Clathromorphum species produce annual growth increments (hereafter "year bands") in their skeleton (Adey et al. 2013), which contain elemental and isotopic signatures that can be used to reconstruct past oceanographic conditions (e.g., Fietzke et al. 2015). We discovered that such year bands also archive urchin grazing scars, allowing us to measure the timing and frequency of past urchin grazing events on C. nereostratum. Wild specimens of C. nereostratum were thus collected and analyzed in order to reconstruct grazing events in the ecosystem.

At each site studied with respect to bioerosion, we collected C. nereostratum specimens (n = 10/site) with hammer and chisel, focusing on individuals without visible signs of grazing on the epithallus. Ship-side, samples were sectioned with a diamond lapidary saw and examined for quality; high quality samples were placed in an oven (50°C) until dry, then archived for subsequent analysis. Samples collected from Amchitka were of poor quality. Hence for Amchitka, we used specimens collected previously (2004) for this exercise. We also augmented our 2014 collections from Attu with samples previously collected (2008) from the same locale, because many of the 2014 samples did not meet our criteria for reconstruction (see below).

Each C. nereostratum specimen (n = 5/island) was mounted to a glass slide, sectioned parallel to the growth axis with a diamond lapidary saw, and polished to 3 microns resolution following established methods (Hetzinger et al. 2009). Each section was then imaged with a camera coupled to a reflected light microscope (GeoTS, Olympus Inc.), which obtains a mosaic of overlapping images and stitches them together to produce a single high-resolution image of the cross-section. This photomosaic was used to carefully identify, age, and count grazing scars that occurred along a transect oriented parallel to the alga's growth axis in the plane of the section. We scored multiple (n = 2-3) transects per sample, given that grazing events do not span the entirety of a year band. The origin of each transect was randomly plotted, then moved to the nearest location on the epithallus that was living and that displayed a flat or convex shape. The transect was then plotted through sequentially older year bands, so long as it: (i) did not cross a fusion between two individual algae; (ii) spanned at least 30 years of growth; and (iii) intercepted year bands that were clearly visible. If any of these criteria were violated, the transect was relocated.

Along each transect, we aged and measured the growth (vertical extension) of year bands in 5-year intervals, repeating this process so long as dating could be rigorously performed and end-dates aligned for all transects within a sample. Within each 5-year interval, we assessed the annual frequency of grazing scars - which manifest as a jagged interruption of the growth margin coupled with serial pitting of the reproductive conceptacles - found 5 mm to either side of the transect. Straight growth lines that lacked conceptacles or the occasional empty conceptacle within an otherwise clean growth band was not considered evidence of grazing. In instances where one side of the transect entered an area that violated the above criteria (e.g., passed under an area that had a concave epithallial surface), we analyzed only one side of the transect (10 mm). With this methodology remains the possibility that urchin grazing may have removed entire years of growth, thus obscuring our estimates of grazing frequency and the timing of each grazing event. To address this issue, we compared (double blind) our age models for a subset of samples from Alaid and Ogliuga to ages produced using Uranium series dating (Fietzke et al. 2005). Our age estimates were very similar to those produced by Uranium series dating, indicating we did not lose entire years to grazing.

We selected the year 1965 as the cutoff for our reconstruction because ecological records are scant prior to this period (Estes et al. 2010). For C. nereostratum samples collected in 2014, we excluded the 10 most recent years of growth (2014-2005) from our analyses due to a known bias in our collection method; since we collected non-grazed specimens from the wild and retained only the highest quality samples, we biased ourselves against finding evidence of grazing in recent years. For samples collected in 2004 and 2008, such a bias was not evident, at least for those records that met our criteria and were used in the study. For the three 2008 samples used, we excluded the first four years to align chronologies with all other specimens.