Ecology, distribution, and life history of Mytilus californianus

Mytilus californianus is dominant in modern rocky intertidal ecosystems spanning 25° of latitude along the northeastern Pacific Coast from the Aleutian Islands of Alaska, USA to Baja California, Mexico [28]. As a foundation species, M. californianus forms a habitat for hundreds of marine taxa, provides a food source for coastal predators, and filters particulate matter from the intertidal water column [28–30]. Mytilus californianus beds are extensive and dominant in exposed, open-ocean environments with heavy wave action due to its strong, tri-layered shell and thick byssal threads [31, 32]. Because of its widespread biogeographic distribution and its role in intertidal food webs, M. californianus has been a foundational study organism in the fields of intertidal ecology and marine invertebrate physiology since the mid-20th century [28, 31, 33–41]. Despite the attention M. californianus has received, its life history remains enigmatic; its lifespan is uncertain, although it is thought to range from 10 to 100 years depending on environmental conditions and disturbance levels [31, 42]. Shell growth rates are variable across both micro-scale (e.g., tidal) and macro-scale (e.g., latitudinal) gradients that control conditions such as water temperature, wave action, and immersion time [29, 40, 43]. Estimating shell growth rates and longevity is further complicated by the lack of time-calibrated periodic growth bands [44]. Despite these complexities, previous studies have documented growth patterns of M. californianus in southern California populations; Ford et al. (2010) found that M. californianus shell length grew continuously over the course of a 382-day outplant experiment in San Diego, which is advantageous for reconstructing minimum and maximum temperatures over the course of a year from shell isotopic records [42]. Additionally, Ferguson et al. (2013) found evidence for continuous calcification without growth shutdown in M. californianus throughout the year at sites in southern California, USA and Baja California, Mexico [45]. There is also general agreement in the mytilid literature that adult M. californianus shells accrete approximately 2–5 mm at their terminal growth margin per month in southern Californian environments [25, 31, 34], where they grow more rapidly in seawater temperatures ranging from 15° to 19°C [29]. Continuous annual growth patterns and the relationship between temperature and growth rate are specific to southern California since the upwelling and temperature regimes differ dramatically for regions north of Point Conception, resulting in significantly slower growth rates for marine calcifying biota in the central and northern portions of the California Current System [29, 40, 44, 46].

Mytilus californianus and human history

In addition to its significant role in structuring the intertidal ecosystem, M. californianus has been a culturally important species and major food source for Indigenous peoples throughout the Holocene. Mytilus californianus shells are found in northeastern Pacific Coast shell middens at a higher density than any other bivalve species [25]. This species is particularly abundant in the shell middens of the Channel Islands, California, USA due to the populous Chumash civilizations that inhabited southern California throughout the Holocene [25, 47, 48]. Prior to European colonization, the Chumash peoples of present-day southern California had established 22 major villages on Santa Cruz, San Miguel, and Santa Rosa Islands, built extensive trade networks throughout coastal California and all nine Channel Islands, and reached the largest populations among hunter-gatherer civilizations known to published records [49]. The dominant presence of M. californianus shells in Channel Islands shell middens across the Holocene indicates that the Chumash groups heavily and continuously harvested M. californianus as a significant food source [48, 50].

Exceptional preservation of Channel Islands shell middens has attracted the attention of archaeologists and anthropologists for the past two centuries, producing an extensive collection of M. californianus shells extracted from the middens. Archaeologists commonly apply stable isotopic analysis to midden shells to reconstruct the season of harvest, which is useful for investigating questions about human migration patterns, site usage, and cultural traditions [51–55]. When analyzed collectively, the ample stable isotope records from archaeological M. californianus shells have the potential to provide an archive of coastal climate variability, seasonality, and shell life-history traits through time. In general, the vast isotopic data from previously excavated collections of archaeological shells have the potential to be valuable paleo-archives. Here, we test the use of M. californianus data from the Channel Islands middens to investigate the utility of this species as a record of environmental conditions across the Holocene.

Shell chemistry of Mytilus californianus

Species-specific relationships between δ¹⁸O_shell and sea surface temperature (SST) and potential covariation in δ¹³C_shell and the δ¹³C of dissolved inorganic carbon (DIC) of seawater allow for paleoenvironmental reconstruction from M. californianus shell carbonate. When sampled serially along the ontogenetic growth trajectory (from umbo to commissure), δ¹⁸O_shell has been found to be a reliable record of mean annual temperature and mean annual temperature ranges in modern M. californianus shells from San Diego, with a slight and consistent temperature-dependent enrichment in ¹⁸O_shell by 0.2 to 0.5‰ relative to predicted δ¹⁸O_equilibrium [42]. Interpretation of δ¹³C_shell is more complex and inconsistent due to the incorporation of metabolic (respired) carbon in the shell mineral. In molluscs, δ¹³C_shell typically decreases with ontogeny as physiology changes with age [56]. For Mytilus in particular, up to 10% of isotopically light respired carbon can be incorporated into the shell during calcification, and metabolic rate can vary throughout ontogeny [45, 57], hindering the use of δ¹³C_shell profiles as direct proxies of the environment. However, significant changes in δ¹³C_shell through time can also indicate broad shifts in ocean carbonate chemistry, as documented by a negative trend in modern δ¹³C_shell in M. californianus relative to δ¹³C_shell of archival and archaeological M. californianus shells collected in Washington State [58]. In addition to assessing long-term changes in ocean carbonate chemistry, δ¹³C_shell profiles from multiple individuals from the same locality could be used to track short-term, upwelling-driven changes in ocean chemistry on a seasonal scale since upwelled water delivers nutrients and respired, isotopically light carbon to surface waters [45, 59], although the relationship between δ¹³C_shell and δ¹³C of DIC in seawater is inconsistent [45, 60, 61]. While Killingley and Berger (1979) found remarkable correlation and covariation between upwelling indices and δ¹³C_shell for San Diego M. californianus shells, more recent work did not corroborate these patterns in modern M. californianus shells from the same locality [45, 59].

Diagenesis is a potential concern when interpreting stable isotope and trace element data from archaeological or fossilized carbonates. All non-modern M. californianus shells analyzed in this study are from shell middens, which structurally and chemically favor preservation since biogenic shells in a midden deposit serve as a natural, carbonate-rich buffer to dissolution and the layers of a midden act as a barrier against erosion [62]. Additionally, M. californianus has an outer layer of calcite [44, 63], which is the more stable form of calcium carbonate and therefore less susceptible to alteration. All sub-samples analyzed here are from the outer calcite layer of M. californianus shells, and all midden shell data were sourced from studies that methodically addressed diagenesis by performing X-ray diffraction analysis and/or HCl etching to avoid sampling diagenetically altered shell carbonate prior to stable isotope analysis [64, 65].

Study area

The California Current System (CCS) influences patterns of sea surface temperature (SST), upwelling intensity, and nutrient concentrations in southern California. The CCS comprises the cold California Current, the warm north-flowing California Undercurrent (subsurface), the warm north-flowing Davidson Current (surface), and the Southern California Eddy [46, 66–68]. The dominant California Current flows southward along the coastline of northern California parallel to the alongshore winds, resulting in strong Ekman transport and intense upwelling near the coast of northern and central California [46, 67]. South of Point Conception, the coastline morphology changes such that the current no longer runs parallel to alongshore winds, which results in weaker Ekman transport and upwelling along southern California. The mild upwelling produces a warmer SST regime along the southern California coast relative to the northern CCS. Wind forcing and upwelling is strongest during the summer in all portions of the CCS [46], but seasonal variability is slightly less pronounced south of Point Conception, resulting in a warm zone and a weaker yet more stable upwelling regime off the coast of southern California.

This study uses only M. californianus shells collected in the southern portion of the CCS, south of Point Conception and north of Baja California (Fig 1), due to the abundance of M. californianus records from this region. Specific collection sites are located on the northern Channel Islands (San Miguel Island, Santa Rosa Island, Santa Cruz Island, and Anacapa Island), Newport Beach, and San Diego. Channel Islands mussels were more likely to experience relatively lower temperatures from the cold California Current, while mussels from Newport Beach and San Diego may have experienced warmer water on the order of ~ 3°C delivered by the California Undercurrent and Davidson Current [69]. The temperature differential between the islands and mainland sites is compounded by considerable small-scale oceanographic variability among the northern Channel Islands due to ocean current activity, which produces a warm-cool SST gradient from east to west [69–72]. At Santa Cruz Island, for example, mean SSTs have been lower on the west coast than the southeast coast throughout the Holocene [72]. Mean SSTs at San Miguel Island are ~ 1 to 4°C lower than those at Anacapa Island [73]. The dynamic nature of the intertidal zone is another source of micro-scale temperature variability across tidal heights even within a single locality. In addition to multi-scaled spatial variability, seasonal patterns result in a ~ 5°C annual temperature range [54, 74]. For example, at Anacapa Island, SST alternates between warming (~ 16°C to 19°C) from late June through early September and cooling (~ 19°C to 15°C) from late September to early March [53]. Seasonal cycles are similar throughout the northern Channel Islands (i.e., cooling and warming periods occur relatively synchronously for all four islands) due to the seasonally driven upwelling regime [73]. No permits were required for our study since no field sites were accessed; all data were sourced from previously published studies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Map of study region in southern California for illustrative purposes only.

Collection sites (red dots) are indicated at northern Channel Islands (San Miguel Island, Santa Rosa Island, Santa Cruz Island, and Anacapa Island), Newport Beach, and San Diego. Inset map of northern Channel Islands modified from USGS National Map Viewer.

https://doi.org/10.1371/journal.pone.0302945.g001

We analyzed Channel Islands shells separately from mainland California shells to determine whether the Channel Islands produced higher mean δ¹⁸O_shell values as expected due to cooler seawater conditions farther offshore. Comparing mainland and island data also provides the opportunity to examine potential δ¹⁸O_shell and δ¹³C_shell shifts related to coastal versus offshore differences in salinity due to precipitation and freshwater runoff.

Mytilus californianus shell records

We used Web of Science to search for relevant literature in geochemical, archaeological, paleontological, paleoceanographic, and sclerochronological journals using combinations of the following keywords: California mussel, Mytilus californianus, stable isotope data, oxygen isotopes (δ¹⁸O), and carbon isotopes (δ¹³C). We included oxygen and carbon isotope data from M. californianus shells that met the following criteria for integration and further analysis:

• all shells were definitively identified by the original authors as M. californianus.
• all shells were sourced from southern California (south of Point Conception), including the Channel Islands, either collected live from the intertidal zone or excavated from a coastal shell midden or outcrop.
• all mussels lived during the Holocene (geologically younger than 11,700 years old).
• all shells have multiple (two or more) stable isotope subsamples per shell along the ontogenetic growth trajectory with a known sampling direction (i.e., stable isotope data is presented in relation to the axis of shell growth to infer the directionality of stable isotope trends in individual shell profiles).
• non-modern shells were examined with X-ray diffraction and/or treated with HCl etching prior to stable isotope analysis to avoid sampling diagenetically altered shell material.
• all shells were complete or nearly complete valves (no fragments or shells made into tools or jewelry).
• non-modern shells were published with geographic locations and associated radiocarbon ages from a dated section of a midden to confidently place specimens in temporal context, ± 30 years.
• modern shells had a known collection date (month/day/year) and latitude/longitude.
  While we intended to include fossil specimens in this study, we found no published records of stable isotope data from fossilized M. californianus shells from this region and time interval (but see Dodd (1966) for stable isotopic analysis of Pleistocene Mytilus shells from other regions of California [75]). We included no summary statistics in our dataset; we only entered shells that reported the full suite of δ¹⁸O_shell and δ¹³C_shell sub-sampled values. The resultant assembled dataset contains 6036 oxygen and carbon stable isotope values from 411 M. californianus shells integrated from 13 published studies (S1 File). The majority (10 of 13 studies) utilized archaeological shells to reconstruct patterns of early human settlement or foraging behavior [51–53, 72, 76–81], while the other three studies used modern M. californianus shells to calibrate stable isotope-based environmental proxies [42, 45] or test the influence of tidal height on shell geochemistry [42, 74]. Summary information, including study objectives and findings, of all 13 papers are presented in Table 1. For each individual shell that met our criteria, we assigned it a unique shell ID number in Excel and then entered δ¹⁸O_shell and δ¹³C_shel values for each subsample in a unique row associated with that shell ID. In addition to stable isotope data, the synthesized dataset also contains the number of subsamples per shell, the citation of the primary publication, subsample label or distance along shell (mm), age in years before present (BP), specimen type (midden or modern), location (islands or mainland), site (specific island or city), latitude, and longitude. The subsample distance along the shell is unrelated to total shell length. In most cases, the total shell length was not provided, although we included it whenever it was reported by primary authors (S1 File). No permits were required for the described study, which complied with all relevant regulations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of the 13 papers from which data are sourced for this study.

https://doi.org/10.1371/journal.pone.0302945.t001

Stable isotope sampling techniques used by primary authors

Shell profiles included in our data synthesis range from 2 to 60 subsamples per individual, representing 2 to 118 mm worth of shell, respectively. Subsamples are spaced evenly, 1 to 3 mm apart, depending on the study. The subsample spacing is reported serially in mm for each individual whenever this information was reported by primary authors.

Many archaeological studies apply the terminal growth band (TGB) sampling approach to infer the season of harvest [77–79, 82]. For individual shells with fewer sub-samples (< 5) analyzed here, the primary authors performed TGB sub-sampling nearest the commissure. To determine whether δ¹⁸O_shell-inferred SST was increasing (indicative of spring-summer) or decreasing (indicative of fall-winter) during the time of collection, primary authors drilled an additional one to three sub-samples of shell material that precipitated just before the TGB, a sampling approach written as TGB + n, where n = number of additional sub-samples [78, 79]. For example, TGB + 1 indicates that one sub-sample was taken at the terminal growth band of the shell, and one additional sub-sample was drilled ~ 2–3 mm umbo-ward of the TGB, producing a two-sample shell profile used to determine the season or conditions occurring at the end of the individual’s life. This commonly used subsampling strategy is based on the estimate of ~ 2–3 mm of growth per 1–2 months in adult M. californianus shells growing in southern California [25, 29, 34, 74, 77].

Categorization and analysis of shell records

After compiling stable isotope data from all shells that met our criteria, we categorized these data for analysis based on sample age, location, and sub-sampling strategy. For geographic categorization, data were binned according to their individual site (San Miguel Island, Santa Rosa Island, Santa Cruz Island, Anacapa Island, San Diego, Newport Beach) and whether they were from an island or the mainland. For temporal comparisons, data were binned both by millennium (e.g., within the past 1000 years, 2000, 3000, etc. up to 9000 BP) and by sub-epoch (early, mid, or late Holocene) according to the Holocene sub-epoch boundaries formally defined by Walker et al. (2019): early Holocene (11,750–8200 BP), mid Holocene (8200–4200 BP), and late Holocene (4200–0 BP) [83]. For late Holocene shell samples, we distinguished between sample types to compare late Holocene midden samples and late Holocene modern, live-collected samples (S1 File).

Individual shells were categorized for analysis based on subsampling strategy: long profiles (15–60 sub-samples per shell, or 28–118 mm of growth), medium profiles (5–14 sub-samples per shell, or 8–26 mm of growth), and short shell profiles (2–4 sub-samples per shell, or the last 4–6 mm of growth per individual shell). Sampling strategies were categorized in this way for three reasons: (1) to run analyses using the longest and highest-resolution shell records to investigate life-history traits and time-averaging effects of sampling methods, (2) to examine patterns in shell data that may contain a full seasonal cycle or more in medium-length and long shell profiles, and (3) to isolate short profiles, which only capture a season or less of growth and therefore would bias the inferred temperature record based on their season of collection.

Using subsampling methods and the directionality of sampling described in each paper, we expressed shell profiles in terms of “distance from growing margin” so that 0 mm represents the most recent shell material to precipitate, closest to the date of harvest or collection. Long profiles were analyzed to examine ontogenetic variability, identify seasonal isotopic trends, and assess variability among individuals. For long-profile modern shells, we compared seasonal interpretations with known instrumental SST records for relevant southern California sites. Both medium and long profiles (five or more subsamples containing multiple seasons of growth per shell) were used for geographic and millennial-scale analyses. Short profiles (< 5 subsamples per shell) were examined in comparison to long profiles and, whenever possible, in conjunction with the annual temperature cycle, to test whether M. californianus shells reliably recorded sub-seasonal δ¹⁸O-inferred temperature trends since short δ¹⁸O_shell profiles from archaeological molluscs are commonly used to infer season of harvest or site occupation [e.g., 82]. Short profiles were also used to investigate the influence of ontogenetic stage and tidal height on terminal edge δ¹⁸O_shell values. We compared δ¹⁸O_shell profiles of different lengths for shells of the same ¹⁴C age and from the same island to determine whether the length of the subsample profile impacts the interpretations of shell chemistry (S1 Fig). Lastly, we compared the mussel record to offshore marine sediment and foraminiferal isotopic records for the same time interval and region.

To analyze ontogenetic trends, local polynomial regression was applied to the longest δ¹⁸O_shell profile. We accessed seawater temperature data for 2000 through 2005 from the National Data Buoy Center (NDBC) Station 46054 (34.273° N, 120.470° W) and Point Dume Shore Station Data provided by the Shore Stations Program at Scripps Institute of Oceanography (34.014° N, 118.822° W) to compare the annual temperature cycle with the longest δ¹⁸O_shell record matching this same range of years. For statistical analysis, Pearson’s correlation was used to test relationships between ontogenetic stage (total shell length) and δ¹⁸O_shell and δ¹³C_shell values in short profiles whenever shell length data were available. Linear regression (Ordinary Least Squares, OLS) and Pearson’s correlation were used to evaluate intra-shell δ¹⁸O_shell-δ¹³C_shell relationships in long-profile individuals. Analysis of variance (ANOVA) was applied to modern shells with medium and long profiles to test for differences of δ¹⁸O_shell and δ¹³C_shell values between island and mainland settings. ANOVA and Tukey HSD were also used to assess differences between individual sites and across tidal heights whenever tidal height data were available. Tukey HSD was used to determine which tidal heights were statistically different from one another in isotopic composition and evaluate the impacts of time-averaged subsampling. ANOVA and Tukey HSD were also used to determine whether δ¹⁸O_shell values were significantly different from one another over each millennium throughout the Holocene. A Welch Two-Sample t-test was used to compare modern and midden δ¹³C_shell values. All statistical analysis was performed in R [84].

Paleotemperature and δ¹⁸O of seawater

To examine temperatures recorded by M. californianus through the Holocene, we applied the δ¹⁸O-SST equation developed by Epstein et al. (1953) [85] and calibrated for M. californianus by Killingley (1981) [86], where δ¹⁸O_w is the stable oxygen isotope value of seawater: (1)

The 13 studies included in our synthesis employed a variety of approaches for δ¹⁸O_w. Multiple studies [72, 76, 80] used the modern measured δ¹⁸O_w value of -0.32‰ from a seawater sample collected by Rick et al. (2006) off the eastern coast of Santa Rosa Island [52]. Three different studies used the δ¹⁸O_w value of -0.32‰ but applied an ice-volume correction [53, 78, 79] for reconstruction of Holocene paleo-temperatures [87]. Two papers did not provide a δ¹⁸O_w value. Since sea level has been relatively constant since 6000 BP [88, 89], we applied an ice-volume correction to δ¹⁸O_w for shells with a radiocarbon age older than 6000 BP only and used the modern δ¹⁸O_w value of -0.32‰ for all other Channel Islands shells younger than 6000 BP. We used established relationships between sea-level change and δ¹⁸O_w; a 10 m increase in sea level results in a 0.1‰ change in δ¹⁸O_w [90, 91]. Assuming a linear change in ice volume during the early Holocene, sea level rose 10 m per millennium [88, 89, 91]. Therefore, we used δ¹⁸O_w = -0.32‰ + 0.3‰ for mussel shells aged 9000–8000 BP, -0.32‰ + 0.2‰ for mussel shells aged 8000–7000 BP, and -0.32‰ + 0.1‰ for shells aged 7000–6000 BP (Table 2). While δ¹⁸O_w can vary on a seasonal scale in addition to a millennial scale, Jazwa et al. (2020) reported that the standard deviation of δ¹⁸O_w values from water samples collected seasonally at Bechers Pier on Santa Rosa Island from August 2015 to May 2019 was 0.1‰, ranging from a minimum of -0.6‰ to a maximum of -0.2‰ [74]. The low seasonal variance of δ¹⁸O_w would yield a temperature difference that is substantially lower than the seasonal temperature range of ~ 10°C, so we used the same modern δ¹⁸O_w value for all data points for which an ice-volume correction was not applicable.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Treatment of δ¹⁸O_w values for this study according to binned age range (years BP).

https://doi.org/10.1371/journal.pone.0302945.t002

Ontogenetic variability

Ontogenetic trends revealed by long profiles.

Of all individual profiles analyzed here, the M. californianus shell with the longest stable isotope profile was collected at Cuyler Harbor, San Miguel Island in 2005 CE [80]. The shell was serially sub-sampled every 2 mm at 60 points along the ontogenetic growth trajectory, representing 118 mm of shell growth [80]. The length and resolution of this record presented the best opportunity to examine ontogenetic variation; the next longest individual shell record from the Channel Islands has 25 subsamples spanning 48 mm of growth.

The San Miguel 2005 shell recorded an average δ¹³C_shell value of -0.51‰, an average δ¹⁸O_shell value of 0.53‰, and six identifiable annual cycles in its δ¹⁸O_shell profile (Fig 2). The δ¹⁸O_shell minima represent warm seasons recorded by the shell and match the two instrumental temperature records from 2000–2005 that we examined for comparison (Fig 2a): National Data Buoy Center (NDBC) Station 46054 located in the Santa Barbara Channel (34.273° N, 120.470° W) and Point Dume Shore Station Data provided by the Shore Stations Program at Scripps Institute of Oceanography (34.014° N, 118.822° W) [80]. Comparing annual temperature profiles to the δ¹⁸O_shell profile indicates that the amount of time represented by evenly spaced (2 mm) subsamples decreases as the shell ages (Fig 2a and 2b). In the years 2000 and 2001, earlier in the individual’s life, the subsamples represent sub-monthly resolution (i.e., ~ 14–18 subsamples capturing 12 months of growth). Following fast growth during earlier stages of life, growth reduction could be linked to the attainment of sexual maturity, when mussels reallocate a greater portion of energy from shell growth to gametogenesis [43, 92]. In the final two years of the individual’s life (2004 and 2005), one full annual cycle is captured within 6–7 subsamples (Fig 2b). This indicates that M. californianus slows its growth from ~ 2.5 mm per month earlier in its life to ~ 1 mm per month within a span of five years. Ontogenetic growth reduction amplifies the effect of time-averaging introduced by the sampling strategy, as shown by the less apparent annual cycles when local polynomial regression was applied to the δ¹⁸O_shell profile with evenly spaced samples (Fig 2c). When M. californianus is growing more slowly in later stages of ontogeny, more subsamples are required to capture environmental periodicity (e.g., temperature seasonality) since the individual is calcifying, and therefore recording, less environmental information in the same amount of time.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Stacked plot comparing seawater temperatures and the longest δ¹⁸O_shell profile.

(A) Seawater temperatures for 2000–2005. Black line is a two-week running mean for NDBC Station 46054 in Santa Barbara Basin. Gray open circles are daily SST. Blue line is two-week running mean for Point Dume provided by the Shore Stations Program sponsored at Scripps Institution of Oceanography. Blue open circles are daily seawater temperatures. Numbers indicate maxima for each year of the record. (B) 118 mm of δ¹⁸O_shell from a M. californianus individual collected in 2005 at San Miguel Island. Gray dashed line represents the midpoint value of 0.49‰. Labeled numbers represent inferred warm seasons best matched with panel (A). (C) Local polynomial regression of δ¹⁸O_shell showing that modeled seasonal cycles become less apparent as the shell ages ontogenetically due to slower growth. Shaded region represents 95% confidence interval. (D) Calculated seasonal amplitude of δ¹⁸O_shell (dark circles and dashed line) and annual standard deviation of δ¹⁸O_shell (black triangles and solid line).

https://doi.org/10.1371/journal.pone.0302945.g002

Since growth reduction throughout ontogeny has been well documented for accretionary carbonates [93–97], we predicted that the M. californianus δ¹⁸O ontogenetic profile from umbo to commissure would contain the following characteristics: (1) sinusoidal in shape, but increasingly cuspate as the individual aged, (2) a progressive decline in amplitude, (3) a progressive increase in frequency, and (4) a progressive decline in wavelength. As predicted, the δ¹⁸O_shell profile was generally sinusoidal with sharp cusps marking seasonal patterns during the last three years of its life (Fig 2b). The increasingly cuspate shape of the δ¹⁸O local minima suggests that growth slowdown occurred as the organism aged, resulting in shorter seasonal cycles [97–99]. Generally, the frequency increased, and the wavelength declined as predicted. Unexpectedly, however, this individual’s δ¹⁸O_shell profile did not exhibit an ontogenetic decline in seasonal amplitude (Fig 2d), as determined by calculating the difference between the local δ¹⁸O_shell extrema for each warm or cool season and the profile’s midpoint δ¹⁸O_shell value (0.49‰). This results in six warm and six cool seasons over the six-year period recorded by this individual. If δ¹⁸O_shell values lower than the midpoint value of 0.49‰ are considered to be warm seasons and δ¹⁸O_shell values higher than this are interpreted as cool seasons, then the shell recorded warm seasons for 41.7% and cool seasons for 58.3% of the 118 mm-long shell record. More fully profiled individuals of this size would help determine whether these percentages reflect a true environmental signal driven by the annual temperature cycle, or whether they are biologically mediated due to growth shutdown/slowdown in warmer conditions, particularly as the organism aged. Throughout ontogeny, the shell continued to record seasonal cyclicity but recorded less sub-seasonal variability (Fig 2b). Additional closely spaced subsamples towards the terminal edge of the shell would likely help capture finer-scale variability (e.g., weekly to monthly) in an aging individual.

We also found a small overall increase in annual standard deviation of δ¹⁸O_shell from 0.27‰ in 2000 to 0.39‰ in 2005 (Fig 2d). This increase was expected since a lower standard deviation earlier in life represents the lower short-term variation between temporally close samples, and a higher standard deviation later in life reflects the greater variability or larger temperature swings from one subsample to the next due to more temporal gaps in the calcified record, and therefore fewer subsamples representing a season of growth.

The ontogenetic δ¹³C_shell profile of the same individual from San Miguel Island was examined separately since δ¹³C is a more complex proxy than δ¹⁸O in biogenic carbonates due to the varying impacts of ocean circulation, upwelling, primary productivity, seawater DIC δ¹³C, and metabolic effects on δ¹³C_shell profiles [45]. In this individual, the ontogenetic δ¹³C_shell profile records six annual cycles with increasing frequency (Fig 3), which could reflect seasonal upwelling/relaxation oscillations in the Santa Barbara Channel. There was an ontogenetic trend towards more positive δ¹³C_shell values (Fig 3), which does not fit the prediction that the organism incorporates a greater amount of isotopically light metabolic carbon as it ages, as found in other marine mussel species including Pinna nobilis and Mytilus edulis [100–102]. The unexpectedly positive δ¹³C_shell trend found in this individual could be associated with ontogenetic growth reduction in this species. Slower shell growth and lower metabolic rates could result in reduced incorporation of respired carbon and therefore higher δ¹³C_shell values with age [102]. Alternatively, the positive δ¹³C_shell trend could be indicative of changes in the timing of growth as the organism aged, such as less growth during periods of upwelling (low δ¹³C) throughout ontogeny.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. δ¹³C_shell profile of San Miguel Island M. californianus individual collected in 2005.

X-axis represents ontogenetic growth from left to right (e.g., 0 mm is closest to time of death). Gray dashed line represents a median δ¹³C_shell value of -0.47‰. Regression line (in red) indicates ontogenetic trend towards positive δ¹³C_shell values.

https://doi.org/10.1371/journal.pone.0302945.g003

Geographic variability

Geographic trends in modern shells.

Analysis of modern shells with five or more subsamples (medium and long profiles) revealed a significant difference in shell chemistry between M. californianus shells collected from the Channel Islands versus mainland southern California (Fig 4, ANOVA, oxygen: F_1,735 = 194.1, p < 0.001, carbon: F_1,679 = 318.2, p < 0.001). In present-day conditions, weak wind forcing and the delivery of warm water by the northbound California Undercurrent and Davidson Current produce a significantly warmer regime along the Southern California Bight than farther offshore (Fig 1), and this pattern is reflected in statistically different M. californianus shell chemistry: lower δ¹⁸O_shell values (mean ± 1σ = -0.32‰ ± 0.47‰) from the mainland sites (Newport Beach and San Diego) relative to higher δ¹⁸O_shell values (mean ± 1σ = 0.33‰ ± 0.42‰) in the Channel Islands shells (Fig 4a). Additionally, the lower δ¹³C values of Channel Islands shells (Fig 4b) may reflect the islands’ proximity and exposure to the biologically productive California Current. The south-facing coastline of the Southern California Bight is protected from intense wind forcing [46], resulting in weaker upwelling. Weaker upwelling is typically associated with higher δ¹³C_shell values, as found in the mainland shells relative to δ¹³C_shell from the offshore islands (Fig 4b).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Box plots showing stable isotopic ranges of all modern shells by location.

Channel Islands includes all data points from all modern shells with five or more subsamples from San Miguel Island and Santa Cruz Island. (There were no modern shells with more than five subsamples from Anacapa or Santa Rosa Island.) Mainland California includes all data points from all modern shells with five or more subsamples from San Diego and Newport Beach. Black bar is the median value. (A) Oxygen isotope range of modern δ¹⁸O_shell in Channel Islands vs. mainland shells. (B) Carbon isotope range of modern δ¹³C_shell in Channel Islands vs. mainland shells. Arrow shows expected directionality of a potential δ¹³C upwelling signal.

https://doi.org/10.1371/journal.pone.0302945.g004

Geographic variability through time.

Oxygen and carbon isotope data from all shells with five or more subsamples (medium and long profiles) were used to calculate summary statistics for each site (San Miguel Island, Santa Rosa Island, Santa Cruz Island, Anacapa Island, Newport Beach, and San Diego) across all time intervals as shown in Table 3. Throughout the study period, M. californianus collectively recorded warmer SSTs from west to east, with stepwise depletion in median δ¹⁸O_shell from San Miguel Island to Anacapa Island (Fig 5a). Differences in δ¹⁸O_shell by site were statistically significant (ANOVA, F_5,2658 = 252.7, p < 0.001). Tukey HSD revealed significant differences between all sites in medium and long δ¹⁸O_shell profiles across the study interval (S1 Table). Overall, mussels from Newport Beach and San Diego recorded the warmest δ¹⁸O_shell signals. Despite most of its shell samples originating within the late Holocene (since 4200 BP), Santa Cruz Island had the widest range of δ¹⁸O_shell values, which is likely a product of its location and oceanographic setting. The California Current delivers cool water to western Santa Cruz Island, while SSTs on the eastern portion of the island are influenced by the warmer Davidson Current and Southern California Countercurrent. The dynamic nature of multiple oceanographic currents at this site may explain greater δ¹⁸O_shell intra-island variability for Santa Cruz Island mussels [72].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Box plots showing stable isotope ranges of each site comprising all shells with five or more subsamples from all Holocene intervals.

Black bar is the median value. (A) Oxygen isotope data from all shells with five or more subsamples for San Miguel Island (SMI), Santa Rosa Island (SRI), Santa Cruz Island (SCI), Anacapa Island (ANI), Newport Beach (NB), and San Diego (SD). Y-axis is inverted to match directionality of temperature proxy as indicated by SST arrow. (B) Carbon isotope data from all shells with five or more subsamples for the same sites. Arrow shows expected directionality of a potential δ¹³C upwelling signal.

https://doi.org/10.1371/journal.pone.0302945.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Summary statistics for all shells with five or more subsamples, sorted by site.

https://doi.org/10.1371/journal.pone.0302945.t003

There is an east-west δ¹³C_shell gradient among the Channel Islands shells. Farthest offshore, San Miguel Island mussels record the lowest δ¹³C_shell values, and overall δ¹³C_shell values increase towards the mainland, with Anacapa Island mussels recording the highest δ¹³C_shell values (Fig 5b). Newport Beach and San Diego mussel shells were isotopically lighter than shells from the eastern Channel Islands, but since the mainland δ¹³C_shell values represent modern mussels only (live-collected in the 2000s) and the island δ¹³C_shell values represent all time points throughout the Holocene, this offset may be a temporal trend rather than geographic. Anthropogenic burning of fossil fuels has changed atmospheric chemistry, and in turn, lowered the δ¹³C composition of seawater DIC and marine biogenic carbonates [58, and references therein]. Additionally, recent increases in upwelling strength and greater nearshore productivity have been documented along coastal mainland California [103], which may also be reflected in the lighter δ¹³C_shell composition of the modern mainland shells. As expected, modern shells were collectively depleted relative to the midden shells (Welch Two Sample t-test, t = -18.589, df = 2010.5, p < 0.0001). Our dataset contains no modern δ¹³C_shell values from Santa Cruz Island or Anacapa Island, but when comparing δ¹³C_shell of modern shells only, Santa Rosa and San Miguel Island shells maintained lower δ¹³C_shell values than the modern mainland shells. While significant terrestrial influences, such as plant decomposition, river input, or agricultural runoff, would likely result in isotopically lighter carbon rather than the higher mainland δ¹³C_shell values found here, it is also possible that the collective analysis of aggregated δ¹³C_shell values obscures the influences of terrestrial and freshwater systems. Collectively, δ¹³C_shell in intertidal mussels may serve as a better indicator of relative oceanographic conditions driven by δ¹³C of the DIC across a geographic region rather than as an indicator of individual events (e.g., a freshwater input pulse due to a storm). The reliability of δ¹³C_shell as an environmental proxy is limited due to the significant and complex role of metabolic δ¹³C incorporation into the shell as well as the competing environmental influences of upwelling, productivity, salinity, freshwater input, and the Suess Effect [56, 58].

Geographic variability by millennium.

We binned oxygen isotope data from shells with five or more subsamples (medium and long profiles) by millennium for each island and converted these values to SST (°C) using Eq 1 to produce a record of temperature snapshots throughout the Holocene. Santa Rosa Island offered the most temporally continuous record with data points spanning every millennium from 9000 BP through the present (Fig 6). Temperatures recorded at Santa Rosa Island were highest (> 14°C) between 9000–8000 BP and lowest (11.67°C) between 4000–3000 BP (Table 4). Both the geologically oldest and youngest shells were from San Miguel Island (8800 BP and 2005 CE), allowing for a comparison between two snapshots at both ends of the record. Inferred San Miguel mean SST was low (11.99°C) during the early Holocene and was overall higher in 2005 CE (12.92°C), although these differences are based on uneven distributions of data and were not statistically significant. Anacapa Island had only one snapshot at 3010 BP but recorded a wide range of temperatures (11.08° to 24.76°C; Table 4). Oxygen isotope data from the three westernmost islands (San Miguel Island, Santa Rosa Island, and Santa Cruz Island) within 6000–5000 BP preserved the geographic east-west SST gradient that characterizes the northern Channel Islands today, indicating that mussels did not collectively record fine-scale changes in California Current and sub-current forcing at 6000 BP that other models and records have documented [104]. The late Holocene is well represented at Santa Cruz Island with snapshots from the past three millennia. Calculated mean temperatures for each millennium at Santa Cruz Island exhibit warm-cool millennial scale variability: 14.13°C during 6000–5000 BP, 13.77°C during 3000–2000 BP, 14.23°C during 2000–1000 BP, 13.05° in midden shells from the past 1000 years, and 14.6°C in modern shells (Table 4). Temperature variability was highest at Santa Cruz Island within 2000–1000 BP (σ = 2.71°C) and lowest at Santa Rosa Island within 3000–2000 BP (σ = 1.34°C). Shells from Santa Cruz and Anacapa Islands recorded the highest overall maximum inferred temperatures, all three of which occurred in the late Holocene (Table 4). San Miguel Island reported the lowest overall minimum inferred temperature (6.29°C) at 8800 BP.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Millennial scale δ¹⁸O_shell-inferred temperature variability for each island.

Data were binned by millennium for each island to produce a range of temperature snapshots. Black bar is the median value.

https://doi.org/10.1371/journal.pone.0302945.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Summary statistics for all shells with five or more subsamples, sorted by both island and millennium.

https://doi.org/10.1371/journal.pone.0302945.t004

Seasonal variability

Seasonality revealed by medium and long profiles.

Shells with long profiles (15 or more subsamples per individual) were used to examine seasonal oscillations. Long oxygen isotope profiles were truncated to the final 40 mm to compare profiles of similar lengths. These profiles exhibited sinusoidal seasonal variability, with some individuals recording more pronounced seasonal extremes than others (Fig 7). Long-profile individuals (n = 28) ranged in geologic age from 8800 BP through 2005 CE, yet all truncated 40-mm profiles recorded at least one annual cycle with a mean of 2.6 ± 0.83 (1σ) warm seasons. Estimating the number of warm seasons serves as a way to estimate the age of an individual at the time of its death if the subsampling profile captures the full ontogenetic history of the shell. However, since the shells analyzed here are primarily archaeological shells harvested by humans, the inferred number of warm seasons indicates the age at which shells were typically harvested for consumption rather than the average lifespan of a mussel. The full potential lifespan of M. californianus is unknown but may range from 10 to as many as 100 years [31].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Individual oxygen isotope profiles of all 28 long profiles (15+ subsamples) from Channel Islands shells.

Each profile is from a unique shell with collection year or calibrated age in years before present (BP) labeled at the top of the plot. Dashed line represents the mean oxygen isotope value recorded by each individual shell. Red shaded bars are estimations of warm seasons. Y-axis is inverted and the scale is uniform for all plots.

https://doi.org/10.1371/journal.pone.0302945.g007

On average, the 28 long-profile individuals recorded a calculated SST range of 5.4°C (δ¹⁸O_shell = 1.35‰), which matches the typical annual temperature range for nearshore waters surrounding the Channel Islands [54, 74]. The highest SST was recorded by an individual at 7315 BP. The warmer inferred conditions are consistent with the oxygen isotope record of G. bulloides planktic foraminifera at ODP Site 893 A/B in Santa Barbara Basin for this same age range [105]. All M. californianus-inferred temperatures lower than 9.5°C were recorded between 5200 and 5099 BP, which was unexpected since this interval is considered to be a warm period for this region [105]. Local-scale changes in ocean circulation may have played a role in this misalignment, particularly since small-scale geographic variability appears to have an influence on shell chemistry (Fig 5).

Seasonality of carbon isotope profiles is less distinguishable. Out of all 28 long-profile shells, only one individual showed clear seasonal cycles that closely aligned with its oxygen isotope seasonal oscillations, and only a few individuals exhibited generally covarying δ¹⁸O_shell and δ¹³C_shell profiles (S2 Fig). While δ¹³C_shell profiles alone are challenging to interpret, strong correlation between δ¹⁸O_shell and δ¹³C_shell can be used to distinguish between the influences of rainfall and upwelling; a positive δ¹⁸O-δ¹³C relationship is indicative of freshwater input, while δ¹⁸O and δ¹³C are negatively correlated during upwelling [106, 107]. The individual with strongly expressed seasonal δ¹³C_shell cycles was collected in 2005 CE from San Miguel Island (Fig 2). In this individual, the oxygen isotope minima (inferred warm seasons) aligned remarkably well with carbon isotope minima (inferred low upwelling), particularly in the last 50 mm of the record (Fig 8). The strong seasonal δ¹³C_shell cycles in this individual are comparable to those found by Killingley and Berger (1979) in M. californianus shells from San Diego [59]. However, the vast majority (27 out of 28 individuals) lacked δ¹³C_shell profiles that aligned with seasonal interpretations of δ¹⁸O_shell cycles, and/or exhibited statistically insignificant relationships between δ¹⁸O_shell and δ¹³C_shell. Similarly, Ferguson et al. (2013) found that δ¹³C_shell in southern California mussels could not be used to reliably reconstruct upwelling, likely due to metabolic carbon contributions to δ¹³C_shell relative to δ¹³C of DIC [45].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. The last 50 mm of δ¹⁸O and δ¹³C profiles in a M. californianus shell collected from Santa Miguel Island in 2005 CE.

In the top plot (δ¹⁸O), y-axis is inverted to match directionality of temperature proxy as indicated by SST arrow. Gray bars in both plots denote δ¹⁸O-inferred warm seasons. In the lower plot, arrow shows expected directionality of a potential δ¹³C upwelling signal.

https://doi.org/10.1371/journal.pone.0302945.g008

We performed OLS regression to investigate whether shells with a particular range of δ¹⁸O_shell values also contained a particular range of δ¹³C_shell variability [108], which could help distinguish between an upwelling signal (high δ¹⁸O_shell, low δ¹³C_shell) and a freshwater input (high δ¹⁸O_shell, low δ¹³C_shell) signal [106–107]. Only three of the 28 long-profile individuals analyzed here exhibited an inferred upwelling signal, or a statistically significant intra-individual negative δ¹⁸O_shell-δ¹³C_shell correlation (Shells L, E, and Z in S2 Fig; Shell L: OLS linear regression, R² = 0.67, F_1,13 = 26.86, p < 0.001, Pearson’s correlation test, t = -5.1828, df = 13, p < 0.001; Shell E: OLS linear regression, R² = 0.51, F_1,15 = 15.7, p = 0.001, Pearson’s correlation test, t = -3.9622, df = 15, p = 0.001; Shell Z: OLS linear regression, R² = 0.38, F_1,13 = 8.118, p = 0.01, Pearson’s correlation test, t = -2.8492, df = 13, p = 0.01) and none of the individuals contained statistically significant positive δ¹⁸O_shell-δ¹³C_shell relationships. The absence of a consistent upwelling signal across multiple individuals could be due to the input of metabolic carbon or the result of growth cessation during upwelling conditions. The three individuals that shared statistically significant negative δ¹⁸O_shell-δ¹³C_shell relationships (Shells L, E, and Z in S2 Fig) have different ¹⁴C ages (7315 BP, 5200 BP, and 4256 BP, respectively) and are from different islands (Santa Rosa Island and Santa Cruz Island). The lack of consistent δ¹⁸O_shell-δ¹³C_shell relationships among mussels from a particular site or time interval is likely influenced by sampling strategy, calcification rate, metabolic activity, and variability in micro-environmental conditions across the dynamic intertidal zone. High-resolution isotope sampling within this species and analysis of additional species from the same and adjacent environments are required to determine which factors influence δ¹⁸O_shell-δ¹³C_shell relationships in molluscan carbonate and obscure upwelling signals.

While seasonality is visually apparent in the long δ¹⁸O_shell profiles (Fig 7), the number of subsamples required to capture seasonal variability depends on the growth rate, and therefore ontogenetic age, of the individual. Typically, ~ 3 mm is thought to represent 1–2 months of growth, and 8–10 closely spaced subsamples are required to identify an annual cycle [51, 74]. Although evenly spaced 2–3 mm increments is a conventional subsampling approach based on field observations of California mussels, we emphasize the need for consistently updated, long-term, and site-specific field studies of M. californianus shell growth rates because this species exhibits highly variable growth rates across sites and through time [29, 40, 44]. Since we observed that profiles with five subsamples contained a local minimum and a local maximum in most of our oxygen isotope profiles (i.e., interpreted as a warm season and a cool season, respectively), we used all individuals with five or more subsamples to infer trends over broader spatial and temporal scales (Figs 4–6). This greatly expanded our sample sizes (both n = number of individual shells and n = number of subsamples) and allowed for additional temporal snapshots and geographic comparisons. Individuals with fewer than five subsamples were excluded from broader scale interpretations since these short, end-of-life profiles capture only a season or less of growth, which could skew a long-term record. For example, if an individual collected in late summer is sampled at three evenly spaced points at its growing margin, its subsamples would likely only capture summer warming. This individual would introduce bias into an aggregated long-term record by indicating that conditions were warmer during its lifespan, so it is important to consider the time of year and length of time represented by the sampling method and interpret shell profiles accordingly.

Millennial variability

Holocene climate in North America is characterized by millennial-scale temperature variability [109]. We aimed to determine whether M. californianus from the Channel Islands collectively recorded such millennial-scale variability despite the additional patterns that emerged in our record, such as local-scale oceanographic differences among the islands (Fig 5) or the ontogenetic and seasonal variability within individuals (Fig 7). To investigate millennial trends, we categorized the δ¹⁸O_shell data into 1000-year time bins using individuals with five or more subsamples (Fig 9), which revealed that median δ¹⁸O_shell values oscillate every 1000 years, with the highest median δ¹⁸O_shell values occurring during 9000–8000 BP and 7000–6000 BP (0.7‰ and 0.73‰, respectively) and the lowest median δ¹⁸O value (0.05‰) occurring at the start of the late Holocene (4000–3000 BP). ANOVA and Tukey HSD showed that millennial δ¹⁸O_shell values were statistically different from the previous millennium, except for from 8000 to 7000 BP and 3000 to 2000 BP. The statistical comparisons between each millennium are in the supporting information (S2 Table). Median δ¹⁸O_shell for modern (non-midden, live-collected) shells is 0.32‰, equal to median δ¹⁸O_shell values from 6000–5000 BP and 2000–1000 BP and lower than the median δ¹⁸O_shell value of six of the nine millennia since 9000 BP (Fig 9a). There has been an overall decline in collective median δ¹⁸O_shell values from the early Holocene to the late Holocene, indicating an overall warming and freshening signal over the three intervals (Fig 9b). Interestingly, the Channel Islands mussels collectively recorded millennial-scale shifts in addition to finer-scale trends (e.g., ontogeny, seasonality, local oceanography, micro-habitat), indicating that M. californianus shells provide a multi-scaled view of past environmental variability depending on the categorization of the data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. All data points from all Channel Islands shells with five or more subsamples binned by time interval.

(A) Each time bin on the x-axis represents all samples within that millennium (e.g., 1000 represents all non-modern shells from 0–1000 BP, etc.). Gray points denote median δ¹⁸O_shell value of each time bin. Error bars represent three times the standard error (3SE) from the median for visibility. (B) All data points from Channel Islands shells in oxygen and carbon isotope space over the Holocene. Centroids (medians) are plotted for each sub-epoch. Both horizontal and vertical error bars represent 3SE from the median. Arrow shows expected directionality of a potential δ¹³C upwelling signal. Seven outliers with values greater than 2‰ or less than -2‰ were omitted from the plot.

https://doi.org/10.1371/journal.pone.0302945.g009

Comparison to regional climate records

We generated a snapshot-based Holocene M. californianus stable isotope record and compared it to other published coeval climate records for the southern California region to determine whether this intertidal species is a comparable archive of broader climate patterns. We used all Channel Islands mussel shells with five or more subsamples (medium and long profiles) to calculate median δ¹⁸O_shell and δ¹³C_shell values for each year of the record to generate 58 annual snapshots over the Holocene (Fig 10). We calculated annual snapshots based on a San Diego study that found M. californianus shells to serve as reliable records of δ¹⁸O_shell-inferred mean annual temperature in southern California [42]. We aimed to assess the reliability of the aggregated M. californianus shell isotope record when analyzed in this way. The annual snapshots were plotted along with 1000-year δ¹⁸O_shell medians (as in Fig 9a) as well as previously published terrestrial precipitation records [110] (Fig 10a) and marine sediment core records [105, 111] (Fig 10b and 10c) from southern California sites to determine whether M. californianus can be interpreted similarly for use as a comparable climate record for this region.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. Age versus climate proxy data in calibrated years before present.

(A) Standardized percent clay as a precipitation proxy from Silver Lake in the Central Mojave, California (Kirby et al., 2015) [110]. (B) Percentage of warm species (Globigerinoides ruber in orange, Globoturborotalita rubescens in red, and Neogloboquadrina dutertrei in yellow) from Santa Barbara Basin ODP Site 893 (Fisler and Hendy, 2008) [111]. (C) Ice-volume corrected oxygen isotope SST record from Globigerina bulloides planktic foraminifera at ODP Site 893 A/B (Kennett et al., 2007) [105]. Y-axis is inverted. Dashed line represents linear trend. (D) Carbon isotope record from M. californianus shells analyzed in this study with 5 or more subsamples with the median for each year of the record plotted. (E) Oxygen isotope record from M. californianus shells analyzed in this study with five or more subsamples with median for each year of the record plotted in blue. Y-axis is inverted. Dashed line represents linear trend. Large gray points represent 1000-year median δ¹⁸O_shell values as in Fig 9. Blue stars on x-axis denote timing of flood events identified from sediment cores in Santa Barbara Basin by Du et al. (2018) [112]. Shaded red bars denote warm periods and shaded gray bar denotes variable conditions inferred by Kennett et al. (2007) [105].

https://doi.org/10.1371/journal.pone.0302945.g010

The annually resolved M. californianus oxygen isotope record (Fig 10e) does not closely match inferred temperatures from δ¹⁸O_foraminifera (Fig 10c) from varved sediments in the Santa Barbara Basin. There are three factors contributing to the misalignment between the M. californianus annual inferred SST record (Fig 10e) and the sediment core record (Fig 10c). Firstly, there are significant environmental differences (e.g., temperature, salinity, nutrients, etc.) between the offshore habitat of the planktic foraminifera G. bulloides and the intertidal habitat of M. californianus. Secondly, there are also major differences between the temporal resolutions of each record; varved marine sediments provide a much more temporally continuous record than M. californianus shells, while M. californianus presents a sparser record of longer-lived shells with nested seasonality and ontogenetic variability. Annual treatment of δ¹⁸O_shell data, such as calculating an annual average or annual median from multiple subsamples and individuals, may obscure the seasonal and ontogenetic patterns that we identified in individual shells. Sampling methods can also play a significant role in observed δ¹⁸O_shell patterns. Lastly, record misalignment is likely influenced by the chronological uncertainty resulting from differences in age model calibrations. While all archaeological shells were from a dated section of a midden, the source material and calibration curves varied. In some cases, the primary authors did not report the ΔR, or used a variety of ΔR values for the purposes of their study. The most commonly used ΔR value was 225 ± 35 years [52, 72, 105]. Annual binning of data as in Fig 10e is based on the author’s original interpretations for the purpose of demonstrating challenges and identifying possible sources of inconsistencies between the mussel record and the sediment core record.

Despite the challenges of interpreting an aggregated M. californianus geochemical dataset, it is still useful to examine the full M. californianus record at coarser resolution using large sample sizes, as supported by the statistically significant millennial-scale oscillations and geographic trends that emerge when M. californianus shells are analyzed collectively (Figs 5 and 9). Additionally, both the δ¹⁸O_shell and δ¹⁸O_foraminifera records underscore recent warming since ~ 1000 BP. In some cases, low annual δ¹⁸O_shell values, which can be indicative of low-salinity conditions, align remarkably well with Santa Barbara Basin flooding events [112], although the alignment is not consistent enough throughout the Holocene to indicate that mussels experience and reliably record freshwater input pulses as they calcify (Fig 10e). The lack of a discernible freshwater signal across the aggregated record corroborates the absence of any statistically significant positive δ¹⁸O_shell-δ¹³C_shell correlation within individual shells (S4 Fig).

Summary of confounding factors

Several environmental and biological factors influence shell chemical compositions, and therefore impact the fidelity of stable isotope data. A complete and reliable shell stable isotope record requires constant and consistent shell growth and high-resolution subsampling along the ontogenetic trajectory of the shell. Constant and consistent growth is not likely in M. californianus due to interruptions in its highly dynamic intertidal habitat and its own biological processes. Therefore, to optimize the paleoceanographic and paleobiological utility of the M. californianus stable isotope record, it is necessary to examine the confounding factors and develop methods to extract the signal or parameter of interest (Table 5). To reconstruct temperature seasonality from an individual shell, is it critical to determine that the controls on δ¹⁸O_shell are temperature-dependent and that the organism was growing throughout the annual cycle. In many cases, environmental parameters (e.g., temperature, seasonality, salinity) interact with biological factors (e.g., calcification, metabolism) to obscure or exaggerate signals. For example, growth cessation during cold conditions would obscure a winter signal, while rapid growth during warm conditions would result in an overrepresented summer signal. Based on our synthesis of stable isotope shell data from a variety of sites, ages, and sampling methods, we identified four recurring factors–calcification, metabolism, ontogeny, and habitat–and proposed corresponding methodological recommendations to investigate and characterize them more accurately (Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Identified factors that may impact environmental interpretations from M. californianus stable isotope data.

https://doi.org/10.1371/journal.pone.0302945.t005

Despite the limitations with environmental interpretations, the M. californianus stable isotope record did align with expected spatial oceanographic conditions and seasonal cycles when the sampling methodology allowed. Long δ¹⁸O_shell profiles provided sufficient temporal context to reveal sub-annual and seasonal trends over the course of a mussel’s life, but short profiles indicate that fine-resolution variability is difficult to extract from 2–4 subsamples from one margin of the shell. For example, mussel shell chemistry across many medium and long subsample profiles from modern shells did appear to reflect the geographic temperature gradient between mainland southern California and the northern Channel Islands. This δ¹⁸O_shell record implied cooler conditions from east to west within the Channel Islands, which matches known seawater temperature datasets and reflects the competing influences of the cold California Current and the warmer California Undercurrent and Davidson Current.

Additionally, we found that five sequential subsamples per individual shell revealed a local δ¹⁸O_shell maximum and a local δ¹⁸O_shell minimum from which a winter low temperature and a summer high temperature could be inferred, respectively. While a five-sample profile appeared to estimate one annual temperature cycle, we emphasize the value of collecting as many sequential subsamples along the growth trajectory of the shell as possible to account for ontogenetic variation and capture multi-season or multi-year δ¹⁸O_shell profiles. A longer δ¹⁸O_shell profile allows for the identification of annual minima and maxima over multiple years, and sampling multiple individuals allows for the cross-correlation and corroboration of seasonal cycles within the same region and could potentially aid in the identification of ontogenetic or metabolic influences on shell growth. Season of harvest information will also be more reliable when there is a multi-seasonal or multi-annual record experienced by the mussel to contextualize end-of-life signals. Conversely, short δ¹⁸O_shell and δ¹³C_shell profiles comprising only a few subsamples at the terminal edge of the shell are less useful for paleo-seasonal and paleoceanographic reconstruction because they reflect shorter periods of time (e.g., weeks to months) and therefore lack the temporal duration required to identify the organism’s life-history patterns and reconstruct broader climate conditions occurring over multiple seasons or years. End-of-life shell chemistry in an aging individual is not necessarily reflective of ambient conditions due to ontogenetic declines in calcification. We also observed high δ¹⁸O_shell variability among specimens growing at the same time and location, which reduces confidence in δ¹⁸O-inferred temperature calculations from shells with terminal-edge sampling only. Lastly, we found inconsistent and unexpected δ¹⁸O_shell-δ¹³C_shell relationships, affirming that the influences of sampling strategy, growth rate, and conditions at a given site or time interval can severely impact shell chemistry, and subsequently, paleoenvironmental interpretations.