Ocean acidification (OA) represents a serious challenge to marine ecosystems. Laboratory studies addressing OA indicate broadly negative effects for marine organisms, particularly those relying on calcification processes. Growing evidence also suggests OA combined with other environmental stressors may be even more deleterious. Scaling these laboratory studies to ecological performance in the field, where environmental heterogeneity may mediate responses, is a critical next step toward understanding OA impacts on natural communities. We leveraged an upwelling-driven pH mosaic along the California Current System to deconstruct the relative influences of pH, ocean temperature, and food availability on seasonal growth, condition and shell thickness of the ecologically dominant intertidal mussel Mytilus californianus . In 2011 and 2012, ecological performance of adult mussels from local and commonly sourced populations was measured at 8 rocky intertidal sites between central Oregon and southern California. Sites coincided with a large-scale network of intertidal pH sensors, allowing comparisons among pH and other environmental stressors. Adult California mussel growth and size varied latitudinally among sites and inter-annually, and mean shell thickness index and shell weight growth were reduced with low pH. Surprisingly, shell length growth and the ratio of tissue to shell weight were enhanced, not diminished as expected, by low pH. In contrast, and as expected, shell weight growth and shell thickness were both diminished by low pH, consistent with the idea that OA exposure can compromise shell-dependent defenses against predators or wave forces. We also found that adult mussel shell weight growth and relative tissue mass were negatively associated with increased pH variability. Including local pH conditions with previously documented influences of ocean temperature, food availability, aerial exposure, and origin site enhanced the explanatory power of models describing observed performance differences. Responses of local mussel populations differed from those of a common source population suggesting mussel performance partially depended on genetic or persistent phenotypic differences. In light of prior research showing deleterious effects of low pH on larval mussels, our results suggest a life history transition leading to greater resilience in at least some performance metrics to ocean acidification by adult California mussels. Our data also demonstrate “hot” (more extreme) and “cold” (less extreme) spots in both mussel responses and environmental conditions, a pattern that may enable mitigation approaches in response to future changes in climate.

Mussel transplants

Using a standard protocol, performance of California mussels (30-45 mm total length) was quantified during 2011 and 2012 upwelling seasons [61,62,64]. We first collected mussels for pre-study measurements and individual tagging. Under permits from the Oregon Department of Fisheries and Wildlife and the California Department of Fish and Game (Oregon Department of Fish and Wildlife, 2010 permit #15122 and California Department of Fish and Wildlife S-183160003-18316-001), mussels were haphazardly collected from the vertically middle portion of M. californianus beds. In 2011 but not 2012, to assess genetic or persistent phenotypic influences on mussel performance, we translocated intermingled local-source (i.e., those from each site) and common-source (mussels from a single site, Bob Creek, Oregon, USA). To distinguish them from local-source mussels, common-source mussels were also marked with a bead of epoxy.

In the lab, translocation mussels were marked with a 1-2 mm triangular notch filed on the posterior shell edge (growing lip) to establish an indicator of initial length. Pre-outplant shell weight was estimated using a buoyant-weight method similar to [65]. Briefly, the process involved collecting separate mussel samples for model calibration at Bob Creek, Bodega Marine Reserve (northern California), Sandhill Bluff in central California and Lompoc Landing (southern California). The buoyant weight of each “calibration” mussel was measured by placing the live mussel on a platform submerged in water. The shells of each mussel were pinched closed during transfer through air, to prevent the confounding effect of air intake on buoyancy. Thus, submerged weight was an estimate of the negatively-buoyant shell weight. Soft tissue was then dissected from the shell and, after drying, shell weight was directly measured. The site-specific relationship between buoyant weight and dry shell was modeled using linear regression. The slope and intercept of each model was then used to estimate pre-study shell weight for translocated mussels. The Bob Creek regression model was used for Fogarty Creek and Strawberry Hill mussels, the Bodega Marine Reserve model for Van Damme and Bodega Marine Reserve, the Sandhill Bluff model for Terrace Point and Hopkins Marine Station, and the Lompoc Landing model for Lompoc Landing and Alegria. 

Mussel translocation

After pre-outplant processing, mussels were translocated back to the field for the April through October upwelling season. In 2011, mussels were sorted into 5 replicate groups of 50 per site, with each group consisting of 25 local- and 25 common-source individuals. For the 2012 season, mussels were sorted into 5 replicate groups of 30 per site. 

Mussel translocation used established methods [59]. Briefly, at each site, mussels were placed ventral side down in cleared plots 2-5 m apart within existing mussel beds.  Because bed heights varied among sites along the coast, tidal height of transplants varied (Table 1). We accounted for these differences by using tidal height as a covariate in data analyses. Mussels were held in place with plastic mesh (1-cm x 1-cm mesh) that was fastened using stainless steel lag screws inserted into pre-drilled holes with wall anchors. Two to four weeks later, the mesh was loosened to encourage more byssal thread production, and then 2-4 weeks later loosened further into a “dome” to allow space for growth while protecting the mussels from predation. 

Sample processing and growth measurements

Within 12 hours of collection, all mussels were placed in seawater tables, then within two days of collection, frozen at -20^oC. During processing, mussels were thawed, measured (length, width, and depth to the nearest 0.01 mm. Epibionts and byssal threads were removed from the shell exterior, and mussels were then dissected into two constituent parts – shell and soft tissue. These were dried separately at 80ºC for ≥ 5 days then weighed to the nearest mg.   

Shell-length growth was measured as mm new shell accumulated between the pre-study notch and the growing edge of the shell. Growth was standardized by dividing by initial length. Shell-weight growth was measured as the difference in pre- and post-study shell weight (g), standardized to the individual’s estimated pre-outplant shell weight and the study-season duration at each site.

Shell-weight growth of each mussel was calculated as the difference between the measured dry shell weight at the end of the season and the pre-season shell weight as determined by the previously described buoyant weighting method. 

The condition index (unitless) of each mussel was measured as the dry tissue mass per total (tissue + shell) dry mass. Higher condition index mussels have proportionately more soft tissue mass and may reflect energy allocation favoring tissue development [66,67]. Higher condition index may also reflect higher resource quality for mussel predators.

Mean shell thickness index (mg/mm²) was estimated by calculating the dry shell mass per shell surface area, with surface area (A ) calculated by the ellipsoid model A=l×h²+w^21/2×π÷2 , where l , h , and w  are mussel length, height and width, respectively [68]. All shell dimensions were measured to the nearest 0.1 mm, and shell weight was measured to the nearest 0.01g. The resulting index assumes a constant crystalline density of the shell structure. Major predators of M. californianus include Nucella whelks consume mussels through holes drilled their shells. Therefore, mean shell thickness index may correspond to drilling susceptibility [69].

Temperature

Temperature data were obtained using mussel biomimetics, which mimic the thermal properties of living mussels [70,71].  Each logger consisted of a thermistor-based temperature recorder (Tidbit logger, Onset Computer Corp., Bourne, MA) embedded in an epoxy mold shaped like an adult mussel. Using Z-spar epoxy, one to two loggers were deployed per site near replicate mussel plots, then covered with a plastic mesh cage to mimic conditions experienced by the transplanted mussels. Loggers recorded temperatures at 10-minute increments. Air and water temperature data were separated [72] and used to calculate mean temperatures by site and upwelling year.

Phytoplankton abundance

Phytoplankton are the primary food of M. californianus [73]. Food availability was quantified using chlorophyll-a concentrations ([Chl-a]) as a proxy for phytoplankton abundance. Chl-a was measured by periodically collecting water samples in opaque bottles during low tide at each site [74-76]. Replicate (n=3) bottle samples were collected at low tide from the shore at ~0.5m below the water surface. In the field, fifty ml of water was passed through 25-mm pre-combusted 0.7-µm Whatman GF/F glass-fiber filters. Filters were placed on ice and taken to the lab where Chl-a concentrations were quantified using a fluorometer. Because discrete sampling was not consistently conducted at all study sites in both study years, for analysis we averaged all bottle samples across all sample years creating site-specific long term mean summaries of Chl-a data. Prior research has shown spatial variability but temporal consistency in the levels of Chl-a among subsets of the sites used in this study [74,77].

pH measurements

pH data were collected at 10-minute intervals using autonomous sensors deployed at each site within 20 meters from the mussel plots. Sensors were attached to the rock using methods similar to the mussel translocations except that they were held down with stainless steel mesh. Care was taken to ensure that the sensing electrode remained wet even at low tide. Details on these custom-designed sensors can be found elsewhere [49,78], but briefly each was based on an ion-sensitive Honeywell Durafet® with an integrated data logger and power supply [79]. Sensors were calibrated either directly against certified reference materials or indirectly using spectrophotometric pH samples that were calibrated using certified reference materials. pH is reported on the total hydrogen ion concentration scale [80]. Calibrations occurred pre- and post-deployment for all sensors. To spot-check sensor performance, sensor data were periodically (2-4 weeks) compared to discrete water samples collected at all sites except at the two southern California sites in 2012 (Lompoc Landing and Alegria ).

To investigate how different aspects of the pH environment might influence mussel performance at each site and in each study year, we compiled summary statistics for the pH mean, standard deviation, and percentages of exposure below two thresholds: pH 7.8 and pH 7.7. These thresholds were chosen for their alignment with model predictions of average global pH conditions by the year 2100 [4,5,6]. Using tide tables, sensor data collected when tides were below the sensor were excluded from analysis. 

Robomussel temperature data

Site information; NA=no data

Regression mussels; NA=no data

Transplant mussels; NA=no data