Insights from Amino Acid Isotope Analysis

University of New Mexico UNM Digital Repository

Biology ETDs Electronic Theses and Dissertations

Fall 12-6-2019

RECONSTRUCTING ENERGY FLOW THROUGH MODERN AND HISTORICAL MARINE COMMUNITIES: INSIGHTS FROM AMINO ACID ISOTOPE ANALYSIS

Emma A. Elliott Smith University of New Mexico

Follow this and additional works at: https://digitalrepository.unm.edu/biol_etds

Part of the Biogeochemistry Commons, Biology Commons, and the Ecology and Evolutionary Biology Commons

Recommended Citation Elliott Smith, Emma A.. "RECONSTRUCTING ENERGY FLOW THROUGH MODERN AND HISTORICAL MARINE COMMUNITIES: INSIGHTS FROM AMINO ACID ISOTOPE ANALYSIS." (2019). https://digitalrepository.unm.edu/biol_etds/341

This Dissertation is brought to you for free and open access by the Electronic Theses and Dissertations at UNM Digital Repository. It has been accepted for inclusion in Biology ETDs by an authorized administrator of UNM Digital Repository. For more information, please contact [email protected], [email protected], [email protected]. Emma A Elliott Smith Candidate Biology Department

This dissertation is approved, and it is acceptable in quality and form for publication: Approved by the Dissertation Committee:

Seth D Newsome , Chairperson

Joseph Cook

Thomas Turner

James A Estes

Torben C Rick

RECONSTRUCTING ENERGY FLOW THROUGH MODERN AND HISTORICAL MARINE COMMUNITIES: INSIGHTS FROM AMINO ACID ISOTOPE ANALYSIS by

EMMA A ELLIOTT SMITH

B.S., Biology, University of California Santa Cruz, 2013 M.S., Biology, University of New Mexico, 2018

DISSERTATION

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy Biology

The University of New Mexico Albuquerque, New Mexico

May 2020

ACKNOWLEDGEMENTS

I would like to extend my deepest appreciation to my advisor, Dr. Seth Newsome, whose mentorship and support has made this dissertation possible. I am also grateful to the members of my committee, Dr. Torben Rick, Dr. Jim Estes, Dr. Joe Cook and Dr. Tom Turner, for their feedback and guidance, and to Dr. Jim Norris, for help sampling historical kelps at the Smithsonian Institution. I would also like to thank Dr Marilyn Fogel, an amazing and inspiring researcher whose pioneering efforts have set a high bar for female isotope ecologists to strive towards.

I would like to acknowledge my funding sources, which made much of this work possible. These include fellowships through the NSF Graduate Research Program (DGE-1418062) North Pacific Research Board, and American Association of University Women–Santa Fe Chapter. UNM sources of funding include the Latin American and Iberian Institute, Graduate and Professional Student Association, Biology Department, Biology Graduate Student Association, and the Center for Stable Isotopes.

I also want to express my sincere gratitude to the many friends and colleagues who have made my time as a graduate student a joy. They are far too many to list, but include Sarah Foster, Alexi Besser, Dr. Cataline Tomé, Dr. Geraldine Busquets, all members of the Newsome Lab and UNM Center for Stable Isotopes. Special thanks also to Drs. Viorel Atudorei and Laura Burkemper for their help and expertise in the analytical portion of this project, as well as their boundless patience.

Lastly, I want to thank my family for their unconditional love and support. In particular my mother, Felisa Smith, who instilled in me an appreciation for the natural world and showed by example that women do belong in science, my father, Scott Elliott who has always challenged me to go the extra mile, my sister, Rosemary Elliott Smith, and my talented and creative grandmother, Maria del Carmen Calvo, who is deeply missed.

iii

RECONSTRUCTING ENERGY FLOW THROUGH MODERN AND HISTORICAL MARINE COMMUNITIES: INSIGHTS FROM AMINO ACID ISOTOPE ANALYSIS

Emma A Elliott Smith

B.S., Biology, University of California Santa Cruz

M.S., Biology, University of New Mexico

Ph.D., Biology, University of New Mexico

ABSTRACT

The fundamental currency of life is energy. Organisms need energy to grow, to survive and to reproduce. Understanding the acquisition of energy by consumers is thus a foundational aspect of biological research. This is especially important in the modern era, as impacts of ongoing anthropogenic effects will be mediated or amplified through food webs. Here, I explore how isotopic analysis of individual amino acids – a technique new to ecological studies – can be used to trace energy flow through animal communities in modern and ancient time periods. In particular, I focus on kelp forest food webs, which are nearshore marine ecosystems highly vulnerable to human impacts. I explore how important kelp

(macroalgae in family Laminariales) are as an energy source to consumers from localities as diverse as Katmai National Park, Alaska, and Antofagasta, Chile. I also examine how

iv the ecology of an important kelp forest species, the sea otter (Enhydra lutris) has changed over time in California. Finally, I employ a biochemical perspective to identify the mechanisms driving differences among producer isotopic values. My dissertation demonstrates that amino acid isotopic measurements can be used to confidently identify kelp-derived energy in consumers across space and time. Using this, I find that across different localities and oceanographic regions, kelp is vitally important to consumers as a source of energy. My work highlights the importance of understanding energy flow through food webs, and how this knowledge can help to better protect and manage biological systems.

TABLE OF CONTENTS

List of Figures ...... vii

List of Tables ...... viii

Chapter 1: The importance of kelp to an intertidal ecosystem varies by trophic level: insights from amino acid d13C analysis ...... 1

Chapter 2: Reductions in the dietary niche of southern sea otters (Enhydra lutris nereis) from the Holocene to the Anthropocene ...... 25

Chapter 3: Multichanneling in a Chilean kelp forest is facilitated by intraspecific specialization ...... 48

Chapter 4: Biochemical mechanisms and environmental influences on the amino acid isotopic composition of marine primary producers ...... 71

Appendix 1: Supplementary Materials for Chapter 1 ...... 101

Appendix 2: Supplementary Materials for Chapter 2 ...... 120

Appendix 3: Supplementary Materials for Chapter 3 ...... 146

Appendix 4: Supplementary Materials for Chapter 4 ...... 154

References ...... 161

LIST OF FIGURES

Figure 1.1 ...... 13

Figure 1.2 ...... 15

Figure 1.3 ...... 16

Figure 1.4 ...... 18

Figure 2.1 ...... 27

Figure 2.2 ...... 31

Figure 2.3 ...... 38

Figure 2.4 ...... 40

Figure 3.1 ...... 61

Figure 3.2 ...... 63

Figure 3.3 ...... 64

Figure 4.1 ...... 86

Figure 4.2 ...... 88

Figure 4.3 ...... 90

Figure 4.4 ...... 94

vii

LIST OF TABLES

Table 1.1 ...... 17

Table 2.1 ...... 32

Table 2.2 ...... 39

Table 3.1 ...... 55

Table 3.2 ...... 65

Table 4.1 ...... 80

Table 4.2 ...... 91

viii

Chapter 1

The importance of kelp to an intertidal ecosystem varies by trophic level – insights from amino acid d13C analysis

Emma A. Elliott Smith1, Chris Harrod2,3, Seth D. Newsome1

1Department of Biology, University of New Mexico, Albuquerque, NM 2Instituto de Ciencias Naturales Alexander von Humboldt, Universidad de Antofagasta, Chile 3Millennium Institute for Invasive Salmonids, Concepción, Chile

Abstract

A fundamental question in ecology is understanding how energy and nutrients move through and between food webs, and which sources of production support consumers. In marine ecosystems, these basic questions have been challenging to answer given the limitation of observational methods. Stable isotope analysis of essential amino acids (EAA d13C) has great potential as a tool to quantify energy and nutrient flow through marine food webs; however, it has been primarily utilized at large spatial scales. Here, we used EAA d13C analysis to test for connectivity between adjacent subtidal and intertidal components of a nearshore ecosystem in south central Alaska. We measured d13C of six

EAA from four marine producer groups: subtidal kelp (Laminaria sp.), offshore particulate organic matter (POM), and intertidal red (Neorhodomela sp.) and green (Ulva sp.) algae.

In addition, we sampled four intertidal invertebrate consumer species spanning a range of trophic/functional groups: Mytilus sp., Strongylocentrotus droebachiensis, Nucella sp., and

Pycnopodia helianthoides. Using canonical analysis of principal coordinates (CAP), and isotope mixing models (MixSIAR), we tested for differences among producer EAA d13C

‘fingerprints’ and quantified the contribution of producer EAA to consumers. We

1 compared these results to previously published EAA d13C data on marine producers to examine the generality of this technique. We found the EAA d13C fingerprints of subtidal kelps (Laminaria), Ulva and Neorhodomela were highly distinct from one another. Further, our measured EAA d13C patterns for kelp and red algae matched those previously reported from other localities, suggesting unique and universal EAA d13C signatures for these groups. However, CAP could not distinguish between microalgae (POM) and Ulva, possibly due to similar biochemical pathways for the synthesis of EAA. Using these producer fingerprints we found upper trophic level invertebrate consumers, Nucella and

Pycnopodia, derived more than 60% of their essential amino acids from subtidal kelps. In contrast, the sampled primary consumers in the system, Mytilus and Strongylocentrotus, relied more heavily on Ulva and/or offshore POM. Our results provide evidence for connectivity between two adjacent nearshore ecosystems and exemplify EAA d13C as a powerful new tool in tracing energy and nutrient flow within and among marine food webs.

Introduction

Understanding how energy and nutrients flow through food webs is a persistent theme in ecology. The types and timing of primary production available for consumers can have strong impacts on the length and stability of food chains, as well as population and community dynamics (Polis and Winemiller 1996, Rooney et al. 2006). Quantifying linkages within and among food webs is thus critical for mitigation of future biodiversity loss. This is especially topical for nearshore marine ecosystems, which are experiencing rates of defaunation and habitat alteration close to that seen in the terrestrial realm during

2 the industrial revolution (Jackson et al. 2001, McCauley et al. 2015). This loss of trophic complexity likely makes marine communities more vulnerable to eutrophication, disease, and climate change (Jackson et al. 2001, Estes et al. 2011).

Among nearshore marine ecosystems, kelp forests are some of the most biodiverse and the most vulnerable (Dayton 1985, Steneck et al. 2002). Serving as foundation taxa, the extensive biomass of kelps (order Laminariales) creates complex and three-dimensional

‘forest’ structures, which serve as both physical habitat and a potentially large, perennial resource for consumer communities within or adjacent to these systems (Mann 1973,

Steneck et al. 2002). This kelp-derived production can enter the food web through direct grazing by primary consumers or, as kelp blades senesce and decay, via detrital material incorporated into particulate organic matter (POM) that is subsequently consumed by suspension feeders in subtidal or intertidal habitats (Duggins et al. 1989). Nearshore food webs can also be supported by seasonal or ephemeral blooms of both phytoplankton and understory macroalgae (Page et al. 2008, von Biela et al. 2016, Docmac et al. 2017). This mixture of (slow) perennial kelp production and (fast) ephemeral production likely adds stability to these ecosystems as consumers can switch resources depending on shifts in availability (Rooney et al. 2006, Rooney and McCann 2012). However, this complexity also makes it challenging to study energy and nutrient flow within kelp forests and their connectivity to adjacent ecosystems. Consequently, the importance of kelp production to consumers is still debated (e.g., Docmac et al. 2017).

Isotope-based approaches are now standard tools in ecological studies, especially in marine ecosystems where observation-based methods are limited (Newsome et al. 2010).

Most isotope studies take advantage of naturally occurring variation in carbon (d13C) and

3 nitrogen (d15N) isotopes among primary producers, which can act as biomarkers, allowing researchers to trace energy and nutrient flow through food webs and between ecosystems

(Ben-David and Flaherty 2012). Traditionally, research has focused on ‘bulk’ d13C or d15N analyses of tissues or whole organisms: however, the inferences that can be drawn from bulk isotope analysis are limited by substantial spatiotemporal isotopic variation among producers and consumers due to biotic and abiotic processes. Much of this variation is driven by the fact that bulk tissue d13C values represent a weighted average of the d13C of all macromolecules within the sample: carbohydrates, lipids and proteins (Wessels and

Hahn 2010, Fox 2013). These molecules exhibit widely different isotopic values and vary substantially in their concentration within tissues (Fox 2013, Larsen et al. 2013, 2015). For example, in California, a 10‰ increase in d13C was observed in rapidly growing giant kelp

(Macrocystis pyrifera) fronds relative to mature blades, likely due to the increased carbohydrate concentration in new tissues (Fox 2013). Marine producer d13C values can

- also vary substantially depending on the availability and utilization of HCO3 vs CO2

(Raven et al. 2002). In some cases, this variability can lead to overlapping isotope values between benthic and pelagic primary producers, limiting the utility of isotopes as tracers in marine food webs (Page et al. 2008). In addition to baseline variation, trophic discrimination factors between consumers and food sources can vary widely depending on consumer tissue type, life history, diet, and physiological state (Ben-David and Flaherty

2012). Combined, these factors make it challenging to determine whether changes in consumer bulk stable isotope values are due to dietary, physiological or baseline shifts, especially in highly dynamic ecosystems such as kelp forests.

Recent developments in compound-specific stable isotope analysis provide a way of untangling complex ecosystem dynamics, while avoiding many of the pitfalls associated with bulk isotope studies. The analysis of individual essential amino acids (EAA) is particularly useful for food web research, as these macromolecules are a major conduit of energy/nutrient flow between trophic levels (Larsen et al. 2013). Recent studies (Scott et al. 2006, Larsen et al. 2013) have demonstrated that different producer taxa (e.g., macroalgae vs. terrestrial plants) have distinct patterns of EAA d13C values depending on their metabolic pathways and primary source of carbon. These unique combinations of individual EAA d13C values associated with certain primary producers act as ‘fingerprints’ that can be detected at higher trophic levels (Larsen et al. 2009, 2013). Moreover, because only photoautotrophs and microbes can synthesize EAA de novo, consumers must acquire them via direct consumption or through gut microbial activity when dietary protein is limiting (Newsome et al. 2011). As a result, producer EAA d13C fingerprints remain largely intact as they move up food chains, passing from prey to consumer with minimal isotopic alteration (Howland et al. 2003, Larsen et al. 2013, McMahon et al. 2015). In addition, studies of diatoms and the giant kelp suggest that variable growth conditions do not substantially alter the patterning between individual EAA d13C values (i.e., the

‘fingerprints’) of primary producers, even when bulk tissue d13C values vary widely

(Larsen et al. 2013, 2015). To date, most publications reporting EAA d13C data are global compilations of taxonomically diverse samples collected from a wide range of ecosystems, and/or cultured in a laboratory (Scott et al. 2006, Larsen et al. 2009, 2013). To our knowledge, no study has fully characterized the EAA d13C fingerprints of primary

5 producers at a local scale, and few have done so at a regional level (Larsen et al. 2012,

McMahon et al. 2016).

13 Here, we use EAA d C analysis to trace linkages between subtidal and intertidal habitats at a site in south central Alaska. Specifically, we ask whether subtidal/offshore derived production in the form of kelp and POM is important for intertidal invertebrate consumers spanning a range of trophic and functional groups. It would be reasonable to expect that consumers are most reliant on local sources of production. Thus, we initially anticipated our sampled intertidal invertebrates would rely more on intertidal red and green algae rather than subtidal kelp and offshore POM. Conversely, a large degree of kelp/POM derived EAA in intertidal consumers would suggest a high degree of connectivity between the adjacent subtidal and intertidal ecosystems. We also compare our results to bulk tissue

13 stable isotope data from the same site and to previous studies that reported EAA d C data from similar producers (Larsen et al. 2013) to examine the generality between local and larger scale applications of this technique.

Materials and Methods

Study Site and Sample Collection

We sampled marine producers and macroinvertebrate consumers from Amalik Bay,

Katmai National Park, Alaska (Appendix 1 Table 1). Macroalgae and invertebrates were sampled in July 2012. Offshore particulate organic matter (POM) samples were collected in July 2012 (n = 1) and July 2013 (n = 4) at locations between 0.5 and 2km from the coast, with the farthest being 40km from Amalik Bay (Appendix 1 Table 1). For the 2012 POM

6 singleton we were only able to obtain bulk d13C and d15N values. For 2013 POM samples, we obtained bulk d13C, d15N, and EAA d13C values.

We analyzed four dominant primary producer groups, and four of the most common invertebrate species spanning a range of trophic levels/functional groups (Dean et al. 2014;

Appendix 1 Tables 1 and 2). We sampled the following producers: perennial kelp

(Laminaria sp.), red algae (Neorhodomela sp.), fast-growing green algae (Ulva sp.), and

POM. Among invertebrate species we sampled: a filter feeder (Mytlius sp.), a grazer/herbivore (Strongylocentrotus droebachiensis), a carnivore (Nucella sp.), and an apex echinoderm predator (Pycnopodia helianthoides; Dean et al. 2014); henceforth, we refer to each species by their genus. Sampling of intertidal red and green macroalgae and sessile/low mobility invertebrates was conducted along two vertical strata placed at 0.5 and

1.5 meters above Mean Low Low Water (MLLW); motile invertebrates (e.g., Pycnopodia) were sampled along transects parallel to shore, with the lower boundary at MLLW. Mature kelps were sampled subtidally via SCUBA. POM samples were collected from the water column; ~1–2 liters of water were pumped through pre-combusted 0.7 μm GF/F filters.

After collection, specimens were kept on ice for a period ranging from ca. 24 to 72 hours before being frozen (-20°C). Specimens were then kept frozen until processing for stable isotope analysis. Collecting permits were issued to the US Geological Survey and US

National Park Service.

Bulk d13C and d15N Analysis

For bulk analysis, macroalgae specimens were rinsed with deionized water, and lyophilized. Dried macroalgae were roughly homogenized, weighed out to ~5mg and

7 sealed in tin capsules for bulk isotope analysis. Pieces of filters containing POM were lyophilized, weighed out to ~20mg and sealed in tin capsules. Mytilus, Strongylocentrotus, and Nucella were shucked, homogenized and subsampled. For Pycnopodia, proteinaceous tube feet were sampled. The whole-body samples (Mytilus, Strongylocentrotus, and

Nucella), or tube feet (Pycnopodia), likely represent at least several months of dietary information (Ebert 1968, Burrows and Hughes 1990, Gooding et al. 2009). Invertebrate samples for bulk isotope analysis were not lipid extracted, were weighed out to 0.5–0.6mg, and sealed in tin capsules. We did not use a posterior d13C lipid correction because current methods do not work well for invertebrate taxa (Kiljunen et al. 2006). Bulk d13C and d15N isotope values, along with mass percent [C] and [N], were measured via EA-IRMS using a Costech elemental analyzer (Costech, Valencia, CA) coupled to a Thermo Scientific

Delta V isotope ratio mass spectrometer (Thermo Fisher Scientific, Bremmen, Germany).

The within-run standard deviation of reference materials was ≤ 0.2‰ for d13C and d15N values.

Amino Acid d13C Analysis (EAA d13C)

We used the same macroalgae and invertebrate subsamples for bulk and EAA d13C analysis, except for two Nucella individuals (NUC-2012-06 and -08; Appendix 1 Table 1).

After bulk preparation, these individuals did not have material remaining, so two additional

Nucella (NUC-2012-02 and -04; Appendix 1 Table 2) were analyzed. For EAA d13C analysis, invertebrate subsamples were lipid extracted via four ~24h soaks in petroleum ether, rinsed five times with deionized water and lyophilized. Approximately 5 and 10mg of lipid-extracted invertebrate and macroalgae tissue respectively was hydrolyzed to

8 constituent amino acids in 1mL of 6N hydrochloric acid (HCl) at 110°C for 20 hours; tubes were flushed with N2 to prevent oxidation during hydrolysis. Whole silica filters containing

POM were hydrolyzed with 1.5mL of 6N HCl. Hydrolyzed producer samples were passed through a cation exchange resin column (Dowex 50WX8 100-200 mesh) to isolate amino acids from other metabolites (Amelung and Zhang 2001). After hydrolysis/Dowex purification, amino acids were derivatized to N- trifluoroacetic acid isopropyl esters following established protocols (O’Brien et al. 2002, Newsome et al. 2011, 2014). Samples were derivatized in batches of 8–17 along with an in-house reference material containing all amino acids we measured.

d13C values of individual derivatized amino acids were measured using a GC-C-

IRMS system. Derivatized samples were injected into a 60m BPx5 gas chromatograph column for amino acid separation (0.32 ID, 1.0μm film thickness, SGE Analytical Science,

Victoria, Australia) in a Thermo Scientific Trace 1300, combusted into CO2 gas via a

Thermo Scientific GC Isolink II, and analyzed with a Thermo Scientific Delta V Plus isotope ratio mass spectrometer. Samples were run in duplicate or triplicate and bracketed with our reference material. Samples were re-run if any EAA exhibited standard deviations

> 0.7‰ across injections. The within-run standard deviations of measured d13C values among EAA in the in-house reference material ranged from 0.2‰ (isoleucine) to 0.8‰

(lysine).

We reliably measured d13C values of six amino acids considered essential for animals: isoleucine (Ile), leucine (Leu), lysine (Lys), phenylalanine (Phe), threonine (Thr) and valine (Val). For one Laminaria and one Ulva sample, we were not able to obtain reliable Phe d13C values (Appendix 1 Table 2), and instead used the mean phenylalanine

9 d13C value for their respective group. The reagents used during derivatization add carbon to the side chains of amino acids, and hence raw d13C values measured via GC-C-IRMS reflect a combination of the intrinsic amino acid carbon and reagent carbon (Silfer et al.

1991). See Appendix 1 for correction equations.

Statistical Analyses

We used methods similar to Larsen et al. (2013) to develop unique EAA d13C patterns (fingerprints) of our sampled producers and used these in concert with Bayesian mixing models (MixSIAR; Stock and Semmens 2016) to identify the most important source of production for consumers. For EAA d13C statistical analyses, we used the mean corrected d13C value across multiple injections of a sample. We tested all data for normality and homogeneity of variance (Tabachnick and Fidell 2013). To test for normality in our producer and invertebrate datasets, we ran multivariate regressions using species as

13 15 13 predictor of either bulk d C and d N, or all EAA d C values, and evaluated the normality of residuals using fitted vs. residuals plots, Q-Q plots, D’Agostino-Pearson’s, and Cramer von Mises Tests. We tested homogeneity of variance via the Brown and Forsythe test. We used MANOVA to test for differences in EAA d13C values among macroalgae and invertebrate groups. We used ANOVA and Tukey’s HSD for individual essential amino acids and individual bulk isotopes to test for differences among producers and consumers.

We ran all statistical analyses, except canonical analysis of principal coordinates (CAP), in program R (v 3.3.1) with RStudio interface (v 0.99.903). We ran CAP using PRIMER7

+ PERMANOVA (Anderson et al. 2008).

We used stable isotope mixing models of both bulk and EAA d13C values

(MixSIAR; Stock and Semmens 2016) to assess the relative contribution of marine producers to intertidal invertebrate consumers. For the bulk model, we used d13C and d15N from the four producer groups (Laminaria, Neorhodomela, POM and Ulva) as source inputs, and ran each invertebrate consumer species separately. We assumed mean ±SD trophic discrimination factors of 1.0 ± 0.5‰ and 3.0 ± 0.5‰ for d13C and d15N respectively, with Mytilus/Strongylocentrotus at TL=1, Nucella at TL=2, and Pycnopodia at TL=3.

Similarly, we ran the model with raw EAA d13C values (corrected for carbon added during derivatization) from the producer groups as source inputs for each invertebrate consumer species. We assumed d13C values of EAA undergo minimal modification through the food chain (Howland et al. 2003, McMahon et al. 2015) and thus used a trophic discrimination factor of 0‰ for all EAA. The model was run at length ‘very long’, at which point diagnostics indicated the MCMC chains had mostly converged (Stock and Semmens 2016).

Because Ulva and POM EAA d13C values were not well differentiated by CAP, we pooled the contribution of these sources a posteriori.

To characterize producer EAA d13C fingerprints and test for these in our consumers, we used CAP (Anderson and Willis 2003), a robust, distance-based alternative to the linear discriminant analysis (LDA) used previously to analyze EAA d13C fingerprints (Larsen et al. 2013). LDA is often advantageous in that it provides a probability of classification for each consumer based on their LD scores. However, the utility of LDA is limited by a set of assumptions (multivariate normality, homogeneity of variance), it is sensitive to small/unbalanced sample sizes, and requires more samples than variables (Tabachnick and

Fidell 2013). The use of a distance-based approach such as CAP removes the issue of

11 balancing sample size with the number of variables by reducing the number of dimensions, increasing its utility in EAA d13C studies where sample size is typically limited due to time and cost considerations.

CAP was run using a Euclidian distance matrix (data untransformed and un- normalized), with 9999 permutations. We first used the EAA d13C data of our primary producers to create a training dataset. This included a leave-one-out and cross validation procedure to obtain an estimate of error rate for our classifier samples. This producer training dataset was then used to classify the EAA d13C values of each individual invertebrate. In addition, we used CAP to compare our producer data to marine producers provided by Larsen et al. (2013). We note that due to the lack of a direct interlab calibration via shared reference materials, this comparison is largely descriptive in nature (i.e., comparing fingerprints across broad taxonomic groups) rather than a test of differences in measured EAA d13C values. Specifically, we compared our data to macroalgae

(Rhodophyta and Phaeophyceae) collected in California, cultured/naturally collected microalgae, and terrestrial plant material from the North Slope of Alaska (see Appendix 1

Table 3; Larsen et al. 2013 for raw data and collection details). We ran the comparison

CAP model two ways: (1) using EAA d13C values from Larsen et al. (2013) as a training dataset to classify our producer samples, and (2) grouping our local samples with appropriate Larsen et al. (2013) producers - Laminaria with Phaeophyceae, Neorhodomela with Rhodophyta, and POM with Microalgae. Larsen et al. (2013) did not report EAA d13C data from green macroalgae, so we left our Ulva as a unique group.

Figure 1.1 Bulk d13C and d15N values (Mean ± SD) for marine producers and consumers collected from Amalik Bay in Katmai National Park, Alaska. Samples were collected in 2012/2013, and all groups except Laminaria and POM. POM = offshore particulate organic matter.

Results

Bulk d13C and d15N Values

Laminaria had the highest bulk d13C values among producers with >5‰ separating

this group from other producers (Appendix 1 Table 1; Figure 1.1). The other producers had

overlapping d13C values, and similar d15N values with the exception of Ulva and POM

(Appendix 1 Tables 1 and 4; Figure 1.1). The two primary consumers Mytilus (filter feeder)

and Strongylocentrotus (grazer) had the lowest and overlapping d13C and d15N values, and

Pycnopodia, (higher carnivore) had the highest values (Appendix 1 Table 1; Figure 1.1).

Invertebrate C:N ratios ranged from 3.5 to 7.8 (Appendix 1 Table 1), indicating contribution of lipids to bulk d13C values (Ambrose 1990).

EAA d13C Values

MANOVA results indicated strong differences in EAA d13C values both among producer (Pillai Trace = 1.77, F39,18 = 3.58, p = 0.0002) and invertebrate groups (Pillai

Trace = 1.92, F72,18 = 7.07, p < 0.00001); see Appendix 1 Table 5 for statistical output of all individual pairwise comparisons. Consistent with the patterns observed in the bulk isotopes, Laminaria had the highest d13C values for all six EAA that we measured

(Appendix 1 Table 5; Figure 1.2). Among consumers, Pycnopodia had the highest

EAA d13C values, and Mytilus and Strongylocentrotus had the lowest (Appendix 1 Tables

2 and 5; Figure 1.2). Three invertebrate samples (two Strongylocentrotus and one Nucella) were outliers as shown by residual versus fitted plots, and the failure of the invertebrate dataset to pass normality when these samples were included. However, running ANOVAs with and without these samples showed little difference in our results, and we opted to include these samples in the remainder of our analyses.

Figure 1.2. Essential amino acid (EAA) d13C values (Mean ± SD) for Amalik producers and consumers. Lower panel (A) shows EAA d13C values for producers. POM = offshore particulate organic matter. Upper panel (B) shows EAA d13C for intertidal consumers. Abbreviations for essential amino acids are as follows: isoleucine (Ile), leucine (Leu), lysine (Lys), phenylalanine (Phe), threonine (Thr), valine (Val). Dotted lines are solely a visualization tool to illustrate amino acid patterns among groups.

Mixing Model Results

The estimated proportion of kelp-derived EAA carbon to consumers varied by

putative trophic level and functional group. Mytilus, had the lowest contribution from

Laminaria (median dietary proportion [95% credibility interval] = 0.33 [0.24, 0.41]), and

highest contribution from Ulva/POM: 0.57 [0.29, 0.72] (Appendix 1 Table 6; Figure 1.3).

Strongylocentrotus had the second highest contribution from Ulva/POM (0.51 [0.22, 0.63];

Appendix 1 Table 6; Figure 1.3). Pycnopodia had the highest proportion of kelp EAA

carbon (0.79 [0.72, 0.87]), and Nucella had the second highest (0.66 [0.59, 0.73]; Appendix

1 Table 6; Figure 1.3). Bulk mixing model results showed similar patterns but with larger

credibility intervals (Appendix 1 Table 8).

Figure 1.3. Mixing model results (MixSIAR) showing the contribution of essential amino acids from different marine producers to Amalik consumers. Results are shown for different intertidal invertebrate species, arranged from left to right in order of increasing trophic level: Mytilus, Strongylocentrotus, Nucella, Pycnopodia. Circles represent the median contribution of each producer group, to the consumer species, with surrounding 95% credibility intervals (Appendix 1 Table 6). POM = particulate organic matter.

CAP Results

CAP of Amalik Bay producer EAA d13C values exhibited an overall successful reclassification rate of 77% (Table 1.1). CAP was most successful at distinguishing kelp from other groups: 100% of Laminaria individuals were correctly reclassified, and no

Ulva, Neorhodomela or POM samples were incorrectly classified as kelp (Table 1.1, Figure

1.4). Similarly, all four Neorhodomela samples were correctly reclassified. 60% of the

Ulva samples were correctly reclassified, with two samples misclassified as POM. Only

25% of the four POM samples were successfully reclassified: two were incorrectly classified as Ulva and one as Neorhodomela (Table 1.1, Figure 1.4).

Table 1.1. Canonical analysis of principal coordinates (CAP) classification table for producers and consumers. Columns represent classification of a sample group as given by CAP. Rows represent the group being classified. For the producer section (top half of table) the cross diagonal (bold text), are correct reclassifications. Numbers falling off this diagonal indicate samples that were incorrectly reclassified by the model. Thus, the sum of the cross diagonal divided by the total number of samples equals the % success of the model (77%). See Figure 1.4 for a visual representation of these data. POM = offshore particulate organic matter. CAP Classification

% Laminaria Neorhodomela POM Ulva Correct

Laminaria 9 0 0 0 100 Amalik Producer Neorhodomela 0 4 0 0 100

POM 0 1 1 2 25

Ulva 0 0 2 3 60

Mytilus 0 0 0 8 –

Amalik Strongylocentrotus 0 0 0 7 –

Consumer Nucella 3 0 0 5 –

Pycnopodia 6 0 0 2 –

Comparisons of our producer data with similar data from Larsen et al. (2013) showed that

kelp, red algae and POM from the two studies had similar EAA d13C fingerprints

(Appendix 1 Table 7, Figure 1). Eight of nine (89%) Laminaria samples classified correctly

with Phaeophyceae from California; one classified with cultured microalgae. Similarly, all

Neorhodomela samples (100%) were correctly classified with Rhodophyta. Three of four

(75%) Amalik POM samples were grouped with microalgae, and one with Rhodophyta.

Figure 1.4. Essential amino acid d13C fingerprinting of Amalik producers and consumers. Circles represent the CAP (canonical analysis of principal coordinates) loading of individual producer samples, and stars represent the centroid for each group. POM = offshore particulate organic matter. Open diamonds are CAP loadings for consumer individuals. Panels are arranged from bottom to top in order of increasing trophic level: (A) Strongylocentrotus (grazer), (B) Mytilus (filter feeder), (C) Nucella (carnivore), and (D) Pycnopodia (apex invertebrate consumer). Where a consumer plots in this space is a reflection of the predominant source of production they ultimately derived their essential amino acids from: the color of each consumer individual represents which producer group they were classified with (Table 1.1).

All Amalik Bay Ulva samples were classified as Microalgae. When Amalik Bay producers were combined with similar taxa from Larsen et al. (2013), the model had a successful reclassification rate of 71%. Respectively, 18/21 Phaeophyceae, 11/13 Rhodophyta, 18/31

Microalgae and 3/5 Chlorophyta were reclassified correctly (Appendix 1 Table 7). None of our samples were classified with terrestrial plants from Larsen et al. (2013; data not shown).

The macroinvertebrate consumer taxa varied in which producer fingerprint they exhibited (Table 1.1, Figure 1.4). Six Pycnopodia individuals (75%) were classified with

Laminaria, and two with Ulva (Figure 1.4). Three Nucella (63%) classified with

Laminaria, and five (38%) with Ulva (Figure 1.4). All Strongylocentrotus and Mytilus were classified with Ulva (Figure 1.4). No sampled macroinvertebrates were classified with

Neorhodomela or offshore POM.

Discussion

Our study provides strong evidence for the connectivity between two adjacent nearshore ecosystems in south central Alaska – a subtidal kelp forest and a rocky intertidal site. We find that our cutting-edge technique (EAA d13C analysis) improves on bulk tissue isotope analysis and is well-suited to quantifying such linkages between kelp forests and adjacent habitats; Laminaria, the dominant kelp at our site, exhibited a highly distinct EAA d13C fingerprint in comparison to all other locally abundant producers that we sampled

(Table 1.1, Figure 1.4). Furthermore, our results suggest that the degree of this connectivity can in part be mediated by the functional or trophic role of a species. Specifically, we find a large degree of subtidal kelp-derived EAA in upper trophic level intertidal invertebrate

19 consumers, whereas grazers and filter feeders likely rely more on local producer groups.

Given the widespread distribution of kelp forests at temperature latitudes, our results have implications in the study of a complex and far reaching suite of habitats (Dayton 1985,

Steneck et al. 2002).

EAA d13C fingerprinting allowed us to confidently distinguish kelp-derived energy/nutrients from other local sources of production. Canonical analysis of principal coordinates (CAP) showed Laminaria individuals significantly differed to the other producer groups (Figure 1.4); Laminaria individuals were correctly reclassified 100% of the time with no other producers incorrectly classified into this group (Table 1.1). It is important here to distinguish EAA d13C fingerprints (as in Larsen et al. 2009) from raw

EAA d13C values. Previous work has found EAA d13C fingerprints (or the patterns among individual EAA d13C values) are driven by highly-conserved differences in the biochemical pathways used to synthesize amino acids de novo. Such pathways are thought to be conserved across variable environmental and physiological conditions (Scott et al. 2006,

Larsen et al. 2009, 2013, 2015). In support of this idea, our comparison with the data of

Larsen et al. (2013) found consistency in EAA d13C fingerprints of kelps across space and time. Seven of our eight sampled Laminaria classified correctly into the Larsen et al.

(2013) Phaeophyceae group, a dataset composed predominantly of samples from the iconic giant kelp (Macrocystis pyrifera; Appendix 1, Table 3). This agreement between EAA d13C fingerprints of Alaskan and Californian kelps hints at a unique EAA d13C signature for this group. We note, however, that the lack of an interlab comparison between the two studies means this result should be taken as suggestive rather than conclusive. The high bulk d13C values of kelps are typically explained as due to their extremely fast growth rates (2-3%

20 per day; Reed et al. 2008) which lead to decreased fractionation between 12C and 13C

- species of aqueous CO2 (Simenstad et al. 1993), and their high utilization of HCO3 (Raven et al. 2002, Page et al. 2008). We suspect the highly distinct EAA d13C fingerprint for kelps is also reflective of these traits, although a mechanistic understanding of EAA biosynthesis in kelps is an avenue for future research.

Among the other three producer groups, there was substantial overlap in EAA d13C fingerprints, predominantly due to the high variance among POM samples. Despite this,

CAP could distinguish the dominant local red algae (Neorhodomela) from the local green algae (Ulva); no individual from either group was misclassified with the other, and these groups separated well along both the first and second canonical axes (Table 1.1, Figure

1.4). These results are in stark contrast with bulk tissue isotope analysis, where

Neorhodomela and Ulva overlapped in d13C and d15N (Figure 1.1), highlighting the power of EAA d13C analysis relative to analysis of bulk tissues. In contrast, POM samples displayed poor classification, with half of the samples incorrectly classified as Ulva, and one incorrectly classified as Neorhodomela (Table 1.1). Consequently, we use caution in interpreting consumers classified with Ulva vs. POM in this study.

This overlap between Ulva and POM EAA d13C fingerprints was not unique to our study site. Microalgae reported in Larsen et al. (2013) and our POM overlapped substantially with Ulva, and these groups had the lowest successful reclassification rates of 58% and 60% respectively (Appendix 1, Table 7). This may reflect similar and relatively simplistic biochemical pathways for EAA synthesis among Ulva and planktonic producers

(Abbott and Hollenberg 1992). Alternatively, the large variability in POM EAA d13C values reported in this study may reflect variability among samples in suspended detrital

21 material. In our system, this may be driven by terrestrial influences as Katmai National

Park has many large glaciers and rivers, and several POM samples were collected near these features (see Methods). Terrestrial plant material tends to have lower d13C bulk values than most phytoplankton or macroalgae (Page et al. 2008). This pattern is suggested by our POM data, as POM collected from Kukak Bay, an area into which three large rivers empty, had the lowest EAA d13C values of any sample we measured (Appendix 1, Table

2). However, we found no evidence that terrestrial material was a substantial component of our POM samples when running a CAP using Alaskan terrestrial plants (data not shown) reported in Larsen et al. (2013). To better understand the drivers of such isotopic variability, future studies should focus on examining POM EAA d13C fingerprints at a single locale with more samples collected over seasons and years.

It is important to note that like linear discriminant analysis (LDA), the classification of consumers by CAP will be sensitive to the producers included in the model, and the variability of each producer EAA d13C fingerprint. With CAP and LDA, classification of a given consumer is strictly by distance to the nearest producer centroid, weighted for the proportion of variance explained by each canonical axis (Tabachnick and Fidell 2013).

This is a significant limitation, as models will force a sample classification and by definition, consumers cannot be a mix of different sources. Thus, it is especially important to pair classification analysis with other statistical approaches, such as mixing models, which permit consumers to be a ‘mixture’ of different sources and incorporate more ecological realism. Although EAA d13C fingerprinting is rapidly becoming a powerful tool to study energy and nutrient flow in food webs, like any tool it needs to be applied appropriately and paired with ecological intuition/ancillary information. For future studies,

22 we urge researchers to continue to sample and analyze local producers, as further work examining the spatiotemporal consistency of EAA d13C fingerprints will be crucial in the continuing development and application of this technique.

EAA d13C fingerprints and mixing model analysis of species in Amalik Bay indicated a large degree of subtidal kelp-derived carbon in intertidal invertebrates. We found functional group to be a strong determinant in whether a species was utilizing this

‘slow’ and perennial energy/nutrient pathway. For the filter feeding bivalve (Mytilus) and grazing sea urchin (Strongylocentrotus), CAP and mixing model analysis indicated these species derived ~30–40% of their EAA from kelp, with the remainder predominantly from

Ulva/POM (Appendix 1, Table 6; Figures 1.3 and 1.4). In contrast, kelp-derived amino acids were the most important for the carnivorous whelk (Nucella) and predatory seastar

(Pycnopodia), with ~65% and ~80% of their EAA derived from kelp respectively

(Appendix 1, Table 6; Figures 1.3 and 1.4). Interestingly, this pattern of trophic-specific reliance on kelp has been previously documented. For example, using both bulk tissue d13C and d15N and gut content analysis, Koenigs et al. (2015) found a positive but noisy relationship between the contribution of kelp-derived energy and the trophic position of fish species in a California kelp forest. Here, we suspect that the greater degree of kelp- derived EAA in the two upper trophic level species is related to their degree of motility.

There are two possible pathways by which subtidal kelps are fueling intertidal Nucella and

Pycnopodia: either (a) kelp detritus washes into the intertidal and is incorporated by primary consumers that are in turn consumed by these carnivores, or (b) Nucella and

Pycnopodia spend a significant portion of time in subtidal habitats foraging on consumers that rely on kelp or kelp detritus. The relatively low degree of kelp-derived EAA in our

23 sampled intertidal filter feeder, Mytilus (Table 1.1, Appendix 1, Table 6; Figures 1.3 and

1.4), a potential prey species for both Nucella and Pycnopodia (Burrows and Hughes 1990,

Dean et al. 2014) provides support for the second pathway.

More broadly, our results lend support to a growing body of literature demonstrating the high degree of connectivity among nearshore marine communities. We show that at our site, connectivity is mediated by the different energy/nutrient pathways utilized by consumers – with subtidal kelp-derived EAA predominant in the sunflower star

(Pycnopodia) and dog whelk (Nucella), and intertidal Ulva/offshore POM most important to the sea urchin (Strongylocentrotus) and mussel (Mytilus) (Table 1.1, Appendix 1, Table

6; Figures 1.3 and 1.4). This coupling of adjacent ecosystems, whether subtidal-intertidal, pelagic-benthic (von Biela et al. 2016, Docmac et al. 2017), or marine-terrestrial (Dunton et al. 2012) has strong implications for conservation efforts of these systems. Most notably, it means that nearshore food webs must be examined in a broader context that includes potential benthic, offshore, and terrestrial sources of energy and nutrients to consumers.

Moreover, the subsidies from each source may be influenced by the trophic position or functional group of a species (Koenigs et al. 2015, von Biela et al. 2016). Here, we provide a case example of a tool – EAA d13C fingerprinting – that can be employed to quantitatively assess connectivity among ecosystems, a task that will be crucial in this modern era of intense and rapid global change.

Chapter 2

Reductions in the dietary niche of southern sea otters (Enhydra lutris nereis) from the

Holocene to the Anthropocene.

Emma A. Elliott Smith1, M. Tim Tinker2,3, Emily L. Whistler4, Douglas J. Kennett5, René

L. Vellanoweth6, Diane Gifford-Gonzalez7, Mark G. Hylkema8, Seth D. Newsome1

1Department of Biology, University of New Mexico, Albuquerque, NM 2Nhydra Ecological Consulting, St Margaret’s Bay, Nova Scotia, Canada 3Department of Ecology and Evolutionary Biology, University of California Santa Cruz, Santa Cruz, CA 4Department of Anthropology, Washington State University, Pullman WA 5Department of Anthropology, University of California, Santa Barbara, CA 6Department of Anthropology, California State University Los Angeles, Los Angeles, CA 7Department of Anthropology, University of California Santa Cruz, Santa Cruz, CA 8California Department of Parks and Recreation, Santa Cruz, CA 95060 USA

Abstract

The sea otter (Enhydra lutris) is a marine mammal hunted to near extinction during the 1800s. Despite their well-known importance as a keystone species in nearshore environments, we still know little about historical sea otter ecology, although such knowledge is imperative for effective conservation efforts. Here, we characterize the ecological niche of ancient southern sea otters (E. lutris nereis) using d13C and d15N analysis of bones recovered from archaeological sites spanning ~7,000 to 350 years before present (N= 112 individuals) at five regions along the coast of California. These data are compared with previously published data on modern animals (N= 165) and potential modern prey items. In addition, we analyze the d15N of individual amino acids for 23 individuals (18 ancient, 5 modern) to test for differences in sea otter trophic ecology through time. After correcting for tissue-specific and temporal isotopic effects, we employ

25 nonparametric statistics and Bayesian niche models to quantify differences among ancient and modern animals. We find ancient otters from the five regions occupied a larger isotopic niche than nearly all modern localities; this likely reflects broader habitat and prey use in pre-fur trade populations. In addition, ancient sea otters at the most southerly sites occupied an isotopic niche that was more than twice as large as ancient otters from northerly regions.

This is likely a consequence of greater invertebrate prey diversity in southern California relative to northern California. Thus, we suggest the potential dietary niche of sea otters in southern California could be larger than in central and northern California. At the two most northerly sites, Año Nuevo and Monterey Bay, ancient otters had significantly higher mean d15N values than modern populations. Amino acid d15N data indicated this resulted from shifting baseline isotope values, which reflects changing oceanographic conditions, rather than a change in sea otter trophic ecology. Our results help in better understanding the contemporary ecological role of sea otters and exemplify the strength of combing zooarchaeological and biological information to provide baseline data for conservation efforts.

Introduction

The sea otter, Enhydra lutris (Figure 2.1), is a marine mammal found in coastal nearshore ecosystems along the North Pacific. Across much of their range – Alaska, British

Columbia, Washington, Oregon, and central California (Bodkin 2015) – sea otters are considered a “keystone species” (Paine 1969), disproportionately influencing ecosystem structure and function through indirect effects of predation (Estes and Palmisano 1974,

Hughes et al. 2016). The control by sea otters of key invertebrate populations prevents the

overgrazing of kelp forests, and clears seagrass of harmful epiphytes, allowing these

primary producers to flourish and provide habitat for a range of diverse taxa (Estes and

Palmisano 1974, Steneck et al. 2002, Hughes et al. 2016). Sea otters thus help maintain

resilient coastal ecosystems, increase nearshore productivity and provide valuable

ecosystem services (Estes and Palmisano 1974, Wilmers et al. 2012, Hughes et al. 2016).

The current global sea otter population and range are severely reduced from

historical levels, due to commercial fur trade activity in the 18th and 19th centuries

(Riedman and Estes 1990). Prior to the initiation of the North Pacific fur trade in the mid-

1700s, the global sea otter population may have been 150,000–300,000 individuals, in a

more or less contiguous distribution around the north Pacific rim from Russia to Baja

California (Riedman and Estes 1990). Over the next ~150 years, sea otters were hunted

Figure 2.1. Southern sea otter (Enhydra lutris nereis) consuming a striped shore crab (Pachygrapsus crassipes). Photo taken in Moss Landing, California, by Joe Tomoleoni. Used with permission.

27 nearly to extinction; by the early 20th century only 1,000–2,000 were left (Riedman and

Estes 1990). In 1911 when the International Fur Seal Treaty afforded sea otters some protection small remnant populations persisted in Russia, southwest Alaska, Haida Gwaii,

Prince William Sound and central California. From the mid 1960s onwards, otters were periodically translocated from these remnant populations to southeastern Alaska, British

Columbia, Washington, Oregon, and other California localities (Bodkin 2015). Today, sea otter populations from eastern Russia to southeastern Alaska and British Columbia have recovered much of their pre-fur trade distribution (Riedman and Estes 1990). In contrast, the coastal areas from Washington to southern California support relatively small and spatially isolated populations; this fragmented distribution reflects the failure of the Oregon translocation (Bodkin 2015) and slow rate of recovery and natural range spread of the

California population.

Despite the conservation success represented by post-fur trade sea otter population recovery, it remains unclear whether sea otters have been fully restored to their historical ecological niches, which we here define as the combination of habitat (e.g. kelp forest vs. soft sediment), and prey species utilized. We propose that industrial-scale exploitation of the marine environment by humans (McCauley et al. 2015) may have led to a constriction of the ecological niche of modern sea otters relative to the past. This should be particularly true for the southern subspecies (Enhydra lutris nereis, Figure 2.1) which is currently found only in central California, an area with high human coastal densities and large-scale fisheries. Recovery in this region has been sluggish, averaging only ~2% annual population growth (Tinker and Hatfield 2018). Reasons suggested include: (1) the linear and narrow coastal shelf of California, that limits access to unoccupied habitats (Lafferty and Tinker

2014), (2) conflicts between otters and macroinvertebrate fisheries (Carswell et al. 2015), and (3) novel threats such as infectious disease and mortality caused by white sharks

(Bodkin 2015, Tinker et al. 2016, Tinker and Hatfield 2018). Defining a pre-industrial ecological baseline for southern sea otters will thus benefit conservation efforts by identifying critical resources or habitats to protect, as well as potential functional roles and species interactions.

It is possible to develop a historical ecological baseline for sea otters because their bones are common in coastal archaeological sites (Misarti et al. 2009, Szpak et al. 2012,

Jones et al. 2011) and isotope-based proxies allow for direct comparison of ancient and modern dietary niche and, indirectly, habitat. In sea otters, the isotopic niche is an established proxy for dietary niche, as otters consume a wide variety of macroinvertebrate prey fueled by two isotopically distinct sources of primary production: phytoplankton and macroalgae (Page et al. 2008, Newsome et al. 2009). Consequently, bulk analysis of vibrissae or other tissues that record dietary inputs over long time scales provides an accurate and high-resolution proxy for actual dietary niche breadth and variation, at both individual and population levels (Newsome et al. 2009, 2015, Elliott Smith et al. 2015).

Further, cutting-edge isotopic techniques for analyzing individual amino acids can identify whether spatiotemporal shifts in bulk tissue isotope values are due to trophic level or baseline ecosystem changes (Chikaraishi et al. 2014, Whiteman et al. 2019). We can thus characterize the ecological niche of southern sea otter populations before and after their near extirpation by using bulk and amino acid isotope analysis of ancient and modern sea otter tissues.

Here, we use zooarchaeological collections and modern tissue samples collected from southern sea otters to: (1) establish an ecological baseline for the species in California and (2) evaluate changes in their dietary niche over the past 7,000 years. We test whether the isotopic niche, measured from bulk tissue d13C and d15N values, occupied by sea otters has remained constant over time at five regions along the central and southern California coastline. We evaluate the extent that modern sea otters occupy the ancient isotopic space as a proxy for the recovery of their historical ecological niche. We also compare these data to the modern isotopic prey space. Finally, we use individual amino acid d15N analysis to examine whether prey choice, environmental conditions (or both) have changed. Our results provide a framework for interpreting the contemporary ecological role of sea otters and for identifying potential areas of their historical niche that are underutilized and could be promoted with conservation efforts.

Material and Methods

Modern Samples

To characterize the modern isotopic niche, we used previously published isotopic data from vibrissae of 158 individuals in five subpopulations in California (Elliott Smith et al. 2015): Monterey Bay, Big Sur Reserve, San Louis Obispo, Santa Barbara Channel and San Nicolas Island (Table 2.1, Figure 2.2). We sampled vibrissae from an additional seven San Louis Obispo individuals captured after 2015 (Appendix 2, Table 2). Our data encompass nearly the entire contemporary range (Tinker and Hatfield 2018) and represent a mix of males and females (Elliott Smith et al. 2015). All sampled individuals were independent foragers (weaned immature otters to aged adults), thus excluding dependent

30 pups. To quantify the potential isotopic niche available for modern sea otters, we used published isotopic data from Newsome et al. (2015) on the prey types most commonly consumed in Monterey Bay and Big Sur (20 species) and San Nicolas Island (11 species) to define a possible ‘prey space’.

ANO

CA-SMA-238

MBY Indicates CA-MNT-234 CA-MNT-831 modern

BSR range

%$S;LO" #(

CA-SLO-2

100 km SBC

CA-SMI-1 CA-SMI-525 SMI CA-SMI-528 CA-SMI-602 CA-SNI-011 CA-SNI-025 CA-SNI-040

( %@#* SNI

Figure 2.2. Locations of ancient and modern sea otter populations. Regions with archaeological southern sea otter remains represented by grey circles. Modern sea otter populations sampled for this study represented by black triangles (Elliott Smith et al. 2015). Modern range redrawn from Tinker and Hatfield (2018). Acronyms for regions as follows: ANO = Año Nuevo, MBY = Monterey Bay, BSR = Big Sur Reserve, SLO = San Louis Obispo, SBC = Santa Barbara Channel, SMI = San Miguel Island, SNI = San Nicolas Island. For a full list of sites and sample sizes see Table 2.1.

Table 2.1. Ancient and modern sea otter samples. For archaeological data, the Age column represents years before present (yBP), based on AMS radiocarbon dates from the listed references. For modern samples this represents the years of sea otter captures/collections (AD). With the exception of CA-SLO-2, all archaeological samples presented were analyzed in the current study. Acronyms for regions: ANO = Año Nuevo, MBY = Monterey Bay, BSR = Big Sur Reserve, SLO = San Louis Obispo, SBC = Santa Barbara Channel, SMI = San Miguel Island, SNI = San Nicolas Island.

Age Site Region N Age (cal yBP) Reference

CA-SMA-238 ANO 12 650–350 Hylkema 1991, 2019; Gifford-Gonzalez 2011 CA-MNT-234 MBY 16 2,100–1600 Gifford-Gonzalez 2007 CA-MNT-831 MBY 6 4,000–600 Breschini and Haversat 2011 CA-SLO-2 SLO 24 7,000–300 Jones et al. 2011 CA-SMI-1 SMI 8 7,000–3,400 Erlandson 1991 CA-SMI-525 SMI 12 3,100–500 Erlandson et al. 2005 CA-SMI-528: Strata 1 SMI 11 1,450–1,200 Walker et al. 2002 CA-SMI-602 SMI 3 500–300 Walker et al. 2002 CA-SNI-011 SNI 5 7,000–510 Rick et al. 2005

CA-SNI-025 SNI 8 740–510 Martz 2008 CA-SNI-040 SNI 8 4,200–3,800 Anis et al. 2014 Ancient Monterey Bay MBY 31 AD 2000-2012 Elliott Smith et al. 2015 Big Sur Reserve BSR 28 AD 2008-2012 Elliott Smith et al. 2015 San Louis Obispo SLO 56 AD 2010-2013 Elliott Smith et al. 2015 Santa Barbara SBC 37 AD 2010-2013 Channel Elliott Smith et al. 2015

San Nicolas Island SNI 13 AD 2004 Elliott Smith et al. 2015 Modern

Archaeological Samples

To characterize the historical isotopic niche, we sampled 107 bones from 10

California archaeological sites in close proximity to where modern sea otters were captured. These included both mainland and island archaeological sites and covered the entirety of the modern southern sea otter range – from Point Año Nuevo to San Nicolas

Island (Table 2.1, Figure 2.2). We also include published isotope data from CA-SLO-2

32 near San Luis Obispo (Jones et al. 2011). Sample sizes and estimated site ages based on

AMS radiocarbon dates are presented in Table 2.1, details on the excavation and identification of faunal remains can be found in the references therein. Care was taken not to resample individuals by considering specimen context (e.g., unit/level) and element.

Within units/levels or for single component sites (e.g., SMA-238), specimens were considered unique if they exhibited distinct (>1.0‰) d13C or d15N values (Clark et al.

2017). Where possible we sampled only adult or subadult individuals. From each specimen, we removed ~100-mg of bone for stable isotope analysis using a Dremel tool.

Comparative Tissue Dataset

Comparison of modern and ancient sea otter samples necessitates isotope data from two distinct proteinaceous tissues: bone collagen and vibrissae keratin. These proteins can exhibit systematic isotopic differences commonly referred to as tissue-specific isotope discrimination (Vanderklift and Ponsard 2003). To quantify and correct for this, we sampled bone collagen, muscle, liver and vibrissae from 29 sea otters stranded in central

California from 2007–2014 (Appendix 2, Table 1). Tissues from stranded sea otter carcasses were collected through the California Sea Otter Stranding Network, a multi- agency program coordinated by the California Department of Fish and Wildlife (CDFW) and USGS. Vibrissae were dry stored at 20ºC, while liver and muscle were stored at -20ºC; skulls were curated at the California Academy of Sciences (Appendix 2, Table 1). We sampled ~100mg from all tissues.

Sample Selection for Amino Acid d15N (AA d15N)

The isotopic analysis of individual amino acids (AA) within proteinaceous tissues is a cutting-edge technique in ecological studies. By breaking down whole protein complexes and applying fundamental biochemical principles, it is possible to disentangle shifts in animals’ trophic ecology from ecosystem level changes associated with changes in the baseline isotopic composition of food webs (Chikaraishi et al. 2014). The d15N of certain ‘source’ amino acids (e.g., phenylalanine and lysine) are not heavily modified by animals during assimilation and tissue synthesis due to their lack of participation in metabolic processes such as deamination. Consequently, source amino acids experience little isotopic alteration as they move through food webs, providing an indicator of the baseline d15N composition. Conversely, ‘trophic’ amino acids, such as glutamic acid and proline, are heavily involved in metabolic processes and exhibit consistent 15N enrichment with each trophic step. Thus, the magnitude of the nitrogen isotope difference between source and trophic amino acids can be used as a proxy for the trophic level of an individual, whereas source amino acids can be used to infer baseline ecosystem d15N composition

(Chikaraishi et al. 2014, Whiteman et al. 2019).

To examine whether spatiotemporal changes in bulk tissue isotope values of sea otters were due to baseline isotopic or dietary shifts, we selected a subsample of 4-5 individuals from each archaeological region (excluding CA-SLO-2) for AA d15N analysis.

In addition, we selected five modern Monterey Bay individuals from stranded otter bone collagen samples. We did not analyze modern sea otter vibrissae for AA d15N analysis because to our knowledge no study has addressed AA-specific tissue discrimination.

Isotopic Analysis

For both bulk and amino acid analysis we report all isotopic results as d values: d13C or d15N = 1000*[(Rsamp/Rstd) − 1], where Rsamp and Rstd are the 13C: 12C or

15N:14N ratios of the sample and standard, respectively (Appendix 2). Prior to isotopic analysis, bone collagen subsamples were cleaned of sediment and then demineralized with

0.25 N hydrochloric acid for 15-72 hours at 5°C. Each sample was then lipid extracted with three sequential 24 hr soaks in 2:1 chloroform:methanol and lyophilized after a deionized water rinse. Muscle and liver samples were also lipid extracted, rinsed, and lyophilized.

Sea otter vibrissae were cleaned with 2:1 chloroform:methanol to remove surface contaminants. For bone, muscle, and liver, 0.5–0.6 mg of each subsample was weighed into 3x5mm tin capsules. Vibrissae were subsampled following Newsome et al. (2009).

For AA d15N analysis, ~5-10mg of extracted collagen was chemically processed following established protocols (Whiteman et al. 2019, Appendix 2).

Data Corrections and Statistical Analysis

We corrected for both tissue-specific discrimination and temporal (Suess Effect;

Cullen et al. 2001) isotopic shifts prior to analysis (Appendix 2, Table 3). The resulting dataset (Appendix 2, Table 2) violated a number of important assumptions of ANOVA.

We thus tested for differences among modern and ancient sea otter d13C and d15N isotope values at each site using the nonparametric Cramér test (Baringhaus and Franz 2004). We also employed Kruskal-Wallis, and pairwise Wilcoxon Signed Rank comparisons with

Bonferroni adjusted p-values; we report pairwise comparisons in Appendix 2, Table 4. We likewise compared d13C and d15N isotopic values of ancient otters from different

35 archaeological sites. Our amino acid d15N dataset did not exhibit deviations from normality

(Appendix 2, Table 7); we used one-way ANOVA to evaluate for differences in source and trophic AA d15N, as well as trophic-source offset among sites. We calculated pairwise trophic-source offsets, as well as average offsets following methods from Bradley et al.

(2015), using glutamic acid (Glu) and hydroxyproline/proline (Hyp-Pro) as trophic AAs and phenylalanine (Phe) and lysine (Lys) as source AAs.

To characterize isotopic/dietary niche space and variability of ancient and modern otter populations we used Bayesian standard ellipse areas (SEAB) (Stable Isotope Bayesian

Ellipses in R; SIBER – Jackson et al. 2011). We ran the model with archaeological sites lumped within regions, and then with each site considered separately. In the latter case we excluded the four samples from SMI-602, and all samples from MNT-831 which had poorly constrained age ranges. We also calculated SEAB for modern California prey items from Monterey Bay/Big Sur and San Nicolas Island. We calculated median SEAB and associated credibility intervals (SD) from 10,000 iterations of the model, as well as the proportion of SEAB iterations from one group (otters or prey) larger than the SEAB of another group. Finally, we tested for the influence of time averaging on isotopic space

(SEAB) using a linear model of median SEAB versus time span occupied by each site (see

Table 2.1); modern samples were given a time span of 10 years. We performed all data corrections and statistical analyses using Program R v.3.2.0.

Results

Bulk d13C and d15N Values

We found ancient and modern otters exhibited different multivariate distributions in all regions (Cramer’s test: ANO/MBY/BSR p=0.00; SLO p=0.00; SMI/SBC p=0.00;

SNI p=0.04). Univariate Kruskal-Wallis confirmed that ancient and modern otters had distinct d15N values (H(9)=159.1, p<0.001) at almost every location (Appendix 2, Table

4). Individual pairwise comparisons indicated ancient sea otter d15N values from Año

Nuevo and Monterey Bay were significantly higher than associated modern populations

(Appendix 2, Table 4; Figure 2.3). In contrast, ancient San Miguel Island otters had lower d15N values than the modern Santa Barbara Channel population (Appendix 2, Table 4;

Figure 2.3). There were no significant differences in d13C values between ancient and modern otters within regions (Appendix 2, Table 4; Figure 2.3). Among the modern sea otter populations, bulk d15N varied with latitude: modern Monterey Bay and Big Sur sea otters had lower average d15N values in comparison to all other modern localities

(Appendix 2, Table 4; Figure 2.3).

Figure 2.3. Boxplots of d15N (panel A) and d13C (panel B) values of modern and ancient sea otters. Grey boxes are ancient populations, black are modern. Bars with asterisks represent ancient-modern pairs that are statistically different at an adjusted p <0.05 (see Appendix 2, Table 4). Acronyms for regions: ANO = Año Nuevo, MBY = Monterey Bay, BSR = Big Sur Reserve, SLO = San Louis Obispo, SBC = Santa Barbara Channel, SMI = San Miguel Island, SNI = San Nicolas Island.

Amino Acid d15N Values

We found regional and temporal differences among otter populations in both source and trophic amino acids (Tables 2.2, Appendix 2, Table 7). One-way ANOVA found significant differences in d15N values between regions for our two source amino acids

([Phe; F(4,18)=5.5, p=0.01]; [Lys; F(4,17)=3.0, p=0.03]), and two commonly measured trophic amino acids ([Glu; F(4,18)=5.1, p=0.01]; [Hyp-Pro; F(4,8)=6.6, p=0.00]).

Table 2.2. d15N values of individual amino acids from a subset of ancient and modern sea otters. Presented here are two commonly reported ‘trophic’ and ‘source’ amino acids, respectively glutamic acid (Glu), hydroxyproline-proline (Hyp-Pro), and phenylalanine (Phe), and lysine (Lys). Also reported is the mean (±SD) Glu-Phe offset for each locale and the average offset between these trophic and source amino acids. Acronyms for regions are as follows: ANO = Año Nuevo, MBY = Monterey Bay, SMI = San Miguel Island, SNI = San Nicolas Island.

Average Glu – Phe Age Locale Glu d15N Hyp-Pro d15N Phe d15N Lys d15N Trophic-Source Offset Offset ANO 24.0 ± 1.6 22.2 ± 1.6 12.3 ± 1.1 10.5 ± 0.7 11.6 ± 2.5 11.7 ± 1.8

MBY 21.9 ± 1.6 19.4 ± 1.8 10.7 ± 0.5 9.1 ± 1.0 11.2 ± 1.7 9.9 ± 2.0

Otters SMI 20.2 ± 0.9 18.4 ± 0.4 9.8 ± 0.9 9.2 ± 0.8 10.5 ± 1.3 9.2 ± 1.2 Ancient

SNI 21.0 ± 0.2 20.0 ± 0.8 12.0 ± 0.9 11.2 ± 0.8 9.0 ± 1.1 8.3 ± 1.8

MBY 22.9 ± 1.7 20.8 ± 0.4 10.3 ± 1.4 9.6 ± 1.2 12.6 ± 1.4 11.7 ± 1.6 Otters Modern Modern

However, we found no statistical difference in the average offset between trophic and source amino acids between ancient and modern otters [F(1,20)=0.29, p=0.60].

Isotopic Niche Breadth

Ancient sea otter populations occupied larger isotopic niches than their modern counterparts at almost every region (Appendix 2, Tables 5 and 6). When archaeological sites within regions were combined, ancient San Louis Obispo, San Miguel Island, and San

Nicolas Island populations exhibited larger standard ellipse areas (SEAB) than all other populations (Figure 2.4). When ancient data were grouped by archaeological site

(excluding SMI-602 and MNT-831) results were very similar (Appendix 2, Table 6). We

Figure 2.4. Isotopic niche space of ancient and modern sea otter populations. Left-side panels show d13C and d15N values of ancient (grey circles) and modern (black diamonds) sea otters. Ellipses represents 50% of the total amount of isotopic space occupied by each population. Right-side shows median Bayesian ellipse size ± credibility intervals for each population. All sea otter data were also compared to modern prey data from Newsome et al. (2015). From top to bottom the ancient/modern comparisons are as follows: ancient Año Nuevo otters (ANO) with modern Monterey Bay otter and prey (MBY), ancient MBY with modern MBY and prey, ancient San Louis Obispo (SLO) with modern SLO otter and MBY prey, ancient San Miguel Island (SMI) with modern Santa Barbara Channel (SBC) and San Nicolas Island prey, and ancient San Nicolas Island (SNI) with modern SNI and prey

40 also found differences in SEAB of ancient and modern sea otter populations relative to the potential prey space produced by analysis of modern prey items. All modern sea otter populations occupied <35% of the median potential prey space, in stark contrast to archaeological sites that ranged from 31% (SMA-238) to 99% (SNI-011) of the modern potential prey space (Appendix 2, Tables 5 and 6).

To evaluate the effect of time averaging, we compared the median SEAB of all ancient and modern sites and modern prey (excluding SMI-602 and MNT-831) to the estimated time span represented by each site (Table 2.1). A resulting linear model found no effect of time span on ellipse area size (R2 = 0.06, p = 0.19; Appendix 2; Figure 2), indicating differences in ellipse area between ancient and modern populations are driven by biological patterns and not statistical artifacts.

Discussion

Our results provide evidence of a reduced dietary (isotopic) niche of southern sea otters in portions of their current range in California. We find that the majority of sampled ancient otter populations in California, ranging in age from 7,000 to 350 years before present, had a wider dietary niche than their modern counterparts. In particular, ancient otters from San Miguel and San Nicolas Islands had the largest dietary niche space of all populations measured. This suggests that the current high diversity of invertebrate prey communities south of Point Conception relative to central California (Graham et al. 2008) has played a role in the past. We also find differences among ancient and modern sea otter populations in the relationship between average d15N values and latitude, which is likely driven by a combination of oceanographic (baseline) and diet composition (trophic level).

Our work provides important context for understanding modern sea otter dietary patterns and allows for predictions of how their ecology will change in the future.

We found striking differences in the isotopic niche space of ancient and modern sea otters in California, particularly in the southern portion of the range. Ancient otters from the five regions occupied a larger isotopic niche than nearly all modern localities

(Appendix 2, Table 5; Figure 2.4). However, at the three most southerly sites, ancient otters exhibited SEAB more than twice as large as modern counterparts (Appendix 2, Table 5;

Figure 2.4). The area near the Santa Barbara Channel is a major zoogeographic transition zone, with nearshore and island ecosystems south of Point Conception having a greater complexity of substrate, milder wave environments, and more variable upwelling regimes in comparison to the rest of California (Graham et al. 2008). As a result, these areas support a greater diversity of invertebrate and vertebrate taxa, and higher rates of endemism than sites in central and northern California (Seapy and Littler 1980, Graham et al. 2008). This plethora of secondary prey species may have contributed to the greater dietary diversity of ancient sea otter populations in this region. In particular, we note the striking variability in d13C values among ancient otters from San Miguel and San Nicolas Islands (-14‰ to -9‰) in comparison with modern counterparts (-12‰ to -10‰; Figure 2.4); ancient populations from these regions may have foraged in both rocky, kelp-dominated habitats, and soft- sediment, phytoplankton-fueled habitats, characterized by high and low d13C values, respectively (Page et al. 2008). The use of soft sediment habitats by otters is now recognized as an important aspect of their ecology (e.g., Hughes et al. 2016), and our results are consistent with a greater historical reliance on these systems in ancient southern

California.

Despite this larger potential ‘prey space’ or isotopic niche, modern sea otters in southern California exhibit low degrees of dietary (isotopic) variation, a pattern we suggest is a function of the time since recolonization and low population densities. Southern sea otters were relocated to San Nicolas Island from 1987–1990 and remained at low population densities for decades; the 12 individuals analyzed here represented nearly a third of the population around the island at the time of collection (Tinker et al. 2008, Elliott

Smith et al. 2015). Similarly, otter captures in the Santa Barbara Channel occurred shortly after a reproductive population had first became naturally established (Elliott Smith et al.

2015, Tinker and Hatfield 2018). Decades of intense observational studies document that recently established sea otter populations at low density target the largest, most energy rich prey, leading to low levels of dietary diversity and a small population-level isotopic niche space (e.g., Estes et al. 2003, Tinker et al. 2008, 2012). As densities increase, high-ranked resources become depleted, and individuals broaden their diet resulting in an increase in population-level dietary/isotopic niche breadth (Tinker et al. 2008, Newsome et al. 2009).

This latter pattern can be seen in modern sea otters in the northerly sites of Monterey Bay and Big Sur, where population densities are high, and otters are likely near carrying capacity (Tinker et al. 2019). In these areas, the size of the ancient sea otter isotopic niche is similar to their modern counterparts, indicating populations in central California have largely recovered their historical dietary breadth (Appendix 2, Table 5; Figure 2.4). In contrast, the isotopic data from archaeological remains of sea otters in southern California suggests that populations in this region could potentially occupy a much larger dietary niche than is currently observed.

Among modern otters, there was a ~4‰ increase in mean d15N with decreasing latitude (Figure 2.3). Some of this can be explained by a ~1–2‰ increase in baseline d15N values along this section of coastline, assumed to be driven by northward flow and upwelling of 15N-enriched intermediate waters of the California Undercurrent (Vokhshoori and McCarthy 2014). However, the majority of the increase in modern d15N values likely results from increased consumption of upper trophic level invertebrates by sea otters in the

Santa Barbara Channel and San Nicolas Island. Behavioral observations suggest otters in these localities have a preference for urchins, large carnivorous crabs, and octopus (USGS, unpublished data; Tinker et al. 2008), which would explain the relatively high d15N values of sea otters at southern sites (Figure 2.3; Newsome et al. 2009). In contrast, high densities of otters in Monterey Bay and Big Sur means they include a greater amount of smaller, lower trophic level invertebrates in the diet (Tinker et al. 2008) and thus have lower d15N values (Newsome et al. 2009).

Among ancient otter populations, we found no overall trend with decreasing latitude, however we did find differences in mean d15N between ancient and modern otters at some sites. Ancient otters from Año Nuevo and Monterey Bay had significantly higher d15N values than modern counterparts (Appendix 2, Table 4; Figure 2.3). Given the time averaging inherent in ancient data—some localities representing >2,000 years—we suspect the density-dependent processes driving the consumption of predominantly high trophic level prey is not likely to be the causal factor. Instead, we suspect a change in the underlying baseline isotope values of the northern regions (Ruiz-Cooley et al. 2014). We tested for this by comparing amino acid d15N data of a subsample of ancient otters from

Año Nuevo, Monterey Bay, San Miguel Island, and San Nicolas Island, to modern otters from Monterey Bay (Table 2.2, Appendix 2, Table 7).

Amino acid d15N analysis indicates that differences in modern and ancient sea otter bulk values (Figures 2.3 and 2.4) are due to a shifting isotopic baseline in the California

Current System over the Holocene. Notably, we did not find a significant difference in the average offset between our trophic and source amino acids between locales or time periods

(Table 2.2), which indicates that the trophic level of sea otters has not significantly changed over the past 7,000 years (Chikaraishi et al. 2014, Whiteman et al. 2019). Instead, we found differences in the baseline d15N values among regions with northerly ancient Año Nuevo

(12.3 ± 1.1‰) otters exhibiting the highest mean (±SD) Phe (source) d15N values, and southerly ancient San Miguel Island otters the lowest (9.8 ± 0.9‰; Table 2.2), a pattern opposite to that observed today (Vohkshoori and McCarthy 2014). In addition, we noted a wide range in the mean (±SD) differences between otter trophic and source d15N values between sites, with ancient San Nicolas Island otters having the lowest (8.3 ± 1.8‰), and modern Monterey Bay (11.7 ± 1.6‰) and ancient Año Nuevo (11.7 ± 1.8‰) the highest, offsets (Table 2.2). This variation likely reflects the trophic flexibility of otters as documented in high-resolution observational datasets (Estes et al. 2003, Tinker et al. 2008).

The analysis of ancient faunal remains from archaeological sites poses a number of quantitative and theoretical issues when comparing them to modern ecological datasets.

Most notably, archaeological sites in even the best scenarios represent time averaged assemblages on the order of hundreds to thousands of years, and so can never realize the high resolution of modern ecological sampling conducted over seasonal to decadal timescales (Rick and Lockwood 2013). Here, our ancient sea otter samples are likely an

45 aggregation of multiple generations. In addition, we cannot rule out the modification of local environments by humans. However, our use of zooarchaeological collections is biologically relevant for two reasons. First, we found no relationship between the isotopic niche space (SEAB) of ancient otters versus the amount of time represented by each archaeological or modern population (Appendix 2, Figure 2), indicating that results are driven by real biological patterns and not a statistical artifact. Second, when characterizing the historical ecological niche, incorporation of data across hundreds or even thousands of years provides a more representative view of a species ecology than a seasonal or multi- annual snapshot. Such an approach, which examines ecological patterns over evolutionarily relevant timescales, is a proven way of establishing conservation baselines and characterizing the plasticity of animals to long-term natural or anthropogenic environmental change (e.g., Jackson et al. 2001, Rick and Lockwood 2013).

Our results have implications for the conservation of southern sea otters, and for efforts to minimize negative interactions between otters and human populations. Our findings suggest sea otters have a much greater potential niche space in southern California than currently occupied, and that this niche may be larger than that utilized by modern, high density, sea otter populations in central California (Tinker et al. 2008, 2019). Southern sea otters are currently expanding their range into southern California where they have not occurred for centuries; our data provide clues as to how their diets will broaden as they continue to establish south of Point Conception. However, the ancient dietary niche space of southern sea otters we have characterized here was prior to the development of commercial fisheries. Over the past two centuries, intensive commercial fisheries have exploited many of the macroinvertebrate species commonly consumed by sea otters,

46 including abalone, red urchins, sea cucumbers and crabs (Dayton et al. 1998, Braje et al.

2007). Consequently, the historical dietary niche occupied by sea otters may be unobtainable until conservation and management efforts restore higher densities of important invertebrate prey. Despite this, our dataset speaks to a long-term history of interactions between humans and sea otters. The coastal archaeological record (e.g., Misarti et al, 2009, Jones et al, 2011, Szpak et al. 2012) demonstrates that humans have been living with, and harvesting, sea otters and their prey items for at least 10,000 years. The isotopic data we present here show that despite this, ancient sea otters had an ecological niche equivalent to, or greater than modern populations, suggesting that they occurred at high density in the past despite being subjected to harvest pressure. Such insights may aid in developing species- or ecosystem-based management plans that promote long-term sustainable relationships between the competing needs of humans and top predators.

Chapter 3

Multichanneling in a Chilean kelp forest is facilitated by intraspecific specialization

Emma A. Elliott Smith1, Chris Harrod2,3, Felipe Docmac2, Seth D. Newsome1

Abstract

A major focus of community ecology is characterizing the factors that contribute to the stability of food webs. Theoretical studies show that the use of multiple types of primary production, or energy channels, can increase community stability. However, the mechanisms that promote and maintain ‘multichannel’ feeding remain poorly understood, especially for marine ecosystems where there are unique logistical and analytical constraints. In particular, one intriguing question is the degree to which multichannel feeding is accomplished vis populations of individual dietary generalists versus specialists.

Here, we use carbon isotope (d13C) analysis of individual amino acids to characterize the relative contribution of pelagic and benthic energy sources to a nearshore kelp forest consumer community in northern Chile. We analyzed bulk tissue and essential amino acid

(EAA) d13C values from nine primary producers and eleven nearshore consumer taxa spanning a range of trophic and functional groups. EAA are an essential nutrient for eukaryotic consumers, and thus a major conduit of carbon flow within a food web.

Consequently, the isotopic composition of these molecules are minimally modified during trophic transfer, enabling us to identify primary producer EAA d13C fingerprints in

48 consumer tissues. Using linear discriminant analysis (LDA) of six EAA d13C values we distinguished among pelagic (near and offshore phytoplankton), and benthic (red algae, green algae, and kelps) endmembers with >87% success. We used this EAA isotopic library to characterize the sources of production for six genera of marine invertebrates and five species of nearshore fishes using a bootstrap-resampling LDA approach. We found evidence for multichannel feeding within the community as a function of trophic level.

Lower level consumers tended to focus on either pelagic or benthic energy, whereas higher order consumers used similar proportions of the two; invertebrate species ranged from deriving 13% to 98% of their EAA from phytoplankton, whereas fish ranged from 21% to

42%. We also found evidence for widespread individual specialization in energy channel usage, with 36 of 56 individual consumers deriving >80% of their EAA from a single source. Our findings suggest that energy channel coupling is mediated by individual level specialization within species. Such data on how consumers use multichannel systems are important for designing conservation and management strategies that focus on preserving stability and resilience in these dynamic food webs.

Introduction

Understanding which physiological, behavioral, or evolutionary factors determine the stability of food webs has defined much of community ecology over the past century

(e.g., Winemiller and Polis 1996). It is a particularly pertinent in this era of anthropogenic global change and rapidly increasing human populations. Recent work suggests that the diversity of energy pathways utilized within and among trophic levels has strong impacts on food web stability and by extension their resilience to perturbations (Rooney et al. 2006,

Wolkovich et al. 2014). Notably, ecosystems that are characterized by asymmetric (i.e., fast and slow) energy channels are theoretically highly stable (Rooney et al. 2006, Rooney and McCann 2012). A diversity of basal resources provides a consistent base for mobile, top consumers who in turn dampen prey oscillations and prevent population overshoots from equilibrium (Rooney et al. 2006, Rooney and McCann 2012). Termed

‘multichanneling’ at the ecosystem scale, or ‘multichannel feeding’ for consumers, this phenomenon is now thought to be widespread across terrestrial, freshwater and marine habitats, and may in fact be responsible for traditional trophic cascades (Wollrab et al.

2012, Wolkovich et al. 2014, Ward et al. 2015). However, one aspect of multichannel feeding that remains understudied is the mechanism underlying this consumer generalism

(Perkins et al. 2018). Is energy channel coupling achieved through populations of individual generalists that switch from fast to slow energy channels as they become available? Or, within a generalist population that appears to use multiple sources of energy, do individual consumers specialize on a particular energy channel? As individual specialization can also increase food web stability (Kondoh 2003), there is reason to believe these processes may be linked.

Kelp forests are highly productive nearshore marine ecosystems, which provide an excellent framework for exploring the processes governing energy channel coupling and multichannel feeding. These systems are characterized by expansive swaths of macroalgae

(Family Laminariales) and occur in nutrient rich waters at temperate latitudes across the globe (Steneck et al. 2002). Consequently, the highly diverse consumer communities within kelp forests are potentially able to utilize a combination of fast (e.g., phytoplankton) and slow (kelps and other benthic macroalgae) energy channels. The degree to which

50 individual members of these communities rely on kelp-derived energy is a highly debated issue (e.g., Duggins et al. 1989, Docmac et al. 2017) and likely depends locally on the invertebrate herbivore community (Yorke et al. 2019), the concentration of secondary metabolites in kelp tissues (Steinberg et al. 1995), and the extent and consistency of adjacent phytoplankton production. For example, in kelp forests at high latitudes, the oceanographic conditions that allow for phytoplankton production are highly seasonal, and local consumers can only occasionally utilize this fast and ephemeral energy channel

(Westberry et al. 2016). Consequently, kelp derived production is an important energy source for consumer communities in these habitats (Duggins et al. 1989, Elliott Smith et al. 2018). In contrast, in low latitude areas that support high amounts of pelagic production nearly year-round, this energy channel provides a more consistent basal resource for consumers, and kelp derived energy may be less important (Theil et al. 2007, Vargas et al.

2007, Docmac et al. 2017). As kelp forests experience escalating anthropogenic impacts

(Catton et al. 2016), quantifying which energy channels are important to local consumers will be vital in characterizing their future resilience and setting management targets.

Traditional methods of studying energy flow have been challenging or entirely intractable in kelp forests. Because direct observations are time and cost intensive in marine ecosystems, researchers have relied on stable isotope analysis as a method for characterizing the transfer of nutrients across trophic linkages (e.g., Newsome et al. 2010,

Layman et al. 2012). In particular, stable carbon (d13C) and nitrogen (d15N) isotope values of bulk tissues can be used as biomarkers with d13C values typically varying among marine producer groups, and d15N values increasing systematically as a function of trophic level

(Fogel and Cifuentes 1993, Hobson et al. 1994). However, the inferences that can be drawn

51 from bulk tissue isotope analysis are limited by substantial spatiotemporal isotopic variation among both producers and consumers (Larsen et al. 2009). In particular, bulk tissue isotopic analysis often fails to adequately resolve pelagic and benthic endmembers in highly productive coastal areas characterized by high degrees of upwelling (Page et al.

2008, Foley and Koch 2010). In these systems, the dynamic nature of nutrient availability and associated variation in growth rates of both phytoplankton and macroalgae produces a large degree of variation in primary producer d13C values related to physiologically- mediated processes that influence discrimination between algae tissues and the inorganic sources of carbon they fix (Page et al. 2008, Foley and Koch 2010). Similarly, d15N values of coastal producers are strongly influenced by the source of inorganic nitrogen, which can change dramatically across seasons (Rau et al. 2003). This can ultimately yield overlapping isotopic values among benthic and pelagic producer taxa, which limits the usefulness of this analysis as a tracer of energy flow (Fredriksen 2003, Reddin et al. 2015). In addition to this baseline variation, trophic discrimination factors between consumers and food sources, for d13C and d15N, can vary widely depending on consumer tissue type, life history, diet, and physiological state, which further complicates estimates of which basal producers are important to consumers (Ben-David and Flaherty 2012).

Recent developments in the isotopic analysis of individual compounds provide an exciting new way of tracking energy flow and characterizing food web structure

(Whiteman et al. 2019). The analysis of individual essential amino acids (EAA) is particularly useful for community ecology, as these molecules are a major conduit of energy flow between trophic levels (Larsen et al. 2009). Because only autotrophs and microbes can synthesize EAA, consumers must acquire them via direct consumption

(Howland et al. 2003) or possibly from their gut microbiome if dietary protein content is insufficient (Newsome et al. 2011). As a result, the isotopic composition of primary producer EAA remains intact as they move up food chains, passing from prey to consumer with minimal alteration (Howland et al. 2003, McMahon et al. 2015). Furthermore, recent work reveals that co-occurring producer taxa within kelp forest ecosystems have distinct patterns of EAA d13C values, or ‘fingerprints’, depending on the metabolic pathways they use to synthesize these compounds de novo (Larsen et al. 2013, Elliott Smith et al. 2018).

These fingerprints do not appear to change across growth or nutrient gradients (Larsen et al. 2013, 2015), and can thus be identified in consumers and used to quantify the importance of distinct energy pathways to any compartment within the food web.

Here, we examine the inter and intra-specific processes governing the coupling of pelagic (phytoplankton) and benthic (kelps, macroalgae) energy channels in a highly productive Chilean kelp forest. We develop a library of essential amino acid d13C values for kelps, red algae, green algae, and phytoplankton, and use this to quantify the importance of pelagic/benthic production to local consumers. We present results for nearly 60 individuals from five fish species, and six invertebrate species spanning a range of trophic and functional groups that provide a good representation of coastal assemblages in the region. For each individual, we were able to characterize the degree to which they engaged in ‘multichannel feeding’, and scaled this up to species, and trophic-level patterns. Our results reveal a high degree of intra-specific variation in energy channel utilization by consumers, and we suggest that such individual-level specialization may play an important role in mediating food web structure and stability.

Material and Methods

Research Sites and Sample Collection

We collected samples from two localities near Antofagasta, off the coast of northern

Chile: Isla Santa María and Coloso (37.0226, 73.5182). Both sites are characterized by rocky substrates, and consistent and high degrees of upwelling (Docmac et al. 2017). All samples were collected between January 2016 to December 2017, and data from both sites were lumped by species.

At each site, we collected four groups of marine producers representing the dominant local taxa (Table 3.1; Fariña et al. 2008). These included kelps (Macrocysistis pyrifera, and Lessonia sp.), red algae (Plocamium cartilagineum, Rhodymenia coralline, crustose coralline algae), green algae (Ulva sp.), and phytoplankton/particulate organic matter (POM) samples from throughout the water column. In addition, we collected two samples of epilithic biofilm from the intertidal. Kelp samples were collected manually by

SCUBA divers at 5, 10 or 12m below the surface. Red algae were similarly collected from the benthos. Samples of Ulva and biofilm were taken from the intertidal. POM and phytoplankton were collected in the pelagic zone (5km offshore) and upper subtidal. To obtain pelagic phytoplankton samples, we took water samples via vertical tow nets with a mesh size of 200 µm to screen out zooplankton and detritus, and then passed ~1–2 liters of water through pre-combusted 0.7 μm GF/F filters. Filters were also checked visually with

Table 3.1. Humboldt Current producer and consumer samples from near Antofagasta, Chile. Presented are details on functional group, species identification, collection notes, and sample sizes for isotopic analysis. The subset of samples analyzed for essential amino acid (EAA) d13C analysis are presented in parentheses.

N Functional Group Species Bulk Collection Notes (EAA) Planktivore (nearshore) Isacia coneptionis 7 (7) Subtidal Collected pelagically by sport Planktivore (offshore) Engraulis ringens 4 (4) fisherman Herbivore Aplodactylus punctatus 8 (8) Subtidal

Fish Carnivore (small Cheilodactylus 8 (6) Subtidal inverts) variegatus Carnivore (large Palabrax humeralis 8 (6) Subtidal inverts) Planktivore Zooplankton 3 (1) Pelagic tows Intertidal/ Filter feeder Perumytilus purpatus 5 (5) upper subtidal

Herbivore Taleipus sp. 6 (6) Subtidal Intertidal/

Inverts Herbivore/Grazer Tegula atra 8 (7) upper subtidal Intertidal/ Herbivore/Grazer Tetrapygus niger 5 (4) upper subtidal Concholepas Intertidal/ Carnivore 2 (2) Concholepas upper subtidal Intertidal/ Epilithic film Biofilm 2 (2) upper subtidal Unidentified red Crustose coralline algae Red Algae 8 (7) Subtidal

Plocamium cartilagineum Rhodymenia coralline Intertidal/

Producers Green Algae Ulva sp. 7 (7) upper subtidal Lessonia sp. Kelp 24 (13) Subtidal at different depths Macrocsystis pyrifera Phytoplankton Pelagic tows w/ intertidal 'POM' Particulate Organic Unknown 16 (14) samples Matter a binocular microscope for any large particles of obvious non-phytoplankton origin.

Particulate organic matter samples from the upper subtidal were collected in 6 liter bottles and filtered through the same process outlined above. For all analyses, phytoplankton/POM samples were lumped into a single ‘pelagic’ category.

At each site we also sampled fish and macroinvertebrates from different phylogenies, trophic levels and feeding modes (Table 3.1). Among the inveterate

55 community we collected: (1) filter feeding mussels (Perumytilus purpatus), and offshore pelagic zooplankton, (2) three herbivore species, including the grazers, Tegula atra

(gastropod), and Tetrapygus niger (sea urchin), and an macroalgae herbivore – Taliepus sp. (kelp crab), and (3) opportunistic samples of a carnivorous gastropod, Concholepas concholepas. In addition, we collected individuals from five fish species belonging to different foraging guilds (Table 3.1): (1) jerguilla (Aplodactylus punctatus) an herbivore,

(2) cabrilla (Palabrax humeralis), an invertebrativore and piscivore, (3) bilagay

(Cheilodactylus variegatus), an invertebrativore carnivore, (4) cabinza (Isacia coneptionis), a nearshore planktivore, and (5) anchoveta (Engraulis ringens) an offshore planktivore. Henceforth we refer to all species by genus. All fish samples were collected via net collection from boats or spearfishing; anchoveta were purchased from local fish markets. Invertebrates, except Taliepus and zooplankton, were collected by hand from the upper subtidal/lower intertidal zone. Zooplankton were collected via vertical tow nets with a mesh size of 200 µm; Taliepus samples were collected from the subtidal as bycatch from gill nets. All samples (producer and consumer) were kept on ice for a period of 2-8 hours immediately post collection, and then either processed and lyophilized immediately or stored frozen at -20° C until analysis.

Sample Preparation and Isotopic Analysis.

In preparation for isotopic analysis, all samples were rinsed with de-ionized water and then lyophilized (producers), or lipid-extracted (consumers). Phytoplankton/POM filters were only lyophilized. Prior to lipid-extraction, invertebrate consumers were shucked, roughly homogenized and subsampled. For fish specimens, a piece of the dorsal

56 muscle was removed for analysis. For all consumer subsamples, lipids were removed via four ~24h soaks in petroleum ether, followed by a DI rinse and lyophilization.

Prior to EAA d13C analysis, we prepared and analyzed samples for bulk d13C and d15N per standard methods (Elliott Smith et al. 2018). For bulk analysis, dried samples were weighed to ~0.5mg (fish/invertebrates), or 3-4mg (macroalgae), and sealed in tin capsules. Pieces of filters containing phytoplankton/POM were weighed out to ~20mg.

After sampling for bulk isotopic analysis, we weighed a subset of samples for EAA d13C analysis (Table 3.1). For this analysis, we measured 2-5mg of consumer tissue, 4-6mg of macroalgae, and ½ to ¾ of phytoplankton/POM filters. For two phytoplankton filters:

HUM-ISM4-FITO and HUM-ISM-PEL3-FITO, initial bulk analysis failed, and thus we only have EAA d13C data for these samples (Appendix 3). In addition, for one intertidal particulate organic matter sample: HUM-POM-ISM1-B2, visual assessment indicated potential macroalgae detritus. Thus, we excluded this sample from LDA analyses. All isotopic measurements – both bulk and EAA d13C – followed established protocols (e.g.,

Newsome et al. 2010; Elliott Smith et al. 2018; Whiteman et al. 2019); see Appendix 3 for details on instrumentation, analysis, and quality control. For all statistical analyses using

EAA d13C data, we used the mean corrected d13C value across multiple injections of a sample for each amino acid.

Statistical Analyses

This goal of this study was to evaluate food web structure in a nearshore Chilean kelp forest. As discussed in the introduction, bulk d13C and d15N isotopic data are of limited use in highly productive coastal areas for a number of biotic and abiotic factors. We thus

57 used bulk isotopic data largely for comparative purposes and focused on EAA d13C values for all statistical analyses. We used MANOVA to compare among producer and consumer groups for all measured EAA d13C values. Individual pairwise contrasts between groups, for each EAA, were subsequently conducted using ANOVA and Tukey’s HSD.

Assumptions of normality and equal variances within and among groups were determined using the Cramer von Mises test with quantile-quantile plots, and Bartlett’s test, respectively. We ran all statistical analyses in program R (v 3.3.1) with RStudio interface

(v 0.99.903).

To characterize the proportional contribution of each producer group to our sampled consumers, we used multivariate techniques following Larsen et al. (2013), Elliott

Smith et al. (2018), and more recently Fox et al. (2019). Specifically, we used a bootstrapping resampling approach to linear discriminant analysis to characterize EAA d13C patterns, or fingerprints, of our producers and test for these in our consumers. Our analysis is thus dependent on the assumption that EAAs are an important and informative metric of energy flow within foodwebs. We first examined the classification error rate for the four producer groups (phytoplankton/POM, green algae, red algae, and kelps) using a leave-one-out, cross-validation procedure to establish if our sources were statistically distinct. We then used this training data set to predict group membership for each consumer sample. However, because LDA only allows for a consumer to be classified with a single producer group, we utilized a bootstrapping approach to quantify the proportional contribution of each producer group to our consumers. Following Fox et al. (2019), we ran

10,000 iterations of the LDA model, using random draws with replacement from our producer training dataset. For each iteration, LDA output included posterior probabilities

58 of individual consumer classification with each producer group, which we used to provide a distribution of possible classifications for that individual. From this distribution we calculated a mean posterior probability of classification with each producer for each individual, and species, with associated error. We took these data to represent the percent reliance on each producer group. We then calculated a posteriori a percent reliance on pelagic (phytoplankton/POM) vs. benthic (Ulva, kelps, red algae) endmembers. We were unable to include epilithic biofilm as a possible source for consumers due to sample size requirements for LDA (Biofilm n =2; Tabachnick and Fidell 2013). Instead, we treated them as a ‘consumer’, group and classified them with our other producer groups purely for comparison. We considered an individual to be an energy channel ‘specialist’ if they exhibited ≥80% posterior probability of classification with a single producer group across all randomized bootstrap iterations.

We ultimately chose not to use Bayesian mixing models to estimate the proportional contribution of pelagic vs. benthic endmembers to our Chilean consumers.

Previous work (e.g., Larsen et al. 2009, 2013, Elliott Smith et al. 2018) has successfully used a combination of Bayesian mixing models and multivariate statistics with EAA d13C data. However, mixing models are most successfully applied when (1) the number of samples for each source is high (i.e., >20), (2) a complete mixing space can be assumed,

(3) the isotope values of each source are statistically distinct from one another, and (4) the geometry of possible sources in isotope space is suitable to classifying consumers (Phillips et al. 2014). Our dataset violates a number of these ‘best practices’. For example, several

EAA did not significantly differ in raw d13C values among our producer groups (Appendix

3, Table 1). Consequently, we were unable to achieve model convergence using either all,

59 or a subset of EAA values in the Bayesian framework of MixSIAR (Stock and Semmens

2013). Instead, we focused on the LDA bootstrap resampling approach described above, which a recent study (Fox et al. 2019) has demonstrated shows remarkable utility and can be preferable to classical isotope mixing models in some scenarios.

Results

Bulk d13C and d15N Values

Local kelp forest species showed substantial variation in d13C and d15N values

(Figure 3.1). Our bulk tissue dataset failed to pass normality checks, thus precluding the use of parametric statistics. However, a number of interesting patterns are still evident

(Figure 3.1). Our sampled consumer groups varied widely in bulk tissue isotope values, particularly in d15N. Mean (±SD) d15N values for invertebrates ranged from 10.0±3.2‰

(zooplankton) to 18.2±1.1‰ (Concholepas), and between 15.4±1.2‰ (Engraulis) to

22.0±0.7‰ (Cheilodactylus) for fish (Figure 3.1). Among primary producers, we found a relatively small degree of variation in mean d15N values (10.8‰ to 12.0‰), with the exception of two offshore phytoplankton samples with values of 4.2‰. In contrast, for d13C, we found a tremendous range in producer bulk tissue values: The red algae

(Rhodymenia and Plocamium) had the lowest mean d13C value (31.9±1.6‰ and

33.0±1.1‰ respectively) whereas green algae (Ulva sp.) and biofilm samples had the highest mean (±SD) d13C values of -13.3±2.3‰ and -9.6±2.0‰, respectively. Local kelps

Genus 1 1 Cheilodactylus 2 Isacia 2 3 4 3 Palabrax 20 4 Aplodactylus 5 Engraulis 6 6 Concholepas 7 Taleipus 9 7 8 8 Perumytilus 5 N 10 15 9

15 Tetrapygus δ 10 Tegula 11 Zooplankton 17 16 15 12 Bioﬁlm 18 13 14 12 13 Ulva 19 14 Phytoplankton 10 11 15 Lessonia 16 Macrocystis 17 Red Algae 18 Rhodymenia 19 Plocamium −35 −30 −25 −20 −15 −10 δ13C Figure 3.1. Variation in bulk d13C and d15N values for producers and consumers sampled near Antofagasta, Chile. Circles represent dominant primary producer taxa at the site, inverted triangles represent invertebrate consumers, triangles represent fish from different trophic levels/functional groups. The group ‘Red Algae’ represents samples of crustose coralline algae and a single unidentified red algal sample.

(Lessonia, Macrocystis) and phytoplankton/POM had intermediate and overlapping mean

(±SD) d13C (phytoplankton/POM = -17.8±1.2‰; kelps = -19.3±3.8‰) values (Figure 3.1).

Amino Acid d13C Values

In contrast to bulk tissue isotope data, we found significant differences in EAA d13C values among producers ([F(3, 36) = 11.99, p < 0.01]) and consumers ([F(10, 45) = 2.58, p < 0.01]; Appendix 3, Table 1). Although all producers were statistically distinct from one another for at least one EAA, we noted differences in which EAA d13C values differed

61 among our four producer groups. For example, our sampled kelps and phytoplankton only exhibited differences in raw d13C values for isoleucine, whereas red algae showed strikingly different d13C values for all six essential amino acids (Appendix 3, Table 1). For consumers, a number of species stood apart in their EAA d13C values. For example, among our sampled fish species the herbivorous Aplodactylus, offshore planktivore Engraulis, and invertivore Palabrax exhibited significantly different d13C values for multiple EAAs in comparison to other species (Appendix 3, Table 1). For the invertebrate community, Tegula showed the greatest number of significant differences with other species (Appendix 3,

Table 1). We note however, that although our consumer EAA d13C dataset passed normality test, it did exhibit mild kurtosis.

Linear Discriminant Analysis

Results from our multivariate analysis of EAA d13C values showed strong separation among producer groups (Appendix 3, Tables 2 and 3; Figure 3.2). The first two linear discriminant axes explained a cumulative 94.9% of the variation in our producer dataset, driven predominantly by differences in isoleucine, lysine, and valine d13C values

(Appendix 3, Tables 2 and 3; Figure 3.2). Importantly for this study, cross-validation of the LDA found an exceptionally high successful reclassification rate (92.3%) among our four primary producer groups according to their EAA d13C values (Appendix 3, Tables 2 and 3; Figure 3.2). Red algae had the highest successful reclassification rate (100%) with

7/7 samples reclassifying correctly, and green algae had the overall lowest (85.7%) with

1/7 samples incorrectly classifying with phytoplankton/POM (Appendix 3, Table 3).

5.0 (a) Fish

2.5

0.0 Phytoplankton/POM

Ulva

Kelps −2.5 Red Algae Cheilodactylus

Isacia

Palabrax −5.0 Aplodactylus Engraulis 5.0 (b) Invertebrates Concholepas Taleipus

Perumytilus

Tetrapygus

2.5 Tegula

Zooplankton Linear Discriminant 2

Bioﬁlm

0.0

−2.5

−5.0

−5 0 5 10 Linear Discriminant 1 Figure 3.2. Fingerprinting of Humboldt Current producers and consumers. Results from a single iteration of linear discriminant analysis using essential amino acid d13C data of producers and consumers. Circles represent the linear discriminant loading of individual producer samples; ellipses surrounding producer groups represent 90% confidence intervals. Triangles (panel a), and inverted triangles (panel b) represent linear discriminant loadings for fish and invertebrate individuals respectively. Where a consumer plots in this space is a reflection of the predominant source of production from which their essential amino acids are sourced.

Phytoplankton/POM and kelps had similar reclassification rates with respectively 1/13

(92.3%) and 1/12 (91.7%) individuals incorrectly reclassifying with other producer groups

(.Appendix 3, Table 3).

The bootstrapped LDA analysis showed strong individual, species, and functional- level partitioning in the energy channels utilized by consumers. Overall, kelps and phytoplankton were the dominant producer groups supporting local consumers (Table 3.2,

Figures 3.2 and 3.3). Among species, however, consumers varied widely in which energy/carbon pathway they were most reliant on. Invertebrates showed a greater degree of species-level partitioning, ranging from an average of 13% (Taleipus) to 98%

(Zooplankton) pelagically (e.g. phytoplankton/POM) derived carbon (Table 3.2, Figures

8 (a) Fish 4 2 6 3 1

4 5 Genus 1 Cheilodactylus 2 Isacia 2 3 Palabrax 4 Aplodactylus 0 5 Engraulis (b)Invertebrates 8 6 Concholepas 10 7 Taleipus 8 Perumytilus 6 7

Number of Individuals 9 Tetrapygus 8 10 Tegula 4 9 11 Zooplankton

2 6 11 0 0 0.25 0.50 0.75 1.00 % Pelagically-Derived Energy Figure 3.3. Individual-level energy channel usage by fish (panel a) and invertebrate (panel b) consumers. Pelagically-derived energy is here defined as essential amino acids within consumer tissues that were ultimately sourced from phytoplankton; a value of ‘0’ on the x- axis thus means the individual(s) derived their energy from benthic macroalgae. For each individual, these percentages are calculated from the average posterior probabilities of 10,000 bootstrapped iterations of linear discriminant analysis using essential amino acid d13C data. Bars represent 10% intervals, circles represent the species average reliance on pelagic energy; error bars around these averages can be found in Table 3.2 but are not presented because of the bimodality of energy channels utilized.

Table 3.2. Percentage of EAA from each producer group routed to consumers. Results represent the species-level average posterior probability of 10,000 bootstrap iterations of linear discriminant analysis using EAA d13C values. POM = particulate organic matter. The % of energy channel specialists was calculated using the number of individuals within each species that had ≥80% posterior probability of classification with a single producer group across all randomized bootstrap iterations (see Appendix 3, Table 4).

% energy Phyto/ Species N Kelp Green Red channel POM ‘specialists’ Cheilodactylus 6 27 ± 37 71 ± 41 3 ± 4 0 ± 0 83 variegatus

Isacia coneptionis 7 47 ± 36 53 ± 36 0.2 ± 0.5 0 ± 0 57

Fish Palabrax humeralis 6 21 ± 21 79 ± 21 0 ± 0 0 ± 0 50 Aplodactylus punctatus 8 24 ± 39 33 ± 42 21 ± 40 21 ± 37 63 Engraulis ringens 4 42 ± 45 23 ± 46 34 ± 45 0 ± 0 75 Concholepas 2 89 ± 15 11 ± 15 0 ± 0 0 ± 0 50 concholepas Taleipus sp. 6 13 ± 26 86 ± 26 0.1 ± 0.3 0.2 ± 0.6 83 Perumytilus purpatus 5 65 ± 31 21 ± 34 14 ± 23 0 ± 0 40 Tetrapygus niger 4 40 ± 37 53 ± 46 7 ± 10 0 ± 0 50 Invertebrates Tegula atra 7 32 ± 38 52 ± 48 16 ± 32 0 ± 0 71 Zooplankton 1 98 2 0 0 100

3.2 and 3.3). At the species level, fish utilized a mixed proportion of pelagic and benthic sources ranging from 24% (Aplodactylus) to 47% (Isacia) phytoplankton derived EAA

(Table 3.2, Figures 3.2 and 3.3). Within each consumer species, the large variation in energy channel used was predominantly driven by variation among, rather than within individuals (Appendix 3, Table 4; Figure 3.3). Of the 56 individual consumers analyzed,

36 had ≥80% posterior probability of classification with a single producer group across all randomized bootstrap iterations (Appendix 3, Table 4; Figure 3.3).

Discussion

Characterizing how biological communities are organized, and what factors contribute to the stability of these systems is a long standing and important area of

65 ecological research (Winemiller and Polis 1996). This has been a challenging task for marine ecosystems, where the very substrate food webs exist in precludes the use of traditional observational methods. Here, we quantified energy channel usage and ecological specialization by eleven species of marine consumers in a highly productive nearshore kelp forest using a cutting-edge analytical technique. Within our sampled community we found evidence for the coupling of benthic (slow) and pelagic (fast) energy channels by consumers, or ‘multichanneling’, that appears to be an important component of food web structure in this system. However, in contrast to theoretical considerations

(e.g., Post et al. 2000), we found that multichannel feeding occurred at both the species and individual level. The majority of our sampled individuals tended to utilize either the pelagic or benthic channel rather than a mixture of the two. Our results provide a mechanism for multichannel feeding in nearshore marine systems based on ecological variation among individuals that likely has important implications for preserving resilience to natural and anthropogenic perturbations.

Multichannel feeding, defined here as utilization of both pelagic (phytoplankton) and benthic (macroalgae) energy channels, was widespread in our sampled fish and invertebrate communities. This phenomenon was clearly evident within the fish taxa, with species averaging 21% to 47% pelagically-derived EAA. (Table 3.2, Figure 3.3).

Invertebrates showed less multichannel feeding overall, with two of six species deriving

>80% of their EAA from either pelagic or benthic energy channels (Table 3.2, Figure 3.3).

These patterns compliment previous empirical and theoretical research demonstrating that the coupling of slow benthic and fast pelagic energy channels is an important component

66 of marine food web dynamics, and further, that this coupling is typically achieved by mobile, top consumers like fish (Post et al. 2000, Wolkovitch et al. 2014, Ward et al. 2015).

The multichanneling we observed at the species level was facilitated by intraspecific differences in energy channel utilization. The majority of individuals within a species tended to use either the pelagic or benthic energy channel rather than a mixture of both channels (Appendix 3, Table 4; Figures 3.2 and 3.3). We quantified energy channel specialization as an individual with ≥80% posterior probability of classification with a single producer group across 10,000 LDA iterations. If a consumer was midway between two sources in this multivariate space, random iterations should result in ~50% probability of classification with each group. This would be the anticipated outcome for a population of energy channel generalists. In contrast to this expectation, we found extremely high posterior probabilities of classification with a single producer group for the majority of individuals (36/56, Appendix 3, Table 4; Figure 3.3). This result provides strong evidence that many of our consumers were ‘specializing’ on a particular energy channel, as they must have been close to a single producer centroid to produce these results (Figure 3.2).

This pattern held across nearly all species examined (Tables 3.2, Figure 3.3). Even within fish taxa that averaged nearly 50% pelagically-derived EAA (e.g., Isacia), 4/7 individuals were considered to be specialists on either pelagic or benthic sources (Tables 3.2, Figure

3.3). The only exception to this was the filter feeding species Perumytilus, where only 2/5 individuals had greater than 80% probability of classifying with a single source group

(Tables 3.2, Figure 3.3). This pattern of specialization within populations along the pelagic–benthic niche axis has been noted for a number of fish species (Robinson and

Wilson 1994), with many of the classic examples being freshwater taxa found in postglacial

67 lakes such as three-spine stickleback (Gasterosteus aculeatus; Matthews et al. 2010,

Ravinet et al. 2013), European white fish (Coregonus lavaretus; Østybe et al. 2006), and

Midas cichlid (Amphilophus tolteca; Kusche et al. 2014). In these species, the ecological variation is thought to serve as an agent of sympatric speciation and parallel evolution.

Traditionally, theoretical work on multichannel omnivory has implicitly or explicitly assumed that energy channels are linked by generalist populations of top predators that switch between prey sources as a function of availability or general preference (Post et al. 2000). Our results demonstrate that in some systems multichannel feeding may in fact be a result of intraspecific dietary variation and individual specialization. However, it is important to distinguish that what we measured here was individual specialization in energy channel usage rather than consumption of a particular prey type. The latter is the traditional metric of individual dietary specialization, which is now recognized as a widespread and ecologically relevant phenomenon, particularly in aquatic communities (e.g., Bolnick et al. 2003, Araújo et al. 2011). We were unable to empirically determine individual-level prey specialization as we did not collect multiple tissues from individuals that would enable us to generate a longitudinal record of dietary variation, nor did we exhaustively collect potential prey items. Despite this limitation, we suggest that specialization of particular functional groups of prey may well be linked to specialization in energy channel. The partitioning among individuals of dietary items can be highly beneficial for consumer populations as it may result in decreased search and handling times and rates of intraspecific competition (Bolnick et al. 2003, Araújo et al.

2011). In turn, favored prey items may be associated with particular habitats or particular functional groups (e.g. filter feeders or grazers). In sea otters, for example, individual

68 specialization on infaunal bivalves requires a particular set of foraging skills (e.g., digging and shucking) that are uniquely suited to soft-sediment habitats that are primarily fueled by phytoplankton (Tinker et al. 2008, Newsome et al. 2015). In contrast, sea otters that specialize on urchins, which in turn are predominantly dependent on macroalgae energy, utilize a very different set of foraging behaviors (Tinker et al. 2008, Newsome et al. 2015).

Thus, individual dietary specialization on a suite of ecologically similar species that belong to particular functional groups may inadvertently lead to specialization on a particular energy channel.

The connection we found between individual-level energy channel specialization and multichannel feeding is one that is typically overlooked in the literature. The study of energy channel coupling by mobile consumers is extensive (e.g., Polis and Strong 1996,

Rooney et al. 2006, Wollrab et al. 2012, Wolkovitch et al. 2014, Ward et al. 2015), however, most of this work has focused on how energy coupling affects food web stability

(Rooney et al. 2006, Blanchard et al. 2011, Rooney and McCann 2012), trophic cascades

(Ward et al. 2015), or how widespread multichanneling is across diverse ecosystems

(Wolkovitch et al. 2014). Little work has been done on the mechanisms by which energy channel coupling is achieved within a consumer species or functional guild (but see Wimp et al. 2013, Perkins et al. 2018). Similarly, there is a large body of literature on individual diet specialization (see reviews by Bolnick et al. 2003 and Araújo et al. 2011), specifically how phenotypic variation (Mathews et al. 2010), behavior (Werner and Sherry 1987), population/community dynamics (Tinker et al. 2008), and environmental conditions mediate intraspecific dietary differences, and how this variation in turn impacts food web stability (Kondoh 2003). Our results are similar to that reported in prior work that found

69 specialization in energy channel usage within sympatric populations (e.g., Matthews et al.

2010, Kusche et al. 2014), however, these studies focused predominantly on evolutionary mechanisms that facilitate adaptive radiations. To our knowledge, we are among the first to empirically show a link between individual specialization and energy channel coupling across a diverse community containing multiple functional groups and species within a single ecosystem.

Our work represents an important contribution to community ecology in providing a mechanism by which individual-level variation in energy channel usage can link community dynamics. Theoretical and empirical work has demonstrated that the coupling of diverse energy channels by mobile top consumers, or ‘multichanneling’, is most likely a stabilizing force in food webs (Rooney et al. 2006). Similarly, intra-specific variation in dietary preferences has also been linked with increased community stability and resilience

(Kondoh 2003). Our results from a diverse and dynamic kelp forest in northern Chile suggests that these processes are linked, with population-level multichannel feeding in both fish and invertebrate species facilitated by intraspecific dietary differences. Potential avenues of future research may include (1) examining how widespread the link between multichannel feeding and individual specialization is across different ecosystems, (2) evaluating differences among functional/trophic groups in energy channel specialization, and (3) assessing the impact of consumer energy specialization on food web stability. An understanding of the latter will be particularly important for vulnerable marine ecosystems in this era of increasing anthropogenic change (Jackson et al. 2001, McCauley et al. 2015).

Chapter 4

Biochemical mechanisms and environmental influences on the amino acid isotopic composition of marine primary producers

Emma A. Elliott Smith1, Michael D. Fox2, Marilyn L. Fogel3, Seth D. Newsome1

1Department of Biology, University of New Mexico, Albuquerque, NM 2Woods Hole Oceanographic Institution, Woods Hole, MA 02543 3Earth and Planetary Sciences, University of California – Riverside, Riverside, CA, 92521 USA

Abstract

The isotopic analysis of individual compounds has emerged as a powerful new tool in the natural sciences. Notably, the measurement of stable isotopes of carbon (d13C) in amino acids can provide higher resolution information on the fundamental source(s) of primary production driving food webs. Often termed amino acid ‘fingerprinting’, this technique relies on isotopic distinctions among producer groups in multivariate space

13 provided by d C values of essential amino acids (AAESS). Although this technique has been widely applied, little work has been done on the biochemical and environmental

13 mechanisms driving producer AAESS d C fingerprints. Here, we present a dataset of amino acid d13C values of 130 marine primary producers from coastal regions along the eastern

Pacific to examine in detail what factors determine these isotopic fingerprint patterns.

Specifically, we analyzed co-occurring red, green, and brown algae as well as particulate organic matter from sites in Alaska, central and southern California, and northern Chile. In addition, we collected historical samples of the giant kelp, Macrocystis pyrifera, that date

13 back to 1896. Hence, we are able to examine if AAESS d C fingerprints vary across, space,

71 taxa, and time. We compiled information on the mode of inorganic carbon acquisition used by each taxa, as well as temperature and seasonality data for each locale, and employed generalized linear models to examine which factors were most important in affecting the amino acid d13C values, and whether these same factors were responsible for determining

13 AAESS d C statistical fingerprints. We find that mode of inorganic carbon acquisition and producer taxonomy (e.g. phylum) are the driving factors in determining amino acid d13C values, followed by mean local temperature and seasonality. The d13C fingerprint relationships were also influenced by the same factors, but not as strongly. Time period

(i.e., historical vs. modern) was not a major influence for the majority of amino acids. By comparing d13C values of amino acids that are closely related biochemically, we were able to identify particular synthesis pathways that may be the driving force behind different

13 primary producer isotope fingerprints. Our results demonstrate that AAESS d C fingerprints are consistent across space and time within taxa. The data and findings in this work provide a solid foundation for using an amino acid d13C fingerprinting technique in marine and potentially terrestrial ecosystem studies.

Introduction

In biology and ecology, the number of natural abundance stable isotope-based publications has risen exponentially in the past few decades (Pauli et al. 2015). Isotopes have been applied to a diverse set of questions from reconstructing paleoenvironments and communities (Yeakel et al. 2013), tracking animal migration (Hobson et al. 2010), defining food web structure (Fry 1988), to tracing energy flow within and among organisms, populations, and habitats (Newsome et al. 2007, 2014, Page et al. 2008, McMahon et al.

2013). Within this interdisciplinary field of isotope ecology, the study of individual compounds (e.g., fatty and amino acids) has emerged as a powerful tool (Fantle et al. 1999,

Fogel and Tuross 2003, Budge et al. 2008, McMahon and Newsome 2019, Whiteman et al. 2019). Compound specific stable isotope analysis (CSIA) takes advantage of the ability to measure individual molecules and relate them according to their biochemical pathways, focusing on processes that conserve, or preferentially modify isotopic compositions as these compounds are assimilated and utilized by organisms (Whiteman et al. 2019). This approach avoids many of the traditional issues with bulk isotopic analysis, such as uncertainties in trophic discrimination factors, and can thus provide higher resolution in answering ecological and physiological questions (Whiteman et al. 2019).

The analysis of carbon isotopes (d13C) of individual amino acids (AA) in particular holds great promise for tracing energy flow within and among organisms over both modern and paleo time scales (Fogel and Tuross 2003, Larsen et al. 2009, 2016, Whiteman et al.

2019). Importantly, the processes driving variation in d13C values are grounded in fundamental biochemical metabolic pathways and physiology. Specifically, the carbon skeleton of essential amino acids (AAESS) can only be synthesized by plants, algae, and certain microbes (Wu 2009). Thus, eukaryotic consumers typically route these compounds directly from dietary sources, resulting in minimal isotopic alteration of AAESS as they are transferred up food chains. As such, AAESS can be considered conservative tracers of the sources of primary production that fuel consumer functional groups across a variety of trophic levels (Hare et al. 1991, Fantle et al. 1999, Howland et al. 2003, McMahon et al.

2015). In addition, recent work has shown that different primary producer taxa (e.g.,

13 bacteria, fungi, plants) impart different, often distinct, d C values on AAESS during de

73 novo synthesis, which implies that these broad classes of organisms synthesize these compounds in slightly different ways (Scott et al. 2006, Larsen et al. 2009, 2013). When viewed in a multivariate space via principle component or linear discriminant analysis, these differences can result in separation among producer taxa, that can subsequently be used to quantify the energy channel(s) utilized by consumers at the individual level (Larsen et al. 2013, McMahon et al. 2016, Elliott Smith et al. 2018, Fox et al. 2019, Rowe et al.

2019). Termed ‘AA fingerprinting’, this technique has been thus far been used to track animal habitat use (Arthur et al. 2014), determine carbon sources to various food web components (McMahon et al. 2016, Elliott Smith et al. 2018, Fox et al. 2019, Rowe et al.

2019), assess linkages among aquatic and terrestrial environments (Larsen et al. 2012,

Thorp and Bowes 2017, Liew et al. 2019), and characterize changes in soil and ocean sediment biogeochemical cycles (Larsen et al. 2015). The greater resolution offered by

CSIA as opposed to bulk tissue isotope analysis has led to an explosion of research using this technique, and a strong interest in testing the assumptions implicit in these new approaches for assessing trophic level and sources of production in food webs (Whiteman et al. 2019).

13 Despite the increasing use of AAESS d C fingerprinting in ecological studies, little fundamental work has been undertaken to evaluate the mechanisms driving differences in

13 primary producer isotope fingerprints or in the AAESS d C values. The first paper to examine and identify amino acid fingerprints showed that using a multivariate framework

AA d13C values could provide insight into the metabolic pathways (e.g., autotrophy vs. heterotrophy) used by microorganisms to synthesize amino acids (Scott et al. 2006).

13 Subsequent work has found that AAESS d C fingerprints can further distinguish among a

74 variety of multicellular primary producer groups including fungi, terrestrial plants, and marine protists (Larsen et al. 2009). Two additional studies have determined that isotopic fingerprints appear to be conserved across a variety of growth conditions in wild samples of giant kelp, Macrocystis pyrifera (Larsen et al. 2013), and in cultured diatoms (Larsen et al. 2015).

13 We lack, however, a comprehensive understanding of how consistent AAESS d C fingerprints are across species and localities, and whether they respond to environmental variation through time. Furthermore, apart from the seminal paper by Scott et al. (2006)

13 that focused primarily on microorganisms, the literature using AAESS d C fingerprints has glossed over the biochemical foundations driving differences among producer fingerprints.

13 With the application of AAESS d C fingerprinting rapidly expanding, research into the underlying mechanisms behind amino acid isotopic compositions is vital to be able to fully and appropriately utilize this cutting-edge technique moving forward.

For plants, protists and bacteria, the carbon skeletons of many AA are synthesized from glucose or through precursors in the citric acid cycle (Nelson et al. 2008), and thus the carbon isotope compositions of the AA are related through central metabolic pathways.

The current methods utilized for CSIA allow confident measurement of d13C values from

13 AA (Silfer et al. 1991), representing 5 main biochemical families derived from different precursors in the metabolism of glucose (Nelson et al. 2008). These are (1) 3-

Phosphoglycerate Family: serine (Ser) and glycine (Gly), (2) Phosphoenolpyruvate

Family: phenylalanine (Phe), and tyrosine (Tyr), (3) Pyruvate Family: alanine (Ala), valine (Val), leucine (Leu), and isoleucine (Ile), (4) a-Ketoglutarate Family: glutamic acid (Glx) and proline (Pro), and (5) Oxaloacetate Family: aspartic acid (Asx), threonine

(Thr), and lysine (Lys). Depending on which pathway each of these AA derives from, and the overall molecular structure, there are vast differences in the molecular costs (Akashi and Gojobori 2002), enzymes involved, and the number of carbon atoms added or gained during synthesis (Nelson et al. 2008). Consequently, there are likely differential isotope fractionations against 13C during biosynthesis among AA families, which should result in the isotopic fingerprints, although this has yet to be explicitly tested. The six AA most commonly used in fingerprinting studies are those that are considered essential for animals and can be reliable measured via CSIA: Ile, Leu, Lys, Phe, Thr, and Val. Thus, we anticipate that subtle differences among primary producers in the synthesis of, or demand

13 for, these amino acids is likely responsible for the documented AAESS d C fingerprints observed to date (e.g., Larsen et al. 2009, 2013, Elliott Smith et al. 2018).

Here we examine the physiological and environmental factors driving variance in nearshore marine producer AA d13C values and associated isotopic fingerprints. We generated and compiled AA d13C data for 117 samples of marine primary producers from

Alaska, central (Monterey Bay) and southern (San Diego) California, and northern Chile.

In addition, we analyzed a series of 13 historical macroalgae samples collected from 1896–

13 1980 in Monterey Bay that enabled us to test how consistent AAESS d C fingerprints are through time at a single locality. From each site we also compiled publicly accessible data on the mean and variance in water temperature, as well as oceanographic conditions (e.g., degree and seasonality of upwelling). We attempt to link biogeographic patterns with known biochemical traits (e.g., presence or absence of carbon-concentrating mechanisms) and how these differ among producer groups. Our work represents an important step

13 towards understanding how observed AAESS d C patterns are produced, and the degree of resolution these analyses can provide in ecological studies.

Methods.

Sample Collection

To assess the consistency of producer AA d13C values and isotopic fingerprints across time and space, we collected 130 samples from four nearshore sites spanning three oceanographic provinces, two continents, and 120 years. Specifically, we sampled in

Katmai National Park, Alaska (58.08°N, 154.47°W), Santa Cruz, CA (36.9497° N,

122.0656° W), La Jolla, CA (32.8634° N, 117.2546° W), and Antofagasta, Chile (23.6509°

S, 70.3975° W). At each site, we collected the dominant macroalgae from three functional groups: kelps (family Laminariales), red algae (families Plocamiaceae, Rhodomelaceae,

Rhodymeniacea, and Corallinaceae), and green algae (family Ulvaceae). At every site except for Santa Cruz, we also collected particulate organic matter (POM), presumed to be composed largely of phytoplankton. Details on collection and AA d13C data for Katmai are presented in Chapter 1 (Elliott Smith et al. 2018), and the same details for Chile are presented in Chapter 3.

Primary producer samples from Santa Cruz were collected in November 2016, and from La Jolla in March of 2018. Briefly, all samples from Santa Cruz, CA were collected from Terrace Point, a long-term intertidal research site monitored by the University of

California Santa Cruz (UCSC). Following protocols from Elliott Smith et al. (2018) we collected live intertidal macroalgae: red algae (Neorhodomela sp.) and fast-growing green algae (Ulva sp.). In addition, we sampled perennial kelp (Macrocystis pyrifera) as fresh

77 drift. From La Jolla, samples of Ulva sp. were collected from the rocky intertidal of La

Jolla Shores Beach. Subtidal samples of kelp (Macrocysits pyrfiera) and red algae

(Plocamium sp.) were collected manually by SCUBA divers from the benthos (15.5 meters) of Mia’s Reef, a site ~5km southwest of Scripps Pier. La Jolla Macrocystis specimens were then sampled sequentially in 1 or 2m intervals from the holdfast to the apical meristem. We selected three of these samples per frond for analysis. After collection, specimens were kept on ice for ca. 24 hours before being frozen (-20°C). Specimens were then kept frozen until processing for stable isotope analysis. La Jolla POM samples were collected from the water column in two locations: off Scripps Pier and Mia’s Reef. In both cases, ~1–2 liters of water were pumped through pre-combusted 0.7 μm GF/F filters to collect POM. Filters containing POM were immediately dried at 50°C and stored at room temperature until analysis. Collecting permits were issued to the University of California

Santa Cruz PISCO/Marine Intertidal Research, and to University of California San Diego

Scripps Institution of Oceanography.

In addition to our modern producers we sampled a series of historical specimens of

Macrocystis pyrifera originally collected from Monterey Bay and curated at the

Smithsonian Institution National Museum of Natural History. The samples were collected from AD 1896 to 1980. Specimens were dried and pressed onto herbarium mounting paper and stored in climate-controlled facilities in manila folders. This resulted in near perfect preservation, and there were no signs of microbial alteration or decay in any of the specimens we analyzed. Importantly, metadata from the specimens showed that these samples were collected from <50km of modern macroalgae we collected from Santa Cruz,

78 making them directly comparable. We selected 13 specimens that span the 1896-1980 time range and subsampled ~150mg of dried tissue for bulk tissue and AA d13C analysis.

Metadata Collection

We compiled data on mean annual water temperature and variance (±SD) for each site from the Coastal Water Temperature Guide maintained by the National Oceanic and

Atmospheric Administration (NOAA: https://www.nodc.noaa.gov/dsdt/cwtg.html), and

Global Sea Temperatures (https://www.seatemperature.org/south-america/ch.htm), which uses satellite sea surface temperature (SST) readings generated by NOAA to estimate monthly water temperatures globally. For Scripps Pier in La Jolla, CA and Terrace Point in Santa Cruz, CA we were able to obtain long-term records of environmental conditions from within 5km of our sample locations. For Katmai National Park, the nearest NOAA station was located in Shelikof Strait, ~ 13km from our intertidal site (Elliott Smith et al.

2018). For Antofagasta, the Global Sea Temperatures provided data for the main port, which was ~20km from our sampling location (see Methods in Chapter 3). We also assigned each locality an oceanographic ‘condition’ (e.g., upwelling vs. downwelling), and noted whether this condition was strongly seasonal: Katmai National Park was assigned a downwelling zone (Reed and Schumacher 1986), Santa Cruz and San Diego, CA were assigned as seasonal upwelling zones (Bograd et al. 2009), and Antofagasta, Chile was assigned as a year-round upwelling zone (Thiel et al. 2007).

In addition to environmental data, we conducted literature searches to assess differences among our sampled primary producers in their physiology, specifically their mode of carbon acquisition and fixation (Table 4.1). In particular, we focused on

Table 4.1. List of samples and associated metadata used in the present study. POM = particulate organic matter, SST = sea surface temperature, CCM = carbon concentrating mechanisms. *Species identities of POM samples are unknown, however, the majority of eukaryotic marine phytoplankton possess CCM, and we thus made the assumption that our POM likely had taxa with CCM. •The presence of CCMSs has not been evaluated in this genus. However, other kelps examined all possess CCM, and we assumed this was also the case for Lessonia. †Species identity of this crustose coralline alga is unknown. We assumed our sample lacked CCM based on data for other crustose Rhodophyta from Stepien et al. (2016).

SST Functional Time CCM CCM Site Mean ± Description Genus N Group Period Present? Reference SD (°C) Stepien et

Laminaria Kelp 9 Modern Yes al. 2016 Stepien et Neorhodomela Red Algae 4 Modern Yes al. 2016 5.0 ± 4.5 Downwelling Stepien et Ulva Green Algae 5 Modern Yes al. 2016

Katmai Alaska Katmai Reinfelder -- POM 4 Modern Yes* 2011 Historic Stepien et Macrocystis Kelp 13 Yes

al al. 2016 Stepien et Macrocystis Kelp 8 Modern Yes 13.4 ± Seasonal al. 2016 1.4 Upwelling Stepien et Neorhodomela Red Algae 7 Modern Yes California

Santa Cruz Cruz Santa al. 2016 Stepien et Ulva Green Algae 7 Modern Yes al. 2016 Stepien et Macrocystis Kelp 15 Modern Yes al. 2016

Stepien et Plocamium Red Algae 6 Modern No 16.6 ± Seasonal al. 2016 2.1 Upwelling Stepien et

La Jolla La Ulva Green Algae 5 Modern Yes California al. 2016 Reinfelder -- POM 7 Modern Yes* 2011 Lessonia Kelp 7 Modern Yes• -- Stepien et Macrocystis Kelp 5 Modern Yes al. 2016 Stepien et Plocamium Red Algae 4 Modern No

al. 2016 Stepien et Rhodymenia Red Algae 1 Modern No al. 2016 18.0 ± Consistent Unidentified Stepien et 2.4 Upwelling Crustose Red Algae 1 Modern No† al. 2016 Coralline Unidentified Antofagasta Chile Antofagasta Red Algae 1 Modern Unknown -- red Stepien et Ulva Green Algae 7 Modern Yes al. 2016 Reinfelder -- POM 14 Modern Yes* 2011

identifying which species possessed carbon concentrating mechanisms (CCM), as these physiological adaptations have been well-documented be major factors in determining patterns in bulk d13C values of marine primary producers (Raven et al. 2008, Stepien et al.

2016). Stepien et al.’s paper was the most extensive source for this information (2016).

They conducted pH drift experiments, reviewed the presence or absence of CCM in the

Rhodophyta and Ochrophyta, and performed a phylogenetic analysis on the ancestral state of CCM in these phyla. For our sampled kelp genus endemic to South America, Lessonia, the presence of CCM have not been assessed. However, because CCM have been identified for all Ochrophyta measured (Stepien et al. 2006), we assumed Lessonia likely possesses them. For our singleton sample of crustose coralline red algae from Chile, where we did not have a confident species identification, we assumed CCM were absent, based on the consistent data for crustose Rhodophyta in Stepien et al. (2016). We do not have data on the phytoplankton taxa represented in our samples. However, because the vast majority of eukaryotic marine phytoplankton possess CCM (Reinfelder 2011), we assumed this was the case for all of our POM samples.

Isotopic Analysis

Where possible, for all samples we measured bulk tissue d13C and d15N values as well as amino acid d13C. For bulk analysis, macroalgae specimens were rinsed with deionized water, and lyophilized. Dried samples were then homogenized, weighed out to

~5mg and sealed in tin capsules for bulk isotope analysis. Pieces of filters containing POM were lyophilized, weighed out to ~20mg and sealed in tin capsules. Bulk d13C and d15N isotope values, along with mass percent [C] and [N], were measured via EA-IRMS using a Costech elemental analyzer (Costech, Valencia, CA) coupled to a Thermo Scientific

Delta V isotope ratio mass spectrometer (Thermo Fisher Scientific, Bremmen, Germany) at the University of New Mexico Center for Stable Isotopes (UNM-CSI). The within-run

81 standard deviation of reference materials was ≤ 0.2‰ for d13C and d15N values. For two

POM samples and a kelp from Antofagasta, Chile (HUM-ISM-PEL3-FITO, HUM-ISM4-

FITO, and HUM-ISM12-MIN3-B3) initial bulk analyses failed due to low peak sizes, and there was not enough remaining material to conduct bulk and AA d13C analysis. Thus, for these samples we only have amino acid data (Appendix 4, Table 1).

We used the same primary producer samples for bulk tissue and EAA d13C analysis.

Approximately 5mg of macroalgae tissue was hydrolyzed to constituent amino acids in

1mL of 6N hydrochloric acid (HCl) at 110°C for 20 hours; tubes were flushed with N2 to prevent oxidation during hydrolysis. Whole silica filters containing POM were hydrolyzed with 1.5mL of 6N HCl. Hydrolyzed primary producer samples were passed through a cation exchange resin column (Dowex 50WX8 100-200 mesh) to isolate amino acids from other metabolites (Amelung and Zhang 2001). After hydrolysis/Dowex purification, amino acids were derivatized to N- trifluoroacetic acid isopropyl esters following established protocols (O’Brien et al. 2002, Newsome et al. 2011, 2014). Samples were derivatized in batches of 8–17 along with an in-house reference material containing all amino acids we measured.

d13C values of individual derivatized amino acids were measured using a GC-C-

IRMS system. Derivatized samples were injected into a 60m BPx5 gas chromatograph column for amino acid separation (0.32 ID, 1.0μm film thickness, SGE Analytical Science,

Victoria, Australia) in a Thermo Scientific Trace 1300, combusted into CO2 gas via a

>0.7‰ across injections. The within-run standard deviations of measured d13C values among AA in the in-house reference material ranged from 0.0‰ (leucine) to 1.2‰

(alanine).

We reliably measured d13C values of 12 amino acids. Six are considered essential for animals: isoleucine (Ile), leucine (Leu), lysine (Lys), phenylalanine (Phe), threonine

(Thr) and valine (Val), and the other six are considered nonessential: alanine (Ala), aspartic acid (Asp), glutamic acid (Glu), glycine (Gly), proline (Pro), and serine (Ser). The reagents used during derivatization add carbon to the side chains of amino acids, and hence raw d13C values measured via GC-C-IRMS reflect a combination of the intrinsic amino acid carbon and reagent carbon (Silfer et al. 1991). See Appendix 4 for correction equations. For 14 of our total 130 samples we were not able to obtain reliable d13C values for either one or two amino acids. These varied by sample, but were mostly nonessential AA: Pro (n=4), Gly

(n=4), Ser (n=3; one of these samples was also missing Gly), Asp (n=1), Glu (n=1), and

Phe (n=2) and Lys (n=1). We excluded the samples missing AA d13C data from multivariate analyses.

Statistical Analyses

To evaluate the factors driving patterns in marine producer AA d13C values, we used (1) generalized linear models (GLM) to assess the influence of different factors from metadata collections on raw AA d13C values, (2) linear discriminant analysis to define AA d13C ‘fingerprints’, and finally (3) a series of linear regressions comparing d13C values of paired AA to look for differences in carbon isotope fractionation during AA synthesis among our sampled producer groups. We assessed all data for normality and homogeneity

13 of variance using multivariate regressions with species as predictor of AA d C values and evaluated the normality of residuals using fitted versus residuals plots, Q-Q plots, and

Cramer von Mises Tests. We tested homogeneity of variance via the Brown and Forsythe test. Our dataset from Antofagasta initially failed normality checks due to the presence of two extreme serine values (+5.6‰, and -45.8‰) for one Ulva and one Rhodymenia sample, respectively. The bulk tissue d13C data corroborated the values, with the Rhodymenia sample having one of the lowest d13C bulk values (-33.0‰) and the Ulva sample having among the highest (-11.7‰) values we measured for producers in Chile. We thus opted to leave these samples and serine values in the dataset as they represent real biological information and not anomalous measurements. We conducted all statistical analyses in program R (v 3.3.1) with RStudio interface (v 0.99.903).

We first employed generalized linear models (GLMs) to evaluate potential predictors of individual AA d13C values. We excluded the non-coralline red algae singleton from Chile for which we did not have a confident species identification. For each amino acid, we compared alternative models within an information theoretic framework, evaluating relative support for models with one to four main effects: presence/absence of

CCM, taxonomy, and time period (all categorical variables), and mean water temperature and standard deviation in temperature (both continuous variables). In addition, we considered an interaction effect between mean water temperature and standard deviation in temperature. For all AA, we also evaluated support for a constant model with no significant effects or interactions, and the model with the lowest Akaike information criterion (AIC) score was considered to be the best supported model while including the

84 smallest number of necessary explanatory variables for a given response; see Appendix 4,

Table 2 for exact model structures.

We used linear discriminant analysis (LDA) to examine differences among primary producer groups in a multivariate framework following Larsen et al. (2013), and Elliott

Smith et al. (2018). We ran two LDA models. The first included all AAs and a second only included AAs considered essential for animals (Ile, Leu, Lys, Phe, Thr and Val). We examined the classification error rate for the four producer functional groups (POM, green algae, red algae, and kelps) using a leave-one-out, cross-validation procedure to establish if our sources were statistically distinct. We also considered the loadings of each AA along the linear discriminant axis to determine which compounds were most useful in distinguishing among different groups. After running the AAESS only LDA model, we pulled the coordinates of each sample along the LD axes and put these data into our GLM framework described above. This allowed us to assess if the same, or different factors drive

13 13 primary producer AAESS d C fingerprints as drive measured AAESS d C values.

After analyzing the broad taxonomic and environmental factors that determine marine primary producer AA d13C values and isotopic fingerprints, we aimed to provide a more mechanistic biochemical framework for these patterns. We thus followed the framework originally developed by Scott et al. (2006) and used a series of linear regression models to compare d13C values among AA pairs derived from the same biochemical precursor. The biosynthesis and remodeling of all AA is connected via two key metabolic pathways: glycolysis and the citric acid cycle. In plants, protists, and some microbes, different combinations of amino acids that are essential or nonessential for animals are synthesized from various intermediaries (precursors) in this pathway (Figure 4.1).

Glucose

3-Phosphoglycerate

Glycine Serine

Phosphoenolpyruvate *Phenylalanine

Pyruvate Alanine *Leucine *Valine Acetyl-CoA

Oxaloacetate Citric Acid Cycle

Aspartic Acid α-Ketoglutarate *Threonine *Isoleucine *Lysine Glutamic Acid Proline

Figure 4.1. Biochemical relationships among amino acids. Shown are the amino acids that can be reliably measured by CISA methods (Silfer et al. 1991), and the intermediaries in glycolysis and citric acid cycle that they are synthesized from (Nelson et al. 2008). Asterisks (*) indicate amino acids that are considered essential for animals.

Consequently, we considered both essential and nonessential amino acids may be important in driving the fingerprinting patterns we analyzed with multivariate techniques.

For each primary producer group, we conducted linear regressions for the following AA pairs (Appendix 4, Table 3): Pro-Glu, Thr-Asp, Lys-Asp, Val-Asp, Leu-Ala, Ser-Gly, Ser-

Ala, and Asp-Ala. We also compared Ile-Asp, which was not considered in the original

Scott et al. (2006) manuscript. The intercepts of each of these relationships can be

13 considered the isotopic discrimination, or D CAA-AA, associated with each synthesis

2 13 pathway. We compared the R , P-value and intercept (D CAA-AA) of these relationships for each producer group, as well as to data reported by Scott et al. (2006). We then evaluated whether these fractionation factors were different among producer functional groups by calculating 95% confidence intervals around the intercept estimates; if confidence intervals

13 did not overlap producer intercepts (D CAA-AA) were considered statistically distinct.

Results

Isotopic Composition of Marine Producers

Nearshore marine primary producers showed substantial variation in isotope values for bulk tissue d13C and d15N, as well as amino acid d13C (Appendix 4, Table 1; Figure

4.2). We found bulk tissue d15N values differed primarily among sites ([F(3, 123) = 50.89, p < 0.01]), with primary producers having the highest d15N in Chile and the lowest in

Alaska. Bulk d13C values varied among producers (([F(3, 123) = 27.48, p < 0.01]), with green algae (Ulva) having the highest and red algae having the lowest values (Figure 4.2).

Our sampled marine primary producers exhibited a high degree of variance in AA d13C values. Serine exhibited the widest variance, from -45.8‰ (Rhodymenia sample) to

+6.8‰ (Laminaria sample), and phenylalanine exhibited the lowest, from -39.2‰

(Plocaminum sample) to -16.0‰ (Macrocystis sample). We also found significant differences in AA d13C values among primary producers (Appendix 4, Table 1). Although all producers were statistically distinct from one another for at least one AA, we noted

12 4

4 4 4 3 2 Group 10 Ulva 2 Kelps POM 2 Red Algae 3 3 N 8 3 Site 15 1 1 Katmai, AK δ 1 1 2 Santa Cruz, CA 3 Santa Diego, CA 6 4 Antofagasta, Chile 1

−35 −30 −25 −20 −15 −10 δ13C Figure 4.2. Average bulk d13C and d15N values for sampled marine primary producers. Colors represent different primary producer taxa sampled; numbers indicate the locality. See Table 4.1 for sample sizes and details on collection and specific taxa included. For individual d13C and d15N values. differences in which AA d13C values differed among our four producer groups. For example, our sampled kelps and green algae only exhibited differences in raw d13C values for valine and threonine, whereas red algae showed strikingly different d13C values for all but one amino acid.

Generalized Linear Models

For the majority of amino acids that we measured, a comparison of generalized linear models showed the strongest support for a model with main effects for the presence

88 of carbon concentrating mechanisms (CCM), producer phyla, and an interaction effect of site-specific mean water temperature and yearly variance around that temperature

(Appendix 4, Table 2). For five amino acids (Leu, Asp, Lys, Pro, and Glu), a model with an additional main effect of time period (historical vs. modern) received marginally more support (Appendix 4, Table 2). None of these factors alone produced a well-supported model, with Akaike information criterion (AIC) scores in most single-effect cases only marginally better than the null model (Appendix 4, Table 2). The effect of site oceanography (e.g., downwelling vs. upwelling) resulted in extremely poor model performance, with AIC scores higher than the null model (Appendix 4, Table 2). We thus removed this effect from all additional models that we tested.

13 After running the LDA model with only AAESS d C data (see results discussed below), we pulled out the linear discriminant coordinates for each primary producer along the two main axes of discrimination and examined which GLM best predicted producer isotopic fingerprints. A model with CCM, phyla, time period, and the same water temperature interaction effect received the best support for predicting the first linear discriminant axis (LD1); however, this model was only marginally more supported than a single effect model with only producer phyla as a predictor (Appendix 4, Table 2). For

LD2, a model with CCM, phyla, and water temperature/variance was the best supported model, although again this was only marginally more significant than the phyla only GLM.

Linear Discriminant Analysis – AA d 13C Fingerprinting

Results from our multivariate analysis of AA d13C values showed strong separation among producer groups (Table 4.2, Appendix 4, Tables 4 and 5; Figure 4.3). For both models (AAALL and AAESS only), the first two linear discriminant (LD) axes cumulatively explained >90% of the variation in our dataset (Appendix 4, Tables 4 and 5). For the model

13 with all AA (AAALL) included, AAESS d C values drove the differences among producer groups. For the first axis (LD1) threonine, isoleucine, and valine d13C values were the most informative for distinguishing among groups, while leucine, isoleucine, and phenylalanine d13C values drove differences along the second discriminant LD2 (Appendix

5.0

4 4 Group 4 4 Ulva 2.5 2 1 3 4 Kelps 2 2 1 2 3 4 2 4 44 44 4 POM 1 44 2 4 4 4 Red Algae 3 3 3 4 4 2 4 1 4 4 223 33 1 1 2 1 3 1 3 2 4 1 3 2 2 3 1 1 1 Site 0.0 3 2 23133 12 3 4 2 1 Katmai, AK

LD2 4 4 2 4 2 2 2 2 2 2 3 2 3 4 4 2 2 4 2 Santa Cruz, CA 4 1 4 3 32 2 1 3 42 2 3 3 3 3 Santa Diego, CA 2 2 4 4 2 1 4 Antofagasta, Chile 331 4 2 3 −2.5 3 43 1

3 3 4 3

−5.0

−5.0 −2.5 0.0 2.5 5.0 LD1 Figure 4.3. Results from linear discriminant analysis using essential amino acid d13C data of marine primary producers. Circles represent the linear discriminant loading of individual producer samples; ellipses surrounding producer groups represent 90% confidence intervals. See Table 4.2 for reclassification rates. Colors represent broad producer taxonomy: Rhodophyta, Ochrophyta, Chlorophyta, and particulate organic matter (POM). Number indicate the locality where each sample was collected. Within the kelp group, numbers in orange font indicate historical Macrocystis samples from close proximity to our modern Santa Cruz locality.

Table 4.2. Linear discriminant analysis (LDA) cross-validation of marine primary producers using all 12 amino acids measured (top half of table) versus only essential amino acids (bottom half). Columns represent classification of a sample group as given by LDA. Rows represent the group being classified. The cross diagonal (bold text), are correct reclassifications. Numbers falling off this diagonal indicate samples that were incorrectly reclassified by the model. Thus, the sum of the cross diagonal divided by the total number of samples equals the % success of the model (All AA: 89.6%, AAESS only: 86.7%). For each model, we removed samples that were missing values for relevant amino acids. See Figure 4.3 for a visual representation of these data. Species codes are as follows: Green = Ulva sp., POM = particulate organic matter and phytoplankton, Kelp = Laminaria sp., Lessonia sp., and Macrocystis pyrifera, Red = Neorhodomela sp., Plocamium sp., Rhodymenia sp., and one singleton each of crustose coralline algae and an unidentified red alga.

LDA Classification

Green Kelp POM Red % Correct

Green 14 0 7 0 67.7

Kelp 1 51 0 0 98.1 All AA: Identity POM 3 1 19 0 82.6 True Producer Red 0 0 0 20 100.0

Green 13 0 9 1 56.6

Kelp 2 54 0 0 96.4

Identity POM 8 1 14 1 58.3 True Producer

Essential AA Only: Red 0 0 0 24 100.0

4, Table 4). Results were very similar for the AAESS only model. Threonine, isoleucine, and valine d13C values had the strongest loadings on LD1, while leucine, isoleucine, and threonine d13C values loaded strongly on LD2 (Appendix 4, Table 5). Cross-validation of the LDA found a high successful reclassification rate among our four primary producer groups for both the AAALL (89.6%) and AAESS only (86.7%) models. Red algae had the highest successful reclassification rate (AAALL: 100%, AAESS: 100%) with 20/20 and 24/24

91 samples reclassifying correctly for AAALL and AAESS models respectively (Table 4.2).

Green algae had the lowest overall reclassification rate (AAALL: 67.7%, AAESS: 56.6%), with a number of samples incorrectly classifying with particulate organic matter (Table

4.2). POM had low reclassification rates for the AAESS model (58.3%), with 8/24 samples incorrectly classifying with green algae, but a high reclassification rate for the all AA model (82.6%; Table 4.2). Kelps had consistently high reclassification rates for the AAALL and AAESS models with respectively only 51/52 (98.1%) and 54/56 (96.4%) individuals correctly reclassifying (Table 4.2). In summary, using AA d13C values in an LDA framework allows us to reliably distinguish kelp and red algae from each other and the other two producer groups (Figure 4.2). Distinguishing between POM and green algae is not as successful, perhaps because their metabolic pathways are more similar.

Pairwise Amino Acid Linear Regression

We found significant relationships for almost all amino acid (AA-AA) pairs we examined among our four producer groups (Appendix 4, Table 3; Figure 4.4). The exceptions were Ser-Gly (R2 = 0.03, p-value = 0.22) and Ser-Ala for green algae (R2 =

0.05, p-value = 0.17), as well as Ser-Ala for particulate organic matter (R2 = 0.08, p-value

= 0.10). Among significant relationships, goodness of fit ranged from a low of R2 = 0.09

(Ser-Ala in kelps) to a high of R2 = 0.90 (Lys-Ala in red algae), with the majority of AA-

AA pairs (20/36) showing linear relationships with R2 >0.6. We also found substantial

13 differences among producer groups in D CAA-AA (calculated intercept) for our examined amino acid pairs (Appendix 4, Table 3; Figure 4.4). For all but one amino acid pairs calculated, at least one producer group exhibited a significantly different intercept from the

92 other three – the exception was Ser-Gly where the calculated intercepts (± 95% confidence intervals) for kelps, red algae, and green algae all overlapped. Regressions between Ile-

Asp and Val-Ala showed the strongest differences among producer groups, with all four of our measured producer groups showing distinct intercepts (Appendix 4, Table 3). The

13 comparison with data from Scott et al. (2006) showed the microbial D CAA-AA data matched with one of our producer groups for only five of eight A-A pairs (excluding Ile-

Asp which was not calculated in the original manuscript): Val-Ala, Leu-Ala, Asp-Ala, Pro-

Glu, and Lys-Asp.

4 b A. Oxaloacetate Derived Thr-Asp Ile-Asp 2 a a a 0 d

-2

-4

a -6

c -8

-10 b AA-AA -12 C

13 -8 B. Pyruvate Derived Val-Ala

Δ b Leu-Ala a -10 c

-12 d

-14 c

a -16 a

-18

-20 b

-22 Greens Kelps POM Reds Bacteria* Figure 4.4. Isotopic fractionation among amino acid pairs for marine producers and published data for bacteria from Scott et al. (2006). Shown are the fractionation and associated 95% confidence intervals for four amino pairs from two main biochemical families. Panel A shows pairs of amino acids linked through a common intermediate in the citric acid cycle: Oxaloacetate, and Panel B shows amino acid pairs derived from Pyruvate. Observed isotopic discrimination and error estimate are derived from multiple linear regression models comparing d13C values for each amino acid pair by marine producer 13 group; we defined the intercept of each relationship as the D CAA-AA. Letters over each symbol represent the statistical grouping for that AA-AA pair and producer: for example, 13 13 green algae and kelps have significantly different D CIle-Asp but not D CThr-Asp. POM = particulate organic matter. Asterisks (*) indicate microbial data from Scott et al. (2016), which did not measure the relationship between Ile-Asp d13C values.

Discussion

With the advent of gas-chromatography-combustion continuous flow techniques

(Silfer et al. 1991), the isotopic analysis of individual amino acids has increased in importance in the natural sciences (Whiteman et al. 2019). The enthusiasm for compound- specific stable isotope analysis (CSIA) is well-founded, because this technique potentially offers a high degree of resolution in addressing a wide variety of ecological and physiological questions (Whiteman et al. 2019). As with many new proxies, however, research into the applications of CSIA have superseded a careful analysis of the limitations

13 inherent in the approach. Notably, the use of AAESS d C ‘fingerprinting’ has been widely applied (Scott et al. 2006, Larsen et al. 2009, 2013, 2015, Thorp and Bowes 2016, Jarman et al. 2017, Elliott Smith et al. 2018, Liew et al. 2019, Fox et al. 2019) despite the fact that we lack a comprehensive understanding of why and how these isotopic fingerprints vary across taxonomic units, geographic space, and time. Here we addressed this by analyzing

13 AA d C values, modeling their relationships on the basis of biochemical synthesis pathways, and characterizing isotopic fingerprints of four broad taxonomic groups of marine primary producers from disparate localities and time periods (Table 4.1). We find that d13C values of individual amino acids are strongly influenced by primary producer physiology and taxonomy, as well as local environmental conditions. We also find remarkable consistencies in the isotopic fingerprints of Rhodophyta and Ochrophyta

(kelps), indicating these statistical patterns are driven by fundamental biochemical pathways and can be reliably identified over broad spatiotemporal scales. However, we were unable to reliably and consistently identify phytoplankton vs. green algae based on isotopic fingerprints, which presents a limitation in using AA d13C analysis for

95 characterizing the importance of benthic and pelagic endmembers. Finally, in comparing d13C values of individual amino acids across biochemical families we identify a number of key synthesis pathways that may be driving these patterns. Our work fills an important gap by providing a mechanistic framework through which to interpret differences among primary producer AA d13C values and isotopic fingerprints.

We identified key physiological and environmental factors driving the d13C values of individual amino acids within nearshore marine primary producers. A comparison of multiple generalized linear models (GLM) found the strongest support for a model which predicted AA d13C values from (1) presence of carbon concentrating mechanisms (CCMs),

(2) primary producer taxonomy, and (3) mean local water temperature (Appendix 4, Table

2). These findings compliment decades of work based on bulk tissue isotope analyses, which has found the mode of inorganic carbon acquisition to be a key predictor of marine primary producer d13C values (Maberly et al. 1992, Raven et al. 2002). In marine

- macroalgae, CCMs primarily involve the use of HCO3 , either directly through an anion transport protein, or indirectly through the production of extracellular carbonic anhydrase

- to increase the rate of conversion of HCO3 into CO2 (Raven and Giordano 2017). There is

13 - a large difference in the d C values of aqueous CO2 (-9‰) and HCO3 (+0.5‰ at 14°C;

Wendt 1968), and these differences in inorganic carbon source values are reflected in the primary producers using them; algae with CCMs tend to have the highest d13C values locally, whereas algae which rely exclusively on diffusive CO2 entry into the cell tend to have bulk d13C values of less than -30‰ (Maberly et al. 1992).

Our work supports these paradigms at the compound level – the presence of CCMs was a major driver of AA d13C values. As expected, red algae genera lacking CCMs

(Plocamium and Rhodymenia) had the lowest AA d13C values. Among algae genera known to use CCMs, there were significant differences among functional groups (Appendix 4,

Table 1). Depending on the site, either green algae (Ulva) or kelps (Laminaria, Lessonia,

Macrocystis) tended to have the highest AA d13C values, while values of POM were consistently lower (Appendix 4, Table 1). This likely relates to the particular CCMs possessed by each group. However, the lack of information about the particular planktonic taxa in our POM samples, as well as baseline knowledge of CCMs in some of our kelps

(e.g., Lessonia), hinders our ability to examine this pattern in more detail. Nonetheless, we conclude that the mode of inorganic carbon acquisition likely sets the baseline d13C value of individual amino acids, with variance around these values driven by local environmental conditions (Appendix 4, Table 2).

Amino acid d13C fingerprints were consistent among producer groups across broad spatiotemporal scales from Alaska to northern Chile. For both the AAALL and AAESS LDA models, the major marine primary producer groups exhibited successful reclassification rates of >80% (Table 4.2, Figure 4.3). Considering this dataset includes over 10 unique genera from two continents and three oceanographic environments collected over the past

120 years (Table 4.1), this finding is truly remarkable. Isotopic fingerprints of red algae and kelp in particular were highly constrained, with >95% of individuals from each group successfully reclassifying in both LDA models (Table 4.2, Figure 4.3). Furthermore, the historical Macrocystis samples, dating back to AD 1896, exhibit the same AA d13C fingerprints as modern samples of the same species collected in the same general locality

(Figure 4.3). Hence, AA d13C fingerprinting analysis provides a way of tracing the importance of kelp- or red-algae derived energy to consumers across both space and time

97 with a high degree of confidence; to our knowledge, no other such analysis currently exists.

Importantly, this finding also highlights the relevance and usefulness of museum collections: the incorporation of historical specimens allowed us to add a temporal perspective, and thus broaden the potential applications of this tool beyond just modern ecological questions.

In contrast to the unique isotopic fingerprints in samples of Rhodophyta and

Ochrophyta, we were unable to discriminate between particulate organic matter (POM) and green algae (Ulva) using AA d13C analysis. For both LDA models, these groups overlapped substantially with one another (Appendix 4, Table 2), particularly for the AAESS model where only 58.3% and 56.6% of POM and Ulva respectively reclassified successfully. This is due to Ulva and POM having statistically different d13C values for only two non-essential

AA, aspartic acid and serine (Appendix 4, Table 1), neither of which were important drivers of isotopic fingerprints (Appendix 4, Tables 4 and 5). These results suggest that biosynthetic pathways for the synthesis of AAs are too similar among eukaryotic phytoplankton and green macroalgae to be distinguished using this technique. Until the variables determining overlap among POM and Ulva are elucidated, we urge caution in using AA d13C analysis in regions where green algae are a major source of production.

The relationships among individual AA d13C values provided insight into the biochemical mechanisms producing isotopic fingerprints (Appendix 4, Table 2). For both the AAALL and AAESS LDA models, the differences among primary producer groups in multivariate space were due to differences in valine, threonine, and isoleucine d13C values

(Appendix 4, Tables 4 and 5). Results from linear regressions of AA pairs found that the

13 calculated isotopic discrimination, D CAA-AA (e.g., intercept of AA1 versus AA2), for Val-

Ala and Ile-Asp were significantly different among all four producer groups. Indeed, these are the only AA pairs for which each producer group exhibited a unique D13C (Appendix

4, Table 2). D13C for Val-Ala ranged from -9.1‰ ± 0.2 (kelps) to -14.9‰ ± 0.5 (POM), and D13C for Ile-Asp ranged from -1.2‰ ± 0.8 (red algae) to -11.4‰ ± 0.2 (kelps). There are two possible interpretations of this result: (1) there are different enzymes used among nearshore marine primary producer groups in the synthesis of valine and isoleucine, or (2) there is differential demand for them among these groups. Either scenario could result in differential isotopic discrimination. For example, in terrestrial primary producers the two primary enzymes used in the capture and fixation of inorganic carbon, PEPC and Rubisco, exhibit very different preferences for 13C and 12C, resulting in different d13C values in the plants utilizing those enzymes (Farquhar et al. 1989, Fogel and Cifuentes 1993). Our data suggest similar processes may be operating in the biosynthesis of amino acids among marine primary producers, with differences in the enzymes critical for valine or isoleucine synthesis used among groups (e.g., dihydroxy-acid dehydratases, and branch-chain-amino- acid-transaminases; Kanehisa and Goto 2000) possibly resulting in differential discrimination against 13C-enriched precursors.

Our study exemplifies the power of using isotopic analysis of individual compounds to address physiological, modern- and paleo- ecological questions. Our findings compliment and expand upon previous research (Larsen et al. 2013, 2015, Liew et al. 2019) in showing that amino acid d13C fingerprints of two major clades of primary producers – kelps and red algae – are unique, and consistent across different genera, oceanographic regions, and time periods. Because essential amino acids appear to be minimally isotopically modified through food webs (Hare et al. 1991, Fantle et al. 1999,

McMahon et al. 2015), AA d13C fingerprinting analysis can thus provide researchers a high-resolution method of tracing the flow of energy through various compartments of nearshore food webs, addressing questions of connectivity between adjacent terrestrial and marine ecosystems, and potentially address the importance of microbially-derived or mediated energy to consumers. However, in line with these important ecological research areas, a number of methodological issues still remain to be addressed, including: (1) the

13 consistency of AA d C fingerprints among terrestrial C3, C4, and CAM producers, (2) how

13 reliably AAESS d C values are transferred up multiple trophic levels within a single food web, and (3) the role of microbial diagenesis of original producer isotopic fingerprints – both within animal guts and in external decomposition. Finally, one of the most powerful applications that we see for AA d13C fingerprinting analysis is in its application to historical and archaeological specimens. These data will allow researchers to assess how vulnerable communities were structured historically, how they have responded to previous environmental change. Given that coastal ecosystems today are currently experiencing high, and escalating degrees of anthropogenic impacts (Jackson et al. 2001, Ceballos et al.

2015, McCauley et al. 2015, Catton et al. 2016), these analyses will provide vital baseline data for conservation efforts to mitigate the loss of biodiversity and ecosystem function.

100

Appendix 1

1. Equations for correcting amino acid d13C data.

The reagents used during derivatization add carbon to the side chains of amino acids, and so the raw d13C measured via GC-C-IRMS is a combination of original amino acid carbon and reagent carbon (O’Brien et al., 2002, Newsome et al., 2014). However, because amino acid standards of known d13C composition were derivatized and run with each batch of samples, we were able to correct for this carbon addition for each amino acid using the following equation (From O’Brien et al., 2002, Newsome et al., 2011; 2014):

,- ,- ( ,- ) ,- δ $/%& − δ $/%0 + δ $%0/ × p%0/ δ $%&'()* = p%0/

13 Here, d Cdsa refers to the measured value of the derivatized amino acid within the sample

13 and d Cdst refers to the measured value of the same derivatized amino acid within the

13 internal standard. d Cstd is the known un-derivatized value of that amino acid in the standard (measured via EA-IRMS), and pstd is the proportion of carbon in the measured derivative that was originally sourced from the amino acid, which varies quite substantially among amino acids.

101 dissolved AA solutions mixed together. For each batch of samples, a small aliquot of this solution was dried and then derivatized alongside each batch of unknown samples.

102

Table 1. Bulk tissue d13C and d15N values and associated weight percent [C]:[N] ratios for all macroalgae and macroinvertebrate samples from Amalik Bay site, as well as all 2012 and 2013 offshore POM samples. In Sample ID’s location codes are as follows: KATM = Katmai National Park, AM = Amalik Bay (58.08°N, -154.47°W), KN = Kinak Bay (58.19°N, -154.47°W), KK = Kukak Bay (58.32°N, -154.21°W), KF = Kaflia Bay (58.26°N, -154.20°W), and TK = Takli Island (58.06°N, -154.48°W). These data are depicted visually in Figure 1.1.

Year Species Sample ID d13C d 15N C:N

2012 Mytilus sp. KATM-AM-MYT-2012-01 -18.8 7.9 4.1

2012 Mytilus sp. KATM-AM-MYT-2012-02 -18.6 7.5 4.2 2012 Mytilus sp. KATM-AM-MYT-2012-03 -18.1 8.0 4.0

2012 Mytilus sp. KATM-AM-MYT-2012-04 -18.3 7.7 4.0

2012 Mytilus sp. KATM-AM-MYT-2012-05 -17.9 8.4 3.6 2012 Mytilus sp. KATM-AM-MYT-2012-06 -19.2 8.1 5.0 2012 Mytilus sp. KATM-AM-MYT-2012-07 -18.2 8.2 3.9 2012 Mytilus sp. KATM-AM-MYT-2012-08 -17.5 9.1 3.7

2012 Mytilus sp. KATM-AM-MYT-2012-09 -18.1 8.7 3.5 2012 Mytilus sp. KATM-AM-MYT-2012-10 -17.9 8.5 3.5 2012 Nucella sp. KATM-AM-NUC-2012-01 -16.8 11.0 4.1

2012 Nucella sp. KATM-AM-NUC-2012-03 -15.7 11.2 3.6 2012 Nucella sp. KATM-AM-NUC-2012-05 -16.2 10.8 4.0 2012 Nucella sp. KATM-AM-NUC-2012-06 -16.2 10.7 3.9 2012 Nucella sp. KATM-AM-NUC-2012-07 -16.0 11.0 3.8

2012 Nucella sp. KATM-AM-NUC-2012-08 -17.4 9.7 4.3

2012 Nucella sp. KATM-AM-NUC-2012-09 -17.5 9.3 4.9 2012 Nucella sp. KATM-AM-NUC-2012-10 -15.8 10.8 3.9

2012 Pycnopodia helianthoides KATM-AM-PYC-2012-01 -13.3 12.8 3.9 2012 Pycnopodia helianthoides KATM-AM-PYC-2012-02 -13.7 12.2 4.0

2012 Pycnopodia helianthoides KATM-AM-PYC-2012-03 -14.1 12.2 3.9

2012 Pycnopodia helianthoides KATM-AM-PYC-2012-04 -12.4 15.1 3.6 2012 Pycnopodia helianthoides KATM-AM-PYC-2012-05 -14.1 12.5 4.1

103

2012 Pycnopodia helianthoides KATM-AM-PYC-2012-06 -13.3 14.4 3.9

2012 Pycnopodia helianthoides KATM-AM-PYC-2012-07 -13.9 12.4 4.2

2012 Pycnopodia helianthoides KATM-AM-PYC-2012-08 -14.0 12.0 3.7

2012 Pycnopodia helianthoides KATM-AM-PYC-2012-09 -13.7 14.1 3.8

2012 Pycnopodia helianthoides KATM-AM-PYC-2012-10 -13.4 12.9 3.8

Strongylocentrotus 2012 KATM-AM-STR-2012-01 -18.8 7.3 7.4 droebachiensis

Strongylocentrotus 2012 KATM-AM-STR-2012-02 -19.4 7.7 6.9 droebachiensis Strongylocentrotus 2012 KATM-AM-STR-2012-03 -17.5 8.0 6.7 droebachiensis Strongylocentrotus 2012 KATM-AM-STR-2012-04 -15.3 7.5 5.2 droebachiensis Strongylocentrotus 2012 KATM-AM-STR-2012-05 -17.8 7.2 5.9 droebachiensis Strongylocentrotus 2012 KATM-AM-STR-2012-06 -19.7 7.6 7.7 droebachiensis

Strongylocentrotus 2012 KATM-AM-STR-2012-07 -17.6 7.6 7.8 droebachiensis

Strongylocentrotus 2012 KATM-AM-STR-2012-08 -18.4 6.8 7.5 droebachiensis Strongylocentrotus 2012 KATM-AM-STR-2012-09 -19.1 7.3 5.9 droebachiensis Strongylocentrotus 2012 KATM-AM-STR-2012-10 -18.4 7.6 5.6 droebachiensis 2012 Laminaria sp. KATM-AM-LAM-2012-01 -15.6 5.0 17.3

2012 Laminaria sp. KATM-AM-LAM-2012-02 -12.6 5.5 18.7 2012 Laminaria sp. KATM-AM-LAM-2012-03 -15.0 6.2 14.3

2012 Laminaria sp. KATM-AM-LAM-2012-04 -12.5 8.5 16.9 2012 Laminaria sp. KATM-AM-LAM-2012-05 -13.4 5.0 23.4

2012 Laminaria sp. KATM-AM-LAM-2012-06 -13.2 8.2 15.5 2012 Laminaria sp. KATM-AM-LAM-2012-07 -13.7 8.4 18.8 2012 Laminaria sp. KATM-AM-LAM-2012-08 -12.9 7.5 18.6

104

2012 Laminaria sp. KATM-AM-LAM-2012-09 -13.5 6.7 16.4

2012 Laminaria sp. KATM-AM-LAM-2012-10 -13.6 7.8 16.0

2012 Neorhodomela sp. KATM-AM-NEO-2012-01 -20.5 7.1 9.5

2012 Neorhodomela sp. KATM-AM-NEO-2012-02 -22.2 7.1 9.0

2012 Neorhodomela sp. KATM-AM-NEO-2012-03 -18.9 7.1 9.3

2012 Neorhodomela sp. KATM-AM-NEO-2012-04 -22.0 6.6 8.4

2012 Neorhodomela sp. KATM-AM-NEO-2012-05 -22.6 6.6 9.6

2012 Neorhodomela sp. KATM-AM-NEO-2012-06 -21.6 7.3 9.2 2012 Ulva sp. KATM-AM-ULV-2012-01 -21.7 7.3 12.5

2012 Ulva sp. KATM-AM-ULV-2012-02 -20.2 7.6 10.6 2012 Ulva sp. KATM-AM-ULV-2012-03 -20.7 7.5 10.0 2012 Ulva sp. KATM-AM-ULV-2012-04 -21.6 7.6 11.0

2012 Ulva sp. KATM-AM-ULV-2012-05 -19.8 7.3 10.2 2012 Ulva sp. KATM-AM-ULV-2012-06 -21.7 8.0 14.7 2012 Ulva sp. KATM-AM-ULV-2012-07 -21.4 7.4 12.4 Offshore Particulate 2012 KATM-KN-POM-2012-OS -21.9 7.8 5.3 Organic Matter

Offshore Particulate 2013 KATM-KK-POM-2013-OS -22.8 2.2 5.2 Organic Matter Offshore Particulate 2013 KATM-KF-POM-2013-OS -20.9 5.3 5.7 Organic Matter Offshore Particulate 2013 KATM-KN-POM-2013-OS -20.2 6.2 5.7 Organic Matter Offshore Particulate 2013 KATM-TK-POM-2013-OS -20.7 5.1 6.2 Organic Matter

105

13 Table 2. Corrected EAA d C values for Amalik Bay producers and consumers. See 13 Appendix 1 for details on the correction of raw EAA d C values and see Methods for sample collection details. POM = Offshore particulate organic matter. Abbreviations for essential amino acids are as follows: isoleucine (Ile), leucine (Leu), lysine (Lys), phenylalanine (Phe), threonine (Thr), valine (Val). For two samples, KATM-AM-LAM- 2012-03 and KATM-AM-ULV-2012-05, we were not able to obtain reliable Phe d13C values. For analysis of these samples we thus used the mean Phe d13C value for other samples in these respective groups. sample identification location codes are as follows: AM = Amalik Bay, KN = Kinak Bay, KK = Kukak Bay, KF = Kaflia Bay and TK = Takli Island. These data are depicted visually in Figure 1.2 Ile Leu Lys Phe Thr Val ID Species d13C d13C d13C d13C d13C d13C

KATM-AM- Laminaria sp. LAM-2012-04 -12.5 -20.8 -6.1 -17.2 1.6 -13.2 KATM-AM- Laminaria sp. LAM-2012-10 -14.9 -21.9 -5.5 -18.3 0.2 -15.4

KATM-AM- Laminaria sp. LAM-2012-03 -17.7 -28.0 -9.7 NA -2.5 -17.8 KATM-AM- Laminaria sp. LAM-2012-06 -14.4 -22.1 -9.8 -19.9 -3.5 -14.3 KATM-AM- Laminaria sp. LAM-2012-01 -18.4 -21.6 -11.7 -19.8 -6.4 -15.8 KATM-AM- Laminaria sp. LAM-2012-05 -8.4 -19.6 -9.8 -17.6 0.7 -13.2 KATM-AM- Laminaria sp. LAM-2012-07 -16.5 -23.7 -10.4 -20.6 -3.2 -18.3 KATM-AM- Laminaria sp. LAM-2012-08 -15.0 -21.8 -9.4 -19.4 -3.1 -15.6

KATM-AM- Laminaria sp. LAM-2012-09 -15.4 -25.7 -8.7 -19.7 -2.9 -18.1 KATM-AM- Neorhodomela sp. NEO-2012-01 -22.9 -27.3 -20.4 -28.2 -17.0 -26.3 KATM-AM- Neorhodomela sp. NEO-2012-03 -21.5 -26.5 -19.5 -27.9 -15.6 -25.7

KATM-AM- Neorhodomela sp. NEO-2012-05 -21.4 -24.0 -18.0 -25.8 -13.0 -24.6

KATM-AM- Neorhodomela sp. -23.2 -26.2 -19.9 -27.7 -15.9 -26.4

106

NEO-2012-02

KATM-KF-POM- POM 2013-OS -19.9 -28.2 -22.2 -27.1 -17.1 -24.1

KATM-KN-POM- POM 2013-OS -19.7 -28.1 -18.5 -28.6 -10.2 -22.0

KATM-TK-POM- POM 2013-OS -18.7 -24.1 -16.9 -23.4 -14.3 -24.6

KATM-KK-POM- POM 2013-OS -24.5 -34.2 -22.4 -31.2 -19.1 -30.5 KATM-AM- Ulva sp. ULV-2012-07 -19.8 -27.2 -17.2 -25.2 -11.4 -22.6 KATM-AM- Ulva sp. ULV-2012-02 -20.3 -27.9 -15.3 -26.4 -13.4 -23.6 KATM-AM- Ulva sp. ULV-2012-05 -19.0 -29.9 -17.0 NA -11.7 -22.7

KATM-AM- Ulva sp. ULV-2012-01 -20.7 -28.2 -19.9 -28.4 -16.6 -22.8

KATM-AM- Ulva sp. ULV-2012-06 -18.6 -26.8 -17.6 -27.5 -13.1 -22.0

KATM-AM- Mytilus sp. MYT-2012-04 -16.6 -26.9 -17.6 -24.8 -9.6 -21.6 KATM-AM- Mytilus sp. MYT-2012-05 -17.9 -27.9 -17.1 -25.2 -10.5 -22.8 KATM-AM- Mytilus sp. MYT-2012-08 -15.1 -25.2 -15.6 -23.0 -6.8 -18.4

KATM-AM- Mytilus sp. MYT-2012-03 -18.5 -27.5 -17.5 -25.3 -9.8 -24.4

KATM-AM- Mytilus sp. MYT-2012-07 -17.1 -28.4 -14.6 -24.6 -8.8 -23.9 KATM-AM- Mytilus sp. MYT-2012-10 -18.2 -27.0 -18.7 -25.3 -10.5 -22.6

KATM-AM- Mytilus sp. MYT-2012-01 -16.2 -27.2 -15.3 -23.8 -8.7 -25.1

KATM-AM- Mytilus sp. MYT-2012-02 -16.4 -26.0 -15.9 -23.8 -8.7 -21.9

107

KATM-AM-STR- Strongylocentrotus 2012-07 droebachiensis -17.4 -25.8 -22.9 -23.3 -9.4 -19.9 KATM-AM-STR- Strongylocentrotus 2012-08 droebachiensis -17.9 -25.9 -23.9 -23.8 -10.7 -20.6

KATM-AM-STR- Strongylocentrotus 2012-10 droebachiensis -18.0 -26.0 -14.9 -23.8 -9.4 -20.4

KATM-AM-STR- Strongylocentrotus 2012-02 droebachiensis -17.5 -24.9 -15.8 -23.4 -9.8 -21.1

KATM-AM-STR- Strongylocentrotus 2012-06 droebachiensis -19.4 -25.8 -17.6 -23.9 -12.4 -20.5

KATM-AM-STR- Strongylocentrotus 2012-05 droebachiensis -18.6 -25.7 -17.4 -23.9 -8.9 -20.6 KATM-AM-STR- Strongylocentrotus 2012-09 droebachiensis -17.8 -26.4 -19.9 -23.0 -8.5 -20.3 KATM-AM- Nucella sp. NUC-2012-03 -15.2 -25.0 -15.9 -22.4 -6.3 -19.6 KATM-AM- Nucella sp. NUC-2012-05 -13.8 -24.4 -13.9 -22.1 -6.0 -18.6

KATM-AM- Nucella sp. NUC-2012-09 -14.7 -25.4 -14.7 -22.4 -7.0 -19.7

KATM-AM- Nucella sp. NUC-2012-02 -14.5 -25.3 -14.7 -22.3 -6.1 -21.2 KATM-AM- Nucella sp. NUC-2012-04 -15.0 -25.8 -11.9 -21.6 -5.0 -25.7 KATM-AM- Nucella sp. NUC-2012-07 -12.8 -24.3 -13.4 -20.6 -6.5 -16.4 KATM-AM- Nucella sp. NUC-2012-01 -13.9 -24.3 -12.8 -21.2 -4.8 -20.5 KATM-AM- Nucella sp. NUC-2012-10 -14.5 -22.6 -14.0 -20.6 -4.7 -16.7

KATM-AM-PYC- Pycnopodia 2012-06 helianthoides -13.6 -23.2 -15.0 -22.2 -3.4 -16.6 KATM-AM-PYC- Pycnopodia 2012-01 helianthoides -14.6 -24.3 -12.8 -22.2 -3.3 -18.8

KATM-AM-PYC- Pycnopodia 2012-07 helianthoides -13.3 -23.6 -11.8 -21.6 -3.4 -18.4

108

KATM-AM-PYC- Pycnopodia 2012-03 helianthoides -14.3 -23.4 -13.9 -22.1 -5.7 -17.5 KATM-AM-PYC- Pycnopodia 2012-09 helianthoides -11.6 -22.2 -10.9 -19.5 -5.3 -15.6

KATM-AM-PYC- Pycnopodia 2012-10 helianthoides -13.1 -22.9 -12.8 -20.6 -5.1 -16.4

KATM-AM-PYC- Pycnopodia 2012-02 helianthoides -11.4 -20.0 -11.9 -20.0 0.2 -14.9

KATM-AM-PYC- Pycnopodia 2012-04 helianthoides -12.2 -21.7 -13.6 -20.7 -0.5 -16.3

109

Table 3. Data from Larsen et al. 2013 used here in comparison with sampled Amalik Bay producers. The column ‘ID’ corresponds exactly to those ID’s used in the Larsen et al. 2013 paper. An asterisk (*) before a location indicates that the individual was a laboratory culture sample. Abbreviations for essential amino acids are as follows: isoleucine (Ile), leucine (Leu), lysine (Lys), phenylalanine (Phe), threonine (Thr), valine (Val).

Type Species Location ID Ile Leu Lys Phe Thr Val d13C d13C d13C d13C d13C d13C

Rhodophyta Priomtis sp. Santa Cruz, R1 -17.3 -20.2 -10.9 -19.2 -6.4 -19.2 CA

Rhodophyta Osmundea Santa Cruz, R2 -15.4 -19.5 -13 -20.7 -4.8 -21.3 spectablilis CA

Rhodophyta Chondracanthus Santa Cruz, R3 -21.7 -25.8 -16.4 -25.3 -9.2 -25.4 canaliculatus CA

Rhodophyta Calliarthon sp. Santa Cruz, R4 -17 -21.9 -14.9 -23.7 -11 -19.8 CA

Rhodophyta Corallina sp. Santa Cruz, R5 -19.5 -22.5 -16.4 -25.3 -15.3 -22.4 CA

Rhodophyta Odonthalia Santa Cruz, R6 -16.5 -21 -13.1 -21.1 -8.7 -20.7 floccosa CA

Rhodophyta Mastocarpus sp. Santa Cruz, R7 -16.7 -21 -11 -21.1 -5.9 -20.3 CA

Rhodophyta Endocladia Santa Cruz, R8 -18 -23.5 -15.8 -22 -7.3 -23.6 muricata CA

Rhodophyta Mazzaella Santa Cruz, R9 -20.4 -23.4 -13.3 -22.5 6.8 -20.9 flaccida CA

Phaeophyceae Macrocystis Big Sur, P1 -17.6 -23.1 -11.1 -19.4 -3.6 -18.8 pyrifera CA

Phaeophyceae Macrocystis Big Sur, P2 -15.2 -21.5 -9.1 -19.1 -1.2 -17 pyrifera CA

Phaeophyceae Macrocystis Big Sur, P3 -19.6 -24.8 -12.7 -22.2 -5.5 -20.7 pyrifera CA

Phaeophyceae Macrocystis Big Sur, P4 -16.8 -20 -9.9 -18.9 -0.4 -15.7 pyrifera CA

Phaeophyceae Macrocystis Big Sur, P5 -21.4 -26.1 -14.1 -23.9 -6.4 -22.5 pyrifera CA

Phaeophyceae Scytosiphon sp. Santa Cruz, P6 -6.3 -13.3 -1.7 -12.8 -2.6 -10.4 CA

Phaeophyceae Laminaria sp. Santa Cruz, P7 -15.6 -23.9 -9.2 -18.8 -1.2 -19.1 CA

Phaeophyceae Silvetia sp. Santa Cruz, P8 -19.5 -25.5 -10.7 -20.4 -0.5 -19.6 CA

110

Phaeophyceae Petrospongium Santa Cruz, P9 -10.9 -18.3 -8.2 -16.7 0.9 -14.5 sp. CA

Phaeophyceae Pelvetiopsis sp. Santa Cruz, P1 -16.9 -23.6 -10.8 -23.6 5.4 -21.2 CA 0

Phaeophyceae Ralfsia sp. Santa Cruz, P1 -7.3 -14.7 -4.5 -12.7 0.8 -10.9 CA 1

Phaeophyceae Cystoseira Santa Cruz, P1 -17.6 -25.2 -9.1 -21.2 -8.6 -22.4 osmundacea CA 2

Microalgae Cyanothece sp. *GEOMA C1 -27.7 -27.8 -18 -24.9 -10.4 -24 R

Microalgae Merismopedia *GEOMA C2 -28.9 -36 -23.8 -34.6 -19.5 -32.8 punctata R

Microalgae Anabaena *SAG C3 -17.3 -23.3 -13.4 -19.8 -5.9 -20.5 cylindrica

Microalgae Nostoc *SAG C4 -20 -25.8 -13.8 -24 -10.1 -24.2 muscorum

Microalgae Adhnanthes *GEOMA D1 -10.2 -17.3 -8.8 -18.1 -2.7 -13.8 brevipes R

Microalgae Amphora *GEOMA D2 -11.5 -18.7 -11.1 -19.6 -2 -16.8 coffaeiformis R

Microalgae Melosia varians *GEOMA D3 -16.5 -22.1 -15.8 -20.6 -7 -19.7 R

Microalgae Phaeodactylum *SAG D4 -16.1 -25.5 -15.2 -23.8 -10.8 -22.9 tricornutum

Microalgae Stauroneis *GEOMA D5 -10.4 -19.3 -8.5 -16 -4.6 -15.5 constricta R

Microalgae Emiliana huxleyi *GEOMA H1 -18.3 -27.2 -15.7 -23.2 -9.1 -24.9 R

Microalgae Emiliana huxleyi *GEOMA H2 -19.2 -23.8 -14.5 -24.6 -8.6 -21.1 R

Microalgae Isochyrsis *GEOMA H3 -18.3 -31.4 -15.4 -25 -8.9 -26.8 galbana R

Microalgae Corcontochrysis *SAG H4 -13.8 -20.5 -12 -21.6 -6.8 -18.8 noctivaga

Microalgae Dunaliella sp. *GEOMA K1 -15.8 -23.7 -13 -19.7 -6.7 -19.2 R

Microalgae Prasinocladus *GEOMA K2 -15.3 -23.3 -12.2 -21.4 -2.2 -20.6 marinus R

Microalgae Ankistrodesmus *SAG K3 -10.4 -15.7 -5.2 -15.9 -8.1 -13 falcatus

Microalgae Chlamydocapsa *SAG K4 -11.4 -16.7 -7.7 -15.7 -4.2 -15.3

111

maxima

Microalgae Chlamydomonas *SAG K5 -16.3 -23.2 -13.9 -22.2 -6.4 -20.4 asymmetrica

Microalgae Chlamydomonas *SAG K6 -11.4 -18.4 -10.5 -18.4 -7.1 -15.9 gigantea

Microalgae Composite Kiel fjord N1 -21 -29.7 -18.1 -27 -11 -26.4 natural sample

Microalgae Composite Kiel fjord N2 -20.6 -29.3 -18 -26.8 -11.2 -27.2 natural sample

Microalgae Composite Kiel fjord N3 -19.8 -29.4 -17.9 -26.9 -10.7 -27.1 natural sample

Microalgae Ochromonas *GEOMA X1 -26.7 -35.1 -20.7 -31.9 -13.6 -30.2 minima R

Microalgae Ochromonas *GEOMA X2 -22.8 -32.7 -19.2 -30 -10.7 -28 villosa R

Microalgae Ochromonas *SAG X3 -19.3 -25.2 -16.4 -25.5 -9.6 -23.9 danica

Microalgae Chromulina sp. *SAG X4 -11.1 -18.2 -6.2 -16.1 -1.3 -14.4

Microalgae Cryptomonas sp. *SAG Y1 -11.8 -19.3 -11.4 -19.5 -4.2 -17.1

Plants Carex utriculata North T9 -27.8 -35.4 -24.3 -28.9 -15.4 -32.6 Slope, AK

Plants Rumex arcticus North T1 -27.6 -37.2 -25.6 -29.9 -18.1 -34.1 Slope, AK 2

Plants Menyanthes North T7 -27.5 -33.6 -23.5 -28.7 -12.6 -32.3 trfoliata Slope, AK

Plants Polygonum North T4 -27.4 -35.3 -23.9 -28.8 -15.8 -33.3 viviparum Slope, AK

Plants Calamagrostis North T6 -27.1 -34.4 -24.9 -28.9 -9.9 -32 canadensis Slope, AK

Plants Carex aquatilis North T5 -26.8 -33.5 -22.5 -27.7 -15.3 -31.6 Slope, AK

Plants Betula nana North T8 -25.2 -35.2 -23.9 -27.8 -8.6 -30.6 Slope, AK

Plants Salix reticulata North T1 -25.1 -34.5 -23.2 -26.4 -11.5 -30.5 Slope, AK 0

Plants Salix sp North T3 -24.2 -34.4 -22.5 -27 -12.6 -31.4 Slope, AK

Plants Eriophorum North T1 -23.2 -33 -21.8 -27 -4.4 -30 angustifolium Slope, AK 1

112

Table 4. ANOVA with posthoc Tukey’s HSD results for pairwise comparisons of producer and invertebrate bulk tissue d13C and d 15N values. Rows are pairwise comparisons, columns are d13C and d15N values. Asterisks (*) indicate significant differences (adjusted p-value of <0.05) between pairs for a given amino acid. Species codes are as follows: LAM = Laminaria sp., NEO = Neorhodomela sp., ULV = Ulva sp., POM = Offshore particulate organic matter, here we grouped 2012 and 2013, MYT = Mytlius sp., STR = Strongylocentrotus droebachiensis, NUC = Nucella sp., PYC = Pycnopodia helianthoides.

13 15 d CBULK d NBULK NEO-LAM 0.0000000* 0.9990088

ULV-LAM 0.0000000* 0.6977763 POM-LAM 0.0000000* 0.1118942 ULV-NEO 0.9605867 0.8359252 POM-NEO 1.0000000 0.1370129 POM-ULV 0.9605867 0.0220675*

STR-MYT 0.9983623 0.0987794 NUC-MYT 0.0002361* 0.0000002* PYC-MYT 0.0000000* 0.0000000* NUC-STR 0.0003718* 0.0000000* PYC-STR 0.0000000* 0.0000000* PYC-NUC 0.0000001* 0.0000001*

113

Table 5. ANOVA with posthoc Tukey’s HSD results for pairwise comparisons of 13 producer and invertebrate EAA d C values. Rows are pairwise comparisons, columns are essential amino acids. Asterisks (*) indicate significant differences (adjusted p-value of <0.05) between pairs for a given amino acid. Species codes are as follows: LAM = Laminaria sp., NEO = Neorhodomela sp., ULV = Ulva sp., POM = Offshore particulate organic matter from 2013, MYT = Mytlius sp., STR = Strongylocentrotus droebachiensis, NUC = Nucella sp., PYC = Pycnopodia helianthoides. Ile Leu Lys Phe Thr Val NEO- 0.0002178* 0.1935537 0.0000003* 0.0000009* 0.0000006* 0.0000010* LAM ULV- 0.0068342* 0.0098384* 0.0000023* 0.0000008* 0.0000024* 0.0000431* LAM POM- 0.0023419* 0.0067482* 0.0000001* 0.0000007* 0.0000007* 0.0000019* LAM ULV- 0.3609212 0.6774403 0.4260327 0.9650634 0.6115888 0.1649379 NEO POM- 0.7817183 0.4920576 0.9799485 0.9990190 0.9995505 0.9905828 NEO POM- 0.9006027 0.9792796 0.2379307 0.9262639 0.6786620 0.2739137 ULV

STR- 0.1649969 0.1100096 0.1157629 0.1911396 0.7947597 0.1847146 MYT NUC- 0.0000500* 0.0004022* 0.0647285 0.0000013* 0.0005038* 0.0392593* MYT PYC- 0.0000001* 0.0000000* 0.0051780* 0.0000000* 0.0000001* 0.0000155* MYT NUC- 0.0000003* 0.1606033 0.0002466* 0.0005161* 0.0000665* 0.9081965 STR PYC- 0.0000000* 0.0000166* 0.0000158* 0.0000177* 0.0000000* 0.0061517* STR PYC- 0.0692654 0.0034595* 0.7074410 0.5600096 0.0119644* 0.0241510* NUC

114

Table 6. EAA d13C MIXSIAR results using essential amino acid (EAA) d13C data; these data are depicted visually in Figure 1.3. Bold values are the median contribution of each producer group to the consumer species, with associated 95% credibility intervals in brackets. The model was run independently for each consumer species at length ‘Very Long’ at which point diagnostics indicated the MCMC chains had largely converged. Trophic discrimination factors for all EAA were assumed to be 0‰. Because Ulva and POM were not well differentiated by canonical analysis of principal coordinates (CAP), we lumped the proportions from each of these groups a posteriori. Proportion Proportion Proportion Proportion Proportion Consumer Laminaria Neorhodomela POM Ulva (Ulva + POM) 0.33 [0.24, 0.10 [0.00, 0.23 [0.03, 0.32 [0.04, 0.57 [0.29, Mytilus 0.41] 0.35] 0.46] 0.62] 0.72] 0.38 [0.31, 0.13 [0.01, 0.07 [0.00, 0.40 [0.10, 0.51 [0.22, Strongylocentrotus 0.46] 0.35] 0.26] 0.62] 0.63] 0.66 [0.59, 0.08 [0.00, 0.11 [0.01, 0.13 [0.01, 0.26 [0.08, Nucella 0.73] 0.25] 0.28] 0.31] 0.38] 0.79 [0.72, 0.06 [0.00, 0.05 [0.00, 0.07 [0.00, 0.15 [0.03, Pycnopodia 0.87] 0.17] 0.17] 0.21] 0.26]

115

Table 7. Comparison of Amalik producer and consumer EAA d13C ‘fingerprints’ with those from Larsen et al. (2013). Producer samples from this study were grouped with appropriate samples from Larsen et al. (e.g., Laminaria with Phaeophyceae); these data are depicted visually in Appendix 1 Figure 1.

CAP Classification Chlorophyta % Phaeophyceae Rhodophyta Microalgae (Ulva) Correct

Phaeophyceae 18 1 1 1 86

Rhodophyta 1 11 0 1 85

Chlorophyta 0 0 3 2 60 (Ulva)

Microalgae 2 5 6 18 58

Combined Producer Groups

116

Table 8. MIXSIAR results using bulk tissue d13C and d15N data. Bold values are the median contribution of each producer group to the consumer species, with associated 95% credibility intervals in brackets. The model was run independently for each consumer species at length ‘Long’ at which point diagnostics indicated the MCMC chains had largely converged. Trophic discrimination factors were assumed to be 1.0 ± 0.5‰ for d13C and 3.0 ± 0.5‰ d15N. Mytilus and Strongylocentrotus were assumed to be at a trophic level of 1, Nucella at a trophic level of 2 and Pycnopodia at a trophic level of 3.

Proportion Proportion Proportion Proportion Consumer Laminaria Neorhodomela POM Ulva

0.52 [0.34, 0.09 [0.01, Mytilus 0.22 [0.16, 0.33] 0.11 [0.00, 0.34] 0.71] 0.30]

0.64 [0.48, 0.05 [0.00, Strongylocentrotus 0.22 [0.07, 0.34] 0.06 [0.00, 0.23] 0.82] 0.21] 0.44 [0.27, 0.03 [0.00, Nucella 0.34 [0.22, 0.44] 0.03 [0.00, 0.27] 0.68] 0.25] 0.28 [0.04, 0.05 [0.00, Pycnopodia 0.59 [0.50, 0.67] 0.05 [0.00, 0.25] 0.43] 0.24]

117

0.25 AK Browns Larsen Browns 0.2 AK Greens AK Reds 0.15 Larsen Reds AK POM Larsen Microalgae 0.1

0.05 2 0 CAP

-0.05

-0.1

-0.15

-0.2

-0.25 -0.35 -0.25 -0.15 -0.05 0.05 0.15 0.25

CAP1

Figure 1. Essential amino acid d13C fingerprinting of Amalik Bay and Larsen et al. (2013) producers. Shown are the coordinates of producers in multivariate space as calculated using canonical analysis of principal coordinates (CAP). Circles represent the CAP loading of individual producer samples, open circles are samples from Amalik Bay, filled circles are from Larsen et al. (2013): brown algae (black), green algae (green), red algae (grey), and microalgae/POM (blue). Among our grouped producer fingerprints, CAP found an overall successful reclassification rate of 71%.

118

Appendix 2

1. Sample preparation and analysis for bulk and amino acid d15N.

For both bulk and amino acid analysis we report all isotopic results as d values: d13C or d15N = 1000*[(Rsamp/Rstd) − 1], where Rsamp and Rstd are the 13C: 12C or

15N:14N ratios of the sample and standard, respectively. The internationally accepted standards for d13C and d15N analysis are Vienna-Pee Dee Belemnite limestone (V-PDB) and atmospheric N2. The units are expressed as parts per thousand, or per mil (‰). For bulk analysis, d13C and ︎d15N values of all samples were analyzed using a Costech 4010 elemental analyzer (Valencia, CA) interfaced with a Thermo Scientific Delta V Plus isotope ratio mass spectrometer (Bremen, Germany) at the University of New Mexico

Center for Stable Isotopes (UNM-CSI, Albuquerque, NM). The standard deviation of organic reference materials within a run was ≤0.2‰ for both d13C and d15N values. As a control for the quality of our ancient collagen samples, we also measured [C]:[N] ratios.

The theoretical weight percent [C]:[N] of unaltered bone collagen falls between 2.8-3.5

(Ambrose 1990), all but two samples fell within this threshold (Appendix 2, Table 2). For all analyses using modern vibrissae, we used the mean d13C and ︎d15N values across all subsampled segments from an individual.

For AA d15N analysis, ~5-10mg of extracted collagen was hydrolyzed to constituent amino acids in 1mL of 6N hydrochloric acid (HCl) at 110°C for 20 hours; tubes were flushed with N2 to prevent oxidation during hydrolysis. After hydrolysis, amino acids

119 were derivatized to N- trifluoroacetic acid isopropyl esters following established protocols

(Whiteman et al. 2019). Samples were derivatized in batches of 8–17 along with an in- house reference material containing all amino acids we measured. d15N values of individual derivatized amino acids were measured using a GC-C-IRMS system at UNM-CSI.

Derivatized samples were injected into a 60m BPx5 gas chromatograph column for amino acid separation (0.32 ID, 1.0μm film thickness, SGE Analytical Science, Victoria,

Australia) in a Thermo Scientific Trace 1300, combusted into CO2 gas via a Thermo

Scientific GC Isolink II, and analyzed with a Thermo Scientific Delta V Plus. Samples were run in duplicate or triplicate and bracketed with our reference material. AA data for a sample were accepted if the standard deviations was <1.0‰ across injections. The within- run standard deviations of measured d15N values among AA in the in-house reference material ranged from 1.7‰ (Glycine) to 0.1‰ (Aspartic Acid). Due to peak coelution, we were unable to reliably distinguish between hydroxyproline (Hyp) and proline (Pro), thus these data are presented as a pooled isotope value (Hyp-Pro).

120

2. Tissue specific and temporal isotope corrections.

A one-way ANOVA of different tissues from stranded sea otters indicated systematic differences in d13C between bone collagen and vibrissae: otter bone collagen had higher mean d13C values by +0.9 ± 0.4‰ ([F (3, 112) = 160, p < 0.001]; Appendix 2,

Table 3). Thus, we added +1.0‰ to modern sea otter vibrissae d13C values so these data could be directly compared to bone collagen. We found no significant difference in d15N between bones and whiskers (Appendix 2, Tables 1 and 3). In addition, we applied a conservative correction of -1.5‰ to ancient sea otter d13C values to account for the

13 decrease in atmospheric CO2 d C over the past 150 years driven by fossil fuel combustion

(i.e., the Suess effect; Cullen et al. 2001). There is no known temporal correction for d15N.

121

Table 1. Isotopic results of paired tissues from stranded sea otters in central California. Vibrissae isotope values are averages across subsamples; N ranged from 7 to 25 subsamples per individual. Acronyms for regions are as follows: SCC = Santa Cruz County, MBY = Monterey Bay, SLO = San Louis Obispo. All tissue samples of stranded sea otters were collected through the California Sea Otter Stranding Network, a multi-agency program coordinated by the California Department of Fish and Wildlife (CDFW) and USGS. For bone data, samples were taken from skulls curated at the California Academy of Sciences (CAS); CAS identifiers are included.

CAS Otter Stranding Bone Muscle Liver Vibrissae Bone Muscle Liver Vibrissae Bone Muscle Liver Vibrissa Region Sex ID ID Site d15N d15N d15N d15N d13C d13C d13C d13C C:N C:N C:N e C:N

SO Salinas MAM 3110- MBY River State M 15.2 16.1 15.6 15.3±1.3 -12.2 -16.4 -15.0 -12.6±1.5 2.7 3.2 3.2 2.9 30425 98 Beach

SO MAM Pacific 4800- MBY F 13.6 14.7 15.3 13.9±0.4 -11.4 -15.4 -14.0 -12.6±0.4 2.7 3.2 3.1 2.9 30417 Grove 06

SO MAM Del Monte 4833- MBY F 16.2 17.9 17.0 17.1±0.3 -11.2 -14.4 -14.0 -11.9±0.2 2.7 3.2 3.2 2.9 30415 Beach 06

SO MAM San Simeon 4910- SLO F 12.3 14.5 14.6 12.8±0.8 -12.1 -15.3 -14.1 -12.9±0.5 2.7 3.2 3.2 2.8 30435 State Beach 07

SO MAM 4923- MBY Otter Point M 14.7 16.0 15.6 14.5±0.8 -11.7 -14.4 -14.2 -12.5±0.4 2.7 3.1 3.2 2.9 30416 07

SO MAM Asilomar 4981- MBY M 13.6 15.3 14.4 13.31.6 -12.5 -15.8 -14.9 -13.6±0.7 2.7 3.2 3.2 2.9 30418 State Beach 07

SO MAM Cannery 5054- MBY F 14.2 15.1 15.3 14.3±1.5 -11.9 -14.8 -13.9 -12.7±0.4 2.7 3.2 3.2 2.9 30421 Row 07

SO MAM 5167- SLO San Simeon F 13.6 14.6 14.7 14.0±0.6 -11.8 -15.8 -14.2 -12.8±0.3 2.7 3.2 3.2 2.9 30436 07

SO MAM Del Monte 5204- MBY F 13.1 14.6 14.3 13.4±0.2 -10.6 -15.0 -13.2 -11.9±0.8 2.7 3.2 3.1 2.9 30437 Beach 08

122

SO MAM Moss 5274- MBY M 14.0 14.5 15.0 13.9±0.3 -10.4 -14.0 -13.7 -11.8±0.5 2.7 3.2 3.2 2.9 30430 Landing 08

SO MAM Monterey 5384- MBY F 13.7 15.4 15.7 14.9±0.4 -11.1 -13.8 -13.0 -11.9±0.3 2.7 3.1 3.2 2.9 30431 Bay Inn 08

SO MAM Wild Cattle 5423- MBY F 12.8 14.5 14.6 13.3±0.9 -12.5 -16.2 -14.6 -13.6±0.7 2.7 3.2 3.1 2.8 30433 Creek 08

SO MAM Moss 5774- MBY M 13.3 15.5 15.0 14.4±0.4 -11.3 -13.9 -13.5 -11.8±0.1 2.7 3.2 3.2 2.9 30434 Landing 10

SO MAM Hopkins 5808- MBY F 11.9 14.3 14.6 13.6±1.2 -11.3 -15.0 -13.7 -12.4±0.6 2.7 3.2 3.2 2.9 30419 Beach 10

SO MAM Pebble 5818- MBY F 14.6 15.5 15.1 15.5±0.9 -12.0 -16.0 -15.3 -13.2±0.3 2.7 3.2 3.2 2.9 30396 Beach 10

SO MAM Anderson 6207- SCC F 13.9 14.3 13.8 12.9±0.3 -11.7 -15.8 -15.0 -13.1±0.3 2.7 3.2 3.2 2.9 30438 Canyon 11

SO MAM Del Monte 6234- MBY F 13.1 15.1 15.4 14.0±0.4 -11.5 -14.6 -13.3 -12.1±0.5 2.7 3.2 3.2 2.8 30420 Beach 11

SO MAM 6678- SLO Cayucos F 16.1 17.0 16.7 15.6±0.5 -12.0 -14.7 -13.6 -12.6±0.6 2.7 3.2 3.1 2.8 30422 13

SO MAM Sea Palm 6690- MBY F 14.1 15.5 15.0 14.6±0.4 -10.6 -14.3 -13.8 -12.3±0.4 2.7 3.2 3.2 2.9 30423 Street 13

SO MAM Pismo 6743- SLO F 15.2 15.1 15.0 16.7±0.5 -12.0 -14.3 -13.6 -12.2±0.5 2.8 3.2 3.1 2.9 30401 Beach 13

SO MAM Elkhorn 6746- MBY M 13.8 17.4 17.0 14.8±0.3 -11.2 -14.0 -13.5 -12.3±0.3 2.7 3.1 3.2 2.8 30402 Slough 13

123

SO Hopkins MAM 6878- MBY Marine F 15.9 17.1 16.7 16.5±0.2 -11.5 -15.5 -15.0 -12.4±0.2 2.7 3.2 3.2 2.8 30426 13 Station

SO MAM Jalama 6990- SBC F 15.4 16.5 16.6 15.8±0.7 -11.4 -15.5 -14.2 -12.9±0.3 2.7 3.2 3.1 2.9 30428 Beach 13

SO MAM Del Monte 7060- MBY F 15.4 15.9 15.7 15.0±0.2 -11.7 -15.9 -15.0 -12.8±0.4 2.8 3.2 3.2 2.8 30439 Beach 13

SO MAM 7064- MBY Point Pinos F 13.0 14.8 13.8 13.8±0.3 -13.3 -15.3 -14.9 -13.0±1.0 2.7 3.2 3.3 2.8 30440 14

SO MAM Del Monte 7089- MBY F 14.6 16.1 16.3 14.9±0.5 -11.7 -15.0 -13.7 -12.5±0.3 2.7 3.2 3.2 2.8 30441 Beach 14

SO MAM Moss 7105- MBY M 14.7 16.7 16.2 15.7±0.4 -11.5 -14.5 -13.6 -12.5±0.5 2.7 3.2 3.2 2.9 30442 Landing 14

SO MAM Elkhorn 7112- MBY F 14.3 16.5 16.0 14.9±0.8 -12.9 -15.3 -14.6 -12.8±1.3 2.7 3.2 3.2 2.8 30443 Slough 14

124

Table 2. Ancient and modern sea otter samples. Data from CA-SLO-2 from Jones et al. (2011). Ancient samples represent bones identified from archaeological sites, modern samples represent sea otter vibrissae. For a specific archaeological site, where applicable/available the exact details on sample location during excavation are provided. See Methods for details on archaeological sample identification, and the preparation and analysis for all samples. For column “d13C Corr” archaeological d13C values were corrected for the Suess Effect by -1.5‰, and modern sea otter d13C values were corrected for tissue specific discrimination (see Appendix 2, Table 3) by +1.0‰. d15N values did not need temporal or tissue specific corrections. Acronyms for regions are as follows: ANO = Año Nuevo, MBY = Monterey Bay, BSR = Big Sur Reserve, SLO = San Louis Obispo, SBC = Santa Barbara Channel, SMI = San Miguel Island, SNI = San Nicolas Island. For MNT-831 we were unable to confidently quantify the number of individuals due to poor sample provenance and excluded these samples from a number of subsequent analyses (see Methods). Note that columns ‘Level (cm)’ and ‘Depth (cm)’ record the same information, but we kept these data consistent with the nomenclature from the original researcher.

Level d13C d13C d15N Depth Sample Type Region Site Sample ID [C]:[N] Unit Loci Stratum (cm) Raw Corr Raw (cm)

Archaeological ANO SMA-238 SO-SMA-343-1 -13.8 -15.3 15.3 2.8 B NA NA NA 0-20

Archaeological ANO SMA-238 SO-SMA-270 -12.6 -14.1 15.4 3.2 4 NA NA NA 0-Sterile

Archaeological ANO SMA-238 SO-SMA-252 -11.9 -13.4 15.6 3.0 NA NA NA

Archaeological ANO SMA-238 SO-SMA-146 -11.1 -12.6 15.9 3.2 2 NA NA NA 0-Sterile

Archaeological ANO SMA-238 SO-SMA-316 -11.4 -12.9 15.2 2.8 B NA NA NA 0-20

Archaeological ANO SMA-238 SO-SMA-739 -10.4 -11.9 15.3 3.1 B NA NA NA 0-20

Archaeological ANO SMA-238 SO-SMA-168 -13.4 -14.9 17 3.0 11 NA NA NA 0-Sterile

Archaeological ANO SMA-238 SO-SMA-751 -11.5 -13 16.8 2.8 D NA NA NA Level 1

Archaeological ANO SMA-238 SO-SMA-706 -11 -12.5 16.8 2.8 B NA NA NA 0-20

Archaeological ANO SMA-238 SO-SMA-315 -9.7 -11.2 17 2.9 B NA NA NA 0-20

Archaeological ANO SMA-238 SO-SMA-857 -14.1 -15.6 17.7 3.2 B NA NA NA Surface

Archaeological ANO SMA-238 SO-SMA-676-7 -12.8 -14.3 17.6 3.0 6 NA NA NA Surface

Archaeological SMI SMI-602 SO-SMI 10663 -9.6 -11.1 17.9 2.8 5 NA 40-50 NA

Archaeological SMI SMI-602 SO-SMI-9702 -9.5 -11 13.8 3.2 5 NA 40-50 NA

125

Archaeological SMI SMI-602 SO-SMI-10664 -7.3 -8.8 13.5 2.9 2 NA E NA

Archaeological SMI SMI-525 SO-SMI-12010 -12.9 -14.4 15.4 2.8 Profile D NA 18-19 NA

Archaeological SMI SMI-525 SO-SMI-12334 -9.8 -11.3 15.3 3.2 Profile D NA 2-4 NA

Archaeological SMI SMI-525 SO-SMI-12404 -8.8 -10.3 14.4 2.9 1-C NA 210-220 4 NA

Archaeological SMI SMI-525 SO-SMI-12017 -8.7 -10.2 14.5 2.9 Profile D NA 18-19 NA

Archaeological SMI SMI-525 SO-SMI-12097 -11.5 -13 13.9 2.8 Profile D NA Upper 20 18-19 NA

Archaeological SMI SMI-525 SO-SMI-12343 -10.3 -11.8 13.9 2.7 Profile D NA 2-4 NA

Archaeological SMI SMI-525 SO-SMI-12405 -10.2 -11.7 13.5 2.7 1-C NA 210-220 4 NA

Archaeological SMI SMI-525 SO-SMI-12414 -9.6 -11.1 13.1 3.0 Profile D NA NA

Archaeological SMI SMI-525 SO-SMI-10717 -9.1 -10.6 13.2 2.9 1-B NA 220-230 4 NA

Archaeological SMI SMI-525 SO-SMI-12406 -7.5 -9 14.1 3.1 1-C NA 210-220 4 NA

Archaeological SMI SMI-525 SO-SMI-12402 -7.9 -9.4 13.8 3.0 1-C NA 220-230 4 NA

Archaeological SMI SMI-525 SO-SMI-12421 -12.3 -13.8 12.5 2.9 Profile D NA NA

Archaeological SMI SMI-528 SO-SMI-10629 -11.3 -12.8 17.9 2.7 1 NA 10-20 I NA

Archaeological SMI SMI-528 SO-SMI-10639 -10.4 -11.9 15.7 2.8 1 NA 40-50 I NA

Archaeological SMI SMI-528 SO-SMI-10618 -10.8 -12.3 16.4 2.8 2 NA 20-30 I NA

Archaeological SMI SMI-528 SO-SMI-10630 -8.7 -10.2 15.9 3.1 2 NA 50-60 I NA

Archaeological SMI SMI-528 SO-SMI-10625 -11.5 -13 13.5 3.2 2 NA 20-30 I NA

Archaeological SMI SMI-528 SO-SMI-10636 -9.7 -11.2 14.2 3.0 1 NA 10-20 I NA

Archaeological SMI SMI-528 SO-SMI-10641 -8.6 -10.1 13.8 2.7 2 NA 10-20 I NA

Archaeological SMI SMI-528 SO-SMI-10632 -12.8 -14.3 12 3.0 2 NA 10-20 I NA

Archaeological SMI SMI-528 SO-SMI-10635 -10.3 -11.8 12.4 2.9 1 NA 20-30 I NA

Archaeological SMI SMI-528 SO-SMI-10626 -9.4 -10.9 12.9 3.1 2 NA 40-50 I NA

Archaeological SMI SMI-528 SO-SMI-10624 -8.4 -9.9 12.9 2.9 1 NA 0-10 I NA

Archaeological SMI SMI-1 SO-SMI-11062 -11.1 -12.6 13.7 2.9 596 NA 6-12 NA

126

Archaeological SMI SMI-1 SO-SMI-11063 -10.7 -12.2 14.6 3.0 206 NA 18-20 NA

Archaeological SMI SMI-1 SO-SMI-11065 -10.1 -11.6 14.4 2.8 824 NA 12-18 NA

Archaeological SMI SMI-1 SO-SMI-11070 -12.3 -13.8 15.9 3.1 567 NA 12-18 NA

Archaeological SMI SMI-1 SO-SMI-11080 -8.8 -10.3 14.7 3.1 387 NA 12-18 NA

Archaeological SMI SMI-1 SO-SMI-11077 -7.4 -8.9 14.4 2.7 342 NA 12-18 NA

Archaeological SMI SMI-1 SO-SMI-11064 -9.5 -11 16.9 2.8 824 NA 12-18 NA

Archaeological SMI SMI-1 SO-SMI-11067 -9.5 -11 17 3.0 737 NA 18-24 NA

Archaeological SNI SNI-011 SO-SNI-003 -14.1 -15.6 17.3 2.8 10.5s/54w Mound A 30-40

Archaeological SNI SNI-011 SO-SNI-002 -12.7 -14.2 16.8 2.8 10.5s/54w Mound A 110-120

Archaeological SNI SNI-011 SO-SNI-006 -8.3 -9.8 15.3 2.8 1.5s/58.5s Mound B 60-70

Archaeological SNI SNI-011 SO-SNI-001 -12.9 -14.4 14.8 2.7 3s/1.5e Mound A 70-80

Archaeological SNI SNI-011 SO-SNI-004 -9.3 -10.8 15.2 2.9 28.5n/9e Mound B 40-50

Archaeological SNI SNI-025 SO-SNI-015 -9.4 -10.9 13.6 2.8 62 South I/3

Archaeological SNI SNI-025 SO-SNI-016 -10.1 -11.6 14.5 2.9 42 West I/2

Archaeological SNI SNI-025 SO-SNI-017 -11.2 -12.7 17 2.8 44 West 40-50

Archaeological SNI SNI-025 SO-SNI-020 -7.6 -9.1 14.2 2.8 7X (Pit A) East II/4

Archaeological SNI SNI-025 EL-SNI-025-002 -9.2 -10.7 14.5 2.9 48 NA 4 II/4

EL-SNI-025-8T- Archaeological SNI SNI-025 -13.7 -15.2 16.7 2.9 8T East South Pit B PB

Archaeological SNI SNI-025 EL-SNI-025-7H-2 -13 -14.5 18.1 2.8 7H East 1 two

Archaeological SNI SNI-025 EL-SNI-025-8X-2 -10.1 -11.6 16.6 2.9 8X East 3 two

Archaeological SNI SNI-040 SO-SNI-008 -9.1 -10.6 14.5 2.8 6 C IIIO/1

Archaeological SNI SNI-040 SO-SNI-009 -12.3 -13.8 15.9 2.9 10A E IIHC-SU/2

Archaeological SNI SNI-040 SO-SNI-010 -12 -13.5 14.8 2.8 11 E II/3

Archaeological SNI SNI-040 SO-SNI-011 -11.6 -13.1 14.8 2.8 10 E II/1

127

Archaeological SNI SNI-040 SO-SNI-012 -8.7 -10.2 14.9 2.8 3a B II

Archaeological SNI SNI-040 SO-SNI-013 -9.6 -11.1 14.6 2.8 2 A II/1

Archaeological SNI SNI-040 EL-SNI-040-004 -11.1 -12.6 15.1 2.8

Archaeological SNI SNI-040 EL-SNI-040-001 -9.1 -10.6 14.9 2.7 6 C 1 IIIO

Archaeological MBY MNT-831 MNT-831-EL-018 -11.9 -13.4 15.3 2.9

Archaeological MBY MNT-831 MNT-831-EL-005 -11.8 -13.3 15.5 2.7

Archaeological MBY MNT-831 MNT-831-EL-014 -10.3 -11.8 15.6 2.7

Archaeological MBY MNT-831 MNT-831-EL-009 -10.3 -11.8 13.7 2.8

Archaeological MBY MNT-831 MNT-831-EL-002 -9.2 -10.7 16.2 2.8

Archaeological MBY MNT-234 EL-MNT-733 -11.2 -12.7 15.2 2.8 1 110-120

Archaeological MBY MNT-234 EL-MNT-1101 -10.9 -12.4 13.9 3.6 Backhoe

Archaeological MBY MNT-234 EL-MNT-1384 -10.4 -11.9 13.7 2.7 1 160-170

Archaeological MBY MNT-234 EL-MNT-1415 -10.4 -11.9 14.1 2.7 4 050-060

Archaeological MBY MNT-234 EL-MNT-2395 -10.8 -12.3 15.1 2.9 1 200-210

Archaeological MBY MNT-234 EL-MNT-2424 -10.7 -12.2 14.5 2.9 1 200-210

Archaeological MBY MNT-234 EL-MNT-2425 -11.1 -12.6 14.4 2.9 1 200-210

Archaeological MBY MNT-234 EL-MNT-2915 -10.6 -12.1 14.8 2.8 1 220-230

Archaeological MBY MNT-234 EL-MNT-2977 -10 -11.5 14.8 3.2 5 160-170

Archaeological MBY MNT-234 EL-MNT-1783 -10.8 -12.3 14.5 2.8 5 110-120

Archaeological MBY MNT-234 EL-MNT-4825 -10 -11.5 15.2 2.9 1 170-180

Archaeological MBY MNT-234 EL-MNT-4851 -11.6 -13.1 14.4 3.7 5 150-160

Archaeological MBY MNT-234 EL-MNT-2518 -10 -11.5 14.1 3.2 1 030-040

Archaeological MBY MNT-234 EL-MNT-4625 -10.8 -12.3 14.9 3.2 1 020-030

Archaeological MBY MNT-234 EL-MNT-2475 -12.1 -13.6 18.1 2.8 1 170-180

Archaeological MBY MNT-234 EL-MNT-2552 -12.9 -14.4 18.6 2.8 5 120-130

128

Archaeological SLO CA-SLO-2 SO-SLO-S4 -10.2 -11.7 14 3.1 N4/W1 30-40

Archaeological SLO CA-SLO-2 SO-SLO-S5 -10.5 -12 13.6 3.2 N4/W2 20-30

Archaeological SLO CA-SLO-2 SO-SLO-S6 -10.1 -11.6 15.3 3.1 N11/W14 50-60

Archaeological SLO CA-SLO-2 SO-SLO-EL3 -10.9 -12.4 13.3 3.2 N1/W1 40-50

Archaeological SLO CA-SLO-2 SO-SLO-EL4 -9.9 -11.4 14 3.2 N1/W1 60-70

Archaeological SLO CA-SLO-2 SO-SLO-EL15 -10.5 -12 17.1 3.0 N4/W1 20-10

Archaeological SLO CA-SLO-2 SO-SLO-EL18 -10.6 -12.1 12.9 3.2 N4/W1 60-70

Archaeological SLO CA-SLO-2 SO-SLO-EL23 -10.7 -12.2 13.4 3.2 N4/W2 20-30

Archaeological SLO CA-SLO-2 SO-SLO-EL24 -11.6 -13.1 13.6 3.3 N4/W2 40-50

Archaeological SLO CA-SLO-2 SO-SLO-EL25 -11.8 -13.3 13.8 3.3 N4/W2 40-50

Archaeological SLO CA-SLO-2 SO-SLO-EL26 -10.1 -11.6 13.3 3.2 N4/W2 40-50

Archaeological SLO CA-SLO-2 SO-SLO-EL27 -10.2 -11.7 12.6 3.0 N4/W2 40-50

Archaeological SLO CA-SLO-2 SO-SLO-EL28 -10.2 -11.7 12.9 3.0 N4/W2 40-50

Archaeological SLO CA-SLO-2 SO-SLO-EL29 -9.9 -11.4 14 3.0 N4/W2 50-60

Archaeological SLO CA-SLO-2 SO-SLO-EL30 -10.8 -12.3 14.2 3.2 N4/W2 60-70

Archaeological SLO CA-SLO-2 SO-SLO-EL21 -9.5 -11 14.5 3.0 N4/W1 70-80

Archaeological SLO CA-SLO-2 SO-SLO-EL22 -11.6 -13.1 16.3 3.2 N4/W1 70-80

Archaeological SLO CA-SLO-2 SO-SLO-EL9 -9.7 -11.2 13.1 3.2 N10/W17 80-90

Archaeological SLO CA-SLO-2 SO-SLO-EL13 -10.6 -12.1 13.5 3.1 N10/W4 70-80

Archaeological SLO CA-SLO-2 SO-SLO-EL31 -9.3 -10.8 14.3 3.1 N1/W1 180-190

Archaeological SLO CA-SLO-2 SO-SLO-EL33 -12 -13.5 17.3 3.2 N10/W4 190-200

Archaeological SLO CA-SLO-2 SO-SLO-EL40 -11 -12.5 15.1 3.2 S4/W9 230-240

Archaeological SLO CA-SLO-2 SO-SLO-EL41 -12.6 -14.1 15.2 3.1 S4/W9 230-240

Archaeological SLO CA-SLO-2 SO-SLO-EL44 -14.5 -16 18.4 3.2 N1/W4 240-250

Modern MBY MBY 289 -16.1 -15.1 10.9 NA

129

Modern MBY MBY 293 -15.5 -14.5 12.2 NA

Modern MBY MBY 938 -14.2 -13.2 12 NA

Modern MBY MBY 946 -14.1 -13.1 13.3 NA

Modern MBY MBY 952 -14.7 -13.7 12.8 NA

Modern MBY MBY 953 -15.4 -14.4 11.5 NA

Modern MBY MBY 956 -14.7 -13.7 12.3 NA

Modern MBY MBY 959 -15.7 -14.7 11.1 NA

Modern MBY MBY 960 -15 -14 11.8 NA

Modern SNI SNI 961 -12.1 -11.1 15.4 NA

Modern SNI SNI 964 -13 -12 14.1 NA

Modern SNI SNI 965 -12 -11 15.5 NA

Modern SNI SNI 966 -13.6 -12.6 14.3 NA

Modern SNI SNI 967 -12 -11 15.2 NA

Modern SNI SNI 968 -13.1 -12.1 15.3 NA

Modern SNI SNI 969 -13.3 -12.3 14.1 NA

Modern SNI SNI 976 -12.2 -11.2 15.1 NA

Modern SNI SNI 977 -12.1 -11.1 16.4 NA

Modern SNI SNI 978 -13 -12 14.1 NA

Modern SNI SNI 979 -12.9 -11.9 14.9 NA

Modern SNI SNI 982 -13.4 -12.4 14.9 NA

Modern SNI SNI 988 -13.4 -12.4 14.7 NA

Modern BSR BSR 1066 -12.4 -11.4 14.7 NA

Modern BSR BSR 1067 -12.8 -11.8 13.6 NA

Modern BSR BSR 1069 -12 -11 12.3 NA

Modern BSR BSR 1070 -12.4 -11.4 12.4 NA

130

Modern BSR BSR 1071 -12.4 -11.4 14.8 NA

Modern BSR BSR 1072 -12.1 -11.1 11.9 NA

Modern BSR BSR 1073 -13.5 -12.5 12.9 NA

Modern BSR BSR 1074 -11.3 -10.3 12.1 NA

Modern BSR BSR 1075 -12.4 -11.4 13.5 NA

Modern BSR BSR 1077 -12.1 -11.1 13 NA

Modern BSR BSR 1081 -11.9 -10.9 12 NA

Modern BSR BSR 1082 -14.7 -13.7 12 NA

Modern BSR BSR 1083 -12.3 -11.3 13.1 NA

Modern BSR BSR 1084 -13.1 -12.1 13.4 NA

Modern BSR BSR 1085 -13.2 -12.2 12.2 NA

Modern BSR BSR 1088 -11.2 -10.2 13.1 NA

Modern BSR BSR 1090 -13.3 -12.3 13.8 NA

Modern BSR BSR 1091 -13.9 -12.9 13.1 NA

Modern BSR BSR 1092 -11.6 -10.6 13.1 NA

Modern BSR BSR 1093 -13.1 -12.1 13.6 NA

Modern BSR BSR 1094 -13.1 -12.1 13.4 NA

Modern BSR BSR 1095 -12.6 -11.6 13.3 NA

Modern BSR BSR 1096 -12 -11 13.4 NA

Modern BSR BSR 1097 -12.9 -11.9 11.4 NA

Modern BSR BSR 1098 -13.6 -12.6 12 NA

Modern BSR BSR 1103 -14.8 -13.8 11.6 NA

Modern BSR BSR 1104 -12.3 -11.3 11.7 NA

Modern BSR BSR 1105 -14.3 -13.3 12.3 NA

Modern SBC SBC 1183 -10.8 -9.8 14.9 NA

131

Modern SBC SBC 1184 -12.1 -11.1 16.3 NA

Modern SBC SBC 1185 -12.3 -11.3 17.1 NA

Modern SBC SBC 1186 -11.9 -10.9 17 NA

Modern SBC SBC 1187 -12.8 -11.8 16.1 NA

Modern SBC SBC 1188 -12.2 -11.2 13.3 NA

Modern SBC SBC 1189 -12.3 -11.3 16.3 NA

Modern SBC SBC 1190 -13.5 -12.5 15.4 NA

Modern SBC SBC 1191 -13.6 -12.6 14.2 NA

Modern SBC SBC 1192 -11.6 -10.6 16 NA

Modern SBC SBC 1196 -12 -11 15.5 NA

Modern SBC SBC 1197 -12.4 -11.4 16.4 NA

Modern SBC SBC 1198 -12.4 -11.4 16.6 NA

Modern SBC SBC 1199 -12.4 -11.4 17.3 NA

Modern SBC SBC 1201 -11.5 -10.5 16.2 NA

Modern SBC SBC 1202 -13.1 -12.1 15 NA

Modern SBC SBC 1203 -12.7 -11.7 15.1 NA

Modern SBC SBC 1204 -12.3 -11.3 15.9 NA

Modern SBC SBC 1264 -10.9 -9.9 16.8 NA

Modern SBC SBC 1265 -12.1 -11.1 16.7 NA

Modern SBC SBC 1267 -12.3 -11.3 16.5 NA

Modern SBC SBC 1268 -11.8 -10.8 16.3 NA

Modern SBC SBC 1269 -12.5 -11.5 16 NA

Modern SBC SBC 1270 -12.3 -11.3 16 NA

Modern SBC SBC 1272 -11.4 -10.4 16 NA

Modern SBC SBC 1273 -11.6 -10.6 16.9 NA

132

Modern SBC SBC 1276 -12 -11 16.1 NA

Modern SBC SBC 1277 -11.4 -10.4 15.9 NA

Modern SBC SBC 1278 -12.2 -11.2 16.2 NA

Modern SBC SBC 1279 -12.7 -11.7 16.1 NA

Modern SBC SBC 1280 -12.3 -11.3 17.4 NA

Modern SBC SBC 1282 -12.1 -11.1 15.5 NA

Modern SBC SBC 1283 -12.8 -11.8 16.5 NA

Modern SBC SBC 1284 -12.3 -11.3 16.2 NA

Modern SBC SBC 1285 -12.3 -11.3 16.2 NA

Modern SBC SBC 1286 -12 -11 16.5 NA

Modern SBC SBC 1287 -12.8 -11.8 16.6 NA

Modern MBY MBY 3110 -14.8 -13.8 11.9 NA

Modern MBY MBY 4687 -13.5 -12.5 11.3 NA

Modern MBY MBY 4690 -13.1 -12.1 11.9 NA

Modern MBY MBY 4696 -14.8 -13.8 10.8 NA

Modern MBY MBY 4724 -14.6 -13.6 10.6 NA

Modern MBY MBY 4732 -13.4 -12.4 11.8 NA

Modern MBY MBY 4742 -13.5 -12.5 12.5 NA

Modern MBY MBY 4748 -12.7 -11.7 12.8 NA

Modern MBY MBY 4749 -13.4 -12.4 12.7 NA

Modern MBY MBY 4751 -14.2 -13.2 11.7 NA

Modern MBY MBY 4772 -14.3 -13.3 13 NA

Modern MBY MBY 4794 -13.1 -12.1 12.7 NA

Modern MBY MBY 4818 -13.4 -12.4 12.7 NA

Modern MBY MBY 4824 -13 -12 12.9 NA

133

Modern MBY MBY 4833 -14.2 -13.2 14.1 NA

Modern MBY MBY 4834 -13.2 -12.2 12.3 NA

Modern MBY MBY 4844 -13.7 -12.7 11.7 NA

Modern MBY MBY 4848 -13.6 -12.6 11.8 NA

Modern MBY MBY 4850 -14.1 -13.1 12.8 NA

Modern MBY MBY 4866 -13.7 -12.7 11 NA

Modern MBY MBY 4872 -13.8 -12.8 13.6 NA

Modern MBY MBY 4874 -13.8 -12.8 12.9 NA

Modern SLO SLO 114331 -13 -12 15.3 NA

Modern SLO SLO 114378 -13.3 -12.3 16.1 NA

Modern SLO SLO 114425 -13.5 -12.5 14.4 NA

Modern SLO SLO 114472 -12.3 -11.3 14.7 NA

Modern SLO SLO 114519 -12.3 -11.3 16.2 NA

Modern SLO SLO 114566 -12.4 -11.4 15.6 NA

Modern SLO SLO 114613 -13.2 -12.2 14.3 NA

Modern SLO SLO 114660 -12.7 -11.7 17 NA

Modern SLO SLO 114707 -13.1 -12.1 15 NA

Modern SLO SLO 114754 -12.8 -11.8 14 NA

Modern SLO SLO 114801 -12.9 -11.9 15.5 NA

Modern SLO SLO 114848 -13.1 -12.1 15.5 NA

Modern SLO SLO 114895 -12.5 -11.5 14.5 NA

Modern SLO SLO 114942 -13.1 -12.1 16.1 NA

Modern SLO SLO 114989 -13.1 -12.1 15 NA

Modern SLO SLO 115036 -12.8 -11.8 16 NA

Modern SLO SLO 117431 -12.8 -11.8 13.6 NA

134

Modern SLO SLO 118690 -11.2 -10.2 15.8 NA

Modern SLO SLO 118736 -12.9 -11.9 14.1 NA

Modern SLO SLO 118782 -12.8 -11.8 15.7 NA

Modern SLO SLO 118828 -12 -11 15.5 NA

Modern SLO SLO 118874 -12.5 -11.5 13.9 NA

Modern SLO SLO 118920 -12.3 -11.3 13.8 NA

Modern SLO SLO 118966 -12.5 -11.5 15.1 NA

Modern SLO SLO 119012 -12.6 -11.6 13.4 NA

Modern SLO SLO 119058 -12.7 -11.7 14.1 NA

Modern SLO SLO 119104 -13 -12 14.5 NA

Modern SLO SLO 119150 -12.9 -11.9 16.2 NA

Modern SLO SLO 119196 -13.6 -12.6 13.4 NA

Modern SLO SLO 119242 -13 -12 15 NA

Modern SLO SLO 119288 -12.6 -11.6 13.1 NA

Modern SLO SLO 119334 -13 -12 14.9 NA

Modern SLO SLO 119380 -12.1 -11.1 14.6 NA

Modern SLO SLO 119472 -12.7 -11.7 13.7 NA

Modern SLO SLO 119518 -12.7 -11.7 15.6 NA

Modern SLO SLO 119564 -12.2 -11.2 15.6 NA

Modern SLO SLO 119610 -13.1 -12.1 15.3 NA

Modern SLO SLO 119656 -10.6 -9.6 15 NA

Modern SLO SLO 119702 -12.5 -11.5 15.6 NA

Modern SLO SLO 119748 -12.9 -11.9 15.9 NA

Modern SLO SLO 119794 -12 -11 15.3 NA

Modern SLO SLO 119840 -13.2 -12.2 14.6 NA

135

Modern SLO SLO 119886 -13.5 -12.5 15.7 NA

Modern SLO SLO 119932 -12.9 -11.9 14.3 NA

Modern SLO SLO 119978 -12.6 -11.6 14.5 NA

Modern SLO SLO 120024 -12.3 -11.3 14.2 NA

Modern SLO SLO 120070 -13 -12 15.5 NA

Modern SLO SLO 120162 -12.1 -11.1 15.7 NA

Modern SLO SLO 120208 -12 -11 14.6 NA

Modern SLO SLO 120254 -12.9 -11.9 14.6 NA

Modern SLO SLO 120300 -13.6 -12.6 15.1 NA

Modern SLO SLO 120346 -11.7 -10.7 14.8 NA

Modern SLO SLO 120392 -11.1 -10.1 15 NA

Modern SLO SLO 120438 -12.2 -11.2 14.5 NA

Modern SLO SLO 120484 -13.1 -12.1 16 NA

Modern SLO SLO 120530 -13.2 -12.2 15.3 NA

136

Table 3. ANOVA results with posthoc Tukey’s HSD results for pairwise comparisons of sea otter bulk tissue d13C and d15N values. Rows are pairwise comparisons, columns are d13C and d15N values. Asterisks (*) indicate significant differences (adjusted p-value of <0.05) between pairwise comparisons. In bold text are the comparisons between vibrissae and bone collagen.

13 15 d C d N Muscle-Bone 0.000* 0.000* Liver-Bone 0.000* 0.000* Vibrissae-Bone 0.000* 0.315 Liver-Muscle 0.000* 0.880 Vibrissae-Muscle 0.000* 0.002* Vibrissae-Liver 0.000* 0.024*

137

Table 4. Kruskal-Wallis and resulting pairwise Wilcoxon Signed Rank results among sea otter d13C and d 15N values (Appendix 2, Table 2). Rows are pairwise comparisons, columns are d13C and d15N values. Asterisks (*) indicate significant differences (Bonferroni adjusted p-value of <0.05) between groups. Acronyms for regions are as follows: ANO = Año Nuevo, MBY = Monterey Bay, BSR = Big Sur Reserve, SLO = San Louis Obispo, SBC = Santa Barbara Channel, SMI = San Miguel Island, SNI = San Nicolas Island. ‘mod’ = modern populations, ‘arc’ = ancient. In the top section of the table, comparisons are population-wide: ancient MBY combines MNT-234 and MNT- 831, ancient SMI combines SMI-1, SMI-525, SMI-528, and SMI-602. Ancient SNI combines SNI-011, SNI-025 and SNI-040. Bottom section of the table compares isotopic values of archaeological sites within regions.

13 15 Comparison d C d N mod:MBY-arc:ANO 1.000 0.000* mod:MBY-arc:MBY 0.105 0.000* mod:SLO-arc:SLO 1.000 0.202 mod:SBC-arc:SMI 1.000 0.001* mod:SNI-arc:SNI 1.000 1.000 mod:MBY-mod:BSR 0.000* 0.2071 mod:MBY-mod:SLO 0.000* 0.000* mod:MBY-mod:SBC 0.000* 0.000* mod:MBY-mod:SNI 0.000* 0.000* mod:BSR-mod:SLO 1.000 0.000* mod:BSR-mod:SBC 0.627 0.000* mod:BSR-mod:SNI 1.000 0.000* mod:SLO-mod:SBC 0.005* 0.000* mod:SLO-mod:SNI 1.000 1.000 mod:SBC-mod:SNI 1.000 0.004*

MNT234-MNT831 1.000 1.000 SMI525-SMI1 1.000 1.000 SMI528-SMI1 1.000 1.000 SMI602-SMI1 1.000 1.000 SMI528-SMI525 1.000 1.000 SMI602-SMI525 1.000 1.000 SMI602-SMI528 1.000 1.000 SNI025-SNI011 1.000 1.000 SNI040-SNI011 1.000 1.000 SNI040-SNI025 1.000 1.000

138

Table 5. Comparison of isotopic niche width between ancient and modern sea otter populations. Each cell represents the proportion of draws (from 10,000 Bayesian iterations) from the row group that are larger than the column group. Acronyms for regions are as follows: ANO = Año Nuevo, MBY = Monterey Bay, BSR = Big Sur Reserve, SLO = San Louis Obispo, SBC = Santa Barbara Channel, SMI = San Miguel Island, SNI = San Nicolas Island. ‘mod’ = modern populations, ‘arc’ = ancient. Here, comparisons are population-wide: ancient MBY combines MNT-234 and MNT-831, ancient SMI combines SMI-1, SMI- 525, SMI-528, and SMI-602. Ancient SNI combines SNI-011, SNI-025 and SNI-040. All samples are also compared to the isotopic niche width of modern potential prey items from Monterey Bay/Big Sur and San Nicolas Island (see Newsome et al. 2015).

Ancient Populations Modern Populations Modern Prey

ANO MBY SLO SMI SNI MBY BSR SLO SBC SNI MBY/BSR SNI

ANO – 0.83 0.42 0.06 0.21 0.98 0.92 1.00 1.00 1.00 0.07 0.01

MBY – – 0.09 0.00 0.02 0.88 0.69 1.00 0.99 1.00 0.00 0.00

SLO – – – 0.04 0.22 1.00 0.97 1.00 1.00 1.00 0.03 0.00

SMI – – – – 0.81 1.00 1.00 1.00 1.00 1.00 0.61 0.10

SNI – – – – – 1.00 1.00 1.00 1.00 1.00 0.20 0.02 Ancient Populations Ancient MBY – – – – – – 0.25 0.92 0.92 0.98 0.00 0.00

BSR – – – – – – – 0.99 0.98 1.00 0.00 0.00

SLO – – – – – – – – 0.56 0.91 0.00 0.00 Populations SBC – – – – – – – – – 0.89 0.00 0.00

SNI – – – – – – – – – – 0.00 0.00 Modern Modern MBY/B – – – – – – – – – – – 0.01 SR

SNI – – – – – – – – – – – – Prey Modern Modern

139

Table 6. Comparison of isotopic niche width between individual archaeological sites and modern sea otter populations. Each cell represents the proportion of draws (from 10,000 Bayesian iterations) from the row group that are larger than the column group. Archaeological sites are listed by name: SMI-602 and MNT-831 excluded because of small sample size and provenance concerns. Modern sites are listed by region. Acronyms for regions are as follows: ANO = Año Nuevo, MBY = Monterey Bay, BSR = Big Sur Reserve, SLO = San Louis Obispo, SBC = Santa Barbara Channel, SMI = San Miguel Island, SNI = San Nicolas Island. All samples are also compared to the isotopic niche width of modern potential prey items from Monterey Bay/Big Sur and San Nicolas Island (see Newsome et al. 2015).

Modern Archaeological Sites Modern Sites Prey

SMA MNT SLO SMI SMI SMI SNI SNI SNI MBY MBY BSR SLO SBC SNI SNI 238 234 2 1 525 528 011 025 040 /BSR

SMA 238 – 0.21 0.01 0.04 0.02 0.00 0.00 0.00 0.59 0.68 0.36 0.75 0.98 0.43 0.00 0.00

MNT 234 – – 0.03 0.14 0.07 0.00 0.01 0.01 0.82 0.94 0.72 0.98 1.00 0.74 0.00 0.00

SLO 2 – – – 0.65 0.59 0.13 0.15 0.13 0.99 1.00 1.00 1.00 1.00 0.99 0.09 0.01

SMI 1 – – – – 0.43 0.12 0.13 0.12 0.96 1.00 0.96 1.00 1.00 0.95 0.12 0.03

SMI 525 – – – – – 0.12 0.14 0.13 0.98 1.00 0.99 1.00 1.00 0.97 0.11 0.02

SMI 528 – – – – – – 0.45 0.47 1.00 1.00 1.00 1.00 1.00 1.00 0.64 0.25

SNI 011 – – – – – – – 0.54 1.00 1.00 1.00 1.00 1.00 1.00 0.66 0.38

SNI 025 – – – – – – – – 1.00 1.00 1.00 1.00 1.00 1.00 0.66 0.30

SNI 040 – – – – – – – – – 0.56 0.30 0.63 0.93 0.35 0.00 0.00 Archaeological Sites Archaeological MBY – – – – – – – – – – 0.15 0.61 0.98 0.25 0.00 0.00

BSR – – – – – – – – – – – 0.93 1.00 0.57 0.00 0.00

SLO – – – – – – – – – – – – 0.97 0.17 0.00 0.00

SBC – – – – – – – – – – – – – 0.01 0.00 0.00

SNI – – – – – – – – – – – – – – 0.00 0.00 Modern Sites Modern

MBY/BSR – – – – – – – – – – – – – – – 0.01

SNI – – – – – – – – – – – – – – – – Prey

140

Table 7. Amino Acid d15N data of ancient and modern sea otters. Each value represents an average ± SD across injections of a sample; see Appendix 2 for details on analysis. Abbreviations for amino acids are as follows: alanine (Ala), glycine (Gly), serine (Ser), valine (Val), leucine (Leu), isoleucine (Ile), hydroxyproline/proline (Hyp-Pro), aspartic acid (Asp), glutamic acid (Glu), phenylalanine (Phe), lysine (Lys). ‘ND’ indicates we were not able to reliably measure the given amino acid. Modern samples are bone collagen from our stranded tissue dataset (Appendix 2, Table 1). Acronyms for Region are as follows: ANO = Año Nuevo, MBY = Monterey Bay, SMI = San Miguel Island, SNI = San Nicolas Island.

SampleID Type Region Ala15N Gly15N Ser15N Val15N Leu15N Ile15N Hyp-Pro15N Asp15N Glu15N Phe15N Lys15N

SO-SMA- Ancient ANO 21.2 ± 0.7 11.7 ± 0.6 12.9 ± 0.4 23.1 ± 0.3 21.3 ± 0.7 20.1 ± 0.9 22.9 ± 0.4 20.7 ± 0.3 23.1 ± 0.5 13.3 ± 0.2 11.1 ± 0.5 146

SO-SMA- Ancient ANO 25.8 ± 0.2 14.3 ± 0.6 15.3 ± 0.5 24.8 ± 0.4 24.7 ± 0.4 23.5 ± 0.3 24.1 ± 0.3 22.2 ± 0.5 26.2 ± 0.6 12.3 ± 0.1 10.0 ± 0.3 168

SO-SMA- Ancient ANO 21.0 ± 0.5 12.3 ± 0.6 12.9 ± 0.3 22.1 ± 0.4 19.3 ± 0.1 20.0 ± 0.3 21.2 ± 0.2 19.3 ± 0.7 22.1 ± 0.4 13.1 ± 0.4 11.1 ± 0.6 252

SO-SMA- Ancient ANO 23.1 ± 0.1 ND 13.4 ± 0.6 23.8 ± 0.2 22.2 ±0.1 23.4 ± 0.6 20.1 ± 0.2 20.2 ± 0.5 23.4 ± 0.0 12.6 ± 0.2 10.8 ± 0.0 739

SO-SMA- Ancient ANO 25.8 ± 0.0 14.5 ± 0.0 13.5 ± 0.4 24.5 ± 0.4 23.3 ± 0.2 23.6 ± 0.9 22.6 ± 0.1 22.0 ± 0.2 25.1 ± 0.2 10.5 ± 0.8 9.5 ± 0.2 751

SO-MNT- Ancient MBY 21.1 ± 0.1 12.0 ± 0.0 11.0 ± 0.3 21.4 ± 0.5 20.2 ± 0.1 21.4 ± 0.3 21.2 ± 0.1 19.8 ± 0.1 22.3 ± 0.1 9.4 ± 0.5 11.2 ± 0.3 234-2724

SO-MNT- Ancient MBY 23.0 ± 0.0 10.9 ± 0.2 12.0 ± 0.6 22.3 ± 0.3 22.5 ± 0.1 22.6 ± 0.7 20.7 ± 0.1 21.4 ± 0.1 23.9 ± 0.2 12.1 ± 0.1 10.1 ± 0.1 234-2977

SO-MNT- Ancient MBY 24.1 ± 0.6 12.4 ± 0.3 10.9 ± 0.9 23.8 ± 0.4 23.5 ± 0.4 23.6 ± 0.7 21.2 ± 0.2 21.7 ± 0.3 25.2 ± 0.2 11.1 ± 0.5 9.1 ± 0.0 234-4825

MNT- 831-EL- Ancient MBY ND 10.3 ± 0.3 11.6 ± 0.3 20.6 ± 0.1 20.4 ± 0.6 20.9 ± 0.9 20.7 ± 0.0 19.7 ± 0.3 21.9 ± 0.5 8.6 ± 0.4 7.9 ± 0.1 005

MNT- 831-EL- Ancient MBY ND 14.3 ± 0.2 13.2 ± 0.3 21.7 ± 0.0 21.3 ± 0.5 21.6 ± 0.4 20.2 ± 0.0 18.8 ± 0.5 21.1 ± 0.8 10.5 ± 0.6 9.7 ± 0.6 018

SO-SMI- Ancient SMI 18.9 ± 0.0 9.7 ± 0.1 11.6 ± 0.5 19.2 ± 0.1 18.7 ± 0.6 17.7 ± 0.3 17.7 ± 0.1 16.7 ± 0.1 19.6 ± 0.3 10.8 ± 0.3 9.4 ± 0.3 9702

141

SO-SMI- Ancient SMI 20.6 ± 0.4 9.8 ± 0.4 10.4 ± 0.2 18.8 ± 0.1 18.3 ± 0.2 18.0 ± 0.2 18.7 ± 0.1 17.3 ± 0.0 19.7 ± 0.0 9.5 ± 0.1 8.7 ± 0.7 10636

SO-SMI- Ancient SMI 20.5 ± 0.3 10.3 ± 0.4 10.0 ± 0.6 20.1 ± 0.5 19.7 ± 0.2 19.0 ± 0.4 18.4 ± 0.0 18.6 ± 0.2 21.5 ± 0.4 10.0 ± 0.8 10.3 ± 0.2 11080

SO-SMI- 257- Ancient SMI 21.0 ± 0.1 11.3 ± 0.2 10.5 ± 0.5 20.2 ± 0.1 19.5 ± 0.1 19.7 ± 0.0 18.7 ± 0.1 17.5 ± 0.0 20.1 ± 0.2 8.7 ± 0.2 8.7 ± 0.3 12400

SO-SNI- Ancient SNI 21.9 ± 0.4 13.4 ± 0.1 12.6 ± 0.0 21.1 ± 0.2 20.4 ± 0.1 20.8 ± 0.5 20.6 ± 0.2 18.9 ± 0.2 21.2 ± 0.3 10.7 ± 0.5 10.4 ± 0.6 001

SO-SNI- Ancient SNI 20.6 ± 0.4 12.2 ± 0.2 13.1 ± 1.0 20.9 ± 0.6 19.7 ± 0.2 20.3 ± 1.0 20.8 ± 0.2 19.4 ± 0.4 21.1 ± 0.3 11.9 ± 0.4 8.8 ± 0.7 006

SO-SNI- Ancient SNI 20.4 ± 0.2 11.7 ± 0.3 13.1 ± 0.1 19.3 ± 0.2 19.3 ± 0.6 18.7 ± 0.8 19.3 ± 0.0 18.8 ± 0.2 20.8 ± 0.6 12.8 ± 0.2 11.9 ± 0.5 008

SO-SNI- Ancient SNI 20.1 ± 0.5 11.9 ± 0.6 13.3 ± 0.9 19.3 ± 0.9 19.5 ± 0.8 18.5 ± 0.1 19.4 ± 0.3 19.4 ± 0.2 21.0 ± 0.5 12.6 ± 0.6 11.4 ± 0.5 012

SO-MBY- Modern MBY 19.6 ± 0.6 11.2 ± 0.1 11.4 ± 0.6 19.4 ± 0.7 18.0 ± 0.4 NA 18.1 ± 0.2 18.5 ± 0.1 21.2 ± 0.3 10.6 ± 0.5 8.6 ± 0.6 7064-14

SO-MBY- Modern MBY 19.9± 0.3 11.0 ± 0.4 10.0 ± 0.4 20.7 ± 0.2 18.7 ± 0.4 17.0 ± 0.3 18.0 ± 0.2 16.9 ± 0.5 20.1 ± 0.5 11.1 ± 0.0 7.8 ± 0.6 5204-08

SO-MBY- Modern MBY 22.0 ±0.4 11.2 ± 0.2 11.7 ± 0.3 22.5 ± 0.5 21.5 ± 0.5 21.0 ± 0.1 21.9 ± 0.3 21.2 ± 0.3 23.2 ± 0.2 10.1 ± 0.5 10.3 ± 0.6 4923-07

SO-MBY- Modern MBY 20.2 ± 0.3 11.8 ± 0.2 11.8 ± 0.1 21.4 ± 0.1 19.4 ± 0.2 19.2 ± 0.8 18.1 ± 0.1 19.4 ± 0.5 20.8 ± 0.1 10.4 ± 0.0 9.9 ± 0.1 5054-07

SO-MBY- Modern MBY 22.5 ± 0.3 10.5 ± 0.4 11.4 ± 0.2 23.1 ± 0.3 22.0 ± 0.4 22.6 ± 0.5 20.7 ± 0.1 20.8 ± 0.6 23.9 ± 0.3 11.2 ± 0.1 9.1 ± 0.3 5818-10

142

18 locale

ANO 16 MBY N 15

d SLO SMI 14 SNI

12 SMA−238 MNT−234 MNT−831CA−SLO−2 SMI−1 SMI−525 SMI−528 SMI−602 SNI−011 SNI−025 SNI−040 Archaeological Site

−10 locale

ANO −12 MBY C 13

d SLO SMI −14 SNI

−16 SMA−238 MNT−234 MNT−831CA−SLO−2 SMI−1 SMI−525 SMI−528 SMI−602 SNI−011 SNI−025 SNI−040 Archaeological Site

Figure 1. Boxplots comparing d15N (top panel) and d13C (bottom panel) values among archaeological sites in each region. Colors represent different regions: red = Año Nuevo, olive = Monterey Bay, green = San Louis Obispo, blue = San Miguel Island, pink = San Nicolas Island.

143

Ancient Otters Modern Otters Modern Prey Median SEA

Time Span (years)

Figure 2. Relationship between median SEAB and time span. Time span for each archaeological site presented in Table 2.1; modern prey and otters were given a time span of 10 years.

144

Appendix 3

1.Procedural details on bulk and amino acid isotope analysis.

For both bulk and amino acid analysis we report all isotopic results as d values: d13C or d15N = 1000*[(Rsamp/Rstd) − 1], where Rsamp and Rstd are the 13C: 12C or 15N:14N ratios of the sample and standard, respectively. The internationally accepted standards for d13C

15 and d N analysis are Vienna-Pee Dee Belemnite limestone (V-PDB) and atmospheric N2 respectively. The units are expressed as parts per thousand, or per mil (‰). The standard deviation of organic reference materials within a run was ≤0.2‰ for both d13C and d15N values. d13C and d15N values of all samples were analyzed using a Costech 4010 elemental analyzer (Valencia, CA) interfaced with a Thermo Scientific Delta V Plus isotope ratio mass spectrometer (Bremen, Germany) at the University of New Mexico Center for Stable

Isotopes (UNM-CSI, Albuquerque, NM). The standard deviation of organic reference materials within a run was ≤0.2‰ for both d13C and d15N values.

For AA d13C analysis, primary producer (kelps, red and green algae) samples were first hydrolyzed with 1.5mL of 6N HCl at 110°C for 20 hours. For POM/phytoplankton samples, approximately ½ of each filter was added to 6N HCl and treated as above.

Hydrolysates were then passed through a cation exchange resin column (Dowex 50WX8

100-200 mesh) to isolate AAs from other contaminant and metabolites (Amelung & Zhang

2001). After Dowex purification, amino acids were derivatized to N-trifluoroacetic acid isopropyl esters following established protocols (O’Brien et al. 2002; Newsome et al. 2011;

Elliott Smith et al. 2018). Samples were derivatized along with an in-house reference material containing all AAs measured for d13C. Derivatized samples were injected into a

145

60m BPX5 gas chromatograph column for AA separation (0.32 ID, 1.0μm film thickness,

SGE Analytical Science, Victoria, Australia) in a Thermo Scientific Trace 1300, then combusted into CO2 with a GC Isolink II interfaced to a Thermo Scientific Delta V Plus isotope ratio mass spectrometer at UNM-CSI. Samples were run in duplicate and bracketed with in-house AA reference material; within-run standard deviations of d13C values in this reference material ranged from 0.3‰ (isoleucine) to 0.5‰ (tyrosine). We reliably measured d13C values of twelve AAs including six considered non-essential: alanine (Ala), aspartic acid (Asp), glutamic acid (Glu), glycine (Gly), proline (Pro), serine (Ser); and six considered essential: isoleucine (Ile), leucine (Leu), lysine (Lys), phenylalanine (Phe), threonine (Thr), and valine (Val).

The reagents used during derivatization add carbon to the side chains of amino acids, and so the raw d13C measured via GC-C-IRMS is a combination of original amino acid carbon and reagent carbon. However, because amino acid standards of known d13C composition were derivatized and run with each batch of samples, we were able to correct for this carbon addition for each amino acid using the following equation (From O’Brien et al., 2002, Newsome et al., 2011; 2014):

,- ,- ( ,- ) ,- δ $/%& − δ $/%0 + δ $%0/ × p%0/ δ $%&'()* = p%0/

13 Here, d Cdsa refers to the measured value of the derivatized amino acid within the sample

13 and d Cdst refers to the measured value of the same derivatized amino acid within the

146 among amino acids. The in-house reference material derivatized with each batch of samples is a mixture of 13 powdered AAs purchased from Sigma Aldrich (Saint Louis,

MO). The d13C values for these underivatized amino acids were measured via EA-IRMS at UNM-CSI. Each amino acid was then dissolved at a concentration of ~250 mM into weak acid and the dissolved AA solutions mixed together. For each batch of samples, a small aliquot of this solution was dried and then derivatized alongside each batch of unknown samples.

147

Table 1. Tukey’s HSD pairwise comparisons of Humboldt Current primary producer EAA d13C values. Rows are pairwise comparisons and columns are essential amino acids; bold number indicate pairwise comparisons that are significant at an adjusted p-value ≤ 0.05. For consumers, we show only pairwise comparisons where species differed in at least one EAA d13C value at an adjusted p-value ≤ 0.1. Species codes: Green = Ulva sp., Phyto = particulate organic matter and phytoplankton, Kelp = Lessonia sp., and Macrocystis pyrifera, Red = samples of Plocamium cartilagineum, Rhodymenia coralline, crustose coralline algae and an unidentified red alga.

Comparison: Ile d13C Leu d13C Lys d13C Phe d13C Thr d13C Val d13C

Green-Phyto 0.09 0.01 0.01 0.19 0.09 0.93

Kelp-Phyto 0.01 0.95 1.00 0.68 0.14 0.93

Red-Phyto 0.00 0.00 0.00 0.00 0.00 0.00

Kelp-Green 0.00 0.00 0.02 0.03 0.94 1.00

Red-Green 0.00 0.00 0.00 0.00 0.00 0.00

Red-Kelp 0.00 0.00 0.00 0.00 0.00 0.00 Producers Engraulis-Cheilodactylus 0.56 0.52 1.00 0.82 0.85 0.00

Aplodactylus- Cheilodactylus 0.00 0.01 0.18 0.06 0.00 0.00

Taleipus- Cheilodactylus 0.02 0.23 1.00 0.80 1.00 0.26

Engraulis-Palabrax 0.99 1.00 1.00 1.00 0.84 0.03

Aplodactylus-Palabrax 0.00 0.61 0.66 0.41 0.00 0.00

Tegula-Palabrax 0.52 0.05 0.02 0.01 0.99 0.62

Aplodactylus-Isacia 0.00 0.21 0.52 0.48 0.00 0.01

Tegula-Isacia 0.58 0.15 0.02 0.00 0.99 0.20

Tegula-Engraulis 0.11 0.03 0.11 0.00 0.26 0.00

Tetrapygus-Engraulis 1.00 0.98 0.71 0.92 1.00 0.10

Taleipus-Aplodactylus 0.84 0.99 0.08 0.92 0.03 0.23

Perumytilus-Aplodactylus 0.00 0.55 0.35 0.79 0.10 0.66

Tegula-Aplodactylus 0.00 0.00 0.00 0.00 0.00 0.00

Tetrapygus-Aplodactylus 0.18 0.28 0.00 0.18 0.45 0.02

Zooplankton- Aplodactylus 0.02 0.96 0.94 0.37 0.07 0.63

Tegula - Taleipus 0.00 0.00 0.29 0.00 0.72 0.02

Tegula - Perumytilus 0.78 0.12 0.13 0.01 0.55 0.01 Consumers

148

Table 2. Coefficients of linear discriminants for essential amino acid d13C values. Individual essential amino acids are listed in alphabetical order. For each LD axis, the two most influential essential amino acids are indicated in bold. The relative proportion of trace explained is: LD1 (0.7775), LD2 (0.1738), LD3 (0.0488).

Essential Amino Acid LD1 LD2 LD3 Isoleucine -0.49 0.42 0.48 Leucine 0.11 0.01 -0.70 Lysine -0.36 0.30 -0.44 Phenylalanine 0.21 0.11 0.28 Threonine -0.05 -0.40 0.00 Valine 0.23 -0.49 0.35

149

Table 3. Linear discriminant analysis (LDA) cross-validation of Humboldt Current producers. Columns represent classification of a sample group as given by LDA. Rows represent the group being classified. The cross diagonal (bold text), are correct reclassifications. Numbers falling off this diagonal indicate samples that were incorrectly reclassified by the model. Thus, the sum of the cross diagonal divided by the total number of samples equals the % success of the model (87.5%). See Figure 3.2 for a visual representation of these data. Species codes are as follows: Green = Ulva sp., Phyto = particulate organic matter and phytoplankton, Kelp = Lessonia sp., and Macrocystis pyrifera, Red = samples of Plocamium cartilagineum, Rhodymenia coralline, and crustose coralline algae and an unidentified red alga.

LDA classification

Phyto Green Kelp Red % Correct

Phyto 12 1 0 0 92.3

Green 1 6 0 0 85.7 Identity

Kelp 1 0 11 0 91.7

Red 0 0 0 7 100.0 True Producer

150

Table 4. Results of LDA bootstrap resampling of Humboldt Current consumers. Rows represent an individual consumer. Columns show the average (± standard deviation) posterior probabilities of classification with a given producer group across 10,000 LDA bootstrap iterations. Primary producer codes are as follows: pGreen = Ulva sp., pPhyto = particulate organic matter and phytoplankton, pKelp = Lessonia sp., and Macrocystis pyrifera, pRed = samples of Plocamium cartilagineum, Rhodymenia coralline, and one singleton each of crustose coralline algae and an unidentified red alga.

ID Species pPhyto pGreen pKelp pRed

HUM-COL-JER3 Aplodactylus punctatus 1.0 ± 1.9 0 ± 0 99.0 ± 2.0 0 ± 0

HUM-COL-JER3-B2 Aplodactylus punctatus 7.2 ± 5.5 0 ± 0 92.8 ± 5.5 0 ± 0

HUM-ISM-JER3 Aplodactylus punctatus 1.9 ± 0.9 0 ± 0 42.2 ± 19.2 56.0 ± 19.8

HUM-COL-JER5 Aplodactylus punctatus 73.9 ± 5.0 0.5 ± 0.1 25.6 ± 5.0 0 ± 0

HUM-COL-JER5-B2 Aplodactylus punctatus 97.8 ± 0.7 0 ± 0 2.2 ± 0.7 0 ± 0

HUM-ISM-JER9 Aplodactylus punctatus 0 ± 0 0 ± 0 0 ± 0 100.0 ± 0

HUM-ISM-JER3-B2 Aplodactylus punctatus 11.3 ± 10.7 72.8 ± 20.4 0 ± 0 16.0 ± 20.5

HUM-ISM-JER9-B2 Aplodactylus punctatus 2.0 ± 1.3 98.0 ± 1.3 0 ± 0 0 ± 0

HUM-ISM-BIL3-B2 Cheilodactylus variegatus 0.5 ± 0.5 0 ± 0 99.5 ± 0.5 0 ± 0

HUM-ISM-BIL5-B2 Cheilodactylus variegatus 1.3 ± 1.2 0 ± 0 98.7 ± 1.2 0 ± 0

HUM-ISM-BIL5 Cheilodactylus variegatus 6.4 ± 2.0 0 ± 0 93.6 ± 2.0 0 ± 0

HUM-ISM-BIL3 Cheilodactylus variegatus 9.6 ± 4.1 0 ± 0 90.4 ± 4.1 0 ± 0

HUM-COL-BIL4-B2 Cheilodactylus variegatus 49.8 ± 10.0 8.3 ± 3.1 41.9 ± 10.2 0 ± 0

HUM-COL-BIL9 Cheilodactylus variegatus 92.2 ± 2.7 7.1 ± 2.6 0.7 ± 0.3 0 ± 0

HUM-ANF-ERI3-B2 Engraulis ringens 7.6 ± 5.1 0 ± 0 92.3 ± 5.1 0 ± 0

HUM-ANF-ERI4-B2 Engraulis ringens 98.3 ± 0.5 0.1 ± 0 1.6 ± 0.5 0 ± 0

HUM-ANF-ERI2-B2 Engraulis ringens 4.4 ± 1.9 95.6 ± 1.9 0 ± 0 0 ± 0

HUM-ANF-ERI1-B2 Engraulis ringens 57.9 ± 8.8 42.1 ± 8.8 0 ± 0 0 ± 0

HUM-ISM-CBZ2-B2 Isacia coneptionis 1.2 ± 0.6 0 ± 0 98.8 ± 0.6 0 ± 0

HUM-COL-CBZ3 Isacia coneptionis 41.8 ± 9.4 0 ± 0 58.2 ± 9.4 0 ± 0

HUM-ISM-CBZ1-B2 Isacia coneptionis 42.0 ± 9.6 1.2 ± 0.5 56.8 ± 9.8 0 ± 0

HUM-COL-CBZ1-B2 Isacia coneptionis 69.8 ± 6.8 0 ± 0 30.2 ± 6.8 0 ± 0

HUM-COL-CBZ1 Isacia coneptionis 81.9 ± 4.8 0 ± 0 18.1 ± 4.8 0 ± 0

HUM-ISM-CBZ1 Isacia coneptionis 0 ± 0 0 ± 0 100.0 ± 0 0 ± 0

HUM-ISM-CBZ2 Isacia coneptionis 90.0 ± 3.9 0 ± 0 10.0 ± 3.9 0 ± 0

HUM-COL-CAB3-B2 Palabrax humeralis 0.2 ± 0.1 0 ± 0 99.8 ± 0.1 0 ± 0

151

HUM-ISM-CAB4-B2 Palabrax humeralis 0.4 ± 0.5 0 ± 0 99.6 ± 0.5 0 ± 0

HUM-COL-CAB6 Palabrax humeralis 8.5 ± 5.3 0 ± 0 91.5 ± 5.3 0 ± 0

HUM-COL-CAB3 Palabrax humeralis 27.1 ± 7.8 0 ± 0 72.9 ± 7.8 0 ± 0

HUM-ISM-CAB4 Palabrax humeralis 39.5 ± 10.2 0 ± 0 60.5 ± 10.2 0 ± 0

HUM-ISM-CAB3 Palabrax humeralis 51.0 ± 12.2 0 ± 0 49.0 ± 12.2 0 ± 0

HUM-COL-CCH2 Concholepas concholepas 78.9 ± 5.5 0 ± 0 21.1 ± 5.5 0 ± 0

HUM-COL-CCH1 Concholepas concholepas 99.4 ± 0.2 0 ± 0 0.6 ± 0.2 0 ± 0

HUM-PPU2-ISM Perumytilus purpatus 20.9 ± 6.7 0.1 ± 0 79.0 ± 6.7 0 ± 0

HUM-COL-PPU2-B2 Perumytilus purpatus 73.1 ± 8.2 0 ± 0 26.9 ± 8.2 0 ± 0

HUM-COL-PPU1 Perumytilus purpatus 83.1 ± 3.6 16.9 ± 3.6 0.1 ± 0 0 ± 0

HUM-COL-PPU2 Perumytilus purpatus 47.5 ± 7.3 52.5 ± 7.3 0 ± 0 0 ± 0

HUM-PPU1-ISM Perumytilus purpatus 99.9 ± 0 0.1 ± 0 0 ± 0 0 ± 0

HUM-G5KC6-B2 Taleipus sp. 0.6 ± 0.2 0 ± 0 99.4 ± 0.2 0 ± 0

HUM-G5KC3-B2 Taleipus sp. 2.3 ± 0.5 0 ± 0 97.7 ± 0.5 0 ± 0

HUM-G15KC3-B2 Taleipus sp. 5.0 ± 1.1 0.1 ± 0 95.0 ± 1.1 0 ± 0

HUM-G5KC9-B2 Taleipus sp. 7.0 ± 4.4 0 ± 0 91.6 ± 4.9 1.4 ± 1.2

HUM-G15KC2-B2 Taleipus sp. 65.1 ± 5.8 0.7 ± 0.2 34.2 ± 5.8 0 ± 0

HUM-G15KC1-B2 Taleipus sp. 0 ± 0 0 ± 0 100.0 ± 0 0 ± 0

HUM-ISM-TAT2-B2 Tegula atra 0.1 ± 0.1 0 ± 0 99.9 ± 0.1 0 ± 0

HUM-COL-TAT1-B2 Tegula atra 3.2 ± 2.0 0 ± 0 96.7 ± 2.0 0 ± 0

HUM-ISM-TAT1-B2 Tegula atra 5.7 ± 4.2 0 ± 0 94.3 ± 4.2 0 ± 0

HUM-TAT1-ISM Tegula atra 35 ± 9.9 0.2 ± 0.1 64.8 ± 10.0 0 ± 0

HUM-COL-TAT2 Tegula atra 66.6 ± 9.3 28.3 ± 8.8 5.1 ± 1.9 0 ± 0

HUM-COL-TAT1 Tegula atra 12.9 ± 5.7 85.8 ± 6.3 1.3 ± 0.8 0 ± 0

HUM-TAT2-ISM Tegula atra 98.5 ± 0.5 0.9 ± 0.4 0.7 ± 0.3 0 ± 0

HUM-ISM1-TNI3-B3 Tetrapygus niger 6.0 ± 1.8 0 ± 0 94.0 ± 1.9 0 ± 0

HUM-ISM2-TNI3-B3 Tetrapygus niger 10.6 ± 4.2 0 ± 0 89.4 ± 4.2 0 ± 0

HUM-COL1-TNI2-B3 Tetrapygus niger 62.5 ± 10.6 9.4 ± 4.4 28.1 ± 9.4 0 ± 0

HUM-COL2-TNI2-B3 Tetrapygus niger 79.5 ± 3.9 20.2 ± 3.8 0.3 ± 0.1 0 ± 0

HUM-ISM-ZOOP-PEL3-B2 Zooplankton 98.2 ± 1.0 0 ± 0 1.8 ± 1.0 0 ± 0

152

Appendix 4

1. AA d13C correction equations.

The reagents used during derivatization add carbon to the side chains of amino acids, and so the raw d13C measured via GC-C-IRMS is a combination of original amino acid carbon and reagent carbon. However, because amino acid standards of known d13C composition were derivatized and run with each batch of samples, we were able to correct for this carbon addition for each amino acid using the following equation (From O’Brien et al., 2002,

Newsome et al., 2011; 2014):

,- ,- ( ,- ) ,- δ $/%& − δ $/%0 + δ $%0/ × p%0/ δ $%&'()* = p%0/

13 Here, d Cdsa refers to the measured value of the derivatized amino acid within the sample

13 and d Cdst refers to the measured value of the same derivatized amino acid within the

The in-house reference material derivatized with each batch of samples is a mixture of 13 powdered AAs purchased from Sigma Aldrich (Saint Louis, MO). The d13C values for these underivatized amino acids were measured via EA-IRMS at UNM-CSI. Each amino acid was then dissolved at a concentration of ~250 mM into weak acid and the dissolved AA solutions mixed together. For each batch of samples, a small aliquot of this solution was dried and then derivatized alongside each batch of unknown samples.

153

Table 1. Pairwise comparisons of marine primary producer AA d13C values. Comparisons were done with Tukey’s HSD following ANOVA. Rows are pairwise comparisons, columns are amino acids organized by biochemical family. Bold number indicate pairwise comparisons that are significant at an adjusted p-value ≤ 0.05. ‘Phosp.’ is short for phosphoenolpyruvate. Abbreviations are as follows: alanine (Ala), valine (Val), leucine (Leu), aspartic acid (Asp), threonine (Thr), lysine (Lys), isoleucine (Ile), glycine (Gly), serine (Ser), proline (Pro), glutamic acid (Glu), phenylalanine (Phe). *Denotes amino acids considered essential for animals. Species codes are as follows: Green = Ulva sp., POM = particulate organic matter and phytoplankton, Kelp = Laminaria sp., Lessonia sp., and Macrocystis pyrifera, Red = Neorhodomela sp., Plocamium sp., Rhodymenia sp., and one singleton each of crustose coralline algae and an unidentified red alga.

3-Phosphoglycerate a-Ketoglutarate Phosp. Pyruvate Family Oxaloacetate Family Pairwise Family Family Family Comparison Ala Val* Leu* Asp Thr* Lys* Ile* Gly Ser Pro Glu Phe*

Kelp-Green 0.87 0.05 0.96 0.23 0.00 0.19 0.80 0.63 1.00 0.71 0.95 0.98

POM-Green 0.24 0.14 0.42 0.01 0.56 0.69 0.62 0.07 0.02 0.34 0.22 0.44

Red-Green 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

POM-Kelp 0.47 0.00 0.09 0.00 0.00 0.01 0.95 0.29 0.00 0.01 0.03 0.13

Red-Kelp 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Red-POM 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.01 0.53 0.00 0.00 0.00

154

Table 2. Summary of results from generalized linear models (GLM). Alternate models and associated Akaike information 13 criterion values are shown for individual AA d C values, as well as linear discriminant coordinates (AAESS only model) for marine primary producers. Bold text represents the lowest AIC, and thus best supported model. Shorthand for the effects: Phyla = taxonomic grouping of each sample, CCM = presence or absence of carbon concentrating mechanisms, Ocea = oceanographic conditions of each site, AvgWaterT = mean water temperature of each site (°C), SDWaterT = standard deviation around mean water temperature, and Time = time period of each sample. Amino acids are organized by the main metabolic intermediary they are derived from: alanine (Ala), valine (Val), leucine (Leu), aspartic acid (Asp), threonine (Thr), lysine (Lys), isoleucine (Ile), glycine (Gly), serine (Ser), proline (Pro), glutamic acid (Glu), phenylalanine (Phe). *Denotes amino acids considered essential for animals. ‘Phosp.’ is short for phosphoenolpyruvate. See Table 4.1 for the values of the main and interaction effects included.

AA 3-Phosphoglycerate Phosp. Pyruvate Derived Oxaloacetate Derived a-Ketoglutarate Derived Fingerprint Derived Derived Coordinates

Ala *Val *Leu Asp *Thr *Lys *Ile *Phe Model: Gly d13C Ser d13C Pro d13C Glu d13C LD1 LD2 d13C d13C d13C d13C d13C d13C d13C d13C

Response ~ 1 (Null) 862.7 848.3 803.6 839.7 896.9 849.2 840.1 840.1 978.0 815.5 837.6 781.9 568.2 471.0

Response ~ Phlya 809.7 763.1 772.0 752.4 788.0 769.1 776.6 810.7 949.7 751.9 767.0 731.1 366.3 366.3

Response ~ CCM 739.5 727.9 677.6 744.2 781.4 726.2 679.2 768.8 917.1 704.2 703.9 653.7 536.7 431.8

Response ~ Ocea 866.3 844.8 803.8 839.4 899.1 852.5 843.5 842.3 970.6 813.6 840.5 783.7 568.9 467.4

Response ~ 833.9 820.2 770.2 814.5 876.4 836.1 817.0 824.6 961.3 791.0 820.9 755.0 565.3 467.5 AvgWaterT*SDWaterT

Response ~ Time 854.1 829.3 792.0 817.6 879.9 838.0 833.7 837.0 977.3 796.3 820.3 774.3 544.9 472.2

Response ~ CCM + AvgWaterT 700.9 694.8 632.6 727.5 761.1 710.9 638.1 760.9 911.4 675.7 689.2 623.3 537.8 419.8 * SDWaterT

Response ~ Phyla + AvgWaterT 756.7 712.0 727.5 707.0 738.7 738.3 723.5 792.8 936.0 734.2 734.2 685.0 362.0 355.4 * SDWaterT

Response ~ CCM + Time + 701.4 686.1 632.9 709.9 750.3 696.3 639.2 762.9 913.4 665.3 669.9 624.7 518.9 420.6 AvgWaterT * SDWaterT

Response ~ CCM + Phyla + 695.8 649.8 631.2 675.8 678.8 678.4 632.3 760.9 901.5 673.1 673.1 621.1 360.2 350.5 AvgWaterT * SDWaterT

155

Response ~ CCM + Time + Phyla + AvgWaterT * 697.0 650.0 630.6 668.6 680.1 674.1 633.9 762.8 903.5 660.0 660.0 623.1 359.4 351.7 SDWaterT

156

Table 3. Results for marine producer groups of linear regression analysis between amino acid – precursors and products. Where regression coefficients are reported, p-values were <0.01. Superscripts for intercept estimates indicate statistical groupings at the 0.05 level. Confidence intervals around the intercept were not reported for Scott et al. (2006), thus we only compared them qualitatively with our measured marine producers. POM = particulate organic matter. *In a few cases, samples were excluded from comparisons due to missing values.

Val-Ala Leu-Ala Pro-Glu

Producer: N* Intercept Intercept 95% CI R2 Intercept Intercept 95% CI R2 Intercept Intercept 95% CI R2

Green 24 -9.9a 0.53 0.72 -16.6a 0.50 0.63 -0.7a 0.68 0.73

Kelp 57 -9.1b 0.22 0.61 -15.8a 0.23 0.59 -2.7b 0.44 0.41

POM 25 -14.9c 0.47 0.64 -20.9b 0.44 0.49 -7.7c 0.97 0.30

Red 24 -12.0d 0.62 0.86 -11.3c 0.59 0.90 -0.8a 0.73 0.89

Microorganisms 29 -13.1e -- 0.9 -14.7 -- 0.8 -1.7 -- 0.9

Thr-Asp Lys-Asp Ile-Asp

Producer: N* Intercept Intercept 95% CI R2 Intercept Intercept 95% CI R2 Intercept Intercept 95% CI R2

Green 24 -0.1a 0.86 0.46 -7.0a 1.29 0.19 -6.7a 0.65 0.57

Kelp 57 0.8a 0.23 0.43 -7.0a 0.17 0.60 -11.4b 0.18 0.56

POM 25 0.6a 0.79 0.63 -5.4a 0.62 0.72 -8.5c 0.48 0.70

Red 24 2.8b 0.85 0.85 -3.6b 0.77 0.86 -1.2d 0.83 0.88

Microorganisms 29 -2.7 -- 0.9 -7.7 -- 0.9 ------

Ser-Gly Ser-Ala Asp-Ala

Producer: N* Intercept Intercept 95% CI R2 Intercept Intercept 95% CI R2 Intercept Intercept 95% CI R2

Green 24 -- -- 0.03 -- -- 0.05 -2.4a 0.32 0.82

Kelp 57 7.0a 0.56 0.45 1.1a 0.85 0.09 2.5b 0.24 0.71

POM 25 5.7a 2.70 0.28 -- -- 0.08 -5.9c 0.61 0.58

Red 24 6.9a 1.48 0.76 10.3b 1.57 0.76 -6.4c 0.63 0.79

Microorganisms 29 -0.5 -- 0.6 4.8 -- 0.7 -2.6a -- 1

157

Table 4. Coefficients of linear discriminants for amino acid d13C values. Presented are results from the model with all amino acids included. Individual amino acids are organized by the main metabolic intermediary they are derived from. For each LD axis, the three most influential essential amino acids are indicated in bold. The relative proportion of trace explained is: LD1 (0.7481), LD2 (0.2153), LD3 (0.0365). Precursor AA LD1 LD2 LD3 Alanine -0.162 -0.093 0.062 Pyruvate Valine 0.322 0.064 -0.045 Leucine -0.199 0.621 -0.077

Aspartic Acid -0.068 0.134 0.366 Threonine 0.409 0.105 -0.213 Oxaloacetate Lysine 0.103 -0.082 -0.067 Isoleucine -0.387 -0.502 -0.038 Glycine -0.125 -0.001 0.118 3-Phosphoglycerate Serine 0.021 0.022 0.061 Proline 0.172 -0.051 0.097 a-Ketoglutarate Glutamic Acid 0.111 -0.128 -0.136

Phosphoenolpyruvate Phenylalanine -0.054 -0.180 -0.052

158

13 Table 5. Coefficients of linear discriminants for amino acid d C values – AAESS only. Presented are results from the model with only essential amino acids included. Individual amino acids are organized by the main metabolic intermediary they are derived from. For each LD axis, the three most influential essential amino acids are indicated in bold. The relative proportion of trace explained is: LD1 (0.7427), LD2 (0.2495), LD3 (0.0079). Precursor AA LD1 LD2 LD3

Valine 0.349 -0.016 0.241 Pyruvate Leucine -0.241 -0.549 0.192

Threonine 0.333 -0.179 -0.204 Oxaloacetate Lysine 0.104 0.119 -0.066 Isoleucine -0.349 0.644 -0.065

Phosphoenolpyruvate Phenylalanine -0.082 0.073 0.046

159

References

Abbott, I. A., A. Isabella, and G. J. Hollenberg. (1992). Marine algae of California. Stanford University Press, Stanford.

Ainis, A. F. et al. (2014). Using non-dietary gastropods in coastal shell middens to infer kelp and seagrass harvesting and paleoenvironmental conditions. Journal of Archaeological Science 49:343–360.

Akashi, H. and T. Gojobori. (2002). Metabolic efficiency and amino acid composition in the proteomes of Escherichia coli and Bacillus subtilis. Proceedings of the National Academy of Sciences 99:3695–3700.

Ambrose, S. H. (1990). Preparation and characterization of bone and tooth collagen for isotopic analysis. Journal of Archaeological Science 17:431–451.

Amelung, W. and X. Zhang. (2001). Determination of amino acid enantiomers in soils. Soil Biology and Biochemistry 33:553–562.

Anderson, M. J., and T. J. Willis. (2003). Canonical analysis of principal coordinates: a useful method of constrained ordination for ecology. Ecology 84:511–525.

Anderson, M. J., R. N. Gorley, and K. R. Clarke. (2008). PERMANOVA+ for PRIMER: Guide to Software and Statistical Methods. PRIMER-E: Plymouth, UK.

Araújo, M. S., D. I. Bolnick, and C. A. Layman. (2011). The ecological causes of individual specialisation. Ecology Letters 14:948–958.

Arthur, K. E., S. Kelez, T. Larsen, C. A. Choy, and B. N. Popp. (2014). Tracing the biosynthetic source of essential amino acids in marine turtles using δ13C fingerprints. Ecology 95:1285–1293.

Baringhaus, L. and C. Franz. (2004). On a new multivariate two-sample test. Journal of Multivariate Analysis 88:190–206.

Ben-David, M. and E. A. Flaherty. (2012). Stable isotopes in mammalian research: a beginner's guide. Journal of Mammalogy 93:312–328.

Blanchard, J. L. et al. (2011). Coupled energy pathways and the resilience of size- structured food webs. Theoretical Ecology 4:289–300.

Bodkin, J. L. (2015). Historic and contemporary status of sea otters in the North Pacific. Pp 43-61 in S. E. Larson, J. L. Bodkin, and G. R. VanBlaricom (Eds.), Sea Otter Conservation, Academic Press, New York.

160

Bograd, S. J. et al. (2009). Phenology of coastal upwelling in the California Current. Geophysical Research Letters 36:L01602

Bolnick, D. I. et al. (2003). The ecology of individuals: incidence and implications of individual specialization. American Naturalist 161:1–28.

Bradley, C. J. et al. (2015). Trophic position estimates of marine teleosts using amino acid compound specific isotopic analysis. Limnology and Oceanography 13:476–93.

Braje, T. J., Erlandson, J. M. and T. C. Rick. (2007). An historic Chinese abalone fishery on California’s northern Channel Islands. Historical Archaeology 41:117–28.

Breschini, G. S. and T. Haversat. (2011). A revised culture sequence for the Monterey Peninsula Area, California. Pacific Coast Archaeological Society Quarterly 44:1–24.

Budge, S. M. et al. (2008). Tracing carbon flow in an arctic marine food web using fatty acid-stable isotope analysis. Oecologia 157:117–129.

Burrows, M. T. and R. N. Hughes. (1990). Variation in growth and consumption among individuals and populations of dogwhelks, Nucella lapillus: a link between foraging behaviour and fitness. Journal of Animal Ecology 59:723–742.

Carswell, L. P., S. G. Speckman, and V. A. Gill. (2015). Shellfish fishery conflicts and perceptions of sea otters in California and Alaska. Pp 333-368 in S. E. Larson, J. L. Bodkin, and G. R. VanBlaricom (Eds.), Sea Otter Conservation, Academic Press, New York.

Catton, C., L. Rogers-Bennett, and A. Amrhein. (2016). Perfect storm decimates northern California kelp forests. California Department of Fish and Wildlife Marine Management News (https://cdfwmarine.wordpress.com/2016/03/30/perfect-storm- decimates-kelp/; accessed 07.01.2019).

Ceballos, G. et al. (2015). Accelerated modern human–induced species losses: entering the sixth mass extinction. Science Advances 1(5):e1400253.

Chikaraishi, Y. et al. (2014). High‐resolution food webs based on nitrogen isotopic composition of amino acids. Ecology and Evolution 4:2423-2449.

Clark, C. T., L. Horstmann, and N. Misarti. (2017). Quantifying variability in stable carbon and nitrogen isotope ratios within the skeletons of marine mammals of the suborder Caniformia. Journal of Archaeological Science 15:393–400.

Cullen, J. T., Y. Rosenthal, and P. G. Falkowski. (2001). The effect of anthropogenic CO2 on the carbon isotope composition of marine phytoplankton. Limnology and Oceanography 46:996–998.

161

Dayton, P. K. (1985). Ecology of kelp communities. Annual Review of Ecology and Systematics 16:215-245.

Dayton, P. K. et al. (1998). Sliding baselines, ghosts, and reduced expectations in kelp forest communities. Ecological Applications 8:309–22.

Dean, T. A, J. L. Bodkin, and H. A. Coletti. (2014). Protocol narrative for nearshore marine ecosystem monitoring in the Gulf of Alaska: version 1.1. Natural Resource Report NPS/SWAN/NRR – 2014/756, Fort Collins.

Docmac, F., M. et al. (2017). Habitat coupling writ large: pelagic-derived materials fuel benthivorous macroalgal reef fishes in an upwelling zone. Ecology 98:2267–2272.

Duggins, D. O., C. A. Simenstad, and J. A. Estes. (1989). Magnification of secondary production by kelp detritus in coastal marine ecosystems. Science 245:170–173.

Dunton, K. H., S. V. Schonberg, and L. W. Cooper. (2012). Food web structure of the Alaskan nearshore shelf and estuarine lagoons of the Beaufort Sea. Estuaries and Coasts 35:416–435.

Ebert, T. A. (1968). Growth rates of the sea urchin Strongylocentrotus purpuratus related to food availability and spine abrasion. Ecology 49:1075–1091.

Elliott Smith, E. A. et al. (2015). The cost of reproduction: differential resource specialization in female and male California sea otters. Oecologia 178:17–29.

Elliott Smith, E. A., C. Harrod, and S. D. Newsome. (2018). The importance of kelp to an intertidal ecosystem varies by trophic level: insights from amino acid δ13C analysis. Ecosphere 9:e02516.

Erlandson, J. M. (1991). The antiquity of CA-SMI-1: A multicomponent site on San Miguel Island. Journal of California and Great Basin Anthropology 13:273–279.

Erlandson, J. M. et al. (2005). Sea otters, shellfish, and humans: 10,000 years of ecological interaction on San Miguel Island, California. Pp 58-69 in D. K. Garcelon and C. Schwemm (Eds.), Proceedings of the Sixth California Islands Symposium, Institute for Wildlife Studies, Arcata.

Estes, J. A. and J. F. Palmisano. (1974). Sea otters: their role in structuring nearshore communities. Science 185:1058–1060.

Estes, J. A. et al. (2003). Individual variation in prey selection by sea otters: patterns, causes and implications. Journal of Animal Ecology 72:144–55.

Estes, J. A. et al. (2011). Trophic downgrading of planet Earth. Science 333:301–306.

162

Fantle, M. S. et al. (1999). A food web analysis of the juvenile blue crab, Callinectes sapidus, using stable isotopes in whole animals and individual amino acids. Oecologia 120:416–426.

Fariña, J. M., A. T. Palma, and F. P. Ojeda. (2008). Subtidal kelp associated communities off the temperate Chilean coast. Pp 79-102 in T. McClananhan and G. M. Branch (Eds.), Food Webs and the Dynamics of Marine Reefs, Oxford University Press, Oxford.

Farquhar, G. D., J. R. Ehleringer, and K. T. Hubick. (1989). Carbon isotope discrimination and photosynthesis. Annual Review of Plant Biology 40:503–537.

Fogel, M. L. and L. A. Cifuentes. (1993). Isotope fractionation during primary production. Pp. 73-98 in M. H. Engel and S. A. Macko (Eds.), Organic Geochemistry, Plenum Press, New York.

Fogel, M. L., and N. Tuross. (2003). Extending the limits of paleodietary studies of humans with compound specific carbon isotope analysis of amino acids. Journal of Archaeological Science 30:535–545

Foley, M. M., and P. L. Koch. (2010). Correlation between allochthonous subsidy input and isotopic variability in the giant kelp Macrocystis pyrifera in central California, USA. Marine Ecology Progress Series 409:41–50.

Fox, M. D. (2013). Resource translocation drives δ13C fractionation during recovery from disturbance in giant kelp, Macrocystis pyrifera. Journal of Phycology 49:811–815.

Fox, M. D. et al. (2019). Trophic plasticity in a common reef‐building coral: insights from δ13C analysis of essential amino acids. Functional Ecology 33:2203–2214.

Fredriksen, S. (2003). Food web studies in a Norwegian kelp forest based on stable isotope (δ13C and δ15N) analysis. Marine Ecology Progress Series 260:71–81.

Fry, B. (1988). Food web structure on Georges Bank from stable C, N, and S isotopic compositions. Limnology and Oceanography 33:1182–1190.

Gifford-Gonzalez, D. (2007). A coordinated approach to site formation analysis in vertically dispersed deposits. 72nd Annual Meeting of the Society for American Archaeology, Austin. Published Abstract.

Gifford-Gonzalez, D. (2011). Holocene Monterey Bay Fur Seals. Pp 221-242 in T. J. Braje and T. C. Rick (Eds.), Human Impacts on Seals, Sea Lions, and Sea Otters: Integrating Archaeology and Ecology in the Northeast Pacific, University of California Press, Berkeley.

163

Gooding, R. A, C. D. Harley, and E. Tang. (2009). Elevated water temperature and carbon dioxide concentration increase the growth of a keystone echinoderm. Proceedings of the National Academy of Sciences 106:9316–9321.

Graham, M. H., B. S. Halpern, and M. H. Carr. (2008). Diversity and dynamics of Californian subtidal kelp forests. Pp 103-134 in T. R. McClanahan and G. M. Branch (Eds.), Food webs and the dynamics of marine reefs, Oxford University Press, Oxford.

Hobson, K. A., J. F. Piatt, and J. Pitocchelli. (1994). Using stable isotopes to determine seabird trophic relationships. Journal of Animal Ecology 63:786-798.

Hobson, K. A., R. Barnett-Johnson, and T. Cerling. (2010). Using isoscapes to track animal migration. Pp 273-298 in J. B. West et al. (Eds.), Isoscapes, Springer, Dordrecht.

Howland, M. R. et al. (2003). Expression of the dietary isotope signal in the compound- specific δ13C values of pig bone lipids and amino acids. International Journal of Osteoarchaeology 13:54-65.

Hughes, B. B. et al. (2016). Trophic cascades on the edge: fostering seagrass resilience via a novel pathway. Oecologia 182:231–241.

Hylkema, M. G. (1991). Prehistoric native American adaptations along the central California coast of San Mateo and Santa Cruz counties. Masters Thesis. San Jose State University, San Jose.

Hylkema, M. G. (2019). The demise of archaeological site CA-SMA-238: sea level rise and its effects to an ancestral Native American archaeological site at Año Nuevo State Park, San Mateo County, California. Manuscript on File with the Santa Cruz District Archaeological Research Facility, California State Parks.

Jackson, J. B. et al. (2001). Historical overfishing and the recent collapse of coastal ecosystems. Science 293:629–637.

Jackson, A. L. et al. (2011). Comparing isotopic niche widths among and within communities: SIBER–Stable Isotope Bayesian Ellipses in R. Journal of Animal Ecology 80:595–602.

Jarman, C. L. et al. (2017). Diet of the prehistoric population of Rapa Nui (Easter Island, Chile) shows environmental adaptation and resilience. American Journal of Physical Anthropology 164:343–361.

Jones, T. L. et al. (2011). Toward a prehistory of the southern sea otter (Enhydra lutris nereis). Pp 243-272 in T. J. Braje, and T. C. Rick (Eds.), Human and Marine Ecosystems: Archaeology and Historical Ecology of Northeastern Pacific Seals, Sea Lions and Sea Otters, University of California Press, Berkeley.

164

Kanehisa, M. and S. Goto. (2000). KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Research 28:27–30.

Kiljunen, M. et al. (2006). A revised model for lipid-normalizing δ13C values from aquatic organisms, with implications for isotope mixing models. Journal of Applied Ecology 43:1213–22.

Koenigs, C., R. J. Miller, and H. M. Page. (2015). Top predators rely on carbon derived from giant kelp Macrocystis pyrifera. Marine Ecology Progress Series 537:1–8.

Kondoh, M. (2003). Foraging adaptation and the relationship between food-web complexity and stability. Science 299:1388–1391

Kusche, H. et al. (2014). Crater lake cichlids individually specialize along the benthic– limnetic axis. Ecology and Evolution 4:1127–1139.

Lafferty, K. D. and M. T. Tinker. (2014). Sea otters are recolonizing southern California in fits and starts. Ecosphere 5:1–11.

Larsen. T. et al. (2009). Stable isotope fingerprinting: a novel method for identifying plant, fungal, or bacterial origins of amino acids. Ecology 90:3526–3535.

Larsen, T. et al. (2012). Can amino acid carbon isotope ratios distinguish primary producers in a mangrove ecosystem? Rapid Communications in Mass Spectrometry 26:1541– 1548.

Larsen, T. et al. (2013). Tracing carbon sources through aquatic and terrestrial food webs using amino acid stable isotope fingerprinting. PLoS ONE 8:e73441.

Larsen, T. et al. (2015). Assessing the potential of amino acid 13 patterns as a carbon source tracer in marine sediments: effects of algal growth conditions and sedimentary diagenesis. Biogeosciences 12:4979–4992.

Layman, C. A. et al. (2012). Applying stable isotopes to examine food-web structure: an overview of analytical tools. Biological reviews of the Cambridge Philosophical Society 87:545–62.

Liew, J. H. et al. (2019). Quantifying terrestrial carbon in freshwater food webs using amino acid isotope analysis—case study with an endemic cave fish. Methods in Ecology and Evolution 10:1594–1605.

Maberly, S. C., J. A. Raven, and A. M. Johnston. (1992). Discrimination between 12 C and 13 C by marine plants. Oecologia 91:481-492.

Mann, K. H. (1973). Seaweeds: their productivity and strategy for growth. Science 182:975–981.

165

Martz, P. C. (2008). 4000 years on Ghalas-at: part nne of the San Nicolas Island index unit analysis program. Report submitted to Naval Air Weapons Station, China Lake.

Matthews, B. et al. (2010). Specialization of trophic position and habitat use by sticklebacks in an adaptive radiation. Ecology 91:1025–1034.

McCarthy, M. D. et al. (2007). Amino acid nitrogen isotopic fractionation patterns as indicators of heterotrophy in plankton, particulate, and dissolved organic matter. Geochimica et Cosmochimica Acta 71:4727-4744.

McCauley, D. J. et al. (2015). Marine defaunation: animal loss in the global ocean. Science 347:1255641.

McClelland, J. W. and J. P. Montoya (2002). Trophic relationships and the nitrogen isotopic composition of amino acids in plankton. Ecology 83:2173–2180.

McMahon, K. W., L. L. Hamady, and S. R. Thorrold. (2013). A review of ecogeochemistry approaches to estimating movements of marine animals. Limnology and Oceanography 58:697–714.

McMahon, K. W. et al. (2015). Carbon and nitrogen isotope fractionation of amino acids in an avian marine predator, the gentoo penguin (Pygoscelis papua). Ecology and Evolution 5:1278–1290.

McMahon, K. W. and M. D. McCarthy. (2016). Embracing variability in amino acid δ15N fractionation: mechanisms, implications, and applications for trophic ecology. Ecosphere 7:e01511.

McMahon, K. W. et al. (2016). Tracing carbon flow through coral reef food webs using a compound-specific stable isotope approach. Oecologia 180:809–821.

McMahon, K. W. and S. D. Newsome. (2019). Amino acid isotope analysis: a new frontier in studies of animal migration and foraging ecology. Pp 173-190 in K. A. Hobson and L. I. Wassenaar (Eds.), Tracking Animal Migration with Stable Isotopes, Academic Press, New York.

Misarti, N. et al. (2009). Changes in northeast Pacific marine ecosystems over the last 4500 years: evidence from stable isotope analysis of bone collagen from archeological middens. The Holocene 19:1139–1151.

Mora, C. et al. (2011). How many species are there on Earth and in the ocean? PLoS Biology 9:e1001127.

Nelson, D. L., A. L. Lehninger, and M. M. Cox. (2008). Lehninger Principles of Biochemistry. Macmillan Press, New York.

166

Newsome, S. D. et al. (2007). A niche for isotopic ecology. Frontiers in Ecology and the Environment 5:429–436.

Newsome, S. D. et al. (2009). Using stable isotopes to investigate individual diet specialization in California sea otters (Enhydra lutris nereis). Ecology 90:961–974.

Newsome, S. D, M. T. Clementz, and P. L. Koch. (2010). Using stable isotope biogeochemistry to study marine mammal ecology. Marine Mammal Science 26:509– 572.

Newsome, S. D. et al. (2011). Contributions of direct incorporation from diet and microbial amino acids to protein synthesis in Nile tilapia. Functional Ecology 25:1051–1062.

Newsome, S. D. et al. (2014). Amino acid δ13C analysis shows flexibility in the routing of dietary protein and lipids to the tissue of an omnivore. Integrative and Comparative Biology 5:890–902.

Newsome, S. D. et al. (2015). The interaction of intraspecific competition and habitat on individual diet specialization: a near range-wide examination of sea otters. Oecologia 178:45–59.

O'Brien, D. M., M. L. Fogel, and C. L. Boggs. (2002). Renewable and nonrenewable resources: amino acid turnover and allocation to reproduction in Lepidoptera. Proceedings of the National Academy of Sciences 99:4413–4418.

O’Connell, T. C. (2017). ‘Trophic’ and ‘source’amino acids in trophic estimation: a likely metabolic explanation. Oecologia 184:317–326.

Østbye, K. et al. (2006). Parallel evolution of ecomorphological traits in the European whitefish Coregonus lavaretus (L.) species complex during postglacial times. Molecular Ecology 15:3983–4001.

Page, H. M. et al. (2008). Assessing the importance of land and marine sources of organic matter to kelp forest food webs. Marine Ecology Progress Series 360:47–62.

Paine, R. T. (1969). A note on trophic complexity and community stability. American Naturalist 103:91–93.

Pauli, J. N., S. A. Steffan, and S. D. Newsome. (2015). It is time for IsoBank. BioScience 65:229–230.

Perkins, M. J. et al. (2018). Multichannel feeding by spider functional groups is driven by feeding strategies and resource availability. Oikos 127:23–33.

Phillips, D. L. et al. (2014). Best practices for use of stable isotope mixing models in food- web studies. Canadian Journal of Zoology 92:823–835.

167

Polis, G. A. and D. R. Strong. (1996). Food web complexity and community dynamics. American Naturalist 147:813–846.

Polis, G. A. et al. (1996). Time, space, and life history: influences on food webs. Pp 435- 460 in G. A. Polis and K. O. Winemiller (Eds.), Food webs. Integration of Patterns and Dynamics. Springer, Boston.

Popp, B. N. et al. (2007). Insight into the trophic ecology of yellowfin tuna, Thunnus albacares, from compound‐specific nitrogen isotope analysis of proteinaceous amino acids. Terrestrial Ecology 1:173–190.

Post, D. M., M. E. Conners, and D. S. Goldberg. (2000). Prey preference by a top predator and the stability of linked food chains. Ecology 81:8–14.

Rau, G. H., M. D. Ohman, and A. Pierrot-Bults. (2003). Linking nitrogen dynamics to climate variability off central California: a 51 year record based on 15N/14N in CalCOFI zooplankton. Deep Sea Research Part II: Topical Studies in Oceanography 50:2431–2447.

Ravinet, M., P. A. Prodöhl, and C. Harrod. (2013). Parallel and nonparallel ecological, morphological and genetic divergence in lake–stream stickleback from a single catchment. Journal of Evolutionary Biology 26:186–204.

Raven, J. A. et al. (2002). Mechanistic interpretation of carbon isotope discrimination by marine macroalgae and seagrasses. Functional Plant Biology 29:355–378.

Raven, J. A., C. S. Cockell, and C. L. De La Rocha. (2008). The evolution of inorganic carbon concentrating mechanisms in photosynthesis. Philosophical Transactions of the Royal Society B: Biological Sciences 363:2641–2650.

Raven, J. A. and M. Giordano. (2017). Acquisition and metabolism of carbon in the Ochrophyta other than diatoms. Philosophical Transactions of the Royal Society B: Biological Sciences 372: 20160400.

Reed, R. K. and J. D. Schumacher. (1986). Physical Oceanography. Pp 57-75 in D. W. Hood and S. T. Zimmerman (Eds.), The Gulf of Alaska: Physical Environment and Biological Resources, Ocean Assessment Division, National Oceanic and Atmospheric Administration.

Reed, D. C., A. Rassweiler, and K. K. Arkema. (2008). Biomass rather than growth rate determines variation in net primary production by giant kelp. Ecology 89:2493–2505.

Reddin, C. J. et al. (2015). Coastal upwelling drives intertidal assemblage structure and trophic ecology. PloS ONE 10: e0130789.

168

Reinfelder, J. R. (2011). Carbon concentrating mechanisms in eukaryotic marine phytoplankton. Annual Review of Marine Science 3:291–315.

Rick, T. C. et al. (2005). From Pleistocene mariners to complex hunter-gatherers: The archaeology of the California Channel Islands. Journal of World Prehistory 19:169– 228.

Rick, T. C. and R. Lockwood. (2013). Integrating paleobiology, archeology, and history to inform biological conservation. Conservation Biology 27:45–54.

Riedman, M. and J. A. Estes. (1990). The sea otter (Enhydra lutris): behavior, ecology, and natural history. Biological Report (USA) 90:1-126.

Robinson, B. W. and D. S. Wilson (1994). Character release and displacement in fishes: a neglected literature. American Naturalist 144:596–627.

Rooney, N. et al. (2006). Structural asymmetry and the stability of diverse food webs. Nature 442:265–269.

Rooney, N. and K. S. McCann. (2012). Integrating food web diversity, structure and stability. Trends in Ecology and Evolution 27:40–46.

Rowe, A. G. et al. (2019). Sources of primary production to Arctic bivalves identified using amino acid stable carbon isotope fingerprinting. Isotopes in Environmental and Health Studies 55:1–19.

Ruiz-Cooley, R. I. et al. (2014). Carbon and nitrogen isotopes from top predator amino acids reveal rapidly shifting ocean biochemistry in the outer California Current. PloS ONE 9:e110355.

Seapy, R. R. and M. M. Littler. (1980). Biogeography of rocky intertidal macroinvertebrates of the Southern California Islands. Pp 307-323 in D. M. Power (Ed.), The California Islands: Proceedings of a Multidisciplinary Symposium. Santa Barbara Museum of Natural History, Santa Barbara.

Scott, J. H. et al. (2006). An examination of the carbon isotope effects associated with amino acid biosynthesis. Astrobiology 6:867–880.

Silfer, J. A. et al. (1991). Stable carbon isotope analysis of amino acid enantiomers by conventional isotope ratio mass spectrometry and combined gas chromatography/isotope ratio mass spectrometry. Analytical Chemistry 63:370–374.

Simenstad, C. A., D. O. Duggins, and P. D. Quay. (1993). High turnover of inorganic carbon in kelp habitats as a cause of δ13C variability in marine food webs. Marine Biology 116:147–160.

169

Steinberg, P. D., J.A. Estes, and F. C. Winter. (1995). Evolutionary consequences of food chain length in kelp forest communities. Proceedings of the National Academy of Sciences 92:8145–8148.

Steneck, R. S. et al. (2002). Kelp forest ecosystems: biodiversity, stability, resilience and future. Environmental Conservation 29:436–459.

Stepien, C. C., C. A. Pfister, and J. T. Wootton. (2016). Functional traits for carbon access in macrophytes. PloS ONE 11:e0159062.

Stock, B. C., and B. X. Semmens. (2016). MixSIAR GUI User Manual. Version 3.1. https://github.com/brianstock/MixSIAR.

Szpak, P. et al. (2012). Historical ecology of late Holocene sea otters (Enhydra lutris) from northern British Columbia: isotopic and zooarchaeological perspectives. Journal of Archaeological Science 39:1553–1571.

Tabachnick, B. G., and L. S. Fidell. (2013). Using multivariate statistics. Pearson Education Inc, Upper Saddle River.

Tinker, M. T., G. Bentall, and J. A. Estes. (2008). Food limitation leads to behavioral diversification and dietary specialization in sea otters. Proceedings of the National Academy of Sciences 105:560–565.

Tinker, M. T. et al. (2012). Structure and mechanism of diet specialisation: testing models of individual variation in resource use with sea otters. Ecology Letters 15:475–483.

Tinker, M. T. et al. (2016). Dramatic increase in sea otter mortality from white sharks in California. Marine Mammal Science 32:309–326.

Tinker, M. T. and B. B. Hatfield. (2018). California sea otter (Enhydra lutris nereis) census results, Spring 2018. US Geological Survey 1067.

Tinker, M. T. et al. (2019). Southern sea otter (Enhydra lutris nereis) population biology at Big Sur and Monterey, California-Investigating the consequences of resource abundance and anthropogenic stressors for sea otter recovery. US Geological Survey 1022.

Thiel, M. et al. (2007). The Humboldt Current System of northern and central Chile: oceanographic processes, ecological interactions and socioeconomic feedback. Oceanography and Marine Biology: An Annual Review 45:195-344.

Thorp, J. H., and R. E Bowes. (2017). Carbon sources in riverine food webs: new evidence from amino acid isotope techniques. Ecosystems 20:1029–1041.

170

Vargas, C. A. et al. (2007). The relative importance of microbial and classical food webs in a highly productive coastal upwelling area. Limnology and Oceanography 52:1495– 510.

Vanderklift, M. A and S. Ponsard. (2003). Sources of variation in consumer-diet δ15N enrichment: a meta-analysis. Oecologia 136:169–182.

Vokhshoori, N. L. and M. D. McCarthy. (2014). Compound-specific δ15N amino acid measurements in littoral mussels in the California upwelling ecosystem: a new approach to generating baseline δ15N isoscapes for coastal ecosystems. PloS ONE 9:e98087. von Biela, V. R. et al. (2016). Widespread kelp-derived carbon in pelagic and benthic nearshore fishes suggested by stable isotope analysis. Estuarine, Coastal and Shelf Science 181:364–374.

Walker, P. L. et al. (2002). Archaeological investigations at the Point Bennett pinniped rookery on San Miguel Island. Pp. 628-632 in D. Browne, K. Mitchell, and H. Chaney (Eds.), Proceedings of the Fifth California Islands Symposium, Santa Barbara Museum of Natural History, Santa Barbara.

Ward, C. L., K. S. McCann, and N. Rooney. (2015). HSS revisited: multi‐channel processes mediate trophic control across a productivity gradient. Ecology Letters 18:1190–1197.

Wendt, I. (1968). Fractionation of carbon isotopes and its temperature dependence in the system CO2-gas-CO2 in solution and HCO3-CO2 in solution. Earth and Planetary Science Letters 4:64–68.

Werner, T. K. and T. W. Sherry. (1987). Behavioral feeding specialization in Pinaroloxias inornata, the “Darwin's finch” of Cocos Island, Costa Rica. Proceedings of the National Academy of Sciences 84:5506–5510.

Wessels, F. J. and D. A. Hahn. (2010). Carbon 13 discrimination during lipid biosynthesis varies with dietary concentration of stable isotopes: implications for stable isotope analyses. Functional Ecology 24:1017–1022.

Westberry, T. K. et al. (2016). Annual cycles of phytoplankton biomass in the subarctic Atlantic and Pacific Ocean. Global Biogeochemical Cycles 30:175–190.

Whiteman, J. P. et al. (2019). A guide to using compound-specific stable isotope analysis to study the fates of molecules in organisms and ecosystems. Diversity 11:8.

Wilmers, C. C. et al. (2012). Do trophic cascades affect the storage and flux of atmospheric carbon? An analysis of sea otters and kelp forests. Frontiers in Ecology and the Environment 10:409–415.

171

Wimp, G. M. et al. (2013). Predator hunting mode influences patterns of prey use from grazing and epigeic food webs. Oecologia 171:505–515.

Winemiller, K. O. and G. A. Polis (1996). Food webs: what can they tell us about the world?. Pp 1-22 in G. A. Polis and K. O. Winemiller (Eds.), Food Webs, Springer, Boston.

Wolkovich, E. M. et al. (2014). Linking the green and brown worlds: the prevalence and effect of multichannel feeding in food webs. Ecology 95:3376–3386.

Wollrab, S., S. Diehl, and A. M. De Roos. (2012). Simple rules describe bottom‐up and top‐down control in food webs with alternative energy pathways. Ecology Letters 15:935–946.

Wu, G. (2009). Amino acids: metabolism, functions, and nutrition. Amino Acids 37:1–17.

Yeakel, J. D. et al. (2013). The impact of climate change on the structure of Pleistocene food webs across the mammoth steppe. Proceedings of the Royal Society B: Biological Sciences 280:20130239.

Yorke, C. E., H. M. Page, and R. J. Miller. (2019). Sea urchins mediate the availability of kelp detritus to benthic consumers. Proceedings of the Royal Society B 286: 20190846.

172