Overview: Trials examining rates of shell dissolution were conducted in seawater of differing carbonate system conditions, using dead shells of sacrificed adult California M. californianus mussel individuals. The relationship between percent surface cover of periostracum and shell dissolution rate under contemporary but chemically stressful seawater conditions was examined. For the current study, adult mussels (42 - 64 mm in length) were collected from Marshall Gulch, California (38.369738 °N, -123.073921 °W) between August 2021 and March 2022 and transported immediately to the University of California Davis’ Bodega Marine Laboratory (< 30 min distance), in Bodega Bay, California. Mussels were held in filtered, flow-through seawater and fed ad libitum until used in experiments.

Shell preparation and periostracum cover measurements: The extent of intact periostracum coverage was determined for one of two valves of each M. californianus mussel. We photographed individual valves with a 12.2-megapixel digital camera (Google Pixel 4a) and then quantified the area of valve covered by periostracum and total valve surface area using ImageJ (software version 1.52a) calibrated to a scale bar. Because we were interested in the effects of the periostracum in protecting the exterior surface of the shell, we sealed the inner, nacre layer of the shell with silicone (Loctite marine silicone sealant) to prevent its contact with seawater.

Seawater manipulation: We used two techniques to establish the different pH treatments employed in our experiments. In the dissolution incubations conducted at pH = 7.5 (n = 49), we modified seawater chemistry using a standard mass-flow control system, bubbling gas of a fixed partial pressure of CO₂ directly into filtered seawater via flow-through sumps. In dissolution trials involving pH = 7.7 (n = 9) and 7.4 (n = 16), we employed an alternative, but equivalent, approach to manipulating seawater pH where we used direct chemical modification to the seawater carbonate system via equimolar additions of 1 M sodium bicarbonate (NaHCO₃) and 1 M hydrochloric acid (HCl). Both methods of seawater manipulation result in an increase of dissolved inorganic carbon (DIC) and reduction of seawater pH without changing total alkalinity (TA).

Dissolution Incubations and Analysis: Following determination of percent cover of periostracum, we incubated each mussel valve in a sealed 250 mL Nalgene bottle filled with modified seawater for 48 hours in a temperature-controlled room, recording seawater properties in each bottle before and at the end of each incubation, including temperature, salinity, and pH (Yellow Springs Instruments Professional Plus Sonde). YSI sonde pH values were calibrated to the total scale based on pH spectrophotometric measurements of m-cresol dye absorbance at the incubation temperature. Incubation bottles were agitated gently once every 8 hours over the course of the incubations to reduce the establishment of strong chemical gradients of their fluid contents. We took discrete seawater samples at the onset and end of each incubation for determination of total alkalinity concentration (TA); these “before-after” measurements of TA enabled quantification of the increase in alkalinity within each bottle over the duration of the incubation, and thereby the rate of dissolution of calcium carbonate (CaCO₃) shell material for individual valves of known periostracum cover. Shell dissolution rates were quantified using standard alkalinity anomaly techniques (analyzing seawater samples in triplicate and selecting median TA for dissolution quantification), which relate calcium carbonate shell loss to an increase in the total alkalinity (TA) of surrounding seawater, per unit time. Across the dissolution trials, control incubations (15% of daily sample size) of modified seawater were conducted to verify minimal background changes in TA.

Data analysis: We performed computations with R statistical software, RStudio version 2023.06.2. We performed carbonate system calculations using the package seacarb. We analyzed the influence of periostracum cover on CaCO₃ shell dissolution in M. californianus adults for trials at pH = 7.5 using a multiple effects linear model where dissolution rate was normalized to shell surface area (mm²) to account for variations in shell size. Normalized dissolution rates were square root transformed to meet linearity requirements for normally distributed residuals. We used backward stepwise model selection using F test significance as the criteria for omitting terms, resulting in a final model that employed both percent periostracum cover and valve length as fixed predictors. Throughout model selection tests, trial date did not significantly explain variability as a random effect so was omitted. Model residual homoscedasticity was verified using a Breusch-Pagan test (BP = 1.0525, df = 2, p-value = 0.5908), and assumptions of linearity and additivity were assessed visually using model residual and QQ-plots.

Additionally, we used a separate multiple effects linear model across the two additional pH levels (pH = 7.4, 7.7) to assess how seawater pH independently affects patterns of dissolution. Here, we included pH as a numerical fixed predictor, alongside periostracum cover. Due to a constrained sampling size and narrow range of shell lengths (52 mm ± 4 mm SD), shell length was not included as a fixed predictor in this model in order to conserve degrees of freedom. Normalized CaCO₃ dissolution rate was again square root transformed, and backward stepwise model selection was used to estimate the significance of fixed predictors. A Breusch-Pagan test confirmed residual homoscedasticity (BP = 3.5024, df = 3, p-value = 0.3204), and other assumptions were assessed visually as before.