Shell collection took place at rocky shore ecosystems spanning ~1,000 kilometers (km) of coastline along the west coast of North America. Shell collections occurred at three wave-exposed rocky headlands in California (Soberanes Point (36°26'50" N, 121°55'44" W), Bodega Marine Reserve (38°19'09" N, 123°04'28" W), Van Damme State Park (39°16'43" N, 123°48'12" W)) and at three wave-exposed rocky headlands in Oregon (Cape Arago (43°18'13" N, 124°24'07" W), Strawberry Hill (44°15'00" N, 124°06'55" W), and Fogarty Creek (44°50'16" N, 124°03'33" W)).

At each of the six study sites, mussels were collected from wave-exposed, mid-intertidal mussel beds. Mussels from 2000-2001 and 2008-2009 were originally collected for different projects. As a result, the sample size and range of mussel shell lengths for each site and time period varied somewhat. The mussel shells from the 2000-2001 time period came from three surveys. First, in 2000, 100 empty/dead mussel shells from Soberanes Point were collected; subsequently, live mussels from a similar range in shell lengths were collected from Strawberry Hill and Fogarty Creek. Second, in 2001, 25 live mussels from a limited size range (55-73 millimeters (mm)) were collected from each site (except for Cape Arago where 50 mussels were collected). Third, in 2001, quadrat surveys of drilled mussels were completed at a subset of the sites. The mussel shells from the 2008-2009 period were originally collected to survey predation in mussel beds (Sanford & Worth 2009). At each site, mussels from ten 0.25-square meter (m²) quadrats were removed, measured, and assessed for drill holes. For the mussel shells from 2019, approximately 100 live mussels were collected from each site. For all collections containing live mussels, the tissue was removed, and the shells were rinsed in fresh water and then air dried.

Mussel shells were cut on a bandsaw at 1/3 the length of the left valve from the anterior end. The mussel shell cross sections were scanned (model: CanoScan LiDE 110) and photos of the anterior section were used to calculate thickness via imaging software (ImageJ; Java 1.8.0_172). Maximum and minimum thickness along each cross section was measured, disregarding the dorsal hinge for maximum thickness and growing lip for minimum thickness, as these regions are known to be significantly thicker and thinner, respectively. Subsequently, average thickness was calculated using these two measurements. Due to the historical nature of the 2000-2001 and 2008-2009 mussel shells, there was some variation in how they were processed. First, some of the archived mussels were drilled/empty shells. Second, some of the 2008-2009 mussels only had the right valve remaining. Third, some of the 2000-2001 and 2008-2009 shells were analyzed on the posterior cross section since these shells were cut for a previous experiment and some shell pieces were missing.

We used the 2019 mussel shells to test the validity of our measurement of thickness to alternate metrics for shell thickness. For all 2019 shells, indirect shell thickness across the entire left valve was calculated as:

1000* valve dry weight / surface area

(where surface area = length x (height² + width²)^0.5 x pi/2) (Freeman & Byers 2006).

We also used 2019 mussels from two sites (Soberanes Point and Strawberry Hill) to determine the integrated thickness of the cross section, measured as the area of the polygon of the scanned image divided by the length of the curved mussel shell.