Indigenous marine stewardship and season of harvest for California mussels (Mytilus californianus) along California's Lost Coast, U.S.A

1 Introduction

Societies around the world have developed various forms of tenure systems—comprising rights, rules, and agreements to govern access to land and resources—that are often deeply embedded in social and cultural frameworks (Garnett et al., 2018; Kasimbazi, 2017; Sauls et al., 2022). Lepofsky and Caldwell (2013, p. 4) list several tenure strategies related to Indigenous marine resource and environmental management systems on the Northwest Coast: (1) rights to harvest specific species; (2) ownership and control of harvesting locales and features; (3) restricted timing and season of harvest; and (4) limits on catch size and who is allowed to harvest. In California, archaeological records emphasize the importance of marine ecosystems throughout the Holocene (Braje et al., 2012; Erlandson et al., 2011, 2020; Hildebrandt and Levulett, 1997; Reeder-Myers et al., 2015), yet specific cultural protocols and practices that Indigenous societies used to steward the long-term productivity of marine environments is broadly understudied in archaeological practice, with a few exceptions (Apodaca et al., 2024; Grone, 2025; McFarland et al., 2025; Sanchez et al., 2023).

As a complement to traditional ecological knowledge, archaeology offers a scientific framework for evaluating past marine stewardship practices by analyzing species abundance and size distributions in zooarchaeological assemblages through time (Lepofsky and Caldwell, 2013) and inferring seasonal harvesting patterns through stable isotope analysis of mollusks shells (Killingley, 1981; Kwiecien et al., 2022; Shackleton, 1973; West et al., 2018). I define stewardship here as a set of cultural practices and protocols that help improve resource abundance while enhancing ecosystems in a way that benefits future generations. Understanding traditional practices that prioritize long-term productivity over short-term gain is important to informing contemporary strategies for resource conservation and environmental protection.

California mussels (Mytilus californianus) are ubiquitous in archaeological assemblages along the west coast of North America, providing a record of at least 12,000 years of coastal occupation patterns (Erlandson et al., 2008, 2009, 2020; Braje et al., 2012), seasonal subsistence strategies (Apodaca et al., 2025; Eerkens et al., 2013; Jazwa et al., 2015, 2020; Jazwa and Kennett, 2016; Jew et al., 2013, 2014; Jones et al., 2008), and environmental conditions (Ferguson et al., 2013; Flores, 2017; Jazwa et al., 2012; Killingley and Berger, 1979; Pfister et al., 2011; Rick et al., 2006; Vriesman et al., 2024) in California. Because mussels are frequently fragmented in archaeological contexts, researchers use size templates or regression techniques to estimate original shell size (White, 1989; Campbell and Braje, 2015; Glassow et al., 2016; McFarland et al., 2023; Singh and McKechnie, 2015). These reconstructions are useful to measure average size changes and assess the health of mussel populations through time, particularly in relation to overharvesting or stewardship practices.

In this study, I examined stewardship practices by testing for the seasonality, frequency, and intensity of harvests of California mussels. To do so, I measured over 9,000 fragmentary mussel shells containing the umbo, collected from eight archaeological sites along the Lost Coast, spanning from Spanish Flat to Shelter Cove in Northwest California (Figure 1). These sites are located south of Cape Mendocino within the King Range National Conservation Area, approximately 450 km north of San Francisco Bay. I developed a multiple linear regression model based on a modern pooled sample of whole shells from two environments with varying wave energy—Mussel Rock and Punta Gorda—to predict total shell length using key morphological features of fragmented archaeological mussels. I then assessed temporal shifts in the predicted average shell sizes across 100-year intervals using radiocarbon age-depth models spanning the past 2,700 years. I compared observed mussel size distributions to theoretical harvesting models developed by Whitaker (2008), applying Kolmogorov–Smirnov (K-S) tests to identify the most likely harvesting strategies, considering various harvest intervals and intensities. Additionally, I developed a model to simulate the average energy yield (kcal) and mean mussel size per harvest cycle across various harvesting strategies, integrating maximum sustainable yield thresholds to assess the long-term viability of each strategy over a 50 year period. Finally, I analyzed 192 carbonate samples from 40 mussel shells using stable isotope analysis to infer the most probable season of shellfish collection at Spanish Flat. These analyses contribute to a broader understanding of seasonal resource scheduling and subsistence strategies.

2 Background

2.1 Cultural setting

The study area lies within the ancestral lands of two Indigenous groups: the Bettōl’ (Mattole) and the Southern Sinkyone (Figure 1). These dialects are related to the Pacific Coast Athabaskan branch of the Na-Dené language family who shares linguistic features with geographically distant communities in Southeast Alaska and Northwest Canada (Golla, 2011). Na-Dené includes Northern Athabaskan languages as well as Tlingit and Haida, highlighting deep historical and linguistic connections across coastal Alaska, British Columbia, and Northwest California (Sapir, 1915, 1921).

Little direct evidence is available about how Indigenous peoples managed marine ecosystems in the region of study prior to European contact. However, ethnographic records offer insights into tenure systems and harvesting practices, even though they may be biased and contain fragmented pieces of history (LeCompte, 1987; Whitaker, 2012). Indigenous communities often pass down cultural protocols through informal knowledge that guide relationships with land and sea. Ethnographic informants likely did not share this knowledge unless asked explicitly, as such practices are often second nature or embedded in unspoken laws. As a result, they may perceive their actions simply as the customary way things have always been done, without explicitly framing them in terms of stewardship.

Nevertheless, cultural protocols embedded within traditional tenure systems played a vital role in regulating access to land and resources along the Lost Coast. These systems are similar to those documented by Lepofsky and Caldwell (2013) along the Northwest Coast, by regulating: (1) ownership and control of harvesting locales and features; (2) timing and season of harvest; and (3) size selective harvesting. In the following, I discuss these cultural practices for Indigenous groups along the Lost Coast using primary ethnographic source materials from Hewes (1940) field notes, reflecting the voices and experiences of his informants rather than his own interpretations. I cite other ethnographic sources for the study area where appropriate.

According to one informant, the Bettōl’ never constructed fish weirs, nets, or dams as permanent fixtures along rivers or streams. This impermanence was intentional to ensure access to spawning grounds further upstream and to maintain equitable access for inland communities, reflecting a broader ethic of shared stewardship. Ownership of specific harvesting locales was deeply rooted in familial and intergenerational ties. Offshore rocks used for gathering seabirds and eggs or sea lion hunting were not communal property but were said to be owned by specific families since the “time the world was made” (Hewes, 1940, p. 4). Individuals passed these rights down through generations, often through partnership-like arrangements with other families. Family members had access to these spaces, but outsiders needed explicit permission to harvest. Anyone who harvested without approval committed a serious offense, similar to poaching or trespassing, which could escalate into intergroup conflict or war (Hewes, 1940, p. 15). Even in resource-rich areas, like the Lost Coast, communities regulated harvesting through kinship-based tenure systems that clearly defined who had the right to gather and when.

Knowledge of seasonal and ecological changes also guided harvesting rights and practices, with communities aligning their efforts to seasonal cycles in behavioral, migratory, and reproductive patterns among different organisms. For instance, coastal harvesting increased when salmon were seasonally unavailable. When resources were plentiful, people worked hard to gather food in mass for winter storage (Hewes, 1940, p. 13). Seafood, such as shellfish and fish, were considered year-round sources of food and dried for winter use (Hewes, 1940; Kroeber, 1959). While shellfish poisoning was reportedly rare south of Cape Mendocino, people still collected larger mussels from the low intertidal zone whenever tides allowed to avoid potentially getting sick, as mussels higher in the water column were smaller and considered less edible except occasionally in the winter (Hewes, 1940; Kroeber and Barrett, 1960).

In spring, families gathered young seabirds and eggs, hunted sea mammals, and fished for smelt that came close to shore to spawn. In autumn, families gathered acorns, wild oats, and seeds—described as filling “six barrels” worth for one house—and preserved deer meat under boards in homes for winter use (Hewes, 1940, p. 12). Although seasonal abundance typically provided plenty of food, individuals were taught to harvest only what they needed and to give extra away to visitors, as it was important to be generous (Hewes, 1940, p. 15). Reliable resources like shellfish and fish helped supplement winter supplies and served as valuable trade goods for inland communities during lean years (Davis, 1961; Hewes, 1940; Nomland, 1938). These practices reflect a deeper worldview rooted in reciprocity, promoting both social relationships and ecological resilience.

2.2 Environmental setting

Paleoenvironmental records from terrestrial and marine contexts in Northwest California reveal notable climatic variability during the Late Holocene, after 4,200 cal BP (Adam and West, 1983; Kirby et al., 2023; Knight et al., 2024; Palmer et al., 2023; Voelker et al., 2018; West, 1990). Offshore diatom and pollen records from a deep-sea sediment core (TN062-O550), located approximately 33 km offshore from the study area, indicate major climatic and oceanographic shifts that potentially influenced mussel population dynamics and harvesting along the Lost Coast. These include: (1) intensified riverine discharge and increased sedimentation in coastal environments from a rise in precipitation beginning around 3,600 cal BP; (2) a rapid change in coastal climates and oceanography including more frequent cool-warm cycles after 2,800 cal BP; and (3) intensified spring upwelling after 2,600 cal BP, which boosted marine productivity by delivering cooler, nutrient-rich waters to nearshore ecosystems (Addison et al., 2018; Barron et al., 2003, 2018).

These long-term changes in oceanographic conditions reflect variable climatic regimes in northern California, where short-term El Niño-Southern Oscillation events modulated interannual precipitation and storm activity, and the broader Pacific Decadal Oscillation imposed multi-decadal shifts in sea surface temperatures (SST) and atmospheric circulation (Barron et al., 2003; Barron and Anderson, 2011; Koutavas and Joanides, 2012). Since 3,500 cal BP, alternating cool upwelling-driven summers and warm wet winters structured both terrestrial and marine environments in northern California linked to these large-scale oceanographic processes centered in the North Pacific (Barron et al., 2003; Barron and Anderson, 2011; Kirby et al., 2023; Palmer et al., 2023).

California mussels are filter feeders that rely on ocean upwelling to deliver nutrients (phytoplankton) to the surface, where they can feed to effectively sustain growth, reproduction, and overall metabolic function (Coe and Fox, 1942). Greater marine productivity following 2,600 cal BP likely benefited mussel beds along the Lost Coast; however, increased sedimentation may have disrupted mussel recruitment and survivorship by altering water quality (Denic and Geist, 2015; Lummer et al., 2016). The Lost Coast is highly prone to both coastal and stream erosion, shaped by tectonic activity, intense winter rainfall, and complex hydrodynamic processes (Dumitru, 1991; Merritts and Vincent, 1989; Moon et al., 2018).

Spanish Flat lies on a slightly elevated marine terrace intersected by numerous stream networks, making it especially vulnerable to high rates of sediment runoff from coastal erosion and landslides during seasonal downpours. The curvature of the coastline periodically deflects prevailing winds and currents away from nearshore zones at Spanish Flat, slowing water movement and accelerating sediment buildup, potentially influencing growth rates for mussel populations. In contrast, Shelter Cove is located on a prominent rocky bluff exposed to prevailing ocean currents. Its resistant geology and limited stream input helps buffer it from erosion and sediment runoff, potentially supporting larger and more resilient mussel beds.

Additionally, water circulation patterns shaped by nearshore bathymetry can create localized variations in SSTs and wave exposure. These differences may alter the thermal tolerances and growth rates of mussel populations, while shifts in SST gradients could also affect mussel larval dispersal and reproductive success (Waite and Sorte, 2022; Widdows, 1991). More frequent cool-warm phases during the Late Holocene may have also altered the size composition of mussel beds and shifted the timing of harvests, especially for mussel populations more vulnerable to environmental constraints, like at Spanish Flat.

2.3 Harvesting strategies: plucking vs. stripping

Plucking and stripping are two mussel harvesting techniques commonly documented in ethnographic records from California, Australia, and South America (Jones and Richman, 1995; White, 1989). Stripping involves the removal of entire mussel sections en masse, facilitated by byssal threads that bind mussels to each other and to the substrate. This method removes mussels of a broad range of sizes and requires extended fallow periods to allow beds to regenerate to harvestable size. Plucking involves the selective removal of larger individuals, typically leaving a proportion of the original stock intact for growth and future harvests. Contemporary ecological studies show that regular harvesting reduces intra-bed competition and promotes individual growth, underscoring the ecological benefits of both selective plucking and cyclical management regimes (Yamada and Peters, 1988).

Whitaker (2008) introduced a statistical method for identifying plucking or stripping and assessing the frequency and intensity of California mussel harvesting in the archaeological record, offering a framework for evaluating potential stewardship practices. His analysis of a coastal rock shelter near Punta Gorda along the Lost Coast in Northwest California revealed evidence of a 24-month stripping cycle between 730 and 530 cal BP, indicating that inhabitants employed a biennial harvesting strategy designed to maximize long-term yields rather than immediate returns, which likely helped prevent size diminution associated with more frequent harvesting. However, Whitaker (2008, p. 1,121) acknowledges that data are lacking to determine if this was the result of stewardship practices or epiphenomenal outcomes from short-term maximizing behavior by harvesting at a patch every other year. His study has also faced some criticism as he used White’s (1989) shell template to measure mussel size, potentially undermining the reliability of his results (Tatlow et al., 2022, p. 135).

As a follow-up study, Kievman (2023) recently demonstrated that inhabitants of CA-HUM-182—a coastal site near Shelter Cove, approximately 35 km south of Punta Gorda—used a plucking strategy to harvest progressively larger California mussels between 730 and 260 cal BP, suggesting potential stewardship practices that maintained mussel bed productivity over multiple generations. Studies in central California have documented similar increases in California mussel sizes linked to either plucking or stripping strategies (Grone, 2025; Tatlow et al., 2022). These patterns are interpreted as potential indicators of low-impact or rotational harvesting practices that may have contributed to long-term mussel bed stability.

While these studies reference Whitaker’s (2008) model to compare mussel size profiles to plucking or stripping strategies, their interpretations are often based on visual comparisons rather than statistical testing. Whitaker (2008) explicitly considers both harvesting methods and the effects of fallow duration and harvesting intensity on mussel populations, offering a robust means of evaluating potential effects of stewardship practices or overharvesting. Although their morphometric and chronological analyses are valid, the absence of systematic comparisons to Whitaker's harvesting models limits the ability to confidently link observed size changes to specific harvesting strategies, considering fallow period and harvest intensity. As a result, it remains unclear whether observed increases in mussel sizes reflect stewardship practices or are incidental outcomes of broader foraging decisions or environmental dynamics. Research integrating accurate size reconstructions, fine-grained temporal analysis, and ecological modeling is needed to clarify the behavioral mechanisms underlying long-term shellfish productivity observed archaeologically.

2.4 Mussel size estimates and sustainability

Various studies have evaluated the accuracy and reliability of mussel size reconstructions using morphological proxies with regression analysis (Campbell and Braje, 2015; Glassow et al., 2016; McKechnie et al., 2015; Singh and McKechnie, 2015), which may vary based on environmental conditions (McFarland et al., 2023). Earlier studies often relied on size templates that estimated total shell length based on hinge width comparisons (Jones and Richman, 1995; Walden-Hurtgen, 2013; Whitaker, 2008; White, 1989); however, experimental work by Bell (2009) demonstrated that such templates yielded inconsistent and imprecise results between analysts. Using either of these methods, archaeologists have documented long-term resource depression in California mussels resulting from high harvesting pressures (Braje et al., 2007, 2012; Erlandson et al., 2008; Glassow et al., 2008; Jazwa et al., 2016; Jones, 1996), while relatively few have examined long-term stewardship (Grone, 2025; Kievman et al., 2023; Tatlow et al., 2022), with some attributing possible sustainability to non-conservative behaviors, epiphenomenal outcomes, or pseudo-aquacultural practices (Jones and Codding, 2019; Whitaker, 2008).

What many of these studies overlook is the question of what constitutes sustainable mussel harvesting and how such practices might be reflected in the archaeological record. According to Code of Federal Regulations (50 CFR § 600.310), stocks are typically considered overfished when standing biomass falls below 50% of minimum stock size threshold, signaling a critical decline in population health and long-term viability. This threshold marks a tipping point beyond which resource depression becomes increasingly likely, often driven by harvesting rates that exceed the maximum sustainable yield (NOAA Fisheries, 2020). Maintaining harvest pressures below this threshold may promote long-term mussel bed productivity, whereas exceeding it increases the probability of reduced population densities and smaller size distributions - archaeological patterns that are often associated with resource depression from either overharvesting or environmental stressors, or both.

3 Hypotheses and expectations

To investigate the long-term sustainability of mussel harvesting in Northwest California, I explore how Indigenous stewardship of California mussel beds shaped ecological outcomes through specific harvesting strategies related to marine tenure systems. My primary hypothesis is that Indigenous communities along the Lost Coast used selective, rotational harvesting practices to prevent overharvesting and enhance the long-term productivity of mussel beds. Specifically, I tested three interrelated hypotheses: (H1) communities used size-selective harvesting techniques to allow smaller mussels to grow to harvestable size; (H2) communities used seasonal, rotational harvesting among mussel beds to limit harvest intensity within sustainable yields and to increase fallow periods, allowing time for recovery and regrowth; and (H3) communities practiced sustainable harvesting to ensure the inheritance of productive mussel beds for future generations. If these hypotheses are correct, I expect the archaeological record to show patterns consistent with harvesting strategies as listed in Table 1.

Table 1. Archaeological expectations and inferences related to each hypothesis for this study.

4 Materials and methods

For this study, I analyzed data from eight archaeological sites along the Lost Coast, representing three permanent villages at Spanish Flat (CA-HUM-277, CA-HUM-279, CA-HUM-281) as well as two permanent villages (CA-HUM-182, CA-HUM-184), two logistical camps (CA-HUM-186, CA-HUM-248), and one food processing station (CA-HUM-185) at Shelter Cove. Archaeological evidence indicates more permanent residential settlement at Spanish Flat and Shelter Cove after 1,500 cal BP, followed by a diversification of site types after 700 cal BP, including core and satellite villages, campsites, and food-processing stations which are identified by the presence or absence of specific dietary, technological, and residential components (McFarland, 2025).

I selected these sites based on prior archaeological investigations documenting a long-term increase in coastal occupation since 2,700 cal BP (Levulett, 1985; Waechter, 1993). Additionally, the majority of these sites are located on the immediate coast within less than 50 m from the rocky intertidal, so I assume that harvesting occurred adjacent to where processing occurred and travel time is not a major consideration in the decisions influencing transport costs (Bettinger et al., 1997). Only CA-HUM-248 is located approximately 800 m (0.5 mi) from the coast, potentially influencing processing decisions, since the transport costs would be greater.

Shellfish assemblages from all eight sites also emphasize California mussels, with mussel shells comprising more than 70% of the shellfish assemblages at Shelter Cove (Table 2). A general increase in shell density during later periods suggests intensive coastal occupation and potentially greater harvesting pressures on mussels. Because all sites occur within broadly comparable coastal settings, I assume that the major taphonomic processes influencing shell preservation operated similarly across the archaeological assemblages. To account for potential variations caused by excavation or sampling differences, I normalize all shell weights by the total volume of excavated soil for each component, identified by distinct stratigraphic units, faunal assemblages, and radiocarbon ages. Depending on the spatial scale of analysis, I summed total volume and shell weight to calculate grams per meter cubed (g/m³). Collectively, these samples provide a robust dataset for comparative analysis of mussel harvesting strategies across spatial and temporal scales.

Table 2. Summary of California mussel and shellfish density for various periods of site occupation included in this study.

My chronological analysis includes 35 new radiocarbon dates from curated charcoal samples, supplementing 18 previously acquired dates to refine age estimates for undated contexts across the study area (Supplementary Dataset S1). For each excavation unit, I targeted both the deepest and shallowest levels to bracket occupational duration, and when these dates did not overlap at the 95.4% confidence level, I sampled intermediate layers with clear transitions in strata, fauna, and artifacts. In contexts lacking direct dates, I constructed age-depth models using the tidypaleo package (Dunnington et al., 2022) in R Studio 4.3.1 (R Core Team, 2020), with at least two calibrated dates from upper and lower layers. I calibrated all samples in OxCal 4.4 (Ramsey, 2009), using the IntCal20 terrestrial calibration curve (Reimer et al., 2020). The Penn State Radiocarbon Laboratory carried out radiocarbon measurements for all newly acquired samples. Dave Rhode at the Desert Research Institute made species-level identifications of submitted charcoal samples when feasible.

4.1 Size reconstructions for Mytilus californianus

Measurements of single morphological features can vary in precision and accuracy across environmental conditions. A multivariable model built from locally pooled samples offers greater precision, accuracy, and generalizability by capturing a broader range of shell morphology linked to local environments (McFarland et al., 2023). Therefore, I estimated total shell length using a multivariable model based on a locally pooled sample of whole shells collected from the beach near two coastal environments within the region of study: Mussel Rock (high-energy, open coast; n = 250) and Punta Gorda (medium to low energy, semi-protected by a bluff; n = 250). In addition to total shell length, I measured two morphological features—anterior-abductor scar (AAS) and umbo thickness (UTH)—for regression analysis as they are key features that: (1) preserve well in archaeological assemblages, (2) generally provide more accurate estimates than other measurements, and (3) are not linearly dependent (McFarland et al., 2023).

I split the total sample into an 80% (n = 400) training set and a 20% (n = 100) test set to evaluate model performance. Using a random stratified sampling strategy, I ensured equal size representation across the two sites and subsets, resulting in comparable size distributions (Supplementary Figure S1). I assessed model performance using 10-fold cross validation on my training set and external validation using the hold-out test set. Performance metrics include three statistical measures of bias, accuracy, and precision. Bias is the difference between estimated shell lengths and observed shell lengths, indicating overestimation (positive) or underestimation (negative). Accuracy is the absolute value of bias where values closer to zero are those closer to their true values. I report precision as the coverage of true values falling within 95% prediction intervals (PI), demonstrating the model's ability to confidently predict future observations. Since the regression model was built with 95% PI's, at least 95% of the observed shell lengths should fall within this range for the training and test set.

4.2 Temporal assessment for Mytilus californianus

To evaluate temporal changes in average mussel sizes associated with Hypothesis 3, I predicted total shell length for over 9,000 mussel shells using the regression model described above. I then grouped shell samples into 100-year intervals using age-depth models for each excavation unit, resulting in 58 distinct components. These components typically correspond to 10 cm depth levels or more. I treated excavation units with upper and lower radiocarbon dates overlapping at the 95.4% confidence level as single components. This approach allowed for high-resolution reconstruction of average mussel sizes through time, capturing harvesting episodes that may have occurred at finer temporal scales than radiocarbon dating can determine. I used a Kruskal–Wallis test (Kruskal and Wallis, 1952) to assess whether the distribution of mussel sizes differed significantly across age groups and three temporal periods: Middle Period (3,500–1,500 cal BP), Late Period I (1,500–700 cal BP), and Late Period II (after 700 cal BP). Dunn's test was used for post hoc comparisons (Dinno, 2024), with p-values adjusted via the Holm method to account for multiple comparisons (Holm, 1979). These analyses provide insight into variation in harvesting pressures over time.

4.3 Empirical modeling for Mytilus californianus

To evaluate the expectations associated with Hypotheses 1 and 2, I constructed empirical cumulative distribution functions (ECDFs) of predicted shell lengths for each archaeological component to characterize the size structure of harvested mussels. I then compared these archaeological ECDFs to theoretical harvesting models developed by Whitaker (2008) to identify specific harvesting strategies. Using Whitaker's model, I simulated plucking and stripping strategies across varying fallow durations (0, 6, 12, 18, and 24 months) and harvest intensities (5%–100%). Stripping assumes 100% of mussels were collected, so harvest intensity does not apply to this scenario and only fallow duration is considered (Whitaker, 2008, p. 1,117). Each model results in a distinct ECDF, reflecting shell size structures generated under specific harvest conditions.

An ECDF is a non-parametric step function representing the cumulative proportion of data points that are less than or equal to each value in a dataset (Dekking et al., 2005). Its shape is determined entirely by the number and spacing of observations. Whitaker’s (2008) harvesting models are based on shell lengths at 15 observations, ranging from 13 mm to 110 mm. Due to the small sample size, the ECDFs show a discrete step-like structure at each value, limiting their comparability to continuous, higher-resolution distributions. Because these are theoretical models intended to represent mussel populations under different harvesting strategies, I assume a continuous distribution. Therefore, I applied a generalized additive model (GAM) using the gam() function with the mgcv package (Wood, 2004, 2011, 2017; Wood et al., 2016) in R Studio 4.3.1 (R Core Team, 2020) to smooth the curves for comparison with archaeological ECDFs.

I assessed similarity between archaeological and theoretical ECDFs, using the Kolmogorov–Smirnov (K-S) test, which is a nonparametric method that quantifies the maximum vertical distance (D statistic) between two cumulative distributions (Massey, 1951). The D statistic serves as the primary metric for identifying the best fit between each archaeological sample and harvesting scenario. Lower D values indicate closer alignment between the archaeological shell size distribution and a given theoretical model. The K-S test also produces a p-value indicating whether the distributions differ significantly, with non-significant values suggesting no statistical difference between the distributions.

Since the K-S test is sensitive to irregularities in small samples (Massey, 1951), I limit my analysis to archaeological samples with more than 20 observations. While this sample size falls below the recommended minimum of 50 (Mishra et al., 2019), I have found that it provides sufficient resolution for comparison when larger archaeological samples are not available. For particularly large samples (n > 200), the K-S test can also become overly sensitive and detect minor deviations that are insignificant (Mishra et al., 2019). I address these issues by bootstrapping each archaeological sample 1,000 times in R Studio 4.3.1 to calculate mean D statistics and p-values across resampled subsets, while iteratively comparing samples to each theoretical model to identify the best match. ChatGPT (Version 5) assisted with code development during this process; however, I meticulously formulated, modified, and reviewed all code for the purpose of this study.

4.4 Simulation model

To test expectations associated with Hypothesis 3, I modeled ecological dynamics across successive harvest cycles in R Studio 4.3.1 (R Core Team, 2020), using a custom simulation developed for this study with assistance from Chat GPT-5 that tracks changes in biomass, average shell length, harvested mass, and energy yield per cycle. The model begins with a population at carrying capacity (K = 10,000 mussels) and an average shell length of 30 mm (harvestable size). During each fallow interval, mussels grow according to two processes: (1) total biomass increases by 5% per month, capped at K; and (2) average shell length increases by 1 mm per month, scaled by the proportion of current biomass relative to carrying capacity (biomass/K) and slowing asymptotically as size approaches a maximum size of 250 mm. A 5% monthly biomass increase is a conservative estimate corresponding to an annual production of roughly 60%–70% similar to production rates reported for Mytilus spp (Cusson and Bourget, 2005.; Dolmer et al., 2012; Seed and Suchanek, 1992), while shell growth of 1 mm per month reflects estimates from California mussel populations along the California Current system, averaging between 0.5–2 mm per month (Blanchette et al., 2007; Coe and Fox, 1942; Smith et al., 2009; Suchanek, 1981; Vriesman et al., 2022).

After each fallow interval, the model incorporates a stochastic mortality event, with the probability of a die-off rising by 2% per month of fallow duration up to a maximum of 100%. If triggered, the event randomly removes 10% of the standing biomass. A density-dependent growth model helps simulate a realistic system which incorporates slower growth as mussels approach their maximum size (250 mm), while stochastic mortality introduces potential risk without relying on assumptions where empirical evidence is lacking (Paine, 1976; Seed and Suchanek, 1992; Suchanek, 1979, 1981).

Each harvest cycle removes a fixed proportion of standing biomass, determined by the specified harvest intensity. The model converts the harvested biomass to meat weight (g) using a mussel length-to-meat equation for dry, non-reproductive tissue mass developed by Singh and McKechnie (2015, p. 179), then scales meat weight to kilocalories (kcal) using a correction factor of 1.72 (USDA FoodData Central, 2018). The function runs across N harvest cycles, each defined by a fallow duration (6, 12, 18, or 24 months) and harvest intensity (5% to 100%). For each cycle, the output records harvested mass, remaining biomass, average mussel size, and energy yield. The model then summarizes these outputs across all cycles to calculate average energy yield per harvest, average mussel size, and records whether biomass falls below 50% of K at any point in the simulation to indicate maximum sustainable yield for each harvesting strategy.

4.5 Seasonality analysis of oxygen isotopic (δ¹⁸O) measurements

Previous estimates of seasonal harvesting along the Lost Coast have relied on the presence of juvenile mammals or migratory bird species at a limited number of sites (Levulett, 1985; Waechter, 1988, 1990, 1991; Whitaker, 2005). However, such interpretations often neglect shellfish, which represent an abundant and available food source throughout the year. To test for season of mussel harvesting associated with Hypothesis 2, I sampled 20 California mussel shells from two Middle Period contexts at Spanish Flat: CA-HUM-277, 3N/5W S.Q. D, 80–90 cmbs (1,920–1,420 cal BP) and CA-HUM-277, 3N/5W, S.Q. D, 140–150 cm (2,710–2,350 cal BP). Because mussels are the greatest contributor to the shellfish assemblage in both contexts, I use these seasonality estimates as proxies for all shellfish harvesting.

After chemically cleaning shells with a 10% HCl etch to remove external contaminants, I obtained powder from each shell at 0-, 1-, 4-, and 9-mm increments along the main axis of growth, using a Sherline Model 5410 drill equipped with a 0.8 mm diameter drill bit. From each level, I selected one shell for additional drilling at 1-mm increments, collecting a total of 20 samples each to generate a sinusoidal curve to reflect a minimum of one year of growth. In total, I drilled 192 carbonate samples. In June 2025, the Stable Isotope Facility (SIF) at University of California, Davis analyzed the samples using a Thermo Scientific GasBench II coupled to a Thermo Scientific Delta V Advantage isotope-ratio mass spectrometer, measuring both δ¹³C and δ¹⁸O. The lab measured replicates every ten samples to determine accuracy, with an expected error of ±0.10 ‰ for δ¹³C and ±0.15 ‰ for δ¹⁸O of carbonate (Supplementary Dataset S2).

I converted δ¹⁸O values to SST (˚C) using an equation from Horibe and Oba (1972):The measured δ¹⁸O value for the sample is δc, and δw represents the δ¹⁸O of ambient seawater. For ambient seawater, I used a regional average of −0.13 ‰ measured from 12 seawater samples collected from intertidal locations ranging from Cape Mendocino to Shelter Cove (Supplementary Dataset S3). In December 2024, SIF performed these measurements and reported an accuracy of ±0.17 ‰ for δ¹⁸O which is better than the expected error of ± 0.20 ‰ for δ¹⁸O of water.

t∘C=17.04−4.34(δc−δw)+0.16(δc−δw)2

5 Results

5.1 Temporal changes in mussel sizes

Full regression results are provided and described in the Supplementary Material. Overall, the model predicts total shell lengths with 95% confidence to within ±20 mm, with mean errors of 5.82–7.94 mm (Supplementary Table S2 and Figure S2). When applied to 9,000 fragmentary archaeological mussel shells measured for this study across multiple site components and time periods, the analysis reveals site-specific temporal patterns in mussel size variation (Figure 2).

At Spanish Flat, mussel sizes fluctuate over time, with periods of growth and decline (Table 3 and Figure 2). Samples from CA-HUM-277 show a significant increase from 57.0 mm during the Middle Period to 63.3 mm during Late Period I, followed by a slight decline to 60.3 mm during Late Period II. Samples from CA-HUM-279 and CA-HUM-281 also show significant increases between Late Period I and II, though at CA-HUM-279 samples generally contain smaller shells (52.1–57.8 mm), except in unit 42SE/12SW. Samples from CA-HUM-281 exhibit some of the largest mean mussel sizes (58.6–72.3 mm), with a significant increase between 500–400 cal BP and 300–200 cal BP, followed by a significant reduction during 200–100 cal BP.

Table 3. Summary statistics for each component from Spanish Flat (n = 3,886).

Sites at Shelter Cove show more consistent size increases and higher overall means than observed at Spanish Flat (Table 4 and Figure 2). Samples from CA-HUM-182 range in mean size from 52.3–71.1 mm, with a significant increase from 62.9 mm to 64.2 mm between Late Period I and II. Samples from both CA-HUM-184 and CA-HUM-186 date to Late Period II and maintain relatively large sizes ranging from 63.1–65.4 mm and 63.2–64.8 mm, respectively. At CA-HUM-185, samples show a significant increase from 62.2 mm (200–100 cal BP) to 73.9 mm (100–0 cal BP) when comparing individual age groups. In contrast, CA-HUM-248 contains the smallest mean sizes (58.1–58.3 mm) observed at Shelter Cove with no significant temporal changes.

Table 4. Summary statistics for each component from Shelter Cove (n = 5,156).

5.2 Archaeological ECDFs vs. theoretical models

Table 5, 6 summarizes the K-S test results comparing archaeological ECDFs with simulated mussel size distributions generated from Whitaker’s (2008) harvesting models for both regions, which predicts expected size structures under varying harvesting intensities and fallow intervals. Across all samples, D values were consistently low (0.07–0.29), and p-values remained above 0.05, indicating no significant differences between observed and modeled size distributions. Sample sizes varied widely from small (20–50 specimens) to large (>400 specimens) collections, yet in nearly all cases the models performed well, with smaller samples generating results consistent with those of larger assemblages from adjacent components.

Table 5. Summary of top kolmogorov–smirnov (K-S) test results for each archaeological sample from Spanish Flat (n = 3,715) based on 1,000 iterations used to estimate D statistics and p-values.

Table 6. Summary of top Kolmogorov–Smirnov (K-S) test results for each archaeological sample from Shelter Cove (n = 5,261), based on 1,000 iterations used to estimate D statistics and p-values.

Spanish Flat samples matched plucking strategies with fallow periods varying from as short as 6 months to as long as 24 months and intensities ranging between 20%–40% (Table 5). Shelter Cove samples also reflect plucking strategies but with more consistent harvesting schedules of 18–24 months at slightly higher intensities ranging between 25%–45% (Table 6). These results highlight a broad reliance on plucking as the dominant harvesting strategy, with Spanish Flat samples showing greater variability in the frequency and intensity of harvests, while Shelter Cove samples reflect more consistent harvesting intervals, often at higher intensities.

5.3 Ecological modeling

Figure 3 shows average energy yield per harvest across combinations of fallow duration and harvest intensity. Yields are unsustainable once standing biomass drops below the maximum sustainable yield (<50% of K) at any point in the simulation. The highest returns occur with a 24-month fallow period at 65% intensity, while major yield reductions first appear at 6 months with 30% intensity, then at 12 months at 45%, 18 months at 60%, and 24 months at 70% intensity. In contrast, sustainable yields involve lower intensities: 6 months at 20%, 12 months at 35%, and 18–24 months at 40% intensity. These strategies represent maximum sustainable yield limits while maintaining biomass above 50% of carrying capacity, a commonly used benchmark for avoiding overharvest. Exceeding these thresholds increases the risk of resource depression and results in diminishing returns.

Figure 4 depicts mussel size trajectories across various harvest cycles, considering fallow period and intensity. Mussels grow asymptotically toward a maximum size of 250 mm over the 600-month (50-year) simulation. Average mussel sizes decline under the same harvesting strategies that first produced lower average energy yields (Figure 3), notably at 6 months with 30% intensity, 12 months at 45%, 18 months at 60%, and 24 months at 70%. Continued harvest pressure at these intensities or higher maintains persistently smaller mussel size distributions, with mean shell lengths dropping to less than half of maximum size (Figure 4). This long-term size reduction ultimately leads to unsustainable yields (Figure 3).

5.4 Season of harvest for CA-HUM-277

The average variance between replicates for the submitted carbonate samples yielded an uncertainty of ±0.05 ‰ for δ¹³C and ±0.09 ‰ for δ¹⁸O. Across all samples from CA-HUM-277, δ¹⁸O averaged +0.55 ‰, corresponding to an estimated temperature of 14.2 ˚C. Observed values ranged from −1.78 ‰ (10 ˚C) to +1.6 ‰ (24.7 ˚C). For shells sampled from 0 mm to 19 mm, earlier samples (2,710–2,350 cal BP) exhibited higher average δ¹⁸O values (+0.79 ‰, 13.2 °C) than those from the later period (1,920–1,420 cal BP), which averaged +0.43 ‰ (14.7 °C). Across all mussel shells from both contexts, the mean δ¹⁸O value for the terminal growth margin was +0.55 ‰ (14.2 °C) in the earlier context and +0.32 ‰ (15.2 °C) in the later context.

Isotopic probability distributions (see Jazwa and Zoellner, 2024) suggest that during the Middle Period (3,500–1,500 cal BP) inhabitants predominantly harvested mussels, and possibly other shellfish, in late spring or early summer with additional activity potentially occurring in late summer or early fall at CA-HUM-277 (Figure 5). These patterns are also consistent with interpretations of spring and summer harvesting of mammals and birds at the site during the Late Period, after 1,500 cal BP (Levulett, 1985).

6 Discussion

The archaeological record for Northwest California contains evidence of human occupation spanning the past 9,000 years (Fitzgerald and Hildebrandt, 2002; Hildebrandt and Levulett, 1997; Tushingham, 2009, 2020), with the majority of the documented coastal settlements dating within the past 3,000 years along the Lost Coast (McFarland, 2025). Understanding how different harvesting strategies generate distinct cycles of decline and rebound in average mussel sizes within the study area offers a way to interpret both periods of resource depression (Braje et al., 2007, 2012; Erlandson et al., 2008; Glassow et al., 2008; Jazwa et al., 2016; Jones, 1996) and resource sustainability (Grone, 2025; Kievman et al., 2023; Tatlow et al., 2022) as documented along the Pacific Coast. When paired with local and regional climatic data, this knowledge can help disentangle cultural decision-making from potential environmental influences.

6.1 Changes in mussel sizes and harvesting strategies

For mussel harvesting along the Lost Coast, plucking of medium to large-sized individuals appears to be the preferred method, as indicated by relatively low standard deviations in mean mussel sizes across all samples (6.96–23.15 mm). The predominance of medium and larger specimens in all archaeological assemblages, ranging from 49 mm to 74 mm on average, suggests that people intentionally practiced size-selective harvesting using a plucking strategy (Expectation 1). Larger mussel shells may have also been used as tools, such as knives commonly used in the region for processing plant fibers to create rope and for processing hides (Driver, 1939; Hewes, 1940; Kroeber and Barrett, 1960).

Across all study areas, however, average mussel sizes differ noticeably (Tables 3,4 and Figure 2). At Spanish Flat, average sizes (59 mm) are 3 mm smaller than the overall population mean of 62 mm. Conversely, sites at Shelter Cove show average sizes (64 mm) that are 5 mm larger than the population mean, resembling larger mussel sizes reported by Kievman et al. (2023). These regional patterns possibly reflect localized differences in SSTs and ecological conditions (Section 2.2). However, the long-term stability in mussel sizes over the past millennium at Shelter Cove, along with increases observed after 700–600 cal BP, appear closely tied to distinctive harvesting strategies.

At Shelter Cove, larger average mussel sizes (Table 4) consistently align with 24-month harvesting intervals across both temporal and spatial scales (Table 6), mirroring findings by Whitaker (2008). In certain contexts, harvesters used 18-month harvest intervals, and when paired with higher intensities (30%–40%), resulted in smaller average mussel sizes (52–61 mm) compared to those observed under 24 month intervals (57–70 mm) at the same harvest intensities. Diminishing returns in average mussel sizes are especially evident between 900–800 cal BP and 700–600 cal BP under this harvesting strategy (Table 4), coinciding with evidence of more intensive harvesting, as reflected in increasing mussel shell densities (Table 2 and Figure 2). In contrast, 24-month harvesting intervals at 25%–40% intensity after 700–600 cal BP sustained average mussel sizes near the baseline mean, and after 200–100 cal BP mussel sizes increased significantly despite higher harvesting pressures (Table 6 and Figure 2).

Site type and distance to the coast are also important variables for interpreting variation in mussel size in archaeological assemblages, particularly because they may reflect differences in how shellfish were processed and transported to central places (Bettinger et al., 1997; Jazwa et al., 2015). For instance, the energetic costs of carrying whole mussels increases with distance from harvesting zones, meaning that foragers are likely to reduce bulk weight by shucking mussel shells at logistical camps and processing stations before transporting meat to central places or they will harvest smaller whole mussels for transport when villages are located farther inland. Considering that many of the sites in this study lie directly on the shoreline and within 800 m (0.5 mi) of core residential areas, travel costs were likely not a major factor in shellfish harvesting and transport. Only one logistical camp (CA-HUM-248) is situated 800 m (0.5 mi) from the shoreline, and its smaller average mussel sizes suggest that distance, and the added cost of transporting larger (heavier) mussels, may have influenced both harvesting decisions and processing strategies.

Samples from Spanish Flat show more variability in average mussel sizes and harvesting strategies in comparison to Shelter Cove (Figure 2). During the Middle Period (2,600–2,500 cal BP to 1,700–1,600 cal BP), plucking strategies mainly involved 12- or 18-month fallow intervals with harvest intensities as high as 35%–40% (Table 5). When sustained over successive 100-year periods, these harvesting strategies reduced average mussel sizes between 2,300–2,200 cal BP and 2,100–2,000 cal BP (Table 3 and Figure 2). A significant rebound occurred during 1,900–1,800 cal BP under a 24-month harvest interval at the same intensity. This period also shows evidence of higher harvesting pressures, as indicated by the overall density of mussel shells (Table 2 and Figure 2). Similar patterns continued into later periods, with average mussel sizes declining between 800–700 cal BP and 600–500 cal BP under 6- to 18-month harvest intervals at 20%–30% intensity (Table 5 and Figure 2). After 600–500 cal BP, mussel sizes stabilized near the baseline mean. This stability in mussel sizes persisted despite mixed strategies of 12- to 24- month harvest cycles at 20%–40% intensity, even under greater harvesting pressure during 100–0 cal BP when mussel shell densities increased substantially (Figure 2).

Overall, the consistent use of 18- to 24-month harvesting intervals at Shelter Cove sustained average mussel sizes over the past millennium, with a gradual increase after 700–600 cal BP (Figure 2). In contrast, archaeological patterns at Spanish Flat suggests that harvesters were balancing immediate returns with long-term sustainability through deliberate use of 12- to 24- month fallow periods (Table 5). Ultimately, these differential harvesting strategies at Spanish Flat produced more variable changes in average mussel sizes. Nevertheless, increases in average mussel sizes occurred in both regions despite intensive coastal occupation after 700 cal BP (Table 2) and periods of higher harvesting pressures (Figure 2). These results support the expectations that people along the Lost Coast used rotational harvesting practices (Expectation 2.2) and place-based knowledge (Expectation 3.2) to sustain, and in some cases enhance, the long-term productivity of mussel beds, particularly after 700 cal BP. Additionally, isotopic profiles for samples from Spanish Flat suggest mussel harvesting occurred predominantly during the spring-summer and potentially early fall (Figure 5), aligning with seasonal harvesting practices (Expectation 2.1).

In general, mussel sizes remained stable over time along the Lost Coast, and harvesting strategies did not exceed maximum sustainable yield limits, even under shorter harvest cycles (Tables 3–6 and Figures 2–5). Although lower shell densities in some contexts indicate potential periods of reduced pressure (Figure 2), average mussel sizes increased even with higher shell densities, pointing to sustainable harvesting practices rather than low predation (Expectation 3.1). However, an exception occurred at CA-HUM-182 (78N/202W, 80–120 cmbs, 900–800 cal BP), where harvest intensity briefly exceeded the maximum sustainable yield threshold (Table 6 and Figure 3), during a 24-month fallow period at 45% intensity, followed by 18-month harvest cycles, that produced the smallest average mussel sizes recorded at Shelter Cove (52–57 mm), spanning 900–800 cal BP to 700–600 cal BP (Table 4). Higher harvest intensities and increasing shell densities during this period suggest that coastal communities were placing greater harvesting pressures on mussel populations, resulting in a temporary reduction in average mussel sizes.

6.2 Mussel harvesting and climate change

Regional hydroclimatic and oceanographic variability since 3,500 cal BP directly influenced both terrestrial and marine environments in Northwest California (Addison et al., 2018; Barron et al., 2018; Kirby et al., 2023; Knight et al., 2024; Palmer et al., 2023; Voelker et al., 2018). These variable conditions likely influenced the productivity of mussel beds in Northwest California and contributed to fluctuations in average mussel sizes observed in archaeological assemblages along the Lost Coast. Specifically, changes in water temperature, salinity, and nutrient availability are all key factors influencing mussel growth rates and reproductive success (Blanchette et al., 2006, 2007; Blanchette and Gaines, 2007; Vriesman et al., 2022), with important implications for long-term resource management and sustainable harvesting practices.

Isotopic δ¹⁸O measurements sampled from 0 mm to 19 mm from the terminal growth band of mussel shells from Spanish Flat suggests that SST fluctuated by as much as 1.5˚C, rising from approximately 13.2 °C to 14.7 °C on average, between 2,710–2,350 cal BP and 1,920–1,420 cal BP. This warming in SSTs during the Middle Period corresponds with evidence of century-long droughts in the Great Basin between 2,000–1,500 cal BP (Mensing et al., 2008, 2013, 2023). These results are also consistent with a 1˚C increase in alkenone SST off the northern California coast, coinciding with a threefold rise in the relative abundance of Pseudoeunotia doliolus diatoms around 3,200 cal BP, a pattern that reflects warmer SSTs potentially driven by increasing wintertime insolation during the Late Holocene (Barron et al., 2003, p. 10; Palmer et al., 2023).

The variability in mussel sizes observed at Spanish Flat between 2,710–2,350 cal BP and 1,920–1,420 cal BP may reflect the combined pressures of environmental stress on mussel populations and coastal communities leading to more intensive mussel harvesting. Fluctuating SSTs, recurrent droughts, and shifts in insolation could produce alternating periods of favorable and constrained growth conditions, potentially influencing mussel populations. These dynamics could have also added to the challenges faced by coastal communities during this time who were likely balancing immediate subsistence needs with the sustainability of mussel harvesting over time.

Similar pressures persisted into later centuries, with variable climatic patterns evident during the Medieval Climatic Anomaly (MCA; ∼1,050–650 cal BP), a period marked by pronounced interannual and interdecadal variability in ocean conditions that contributed to prolonged droughts and heightened aridity across western North America (Jones and Schwitalla, 2008; Kirby et al., 2023; Voelker et al., 2018). Isotopic analyses of mussel shells from central California also reveal greater seasonal variation in SSTs between 700 and 500 cal BP (Jones and Kennett, 1999), potentially placing additional stress on coastal communities and mussel populations.

Towards the end of the MCA, average mussel sizes declined at both Spanish Flat and Shelter Cove (Tables 3,4), which also corresponds with intensive harvesting pressures (Figure 2). Archaeological samples from both regions during this time indicate harvesting strategies operated on 18- to 24-month cycles at much higher intensities ranging from 30%–45% on average (Tables 5,6), suggesting that coastal populations were perhaps adapting to changes within terrestrial environments by intensively harvesting mussels while attempting to remain within sustainable limits. However, the combination of intensive harvesting and increasingly variable ocean conditions potentially constrained the short-term recovery of mussel beds, resulting in measurable decreases in average size.

In comparison, the Little Ice Age (LIA; ∼650–50 cal BP) was characterized by cooler and wetter terrestrial climates (Bradley and Jones, 1995; Palmer et al., 2023), with SSTs 2–3 °C below present between 500 and 300 cal BP, as indicated by isotopic analyses of mussel shells from the central California coast (Jones and Kennett, 1999). These cooler ocean conditions potentially reflect enhanced ocean upwelling that boosted marine productivity, creating more favorable ecological conditions for mussel growth. However, high sedimentation from upstream erosion may have also impacted water quality and stressed mussel populations when initially combined with higher harvest intensities, potentially contributing to noticeable declines in average mussel sizes. After 700 cal BP, lower harvest intensities with longer fallow periods might have allowed mussel sizes to recover and stabilize. Further evaluating how environmental conditions influenced mussel productivity and cultural adaptations in the study area would require more comprehensive nearshore oceanographic records from the Lost Coast, as existing data are limited to the isotopic δ¹⁸O measurements presented here for Spanish Flat.

Harvesting strategies that incorporated consistently longer 24-month fallow periods appear to have stabilized mussel sizes near baseline averages in both regions, even under sustained pressure with intensities ranging from 25%–40% during the LIA (Tables 3–6 and Figure 2). Mussel shell densities, for example, indicate intensive harvesting between 600–500 cal BP at Shelter Cove and 100–0 cal BP at Spanish Flat, yet mussel sizes remained stable (Figure 2). The consistent use of longer fallow periods could have enhanced mussel bed productivity at Shelter Cove between 300–200 cal BP and 100–0 cal BP where there is evidence of increasing harvest pressures, with greater mussel shell densities (Table 2 and Figure 2) and higher harvest intensities (Table 6). The long-term stability as well as brief cycles of decline and rebound in mussel sizes associated with distinct harvesting strategies suggest some form of place-based stewardship along the Lost Coast (Expectation 3.2).

These patterns of marine stewardship also coincides with evidence of Indigenous management of terrestrial ecosystems through landscape burning practices in the Klamath Mountains of Northwest California—located approximately 120 km to the northeast of the Lost Coast—resulting in significant ecosystem modifications during the LIA (Crawford et al., 2015; Knight et al., 2022). Similar evidence of terrestrial management is also apparent in the Sierra Nevada's after 650 cal BP (Klimaszewski-Patterson et al., 2021; Klimaszewski-Patterson and Mensing, 2016) as well as marine stewardship along the central California coast (Apodaca et al., 2024; Grone, 2020, 2025; Lightfoot et al., 2025; Sanchez et al., 2023), suggesting management of both marine and terrestrial ecosystems was perhaps widespread throughout California during the Late Holocene.

6.3 Indigenous tenure systems and stewardship

During the Late Period (after 1,500 cal BP), current archaeological evidence indicates that settlement at Spanish Flat and Shelter Cove involved more permanent residential occupation (McFarland, 2025), and after 700 cal BP, total shell densities indicate increasingly intensive harvesting of shellfish resources, with mussel shells comprising over 50% of assemblages in many contexts (Table 2). As coastal communities became more sedentary, implementing stewardship practices for key food staples near residential areas became important to sustaining the ecological productivity of the nearshore environments they depend on (McFarland, 2025). Such strategies require forms of tenure systems that restrict the timing and location of harvests to prevent localized population depletion and support natural regrowth.

Since shellfish and other coastal foods are considered common pool resources that are difficult to exclude others from and vulnerable to depletion (Jones and Codding, 2019), resource governance is important, especially for primary food staples like mussels. Ethnographic evidence suggests that kin-based territorial rights governed coastal harvesting areas along the Lost Coast (Hewes, 1940; Nomland, 1935, 1938), which potentially extended into intertidal zones for shellfish harvesting too (McFarland, 2025). Familial ownership could have controlled and limited access to mussel beds, like they did for offshore colonies of sea mammals and birds (Hewes, 1940; Nomland, 1935, 1938).

Additionally, the emergence of permanent residential communities at Spanish Flat and Shelter Cove, after 1,500 cal BP, may have fostered stronger social cohesion in these areas, enabling the development of shared resource ethics, coordinated management of mussel beds, and other marine management practices (McFarland, 2025). This cultural system could have mitigated potential risks associated with localized resource depression in mussel beds, resulting from either periodic climatic stressors or overharvesting, or both. By using a cooperative-collectivism form of social organization—characterized by high inter-household collaboration and social conformity (Gallyamova and Grigoryev, 2025, p. 219)—communities likely embedded sustainable mussel harvesting strategies through intergenerational knowledge tied to specific places along the Lost Coast (McFarland, 2025).

Selective and rotational harvesting of larger mussels may have also served as adaptive responses to seasonal ecological risks that were ever more common during the Late Holocene, including greater climatic variability and environmental changes within terrestrial environments (Addison et al., 2018; Barron et al., 2018; Kirby et al., 2023; Palmer et al., 2023). For example, CA-HUM-277 at Spanish Flat provides the earliest evidence of intensive marine resource use in Northwest California, spanning the Middle Period from 2,710–2,350 cal BP to 1,920–1,420 cal BP. Higher mussel shell densities during this time suggest more intensive coastal occupation and greater harvesting pressures (Table 2 and Figure 2), potentially reflecting a strategic intensification of marine resource use during the Late Holocene dry period (2,000–1,500 cal BP). Variations in average mussel sizes during this period could reflect an initial phase of coastal adaptation at the site. Inhabitants likely considered the trade-offs between two strategies: (1) harvesting at higher intensities (30%–40%) with shorter fallow periods (6–12 months), which would yield higher immediate returns but lead to smaller mussels over time, or (2) harvest at the same intensities while extending fallow periods (18–24 months), which would allow mussels to grow larger and help sustain productivity over time (Table 5 and Figures 3–6).

By consistently harvesting mussels at an intensity below maximum sustainable yields with longer fallow periods (Table 5 and Figure 3), average mussel sizes at Spanish Flat gradually stabilized near their long-term mean while mussel sizes increased at Shelter Cove (Table 3 and Figure 2), despite evidence of more intensive coastal occupation and greater mussel harvesting in both regions after 700 cal BP (Table 2). This suggests sustainable mussel harvesting for over a millennium, likely supported by stewardship practices and place-based ecological knowledge involving an intimate understanding of tidal cycles and local environmental limits, as well as seasonal changes in oceanographic and ecological conditions.

For instance, when harvest intensity in a mussel bed approached levels near the maximum sustainable yield, foragers likely shifted to alternative beds, allowing time for local recovery and regrowth. Rotational harvesting with longer fallow periods, such as 18–24 months, would then maximize long-term returns while keeping harvests within sustainable limits, promoting larger and healthier mussel populations for future generations. Seasonal harvesting restricted to the spring and summer months also likely kept annual pressure low, with potentially modest fall harvesting occurring when conditions allowed for sustainable take (Figure 5). In contrast, harvesting at intensities beyond sustainable thresholds increases the probability of resource depression and eventually produces sharp diminishing returns (Figure 3), with average mussel sizes dropping to less than half their potential size (Figure 4). Figure 6 illustrates this scenario, using the results of this study.

In Optimal Foraging Theory (OFT), some scholars argue that outcomes which appear to be the result of stewardship can also emerge from foraging strategies that maximize return rates, or from low human population densities, which naturally keep harvest pressure low (Alvard, 1993, 1994, 1998; Low, 1996; Stephens and Krebs, 1986; Thomas, 2007; Venkataraman et al., 2017; Winterhalder and Lu, 1997). Under the Marginal Value Theorem (MVT), for instance, foragers are expected to leave a patch when the instantaneous return rate falls to the long-term average return rate for the environment, including travel time, making it more profitable to move to a new patch than to continue foraging in the current one (Charnov, 1976). This model assumes that travel and search time impose meaningful costs, and it predicts that foragers will remain longer in patches when those costs are higher.

Along the Lost Coast, however, travel and search times were likely minimal. All of the village sites included in this study are situated directly adjacent to harvesting zones, and foragers using selective harvesting would have likely known the precise locations of mussel beds, considering some portion would remain for future harvest if managed sustainably. In this context, the relevant “cost” shaping foraging decisions is not travel or search time but the long-term consequences of overharvesting and the short-term opportunity cost of refraining from exceeding the maximum sustainable yield. Although higher harvest intensities may produce greater immediate returns, they increase the long-term risk of resource depression (e.g., biomass < 50%) and guarantee sharp diminishing returns beyond a certain threshold (e.g., mussel sizes reduced by half).

Harvesting strategies in this study consistently fell below the level that would have produced maximum returns in the simulation, even while remaining within sustainable limits during periods of intensified shellfish harvesting and increasing coastal occupation after 700 cal BP. This pattern suggests that people were not maximizing short-term returns, but instead, they were likely incorporating long-term costs of overharvesting into their decisions, involving less intense, less prolonged, and less extractive harvesting strategies. Given the finite number of available mussel beds within a foraging radius around each residential area, people returned to the same locations on daily, seasonal, and annual cycles to sustain the broader population over millennia. These long-term relationships with specific places would have ultimately encouraged the use of stewardship practices, as failing to do so could risk resource depression and lead to unsustainable long-term yields while harvesting mussels.

Additionally, after 1,500 cal BP, coastal settlement densities and shellfish harvesting increased (McFarland, 2025), creating either competition or cooperation among villages. If competition is high, the number of accessible patches would have been further constrained around core residential areas, increasing patch residence times and restricting foraging to possibly less than an 800 m (0.5 mi) radius, as that is the average distance between each village. This potential creates a strong incentive for cooperative management, where communities can expand their foraging radii and harvest opportunities by sharing access to land and resources, strengthening inter-village relationships, and engaging in reciprocal exchange during periods of scarcity (Jones and Ready, 2022; Kaplan et al., 1985; Smith, 1988; Smith and Boyd, 2019). Such practices are well documented in ethnographic records for the Lost Coast (Driver, 1939; Hewes, 1940; Nomland, 1935, 1938), and elsewhere along the Northwest Coast of North America (Ban et al., 2020; Trosper, 2009). By distributing the short-term opportunity costs among villages, by not maximizing immediate returns, communities could benefit from long-term strategies involving longer harvest intervals and lower harvest intensities that promote larger and more stable mussel beds, ultimately increasing patch residence times and long-term yields.

Sustainable returns can also be supported by diversifying shellfish harvesting within the intertidal zone (McFarland, 2025) and by managing key mussel predators (Salomon and McKechnie, 2025), reducing overall pressure on mussel populations. Archaeological evidence from Spanish Flat, for example, shows that mussel shell densities declined by more than half after 1,500 cal BP despite greater coastal settlement densities and increasing mussel sizes (Table 2 and Figure 2). During this same period, shell densities of secondary taxa including chitons (Cryptochiton stelleri, Katharina tunicata, Mopalia muscosa), barnacles (Balanus spp.), turban snails (Tegula spp.), and rock whelks (Nucella spp.) increased. This pattern may reflect strategic diversification intended to reduce harvesting pressure on primary food staples such as mussels (McFarland, 2025, p. 99).

At Shelter Cove, mussel shell densities remained high (71%–93%), indicating sustained pressure on mussel populations during periods of intensified coastal occupation (Table 2). However, despite these pressures, stable and increasing mussel sizes suggest that strategies involving longer fallow periods facilitated sustainable harvesting over the past millennium. Additionally, higher densities of sea otter (Enhydra lutris) bones represented in mammal assemblages from Shelter Cove, relative to other areas along the Lost Coast (McFarland, 2025), are consistent with predator management practices documented elsewhere along the Northwest Coast (Slade et al., 2022). Regions where sea otters were harvested less frequently may have contributed to the temporal variability in mussel sizes observed in this study, which may also be because of the difficulty in hunting sea otters and their prestigious value (Hewes, 1940, p. 3).

In summary, stewardship of mussel beds can involve a multitude of strategies. People can maximize long-term returns of mussels by: (1) selectively harvesting a narrower size range of medium to large-sized mussel, allowing smaller mussels to grow to harvestable size, (2) rotating between mussel beds to limit the intensity of harvest within sustainable yields, (3) restricting season and timing of harvest to increase fallow periods, allowing time for recovery and regrowth, and (4) diversifying shellfish harvesting as well as (5) actively excluding predators from primary mussel harvesting zones to reduce pressures on mussel beds. These stewardship practices, likely embedded in place-based knowledge and various cultural protocols associated with marine tenure systems, enhanced and sustained mussel productivity near core residential areas in the study area. In doing so, they likely helped reduce the risk of shortfalls amid increasing coastal settlement densities, intensive shellfish harvesting, and variable climatic conditions during the Late Holocene (McFarland, 2025).

7 Conclusion

In this study, I examined temporal changes in mussel sizes across archaeological samples from Northwest California to evaluate how mussel productivity responded to different harvest intervals and intensities along the Lost Coast, since 2,700 cal BP. I identified sustainable harvesting in the archaeological record by comparing empirical size distributions with theoretical distributions generated under various harvesting conditions while simulating long-term mussel bed dynamics to evaluate recovery potential and strategies that produce maximum sustainable yields. The results of these analyses demonstrate statistically significant increases in mussel sizes, despite periods of higher harvesting pressures and potentially local environmental stressors. Differences in harvesting and oceanographic conditions during the Late Holocene contributed to fluctuations in average mussel sizes across spatial and temporal scales. Yet average mussel sizes increased and stabilized over the past millennium during periods associated with longer harvest intervals (18–24 months) and intensities (20%–40%) that did not exceed maximum sustainable yields.

By integrating archaeological and ethnographic records, I show how Indigenous communities in Northwest California sustainably harvested mussels through selective, rotational practices likely embedded within place-based knowledge and marine tenure systems along the Lost Coast. These systems regulated when, where, and who could harvest, potentially extending into the intertidal zone through restrictions on harvest timing, location, and size limits to ensure mussel beds had the time and space needed for regrowth. I further suggest that Indigenous communities reduced predation pressures on local mussel beds by diversifying shellfish harvesting and managing sea otter populations. Understanding how past societies organized themselves to sustain marine resources is important, as the collective choices to safeguard the long-term health of mussel beds result in distinct archaeological signatures, such as long-term stabilization or increases in mussel sizes, reflecting the ecological outcomes of these practices.

Using the methods presented in this study, future archaeological investigations can identify mussel harvesting strategies and contextualize long-term changes in mussel sizes within unique social, cultural, and political settings along the Pacific Coast. This information helps distinguish ecological outcomes from the effects of climate change, overharvesting, and stewardship practices on timescales relevant to resource managers. Although cultural decisions can sometimes lead to negative outcomes, sustainable harvesting depends on embedding stewardship within place-based knowledge and tenure systems to prevent overexploitation and maintain productive ecosystems for future generations. In this study, I present multiple lines of evidence to contextualize sustainable harvesting and Indigenous stewardship of mussel beds in Northwest California, offering insights relevant to contemporary marine management.

7.1 Management implications

Respectfully integrating archaeological data, ethnographic records, ecological modeling, and the knowledge of Indigenous communities who have sustained marine ecosystems for generations provides a culturally informed framework with several management implications:

• Recognize that archaeological and ethnographic approaches open new avenues for future ecological research, especially in places where long-term relationships between people and marine ecosystems have persisted for millennia. Remote regions such as the Lost Coast remain largely unmonitored and understudied, yet archaeological records can help identify vulnerable and ecologically resilient areas that may merit future conservation attention (McFarland et al., 2025).
• Integrate place-based, community-led resource governance into management plans. Kin-based territorial rights present a compelling precedent for co-management through social licenses that respects the rights, rules, and agreements that Indigenous communities have held for generations and whose cultural and spiritual ties to these environments foster a commitment to long-term reciprocity rather than more extractive models of management (Lyons et al., 2023). Integrating diverse knowledge systems can also guide managers toward places where communities have sustained ecosystems for generations, offering a starting point for targeted monitoring and protection under shared stewardship (Ban et al., 2020; Lepofsky et al., 2017).
• Consider rotational closure of areas not formally designated as Marine Protected Areas (MPAs) to help mitigate overharvesting and support long-term ecological resilience across the wider coastline. Permanent MPAs could play a role in sustaining the productivity of resilient marine ecosystems and facilitating spillovers into adjacent areas, while rotational MPAs can support species and ecosystems vulnerable to depletion. Spatial and temporal restrictions on harvesting across a wider scale may offer a more balanced approach to marine management, particularly for species that respond differently to high rates of fishing and collecting (André et al., 2022; Frid et al., 2016; Vave, 2022).
• Monitor vulnerable marine resources on a regular basis, such as high-trafficked mussel beds. California mussels benefit from rotational harvesting (Yamada and Peters, 1988), but overharvesting can rapidly reduce local populations and mussel sizes as simulated in this study. Community participation and place-based knowledge can allow for adaptive planning and cyclical management, involving occasional closures (André et al., 2022; Ban et al., 2020; Frid et al., 2016; Vave, 2022).

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. I acknowledge the assistance of Microsoft CoPilot to edit and improve the wording, clarity, and precision in the text. ChatGPT (Version 5) assisted with code development, as referenced in text where applicable.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fearc.2026.1771926/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Gabriel Sanchez, University of Oregon, United States

Reviewed by: Brian F. Codding, The University of Utah, United States

Iain McKechnie, University of Victoria, Canada