Stress bands and annual growth bands were identified in the CT scan images of each core and quantified using the automated code coralCT for MATLAB which traces the density of individual corallites within the 3 dimensional core (DeCarlo and Cohen 2016).  Revisions to version 1.1 of coralCT, which was designed primarily for Porites spp., were made to accommodate the skeletal architecture of the Atlantic corals which have more prominent thecal walls than Porites. Specifically, a linear quadratic estimation algorithm (i.e. Kalman Filtering) improved polyp tracing along the core axis.  The code was also revised to enable automated identification of stress bands (Barkley et al. 2018). Specifically, the density time-series for all corallites in the core (see DeCarlo et al. 2015, Data Repository Figure DR2) were averaged to create an "ensemble" mean density time series for all traceable corallites. The averaged density profile time-series was then detrended to account for shifts in mean density that may occur over time.

Regions where skeletal density exceeded 2 standard deviations above the whole-core mean, a threshold chosen to account for the range in natural seasonal and inter-annual density variability, were identified as stress bands. To exclude fine-scale density anomalies such as worm borings, stress bands were defined as having a minimum width of 1 mm. Each CT scan was visually inspected to validate the presence and location of stress bands identified by the automated program.

BCO-DMO Processing Notes:

• Modified parameter names to conform with BCO-DMO naming conventions
• converted year/month  format to yyyy-mm