, we averaged bleaching data from the two or three sites that both fell within a 10-km radius of each core site and that represented the same environment type to calculate a community bleaching estimate. Bleaching information about specific coral colonies from which we collected cores was not available.