PICKY EATER OR GENERALIST FEEDER? DIET DIVERSITY AND FUNCTIONAL HOMOGENIZATION IN HERBIVOROUS REEF FISHES

A DISSERTATION SUBMITTED TO THE GRADUATE DIVISION OF THE UNIVERSITY OF HAWAIʻI AT MĀNOA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

IN

MARINE BIOLOGY (SPECIALIZATION IN ECOLOGY, EVOLUTION, AND CONSERVATION BIOLOGY)

May 2020

By Eileen M. Nalley

Dissertation Committee: Robert J. Toonen, Chairperson Megan J. Donahue Mark A. Hixon Brian W. Bowen Amber N. Wright

Keywords: herbivore, specialization, metabarcoding, homogenization

© Copyright 2020 – Eileen Nalley

All Rights Reserved.

i

DEDICATION

This work is dedicated to Dr. Stephen Karl, who gave me the chance to get started, to Ms. Susan Tittlebaum and Dr. Mary Gubala, who inspired me to pursue a career in science, and to my family, who supported and encouraged me every step of the way.

ii ACKNOWLEDGEMENTS

I would first like to thank Steve Karl for welcoming me into his lab and for opening the door for me to the University of Hawaiʻi. I will always be grateful for this opportunity. I also appreciate the mentorship of Rob Toonen and Megan Donahue. Without them, none of this work would have been possible. Rob always has an answer in a time of crisis or a joke in times of calm. He gives so much to his work and to his students, and it has been such a privilege to work with him. Megan’s willingness to tackle new ideas and challenges serves as a constant reminder that the joy of learning never ends, and I am so thankful for her humor and guidance. I would also like to thank the other members of my committee. Brian, who along with Rob, welcomed me into the ToBo Lab and provided an amazing example of what it means to be a great person, in addition to being a great scientist. Mark’s willingness to communicate outside of academia is an inspiration to students trying to navigate the murky boundaries of science and policy that we face at present. Amber has been supportive and encouraging from the beginning, giving feedback that is always insightful and transcends the land-sea boundary. In addition, I would like to thank Adel Heenan, who has been an integral part of much of this dissertation work. Her positivity is a wonder and an inspiration, and I am so fortunate to have had the chance to work with her. Molly Timmers and Rachael Wade have also provided so much feedback and guidance throughout this process. They blazed the way in metabarcoding, and without them I would have been lost. I also owe a huge debt of gratitude to Richard Coleman for being willing to share and to Derek Kraft and Julie Zill for their expert fishing skills. I also received help from dedicated undergraduate research assistants (Luis Espinosa, Ryan Shiesha, Sydney Luitgaarden, Madeline Schmidbauer, Alex Mathers, Morgan Pugh, and the OPIHI crew). In addition, I want to thank the residents, facilities, fiscal, and administrative teams at the Hawaiʻi Institute of Marine Biology for all that they do. Nothing would happen without them. I received daily support in the form of laughs, commiseration, enthusiasm, and friendship from the members of the Karl, ToBo, and Donahue Labs. Thank you all for being so wonderful. In addition, I want to thank the Vaughan Lab for welcoming me into their piko and teaching me so much. I also want to thank the KaPAʻA crew for all of the lessons and for the inspirational work that you are doing. In addition, thank you to the TRHT ʻOhana for filling my mind with

iii compassion, hope, and a new sense of community. To all the teachers who taught me lessons that fell far outside the scope of required classes but that have been integral to my personal and professional development, thank you so very much. I am also fortunate to have amazing friends who have lifted me up and inspired me. I have watched them become partners, parents, professionals, and most importantly, beautiful people, and that has been such a gift. To my best friend of all – Spencer Bates – thank you for filling our home and my life with levity, love, and delicious food. I know this has been a long road for you as well, so thank you for your patience. And to my family – you are the reason for all that I do. You have always been my biggest fans, defenders, confidants, sounding boards, inspiration, motivation, and source of joy. There are no sufficient words to express my gratitude. This research and my time in school was funded by the following sources: National Science Foundation Graduate Research Fellowship Program (Grant #: 1329626); Hawaiʻi Sea Grant; the UH Mānoa Marine Biology Graduate Program (Small Grants Award & Achievement Scholarship); the UH Mānoa Ecology, Evolution, and Conservation Biology program (Yoshimoto Fellowship); the Hawaiʻi Institute of Marine Biology (Lord Scholarship & Castro Scholarship); and the American Museum of Natural History (Lerner-Gray Memorial Fund).

iv ABSTRACT Herbivorous fishes are critical to the function and resilience of coral reefs, but their diets and functional roles can show great variability, even among closely related taxa. I assessed the diet specialization of herbivorous reef fishes at multiple scales ranging from an examination of diet variation on an individual scale to a Pacific wide analysis of functional homogenization in herbivore assemblages. Using a molecular metabarcoding approach, I first identified the algal diets of two common Hawaiian surgeonfishes, Acanthurus triostegus and A. nigrofuscus, which consume large amounts of turf algae. Because turf algae are difficult to identify visually, especially in the field, our understanding of the true diet breadth of these species has been limited. A. nigrofuscus exhibited greater diet diversity, more variability between individuals, and less variation between sites than A. triostegus. I then used this same metabarcoding approach to examine the diets of eight other herbivores and conducted a systematic literature review of studies examining herbivorous reef fish diets. I combined these data to generate a standardized index of diet diversity for herbivorous reef fishes, which indicated that parrotfishes and detritivorous surgeonfishes have more limited diets, but there is also a range of specialization within each functional group and taxonomic family. I combined the index of diet diversity with fish abundance and trait data to examine functional homogenization in herbivore assemblages throughout the Pacific. Using statistical models, I examined the direct and indirect effects of a suite of ecological, biophysical, and anthropogenic drivers on herbivores. Differences in herbivore assemblage composition between islands, regions, and human population densities were apparent, but local and indirect effects drive variability in the relationship. This work will assist in reevaluating the functional role of herbivores to guide effective management by examining herbivore specialization at scales ranging from an individual to an assemblage.

v TABLE OF CONTENTS

Acknowledgements ...... iii

Abstract ...... v

List of Tables ...... ix

List of Figures ...... x

Chapter 1: Introduction ...... 1

Chapter 2: Metabarcoding as a tool to examine cryptic algae in the diets of two common grazing surgeonfishes, Acanthurus triostegus and A. nigrofuscus ...... 4 Abstract ...... 4 Introduction ...... 4 Methods ...... 6 Results ...... 10 Discussion ...... 13 Conclusions ...... 14 Acknowledgements ...... 15 Tables ...... 16 Figures ...... 18

Chapter 3: Picky eater or generalist feeder? Comparing diet diversity of herbivorous coral reef fishes in the Pacific ...... 23 Abstract ...... 23 Introduction ...... 23 Methods ...... 26 Results ...... 32 Discussion ...... 34 Conclusions ...... 37 Acknowledgements ...... 38 Tables ...... 39 Figures ...... 41

vi Chapter 4: Functional homogenization in herbivorous coral reef fish assemblages in response to anthropogenic impacts ...... 47 Abstract ...... 47 Introduction ...... 47 Methods ...... 49 Results ...... 53 Discussion ...... 55 Conclusions ...... 58 Acknowledgements ...... 58 Tables ...... 59 Figures ...... 63

Chapter 5: Conclusions ...... 70 Primary Findings ...... 70 Future directions ...... 71

Appendices ...... 72 Appendix A: Chapter 2 Supplemental Figures ...... 72 Appendix B: Chapter 3 Supplemental Tables ...... 81 Appendix C: Chapter 3 Supplemental Figures ...... 82 Appendix D: Chapter 3 Supplemental Text ...... 85 Appendix E: Chapter 4 Supplemental Tables ...... 96 Appendix F: Chapter 4 Supplemental Figures ...... 99

References ...... 114

vii LIST OF TABLES

Table 2.1. Measures of individual variation in Acanthurus triostegus and A. nigrofuscus. Values of mean pairwise diet overlap range from 0 to 1, with 1 signifying complete diet overlap. Individual variation at the population level ranges from 0 to 1, with 1 signifying an average individual diet that mirrors that of the total population. Significant differences between sites are indicated with grouping letters for A. triostegus. There were three groups for mean pairwise diet overlap in A. nigrofuscus (a: Ala Moana, Haleʻiwa, Kakaʻako, Lāiʻe, and Maunalua; b: Mokuleʻia; ab: Chuns and Olowalu), but there were no significant differences between sites in individual specialization as compared to the population...... 16

Table 3.1. Mean Diet Diversity Index (DDI) values for each species are listed along with sources used to generate the values. The number of individuals used in each study is noted in parentheses next to the source...... 39

Table 4.1. Linear Mixed Effects Model results for models testing the direct effects of human population and human impacts on metrics of functional homogenization in herbivores ...... 59

Table 4.2. Piecewise Structural Equation Model significant results grouped by response and in order of decreasing standard estimate absolute value...... 61

Table S3.1. Summary statistics for the ten herbivore species that were included in the metabarcoding dietary analysis...... 81

Table S4.1. Description of predictor variables with reason for inclusion ...... 96

viii LIST OF FIGURES

Figure 2.1. (A) Diet composition of Acanthurus triostegus; each bar represents a unique sample. (B) The distribution of individual A. triostegus diet diversity, as measured by Simpson’s Index on each sample. (C) The distribution of individual A. triostegus diet richness ...... 18

Figure 2.2. (A) Diet composition of Acanthurus nigrofuscus; each bar represents a unique sample. (B) The distribution of individual A. nigrofuscus diet diversity, as measured by Simpson’s Index on each sample. (C) The distribution of individual A. nigrofuscus diet richness ...... 19

Figure 2.3. Non-metric multidimensional scale plot using Bray-Curtis similarity matrix for Acanthurus triostegus by location (color). Stress was 0.091 with k = 4. The large amount of variability between individuals at Salt Pond, Kauaʻi obscured the differences between other sites, so it was not included in this plot. All sites are on Oʻahu, except for Waiheʻe, which is on Maui...... 20 (a) NMDS plot overlaid with ellipses that show a 95% confidence interval on site clusters...... 20

(b) NMDS plot overlaid with diet items (i.e., algal taxa) ...... 21

Figure 2.4. Individual pairwise diet overlap by site is shown for Acanthurus triostegus (blue) and A. nigrofuscus (orange). Higher values indicate less individual diet specialization...... 22

Figure 3.1. Species diet diversity measured by the Simpson’s Diet Diversity Index plotted by functional group. Higher values indicate a more diverse diet. The same figure organized by taxonomic family is available in Appendix C ...... 41

Figure 3.2. The relative abundance of diet items in the species that were incorporated into the diet diversity index are shown, with color indicating diet item type (legend on bottom) ...... 42

Figure 3.3. Non-metric multidimensional scale plot using Bray-Curtis similarity matrix based on diet composition with functional groups coded by color ...... 43 (a) Ellipses show a 95% confidence interval on functional group clusters ...... 43

(b) Significant (p < 0.05) diet items are overlaid onto the NMDS plot ...... 44

Figure 3.4. Dendrogram of relationships between species and functional groups based on Gower’s Dissimilarity Matrix values from trait matrices that included diet diversity ranking. Colors represent functional groups ...... 45

Figure 3.5. Mean diet composition as determined by metabarcoding analysis grouped by functional group and in order of increasing diet diversity within functional groups based on Simpson’s Diversity Index ...... 46

ix

Figure 4.1. Herbivore assemblage composition in the Hawaiian Islands with herbivore diversity increasing from left to right (i.e., most diverse assemblages as measured by Simpson’s Diversity Index on the right). Only species that made up at least 5% of the relative abundance of herbivores present at an island were included ...... 63

Figure 4.2. NMDS (stress = 0.16) of herbivore assemblage composition at the island scale across regions in the Pacific. Ellipses show 95% confidence intervals ...... 64

Figure 4.3. Regional groupings of homogenization metrics ...... 65 (a) Herbivore assemblage diet specialization by island, grouped by region. Dark blue lines represent the mean value for the region...... 65

(b) Herbivore species diversity (measured by Simpson’s Diversity Index) by island, grouped by region. Dark blue lines represent the mean value for the region ...... 66

Figure 4.4. Non-metric multidimensional scaling plot using Bray-Curtis dissimilarity between herbivore assemblages grouped by human population within 20 km. Minimally inhabited sites had less than 100 human residents within 20 km. Populated sites had more than 100 but less than 10,000 human inhabitants within 20 km, and densely populated sites had more than 10,000 human inhabitants within 20 km. Some islands, such as Aguijan, are uninhabited but are within 20 km of densely populated islands ...... 67

Figure 4.5. Functional dissimilarity of herbivores compared to NCEAS Human Impact Score. The shaded area shows a 95% confidence interval. Colors correspond to regions. NCEAS Human Impact Scores were not available for the Papahānaumokuākea Marine National Monument (Northwestern Hawaiian Islands)...... 68

Figure 4.6. Piecewise structural equation model results. The arrow is pointing from the response variable to the predictor variable. Only significant interactions are listed (P < 0.05), and the width of the line indicates the magnitude of the standardized estimate ...... 69

Figure S2.1. Map of sample collection locations. Left: Oʻahu. Top right: Maui. Bottom right: Kauaʻi ...... 72

Figure S2.2. Species accumulation curves generated using the Vegan package in R for Acanthurus triostegus (grey) and A. nigrofuscus (brown) samples, with the extrapolated sampling curve for all samples combined in pink. Vertical lines represent the standard deviation of the extrapolated sampling curve...... 73

Figure S2.3. A. triostegus diet composition broken down by order ...... 74

Figure S2.4. A. nigrofuscus diet composition broken down by order ...... 75

x Figure S2.5. Diet richness (A: A. triostegus & B: A. nigrofuscus) and diversity (C: A. triostegus & D: A. nigrofuscus) with significant differences between sites indicated with letters. There were no significant differences in richness or diversity between sites for A. nigrofuscus ...... 76

Figure S2.6. Linear mixed effects model diagnostic plots for model examining diet overlap among individuals of Acanthurus triostegus in the population. (A) Plot of residual versus fitted values. (B) Random effect (site) estimate plot. (C) Fixed effects estimate plots. Weight was squared in the model to account for the nonlinear distribution ...... 77

Figure S2.7. Linear mixed effects model diagnostic plots for model examining diet overlap between Acanthurus triostegus individuals and the overall population. (A) Plot of residual versus fitted values. (B) Random effect (site) estimate plot. (C) Fixed effects estimate plots. Weight was squared in the model to account for the nonlinear distribution ...... 78

Figure S2.8. Linear mixed effects model diagnostic plots for model examining diet overlap among Acanthurus nigrofuscus individuals in the population. (A) Plot of residual versus fitted values. (B) Random effect (site) estimate plot. (C) Fixed effects estimate plots ...... 79

Figure S2.9. Linear mixed effects model diagnostic plots for model examining diet overlap between Acanthurus nigrofuscus individuals and the overall population. (A) Plot of residual versus fitted values. (B) Random effect (site) estimate plot. (C) Fixed effects estimate plots. .... 80

Figure S3.1. Collection locations of fish used in this metabarcoding dietary study where the size of the box reflects the sample size at a given location, and the color refers to the box surrounding the species. Locations not included on the map are listed below the species. The species are (from left to right, top to bottom) Acanthurus triostegus, A. nigrofuscus, Ctenochaetus strigosus, A. nigroris, Zebrasoma flavescens, A. leucopareius, A. guttatus, Naso unicornis, A. olivaceus, and N. lituratus ...... 82

Figure S3.2. Species Diet Diversity Index (DDI) values are plotted by functional group with the mean value decreasing from left to right. The graph shows the median (line), first and third quartiles (hinges), the 95% confidence interval of the median (whiskers), and outliers (circles). Image Credit: Keoki Stender and obtained with permission from his website (MarineLifePhotography.com) ...... 83

Figure S3.3. Species diet diversity generated using Simpson’s Diversity Index based on diet data from existing literature grouped by taxonomic family. Larger values indicate a more diverse diet (i.e., diet diversity increases moving towards bottom of the page) ...... 84

Figure S4.1. Proportion of herbivores present at a site (i.e., total herbivore richness) that were included in the herbivore assemblage diet specialization index based on the Diet Diversity Index from Chapter 3 ...... 99

xi Figure S4.2. Assemblage composition based on the relative abundance of herbivores present averaged by island and grouped by region. Only species that comprised at least 5% of the overall abundance were included. Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation)...... 100

Figure S4.3. Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation). Species codes in the legend are ordered from least diverse diet to most diverse diet (top to bottom) ...... 101 (a). The frequency with which a given species was the most abundant species at a site, grouped by island in the Main Hawaiian Islands ...... 101

(b). The frequency with which a given species was the most abundant species at a site, grouped by island in the Papahānaumokuākea Marine National Monument ...... 102

(c). The frequency with which a given species was the most abundant species at a site, grouped by island in the Pacific Remote Island Areas ...... 103

(d). The frequency with which a given species was the most abundant species at a site, grouped by island in Samoa ...... 104

(e). The frequency with which a given species was the most abundant species at a site, grouped by island in the Northern Marianas Islands ...... 105

(f). The frequency with which a given species was the most abundant species at a site, grouped by island in the Southern Marianas Islands ...... 106

Figure S4.4. Correlation plot for a combination of variables of interest for use in a linear mixed effects model. Variables with r > 0.7 were not included together ...... 107

Figure S4.5. Herbivore functional dissimilarity by island, grouped by region. The dark blue lines represent the mean value for the region ...... 108

Figure S4.6. The relative abundance of nominally herbivorous reef fishes grouped by functional groups were compared to the NCEAS Human Impact Score (38 islands). The NCEAS Human Impact Score is developed using 17 anthropogenic stressors that affect coral reefs, and impacts increase from left to right. Gray areas show 95% confidence interval ...... 109

Figure S4.7. The relative abundance of Acanthurus spp. were compared to the NCEAS Human Impact Score (left, 38 islands) and the population of humans within a 20 km radius (right, plotted on a log scale, 42) throughout the Pacific. When examined independently Acanthurus nigrofuscus had a significant positive relationship with the Human Impact Score (P < 0.0001, Estimate = 0.29, R2 marginal = 0.09, R2 conditional = 0.23) ...... 110

xii Figure S4.8. NMDS of herbivore assemblage composition at the island scale across regions in the Pacific with significant environmental variables overlaid (k = 2, stress = 0.14)...... 111

Figure S4.9. NMDS of herbivore assemblage composition at the island scale across regions in the Pacific with significant herbivore species overlaid (k = 2, stress = 0.14). Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation) ...... 112

Figure S4.10. Herbivore diversity compared to NCEAS Human Impact Score. Colors correspond to regions. NCEAS Human Impact Scores were not available for the Papahānaumokuākea Marine National Monument (Northwestern Hawaiian Islands) ...... 113

xiii CHAPTER 1 INTRODUCTION

Niche theory and optimal foraging models have provided a foundation for much ecological research over the past century1–4, yet in a modern context they are still highly relevant and critical to understanding an organism’s response to increasing anthropogenic impacts. In diverse communities, such as those found on coral reefs, species constantly compete for limited habitat and resources, which drives niche diversification and may enable the persistence of specialists. In the past these conditions allowed for the coexistence of a diverse array of fishes on coral reefs; however, as these communities experience continuous degradation and loss of diversity as a result of human impacts, their ability to sustain unique assemblages of specialized organisms is reduced. This dissertation research examines diet breadth and relative specialization in herbivorous reef fishes at multiple scales to contribute to our understanding of how human impacts may be affecting the niches and foraging of herbivores. Coral reefs provide an interesting setting in which to examine questions of specialization. Though they cover a small fraction of the total ocean habitat, coral reefs house a tremendous amount of the ocean’s biodiversity5. As highly complex structures, coral reefs provide the opportunity for organisms with a multitude of niches to thrive and coexist, but coral reefs face mounting stressors globally that threaten their persistence. Warming ocean temperatures have resulted in repeated mass coral bleaching events, and ocean acidification weakens the structural integrity of calcifying organisms, including reef-building corals6,7. These processes threaten the physical structure, and consequently the complexity, of coral reef ecosystems8. In addition, many reefs also face a variety of local stressors that may range from sedimentation to pollution to an overgrowth of invasive algae, which may locally exacerbate the global impacts of climate change9,10. Nominally herbivorous reef fishes, which often consume sediment, decomposing organic matter, and microorganisms along with the more obvious macroalgae and turf algae, serve a critical function in terms of mediating the impacts of some of these local stressors. For example, herbivores may consume invasive macroalgae that would otherwise compete with corals for space on the benthos11. They also consume sediment and transport it off of the reef, inadvertently

1 preventing the accumulation of problematic amounts of sediment12. Other herbivores scrape reef surfaces, and in the process, provide new substrate on which corals can settle13,14. On some reefs, herbivores consume more than 70% of the turf algae produced on a daily basis, representing a massive transfer of energy within coral reef food webs15. Within this one group of organisms, however, a wide range of ecological function and dietary specialization exists. Herbivory on coral reefs is often understood in terms of the relationship between fish, algae, and corals16. In a simplified version, herbivores consume macroalgae, and in the process they contribute to the balance of the benthos, creating the space for corals to grow on the reef. This creates a feedback loop in which the reef provides habitat for fish that consume algae which reduces the competitive advantage that fast growing algae may have over corals. When herbivores are removed from this relationship, through processes such as overfishing or loss of habitat, the balance can be disrupted, and macroalgae can grow unchecked, potentially leading to a shift to algae dominated reefs17. As this interpretation of the herbivory paradigm often guides management practices, it is important to understand the true impact of herbivores on reefs. As a group with a variety of ecological functions, one herbivore is not interchangeable for another. There are five main functional groups that are typically considered within nominally herbivorous reef fishes: grazers, browsers, scrapers, excavators, and detritivores. Grazers, which can be represented by surgeonfishes such as Acanthurus triostegus, primarily consume the heterogeneous mix of taxa that is referred to as turf algae. This algae is typically filamentous and may only grow to be a few millimeters tall, carpeting the surface of the reef. While turf algae is a natural element of healthy reefs, it proliferates on many degraded reefs18. Browsers, often represented by larger bodied fishes such as Naso unicornis, consume large, fleshy macroalgae, making them important for the control of invasive macroalgae. Parrotfishes are grouped into the scrapers, which scrape the reef leaving an intact substrate, and excavators, which take bites of substrate19. Though they actually target microalgae on or within the reef14, these fish inadvertently clear substrate that may assist in coral settlement. Finally, some acanthurids which are effectively detritivores, such as Ctenochaetus strigosus, are typically grouped together with herbivores, as they comb through turf algae to retrieve detritus20.

2 Each of these functional groups serve a unique role in the ecosystem, and so in many ways, they need to be considered independently when developing management frameworks. To truly understand the role of herbivores on coral reefs, researchers must begin to unravel the diversity of functions that they exhibit and the range of specialization that may exist within each function. In some cases where species have extremely unique functional roles, the loss of one species may result in the complete loss of an ecosystem function21. In other cases there may be functional overlap and redundancy that afford the community resilience22–24. This mix of diversity and redundancy becomes particularly crucial when taxa experience a drastic reduction in abundance, as can happen for species that are fishing targets. Organisms that are very specialized and have narrower niches may suffer more from habitat loss than those that are able to thrive in a variety of conditions. It is expected that as habitat degradation increases globally, biotic homogenization may begin to occur, in which weedy generalists that are able to consume a wider breadth of items and are able to live in a broader range of habitats will be able to persist while those taxa that require more specific habitats and diet items may experience significant population reductions25–29. A shift in the relative abundance of specialists versus generalists would be expected as habitat conditions shift or degrade, but this process has not been thoroughly examined in coral reef fishes. This dissertation aims to assess the range of specialization seen within and among groups of herbivorous reef fishes, using diet as a proxy for niche breadth. Using a variety of methods, this question was addressed at multiple spatial scales. First, metabarcoding, a molecular method, was used to examine diet variability at the individual scale between and within two species of closely related grazing surgeonfishes, A. triostegus and A. nigrofuscus (Chapter 2). Next, diet differences were examined between and within the previously defined functional groups of herbivores using a meta-analysis combined with metabarcoding data (Chapter 3). Finally, functional homogenization was examined using the data generated in the previous chapters integrated with fish abundance data collected at sites across the Pacific to better understand the direct and indirect effects of human impacts on herbivore assemblages (Chapter 4).

3 CHAPTER 2 METABARCODING AS A TOOL TO EXAMINE CRYPTIC ALGAE IN THE DIETS OF TWO COMMON GRAZING SURGEONFISHES, ACANTHURUS TRIOSTEGUS AND A. NIGROFUSCUS

Abstract We employed metabarcoding to identify and compare the algal diets of two common Hawaiian surgeonfishes, the convict tang (Acanthurus triostegus) and the brown surgeonfish (A. nigrofuscus). Diet richness was not significantly different between the species, with 64 unique taxa identified from the gut contents of 96 A. triostegus and 57 taxa from 81 A. nigrofuscus. Although the two species consumed many of the same algae, A. nigrofuscus consumed a greater proportion of brown algae, cyanobacteria, and material from the epilithic algal matrix. In contrast, A. triostegus had a taxonomically more specialized diet, consuming primarily red algae. A. triostegus showed greater variability between sites in both algal diet diversity and individual diet overlap, whereas A. nigrofuscus exhibited greater variation among individuals with less diet variability between sites. Findings indicate that individuals of the generalist A. nigrofuscus feed opportunistically whereas individuals of the larger, schooling A. triostegus are more capable of accessing preferred resources at each location. This study is the most comprehensive application of metabarcoding to examine herbivorous reef fish diets and provides the most complete examination of diet composition, individual variability, diet overlap, and spatial variability among grazing Hawaiian herbivorous surgeonfish to date. This work sheds light on how these important coral reef grazers partition resources and may help managers anticipate the functional responses of herbivorous reef fishes in the face of a changing climate.

Introduction Diet breadth, composition, and specialization are foundational to our understanding of organismal ecology, nutrition, and metabolic function. On a population level, diet shapes habitat use and the way individuals interact through competition for limited resources, and trophic interactions form the basis of our understanding of functional roles and ecological relationships. While dietary specialization is generally considered on a species level, individual specialization

4 in a population of generalists may also be quite prevalent30–33. Anthropogenic impacts have become increasingly apparent across a range of habitats, and concerns have mounted over functional homogenization, or the widespread replacement of specialized species with weedy generalists27–29,34–36. Consequently, the functional homogenization of diet and, more specifically, the loss of diet specialization, is emerging as a global concern. Interpretation of diet breadth and composition may vary depending on the approach used. Recent work shows that in some cases visual methods underestimate the diversity of taxa as compared to molecular approaches, such as DNA barcoding or metabarcoding, a molecular tool that uses reference sequences to identify species composition37–41. Metabarcoding has been used to answer fundamental ecological questions about niches and diet42–48, assist in conservation and monitoring efforts49–54, identify pollen55–57, detect invasive species58–60, and conduct community surveys61–63. Metabarcoding also serves as the foundation of the rapidly expanding realm of environmental DNA studies52,64–66. Though the quantitative aspect of metabarcoding has presented a challenge in the past, recent advances have led to useful estimates of relative biomass67–70. Similarly, the dearth reference sequences, which are necessary to identify individual species, has been a limitation historically64,71–73, but great advances have also been made in recent years in generating reference barcodes74–79. Hawaiian algae, which are largely rhodophytes, have received a fair amount of expert attention in comparison to algae in other locations. This has resulted in a relatively rich database of reference sequences80–82. Thus, Hawaiʻi is uniquely positioned to leverage these reference sequences in herbivore diet metabarcoding studies83. Given the diversity and proliferation of turf algae on impacted reefs and the degraded nature of algal material in gut contents, metabarcoding provides a particularly appealing way of broadening our understanding of the diversity and breadth of algae consumed by grazing herbivorous fishes, a group that is dominated by surgeonfishes in Hawaiʻi. This project applies metabarcoding to differentiate the heterogeneous mix of turf algae in the gut contents of two common grazing surgeonfishes, Acanthurus triostegus (manini or convict tang) and A. nigrofuscus (maʻiʻiʻi or brown surgeonfish). A. triostegus and A. nigrofuscus are found throughout the Indo-Pacific and are both common and generally abundant throughout their range84–86. A. triostegus is far more likely to be targeted as a food fish than A. nigrofuscus, though neither species is generally considered to be overfished85,87.

5 A. triostegus schools in large aggregations that can contain hundreds of individuals84, and with these groups they can outnumber, and consequently outcompete, other grazing species. Using metabarcoding to examine the diets of A. triostegus and A. nigrofuscus, we set out to answer the following ecological questions: (1) How similar are the diet breadth and composition of two congeneric species that have overlapping ranges, particularly when feeding in the same habitats? (2) Is there evidence of individual specialization in these two widely distributed, common species? (3) Can metabarcoding assist in the identification of cryptic prey in diet studies?

Methods I. Sample Collection & Processing 96 Acanthurus triostegus samples were collected from locations on Oʻahu (Mākaha, Kahe Point, Kakaʻako, Maunalua, Kailua, Kāneʻohe, Kahana, and Haleʻiwa), Maui (Waiheʻe), and Kauaʻi (Salt Pond) (Figure S2.1). 81 A. nigrofuscus samples were collected from locations on Oʻahu (Kakaʻako, Ala Moana, Maunalua, Lāʻie, Hāleʻiwa, and Mokulēʻia). Fishes were collected with spears from July 2015 and July 2017 between the hours of 8:00 and 16:00. Whole fish were frozen as quickly as possible following collection and stored at - 20 °C. Fish were measured, weighed, and dissected at the Hawaiʻi Institute of Marine Biology (metadata on GitHub: enalley/Dissertation/SupplementalFile2.1). During this process 0.25 g of stomach content material was removed and placed into a PowerBead collection tube provided in the MoBio PowerSoil DNA Extraction Kits (now Qiagen DNEasy PowerSoil Kits – Qiagen, Hilden, Germany). Stomach rather than intestinal contents were used to minimize the degree of digestion and to standardize the sampling methods. Extractions were completed using the MoBio PowerSoil DNA Extraction Kits.

II. Library Preparation The protocol for MoBio PowerSoil DNA Extraction Kits was followed with an additional elution step with 50 µl of PCR water. The DNA from this second water-based elution was used for downstream processing. The Illumina 16S Metagenomic Sequencing Library Preparation guide was followed88, substituting 23S rRNA Universal Plastid Amplicon (UPA) for 16S80. A

6 two-step PCR process was used in which Illumina adaptors (forward: 5’- TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG, reverse: 5’- GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG)88 were added to the 5’ end of the forward (p23SrV_f1: 5’-GGACAGAAACACCCTATGAA)80 and reverse (p23SnewR: 5’- TCAGCCTGTTATCCCTAGA)89 primers in the first PCR step. Individual indices were added to each sample in the second PCR step. For the first PCR (i.e., Illumina adaptors and primers), 25 µl reactions had the following components: 12.5 µl MyTaq Red 2x Master Mix, 9.5 µl ultrapure PCR water, 1 µl forward primer with attached Illumina adaptors, 1 µl reverse primer with attached Illumina adaptors, and 1 µl of DNA. The first PCR protocol was as follows: 2 minutes of denaturation at 95 °C; 35 cycles of 94 °C for 30 seconds, 66 °C for 30 seconds with the temperature reduced by 0.5 °C every cycle until reaching 58 °C, and finally, 72 °C for 30 seconds; and 5 minute extension at 72 °C 90. Extraction and PCR negative controls were examined using gel electrophoresis, and any runs with bands in the control lane were discarded. Before the indexing PCR, products were cleaned using PCRClean DX magnetic beads following the recommended guidelines of a bead volume 1.8 times the PCR volume (Aline Biosciences, Woburn, Massachusetts, United States). All indexing PCR reactions used KAPA HiFi HotStart Ready Mix (Roche, Basel, Switzerland). For one of the three libraries, Nextera XT Indices (Illumina, San Diego, California, United States) were used in 50 µl reactions that were comprised of 25 µl of KAPA HiFi HotStart Ready Mix, 5 µl of Nextera XT Index 1 Primers, 5 µl of Nextera XT Index 2 primers, 10 µl of ultraclean PCR water, and 5 µl of PCR product. The following PCR protocol was used for this library: 95 °C for 3 minutes; 8 cycles of 95 °C for 30 seconds, 55 °C for 30 seconds, and 72 °C for 30 seconds; and 72 °C for 5 minutes. Similar indices designed by Kozich et al. (2013) were used in 25 µl reactions (12.5 µl KAPA HiFi HotStart Ready Mix, 2.5 µl i5 indices, 2.5 µl i7 indices, 5 µl ultrapure PCR water, and 2.5 µl of PCR product)91. The PCR protocol for the Kozich et al. (2013) indices was as follows: 95 °C for 3 minutes; 8 cycles of 95 °C for 30 seconds, 55 °C for 30 seconds, 72 °C for 30 seconds; and 72 °C for 30 seconds. Products were again cleaned using PCRClean DX to remove primer dimers and small fragments as above, and they were then quantified on a SpectraMax Microplate Reader (Molecular Devices, San Jose, California, United States) with the Accuclear Ultra HS dsDNA Quantitation Kit (Biotium, Fremont, California, United States) following the manufacturer’s

7 protocol. Twenty one negative controls were haphazardly selected for sequencing from the extraction, PCR, and indexing controls. Libraries were pooled based on their concentrations and were sequenced at the Genetics Core Facility at the Hawaiʻi Institute of Marine Biology (Kāneʻohe, Hawaiʻi, United States) and the Genetics Core Facility at the Advanced Studies in Genomics, Proteomics, and Bioinformatics at the University of Hawaiʻi at Mānoa (Honolulu, Hawaiʻi, United States). An Illumina MiSeq platform with the V3-600 cycle kit (2x300 bp) was used for all sequencing. Pooled libraries were demultiplexed by the Genetics Core Lab facilities.

III. Bioinformatic Analyses The DADA2 program in R was used to filter, trim, align, and remove chimeras from raw unpaired reads92,93. DADA2 outputs an amplicon sequence variant (ASV) table, akin to an operational taxonomic unit (OTU) table, which provides information regarding the frequency with which each ASV appeared in each sample92. A pseudo-pooling approach was used, which allows sequences in pooled samples (i.e., the rest of the sequencing library) to provide Bayesian- like priors to improve the recognition of ASVs. Taxonomic assignments of ASVs were made using the informatic sequence classification trees (INSECT) pipeline in R, which references a classification tree of 23S UPA sequences developed using training sequences from a reference database and hidden Markov models for grouping93,94. Assignments are made by comparing the Akaike weight of the likelihood at each competing node until the weight falls below a 0.9 threshold or an identification is made at the lowest taxonomic level possible94. Following assignment, INSECT yielded an ASV table with the following: amplicon sequence, ASV number, taxonomic identification number, assignment score, taxonomic assignment (i.e., kingdom, phylum, class, order, family, genus, species), and read abundance for each sample94. This ASV table was condensed to show one row per unique taxonomic identification, meaning unique ASVs given the same taxonomic assignment were clustered. Controls were examined, and non-algal ASVs were removed from all samples (GitHub: enalley/Dissertation/SupplementalFile2.2). All raw data and codes for bioinformatic and statistical analyses are available for reference on GitHub (in enalley/Dissertation).

8 IV. Statistical Analyses A. Sampling Coverage Technical replicates (i.e., multiple sequencing runs of the same sample) were added together, and the data was then filtered to include only samples containing over 10,000 reads. The diet richness and diversity (using Simpson’s Diversity Index) were determined for each sample using the Vegan package in R95. When there were biological replicates, the biological replicate (i.e., independent extractions and complete protocols performed for replicates of the same sample) with the highest richness was selected for analysis. The data was then imported into the phyloseq package in R to be rarefied to an even sampling depth of 10,000 reads to reduce sequencing and amplification biases between samples and to generate relative read abundance values96 (GitHub: enalley/Dissertation/SupplementalFile2.3). Species accumulation curves were plotted for each species using the Vegan package in R to evaluate sampling coverage95 (Figure S2.2).

B. Diet Breadth and Composition (Differences Between Species & Sites) Diet richness was generated for each species and was compared between the species and sites using the Wilcoxon signed rank test and permutation tests in the package rcompanion in R97. The final ASV table was further classified into relative read abundance (RRA) in order to generate quantitative values in addition to presence/absence data98,99. Because of the frequency of null values (i.e., zero abundance of a particular prey item in a stomach), a Bray-Curtis similarity matrix, which is unaffected by variable numbers of zero values between samples, was used with non-metric multidimensional scaling ordination to compare diet composition between individuals of a species at a site using the Vegan package in R.

C. Individual Variation within a Population The proportional similarity index, developed by Bolnick et al. (2003) and based on mean pairwise diet overlap, was calculated for each individual in the population and for the mean of the population using the RInSp package in R30,100 (Table 2.1). This value measures the overlap between the individual’s diet and the overall population diet. With values falling between 0 and

9 1, higher values indicate a diet that more closely mirrors the overall population diet. The individual overlap value indicates the prevalence of individual variation in the population. The mean pairwise overlap of all individuals of a species at each site was calculated using the RInSp package, and this value was compared between sites for a given species using permutation tests in the package rcompanion in R (Table 2.1)97. To examine potential drivers of individual variation, the length, weight, body condition using Fulton’s K (weight/(total length)3), hepatosomatic index (weight of liver/ weight of fish), and fullness index (weight of intestines/weight of fish) were compared between species using a Mann-Whitney test to account for the non-normal distribution of the data95,101. Drivers of individual variation were examined using linear mixed effects models in the lme4 package102. The proportional similarity index and mean diet overlap between individuals were used as response variables in independent models. In each of the models weight (squared to account for a nonlinear relationship), diet richness, and diet diversity were used as fixed effects. Site was included as a random effect. All quantitative predictor variables were centered and scaled, and variance inflation factor values were less than 2.

Results I. Differences in Diet Breadth and Composition Between Species Acanthurus triostegus had a greater total diet richness than A. nigrofuscus at the population level (64 unique identifications versus 57, across all individuals), but individuals of the two species exhibited no significant differences in their mean diet richness (Wilcox: P = 0.42). Despite having similar mean diet richness, A. nigrofuscus exhibited significantly greater diet diversity than A. triostegus at the population level, as measured by the Simpson’s Diversity Index (Wilcox: P = 0.04). Though individuals of both species consumed a similar number of taxa, A. triostegus primarily consumed red algae (Rhodophyta; Figure 2.1), while the relative read abundance of taxa consumed by A. nigrofuscus was more even and included brown algae (Phaeophyceae) and cyanobacteria along with rhodophytes (Figures 2.2). The higher proportion of cyanobacteria in A. nigrofuscus indicated an increased abundance of detritus (i.e., organic matter) in their diets. A complete list of identified diet items for both species is provided in the

10 Supplemental Files (GitHub: enalley/Dissertation/SupplementalFile2.4), along with a taxonomy table (GitHub: enalley/Dissertation/SupplementalFile2.5). Within Rhodophyta, A. triostegus primarily targeted Ceramiales, which are turf algae, with a smaller portion of the red algae being unidentified rhodophytes (Figure S2.3). At nearly every site, A. triostegus also consumed Asparagopsis taxiformis. At Maunalua Bay, Kahana Bay, and Hāleʻiwa (Puʻena Point) (all on the island of Oʻahu), A. triostegus also consumed Acanthophora spicifera, an invasive alga. The small amount of green algae consumed by A. triostegus was Ulva sp. and Bryopsidales. Similarly, the small amount of brown algae consumed by A. triostegus was mostly Dictyotaceae and Sphacelaria tribuloides. Though they represented a very small portion of their diet, Oscillatoriales and other cyanobacteria were also detected. A. nigrofuscus, which had a far greater proportion of brown algae than A. triostegus, still had a diet dominated by rhodophytes (Figure 2.2). Like A. triostegus, Ceramiales were targeted, but A. nigrofuscus also consumed Asparagopsis taxiformis at every site examined (Figure S2.4). A. nigrofuscus consumed the invasive alga Acanthophora spicifera only at Hāleʻiwa (Puʻena Point). A. nigrofuscus also had higher relative read abundances from calcareous and coralline algae than A. triostegus. Though A. nigrofuscus also consumed very little green algae, the green algae present were predominantly Ulva sp. and Bryopsidales. As mentioned previously, A. nigrofuscus had far more brown algae sequences than A. triostegus, but as with A. triostegus, the brown algae consumed was again dominated by Dictyotaceae and Sphacelaria tribuloides. In some locations such as the urban and heavily populated sites at Ala Moana, Kakaʻako, and Maunalua, A. nigrofuscus also consumed substantial amounts of detritus, as indicated by the presence of Oscillatoriales, unidentified cyanobacteria, and other taxa found in the EAM. There were morphological differences between the two surgeonfishes as well. In this study, which targeted adult fish, A. nigrofuscus had a mean weight of 59 g and a maximum weight of 113 g, which was significantly lower (Wilcox: P < 2.2e-16) than the mean weight of 101 g and maximum weight of 220 g in A. triostegus. In addition, A. nigrofuscus had a significantly lower Fulton’s K (P < 2.2e-16) and gut fullness index (P = 3.31e-09) than A. triostegus. There was no significant difference in the hepatosomatic index between the two species (P = 0.22).

11

II. Differences in Diet Breadth and Composition Between Sites Acanthurus triostegus, which is more specialized than A. nigrofuscus based on its overall diet diversity, also had greater differentiation in diet between sites (Figures 2.3 and 2.4). A. triostegus showed significant differences in diet richness (permutation test: P < 0.05) and diet diversity between sites (permutation test: P < 0.05, Figure S2.5). Populations from Maunalua, Kailua, Kahana, and Hāleʻiwa (Puʻena Point) all had relatively high richness, and those in Kakaʻako, Makaha, Kahe Point (Electric Beach), Waiheʻe (Maui), and the Salt Pond (Kauaʻi) all had moderate richness (Figure S2.5). A. nigrofuscus did not show significant differences between sites in diet richness or diversity (permutation test: all p > 0.05, Figure S2.5). There was less differentiation between sites in A. nigrofuscus than in A. triostegus, so there was also less evidence of strong associations between certain species of algae and specific locations.

III. Individual Variation within a Population There was greater overlap between the diets of individuals, as well as between the diet of an individual with overall population diet in A. triostegus than in A. nigrofuscus (Figure 2.4). Measures of individual variation within the population also showed less variability across sites in A. nigrofuscus. Acanthurus triostegus showed significant differences between sites in the mean diet overlap between individuals within the population (Table 2.1). Similarly, there were significant differences between sites in the degree of mean overlap between an individual and the population. In A. nigrofuscus, there were no significant differences between sites in the diet overlap between individuals and the overall population. In A. triostegus, the mean individual diet overlap was influenced almost exclusively by site level differences (R2 marginal: 0.03, R2 conditional: 0.88). Mean individual diet overlap also increased significantly with increased diet richness (estimate: 0.02, P = 0.04) and decreased as the weight of the fish increased (estimate: -0.02, P < 0.001) (Figures S2.6). Similarly, in A. triostegus, site level differences drove variation in the proportion of the population diet breadth represented by an individual (R2 marginal: 0.08, R2 conditional: 0.72) (Figure S2.7). Far less of the diet variation was explained by site level differences in A. nigrofuscus as compared to A. triostegus (Figures S2.8 and S2.9). Specifically, in the model examining mean

12 diet overlap between individuals, only 10% of the variation was explained by random effects (R2 marginal: 0.39, R2 conditional: 0.49). There was a significant negative relationship between the weight of the fish and the diet overlap (estimate: -0.02, P = 0.02) and a significant positive relationship with diet richness (estimate: 0.05, P = 1.13e-07). Similarly, when individual diet breadth was compared to the population, site level differences did not explain much of the variation (R2 marginal: 0.38, R2 conditional: 0.41).

Discussion Metabarcoding provided a detailed look at the composition, breadth, and diversity of the diets of two grazing surgeonfishes and facilitated the identification of more species than would have been possible using visual identification of stomach contents alone. While A. triostegus and A. nigrofuscus have considerable overlap in the algae they target, A. nigrofuscus has a more diverse diet that incorporates more of the epilithic algal matrix and brown algae. Previous accounts have emphasized that unlike congeners A. nigroris or A. guttatus, A. triostegus grazes on the epilithic algal matrix only incidentally, instead, directly biting off pieces of filamentous algae84. Unlike some other surgeonfishes, A. triostegus lacks the thick walled stomach or gizzard that indicates sediment consumption84,103. Our metabarcoding analysis showed little evidence of detritus consumption from the epilithic algal matrix by A. triostegus in most sites (i.e., there was little cyanobacteria, Miozoa, Chromerida, Cercozoa, or Bacillariophyta) (Figures 2.1, S2.3). In comparison, A. nigrofuscus has a more diverse diet and consumes more detritus than A. triostegus, as indicated by the greater proportion of cyanobacteria, Miozoa, Chromerida, and Cercozoa, in addition to green and brown algae (Figures 2.2, S2.4). Schooling behavior in A. triostegus may help explain the greater diet overlap and lower levels of individual specialization in A. triostegus as compared to A. nigrofuscus. Being a larger, schooling species may afford A. triostegus better access to preferred resources on the reef. In comparison, A. nigrofuscus, a smaller fish, is aggressive, and while it may school, it tends to be solitary103. Though A. nigrofuscus has a narrower depth range86, it can thrive in a variety of habitats and appears to be relatively unaffected by typical shallow water disturbances, like increased sediment loads, which is particularly relevant for the impacted reefs of our study sites in Hawaiʻi104,105. These behaviors may contribute to its apparent opportunistic feeding approach.

13 These same ecological factors likely explain why A. triostegus showed greater variability in specialization among individuals in the population than A. nigrofuscus, which had lower and more consistent levels of individual specialization across sites. Larger individuals often have a greater competitive advantage among conspecifics, which can result in access to preferred resources. The degree of diet variability between individuals of A. nigrofuscus followed this trend, as expected, with larger individuals having less diet overlap with other individuals in the population (Figures S2.8 and S2.9). Individuals of A. triostegus, however, showed the greatest diet overlap with one another and with the overall population at an intermediate size (Figures S2.6 and S2.7). This may be due to the schooling behavior of this species, which can provide access to preferred resources that are typically defended by territorial species, but which consequently reduces diet variability among individuals106. Populations of A. nigrofuscus exhibited less diet overlap and far less variability between sites than A. triostegus (Figure 2.4). A. triostegus also displays greater diet differentiation between sites than A. nigrofuscus (Figure 2.3), likely indicating a targeted feeding strategy that varies by the availability of resources in different locations. This may be influenced by its increased depth range as compared to A. nigrofuscus86. Specifically, it seems that A. triostegus individuals from the Salt Pond on Kauaʻi, from Maunalua on Oʻahu, and from Kailua on Oʻahu have diets that are somewhat distinct from individuals in other locations. Though the exact reasons for these patterns in A. triostegus are not clear, they likely indicate some preferential consumption of certain species based on their relative availability at those locations. A. nigrofuscus had far greater diet overlap between populations in different locations, which indicates its role as an opportunistic feeder with a broader diet. Though A. nigrofuscus were also sampled from a variety of locations, neither their diet nor their degree of diet overlap vary spatially to the extent seen in A. triostegus (Figures 2.4, S2.6-S2.9).

Conclusions This study is to our knowledge the most comprehensive survey to date to examine the diet of herbivorous reef fishes using metabarcoding and sheds light on how they partition resources. We demonstrated that A. nigrofuscus appears to be a more opportunistic feeder with less variability between locations and greater variability among individuals than A. triostegus,

14 despite the two species having considerable overlap in their diet. As metabarcoding and environmental DNA become increasingly common tools for management and conservation, such studies will be key to anticipate functional responses of coral reef herbivores under changing future environmental conditions on coral reefs.

Acknowledgements All collections were made under the University of Hawaiʻi Institutional Animal Care and Use Committee permits 15-2145-2 and 15-2271. Many thanks to Rachael Wade, Molly Timmers, Annick Cros, Ingrid Knapp, Zac Forsman, Richard Coleman, Derek Kraft, and the ToBo and Donahue Labs at the Hawaiʻi Institute of Marine Biology for your advice and helpful conversations. We also acknowledge the special contributions of Stephen Karl to diverse components of this project. Brian Bowen, Mark Hixon, and Amber Wright provided valuable feedback and suggestions throughout the process.

15 Tables Table 2.1. Measures of individual variation in Acanthurus triostegus and A. nigrofuscus. All calculations made using the RInSp package. Values of mean pairwise diet overlap range from 0 to 1, with 1 signifying complete diet overlap. Individual variation at the population level ranges from 0 to 1, with 1 signifying an average individual diet that mirrors that of the total population. Significant differences between sites are indicated with grouping letters for A. triostegus. There were three groups for mean pairwise diet overlap in A. nigrofuscus (a: Ala Moana, Haleʻiwa, Kakaʻako, Lāiʻe, and Maunalua; b: Mokuleʻia; ab: Chuns and Olowalu), but there were no significant differences between sites in individual specialization as compared to the population.

Individual Mean Specialization Number of Pairwise Island Shore Site Species at the Individuals Diet Population Overlap Level Acanthurus Kauaʻi West Salt Pond triostegus 5 0.32 (e) 0.52 (f) Acanthurus Maui North Waiheʻe triostegus 3 0.62 (abd) 0.76 (ace) Acanthurus West Olowalu nigrofuscus 4 0.53 0.69 Acanthurus Oʻahu East Kahana Bay triostegus 14 0.78 (c) 0.84 (d) Acanthurus Kailua Bay triostegus 16 0.62 (b) 0.72 (ae) Acanthurus Laiʻe nigrofuscus 7 0.56 0.69 Acanthurus North Chuns nigrofuscus 4 0.53 0.7 Acanthurus Hāleʻiwa nigrofuscus 7 0.58 0.7 Acanthurus triostegus 11 0.66 (ab) 0.75 (a) Acanthurus Mokulēʻia nigrofuscus 8 0.45 0.6 Acanthurus South Ala Moana nigrofuscus 22 0.54 0.67 Acanthurus Kakaʻako nigrofuscus 4 0.58 0.72

16 Table 2.1. (Continued) Measures of individual variation in Acanthurus triostegus and A. nigrofuscus. All calculations made using the RInSp package. Values of mean pairwise diet overlap range from 0 to 1, with 1 signifying complete diet overlap. Individual variation at the population level ranges from 0 to 1, with 1 signifying an average individual diet that mirrors that of the total population. Significant differences between sites are indicated with grouping letters for A. triostegus. There were three groups for mean pairwise diet overlap in A. nigrofuscus (a: Ala Moana, Haleʻiwa, Kakaʻako, Lāiʻe, and Maunalua; b: Mokuleʻia; ab: Chuns and Olowalu), but there were no significant differences between sites in individual specialization as compared to the population.

Acanthurus triostegus 4 0.77 (c) 0.85 (bd) Maunalua Acanthurus Bay nigrofuscus 16 0.59 0.71 Acanthurus triostegus 14 0.53 (d) 0.66 (e) Kahe Point (Electric Acanthurus West Beach) triostegus 12 0.69 (a) 0.78 (abc) Acanthurus Makaha triostegus 10 0.75 (c) 0.83 (bcd)

17 Figures Rhodophyta Chlorophyta Ochrophyta Cyanobacteria Miozoa Chromerida Cercozoa Bacillariophyta Diet Items The of distribution

Waihee ) B

(Maui) (

. Salt Pond The distribution of of distribution The

(Kauai) ) C ( . Diet Richness

Maunalua

Makaha 20 30 40 50 60 Kaneohe 5 0 10 Kakaako C ; bareach a represents unique sample

Kailua

Kahana triostegus Acanthurus

Haleiwa diet Indexsample by each measured diversity, as Simpson’s on diet richness. Diet Diet composition of

) A Diet Diversity (Simpson’s Diversity Index) (Simpson’s Diversity Diet Diversity Electric Beach ( A. triostegus A. triostegus A. triostegus

0.4 0.6 0.8 1.0 0%

5 0 75% 50% 25%

10 100%

Number of Individuals of Number Percent of Diet of Percent Figure 2.1. individual individual A B 18 Rhodophyta Chlorophyta Ochrophyta Cyanobacteria Miozoa Chromerida Cercozoa Bacillariophyta Diet Items

Olowalu (Maui) The distribution distribution The

distribution of )

B (

Mokuleia . The

) C ( Diet Richness

Maunalua 20 30 40 50 60

9 6 3 0 Laie Individuals of Number C ; bareach a represents unique sample

Kakaako

Haleiwa

Electric Beach

Chuns Acanthurus nigrofuscus Acanthurus diet diversity, as measured by Simpson’s diet Indexsample. by each measured diversity, as Simpson’s on

diet richness.

Ala Moana Diet Diversity (Simpson’s Diversity Index) (Simpson’s Diversity Diet Diversity Diet Diet composition of

) A. A. nigrofuscus A ( A. A. nigrofuscus 0.4 0.6 0.8 1.0 0%

5 0 75% 50% 25%

15 10 100% Percent of Diet of Percent Individuals of Number Figure 2.2. of individual individual A B 19 Laurencia sp.

1.0 Sphacelaria tribuloides

0.4 Champiaceae Laurencia sp. Laurencia sp. Laurencia sp. Oscillatoriales Gigartinales Dasya kristeniae Sphacelaria tribuloides Dictyotaceae Asparagopsis taxiformis Ceramium sp.Champiaceae Oscillatoriales Nemaliophycidae Laurencia sp. Ceramiaceae0.4 DictyotaceaeEuptilocladia magruderi Dasya kristeniae 1.0 Sphacelaria tribuloides Erythrocolon podagricum Ceramiaceae 0.5 Oscillatoriophycideae GayliellaCeramium sp. sp. Species 0.4 Champiaceae Laurencia sp. Laurencia sp. Nemaliophycidae Gigartinales Oscillatoriales Herposiphonia sp. Euptilocladia magruderi Gigartinales Dasya kristeniae Rhodophyta Oscillatoriophycideae Asparagopsis taxiformisHerposiphonia sp. Electric Beach Sphacelaria tribuloides Dictyotaceae Asparagopsis taxiformis Ceramium sp.Champiaceae Oscillatoriales Erythrocolon podagricum Gayliella sp. Puena Point Nemaliophycidae HerposiphoniaLaurencia sp. sp. Ceramiaceae0.4 DictyotaceaeEuptilocladia magruderi Dasya kristeniae Rhodophyta Erythrocolon podagricum Ceramiaceae Herposiphonia sp. Kahana Bay 0.5 Oscillatoriophycideae GayliellaCeramium sp. sp. Species 0.0 Nemaliophycidae Gigartinales 0.0 Herposiphonia sp. Euptilocladia magruderi Rhodophyta Oscillatoriophycideae Asparagopsis taxiformisHerposiphonia sp. Electric Beach Kailua Bay 0.0 Erythrocolon podagricum Gayliella sp. Herposiphonia sp. NMDS2 NMDS2 Puena Point Kakaako NMDS2 Rhodophyta Herposiphonia sp. Kahana Bay Kaneohe Bay 0.0 0.0 Kailua Bay 0.0 Makaha NMDS2 NMDS2 Kakaako NMDS2 Kaneohe Bay Polysiphonia sp. Maunalua Ceramiales −0.4 Makaha Chondracanthus acicularis Waihee −0.5 Polysiphonia sp. Maunalua Ulva sp. Ceramiales −0.4 Chondracanthus acicularis PolysiphoniaWaihee sp. Ulva sp. −0.5 −0.4 Ceramiales Polysiphonia sp. Chondracanthus acicularis −0.4 Ceramiales Chondracanthus acicularis Ulva sp. Rhodymeniophycidae Ulva sp. Rhodymeniophycidae −0.8 −0.8 −0.5 0.0 0.5 1.0 −0.5 0.0 0.5 1.0 −0.5 0.0 0.5 1.0 −0.5 0.0 0.5 1.0 NMDS1 NMDS1 NMDS1 NMDS1

Figure 2.3. (a) Non-metric multidimensional scale (NMDS) plot using Bray-Curtis similarity −0.8 matrix for Acanthurus triostegusRhodymeniophycidae by location (color). Stress was 0.091 with k = 4. The large amount of variability−0.8 between individuals at Salt Pond, Kauaʻi obscured the differencesRhodymeniophycidae between −0.5 0.0 0.5 1.0 other sites, so it was notNMDS1 included in −this0.5 plot. All sites are on Oʻahu, 0.0except for Waiheʻe, which 0.5 1.0 is on Maui. NMDS plot overlaid with ellipses that show a 95% confidence intervalNMDS1 on site clusters.

20 Laurencia sp.

1.0 Sphacelaria tribuloides

0.4 Champiaceae Laurencia sp. Laurencia sp. Oscillatoriales Gigartinales Dasya kristeniae Sphacelaria tribuloides Dictyotaceae Asparagopsis taxiformis Ceramium sp.Champiaceae Oscillatoriales Nemaliophycidae Laurencia sp. Ceramiaceae0.4 DictyotaceaeEuptilocladia magruderi Dasya kristeniae Erythrocolon podagricum Ceramiaceae 0.5 Oscillatoriophycideae GayliellaCeramium sp. sp. Species Nemaliophycidae Gigartinales Herposiphonia sp. Euptilocladia magruderi Rhodophyta Oscillatoriophycideae Asparagopsis taxiformisHerposiphonia sp. Electric Beach Erythrocolon podagricum Herposiphonia sp. Gayliella sp. Puena Point Herposiphonia sp. Rhodophyta Kahana Bay 0.0 0.0 Kailua Bay 0.0 NMDS2 NMDS2 Kakaako NMDS2 Kaneohe Bay Makaha Polysiphonia sp. Maunalua Ceramiales −0.4 Chondracanthus acicularis Waihee −0.5 Ulva sp. Polysiphonia sp. −0.4 Ceramiales Chondracanthus acicularis

Ulva sp. Rhodymeniophycidae −0.8 −0.5 0.0 0.5 1.0 −0.5 0.0 0.5 1.0 NMDS1 NMDS1

Figure 2.3. (b) Non-metric multidimensional scale (NMDS) plot using Bray-Curtis similarity matrix for Acanthurus triostegus by location (color). Stress was 0.091 with k = 4. The large −0.8 amountRhodymeniophycidae of variability between individuals at Salt Pond, Kauaʻi obscured the differences between −0.5 0.0 other sites, so it was not included0.5 in this plot. All sites are on1.0 Oʻahu, except for Waiheʻe, which is on Maui.NMDS1 NMDS plot overlaid with diet items (i.e., algal taxa).

21

Mokuleia Mokuleia

MaunaluaBay MaunaluaBay

A. A. nigrofuscus

Makaha Makaha

Laie Laie

(blue) and (blue) and

Kakaako Kakaako s s triostegus

Oahu Oahu

KailuaBay KailuaBay

Acanthurus triostegus Acanthurus triostegus

Acanthuru

KahePoint KahePoint KahePoint KahePoint

KahanaBay KahanaBay

Haleiwa Haleiwa

Acanthurus nigrofuscus Acanthurus nigrofuscus

Chuns Chuns

Species Species

AlaMoana AlaMoana

Waihee Waihee

Maui Maui

Olowalu Olowalu Individual pairwise diet overlap by site is for shown pairwise Individual diet by overlap site

. 4

Kauai Kauai

SaltPond SaltPond

Figure 2. individual specialization. (orange). less diet Higher indicate values 0.8 0.6 0.4 0.8 0.7 0.6 0.5 0.4 0.3 0.8 0.6 0.4 0.8 0.7 0.6 0.5 0.4 0.3 Diet Overlap Diet Overlap Diet Diet Overlap Diet Overlap Diet

Individual Pairwise Pairwise Individual Pairwise Mean Individual Pairwise Pairwise Individual Pairwise Mean A B A B

22 CHAPTER 3 PICKY EATER OR GENERALIST FEEDER? COMPARING DIET DIVERSITY OF HERBIVOROUS CORAL REEF FISHES IN THE PACIFIC

Abstract Diet is an important component of an animal’s ecological role and can be used as a measure of specialization. Herbivore diets have been examined using a myriad of methods, making it difficult to compare between analyses and to examine specialization across large spatial scales and heterogeneous environments. We addressed these limitations by compiling a database of information on Pacific herbivorous reef fishes through a systematic review of diet data. The diet diversity of each species was assessed, and a single consumption diversity index value was generated to represent diet specialization. In addition, we independently examined the diet diversity of ten species using metabarcoding to sequence the algae present in their stomach contents. We compared the results of the two approaches and showed that considerable diet variability exists within functional groups and between closely related species. Parrotfishes and detritivorous surgeonfishes have more limited diets than other nominally herbivorous fishes, but within each functional and taxonomic group there is a range of diet specialization. Finally, we demonstrate how our diet specialization index can be incorporated into trait matrices that offer valuable insight into the dissimilarity between species, regardless of their taxonomic and functional classification. This analysis offers a simplified index that can be applied at large scales to examine ecological roles across reef systems and can assist in the reevaluation and management of herbivores and their functional effects on changing, degraded reefs.

Introduction A compelling case has been made in recent years for prioritizing the maintenance of functional traits, and, especially, critical ecosystem functions over traditional conservation metrics such as species richness or biodiversity107. The rate at which habitats and ecological communities are changing in response to anthropogenic impacts is outpacing our ability to adequately describe, classify, and most importantly, protect the full range of biodiversity observed in many systems. The ecosystem function performed by an organism, defined by

23 Bellwood et al. (2019) as the way an organism transports or stores energy, can thus be a particularly useful way of summarizing the ecological role and potential importance of an organism in its community. Based on this definition of function, an organism’s diet is a critical defining trait, along with behavioral and morphological characteristics, and these descriptive features can be used to categorize the species’ role in its community. This trait-based approach has been successfully incorporated into analyses of ecosystem function and response to stress in both terrestrial108–111 and aquatic systems112–116. For most animals, a fundamental component of their functional role is by definition their diet107, yet in many cases our understanding of the degree of diet partitioning and variability that exists among closely related species has remained quite limited117–123. In this study we contribute to the recent advances that have been made in understanding functional diversity between and within herbivorous coral reef fishes14,118,119,124 by compiling and collating existing diet descriptions, contributing new diet data generated using metabarcoding, and combining these data to generate an index of diet specialization that can be incorporated into trait matrices used to examine ecosystem function. Herbivorous reef fishes serve a vital and irreplaceable role in coral reef ecosystems16. They have received much research attention and for many years have had commonly referenced (though not entirely accurate) functional group assignments107,125. Nominally herbivorous reef fishes are critical to ecosystem function on coral reefs, as they consume fleshy macroalgae that compete with corals for space126, turf algae that proliferate on degraded reefs15, and detritus that can smother corals127. In doing so, they help maintain a diverse assemblage of benthic organisms and promote the structural complexity of one of the most diverse ecological communities on the planet5,128. The diversity of functional roles and disturbance responses exhibited by organisms can facilitate greater resilience in the face of anthropogenic impacts129,130. Thus, it is unsurprising that herbivore overfishing has been linked to dramatic shifts in reef assemblages and long-lasting consequences to ecosystem function7,131,132. The degree of functional redundancy varies across groups and may in some cases be a coincidence of evolutionary history that is not obvious at first glance. In parts of the Atlantic, gobies and blennies dominate reef fish diversity as a result of species radiations but fill a limited number of functional roles133,134. Herbivorous fishes were historically assumed to have high dietary overlap within their functional group135, but recent studies have highlighted diet and

24 functional variability within this group124,136–139. For example, the Green Humphead Parrotfish (Bolbometopon muricatum), has been identified as a critical bioeroder on coral reefs, and the loss of this one species has been undeniably linked to loss of ecosystem function21. Furthermore, the definition of true ecological function remains a topic of debate within the field of herbivory. Some suggest that the function of an animal is integrally linked to its nutritional mode14, but for the purposes of this study we are using the definition of function which encompasses an animal’s interaction with its environment as it relates to the transfer of energy107. While both interpretations are valid, the latter definition is more appropriate for understanding how the relationships between herbivores and their environment may be affected by human impacts such as habitat degradation. A variety of approaches have been used to examine the diets of herbivorous reef fishes, and understandably, different approaches may color the interpretation of diet and function. For example, feeding choice assays can assess numerous characteristics ranging from diet preference to bite rate to consumer identity126,140,141. Others have used visual identification of consumed species or substrates, which can be accomplished in situ142 or remotely via video143. A classic approach has been to use gut content analysis, and this can be done using visual inspection144 or, more recently, using molecular approaches such as DNA barcoding145 and metabarcoding83,146. Dietary overlap has also been examined using chemical approaches such as stable isotope analysis of consumer tissues and prey items117. Advances have also been made recently using short-chain fatty acid analysis, which can illuminate resource partitioning among even closely related fishes14. While each approach offers a unique view into the complexity of herbivory on coral reefs, drawing comparisons between different approaches can be challenging, making analyses that extend beyond a local scale extremely difficult and, therefore, infrequent in the literature. To understand the trends, responses, and habitat shifts of functionally important herbivores on large spatial scales, we must examine the responses of not only single animals but also whole assemblages and communities. Compiling the rich data that has been acquired in site- specific studies can facilitate comprehensive analyses that address questions on a regional or even global scale. Assessing processes that occur across the full geographic range of a species, however, may require some simplifying assumptions about variability on a local scale. To that

25 end, we generated a standardized index of diet diversity that could be used to compare herbivore diet specialization among locations and studies. The conceptual framework was inspired by terrestrial community-level indices of specialization that can be used to examine functional homogenization in response to human impacts25–27. In these cases, functional homogenization implies that generalist species are able to outcompete more specialized species in degraded habitats, which means that specialists will decrease in abundance with increasing human impacts. To assess homogenization at the community scale, however, a measure of species specialization is first required. By conducting a literature review of existing diet data for herbivorous reef fishes in the Pacific and compiling this information into a database, we generated an index of diet diversity that can be compared among species. We then compared the results of this specialization index, based on existing sources, to metabarcoding diet data for ten species of herbivores that are commonly found in Hawaiʻi that was generated using metabarcoding. Finally, we demonstrated how these data can be incorporated into a trait matrix that sheds light on functional dissimilarity between species. Together, these complementary approaches provide an overview of the range of specialization within and among functional groups of herbivores, demonstrate the utility of new metabarcoding approaches, and provide context and references for future studies to analyze shifts in assemblages and community function at large spatial scales.

Methods I. Examination of Existing Data A. Literature Review A list of herbivorous fish species was developed based on a NOAA coral reef monitoring dataset from the Pacific Reef Assessment and Monitoring Program (RAMP)147. RAMP monitors fishes by underwater visual surveys via the stationary point method in nearshore shallow reefs at U.S. affiliated islands and atolls throughout the western central Pacific148,149. In nearly 4000 surveys, 78 herbivorous fish species were recorded (GitHub: enalley/Dissertation/SupplementalFile3.1). To generate a species-specific diet diversity index, we conducted a systematic review of peer-reviewed diet data for those 78 herbivore species (GitHub: enalley/Dissertation/SupplementalFile3.2). The review approach incorporated

26 recommendations from the Collaboration for Environmental Evidence’s Guidelines for Systematic Reviews in Environmental Management150. Consistent search terms were used in Web of Science (i.e., (“Genus species”) AND (diet OR alga* OR herb* OR graz* OR brow* OR excav*)). Additional sources were examined from FishBase using rfishbase package in R86,151. Inclusion was limited to sources from the Pacific that had quantitative, proportional gut content data or data that could be converted into this form (GitHub: enalley/Dissertation/SupplementalFile3.3). Studies using laboratory experiments, feeding-choice assays, or field observation of feeding were not included. The literature review generated 376 peer-reviewed article abstracts that were examined based on key factors including relevance, location, and type of study (GitHub: enalley/Dissertation/SupplementalFile3.2). The search found no diet data for Acanthurus leucocheilus, Calotomus zonarchus, Centropyge fisheri, Centropyge heraldi, Centropyge loricula, Centropyge shepardi, Cetoscarus ocellatus, Naso brachycentron, or Scarus xanthopleura. Further inspection of the abstracts highlighted 131 of these 376 papers that were potentially applicable to this study as they contained quantitative, proportional diet data (or data that could be converted into this form). After more detailed review of the type of study (e.g., feeding assay, morphological study, lab experiment, gut content analysis), 66 of those 131 papers were identified for potential inclusion in the analysis. As described above, inclusion was limited to studies that had quantitative, proportional gut content analysis data that could be compared between studies. Ultimately, diet data from 6 publications encompassing 31 species were used to generate the species level diet diversity index (Table 3.1). One set of authors identified two distinct habitats in their publication, so these were included as two separate studies152. A discussion of descriptive diet data available for each species is available in the Appendices (Text S3.1).

B. Generation of the Species Diet Diversity Index Diet data were first combined into a datasheet that mirrored the output of FishBase diet queries so that they could be used in that format by others in the future (GitHub: enalley/Dissertation/SupplementalFile3.2). We then generated a matrix consisting of the herbivore species, study author, number of samples, and proportional diet data for each diet item

27 included in that study. This resulted in a matrix that resembles a relative abundance matrix for a community dataset. When the original diet data were presented as relative abundance, they were input directly into the matrix described above. If information was available only in the form of a graph, WebPlotDigitizer was used to translate it into numerical values153. When diet data were provided in a different form (e.g., Debenay et al. 2011) they were converted into relative abundance values. The conversion methods used are described in detail for each source, alongside the original data (GitHub: enalley/Dissertation/SupplementalFile3.3). Though each study provided slightly different diet categories, we standardized the categories for this dataset by converting them to higher taxonomic classifications which were comparable between studies (GitHub: enalley/Dissertation/SupplementalFile3.4). The diet categories used were as follows: Rhodophyta (macroalgae), Chlorophyta (macroalgae), Ochrophyta (macroalgae), turf algae, epilithic algal matrix (EAM) (e.g., sediment, organic matter, cyanobacteria), as well as benthic invertebrates, and zooplankton for more omnivorous species. Simpson’s Diversity Index, which accounts for evenness and abundance and controls for the number of diet components, was applied to the relative abundance of diet items in each species’ diet using the Vegan package in R95. A single value between 0 and 1 was generated that represented the diversity of that species’ diet in that particular study. Higher values signified greater diversity in the diet (i.e., a more generalized diet). For species included in multiple studies, the diet diversity was calculated for each study, and then a single diet diversity value was calculated for each species by averaging across studies weighted by the sample size used in each study. This generated a value of the species’ mean diet diversity (i.e., mean Simpson’s Diversity Index) (Table 3.1).

C. Comparing Diet Diversity between Groups The variability in the mean diet diversity values between functional and taxonomic groups was examined (GitHub: enalley/Dissertation/SupplementalFile3.5). Mean diet diversity values were not normally distributed (Shapiro-Wilk normality test: W = 0.827, P = 0.0002). The non-normality and small number of groups (6 functional groups, 6 families, and 11 genera) may result in a large number of ties in the post-hoc Dunn test for group comparisons. Therefore, the differences between the within-group mean and median diet diversity values of functional

28 groups, families, and genera were examined using a permutation test in rcompanion in R97. Differences in diet composition between functional groups were examined using a non-metric multidimensional scale plot with significant (P < 0.05) diet items overlaid. All analyses were performed in R statistical software93. The code and data used to complete this analysis are available on GitHub (in enalley/Dissertation).

D. Trait Based Dissimilarity Matrix The diet diversity values of each species were used to rank diet diversity into four categories. (i) Very low diversity diets had a mean value less than 0.1; (ii) low diversity diets had a value between 0.1 and 0.4, (iii) medium diversity diets had a mean value between 0.4 and 0.65, and (iv) high diversity diets exceeded 0.65. These ranks were incorporated into a trait matrix that also included the functional group, taxonomic family, dominant food type, territoriality, typical habitat, size class, schooling behavior, depth range, and geographic distribution (GitHub: enalley/Dissertation/SupplementalFile3.6)86,107. Dissimilarity between species based on their trait matrix was measured using Gower’s distance to better understand the relationships between different species, without being confined by their functional and taxonomic groups, and the results were clustered into a dendrogram using Gower’s distance metric in the cluster package in R154.

II. Integrating Metabarcoding Diet Data A. Sample Collection & Processing Whole fish samples were collected in Hawaiʻi between July 2015 and July 2017 at 18 locations on Oʻahu (14), Kauaʻi (1), and Maui (3) during daylight hours (i.e., 8:00-16:00) (Figure S3.1). This sampling included species that largely feed on detritus and the EAM (i.e., Ctenochaetus strigosus and Acanthurus olivaceus) and large browsers that feed on macroalgae (i.e., Naso lituratus and N. unicornis), as well as grazing surgeonfishes including A. guttatus, A. leucopareius, A. nigrofuscus, A. nigroris, A. triostegus, and Zebrasoma flavescens that primarily target filamentous turf algae. Fish were placed on ice immediately after collection and were frozen at -20 °C until further processing. Metadata for each sample are available on GitHub (enalley/Dissertation/SupplementalFile3.7). Fish were dissected, and stomach contents were

29 removed. For each individual, 0.25 g of stomach content material was obtained and used for further analysis. Stomach content material was put directly into PowerBead tubes provided in the MoBio PowerSoil DNA Extraction Kits (now Qiagen DNEasy PowerSoil Kits – Qiagen, Hilden, Germany), which were used to extract DNA from the stomach contents.

B. Metabarcoding The metabarcoding methods used in this study are outlined in Chapter 2, as some individuals were used in both studies. To summarize, the MoBio PowerSoil DNA Extraction Kit factory protocol was followed with the addition of a second elution step with 50 µl of PCR water. The Illumina 16S Metagenomic Sequencing Library protocol was followed with minor changes as described in Chapter 288. The 23S rRNA Universal Plastid Amplicon (UPA) marker developed by Sherwood and Presting (2007) was used in place of 16S for this study; this marker is conserved across algal lineages but is differentiated among closely related taxa, making it an excellent marker for metabarcoding studies focused on algae80. As described in Chapter 2, Illumina adaptors were added to the 5’ end of the forward 23S UPA primer from Sherwood and Presting (2007) and to the 5’ end of the reverse 23S UPA primer in Saunders and Moore (2013) (forward: 5’-TCGTCGGCAGCGTCA- GATGTGTATAAGAGACAGGGACAGAAACACCCTATGAA, reverse: 5’ –GTCTCGTGG- GCTCGGAGATGTGTATAAGAGACAGTCAGCCTGTTATCCCTAGA). The 23S Universal Plastid Amplicon marker is especially good for the identification of red algae. In Hawaiʻi where these samples were taken, the majority of algae, especially turf algae, falls within Rhodophyta, so this choice was particularly important for grazing surgeonfishes (A. triostegus, A. nigrofuscus, A. leucopareius, A. nigroris, A. guttatus, and Z. flavescens). 25 µl reactions were used with the following touchdown PCR profile90: 2 minutes at 95 °C, 35 cycles of 94 °C for 30 seconds, 66 °C for 30 seconds with the temperature reduced by 0.5 °C every cycle until reaching 58 °C, 72 °C for 30 seconds, and then a 5 minute extension at 72 °C. PCRClean DX magnetic beads (Aline Biosciences, Woburn, Massachusetts, United States) were used to clean the PCR products before the indexing PCR. Kapa HiFi HotStart Ready Mix was used for all indexing PCRs (Roche, Basel, Switzerland). Nextera XT Indices (Illumina, San Diego, California, United States) were used following the manufacturer’s protocol for one

30 sequencing run with the following indexing PCR protocol: 95 °C for 3 minutes; 8 cycles of 95 °C for 30 seconds, 55 °C for 30 seconds, and 72 °C for 30 seconds; and 72 °C for 5 minutes. Kozich et al. (2013) indices were used for the rest of the samples in 25 µl reactions with the following PCR protocol: 95 °C for 3 minutes; 8 cycles of 95 °C for 30 seconds, 55 °C for 30 seconds, 72 °C for 30 seconds; and 72 °C for 30 seconds. Indexed PCR products were cleaned again with PCRClean DX magnetic beads and were quantified using Accuclear Ultra HS dsDNA Quantitation Kit (Biotium, Fremont, California, United States) on a SpectraMax Microplate Reader (Molecular Devices, San Jose, California, United States). Normalized, pooled libraries were sequenced at the Genetics Core Facility at the Hawaiʻi Institute of Marine Biology (Kāneʻohe, Hawaiʻi, United States), part of the University of Hawaiʻi at Mānoa, and at the Genetics Core Facility at the Advanced Studies in Genomics, Proteomics, and Bioinformatics at the University of Hawaiʻi at Mānoa (Honolulu, Hawaiʻi, United States) on the Illumina MiSeq platform with the V3-600 cycle kit (2x300 bp). Libraries were demultiplexed by the sequencing lab, and DADA2 algorithm in R was used to filter, trim, and align raw, unpaired reads92. Amplicon sequence variant (ASV, akin to operational taxonomic unit) tables were generated by DADA2, and a pseudo pooling approach was used to aid in the recognition of ASVs. The informatic sequence classification trees (INSECT) program in R was used to make taxonomic assignments of ASVs using a reference classification tree of 23S UPA sequences50 (GitHub: enalley/Dissertation/SupplementalFile3.8). Technical replicates were combined and only samples with over 10,000 reads were retained. The Vegan package in R was used to determine the richness and diversity of each sample, and for samples that had biological replicates, the replicate with the highest diversity was retained95. The phyloseq package in R was used to rarefy each sample to an even depth of 10,000 reads to reduce biases between samples and to generate the relative read abundance for each sample51. Species diet overlap was examined using a Bray-Curtis similarity matrix with a non-metric multidimensional scaling ordination using the Vegan package95. The R code and all of the associated raw data and input files required to run the script are provided for reference on GitHub (in enalley/Dissertation). Though one of the main interests of the metabarcoding analysis was to develop a more detailed understanding of the taxa that comprise turf algae and their consumption by grazing

31 surgeonfishes, for the purposes of comparison to data from the literature review, which primarily relies on the category ‘turf algae’, the diversity of taxa representing turf were condensed into one group. Algae that were not identified to a taxonomic level that could be attributed to turf were left in the higher taxonomic classifications, which largely fell within Rhodophyta. While some of these reads are assigned to red macroalgae, such as Acanthophora spicifera or Asparagopsis taxiformis, some of them almost certainly are turf algae that lack reference sequences.

Results I. Examination of Existing Data A. Diet Diversity Taxonomic groupings were not clear indicators of diet diversity. There were no mean differences between families (permutation tests: all adj. P > 0.05) or genera (permutation tests: all adj. P > 0.05), and several groups had very high within-taxon variability (Figures 3.1, 3.2, S3.2, S3.3). Acanthurids exhibit a large range of within-group diet diversity (mean diet diversity = 0.50, range = 0.03-0.77) and include detritivores (mean diet diversity = 0.21, range = 0.03- 0.46), which have a significantly lower diet diversity than browsers (mean diet diversity = 0.60, range = 0.50-0.68, P = 0.03), grazers (mean diet diversity = 0.64, range = 0.41-0.70, P = 0.00), or omnivores (mean diet diversity = 0.70, range = 0.61-0.77, P = 0.02) (Figure 3.1). For example, Acanthurus nigricauda had the least diverse diet in the index (diet diversity = 0.03), primarily eating calcareous sediment and organic matter from the EAM119,144. However, its congener, A. dussumieri, had one of the most diverse diets (diet diversity = 0.775) and ate macroalgae, turf algae, and a variety of benthic invertebrates, along with cyanobacteria, sediment, and foraminifera from the EAM155. A. lineatus exhibited a different but also diverse diet (diet diversity = 0.703), consuming red thallate and filamentous algae, along with smaller amounts of green and brown algae, as well as organic matter and sediment from the EAM and a small amount of benthic invertebrates144. Conversely, adults of both Naso annulatus and N. brevirostris feed predominantly on zooplankton144. Considerable variability also existed in the dietary breadth of the different functional groups examined (Figures 3.2-3.4). Omnivores, of course, had the broadest diets and consumed all diet categories. Grazers also had on average greater diet breadth that overlapped with

32 omnivores and browsers and was dominated by red algae and turf. The diets of browsers overlapped with omnivores, grazers, and planktivores. Both planktivores considered in this study were Naso spp. that exhibited overlap with browsers, despite distinct differences in their consumption of plankton. Though few excavators and scrapers were included in the meta- analysis, their diets overlapped with detritivores, which indicates related ecological functions though the groups have distinct nutritional modes.

B. Cluster Analysis of Herbivorous Fish Using a Dissimilarity Matrix Parrotfishes clustered together distinct from all other herbivores examined, despite the unique diet of Scarus ghobban (Figure 3.4). Next, all grazing fishes, with the exception of Zebrasoma scopas and Siganus argenteus, which are both medium sized schooling herbivores with high diversity diets that are found down to 40-60 m throughout the Indo-Pacific, then clustered together as distinct from the remaining species. Z. scopas and S. argenteus clustered most closely with S. punctatus and A. dussumieri, which were classified as omnivores. These species were most closely related to three browsing species, Kyphosus vaigiensis, K. cinarescens, and Naso lituratus. The other Naso spp. (N. unicornis, N. brevirostris, and N. annulatus) all clumped together, despite the fact that N. unicornis is classified as a browser while the other two primarily consume plankton. Finally, all the detritivorous fishes clumped together, despite the fact that they spanned two genera.

II. Comparing Diet Diversity Based on Existing Data to Metabarcoding Data The diets of ten species of nominally herbivorous surgeonfishes were examined using DNA metabarcoding to help ground-truth the diet diversity values (Figure 3.5; Table S3.1). Complete lists of all of the diet items identified using metabarcoding can be found on GitHub (enalley/Dissertation/SupplementalFile3.9-10). The actual amount of detritus or sediment that was detected in the gut contents was not quantified as part of this study, so the category of the epilithic algal matrix (EAM) reflects the relative abundance of taxa typically found in the EAM and is likely an underestimate. The presence of cyanobacteria, Chromerida, Miozoa, Cercozoa, and Bacillariophyta was generally considered to be indicative of consumption of the EAM. This

33 was confirmed by the fact that C. strigosus and A. olivaceus (detritivores) had significant proportions of sequences from those diet categories (Figure 3.5). The metabarcoding results supported the findings of the meta-analysis but provided significantly more taxonomic resolution (Text S3.1; GitHub: enalley/Dissertation/SupplementalFile3.9-10). The two browsers examined, Naso lituratus and N. unicornis both had brown algae relative read abundances of more than 50%, with more than 10% represented by turf algae. The remaining sequences were from red macroalgae or unidentified rhodophytes, with a very small proportion of EAM represented. More than 10% of the relative read abundances in the two detritivores, Ctenochaetus strigosus and Acanthurus olivaceus, were from taxa representing the EAM, though it is understood that this is an underestimate, as explained above. Turf algae represented ~20% of the reads in both taxa, and in A. olivaceus more than 25% were from brown algae, with the rest being assigned to red macroalgae or unidentified rhodophytes. In C. strigosus more than 50% of the reads were assigned to red macroalgae or unidentified rhodophytes, with a small portion assigned to brown and green algae. Among the grazers A. guttatus, A. nigroris, and Zebrasoma flavescens consumed the most EAM (~15%, ~5%, and ~5% respectively). Approximately 40% of the reads in A. nigroris were assigned to brown algae, with the remaining sequences assigned to red macroalgae or unidentified rhodophytes. A. nigrofuscus and A. leucopareius both had ~15% brown algae reads with a majority of reads being assigned between red and turf algae (more than 25% of each). Almost all of the reads in A. triostegus were split between red and turf algae, with less than 15% assigned to brown, green, and EAM.

Discussion Herbivorous reef fishes are often lumped together as a homogenous trophic level or differentiated into a few functional groups such as grazers, scrapers, browsers, and excavators. While these divisions can be useful for broad assessments of ecosystems and communities, they can also result in overgeneralizations about the similarities among taxa. Combining diet data from multiple sources for 31 herbivores, we have demonstrated that variability exists in the degree of specialization within functional groups of herbivorous reef fishes and even between species in the same genus. Herbivores occur at high densities on coral reefs, often forming mixed

34 species schools, so this work further illuminates the ways in which closely related taxa partition resources and occupy numerous, though sometimes overlapping, functional roles. Though herbivore assemblages on coral reefs do exhibit some functional redundancy in a broad sense135,156, our understanding of resource partitioning within and among functional groups of herbivores has been shifting. Several studies have shown that certain species may serve unique and vital functional roles within their communities136,157,158. Some functional groups, such as grazers, which eat predominantly fine filamentous turf algae, appear to have far more diet and functional heterogeneity than previously understood124,139, and they may feed on species unobserved in visual surveys of the benthos83. Because of dietary partitioning even among closely related taxa, morphological specialization is a better indicator of the microhabitat that a fish will use for foraging rather than the actual degree of diet specialization of that fish119. In addition, our collective understanding of the true variety of feeding habits of nominally herbivorous fishes is becoming more nuanced. For example, scraping and excavating parrotfishes, which have been previously shown to have a good deal of diet variability within genera137, appear to target microorganisms within the reef matrix14. Resource partitioning has also been documented through differences in foraging strategy and access to resources within the reef matrix, perhaps allowing fishes with outwardly similar functional roles to coexist118. Additionally, specialized behavior, such as coordinated vigilance in rabbitfishes, allows for the exploitation of risky resources and can expand species’ resource use159. Although these results largely support the general expectations of certain functional groups being more or less specialized feeders, there is considerable variation among species within groups making it difficult to make inferences based on these broad functional or taxonomic groups. For example, we found that despite the fact that parrotfishes are more specialized in general than other herbivores, one species (Scarus ghobban) stood out as having very high diet diversity (Figure 3.1). Scarus ghobban is also unique within its genus in other traits: along with a generalist diet, it also exhibits a wide depth and salinity range, which likely contributed to its success in a Lessepsian migration into the Mediterranean86,160. Past research has also found that with the exception of Chlorurus microrhinos, parrotfishes exhibiting higher bite rates also appear to have more specialized diets156.

35 Similarly, among the rabbitfishes there appeared to be a good deal of variability in diet diversity. Siganus spinus, the most specialized of the three Siganus spp. included in this analysis, is prevalent in reef flats and lagoons where resources are more limited118. S. argenteus, which has greater diet diversity, is a schooling species, which can enable access to a greater array of resources118. S. punctatus, which has the highest diet diversity of the three siganids, is a pairing species, which means it may also exhibit coordinated vigilance that facilitates access to a variety of resources118,159. Acanthurids showed a great range of specialization (Figure 3.1), which is to be expected as they represent numerous functional groups (i.e., detritivores, grazers, omnivores)124,144. Acanthurus nigricauda exhibited the highest degree of specialization in these studies, feeding predominantly on the EAM119,144, but its congener A. leucopareius exhibited a great deal of diet variability124. Though A. olivaceus and Ctenochaetus spp. both consume sand and detritus, A. olivaceus supplements its diet with a variety of other items124,144,156, resulting in a higher diet diversity value (Figure 3.1). A. nigroris, which has a thick walled stomach unlike some of its congeners, consumed a higher proportion of detritus as well, which has been reported in other studies161. A. nigroris is likely a more opportunistic feeder and less specialized than some of the other surgeonfishes, such as A. triostegus. Similarly, though Naso unicornis in many cases focuses on macroalgae84,155, other Naso spp. generally exploit a wider range of resources than other acanthurids119,144. These comparisons can be complicated, however, by the foraging microhabitats exploited by different species119. It is thus important to keep in mind that while diet represents an important aspect of an organism’s functional role, behavior may influence its overall ecological role in any particular community107. Our literature review illuminated a key need for fundamental descriptions of species’ diets. Diet data were not available for 14% of the species included in the literature review search, and only about half of the species had diet data available that could be incorporated into a standardized comparison (GitHub: enalley/Dissertation/SupplementalFile3.2). Some species exhibited diet variability between studies, which in some cases was due to differences in habitat, but it may also be indicative of measurement error or variation among sites and times at which data were collected. If we are to understand the response of coral reef communities to human impacts, it is crucial that we continue to gather basic diet data for reef fishes in an array of

36 habitats. Historic diet studies have been the cornerstone of our understanding of trophic dynamics on coral reefs, yet, as reefs change, we cannot rely exclusively on descriptions of the reefs of the past. Invasive species outbreaks, severe coral bleaching events, and numerous other impacts have profoundly altered coral reef communities in the past twenty years7,162,163. It follows that trophic relationships and fish behavior may also be modified on reefs that have transitioned from coral dominated to algae dominated in a matter of years164.

Conclusions The commonly used categories of grazers, browsers, scrapers, and excavators are not sufficient in all circumstances to encompass the range of dietary specialization among these herbivores. Even among closely related species such as parrotfishes, a broad range of diet specialization can still exist. Our analysis contributes a straightforward, standardized way of comparing the diet diversity of multiple species of herbivorous reef fishes using data from different sources25,26. The database of diet composition and herbivore diet studies we generated will facilitate and expedite future work that seeks to examine herbivore specialization at large spatial scales. In addition, using a standardized index to compare diet diversity among taxa allows for expansion into functional trait based analyses107 or assessments of biotic homogenization and shifts in assemblage composition in response to human impacts26,165. Further, this examination of diet diversity supports an expanding body of work indicating that there is variability in the degree of diet specialization and resource partitioning within herbivore assemblages, and even within functional groups and taxonomic families23,83,124,137,139,158. Management actions aimed at protecting herbivores have become increasingly urgent and require a deeper understanding of herbivore diets and whether their diet changes with habitat, through time, or as a consequence of human activities. To draw such inferences, however, we need standardized ways of comparing local studies to examine trends at large spatial scales. We have provided a transparent template by which diet diversity can be compared by other researchers, enhancing the transferability of new diet data. This will catalyze future research on the functional roles of fishes and has the potential to inform management policies.

37 Acknowledgments Collections were approved under an Institutional Animal Care and Use Committee permits (IACUC #: 15-2145-2 and 15-2271). Thank you to members of the Donahue and ToBo Labs at the Hawaiʻi Institute of Marine Biology for their support and feedback. We also particularly thank Molly Timmers and Rachael Wade for their guidance with metabarcoding, and we are very grateful to Shaun Wilkinson for the development of the INSECT pipeline. Brian Bowen, Mark Hixon, and Amber Wright also provided valuable feedback throughout the process. Finally, thank you to the many authors of diet studies on reef fishes who made this work possible.

38 Tables Table 3.1. Mean Diet Diversity Index (DDI) values for each species are listed along with sources used to generate the values. The number of individuals used in each study is noted in parentheses next to the source. Species DDI Sources Acanthurus blochii 0.460 Brandl et al. 2015 (12), Debenay et al. 2011 (1) Acanthurus dussumieri 0.775 Debenay et al. 2011 (4) Acanthurus leucopareius 0.698 Kelly et al. 2016 (2) Acanthurus lineatus 0.703 Brandl et al. 2015 (10), Choat et al. 2002 (8) Acanthurus nigricans 0.637 Choat et al. 2002 (12) Acanthurus nigricauda 0.032 Brandl et al. 2015 (4), Choat et al. 2002 (5) Acanthurus nigrofuscus 0.682 Brandl et al. 2015 (3), Kelly et al. 2016 (17) Acanthurus nigroris 0.700 Kelly et al. 2016 (2) Acanthurus olivaceus 0.227 Brandl et al. 2015 (6), Choat et al. 2002, Kelly et al. 2016 (10) Acanthurus pyroferus 0.249 Eagle & Jones 2004 (5) Acanthurus triostegus 0.603 Kelly et al. 2016 (6) Acanthurus xanthopterus 0.607 Debenay et al. 2011 (1) Centropyge bispinosa 0.611 Debenay et al. 2011 (1) Centropyge vrolikii 0.713 Eagle & Jones 2004 (5) Chlorurus microrhinos 0.145 Choat et al. 2002 (4) Chlorurus sordidus 0.114 Choat et al. 2002 (4), Debenay et al. 2011 (2) Ctenochaetus striatus 0.208 Brandl et al. 2015 (10), Choat et al. 2002 (10), Debenay et al. 2011 (4) Ctenochaetus strigosus 0.074 Kelly et al. 2016 (6) Kyphosus cinerascens 0.681 Choat et al. 2002 (8) Kyphosus vaigiensis 0.665 Choat et al. 2002 (8) Naso annulatus 0.647 Choat et al. 2002 (6) Naso brevirostris 0.604 Choat et al. 2002 (16)

39 Table 3.1. (Continued) Mean Diet Diversity Index (DDI) values for each species are listed along with sources used to generate the values. The number of individuals used in each study is noted in parentheses next to the source. Naso lituratus 0.496 Brandl et al. 2015 (3) Naso unicornis 0.558 Brandl et al. 2015 (11), Choat et al. 2002 (9), Debenay et al. 2011 (2) Scarus ghobban 0.720 Debenay et al. 2011 (4) Scarus schlegeli 0.055 Choat et al. 2002 (7) Siganus argenteus 0.701 Hoey et al. 2013 (23) Siganus punctatus 0.754 Debenay et al. 2011 (1), Hoey et al. 2013 (31) Siganus spinus 0.567 Hoey et al. 2013 (21) Zebrasoma scopas 0.676 Brandl et al. 2015 (5), Choat et al. 2002 (8) Zebrasoma veliferum 0.408 Brandl et al. 2015 (7)

40 Figures

Acanthurus nigricauda Scarus schlegeli Ctenochaetus strigosus RecFG Chlorurus sordidus Browser Chlorurus microrhinos Detritivore Ctenochaetus striatus Excavator Acanthurus olivaceus Grazer Acanthurus pyroferus Omnivore Zebrasoma veliferum Planktivore Acanthurus blochii Scraper Naso lituratus Naso unicornis Siganus spinus Acanthurus triostegus Naso brevirostris Acanthurus xanthopterus Centropyge bispinosa Acanthurus nigricans Naso annulatus Kyphosus vaigiensis Zebrasoma scopas Kyphosus cinerascens Acanthurus nigrofuscus Acanthurus leucopareius Acanthurus nigroris Siganus argenteus Acanthurus lineatus Centropyge vrolikii Scarus ghobban Siganus punctatus Acanthurus dussumieri 0.0 0.2 0.4 0.6 0.8 Species Diet Diversity Index Figure 3.1. Species diet diversity measured by the Simpson’s Diet Diversity Index plotted by functional group. Higher values indicate a more diverse diet. The same figure organized by taxonomic family is available in Appendix C.

41 Scraper& Detritivore Browser Planktivore Grazer Omnivore Excavator

100%

75%

50% Relative Abundance Relative

25%

0%

Naso lituratus Naso unicornis Naso annulatus Siganus spinus Scarus schlegeli Naso brevirostris Scarus ghobban Chlorurus sordidus Acanthurus blochii Zebrasoma scopas Siganus argenteus Centropyge vrolikiiSiganus punctatus Kyphosus vaigiensis Acanthurus nigrorisAcanthurus lineatus AcanthurusAcanthurus olivaceus pyroferus Zebrasoma veliferum Acanthurus nigricans Chlorurus microrhinosAcanthurus nigricaudaCtenochaetus striatus Kyphosus cinerascens Acanthurus triostegus Centropyge bispinosa Ctenochaetus strigosus Acanthurus nigrofuscus Acanthurus dussumieri Acanthurus leucopareius Acanthurus xanthopterus

RHODOPHYTA CHLOROPHYTA EAM ZOOPLANKTON OCHROPHYTA TURF BENTHICINVERT

Figure 3.2. The relative abundance of diet items in the species that were incorporated into the diet diversity index are shown, with color indicating diet item type (legend on bottom).

42 Stress = 0.08, k = 3

1.5

1.0

N. unicornis K. vaigiensis

N. unicornis N. brevirostris

N. annulatus N. unicornis

0.5 C. sordidus N. lituratus

A. nigricauda C. striatus S. ghobban A. olivaceus A. pyroferus S. punctatus Z. veliferum A. nigricauda S. punctatus C. striatus C. striatus NMDS2 C. sordidus S. punctatus A. olivaceus K. cinerascens 0.0 C. microrhinos S. schlegeli A. blochii C. strigosus A. olivaceus C. vrolikii Z. scopas

A. dussumieri S. spinus Z. scopas

A. leucopareius A. nigrofuscus A. nigrofuscus C. bispinosa A. lineatus −0.5 S. argenteus A. nigricans A. blochii A. nigroris A. lineatus A. xanthopterus

A. triostegus

−1.0

Stress = 0.08, k = 3 −1.0 −0.5 0.0 0.5 1.0 1.5 1.5 NMDS1

TURF

1.0 RHODOPHYTA N. unicornis K. vaigiensis 0.5 N. unicornis N. brevirostris

N. annulatus N. unicornis

0.5 C. sordidus N. lituratus

A. nigricauda C. striatus S. ghobban A. olivaceus A. pyroferus S. punctatus Z. veliferum A. nigricauda S. punctatus C. striatus C. striatus 0.0 NMDS2 C. sordidus S. punctatus A. olivaceus K. cinerascens

NMDS2 0.0 C. microrhinos S. schlegeliEAM A. blochii C. strigosus A. olivaceus C. vrolikii Z. scopas

A. dussumieri S. spinus Z. scopas

A. leucopareius A. nigrofuscus A. nigrofuscus C. bispinosa CHLOROPHYTAA. lineatus −0.5 S. argenteus A. nigricans A. blochii A. nigroris A. lineatus −0.5 A. xanthopterus ZOOPLANKTON A. triostegus OCHROPHYTA

−1.0

−1.0 −0.5 0.0 0.5 1.0 1.5 −1.0 −0.5 NMDS10.0 0.5 1.0 NMDS1 TURF Browser Excavator Omnivore Scraper Detritivore Grazer Planktivore

Figure 3.3. (a) Non-metric multidimensional scale plot using Bray-RHODOPHYTACurtis similarity matrix based on diet0.5 composition with functional groups coded by color. Ellipses show a 95% confidence interval on functional group clusters.

0.0

NMDS2 EAM

CHLOROPHYTA

−0.5 43 ZOOPLANKTON

OCHROPHYTA

−1.0 −0.5 0.0 0.5 1.0 NMDS1

Browser Excavator Omnivore Scraper Detritivore Grazer Planktivore Stress = 0.08, k = 3

1.5

1.0

N. unicornis K. vaigiensis

N. unicornis N. brevirostris

N. annulatus N. unicornis

0.5 C. sordidus N. lituratus

A. nigricauda C. striatus S. ghobban A. olivaceus A. pyroferus S. punctatus Z. veliferum A. nigricauda S. punctatus C. striatus C. striatus NMDS2 C. sordidus S. punctatus A. olivaceus K. cinerascens 0.0 C. microrhinos S. schlegeli A. blochii C. strigosus A. olivaceus C. vrolikii Z. scopas

A. dussumieri S. spinus Z. scopas

A. leucopareius A. nigrofuscus A. nigrofuscus C. bispinosa A. lineatus −0.5 S. argenteus A. nigricans A. blochii A. nigroris A. lineatus A. xanthopterus

A. triostegus

−1.0

Stress = 0.08,− k1.0 = 3 −0.5 0.0 0.5 1.0 1.5 NMDS1 1.5 TURF

RHODOPHYTA

1.00.5

N. unicornis K. vaigiensis

N. unicornis N. brevirostris

N. annulatus N. unicornis

0.5 N. lituratus 0.0 C. sordidus

NMDS2 A. nigricaudaEAM C. striatus S. ghobban A. olivaceus A. pyroferus S. punctatus Z. veliferum A. nigricauda S. punctatus C. striatus C. striatus NMDS2 C. sordidus S. punctatus A. olivaceus K. cinerascens 0.0 C. microrhinos CHLOROPHYTA S. schlegeli A. blochii C. strigosus Z. scopas −0.5 A. olivaceus C. vrolikii ZOOPLANKTON A. dussumieri S. spinus Z. scopas OCHROPHYTA A. leucopareius A. nigrofuscus A. nigrofuscus C. bispinosa A. lineatus −0.5 S. argenteus A. nigricans A. blochii A. nigroris A. lineatus −1.0 −0.5 A. xanthopterus 0.0 0.5 1.0 NMDS1

A. triostegus Browser Excavator Omnivore Scraper Detritivore Grazer Planktivore −1.0 Figure 3.3. (b) Non-metric multidimensional scale plot using Bray-Curtis similarity matrix based on diet composition with functional groups coded by color. Significant (p < 0.05) diet items are overlaid onto the NMDS−1.0 plot. −0.5 0.0 0.5 1.0 1.5 NMDS1

TURF

RHODOPHYTA

0.5

0.0

NMDS2 EAM 44

CHLOROPHYTA

−0.5 ZOOPLANKTON

OCHROPHYTA

−1.0 −0.5 0.0 0.5 1.0 NMDS1

Browser Excavator Omnivore Scraper Detritivore Grazer Planktivore Siganus punctatus

Acanthurus dussumieri

Zebrasoma scopas Omnivore Siganus argenteus

Kyphosus vaigiensis Grazer Kyphosus cinerascens Browser Naso lituratus Detritivore Acanthurus pyroferus Planktivore Acanthurus olivaceus Scraper Ctenochaetus striatus

Acanthurus nigricauda Excavator

Acanthurus blochii

Ctenochaetus strigosus

Naso unicornis

Acanthurus xanthopterus

Naso brevirostris

Naso annulatus

Acanthurus triostegus

Acanthurus leucopareius

Acanthurus nigroris

Acanthurus nigrofuscus

Acanthurus lineatus

Zebrasoma veliferum

Acanthurus nigricans

Siganus spinus

Scarus ghobban

Chlorurus sordidus

Chlorurus microrhinos

Scarus schlegeli

0.00 0.25 0.50 0.75 1.00

Figure 3.4. Dendrogram of relationships between species and functional groups based on Gower’s Dissimilarity Matrix values from trait matrices that included diet diversity ranking. Colors represent functional groups.

45 Detritivore Browser Grazer

100%

75%

50% Mean Relative Read Abundance Mean Relative 25%

0%

Naso lituratus Naso unicornis Acanthurus nigroris Acanthurus guttatus Acanthurus olivaceus Acanthurus triostegusZebrasoma flavescens Ctenochaetus strigosus Acanthurus nigrofuscusAcanthurus leucopareius

Rhodophyta Ochrophyta Chlorophyta Turf EAM

Figure 3.5. Mean diet composition as determined by metabarcoding analysis grouped by functional group and in order of increasing diet diversity within functional groups based on Simpson’s Diversity Index.

46 CHAPTER 4 FUNCTIONAL HOMOGENIZATION IN HERBIVOROUS CORAL REEF FISH ASSEMBLAGES IN RESPONSE TO ANTHROPOGENIC IMPACTS

Abstract Habitat and biodiversity loss resulting from human impacts can lead to functional homogenization, whereby natural communities become less diverse. Herbivorous coral reef fishes display different degrees of specialization and unique functional roles, even among closely related taxa, which makes homogenization particularly relevant to ecosystem function. To explore the relationship between humans and functional homogenization in herbivores, we combined an index of diet specialization and a trait matrix that included functional traits (e.g., size, behavior, depth range) with fish abundance data for >3000 sites across the Pacific that encompass a spectrum of intact to degraded coral reefs. This was used to generate a measure of the relative abundance of specialized species and to determine the functional dissimilarity in an assemblage. Hierarchical models were integrated into a piecewise structural equation model to examine the indirect and direct effects of a suite of environmental and anthropogenic drivers on herbivore assemblages. Herbivore assemblage composition varied between islands, regions, and human population densities, and generalist species such as Acanthurus nigrofuscus were dominant in highly populated locations. Functional dissimilarity decreased with human impacts, but spatial patterns and variability in the relationship indicated the presence of local and indirect effects as well. Humans had a strong direct effect on variables such as complexity, benthic cover, and predator biomass which in turn had strong influences on functional dissimilarity and the relative abundance of herbivores. Cascading indirect effects can alter the composition of herbivorous reef fish assemblages and lead to functional homogenization.

Introduction Humans fundamentally alter ecosystems, and these impacts are generally assessed through the analysis of habitat loss, diminished biomass, or changing community composition166– 168. These factors can be used to evaluate community function following an acute or prolonged disturbance. Though traditional measures of ecosystem integrity (e.g., species richness, diversity,

47 or biomass) are important, they may not capture the functional changes in an ecosystem in response to human stressors169. The multiplicity of pressures on natural systems requires new ways of examining functional changes in ecological communities. Several terrestrial studies have employed community specialization indices to examine biotic homogenization (i.e., the increasing abundance of generalists versus specialists) in the context of habitat degradation25–27. This has proven to be an effective way of examining homogenization on large spatial scales, but to our knowledge, a similar approach has not yet been used in marine ecosystems. In this study, we adopted this assemblage specialization approach, in conjunction with more common approaches including species diversity and functional dissimilarity, to examine biotic homogenization in herbivorous fish assemblages throughout the Pacific. In the past few years, coral reefs have experienced unprecedented losses in structural complexity and benthic diversity as a result of repeated mass bleaching events7,170. For some fishes, such as obligate corallivores or coral-associated fishes, the consequences of bleaching and a reduction in coral cover on fishes are obvious171–173. But even organisms that depend indirectly on the reef structure, such as fish that use reef crevices for shelter while foraging, can be profoundly affected by loss of complexity174. Herbivorous reef fishes have received attention in recent years because of their role in mediating competition between corals and algae16, and several studies have highlighted the complementarity that exists among these fishes (Chapter 3). While some metrics of community health, such as species richness, may not vary following disturbance events, more nuanced contributors to resilience such as community composition and the distribution of individuals fulfilling certain functional roles may be altered175,176. “Response diversity”, whereby functionally redundant species have unique reactions to disturbance events, may be a necessary component of resilience24. While certain characteristics, such as diet or size, are often used to represent function, the working definition of an ecological function has been evolving in recent years and in many ways reflects an organism’s resource use24,107,116,177,178. Even among fish that are closely related, the functional roles filled by each species can vary quite dramatically (Chapters 2-3). This can in turn have profound effects on their response to climate stress and habitat degradation. A generalist species that has the capacity to use a variety of resources or live in a degraded, less complex habitat may

48 be less affected by stressors, such as sedimentation or bleaching events, than a more specialized species. In the case of herbivorous reef fishes, certain species, such as Acanthurus nigrofuscus, stand out as more likely to thrive in a variety of conditions which can include degraded reefs because of its very broad diet (Chapter 2). In contrast, Ctenochaetus striatus feeds on detritus, but the deposition of fine sediment on the reef, as can occur with excessive terrestrial runoff, actually inhibits its specialized ability to feed on this resource179. This species transports much of its ingested sediment off the reef into deeper water, so reduced consumption can in turn have negative consequences for sediment removal on reefs127,180. This feeding selectivity has also been documented in other species that feed on the epilithic algal matrix, such as Scarus rivulatus, despite the fact that the two species have distinct feeding strategies181. To test for biotic homogenization in reef fish communities, we integrated an index of diet diversity in herbivores (Chapter 3) with abundance estimates based on visual surveys, to generate a measure of the relative abundance of diet specialists at a site. To understand the direct and indirect effects on functional homogenization in herbivores, this measure was then included as a response variable in a model with environmental and human drivers. Simpson’s Diversity, an established metric, and functional dissimilarity, a more comprehensive metric, were also used as response variables in an additional set of models, using the same environmental and human drivers. These models were incorporated into a piecewise structural equation model. Our aim was to determine (1) the relationship between the functional homogenization of herbivore assemblages and a suite of human impacts, (2) how herbivore assemblage composition reflects these changes, and (3) which environmental and human drivers have the greatest effect on herbivore assemblage specialization, diversity, and functional dissimilarity.

Methods I) Quantifying Herbivore Assemblage Homogenization A. Relative Abundance of Diet Specialists in Herbivore Assemblage To create a site-level diet specialization index for the assemblage (Community Specialization Index in Devictor et al. 2008), the species-level diet diversity index developed in Chapter 3 was used as a base. Diet diversity estimates were added for some species based on the

49 known values of ecologically similar species for which diet data was available, which ensured that all islands had at least 80% of their total herbivore assemblage represented in the index (Figure S4.1). A total of 79 herbivorous reef fish species were included from the Hawaiian Islands (Main Hawaiin Islands and the Papahānaumokuākea Marine National Monument), the Marianas Islands (Northern and Southern), the Pacific Remote Islands Marine National Monument (also known as the Pacific Remote Island Areas, PRIAs), and Samoa. Data from ~4000 stationary point count fish surveys from U.S. islands and atolls in the western central Pacific were used to represent the relative species abundance of herbivores93,95,149,182. If a site did not have any species with diet diversity values, then it was removed from the analysis (<2% of the sites). A vector of diet specialization values for every species was multiplied by the species’ relative abundance at each site and summed to get a single measure of the relative abundance of diet specialists for that site. Higher values indicate a greater proportion of species with diverse diets, so we used the inverse to indicate higher specialization. The relative abundances of influential functional groups and species, such as the generalist Acanthurus nigrofuscus, were also examined independently as response variables to investigate how certain taxa may serve as indicator species.

B. Metrics of Functional Homogenization The following three metrics were used in this analysis to test for functional homogenization in herbivore assemblages: (1) The relative abundance of diet specialists, determined as described above, was used to examine homogenization through the lens of diet exclusively. Diet is a foundational component of an animal’s ecosystem function, so diet diversity offers great insight into their degree of specialization. This metric can then be used to test the hypothesis that sites experiencing greater human impacts will have fewer specialists. (2) The species diversity of herbivores present in an assemblage was also used as a metric of functional homogenization, and this enabled us to test the hypothesis that humans are having a direct effect on herbivore diversity. The diversity of herbivores present, as measured by the Simpson’s Diversity Index using the Vegan package in R, was determined for each site, which

50 allowed for direct comparisons to site and island level measures of human impacts, such as human population or the benthic composition95. (3) The functional dissimilarity of herbivores present in an assemblage was used as the most comprehensive metric of homogenization, where less dissimilarity among herbivores indicated greater homogenization. A matrix of functional traits for each species was generated using data collected in Chapter 3 and from FishBase86. The traits included were the species’ diet diversity, functional group, taxonomic family, main food types, territoriality, schooling behavior, size class, depth range, and geographic distribution. These were combined with the species abundance data to generate a value of Rao’s Quadratic Entropy at each site, which uses the Gower’s Dissimilarity Index combined with abundance data to determine the assemblage functional dissimilarity, using the SYNCSA package in R183–185. Data and code are available on GitHub (in enalley/Dissertation).

C. Species Composition of Herbivore Assemblages In addition to the metrics to quantify functional homogenization in herbivore assemblages (i.e., the relative abundance of diet specialists, herbivore species diversity, and functional dissimilarity), the species composition of these assemblages was also examined (Figures S4.2, S4.3a-S4.3f). The species that were most influential in distinguishing between clusters were determined using a nonmetric multidimensional scaling (MDS) analysis with a Bray-Curtis dissimilarity index using the Vegan package in R95. Environmental and ecological variables that could be significant in determining differences in herbivore assemblage composition between regions were also examined. The relative abundances of influential functional groups and species, such the generalist Acanthurus nigrofuscus, were examined independently as response variables to investigate how certain taxa may serve as indicator species. The National Center for Ecological Analysis and Synthesis’ (NCEAS) Human Impact Score, developed using 17 different anthropogenically derived stressors that affect the condition of marine ecosystems, was used as one composite measure of human impacts186,187. This value was available at the island scale for 38 islands. Data on human populations within 20 km were available at a site scale for 41 islands. Linear mixed- effects models were fitted using the lmer function in the lme4 package in R to test the

51 relationship between herbivore abundances and the two measures, with region and island included as random effects to account for the nested spatial structure of the data102.

II. Assessing Drivers of Homogenization Linear mixed-effects models were used to test the hypothesis that the presence of humans (measured by the Human Impact Score and human population) would be associated with lower herbivore diet specialization and increased functional homogenization (measured by relative abundance of diet specialists, species diversity, and functional dissimilarity) (Table 4.1). Linear mixed-effects models were also used to test the hypothesis that human influences would outweigh inherent environmental differences between sites, using environmental covariates previously shown to be important drivers in coral reef fish assemblages (Table S4.1)102. In all mixed effects models, region and island were included as random effects to account for the nested spatial structure of the data. All predictors were scaled and centered on a normal scale to facilitate direct comparison of their effect sizes. Variables were examined for collinearity using Pearson’s correlation, and variables with r > 0.70 were not included in the model together (Figure S4.4).

III. Understanding Indirect Effects Using a Structural Equation Model Clear direct effects of humans on the functional homogenization metrics were not evident, so the models were integrated into a piecewise structural equation model, also known as a confirmatory path analysis, to assess the interconnected and indirect effects of the variables upon one another. The piecewiseSEM package in R integrates different models and model types to facilitate hypothesis testing causal links between variables188. Five hierarchical mixed effects models examining the effects of environmental and anthropogenic drivers on predator biomass, herbivore abundance, benthic complexity and cover, and herbivore functional dissimilarity were integrated into one piecewise structural equation model (GitHub: enalley/Dissertation/SupplementalFile4.1). Region was included as a random effect in each model. Linear mixed effects models were assessed independently using marginal and conditional R2 values189, as well as effects plots, residual plots, and quantile-quantile plots. The most parsimonious piecewise structural equation model that met the assumptions of directed

52 separation with covariances all under 0.7 was chosen, and the overall model goodness of fit was validated (P > 0.05).

Results I. Quantifying Herbivore Assemblage Homogenization Herbivore assemblages differed between islands and between regions (Figures 4.1, 4.2, S4.2). Differences among islands were particularly evident in the Hawaiian Islands (Figure 4.1). Permutation tests indicated significant differences in the herbivore assemblage diet specialization index between all regional pairs (all P < 0.05), except between the PRIAs and the Southern Marianas (Figure 4.3a). There were also significant differences between regions in herbivore diversity (Figure 4.3b); all regions were significantly different from each other (all P < 0.05), except the PRIAs (mean = 0.73 +/- 0.12), which were not significantly different from the Papahānaumokuākea Marine National Monument (Northwestern Hawaiian Islands, PMNM) (mean = 0.72 +/- 0.18) or from Samoa (0.75 +/- 0.11). Functional dissimilarity was also lower in the Main Hawaiian Islands as compared to all other regions (mean = 0.47 +/- 0.17, all P < 0.05) (Figure S4.5). Herbivore assemblage composition also differed between locations (Figure 4.4). The Hawaiian Islands, which have the largest gradient of human population within a 20 km radius (ranging from 0 to 723,248 people), exhibited clear differences in community composition (Figure 4.1; other regions in Figure S4.2). Generalists like Acanthurus nigrofuscus dominated islands with greater habitat degradation and less herbivore diversity like Oʻahu, Kauaʻi, Maui, and Guam (Figures S4.3, S4.6, & S4.7). The relative abundance of grazers had a significant negative relationship with the Human Impact Score (P < 0.0001, Estimate = -0.13, R2 marginal = 0.02, R2 conditional = 0.12), but when examined independently Acanthurus nigrofuscus had a significant positive relationship (P < 0.0001, Estimate = 0.29, R2 marginal = 0.09, R2 conditional = 0.23) (Figure S4.7).

II. Assessing Drivers of Homogenization The species composition of herbivore assemblages varied by region (Figure 4.2). The relative abundance of diet specialists, which increased with herbivore species diversity (P <

53 0.001, Estimate = 0.34, R2 marginal = 0.11, R2 conditional = 0.30), also varied by region (Figure 4.3), but it did not have a significant relationship with the Human Impact Score (P = 0.57). Human settlement on islands is not random and in some cases has implications about the site characteristics (e.g., reef area, high island, protected back reefs, abundant fish), and in other cases, the presence of humans can influence expected biophysical processes190. When all islands were included (i.e., populated and unpopulated), there was no significant relationship between the Human Impact Score and herbivore diversity or functional dissimilarity (P = 0.11 and P = 0.22 respectively). When only populated sites were examined, however, both homogenization metrics had a significant negative relationship with the Human Impact Score (P = 0.01, Estimate =-0.53, R2 marginal = 0.31, R2 conditional = 0.64 and P = 0.03, Estimate = -0.49, R2 marginal = 0.24, R2 conditional = 0.63 respectively) (Figure 4.5). A Bray-Curtis dissimilarity matrix was used with non-metric multidimensional scaling plots where the human population within 20 km of a site was binned into groups designated as unpopulated (no humans), minimal inhabitants (less than 100 humans), populated (more than 100 and less than 10,000 humans), and densely populated (more than 10,000 humans). Herbivore assemblages at unpopulated islands had much greater dissimilarity than those at densely populated or populated islands (Figure 4.4). The influence of specific variables and species on the differences seen in the dissimilarity matrix were also examined (Figures S4.8, S4.9). Total fish biomass, hard coral cover, crustose coralline algae (CCA) cover, functional dissimilarity, and herbivore diversity all were higher in the unpopulated sites (P = 0.001, 0.005, 0.003, 0.002, 0.002 respectively). Among surgeonfishes, the generalist Acanthurus nigrofuscus, A. leucopareius, and A. dussumieri were more abundant at populated sites (P = 0.001, 0.003, and 0.001 respectively), and the more specialized A. triostegus was associated with unpopulated sites (P = 0.002). Scarus frenetus, S. tricolor, and Zebrasoma scopas, which have been linked to high coral cover in other studies191, all had significantly higher abundance in unpopulated sites (P = 0.001, 0.001, 0.010 respectively).

III. Understanding Indirect Effects on Functional Homogenization Indirect relationships between anthropogenic, ecological, and biophysical variables undoubtedly influence the prevalence of functional homogenization (Figure 4.6). Based on the

54 structural equation model results, benthic complexity and predator biomass both had a direct, significant positive effect on herbivore functional dissimilarity (pSEM: P < 0.001, Standard Estimate = 0.11 and P < 0.001, Standard Estimate = 0.28 respectively), but the percent cover of macroalgae and EAM had smaller but significant negative effects (P < 0.001, Standard Estimate = -0.08 and P < 0.001, Standard Estimate = -0.09 respectively) (Table 4.2). Though the Human Impact Score did not have a significant direct effect on herbivore functional dissimilarity, it did have a significant negative relationship with benthic complexity (P < 0.001, Standard Estimate = -0.12), predator biomass (P < 0.001, Standard Estimate = -0.36), and herbivore abundance (P < 0.001, Standard Estimate = -0.09). Benthic complexity and predator biomass had significant positive relationships with herbivore functional dissimilarity, indicating an indirect effect on the Human Impact Score and functional dissimilarity. The Human Impact Score also had a significant positive relationship with EAM percent cover (P < 0.001, Standard Estimate = 0.20), which had a significant negative relationship with herbivore functional dissimilarity. Macroalgae, which is often associated with herbivores, had a significant negative relationship with complexity (P < 0.001, Standard Estimate = -0.25), which creates an indirect negative effect on herbivore abundance and functional dissimilarity.

Discussion Coral reef ecosystems have been severely modified by human activities, threatening the long-term persistence of specialized organisms with particular habitat or diet requirements. This can result in functional homogenization and assemblages dominated by generalists that thrive in degraded habitats – a situation that has been well described in terrestrial26, freshwater192, and other marine assemblages165. This study examined the relationship between human stressors and functional homogenization in herbivore assemblages throughout the Pacific. Our findings indicate greater functional dissimilarity in sparsely populated locations. In light of the spatial variability and complexity of these systems, it is essential that these potential impacts on ecosystem function are considered alongside both the natural drivers of variation that exist in herbivorous fish community composition, as well as the indirect effects of human activities mediated through the habitat.

55 Herbivore functional dissimilarity decreased with increasing human impacts, which is indicative of biotic homogenization (Figure 4.5)193,194, and less degraded reefs had a greater relative abundance of specialists (Figure 4.3a). Conversely, islands like Oʻahu with dense human populations had a greater proportion of herbivores with more generalized diets (Figures 4.1). For example, the generalist A. nigrofuscus was the most abundant species on Oʻahu, and though it was present at less populated islands, it was far less abundant. The significant positive relationship between the relative abundance of diet specialists and the Simpson’s Diversity Index of herbivores further suggests that biodiversity loss contributes to functional homogenization (Table 4.1). We also assessed shifting species composition in herbivore assemblages in sites that vary greatly in human population. There were clear differences in the composition of herbivore assemblages between regions (Figure 4.2) and between islands within regions (Figure 4.1). For example, in the Main Hawaiian Islands, which are the most densely populated islands in this analysis, Acanthurus nigrofuscus, a known generalist, dominated more sites than any other species (Figure S4.3a). In the virtually uninhabited Papahānaumokuākea Marine National Monument, which is the closest geographically, not a single island exhibited this trend (Figure S4.3b). Guam, Saipan, and Rota in the Southern Marianas all had a much higher frequency of sites dominated by A. nigrofuscus than by any other species, but Guam did also have a number sites where more specialized detritivores (e.g.,Ctenochaetus striatus) and parrotfishes were the dominant species (Figure S4.3f). Spatial differences can be further complicated by varied land use and management in each location. For example, highly populated locations may have greater diversity and specialization if there are protected areas where fishing and other direct, local stressors are restricted. Conversely, less densely populated islands may still be heavily influenced by land based human activities that were not explicitly accounted for in these models, such as industrial farming (e.g., Kauaʻi in Figure 4.5) or intensely developed tourist areas (e.g., Maui in Figure 4.5). There is also historic and current military infrastructure on many of the uninhabited or sparsely inhabited islands (e.g., Wake and Johnston) in the Pacific, but this type of use and the resultant changes to community composition can be difficult to characterize and quantify195–197.

56 Spatial variation seen at an island and regional scale suggests local drivers influence functional dissimilarity, diversity, specialization 198. On an ocean-wide scale human population density can be a predictor of reef fish biomass199, but human populations are, in the absence of more specific data, only a proxy for more direct stressors such as fishing200. However, not all herbivores are targeted, and the effectiveness of fishing intensity as a predictor of declines in herbivore biomass varies by functional group and species201–203. Herbivore biomass may also vary with the local environment204, so it is also important to account for spatial variation in environmental conditions in any analyses. Incorporating multiple mixed effects models into a piecewise structural equation model allowed for the known spatial variability and the unknown indirect relationships to be included in our assessment of functional homogenization in herbivorous reef fishes in response to human impacts. Degraded Pacific reefs are often covered by turf algae and sediment18, and while many herbivores consume turf, degraded turf covered reefs do not necessarily offer greater resource opportunities164. Similarly, reefs with high rates of sedimentation are not preferred by species that feed off of detritus and the epilithic algal matrix, such as detritivores, scrapers, and excavators105. On reefs that have become overgrown with macroalgae the likelihood of algal consumption may vary widely depending on physical and chemical defenses. Though large browsers consume macroalgae158,205,206, habitats dominated by macroalgae do not host high herbivore diversity or abundance207–210. Different functional groups203, and species within functional groups (Chapter 3) respond uniquely to anthropogenic impacts (Figures S4.6, S4.7). Response diversity may thus be an important additional metric of resilience107,129. Finally, we determined that predator biomass, benthic cover, and complexity had the greatest direct effects on herbivore assemblage functional dissimilarity in the locations studied, while human impacts have a series of indirect effects (Figure 4.6; Table 4.2). Though some variables showed a different response when only populated islands were included, the range of human stressors extends far beyond fishing pressure and is thus well represented in the composite Human Impact Score developed by NCEAS. Decoupling between biophysical site characteristics and benthic cover in populated sites has been shown for many of these islands190. Benthic characteristics (i.e., complexity and benthic cover) showed significant relationships with homogenization metrics, suggesting that they are robust predictors of herbivore assemblage

57 specialization on coral reefs. Habitat and species richness are also linked to genetic diversity at the seascape scale, suggesting that these drivers have a profound influence not only on specialization but also on diversity at the most fundamental level211. Other biophysical variables that did not have a direct effect on homogenization metrics, such as wave energy, did have direct negative effects on hard coral cover, which in turn has negative indirect effects on herbivore functional dissimilarity (Figure 4.6).

Conclusions Humans are undoubtedly altering coral reef habitats in what some have called a “flattening”, or the loss of reef building corals and the structural complexity that they provide212. Our study demonstrates that there is regional variability in the relationship between human population and the relative abundance of herbivores with specialized diets. Functional homogenization is also more visible in some regions, such as the Hawaiian Islands, which may be attributable to a greater range of human impacts. Communities that are dominated by just a few generalists exist without the functions performed by specialized species, which can in turn exacerbate degradation via feedback loops171,213. Other local processes also affect the composition of coral reef communities, such as land use, runoff, and pollution. As coral reefs continue to deteriorate at an exceptional rate7, it is critical that we take stock of how these stressors may be interrelated and cascading, and this analysis underscores the importance of maintaining biomass, diversity, and intact benthic structure for the perseverance of specialized organisms.

Acknowledgements We thank the Ecosystem Sciences Division at NOAA for their hard work collecting abundance data and for sharing the data with us. We thank the members of the Donahue and ToBo Labs at the Hawaiʻi Institute of Marine Biology, as well as Brian Bowen, Mark Hixon, Amber Wright, and Stephen Karl for their insight and feedback throughout the project.

58 Tables Table 4.1. Linear Mixed Effects Model results for models testing the direct effects of human population and human impacts on metrics of functional homogenization in herbivores. Random Effects & Estimate Response Fixed Effect Significance Residual (Effect Size) Variance Explained Herbivore Human P < 0.0001 Assemblage Diet Population within 0.45 Region: 16% t = 5.16 Specialization 20 km Herbivore Human Impact P = 0.57 Assemblage Diet Score t = 0.57 Specialization Herbivore Herbivore P < 0.0001 Region: 9% Assemblage Diet 0.34 Species Diversity t = 22.55 Island: 12% Specialization Human Herbivore P = 0.07 Population within Species Diversity t = -1.83 20 km Herbivore Human Impact P = 0.11

Species Diversity Score t = -1.65 Herbivore Human P = 0.45 Functional Population within t = -0.76 Dissimilarity 20 km Herbivore Human Impact P = 0.22 Functional Score t = -1.26 Dissimilarity

Human Grazer Relative P < 0.01 Region: 14% Population within -0.23 Abundance t = -3.02 Island: 10% 20 km Grazer Relative Human Impact P < 0.0001 -0.13 Region: 10% Abundance Score t = -4.50 Acanthurus Human nigrofuscus P = 0.04 Region: 24% Population within 0.13 Relative t = 2.12 Island: 5% 20 km Abundance

59 Table 4.1. (Continued) Linear Mixed Effects Model results for models testing the direct effects of human population and human impacts on metrics of functional homogenization in herbivores.

Acanthurus nigrofuscus Human Impact P < 0.0001 0.29 Region: 16% Relative Score t = 10.80 Abundance

Detritivore Human P = 0.76 Relative Population within t = -0.31 Abundance 20 km Detritivore Human Impact P < 0.0001 Relative -0.18 Region: 17% Score t = -6.05 Abundance

Human Scraper Relative P < 0.0001 Region: 24% Population within 0.63 Abundance t = 7.40 Island: 23% 20 km Scraper Relative Human Impact P < 0.0001 0.36 Region: 17% Abundance Score t = 14.11

Excavator Human Region < P = 0.002 Relative Population within -0.20 0.001% t = -3.25 Abundance 20 km Island: 18% Excavator Human Impact P = 0.09 Relative Score t = -1.68 Abundance

Human Browser Relative P = 0.30 Population within Abundance t = -1.05 20 km Browser Relative Human Impact P < 0.0001 0.19 Region: 26% Abundance Score t = 6.80

Omnivore Human P = 0.09 Relative Population within t = -1.69 Abundance 20 km Omnivore Human Impact P = 0.09 Relative Score t = -1.75 Abundance

60 Table 4.2. Piecewise Structural Equation Model significant results grouped by response and in order of decreasing standard estimate absolute value. Std. Critical P Std. Response Predictor Estimate Error Value Value Estimate Functional Predator Dissimilarity Biomass 0.081 0.006 13.820 0.000 0.284 Functional Dissimilarity Complexity 0.048 0.007 6.759 0.000 0.111 Epilithic Functional algal Dissimilarity matrix -0.001 0.000 -4.525 0.000 -0.092 Functional Dissimilarity Macroalgae -0.001 0.000 -4.342 0.000 -0.080 Human Functional Impact Dissimilarity Score 0.002 0.001 1.187 0.236 0.034 Functional Herbivore Dissimilarity abundance 0.041 0.047 0.884 0.377 0.015 Complexity Macroalgae -0.007 0.001 -13.319 0.000 -0.248 Epilithic algal Complexity matrix -0.003 0.000 -11.834 0.000 -0.248 Human Impact Complexity Score -0.013 0.003 -4.199 0.000 -0.124 Human Predator Impact Biomass Score -0.058 0.004 -15.088 0.000 -0.357 Predator Biomass Complexity 0.152 0.020 7.624 0.000 0.101 Epilithic Predator algal Biomass matrix -0.001 0.000 -3.505 0.001 -0.054 Herbivore abundance Macroalgae -0.001 0.000 -15.010 0.000 -0.282 Epilithic Herbivore algal abundance matrix 0.000 0.000 -8.903 0.000 -0.187 Herbivore abundance Complexity 0.019 0.003 7.198 0.000 0.123

61 Table 4.2. (Continued) Piecewise Structural Equation Model significant results grouped by response and in order of decreasing standard estimate absolute value.

Herbivore Predator abundance Biomass 0.012 0.002 5.490 0.000 0.118 Human Herbivore Impact abundance Score -0.002 0.001 -3.122 0.002 -0.093 Epilithic algal matrix Macroalgae -0.685 0.029 -23.795 0.000 -0.338 Human Epilithic Impact algal matrix Score 1.555 0.196 7.930 0.000 0.195

62 Main HawaiianMHI Islands Papahānaumokuākea MarineNWHI National Monument

100%

75%

Figures

Main HawaiianMHI Islands Papahānaumokuākea MarineNWHI National Monument 50% 100% Abundance

75%

Relative Abundance (> 5%) 25%

50% Abundance

Relative Abundance (> 5%) 25%

0%

0% Maui Kure Oahu Kauai Niihau Hawaii Lanai Nihoa Maro Molokai Laysan Gardner Midway Necker Lisianski Maui Kure Oahu Kauai NiihauKahoolaweHawaii Lanai Nihoa Maro Molokai Laysan Gardner Midway Necker Lisianski FrenchKahoolawe Frigate French Frigate Pearl & Hermes Pearl & Hermes

A. blochii A. olivaceus C. sordidus N. lituratus A. blochii A. leucopareiusA. olivaceusA. triostegus KyphosusC. sordidus spp. N. unicornis N. lituratus A. nigrofuscus C. potteri C. strigosus S. dubius A. leucopareius A. nigrorisA. triostegusC. perspicillatus N.Kyphosus brevirostris spp.Z. flavescens N. unicornis A. nigrofuscus C. potteri C. strigosus S. dubius A. nigroris C. perspicillatus N. brevirostris Z. flavescens

Figure 4.1. Herbivore assemblage composition in the Hawaiian Islands with herbivore diversity increasing from left to right (i.e., most diverse assemblages as measured by Simpson’s Diversity Index on the right). Only species that made up at least 5% of the relative abundance of herbivores present at an island were included.

63 Stress = 0.16

Stress = 0.16

Swains 1 Midway

Gardner Kure Baker Howland Wake Pearl & Hermes Lisianski Swains Jarvis 1 Midway Kingman Maro Laysan Palmyra Region Gardner Farallon de Pajaros French Frigate Kure a MHI 0 Maug Nihoa Necker Baker Asuncion a NWHI Howland Johnston Rose Niihau Molokai a PRIAs Wake Lisianski Agrihan Pearl & Hermes NMDS2 Jarvis Tau Alam−Gug−Sarig Oahu a SAMOA Ofu & Olosega Pagan Maui a S.MARIAN Tutuila Kingman Maro Aguijan Rota Kauai Lanai Kahoolawe a N.MARIAN Saipan Laysan Tinian Guam Hawaii Palmyra Region Farallon de Pajaros French Frigate −1 a MainMHI Hawaiian Islands 0 Maug Nihoa Necker Asuncion a PapahānaumokuākeaNWHI Marine National Monument Johnston Rose Niihau Molokai a PRIAs Agrihan Pacific Remote Island Areas NMDS2 Tau Alam−Gug−Sarig Oahu a SamoaSAMOA Ofu & Olosega Pagan Maui a SouthernS.MARIANMarianas Tutuila Aguijan Rota Kauai Lanai Kahoolawe a NorthernN.MARIANMarianas Saipan −3 −2 −1 0 1 Tinian Guam Hawaii NMDS1

−1 Figure 4.2. NMDS (stress = 0.16) of herbivore assemblage composition at the island scale across regions in the Pacific. Ellipses show 95% confidence intervals.

−3 −2 −1 0 1 NMDS1

64 MHI PMNM PRIMNM Samoa S. Marianas N. Marianas A

0.8

REGION.x MHI 0.6 NWHI PRIAs SAMOA S.MARIAN N.MARIAN

Herbivore Assemblage Diet Specialization Assemblage Herbivore 0.4

0.2

Tau LanaiMaui Oahu Kure Maro Rose Rota Sarig Hawaii Kauai Niihau Nihoa Baker Jarvis Wake Tutuila Guam Tinian − MaugPagan Molokai Laysan MidwayNecker Palmyra Swains Aguijan Saipan Agrihan Gardner Lisianski Howland JohnstonKingman Asuncion Gug Kahoolawe − French Frigate Pearl & Hermes Ofu & Olosega Alam Farallon de Pajaros

Figure 4.3. (a) Herbivore assemblage diet specialization by island, grouped by region. Dark blue lines represent the mean value for the region.

65 MHI PMNM PRIAs Samoa S. Marianas N. Marianas B

0.75

0.50

Herbivore Species Diversity (Simpson's) Species Diversity Herbivore 0.25

0.00

Tau Lanai Maui Oahu Kure Maro Rose Rota Sarig Hawaii Kauai Niihau Nihoa Baker Jarvis Wake Tutuila Guam Tinian − MaugPagan Molokai Laysan MidwayNecker Palmyra Swains Aguijan Saipan Agrihan Gardner Lisianski Howland JohnstonKingman Asuncion Gug Kahoolawe South Bank − French Frigate Pearl & Hermes Ofu & Olosega Alam Farallon de Pajaros

Figure 4.3. (b) Herbivore species diversity (measured by Simpson’s Diversity Index) by island, grouped by region. Dark blue lines represent the mean value for the region.

66 Stress = 0.14

Johnston

1.0 Jarvis Midway Baker Howland Gardner Kure Swains 0.5 Lisianski Wake Human Population Kingman Palmyra Maro Pearl & Hermes within 20 km Laysan a unpopulated a minimal inhabitants

NMDS2 Rose French Frigate 0.0 Necker a populated Maug Farallon de Pajaros Ofu & Olosega Niihau a densely populated Agrihan Asuncion Nihoa Stress = 0.14 Tau Molokai Tutuila Alam−Gug−Sarig Lanai Maui −0.5 Aguijan Pagan Rota Oahu Kahoolawe Tinian Kauai Johnston Saipan Guam Hawaii

1.0 −1.0 Jarvis Midway −1 0 1 2 Baker Howland Gardner NMDS1 Kure Swains 0.5 Lisianski Wake Human Population Kingman Palmyra Maro Pearl & Hermes within 20 km Laysan a unpopulated a minimal inhabitants

NMDS2 Rose French Frigate 0.0 Necker a populated Maug Farallon de Pajaros Ofu & Olosega Niihau a densely populated Agrihan Asuncion Nihoa Tau Molokai Tutuila Alam−Gug−Sarig Lanai Maui −0.5 Aguijan Pagan Rota Oahu Kahoolawe Figure 4.4. Non-metric Multidimensional Scaling plot using Bray-Curtis dissimilarity between Tinian herbivore assemblages grouped by human population within 20 km. Minimally inhabited sites Saipan Guam Kauai Hawaii had less than 100 human residents within 20 km. Populated sites had more than 100 but less than 10,000 human inhabitants within 20 km, and densely populated sites had more than 10,000

−1.0 human inhabitants within 20 km. Some islands, such as Aguijan, are uninhabited but are within 20 km of densely populated islands. −1 0 1 2 NMDS1

67 Maug Wake Maug Palmyra Kingman Wake FarallonPalmyra de Pajaros Kingman Swains AsuncionFarallon deJohnston Pajaros Swains 0.6 Alam−GugAsuncion−Sarig RoseJohnston 0.6 Alam−Gug−Sarig Rose Tau size Tutuila Agrihan Tau sizea 15 Howland Tutuila Agrihan a 15 BakerHowland Ofu & Olosega Rota Baker Ofu & Olosega factor(REGION) Rota Saipan Guam factor(REGION)a MHI Jarvis Tinian Saipan Jarvis LanaiGuam a MainMHINWHI Hawaiian Islands Tinian Hawaii Aguijan MolokaiLanai a PacificNWHIPRIAsRemote Island Areas 0.5 Hawaii a Aguijan Molokai a SamoaPRIAsSAMOA 0.5 Niihau a SouthernSAMOAS.MARIANMarianas Niihau a NorthernS.MARIANN.MARIANMarianas Maug Wake a N.MARIAN Palmyra Maui

Functional Dissimilarity of All Herbivores Kingman Farallon de PajarosMauiSwains Functional Dissimilarity of All Herbivores Asuncion Johnston 0.6 Alam−Gug−Sarig Rose

Tau size 0.4 Tutuila Agrihan a 15 0.4 Howland Kauai Baker Ofu & OlosegaKauai Oahu Rota Oahu factor(REGION) Saipan Guam a MHI 4 8Jarvis 12 Tinian 16 4 NCEAS8 Human Impact Score12 16 Lanai a NWHI Hawaii NCEAS Human Impact Score Aguijan Molokai a PRIAs 0.5 a SAMOA Niihau a S.MARIAN a N.MARIAN

Maui Functional Dissimilarity of All Herbivores

0.4 Kauai Oahu

4 8 12 16 NCEAS Human Impact Score

Figure 4.5. Functional dissimilarity of herbivores compared to NCEAS Human Impact Score. The shaded area shows a 95% confidence interval. Colors correspond to regions. NCEAS Human Impact Scores were not available for the Papahānaumokuākea Marine National Monument (Northwestern Hawaiian Islands).

68 > 0.25 c = 0.38 2 R Biomass Predator > 0.10 & < 0.25 Human Impacts Standardized Effect Size: < 0.10 Effect Standardized c = 0.22 2 R Herbivore Abundance < 0.05), and the width of the line indicates the indicates the the of line width < and 0.05), P = 0.22 c 2 R Functional Dissimilarity

EAM Piecewise structural equation model results. The arrow is pointing from the response variable to variable the The structural response Piecewise results. from is equation arrow pointing the model

. c = 0.38 6 2 R Complex c = 0.17 2 Macroalgae Figure 4. predictor significant listed ( interactions Only variable. are magnitude ofestimate. standardized the R Significant Relationship Positive Significant Relationship Negative

69 CHAPTER 5 CONCLUSIONS

Diet defines an essential component of an animal’s ecological function, and this research contributes to an increased understanding of the complexity that exists within the diets of herbivorous reef fishes. Coral reefs are changing before our eyes, however, lending urgency to the task of assessing the role of diet specialization in maintaining diversity, redundancy, and resilience in coral reef communities. This work begins to answer those questions in a scalable way that lends insight into the future of herbivores on coral reefs, while building upon a body of existing research to address data gaps that exist in our primary understanding of herbivore diets.

Primary Findings: Though Acanthurus triostegus and A. nigrofuscus exhibit diet overlap, A. nigrofuscus has significantly greater diet diversity. This suggests that A. nigrofuscus is likely an opportunistic feeder, consuming a greater proportion of the EAM, while A. triostegus primarily consumes red algae. A. nigrofuscus exhibited greater individual specialization than A. triostegus, but A. triostegus showed more variability between locations. The diet differences seen in these two species may be due to behavioral differences, such as schooling behavior, which is exhibited more commonly by A. triostegus. In this study, which to our knowledge was the most comprehensive herbivore metabarcoding diet study to date, metabarcoding was successfully used as a tool to elucidate diet differences in closely related grazing surgeonfishes. Many herbivores lack quantitative diet data references, indicating that there is a great need for additional research examining fundamental components of functional roles, such as an animal’s diet. It is additionally important that data be collected in a standardized way so that comparisons may be made between studies. Variability in diet diversity exists between and within functional groups of herbivorous reef fishes. Surgeonfishes, which include detritivores, grazers, and omnivores, exhibited large range in their diet diversity, with some of the most specialized (e.g., A. nigricauda) and least specialized (e.g., A. leucopareius) species. Parrotfishes exhibited more specialized diets, with the exception of the generalist Scarus ghobban. Rabbitfishes also displayed a range of specialization which may be attributable to behavioral

70 differences among species (e.g., schooling, coordinated vigilance). This analysis offers a reproducible, straightforward way to compare diet data from multiple studies to make comparisons at large, ecologically relevant scales. It also demonstrated that metabarcoding can be used in combination with observational diet studies to offer greater taxonomic resolution. Coral reef communities are complex systems that have been highly modified in many locations. As a result, herbivore assemblages in some highly populated locations appeared to display functional homogenization, with assemblages dominated by diet generalists, such as A. nigrofuscus. The relative abundance of diet specialists also decreased with decreasing species diversity of herbivores, highlighting the importance of maintaining biodiversity for sustained ecosystem function. When additional biological and behavioral traits were considered, functional dissimilarity in herbivore assemblages also decreased with increasing human impacts. Despite these clear trends, there was considerable spatial variability, and it was also clear that indirect effects have a significant role in shaping coral reef communities. We used a piecewise structural equation model to determine that predator biomass, benthic cover, and complexity had the greatest direct influence on functional dissimilarity in herbivore assemblages, but human impacts had numerous indirect effects that are undoubtedly influencing herbivore assemblages.

Future Directions: The body of work presented in this dissertation was developed through an iterative, scalable process. Starting with a detailed examination of diet variability between individuals within two species using a molecular approach yielded the opportunity to eventually examine trends in the homogenization of herbivore assemblages across the Pacific. This approach offers an apt metaphor for how we might consider human impacts and the resultant ecosystem changes moving forward. While some stressors occur at a global scale and require comprehensive methods of management, there are many stressors that occur at a local scale and require place- specific consideration. Much as understanding diet on the scale of an individual can help to assess trends occurring across an ocean basin, addressing local environmental stressors, such as pollution and sedimentation, with effective, place-based management offers the opportunity for coastal ecosystems to build their capacity for resilience to global stressors. Through these combined efforts we can begin to anticipate and plan for the changing coral reefs of the future.

71 APPENDIX A CHAPTER 2 SUPPLEMENTAL FIGURES

Waiheʻe

Haleʻiwa Lāiʻe Mokulēiʻa Kahana

Kāneʻohe

Mākaha Kailua

Kahe Point

Kakaʻako Maunalua Ala Moana

Salt Pond

Figure S2.1. Map of sample collection locations. Left: Oʻahu. Top right: Maui. Bottom right: Kauaʻi.

72 120 100 80 60 Identified Diet Taxa 40 20 0

0 50 100 150 Individuals Sampled

Figure S2.2. Species accumulation curves generated using the Vegan package in R 93,95 for Acanthurus triostegus (grey) and A. nigrofuscus (brown) samples, with the extrapolated sampling curve for all samples combined in pink. Vertical lines represent the standard deviation of the extrapolated sampling curve.

73 Cyanobacteria Chroococcales Nostocales Oscillatoriales Oscillatoriophycideae Chromerida Chlorarachniales Chlorodendrales Bryopsidales Ulvales Dictyotales Ectocarpales Phaeophyceae Sphacelariales Ceramiales Corallinales Florideophyceae Rhodymeniophycidae Rhodophyta Diet Items Diet Items Diet Items Diet Items

Waihee (Maui)

Salt Pond (Kauai)

Maunalua

Makaha

Kaneohe

Kakaako

Kailua 0% 0% 75% 50% 25% 75% 50% 25% 100% 100% 0% 75% 50% 25% 0% 100% 75% 50% 25% 100%

Kahana

Haliewa down by diet order. composition broken

Electric Beach A. triostegus

. 3 S2. 0% 0% 0% 0%

75% 50% 25% 75% 50% 25% 75% 50% 25% 75% 50% 25%

100% 100% 100% 100%

% of Chlorophyta of % Ochrophyta of % Other of % % of Rhodophyta of %

Figure 74 Dictyotales Ectocarpales Phaeophyceae Sphacelariales Cyanobacteria Nostocales Oscillatoriales Oscillatoriophycideae Dinophyceae Chromerida Chlorarachniales Chlorodendrales Bryopsidales Ulvales Bonnemaisoniales Ceramiales Corallinales Florideophyceae Rhodymeniales Rhodymeniophycidae Rhodophyta Diet Items Diet Items Diet Items Diet Items

Olowalu (Maui)

Mokuleia

Maunalua

Laie

Kakaako

Haleiwa 0% 0% 75% 50% 25% 75% 50% 25% 100% 100% 0% 75% 50% 25% 0% 100% 75% 50% 25% 100% Electric Beach

Chuns diet composition broken diet down composition broken by order.

Ala Moana A. nigrofuscus A. nigrofuscus

. 4 Figure S2. 0% 0% 0% 0%

75% 50% 25% 75% 50% 25% 75% 50% 25% 75% 50% 25%

100% 100% 100% 100% % of Rhodophyta of % % of Other of % % of Chlorophyta of % Ochrophyta of %

75

A c abc ac ac abcd ab ab b bd d B

60 50

50 40

40 Diet Richness Diet Richness 30

30

20 Kailua KahanaHaleiwa KakaakoMakaha Laie MaunaluaKaneoheKahe Point Haleiwa Kakaako Chuns Waihee, Maui Kahe PointAla MoanaMaunalua Mokuleia Salt Pond, Kauai Olowalu, Maui

C d bd b abcd bc abcd bc abc a ac D 0.9 0.9

0.8 0.8

0.7 0.7

0.6 0.6 Diet Diversity (Simpson's) Diet Diversity 0.5 (Simpson's) Diet Diversity

0.5

Kailua Kahana Haleiwa MakahaKakaako Laie Maunalua Kaneohe Kahe Point Kakaako Haleiwa Chuns Waihee, Maui Maunalua MokuleiaAla MoanaKahe Point Salt Pond, Kauai Olowalu, Maui

Figure S2.5. Diet richness (A – A. triostegus & B – A. nigrofuscus) and diversity (C – A. triostegus & D – A. nigrofuscus) with significant differences between sites indicated with letters. There were no significant differences in richness or diversity between sites for A. nigrofuscus.

76 A. 0.10 Model:

lmer ( 0.05 Diet Overlap with other Individuals in the Population 0.00

~ 0.05 − Weight2 resid(mod.overlap_ACTR) 0.10 + Diet Richness − + Diet Diversity 0.15 − + (1 | Site)) 0.3 0.4 0.5 0.6 0.7 0.8

fitted(mod.overlap_ACTR)

Site Wt_g effect plot Richness effect plot

B. (Intercept) C. 0.75 0.80

Kakaako 0.70 0.75

0.65 0.70 KahanaBay 0.60 0.65 0.55 Makaha

(MeanIndivOverlap) (MeanIndivOverlap) 0.60 0.50 0.55 ElectricBeach 0.45

50 100 150 200 30 40 50 60 70 Haleiwa Wt_g Richness

Diversity.Simp. effect plot

Waihee 0.75

Kailua 0.70

Maunalua 0.65

0.60 SaltPond (MeanIndivOverlap) 0.55 −0.3 −0.2 −0.1 0.0 0.1 0.2

0.5 0.6 0.7 0.8 0.9 Diversity.Simp.

Figure S2.6. Linear mixed effects model diagnostic plots for model examining diet overlap among individuals of Acanthurus triostegus in the population. (A) Plot of residual versus fitted values. (B) Random effect (site) estimate plot. (C) Fixed effects estimate plots. Weight was squared in the model to account for the nonlinear distribution.

77 A. Model: lmer ( 0.10

Diet Overlap of an Individual with the 0.05 Mean of the Population ~ 0.00 0.05 − Weight2 resid(mod.PSi_ACTR)

+ Diet Richness 0.10 − + Diet Diversity 0.15 + (1 | Site)) −

0.55 0.60 0.65 0.70 0.75 0.80 0.85

fitted(mod.PSi_ACTR)

Site Wt_g effect plot Richness effect plot

B. (Intercept) C. 0.85 0.90 Kakaako 0.80

0.85 0.75

Makaha 0.80 0.70 (PSi) (PSi) 0.75 0.65 KahanaBay 0.70 0.60

Waihee 0.65 0.55 50 100 150 200 30 40 50 60 70 Wt_g Richness ElectricBeach Diversity.Simp. effect plot

Haleiwa 0.85

0.80 Kailua

0.75

Maunalua (PSi) 0.70

SaltPond 0.65

0.60 −0.2 −0.1 0.0 0.1 0.5 0.6 0.7 0.8 0.9 Diversity.Simp.

Figure S2.7. Linear mixed effects model diagnostic plots for model examining diet overlap between Acanthurus triostegus individuals and the overall population. (A) Plot of residual versus fitted values. (B) Random effect (site) estimate plot. (C) Fixed effects estimate plots. Weight was squared in the model to account for the nonlinear distribution.

78 A.

Model: 0.10 lmer ( Diet Overlap with other Individuals in 0.05

the Population 0.00 ~ 0.05 − Weight2 resid(mod.overlap_ACNF) 0.10 + Diet Richness − 0.15

+ Diet Diversity − + (1 | Site)) 0.40 0.45 0.50 0.55 0.60 0.65

fitted(mod.overlap_ACNF)

Site Wt_g effect plot Richness effect plot

(Intercept) 0.70 B. C. 0.60 0.65 Maunalua 0.60 0.55 0.55 Kakaako 0.50 0.50 0.45 (MeanIndivOverlap) (MeanIndivOverlap) Chuns 0.40 0.45 0.35 Olowalu 40 50 60 70 80 90 100 20 30 40 50 60 Wt_g Richness

Diversity.Simp. effect plot Haleiwa

0.60

AlaMoana

0.55 Laie

0.50

Mokuleia (MeanIndivOverlap)

−0.05 0.00 0.05 0.45 0.5 0.6 0.7 0.8 0.9 Diversity.Simp.

Figure S2.8. Linear mixed effects model diagnostic plots for model examining diet overlap among Acanthurus nigrofuscus individuals in the population. (A) Plot of residual versus fitted values. (B) Random effect (site) estimate plot. (C) Fixed effects estimate plots.

79 A. 0.2 Model: lmer ( 0.1 Diet Overlap of an Individual with the Mean of the Population ~ 0.0 Weight resid(mod.PSi_ACNF) 0.1 + Diet Richness − + Diet Diversity 0.2 + (1 | Site)) − 0.50 0.55 0.60 0.65 0.70 0.75

fitted(mod.PSi_ACNF)

Site Wt_g effect plot Richness effect plot (Intercept) C. B. 0.75

Chuns 0.8 0.70

0.7 Maunalua 0.65 (PSi) (PSi) 0.6

Kakaako 0.60 0.5

Olowalu 40 50 60 70 80 90 100 20 30 40 50 60 Wt_g Richness

Diversity.Simp. effect plot Haleiwa

0.70 Mokuleia

0.65

AlaMoana (PSi) 0.60

0.55 Laie

−0.04 −0.02 0.00 0.02 0.04 0.5 0.6 0.7 0.8 0.9 Diversity.Simp.

Figure S2.9. Linear mixed effects model diagnostic plots for model examining diet overlap between Acanthurus nigrofuscus individuals and the overall population. (A) Plot of residual versus fitted values. (B) Random effect (site) estimate plot. (C) Fixed effects estimate plots.

80 APPENDIX B CHAPTER 3 SUPPLEMENTAL TABLES

Table S3.1. Summary statistics for the ten herbivore species that were included in the metabarcoding dietary analysis. Std. Std. Mean Max Min. Mean Max Min. Dev. Dev. Sample Species Diet Diet Diet Diet Diet Diet Diet Diet Size Rich. Rich. Rich. Div. Div. Div. Rich. Div. Acanthurus guttatus 37 5.66 41 33 0.845 0.014 0.855 0.835 2 Acanthurus leucopareius 37.5 9.96 49 14 0.806 0.112 0.91 0.519 13 Acanthurus nigrofuscus 42.6 7.24 57 21 0.826 0.0686 0.909 0.506 73 Acanthurus nigroris 39.4 2.97 43 36 0.758 0.159 0.897 0.521 5 Acanthurus olivaceus 33 NA 33 33 0.859 NA 0.859 0.859 1 Acanthurus triostegus 42 8.52 64 11 0.797 0.0838 0.914 0.464 90 Ctenochaetus strigosus 33.5 5.92 38 25 0.801 0.0526 0.842 0.723 4 Naso lituratus 26 3.63 32 23 0.624 0.184 0.819 0.346 6 Naso unicornis 19 NA 19 19 0.553 NA 0.553 0.553 1 Zebrasoma flavescens 43.9 4.12 53 38 0.841 0.0393 0.895 0.735 20

81 APPENDIX C CHAPTER 3 SUPPLEMENTAL FIGURES

Haleʻiwa Lāʻie Mokulēʻia

Kahana

Kāneʻohe Mākaha Kailua

- Olowalu, Maui Kahe Point - South Shore Viewpoint, Maui Kakaʻako - Waiheʻe, Maui Maunalua - Salt Pond, Kauaʻi - Fish Market in Waipahu Ala Wai A. B.

Figure S3.1. Collection locations of fish used in this metabarcoding dietary study where the size of the box reflects the sample size at a given location, and the color refers to the box surrounding the species. Locations not included on the map are listed below the species. The species are (from left to right, top to bottom) Acanthurus triostegus, A. nigrofuscus, Ctenochaetus strigosus, A. nigroris, Zebrasoma flavescens, A. leucopareius, A. guttatus, Naso unicornis, A. olivaceus, and N. lituratus.

82 1.00

0.75

0.50

Species Diet Diversity Index Species Diet Diversity 0.25

0.00

Grazer Browser Scraper Omnivore Excavator Planktivore Detritivore

Figure S3.2. Species Diet Diversity Index (DDI) values are plotted by functional group with the mean value decreasing from left to right. The graph shows the median (line), first and third quartiles (hinges), the 95% confidence interval of the median (whiskers), and outliers (circles). Image Credit: Keoki Stender and obtained with permission from his website (MarineLifePhotography.com).

83 Acanthurus nigricauda Scarus schlegeli Ctenochaetus strigosus Chlorurus sordidus Family Chlorurus microrhinos Acanthuridae Ctenochaetus striatus Kyphosidae Acanthurus olivaceus Pomacanthidae Acanthurus pyroferus Scaridae Zebrasoma veliferum Siganidae Acanthurus blochii Naso lituratus Naso unicornis Siganus spinus Acanthurus triostegus Naso brevirostris Acanthurus xanthopterus Centropyge bispinosa Acanthurus nigricans Naso annulatus Kyphosus vaigiensis Zebrasoma scopas Kyphosus cinerascens Acanthurus nigrofuscus Acanthurus leucopareius Acanthurus nigroris Siganus argenteus Acanthurus lineatus Centropyge vrolikii Scarus ghobban Siganus punctatus Acanthurus dussumieri 0.0 0.2 0.4 0.6 0.8 Species Diet Diversity Index

Figure S3.3. Species diet diversity generated using Simpson’s Diversity Index based on data from existing literature grouped by taxonomic family. Larger values indicate a more diverse diet (i.e., diet diversity increases moving towards bottom of the page).

84 APPENDIX D CHAPTER 3 SUPPLEMENTAL TEXT

Text S3.1. This is a qualitative description of the diet and ecology of the species that are included in the Diet Diversity Index based on a literature review. Many of these sources were not able to be incorporated into the Diet Diversity Index because of issues with standardizing quantitative data. Comparisons to data generated in this study using metabarcoding are included below the literature review. Complete lists of diet items found in the metabarcoding study can be found in the Supplemental Materials (GitHub: enalley/Dissertation/SupplementalFile3.9-10).

Acanthurus blochii: Among the surgeonfishes that primarily feed on detritus, A. blochii has a more diverse diet119,155. A. blochii is primarily found in sandy bottomed habitats161, but it is reported to consume algal fronds and filaments, turf algae, and detritus, along with the associated diatoms, foraminifera, and invertebrates84,214.

Acanthurus dussumieri: A. dussumieri is typically found on sandy patches, like other acanthurid detritivores but has a more diverse diet than more exclusive detritivores like A. olivaceus161. Based on our assessment of existing literature, A. dussumieri has an extremely diverse diet, consumes considerable amounts of benthic invertebrates, and should likely be reclassified as an omnivore155.

Acanthurus guttatus: A. guttatus is one of the larger species of surgeonfish, and while it has a thick walled stomach, it lacks the gizzard present in some congeners84. This would suggest that the species consumes some sediment but is not an exclusive detritivore. Typically found in surge zones with a population size limited by habitat availability161, it has been reported that these fish crop filamentous algae and larger fronds but also scrape the EAM, consuming calcareous material and cyanobacteria in the process84. A generalist strategy may allow A. guttatus to eat whatever is available as it gets moved back and forth over the reef in the surge. § Metabarcoding comparison: Our metabarcoding results match these findings and indicate that A. guttatus consumes EAM (e.g., unidentified cyanobacteria, Nostocales, and Oscillatoriales) at a proportion that is on par with surgeonfishes that focus their feeding on detritus and the EAM, such as Ctenochaetus strigosus and A. olivaceus (Figures 3.5,

85 S3.2, and S3.3). Despite this increased abundance of EAM, the largest relative read abundances were from red macroalgae (e.g., Asparagopsis taxiformis, Dasya sp., Laurencia sp., and Corallinales) and from turf algae (e.g., Ceramium sp., Heterosiphonia crispella, Herposiphonia sp., Polysiphonia sp., and Tolypiocladia glomerulata), as seen in other grazing surgeonfishes. There were a smaller proportion of reads for green (Bryopsidales) and brown algae (e.g., Asteronema breviarticulatum and Sphacelaria tribuloides).

Acanthurus leucopareius: Based on the results of existing studies and our own metabarcoding analysis, A. leucopareius appears to have a diverse diet as compared to other surgeonfishes124. While the closely related generalist A. nigrofuscus is typically found in calmer water below the surge zone, A. leucopareius is more likely to be found in the surge habitat, which perhaps facilitates niche differentiation among these species103. A. leucopareius also has a very long intestinal tract, which may allow it to consume a variety of food types and obtain energy from a diverse array of less desirable foods103. This species schools and sometimes mixes with schools of A. triostegus. § Metabarcoding comparison: A. leucopareius had the highest proportion of reads from red macroalgae (e.g., Asparagopsis taxiformis, Dasya sp., Chondria sp., Laurencia sp., Galaxaura rugosa, and Corallinales) and turf algae (e.g., Ceramium sp., Herposiphonia sp., and Polysiphonia sp.) (Figure 3.5). Brown algae were also detected (e.g., Asteronema breviarticulatum, Sphacelaria tribuloides, Lobophora variegata, and Turbinaria ornata), though in smaller proportions than red macroalgae or turf, but this was not seen in studies referenced for the meta-analysis. A very small proportion of reads were detected from green algae (e.g., Bryopsidales) and from the EAM as indicated by reads from taxa such as unidentified cyanobacteria and Oscillatoriales.

Acanthurus lineatus: Though we were not able to examine A. lineatus samples for the metabarcoding analysis, historic data indicates that the species targets fronds and filaments of algae, largely avoiding detritus or EAM in the process84. The studies that were included in the diet diversity index similarly found that at least 90% of the A. lineatus diet was algal filaments

86 and fronds, though this species appears to consume a greater diversity of taxa than many other surgeonfishes, apparently eating brown, green, and red macroalgae alongside filamentous turf algae119,144. This diet diversity consequently places A. lineatus among the more generalist surgeonfishes examined.

Acanthurus nigricans: A. nigricans, which inhabits the surge zone, appears to have a diverse diet, despite having a unique mouth morphology103,144. Other studies have found that specialized morphology does not necessarily correlate with specialized diets, so A. nigricans may provide an example in support of that theory119,215,216. It has been suggested as well that the abundance of A. nigricans is limited by competitive exclusion by A. achilles in locations where both are present; when A. nigricans is abundant, A. achilles typically is not103.

Acanthurus nigricauda: As the species with the most limited diet in the references examined in this study, A. nigricauda can be described as quite specialized among herbivores. Other sources describe this species as feeding almost entirely on the EAM, as opposed to feeding on algae directly217. Unlike other surgeonfishes, which tend to be present in very high numbers, A. nigricauda does not make up much of the overall abundance of herbivores when found218. This species is also harvested in parts of its range, so when combined with low abundances and high diet specialization, this may make populations more vulnerable in the face of habitat disturbances or environmental perturbations219,220.

Acanthurus nigrofuscus: Though A. nigrofuscus is a small fish as compared to many of the other surgeonfishes, it is quite aggressive even towards larger fishes, with noticeable changes in coloration as an indicator103. This behavior may be necessary as the species does not tend to school as much as other surgeonfishes, instead feeding independently on a variety of resources119,124. § Metabarcoding comparison: A. nigrofuscus had one of the most diverse diets among grazers (Figure 3.5). This species consumes a majority of turf algae (e.g., Ceramium spp., Heterosiphonia crispella, Herposiphonia sp., and Polysiphonia sp.) and red macroalgae (e.g., Asparagopsis taxiformis, Dasya sp., Acanthophora spicifera, and Laurencia spp.),

87 which was also found in the studies used to generate the diet diversity values. A. nigrofuscus also consumes the EAM, as indicated by the presence of cyanobacteria, Oscillatoriales, Chromerida, and a variety of other taxa found in benthic organic matter. A. nigrofuscus also consumes a small portion of green algae (e.g., Ulva sp. and Bryopsidales) and brown algae (e.g., Lobophora variegata, Colpomenia sinuosa, and Sphacelaria tribuloides).

Acanthurus nigroris: A. nigroris is extremely abundant in some locations (e.g., Marshall Islands as reported by Schultz et al. 1953) and can be found in a variety of reef habitats84. Though it also focuses primarily on algal filaments, it appears that A. nigroris also consumes detritus and organic matter from both sandy and rocky areas on reefs below the surge zone103, as indicated by the thicker walled stomach to assist with mechanical digestion84. § Metabarcoding comparison: Unexpectedly, this species consumed more brown algae (e.g., Sphacelaria tribuloides and Padina gymnospora), than the other grazing surgeonfishes (Figure 3.5). Though approximately half of its diet does come from turf algae (e.g., Ceramium spp. and Herposiphonia sp.) and red macroalgae (e.g., Asparagopsis taxiformis and Corallinales) with a very small portion coming from green algae (Bryopsidales), a substantial portion of reads are derived from taxa in the EAM (e.g., unidentified cyanobacteria and Oscillatoriales).

Acanthurus olivaceus: According to historic data from the Marshall Islands, A. olivaceus can be found predominantly in areas with sandy cover or coral rubble respectively grazing or scraping on the epilithic algal matrix84,103. Though they may consume some invertebrates, as with many other surgeonfishes that primarily feed from the epilithic algal matrix, invertebrates are seldom the primary target84. A. olivaceus has a thick walled, gizzard-like stomach and a long intestinal tract, which allows it to effectively perform both mechanical and chemical digestion84. Though this species is not usually taken for human consumption, it is targeted in some locations for food and for aquarium sales making it potentially vulnerable to exploitation220,221. § Metabarcoding comparison: A. olivaceus had a diet similar to C. strigosus in the meta- analysis: it predominantly targeted the EAM, with some additional turf, red, and brown

88 algae. The metabarcoding results were similar to existing literature in that A. olivaceus had a high proportion of reads attributable to the EAM compared to other taxa, demonstrated by high relative read abundances of cyanobacteria (Figure 3.5). A. olivaceus also had reads from red macroalgae (e.g., Bangiaceae and Nemaliophycidae), turf algae (e.g., Ceramium sp. and Herposiphonia sp.), and brown algae (e.g., Sphacelaria tribuloides).

Acanthurus pyroferus: This species is known for mimicry behavior in juveniles, which varies by geographic location but can offer a competitive advantage when feeding in damselfish territories222,223. As an adult, it feeds on sediment and detritus, akin to A. olivaceus or C. striatus224. This species is exploited in parts of its range and has shown recovery in response to protected areas225.

Acanthurus triostegus: This species was reported to be more specialized in cropping filamentous red algae than many of its congeners, which also appear to intentionally graze on the epilithic algal matrix84,226. This observation is further supported by 1) the nature of their intestinal tract, which does not include the thick walled portion that is present in many other surgeonfishes but is quite long, akin to A. leucopareius, and 2) feeding morphology, which appears to be very efficient for consuming filamentous algae84,103. The ability to be more specialized may be conferred by the safety that comes from schooling in large numbers103. § Metabarcoding comparison: The greatest proportion of reads seen in A. triostegus were from turf algae (e.g., Ceramium spp., Herposiphonia sp., Polysiphonia sp., and Gayliella sp.) and red macroalgae (e.g., Asparagopsis taxiformis, Acanthophora spicifera, Dasya spp., Laurencia spp., and Corallinales) algae (Figure 3.5). There was a much smaller proportion of green algae (e.g., Ulva sp. and Bryopsidales) and taxa found in the EAM (e.g., unidentified cyanobacteria and Oscillatoriales). A. triostegus also consumed a small amount of brown algae (e.g., Sphacelaria tribuloides and Lobophora variegata), which was not reported in the other studies that we referenced for the diet diversity index, but has been observed in other studies in Hawaiʻi227.

89 Acanthurus xanthopterus: This species has a gizzard-like stomach that aids in the digestion of coarser material and detritus84. It is found in habitats that have a sandy benthic environment, and it appears to graze on the EAM and algal filaments within this sandy habitat84,103. In the study that we incorporated into the diet diversity analysis, A. xanthopterus had a diet that was primarily turf algae and EAM, along with benthic invertebrates155. There has been some question about whether or not A. xanthopterus consumes invertebrates (e.g., hydroids) incidentally but regardless, they appear to consume a greater amount of invertebrates as compared to other species that mainly target the epilithic algal matrix84.

Centropyge bispinosa: This species is wide ranging and widespread, consuming a large amount of detritus and also benthic invertebrates, including sponges228.

Centropyge vrolikii: This species, which is mimicked by juvenile Acanthurus pyroferus to gain access to defended damselfish territories, has a diverse diet that includes red and green algae, along with sponges, detritus, and sediment222.

Chlorurus sordidus: C. sordidus appears to be more resilient to stressors because of its smaller size, quick maturation and growth, and high gonado-trophic index (similar to Scarus schlegeli)229. Though this species can be long lived, they undergo sex transition quite early on which can increase their growth rate and likely enhance resilience230. When coral cover is high, C. sordidus appears to target epiphytes on macroalgae, such as Sargassum, making it one of the dominant “grazers”, but on reefs with lower coral cover, it was not seen targeting this food source126.

Ctenochaetus striatus: Ctenochaetus striatus is acknowledged in both historic and contemporary diet studies as primarily targeting detritus in the epilithic algal matrix84,127,217. Though it appears to consume some benthic invertebrates, which is likely incidental, the vast majority of its diet is the EAM119,144,155. C. striatus appears to avoid feeding on biofilms containing ciliate toxins, indicating that even within the EAM, there may be more selectivity than previously imagined231. This species plays an important role on coral reefs by transporting sediment off of the reef,

90 thereby creating greater access to turf algae food resources for other fishes127,180. It seems that C. striatus is also more sensitive than some other species to high sediment loads which limit its ability to feed, so as sedimentation becomes an increasingly common issue globally, this disturbance may pose a significant threat to the persistence of this species and its ecosystem service105. In addition this species is targeted as a food fish in parts of its range, but it has shown population recovery in protected areas225.

Ctenochaetus strigosus: This species is endemic to Hawaiʻi and fills a similar ecological niche as its sister species C. striatus. Unlike some of the Acanthurus spp. detritivores which are often found in sandy areas, C. striatus and C. strigosus are both found on reef habitat below the surge zone103. C. strigosus has been reported to consume organic matter from the EAM almost exclusively124. § Metabarcoding comparison: Because detritus is not specifically captured by metabarcoding, the relative read abundance of algae from the gut contents of C. strigosus overrepresents the amount of algae in their diet. Despite that, approximately 20% of their diet as measured by relative read abundance was attributable to organisms found in the EAM including unidentified cyanobacteria, Oscillatoriales, Chromerida, and others (Figure 3.5). The rest of the reads were assigned to turf algae (e.g., Ceramiaceae and Herposiphonia sp.) and red macroalgae (e.g., Corallinales, Peyssonnelia sp., and Gigartinales), with a small portion also being assigned to brown (e.g., Sphacelaria tribuloides) and green algae.

Kyphosus cinerascens: As compared to other browsers, K. cinerascens exhibits a fairly broad diet144,232. Though they do not consume detritus or invertebrates and only seem to eat minimal turf algae, they do consume a small amount of brown, green, and red algae, as opposed to some other herbivores that focus more exclusively on brown (e.g., Naso unicornis), green (e.g., Zebrasoma veliferum), or red (e.g., Acanthurus triostegus) algae. It was reported in a historic study that one specimen examined had only consumed garbage84.

91 Kyphosus vaigiensis: K. vaigiensis has a relatively diverse diet among herbivores, but it consumes far more brown and green algae than red, especially as compared to K. cinerascens144.

Naso annulatus: This species tends to feed on both green algae and zooplankton224. It has also been observed in deeper depths than some other surgeonfishes233.

Naso brevirostris: Unlike N. unicornis, N. brevirostris adults are reported to consume more green algae than brown and also significant amounts of zooplankton224,234. Subadults consume a diet more similar to N. unicornis, though N. unicornis typically occurs at greater abundances than N. brevirostris103.

Naso lituratus: Unlike other surgeonfishes, Naso lituratus mostly targets larger macroalgae fronds, typically brown algae, without eating the lower portions that are likely to be in contact with the EAM84,217,236. Hence they do not consume much sediment or EAM and therefore have a more specialized diet than most of the surgeonfishes examined. These findings are in keeping with the anatomy of Naso spp., which includes teeth that are more compact and blunt than acanthurids, along with thin walled stomachs84. These species are harvested for food and by aquarium collectors, and they have been shown to increase in abundance in protected areas237–239. § Metabarcoding comparison: Metabarcoding results indicate that N. lituratus primarily consumes brown algae (e.g., Sphacelaria tribuloides and Dictyotaceae) but also consumes a smaller amount of red macroalgae (e.g., Asparagopsis taxiformis, Laurencia sp., Corallinales, and Gigartinales) and turf algae (e.g., Ceramium spp. and Tolypiocladia glomerulata). N. lituratus also consumes a small amount of green algae (Bryopsidales) and EAM indicated by the presence of cyanobacteria and other taxa found in benthic detritus (Figure 3.5).

Naso unicornis: N. unicornis is also considered a browser of brown algae119,144,155,217. As with N. lituratus, N. unicornis is often exclusive in its consumption of algae without accompanying detritus236. Past work suggests that N. unicornis targets Sargassum and Dictyota, while N. lituratus consumes more Pocockiella, which allows for resource partitioning within the same

92 habitat103,126,217. N. unicornis is consistently recognized as an important consumer of brown algae126,217, though it seems other species, such as N. elegans which is a sister species of N. lituratus, may dominate macroalgae consumption in some degraded habitats, perhaps indicating additional complexity in consumer dynamics on degraded reefs126. It is important to note that N. unicornis was also observed consuming Gracilaria salicornia, an invasive red algae that has overgrown some reefs in Hawaiʻi, and it transfers viable fragments of the algae11. This provides an interesting example of the potential capacity for species, even those with seemingly highly specialized diets, to shift their consumption based on the availability of preferred resources and changing environmental conditions. N. unicornis is heavily targeted as a food fish in much of its range239,240. § Metabarcoding comparison: The N. unicornis examined primarily consumed brown algae (unidentified Dictyotaceae and Turbinaria ornata), with small amounts of red macroalgae (e.g., Spyridia sp., Acanthophora spicifera, and Chondria sp.) and turf algae (e.g., Ceramiales) (Figure 3.5).

Scarus ghobban: Among parrotfish, S. ghobban has a very diverse diet. Even in immature individuals, this species exhibits isotopic signatures that group it more closely with excavating parrotfishes than with other scrapers117. It was the first Scarid recorded to complete a Lessepsian migration into the Mediterranean, where it was feared as a threat to the native ecosystem given its size and versatile diet160. Despite being a generalist, this species has shown a significant increase in abundance with live coral and crustose coralline algae cover241.

Scarus schlegeli: Though S. schlegeli is quite specialized in its feeding approach, it appears to consume some epiphytes from larger macroalgae in addition to its consumption of EAM217. While S. schlegeli does not appear to be deterred by physical defenses (e.g., mineralization) when feeding, experimental trials suggest that it is deterred by chemical defenses in certain species of algae242. Despite its relative specialization, past work has found that S. schlegeli is somewhat resilient to anthropogenic stressors as compared to some other parrotfish species because of its smaller size, fast growth, quick transition to sexual maturity, and high gonado- trophic index229.

93

Siganus argenteus: In feeding trials completed by Paul et al. (1990), S. argenteus preferred the green algae Caulerpa racemosa, Cladophoropsis membranacea, Chlorodesmis fastigiata, Ulva clathrata, and Valonia fastigiata, but they also consumed a variety of cyanobacteria and sea grasses, along with brown and red algae243. Many of the taxa consumed were physically or chemically defended, indicating that these defense mechanisms did not have a great effect on this species243.

Siganus punctatus: The diet of this species is consistent in both inner and mid-reef shelf environments, as it has a diverse and variable diet in both cases152. Specifically, it consumes foliose brown algae, filamentous and corticated red algae, jointed calcareous algae, cyanobacteria, organic material, sediment, and invertebrates, thereby exhibiting feeding patterns that are characteristic of several different functional groups152.

Siganus spinus: S. spinus preferentially consumes both red and green algae, though selection may vary by algal availability244. In experimental feeding trials this species and its congener, Siganus argenteus, appear to be unaffected by a variety of secondary metabolites and algal extracts that inhibit feeding by other herbivores245. Though S. spinus appears to have a more specialized diet as compared to its congeners, S. argenteus and S. punctatus, it still has a relatively diverse diet as compared to many other nominally herbivorous reef fishes.

Zebrasoma flavescens: Z. flavescens and Z. scopas are sister species246. Z. veliferum and Z. flavescens overlap in their range, with Z. flavescens being very abundant in Hawaiʻi and likely outcompeting Z. veliferum in this area103. § Metabarcoding comparison: Though we were not able to examine Z. veliferum or Z. scopas in our metabarcoding study, Z. flavescens’ diet is dominated by rhodophytes, with the majority falling into the category of turf algae (e.g. Ceramium spp., Heterosiphonia crispella, and Herposiphonia sp.) (Figure 3.5). As with other grazers, Z. flavescens also consumes red macroalgae (e.g., Asparagopsis taxiformis, Dasya spp., Laurencia sp., Corallinales, and Gigartinales) but also appears to consume more green algae (e.g., Ulva

94 sp., and Bryopsidales) than the other species examined along with cyanobacteria, which is indicative of the EAM. Z. flavescens individuals consumed the least brown algae of all the species included in the metabarcoding analysis, but they consumed the most green algae, which is in keeping with the other Zebrasoma spp. included in the diet diversity index. Anecdotally, a considerable amount of coarse sediment was observed in the stomachs of the individuals used for this study, though those amounts were not explicitly quantified. We would recommend the placement of Z. flavescens near Z. scopas in the diet diversity index.

Zebrasoma scopas: Though it does eat more green algae than many other grazing surgeonfishes, Z. scopas targets far less green algae than Z. veliferum119,144. It instead consumes a greater proportion of red algae, turf algae, and brown algae with little indication of EAM119,144.

Zebrasoma veliferum: Z. veliferum does not exhibit a gizzard or thick walled stomach, as is seen in many other surgeonfishes84. They do appear to consume some detritus incidentally while feeding on the EAM, but they target turf algae15,84,224. They consume more green algae than other surgeonfishes examined in this study, with a comparably small proportion of red algae119.

95 APPENDIX E CHAPTER 4 SUPPLEMENTAL TABLES

Table S4.1. Description of predictor variables with reason for inclusion. Driver Units Source Importance Scale

Sea Surface Lower ID Williams et al. SST has been shown Island Temperature climatological 2015 – based on to have a significant mean - °C Gove et al. 2013 relationship with herbivore biomass (Heenan et al. 2016)

Mean wave kW m-1 ID Williams et al. Moderate wave energy Island energy 2015 – based on is associated with high Gove et al. 2013 grazer biomass (Heenan et al. 2016)

Structural m NOAA Coral Reef Complexity is Site Complexity Ecosystem decreasing on many Program; Pacific impacted coral reefs Islands Fisheries and is an important Science Center habitat requirement for (2016) many fish (Alvarez-Filip et al. 2009; Graham et al. 2007; Emslie et al. 2014; Heenan et al. 2016)

Island Ha ID Williams et al. High forereef area is Island forereef area 2015 associated with higher detritivore biomass (Heenan et al. 2016)

Human Individuals NOAA Coral Reef Human population is a Site Population within a 20 Ecosystem simple and often used km radius Program; Pacific measure of human Islands Fisheries impacts (Mora et al. Science Center 2011; GJ Williams et (2016) - based on al. 2015; Duffy et al. US Census data 2016)

96 Table S4.1. (Continued) Description of predictor variables with reason for inclusion. Total fish g m-2 NOAA Coral Reef Lower fish biomass, Site biomass Ecosystem which can be Program; Pacific indicative of a loss of Islands Fisheries large predator Science Center biomass, is associated (2016) with more impacted areas (DeMartini et al. 2008)

Herbivore Individuals m-2 NOAA Coral Reef Herbivore biomass has Site abundance Ecosystem been shown to Program; Pacific decrease with human Islands Fisheries density Science Center (ID Williams et al. (2016) 2015)

Herbivore Simpson’s Based on Diversity can Site diversity Diversity herbivore ameliorate the effects Index (1-D) abundance data of environmental from 0 to 1 from NOAA Coral impacts on fish where 1 is the Reef Ecosystem biomass most diverse Program; Pacific (Duffy et al. 2016) Islands Fisheries Science Center (2016)

Hard coral Proportion of NOAA Coral Reef Reef fish biomass, Site cover benthic cover Ecosystem richness, and diversity (0 to 1) Program; Pacific have been found to Islands Fisheries vary with benthic Science Center complexity and coral (2016) cover (Friedlander et al. 2003; Graham & Nash 2013)

97

Table S4.1. (Continued) Description of predictor variables with reason for inclusion. Proportion of Proportion Generated using The biomass of Site Herbivores in (0 to 1) data from NOAA herbivores as part of Total Fish Coral Reef the total fish biomass Biomass Ecosystem shifts with human Program; Pacific impacts and can be Islands Fisheries informative for Science Center understanding (2016) community composition (I.D. Williams 2015) Human Score (0 to National Center for The Human Impact Island Impact Score 16) Ecological Analysis Score, developed and Synthesis using 17 different (NCEAS) (Halpern anthropogenically et al. 2015). Island derived stressors that mean values were affect the condition of taken from marine ecosystems, (McDole Somera et was used as one al. 2016) measure of human impacts 186,187

98

APPENDIX F CHAPTER 4 SUPPLEMENTAL FIGURES

1.00

0.95

0.90

0.85 Proportion of Total Herbivore Richness Herbivore Proportion of Total

0.80

Tau Sarig KureLanai Maro Maui Oahu RoseRota − Baker GuamHawaii Jarvis Kauai Maug NihoaNiihau Pagan TinianTutuilaWake AgrihanAguijan Laysan MidwayMolokaiNecker Palmyra Saipan Swains Asuncion Gardner HowlandJohnston Kingman Lisianski Gug − Kahoolawe South Bank French Frigate Ofu & OlosegaPearl & Hermes Alam Farallon de Pajaros Island Figure S4.1. Proportion of herbivores present at a site (i.e., total herbivore richness) that were included in the herbivore assemblage diet specialization index based on the Diet Diversity Index from Chapter 3.

99 MHI NWHI PRIAs SAMOA S.MARIANAS N.MARIANAS

100%

75%

50% Abundance (>5%) Abundance

25%

0%

Tau OahuKauaiMaui Lanai Nihoa Kure Maro JarvisBaker Rose Rota Sarig NiihauHawaii Laysan MidwayNecker Tutuila AguijanTinianSaipanGuam Pagan − Molokai Gardner Lisianski HowlandJohnstonKingmanPalmyra Agrihan Asuncion Kahoolawe −Gug French Frigate Ofu & Olosega Pearl & Hermes Alam Farallon de Pajaros

ACAC_abund ACNR_abund CELO_abund CTBI_abund NALI_abund ACBL_abund ACOL_abund CEPO_abund CTCY_abund NAUN_abund ACLE_abund ACTR_abund CESH_abund CTMA_abund SCDU_abund ACLI_abund CEBI_abund CHPE_abund CTSR_abund ZEFL_abund ACNC_abund CEFL_abund CHSO_abund CTST_abund ACNF_abund CEHE_abund CHUB_abund NABR_abund

Figure S4.2. Assemblage composition based on the relative abundance of herbivores present averaged by island and grouped by region. Only species that comprised at least 5% of the overall abundance were included. Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation).

100 Hawaii Kahoolawe Kauai Lanai Maui Molokai Niihau Oahu 150

SCPS CTHA CTST CHUB 100 CHSO SCSP SCRU ACOL NALI NAUN ACTR NABR ZEFL ACNF

50 ACLE

Frequency of Sites with Max Abundance ACNR CEFI CEPO ABSO ACDU

0

Figure S4.3 (a) The frequency with which a given species was the most abundant species at a site, grouped by island in the Main Hawaiian Islands. Species codes in the legend are ordered from least diverse diet to most diverse diet (top to bottom). Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation).

101 French Frigate Gardner Kure Laysan Lisianski Maro Midway Necker Nihoa Pearl & Hermes

40

SCDU SCPS CTST CHUB CHSO 30 SCSP CHPE SCRU ACOL NAUN ACTR NABR 20 ACNF ACLE ACNR

Frequency of Sites with Max Abundance CEFI CEFL CEPO 10 ABSO ACDU

0

Figure S4.3. (b) The frequency with which a given species was the most abundant species at a site, grouped by island in the Papahānaumokuākea Marine National Monument. Species codes in the legend are ordered from least diverse diet to most diverse diet (top to bottom). Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation).

102 Baker Howland Jarvis Johnston Kingman Palmyra

60

CTST CHUB CHSO CHFN CTCY CTMA

40 CTSR ACOL ZERO ACTR NABR CELO ACAC ACNC ACNF Frequency of Sites with Max Abundance 20 ACNR ACLI CEFL

0

Figure S4.3. (c) The frequency with which a given species was the most abundant species at a site, grouped by island in the Pacific Remote Island Areas. Species codes in the legend are ordered from least diverse diet to most diverse diet (top to bottom). Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation).

103 Ofu & Olosega Rose Tau Tutuila

200

150 SCFO NALI SCPS ACTR SCSC CEBI SCTR CELO CHSO ACAC SCSP ACNC CHFN ACNF 100 CHJA ACGU CHMC ACNR CTBI ACLI CTCY CEFI CTSR CEFL

Frequency of Sites with Max Abundance ACOL CEHE

50

0

Figure S4.3. (d) The frequency with which a given species was the most abundant species at a site, grouped by island in Samoa. Species codes in the legend are ordered from least diverse diet to most diverse diet (top to bottom). Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation).

104 Agrihan Alam−Gug−Sarig Asuncion Farallon de Pajaros Pagan 25

20

CTHA CHUB CTBI CTCY 15 CTSR ACOL NALI ACTR CESH ZEFL 10 ACNC ACNF ACLE ACGU Frequency of Sites with Max Abundance ACLI CEFL 5

0

Figure S4.3. (e) The frequency with which a given species was the most abundant species at a site, grouped by island in the Northern Marianas Islands. Species codes in the legend are ordered from least diverse diet to most diverse diet (top to bottom). Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation).

105 Aguijan Guam Rota Saipan Tinian

150

100 SCPS ACTR CHSO CESH SCSP ZEFL CHFN ACNC SCAL ACNF CTBI ACGU CTCY ACLI CTSR CEFL ACOL CEHE ACPY ABSO 50 NALI Frequency of Sites with Max Abundance

0

Figure S4.3. (f) The frequency with which a given species was the most abundant species at a site, grouped by island in the Southern Marianas Islands. Species codes in the legend are ordered from least diverse diet to most diverse diet (top to bottom). Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation).

106 Hard Coral Benthic Complexity Diversity Herb. Diet Diversity Herb. of Biomass Prop. Herb. Human Impact Score Fish Biomass Total Energy Wave Abundance Herbivore Human Population 1

0.8 0.2 0.6 0.2 0.24 0.4 −0.25 −0.12 −0.39 0.2 −0.05 0.19 0.14 −0.17 0 −0.16 −0.13 −0.29 0.05 0.28 −0.2 0.19 0.25 0.4 −0.17 −0.2 −0.59 −0.4 −0.12 −0.07 −0.2 0.06 −0.08 0.12 0.03 −0.6 0.16 0.23 0.12 −0.2 0.34 −0.12 0.31 −0.05 −0.8 −0.12 −0.14 −0.24 0 0.33 0.9 −0.57 −0.08 −0.09 −1

Figure S4.4. Correlation plot for a combination variables of interest for use in a linear mixed effects model. Variables with r > 0.7 were not included together.

107 MHI PMNM PRIAs Samoa S. Marianas N. Marianas

0.6

0.4 Herbivore Functional Dissimilarity Herbivore

0.2

0.0

Tau Lanai Maui Oahu Kure Maro Rose Rota Sarig Hawaii Kauai Niihau Nihoa Baker Jarvis Wake Tutuila Guam Tinian − MaugPagan Molokai Laysan MidwayNecker Palmyra Swains Aguijan Saipan Agrihan Gardner Lisianski Howland JohnstonKingman Asuncion Gug Kahoolawe South Bank − French Frigate Pearl & Hermes Ofu & Olosega Alam Farallon de Pajaros Figure S4.5. Herbivore functional dissimilarity by island, grouped by region. The dark blue lines represent the mean value for the region.

108 0.6

0.4

Detritivores Scrapers & Excavators Grazers Browsers Omnivores

Relative Abundance Relative 0.2

0.0

4 8 12 16 NCEAS Human Impact Score

Figure S4.6. The relative abundance of nominally herbivorous reef fishes grouped by functional groups were compared to the NCEAS Human Impact Score (38 islands). The NCEAS Human Impact Score is developed using 17 anthropogenic stressors that affect coral reefs, and impacts increase from left to right. Gray areas show 95% confidence intervals.

109 A. nigricauda A. blochii A. achilles A. leucopareius A. lineatus A. olivaceus A. triostegus A. nigricans A. guttatus A. dussumieri A. pyroferus A. xanthopterus A. nigrofuscus A. nigroris

0.4 0.4

0.3 0.3

0.2 0.2

Relative Abundance Relative 0.1 0.1

0.0 0.0

4 8 12 16 1e+02 1e+04 1e+06 NCEAS Human Impact Score Human Population within 20 km

Figure S4.7. The relative abundance of Acanthurus spp. were compared to the NCEAS Human Impact Score (left, 38 islands) and the population of humans within a 20 km radius (right, plotted on a log scale, 42) throughout the Pacific. When examined independently Acanthurus nigrofuscus had a significant positive relationship with the Human Impact Score (P < 0.0001, Estimate = 0.29, R2 marginal = 0.09, R2 conditional = 0.23).

110 Ordination with ecological and environmental vectors Stress = 0.14

1.0

Fish Biomass

0.5 Chlorophyll Functional Dissimilarity Human Population Mean Irradiance CCA Waves within 20 km Herbivore Diversity unpopulated minimal inhabitants

NMDS2 Hard Coral populated densely populated

0.0

SST EAM

Functional Redundancy Forereef Area

−0.5 Herbivore Relative Biomass Human Impact Score

−1.0 −0.5 0.0 0.5 1.0 NMDS1

Figure S4.8. NMDS of herbivore assemblage composition at the island scale across regions in the Pacific with significant environmental variables overlaid (k = 2, stress = 0.14).

111 Ordination with species vectors Stress = 0.14

1.0

CELO ACNR CTMA ACTR 0.5 SCTR ZERO CHPE CEPO ACLU CAZO SCFR Human Population ACAC ZEVE SCDU within 20 km ACNC CTFL NAUN CTSP unpopulated CEFL ZESC CTCY minimal inhabitants NMDS2 CHMC ACOL populated SCNI densely populated 0.0 SCSI CTST CHFN ACNI

SCOV ACBL SCGL SCFO CEFI CEOC ACDU ACLI NATO SCDM ACLE ACGU SCPS CTSR CESH ZEFL SCSC −0.5 ACPY CEHE CTBI

ACNF

−1.0 −0.5 0.0 0.5 1.0 NMDS1 Figure S4.9. NMDS of herbivore assemblage composition at the island scale across regions in the Pacific with significant herbivore species overlaid (k = 2, stress = 0.14). Species names corresponding to abbreviations are available in raw data files on GitHub (in enalley/Dissertation).

112 Maug Wake Maug Palmyra Kingman Wake FarallonPalmyra de Pajaros Kingman Swains AsuncionFarallon deJohnston Pajaros Swains 0.6 Alam−GugAsuncion−Sarig RoseJohnston 0.6 Alam−Gug−Sarig Rose Tau size Tutuila Agrihan Tau sizea 15 Howland Tutuila Agrihan a 15 BakerHowland Ofu & Olosega Rota Baker Ofu & Olosega factor(REGION) Rota Saipan Guam factor(REGION)a MHI Jarvis Tinian Saipan Jarvis LanaiGuam a MainMHINWHI Hawaiian Islands Tinian Hawaii Aguijan MolokaiLanai a PacificNWHIPRIAsRemote Island Areas 0.5 Hawaii a Aguijan Molokai a SamoaPRIAsSAMOA 0.5 Niihau a SouthernSAMOAS.MARIANMarianas NiihauMaug a NorthernS.MARIANN.MARIANMarianas a N.MARIAN 0.8 Maui

Functional Dissimilarity of All Herbivores Palmyra SwainsMaui Functional Dissimilarity of All Herbivores Alam−Gug−Sarig Rose Wake Kingman Farallon de Pajaros Tau Asuncion Johnston Agrihan Tutuila size 0.4 0.4 Howland Ofu &Kauai Olosega a 15 Kauai Rota Oahu 0.7 Oahu Guam factor(REGION) Jarvis Baker 4 8 12 Aguijan Saipan16 a MHI 4 NCEAS8 Human Impact Score12 16Hawaii a NWHI Tinian NCEAS Human Impact Score Lanai a PRIAs a SAMOA Niihau Molokai a S.MARIAN a N.MARIAN

0.6 Simpson's Diversity of All Herbivores Simpson's Diversity Maui

Kauai 0.5 Oahu

4 8 12 16 NCEAS Human Impact Score

Figure S4.10. Herbivore diversity compared to NCEAS Human Impact Score. Colors correspond to regions. NCEAS Human Impact Scores were not available for the Papahānaumokuākea Marine National Monument (Northwestern Hawaiian Islands).

113 REFERENCES

1. Hutchinson, G. E. Concluding remarks. Cold Springs Harbor Symposium. Quant. Biol. 22, 415–427 (1957). 2. Roughgarden, J. Evolution of niche width. Am. Nat. 106, 683–718 (1972). 3. Werner, E. E. & Mittelbach, G. G. Experimental tests of optimal habitat use in fish: The role of relative habitat profitability. Ecology 64, 1525–1539 (1983). 4. Tilman, D. Niche tradeoffs, neutrality, and community structure: A stochastic theory of resource competition, invasion, and community assembly. Proc. Natl. Acad. Sci. 101, 10854–61 (2004). 5. Roberts, C. M. et al. Marine biodiversity hotspots and conservation priorities for tropical reefs. Science (80-. ). 295, 1280–1284 (2002). 6. Hoegh-Guldberg, O. et al. Coral reefs under rapid climate change and ocean acidification. Science 318, 1737–42 (2007). 7. Hughes, T. P. et al. Global warming and recurrent mass bleaching of corals. Nature 543, 373–377 (2017). 8. Alvarez-Filip, L., Côté, I. M., Gill, J. A., Watkinson, A. R. & Dulvy, N. K. Region-wide temporal and spatial variation in Caribbean reef architecture: Is coral cover the whole story? Glob. Chang. Biol. (2011). doi:10.1111/j.1365-2486.2010.02385.x 9. Fabricius, K. E. Effects of terrestrial runoff on the ecology of corals and coral reefs: Review and synthesis. Mar. Pollut. Bull. 50, 125–146 (2005). 10. Martinez, J. a., Smith, C. M. & Richmond, R. H. Invasive algal mats degrade coral reef physical habitat quality. Estuar. Coast. Shelf Sci. 99, 42–49 (2012). 11. Bierwagen, S. L., Price, D. K., Pack, A. A. & Meyer, C. G. Bluespine unicornfish (Naso unicornis) are both natural control agents and mobile vectors for invasive algae in a Hawaiian Marine Reserve. Mar. Biol. 164, 1–14 (2017). 12. Goatley, C. & Bellwood, D. Biologically mediated sediment fluxes on coral reefs: sediment removal and off-reef transportation by the surgeonfish Ctenochaetus striatus. Mar. Ecol. Prog. Ser. 415, 237–245 (2010). 13. Bonaldo, R. M. & Bellwood, D. R. Dynamics of parrotfish grazing scars. Mar. Biol. 156,

114 771–777 (2009). 14. Clements, K. D., German, D. P., Piché, J., Tribollet, A. & Choat, J. H. Integrating ecological roles and trophic diversification on coral reefs: multiple lines of evidence identify parrotfishes as microphages. Biol. J. Linn. Soc. 120, 729–751 (2017). 15. Marshell, A. & Mumby, P. J. The role of surgeonfish (Acanthuridae) in maintaining algal turf biomass on coral reefs. J. Exp. Mar. Bio. Ecol. 473, 152–160 (2015). 16. Hixon, M. A. Reef fishes, seaweeds, and corals: a complex triangle. in Coral reefs in the Anthropocene (ed. Birkeland, C.) 195–215 (Springer Netherlands, 2015). 17. Hughes, T. P. et al. Phase shifts, herbivory, and the resilience of coral reefs to climate change. Curr. Biol. 17, 360–365 (2007). 18. Jouffray, J.-B. et al. Human and natural drivers of multiple coral reef regimes across the Hawaiian Archipelago. Philos. Trans. B (2014). doi:10.1098/rstb.2013.0268 19. Ong, L. & Holland, K. N. Bioerosion of coral reefs by two Hawaiian parrotfishes: species, size differences and fishery implications. Mar. Biol. 157, 1313–1323 (2010). 20. Tebbett, S. B., Goatley, C. H. R., Huertas, V., Mihalitsis, M. & Bellwood, D. R. A functional evaluation of feeding in the surgeonfish Ctenochaetus striatus: The role of soft tissues. R. Soc. Open Sci. 5, 171111 (2018). 21. Bellwood, D. R., Hoey, A. S. & Choat, J. H. Limited functional redundancy in high diversity systems: resilience and ecosystem function on coral reefs. Ecol. Lett. 6, 281–285 (2003). 22. Micheli, F. & Halpern, B. S. Low functional redundancy in coastal marine assemblages. Ecol. Lett. 8, 391–400 (2005). 23. Bellwood, D. R., Wainwright, P. C., Fulton, C. J. & Hoey, A. S. Functional versatility supports coral reef biodiversity. Proc. R. Soc. B Biol. Sci. 273, 101–107 (2006). 24. Nash, K. L., Graham, N. A. J., Jennings, S., Wilson, S. K. & Bellwood, D. R. Herbivore cross-scale redundancy supports response diversity and promotes coral reef resilience. J. Appl. Ecol. 646–655 (2015). doi:10.1111/1365-2664.12430 25. Julliard, R., Clavel, J., Devictor, V., Jiguet, F. & Couvet, D. Spatial segregation of specialists and generalists in bird communities. Ecol. Lett. 9, 1237–1244 (2006). 26. Devictor, V. et al. Functional biotic homogenization of bird communities in disturbed

115 landscapes. Glob. Ecol. Biogeogr. 17, 252–261 (2008). 27. Clavel, J., Julliard, R. & Devictor, V. Worldwide decline of specialist species: toward a global functional homogenization? Front. Ecol. Environ. 9, 222–228 (2011). 28. Abadie, J. C., Machon, N., Muratet, A. & Porcher, E. Landscape disturbance causes small- scale functional homogenization, but limited taxonomic homogenization, in plant communities. J. Ecol. 99, 1134–1142 (2011). 29. Merckx, T. & Van Dyck, H. Urbanization-driven homogenization is more pronounced and happens at wider spatial scales in nocturnal and mobile flying insects. Glob. Ecol. Biogeogr. 28, 1440–1455 (2019). 30. Bolnick, D. I. et al. The ecology of individuals: incidence and implications of individual specialization. Am. Nat. 161, 1–28 (2003). 31. Newsome, S. D. et al. Using stable isotopes to investigate individual diet specialization in California sea otters ( Enhydra lutris nereis ). Ecology 90, 961–974 (2009). 32. Hammerschlag-Peyer, C. M. & Layman, C. a. Intrapopulation variation in habitat use by two abundant coastal fish species. Mar. Ecol. Prog. Ser. 415, 211–220 (2010). 33. Araújo, M. S., Langerhans, R. B., Giery, S. T. & Layman, C. a. Ecosystem fragmentation drives increased diet variation in an endemic livebearing fish of the Bahamas. Ecol. Evol. 4, 3298–3308 (2014). 34. Olden, J. D. Biotic homogenization: a new research agenda for conservation biogeography. J. Biogeogr. 33, 2027–2039 (2006). 35. Devictor, V. et al. Defining and measuring ecological specialization. J. Appl. Ecol. 47, 15–25 (2010). 36. Filippi-Codaccioni, O., Devictor, V., Bas, Y. & Julliard, R. Toward more concern for specialisation and less for species diversity in conserving farmland biodiversity. Biol. Conserv. 143, 1493–1500 (2010). 37. Côté, I., Green, S., Morris, J., Akins, J. & Steinke, D. Diet richness of invasive Indo- Pacific lionfish revealed by DNA barcoding. Mar. Ecol. Prog. Ser. 472, 249–256 (2013). 38. Berry, O. et al. Comparison of morphological and DNA metabarcoding analyses of diets in exploited marine fishes. Mar. Ecol. Prog. Ser. 540, 167–181 (2015). 39. Harms-Tuohy, C. A., Schizas, N. V. & Appeldoorn, R. S. Use of DNA metabarcoding for

116 stomach content analysis in the invasive lionfish pterois volitans in Puerto Rico. Mar. Ecol. Prog. Ser. 558, 181–191 (2016). 40. McInnes, J. C. et al. High occurrence of jellyfish predation by black-browed and Campbell albatross identified by DNA metabarcoding. Mol. Ecol. 26, 4831–4845 (2017). 41. Kaunisto, K. M., Roslin, T., Sääksjärvi, I. E. & Vesterinen, E. J. Pellets of proof: First glimpse of the dietary composition of adult odonates as revealed by metabarcoding of feces. Ecol. Evol. 7, 8588–8598 (2017). 42. Pompanon, F. et al. Who is eating what: diet assessment using next generation sequencing. Mol. Ecol. 21, 1931–50 (2012). 43. Arrizabalaga-Escudero, A. et al. Trophic requirements beyond foraging habitats: The importance of prey source habitats in bat conservation. Biol. Conserv. 191, 512–519 (2015). 44. Hawkins, J. et al. Using DNA metabarcoding to identify the floral composition of honey: A new tool for investigating honey bee foraging preferences. PLoS One 10, 1–20 (2015). 45. McClenaghan, B., Gibson, J. F., Shokralla, S. & Hajibabaei, M. Discrimination of grasshopper (Orthoptera: Acrididae) diet and niche overlap using next-generation sequencing of gut contents. Ecol. Evol. 5, 3046–3055 (2015). 46. Berry, T. E. et al. DNA metabarcoding for diet analysis and biodiversity: A case study using the endangered Australian sea lion (Neophoca cinerea). Ecol. Evol. 7, 5435–5453 (2017). 47. Siegenthaler, A., Wangensteen, O. S., Benvenuto, C., Campos, J. & Mariani, S. DNA metabarcoding unveils multiscale trophic variation in a widespread coastal opportunist. Mol. Ecol. 28, 232–249 (2018). 48. Walters, T. L. et al. Diet and trophic interactions of a circumglobally significant gelatinous marine zooplankter, Dolioletta gegenbauri (Uljanin, 1884). Mol. Ecol. 176–189 (2018). doi:10.1111/mec.14926 49. Ji, Y. et al. Reliable, verifiable and efficient monitoring of biodiversity via metabarcoding. Ecol. Lett. 16, 1245–1257 (2013). 50. Yang, C. et al. Using metabarcoding to ask if easily collected soil and leaf-litter samples can be used as a general biodiversity indicator. Ecol. Indic. 46, 379–389 (2014).

117 51. Ardura, A. et al. Environmental DNA evidence of transfer of North Sea molluscs across tropical waters through ballast water. J. Molluscan Stud. 81, 495–501 (2015). 52. Thomsen, P. F. & Willerslev, E. Environmental DNA - An emerging tool in conservation for monitoring past and present biodiversity. Biol. Conserv. 183, 4–18 (2015). 53. Andruszkiewicz, E. A. et al. Biomonitoring of marine vertebrates in Monterey Bay using eDNA metabarcoding. PLoS One 12, 1–20 (2017). 54. Seeber, P. A. et al. Terrestrial mammal surveillance using hybridization capture of environmental DNA from African waterholes. Mol. Ecol. Resour. 19, 1486–1496 (2019). 55. Richardson, R. T. et al. Rank‐based characterization of pollen assemblages collected by honey bees using a multi‐locus metabarcoding approach. Appl. Plant Sci. 3, 1500043 (2015). 56. Richardson, R. T. et al. Application of ITS2 metabarcoding to determine the provenance of pollen collected by honey bees in an agroecosystem. Appl. Plant Sci. 3, 1400066 (2015). 57. Bell, K. L. et al. Applying pollen DNA metabarcoding to the study of plant–pollinator interactions. Appl. Plant Sci. 5, 1600124 (2017). 58. Dejean, T. et al. Improved detection of an alien invasive species through environmental DNA barcoding: The example of the American bullfrog Lithobates catesbeianus. J. Appl. Ecol. 49, 953–959 (2012). 59. Wade, R. M. & Sherwood, A. R. Updating Plakobranchus cf . ianthobapsus (Gastropoda, Sacoglossa) host use: Diverse algal-animal interactions revealed by NGS with implications for invasive species management. Mol. Phylogenet. Evol. 128, 172–181 (2018). 60. Mychek-londer, J. G., Balasingham, K. D. & Heath, D. D. Using environmental DNA metabarcoding to map invasive and native invertebrates in two Great Lakes tributaries. Environ. DNA 1–15 (2019). doi:10.1002/edn3.56 61. Deiner, K. et al. Environmental DNA metabarcoding: Transforming how we survey animal and plant communities. Mol. Ecol. 26, 5872–5895 (2017). 62. Yamamoto, S. et al. Environmental DNA metabarcoding reveals local fish communities in a species-rich coastal sea. Sci. Rep. 7, 1–12 (2017).

118 63. Sawaya, N. A. et al. Assessing eukaryotic biodiversity in the Florida Keys National Marine Sanctuary through environmental DNA metabarcoding. Ecol. Evol. 9, 1029–1040 (2019). 64. Coissac, E., Riaz, T. & Puillandre, N. Bioinformatic challenges for DNA metabarcoding of plants and animals. Mol. Ecol. 21, 1834–47 (2012). 65. Taberlet, P., Coissac, E., Pompanon, F., Brochmann, C. & Willerslev, E. Towards next- generation biodiversity assessment using DNA metabarcoding. Mol. Ecol. 21, 2045–50 (2012). 66. Taberlet, P. et al. Soil sampling and isolation of extracellular DNA from large amount of starting material suitable for metabarcoding studies. Mol. Ecol. 21, 1816–1820 (2012). 67. Deagle, B. E., Thomas, A. C., Shaffer, A. K., Trites, A. W. & Jarman, S. N. Quantifying sequence proportions in a DNA-based diet study using Ion Torrent amplicon sequencing: Which counts count? Mol. Ecol. Resour. 13, 620–633 (2013). 68. Ransome, E. et al. The importance of standardization for biodiversity comparisons: A case study using autonomous reef monitoring structures (ARMS) and metabarcoding to measure cryptic diversity on Mo’orea coral reefs, French Polynesia. PLoS One 12, e0175066 (2017). 69. Lamb, P. D. et al. How quantitative is metabarcoding: A meta-analytical approach. Mol. Ecol. 28, 420–430 (2019). 70. Nichols, P. K. & Marko, P. B. Rapid assessment of coral cover from environmental DNA in Hawai’i. Environ. DNA 1, 40–53 (2019). 71. Bik, H. M. et al. Sequencing our way towards understanding global eukaryotic biodiversity. Trends Ecol. Evol. 27, 233–243 (2012). 72. McGee, K. M., Robinson, C. V. & Hajibabaei, M. Gaps in DNA-based biomonitoring across the globe. Front. Ecol. Evol. 7, 1–7 (2019). 73. Ruppert, K. M., Kline, R. J. & Rahman, S. Past, present, and future perspectives of environmental DNA (eDNA) metabarcoding: A systematic review in methods , monitoring , and applications of global eDNA. Glob. Ecol. Conserv. 17, e00547 (2019). 74. Leray, M. & Knowlton, N. DNA barcoding and metabarcoding of standardized samples reveal patterns of marine benthic diversity. Proc. Natl. Acad. Sci. 112, 201424997 (2015).

119 75. Miya, M. et al. MiFish, a set of universal PCR primers for metabarcoding environmental DNA from fishes detection of more than 230 subtropical marine species. R. Soc. Open Sci. 2, 150088 (2015). 76. Dopheide, A. et al. Estimating the biodiversity of terrestrial invertebrates on a forested island using DNA barcodes and metabarcoding data. Ecol. Appl. 29, 1–14 (2019). 77. Macheriotou, L. et al. Metabarcoding free-living marine nematodes using curated 18S and CO1 reference sequence databases for species-level taxonomic assignments. Ecol. Evol. 9, 1211–1226 (2019). 78. Morín, J. G. et al. Reference DNA barcodes and other mitochondrial markers for identifying Caribbean Octocorals. Biodivers. Data J. 7, (2019). 79. Weigand, H. et al. DNA barcode reference libraries for the monitoring of aquatic biota in Europe: Gap-analysis and recommendations for future work. Sci. Total Environ. 678, 499–524 (2019). 80. Sherwood, A. R. & Presting, G. G. Universal primers amplify a 23S rDNA plastid marker in eukaryotic algae and cyanobacteria. J. Phycol. 43, 605–608 (2007). 81. Wang, N. et al. The Hawaiian Algal Database: a laboratory LIMS and online resource for biodiversity data. BMC Plant Biol. 9, 117 (2009). 82. Sherwood, A. R., Kurihara, A., Conklin, K. Y., Sauvage, T. & Presting, G. G. The Hawaiian Rhodophyta Biodiversity Survey (2006-2010): a summary of principal findings. BMC Plant Biol. 10, 258 (2010). 83. Stamoulis, K. A., Friedlander, A. M., Meyer, C. G., Fernandez-Silva, I. & Toonen, R. J. Coral reef grazer-benthos dynamics complicated by invasive algae in a small marine reserve. Sci. Rep. 7, 43819 (2017). 84. Hiatt, R. W. & Strasburg, D. W. Ecological relationships of the fish fauna on coral reefs of the Marshall Islands. Ecol. Monogr. 30, 65–127 (1960). 85. Choat, J. H. et al. Acanthurus nigrofuscus. The IUCN Red List of Threatened Species 2012: e.T178019A1523035. (2012). doi:10.2305/IUCN.UK.2012.RLTS.T178019A1523035.en 86. Froese, R. & Pauly, D. FishBase. (2017). Available at: https://www.fishbase.de/. 87. McIlwain, J. et al. Acanthurus triostegus. The IUCN Red List of Threatened Species

120 2012: e.T177965A1504553. (2012). doi:10.2305/IUCN.UK.2012.RLTS.T177965A1504553.en 88. Illumina. 16S Metagenomic Sequencing Library Preparation. (2019). Available at: https://support.illumina.com/content/dam/illumina- support/documents/documentation/chemistry_documentation/16s/16s-metagenomic- library-prep-guide-15044223-b.pdf. 89. Saunders, G. W. & Moore, T. E. Refinements for the amplification and sequencing of red algal DNA barcode and RedToL phylogenetic markers: A summary of current primers, profiles and strategies. Algae 28, 31–43 (2013). 90. Sherwood, A. R., Chan, Y. L. & Presting, G. G. Application of universally amplifying plastid primers to environmental sampling of a stream periphyton community. Mol. Ecol. Resour. 8, 1011–1014 (2008). 91. Kozich, J. J., Westcott, S. L., Baxter, N. T., Highlander, S. K. & Schloss, P. D. Development of a dual-index sequencing strategy and curation pipeline for analyzing amplicon sequence data on the MiSeq. Appl. Environ. Microbiol. 79, 5112–5120 (2013). 92. Callahan, B. J. et al. DADA2: High-resolution sample inference from Illumina amplicon data. Nat. Methods 13, (2016). 93. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.r-project.org/. (2017). 94. Wilkinson, S. P., Davy, S. K., Bunce, M. & Stat, M. Taxonomic identification of environmental DNA with informatic sequence classification trees. PeerJ Prepr. (2018). 95. Oksanen, J. et al. vegan: Community Ecology Package. R package version 2.4-4. https://cran.r-project.org/web/packages/vegan/index.html. (2017). 96. McMurdie, P. J. & Holmes, S. Phyloseq: An R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One 8, (2013). 97. Mangiafico, S. S. Summary and analysis of extension program evaluation in R, version 1.13.0. http://rcompanion.org/handbook/. (2016). 98. Deagle, B. E. et al. Counting with DNA in metabarcoding studies: How should we convert sequence reads to dietary data ? Mol. Ecol. 391–406 (2018). doi:10.1111/mec.14734 99. Casey, J. M., Meyer, C. P., Morat, F., Brandl, S. J. & Casey, J. M. Reconstructing

121 hyperdiverse food webs : Gut content metabarcoding as a tool to disentangle trophic interactions on coral reefs. Methods Ecol. Evol. 10, 1157–1170 (2019). 100. Zaccarelli, N., Bolnick, D. I. & Mancinelli, G. RInSp: An r package for the analysis of individual specialization in resource use. Methods Ecol. Evol. 4, 1018–1023 (2013). 101. Tootell, J. S. & Steele, M. A. Distribution, behavior, and condition of herbivorous fishes on coral reefs track algal resources. Oecologia 181, 13–24 (2016). 102. Bates, D., Mächler, M., Bolker, B. M. & Walker, S. C. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67, (2015). 103. Nathan, A. J. & Scobell, A. How China sees America. Foreign Aff. 91, 309–361 (2012). 104. Rodgers, K. S., Jokiel, P. L., Brown, E. K., Hau, S. & Sparks, R. Over a decade of change in spatial and temporal dynamics of Hawaiian coral reef communities. Pacific Sci. 69, 1– 13 (2015). 105. Tebbett, S. B., Goatley, C. H. R. & Bellwood, D. R. The effects of algal turf sediments and organic loads on feeding by coral reef surgeonfishes. PLoS One 12, 1–14 (2017). 106. Lawson, G. L., Kramer, D. L. & Hunte, W. Size-related habitat use and schooling behavior in two species of surgeonfish (Acanthurus bahianus and A. coeruleus) on a fringing reef in Barbados, West Indies. Environmental Biology of Fishes 54, 19–33 (1999). 107. Bellwood, D. R., Streit, R. P., Brandl, S. J. & Tebbett, S. B. The meaning of the term ‘function’ in ecology: a coral reef perspective. Funct. Ecol. 1–14 (2019). doi:10.1111/1365-2435.13265 108. Gross, N. et al. Functional trait diversity maximizes ecosystem multifunctionality. Nat Ecol Evol. 1, 1–24 (2017). 109. Normandin, É., Vereecken, N. J., Buddle, C. M. & Fournier, V. Taxonomic and functional trait diversity of wild bees in different urban settings. PeerJ 2017, 1–35 (2017). 110. Astor, T., von Proschwitz, T., Strengbom, J., Berg, M. P. & Bengtsson, J. Importance of environmental and spatial components for species and trait composition in terrestrial snail communities. J. Biogeogr. 44, 1362–1372 (2017). 111. Wong, M. K. L., Guénard, B. & Lewis, O. T. Trait-based ecology of terrestrial arthropods. Biol. Rev. 94, 999–1022 (2019).

122 112. Madin, J. S. et al. A trait-based approach to advance coral reef science. Trends Ecol. Evol. 31, 419–428 (2016). 113. Darling, E. S. et al. Relationships between structural complexity, coral traits, and reef fish assemblages. Coral Reefs 36, 561–575 (2017). 114. Mouillot, D. et al. Rare species support vulnerable functions in high-diversity ecosystems. PLoS Biol. 11, e1001569 (2013). 115. Villéger, S., Brosse, S., Mouchet, M., Mouillot, D. & Vanni, M. J. Functional ecology of fish : current approaches and future challenges. Aquat. Sci. 79, 783–801 (2017). 116. McLean, M. et al. Trait structure and redundancy determine sensitivity to disturbance in marine fish communities. Glob. Chang. Biol. 25, 3424–3437 (2019). 117. Plass-Johnson, J. G., McQuaid, C. D. & Hill, J. M. Stable isotope analysis indicates a lack of inter- and intra-specific dietary redundancy among ecologically important coral reef fishes. Coral Reefs 32, 429–440 (2013). 118. Fox, R. J. & Bellwood, D. R. Niche partitioning of feeding microhabitats produces a unique function for herbivorous rabbitfishes (Perciformes, Siganidae) on coral reefs. Coral Reefs 32, 13–23 (2013). 119. Brandl, S. J., Robbins, W. D. & Bellwood, D. R. Exploring the nature of ecological specialization in a coral reef fish community: morphology, diet and foraging microhabitat use. Proc. R. Soc. B Biol. Sci. 282, (2015). 120. Barger, C. P., Young, R. C., Will, A., Ito, M. & Kitaysky, A. S. Resource partitioning between sympatric seabird species increases during chick-rearing. Ecosphere 7, 1–15 (2016). 121. Crowell, M. M., Shipley, L. A., Forbey, J. S., Rachlow, J. L. & Kelsey, R. G. Dietary partitioning of toxic leaves and fibrous stems differs between sympatric specialist and generalist mammalian herbivores. J. Mammal. 99, 565–577 (2018). 122. Sato, J. J. et al. Dietary niche partitioning between sympatric wood mouse species (Muridae: Apodemus) revealed by DNA meta-barcoding analysis. J. Mammal. 99, 952– 964 (2018). 123. Mori, E., Ferretti, F. & Fattorini, N. Alien war: ectoparasite load, diet and temporal niche partitioning in a multi-species assembly of small rodents. Biol. Invasions 21, 3305–3318

123 (2019). 124. Kelly, E. L. A. et al. Investigating functional redundancy versus complementarity in Hawaiian herbivorous coral reef fishes. Oecologia 182, 1151–1163 (2016). 125. Bellwood, D. R., Hughes, T. P., Folke, C. & Nyström, M. Confronting the coral reef crisis. Nature 429, 827–33 (2004). 126. Chong-Seng, K. M., Nash, K. L., Bellwood, D. R. & Graham, N. A. J. Macroalgal herbivory on recovering versus degrading coral reefs. Coral Reefs 33, 409–419 (2014). 127. Goatley, C. H. R. & Bellwood, D. R. Biologically mediated sediment fluxes on coral reefs: sediment removal and off-reef transportation by the surgeonfish Ctenochaetus striatus. Mar. Ecol. Prog. Ser. 415, 237–245 (2010). 128. Connell, J. H. Diversity in tropical rain forests and coral reefs. Science (80-. ). 199, 1302– 1310 (1978). 129. Elmqvist, T. et al. Response diversity, ecosystem change, and resilience. Front. Ecol. Environ. 1, 488–494 (2003). 130. Folke, C. et al. Regime shifts, resilience, and biodiversity in ecosystem management. Annu. Rev. Ecol. Evol. Syst. 35, 557–581 (2004). 131. Jackson, J. B. C. What was natural in the coastal oceans? Proc. Natl. Acad. Sci. U. S. A. 98, 5411–5418 (2001). 132. Mumby, P. J. Stratifying herbivore fisheries by habitat to avoid ecosystem overfishing of coral reefs. Fish Fish. 17, 266–278 (2014). 133. Halpern, B. & Floeter, S. Functional diversity responses to changing species richness in reef fish communities. Mar. Ecol. Prog. Ser. 364, 147–156 (2008). 134. Bowen, B. W. et al. Species radiations in the sea: What the flock? J. Hered. 1–14 (2020). doi:10.1093/jhered/esz075 135. Sale, P. F. Maintenance of high diversity in coral reef fish communities. Am. Nat. 111, 337–359 (1977). 136. Burkepile, D. E. & Hay, M. E. Herbivore species richness and feeding complementarity affect community structure and function on a coral reef. Proc. Natl. Acad. Sci. U. S. A. 105, 16201–16206 (2008). 137. Burkepile, D. E. & Hay, M. E. Feeding complementarity versus redundancy among

124 herbivorous fishes on a Caribbean reef. Coral Reefs 30, 351–362 (2011). 138. Burkepile, D. E. & Hay, M. E. Impact of herbivore identity on algal succession and coral growth on a Caribbean reef. PLoS One 5, e8963 (2010). 139. Brandl, S. J. & Bellwood, D. R. Individual-based analyses reveal limited functional overlap in a coral reef fish community. J. Anim. Ecol. 83, 661–670 (2014). 140. Hoey, A. S. Size matters: macroalgal height influences the feeding response of coral reef herbivores. Mar. Ecol. Prog. Ser. 411, 299–302 (2010). 141. Michael, P. J., Hyndes, G. A., Vanderklift, M. A. & Vergés, A. Identity and behaviour of herbivorous fish influence large-scale spatial patterns of macroalgal herbivory in a coral reef. Mar. Ecol. Prog. Ser. 482, 227–240 (2013). 142. Basford, A. J. et al. Feeding habits of range-shifting herbivores: tropical surgeonfishes in a temperate environment. Mar. Freshw. Res. 67, 75–83 (2016). 143. Fox, R. J. & Bellwood, D. R. Direct versus indirect methods of quantifying herbivore grazing impact on a coral reef. Mar. Biol. 154, 325–334 (2008). 144. Choat, J. H., Clements, K. D. & Robbins, W. D. The trophic status of herbivorous fishes on coral reefs, 1: Dietary analyses. Mar. Biol. 140, 613–623 (2002). 145. Oyafuso, Z. S., Toonen, R. J. & Franklin, E. C. Temporal and spatial trends in prey composition of wahoo Acanthocybium solandri: a diet analysis from the central North Pacific Ocean using visual and DNA bar-coding techniques. J. Fish Biol. 88, 1501–1523 (2016). 146. Leray, M., Meyer, C. P. & Mills, S. C. Metabarcoding dietary analysis of coral dwelling predatory fish demonstrates the minor contribution of coral mutualists to their highly partitioned, generalist diet. PeerJ 3, e1047 (2015). 147. Coral Reef Ecosystem Program; Pacific Islands Fisheries Science Center. National Coral Reef Monitoring Program: stratified random surveys of reef fish, including benthic estimate data of the U.S. Pacific Reefs since 2007. NOAA National Centers for Environmental Information. https://inport.nmfs.noaa.gov/inport/item/24447. (2016). 148. Williams, I. D. et al. Human, oceanographic and habitat drivers of Central and Western Pacific coral reef fish assemblages. PLoS One 10, 1–19 (2015). 149. Heenan, A. et al. Data Descriptor: Long-term monitoring of coral reef fish assemblages in

125 the western Central Pacific. Sci. Data 4, (2017). 150. Guidelines for Systematic Review and Evidence Synthesis in Environmental Management. Version 4.2. Collaboration for Environmental Evidence. www.environmentalevidence.org/Documents/Guidelines/Guidelines4.2.pdf. (2013). 151. Boettiger, C., Lang, D. T. & Wainwright, P. C. Rfishbase: Exploring, manipulating and visualizing FishBase data from R. J. Fish Biol. 81, 2030–2039 (2012). 152. Hoey, A. S., Brandl, S. J. & Bellwood, D. R. Diet and cross-shelf distribution of rabbitfishes (f. Siganidae) on the northern Great Barrier Reef: implications for ecosystem function. Coral Reefs 32, 973–984 (2013). 153. Rohatgi, A. WebPlotDigitizer, Version 4.0. https://automeris.io/WebPlotDigitizer/. (2017). 154. Maechler, M. et al. Finding groups in data: cluster analysis extended Rousseeuw et al. R package version 2.1.0. https://svn.r-project.org/R-packages/trunk/cluster. (2019). 155. Debenay, J. P., Sigura, A. & Justine, J. Lou. Foraminifera in the diet of coral reef fish from the lagoon of New Caledonia: predation, digestion, dispersion. Rev. Micropaleontol. 54, 87–103 (2011). 156. Hamilton, S. L., Smith, J. E., Price, N. N. & Sandin, S. A. Quantifying patterns of fish herbivory on Palmyra Atoll (USA), an uninhabited predator-dominated central Pacific coral reef. Mar. Ecol. Prog. Ser. 501, 141–155 (2014). 157. Bellwood, D. R., Hughes, T. P. & Hoey, A. S. Sleeping functional group drives coral-reef recovery. Curr. Biol. 16, 2434–2439 (2006). 158. Hoey, A. S. & Bellwood, D. R. Limited functional redundancy in a high diversity system: single species dominates key ecological process on coral reefs. Ecosystems 12, 1316–1328 (2009). 159. Brandl, S. J. & Bellwood, D. R. Coordinated vigilance provides evidence for direct reciprocity in coral reef fishes. Sci. Rep. 5, 14556 (2015). 160. Goren, M. & Aronov, A. First record of the Indo-Pacific parrotfish Scarus ghobban in the Eastern Mediterranean. Cybium 26, 239–240 (2002). 161. Jones, R. S. Ecological relationships in Hawaiian and Johnston Island Acanthuridae (surgeonfishes). Micronesia 4, 309–361 (1968).

126 162. Kayal, M. et al. Predator crown-of-thorns starfish (Acanthaster planci) outbreak, mass mortality of corals, and cascading effects on reef fish and benthic communities. PLoS One 7, (2012). 163. Bahr, K. D., Jokiel, P. L. & Toonen, R. J. The unnatural history of Kāne‘ohe Bay: coral reef resilience in the face of centuries of anthropogenic impacts. PeerJ 3, e950 (2015). 164. Letourneur, Y., Briand, M. J. & Graham, N. A. J. Coral reef degradation alters the isotopic niche of reef fishes. Mar. Biol. 164, 1–10 (2017). 165. Magurran, A. E., Dornelas, M., Moyes, F., Gotelli, N. J. & McGill, B. Rapid biotic homogenization of marine fish assemblages. Nat. Commun. 6, 8405 (2015). 166. Sala, E. & Knowlton, N. Global marine biodiversity trends. Annu. Rev. Environ. Resour. 31, 93–122 (2006). 167. De’ath, G., Fabricius, K. E., Sweatman, H. & Puotinen, M. The 27 – year decline of coral cover on the Great Barrier Reef and its causes. PNAS 109, 17995–17999 (2012). 168. Alvarez-Filip, L., Paddack, M. J., Collen, B., Robertson, D. R. & Côté, I. M. Simplification of Caribbean reef-fish assemblages over decades of coral reef degradation. PLoS One 10, e0126004 (2015). 169. Devictor, V. & Robert, A. Measuring community responses to large-scale disturbance in conservation biogeography. Divers. Distrib. 15, 122–130 (2009). 170. Couch, C. S. et al. Mass coral bleaching due to unprecedented marine heatwave in Papahānaumokuākea Marine National Monument (Northwestern Hawaiian Islands). PLoS One 12, 1–27 (2017). 171. Munday, P. L. Habitat loss, resource specialization, and extinction on coral reefs. Glob. Chang. Biol. 10, 1642–1647 (2004). 172. Cole, A. J., Pratchett, M. S. & Jones, G. P. Diversity and functional importance of coral- feeding fishes on tropical coral reefs. Fish Fish. 9, 1–22 (2008). 173. Pratchett, M. S. et al. Effects of climate-induced coral bleaching on coral-reef fishes - Ecological and economic consequences. Oceanogr. Mar. Biol. An Annu. Rev. 46, 251–296 (2008). 174. Gratwicke, B. & Speight, M. Effects of habitat complexity on Caribbean marine fish assemblages. Mar. Ecol. Prog. Ser. 292, 301–310 (2005).

127 175. Bellwood, D. R. et al. Coral recovery may not herald the return of fishes on damaged coral reefs. Oecologia 170, 567–73 (2012). 176. Goatley, C. H. R., Bonaldo, R. M., Fox, R. J. & Bellwood, D. R. Sediments and herbivory as sensitive indicators of coral reef degradation. Ecol. Soc. 21, (2016). 177. Brandl, S. J., Emslie, M. J. & Ceccarelli, D. M. Habitat degradation increases functional originality in highly diverse coral reef fish assemblages. Ecosphere 7, (2016). 178. Williams, G. J. & Graham, N. A. J. Rethinking coral reef functional futures. Funct. Ecol. 33, 942–947 (2019). 179. Tebbett, S. B., Goatley, C. H. R. & Bellwood, D. R. Fine sediments suppress detritivory on coral reefs. Mar. Pollut. Bull. 114, 934–940 (2017). 180. Krone, R., Paster, M. & Schuhmacher, H. Effect of the surgeonfish Ctenochaetus striatus (Acanthuridae) on the processes of sediment transport and deposition on a coral reef in the Red Sea. Facies 57, 215–221 (2011). 181. Gordon, S. E., Goatley, C. H. R. & Bellwood, D. R. Low-quality sediments deter grazing by the parrotfish Scarus rivulatus on inner-shelf reefs. Coral Reefs 35, 285–291 (2016). 182. Coral Reef Ecosystem Program. Pacific Islands Fisheries Science Center. National Coral Reef Monitoring Program: stratified random surveys (StRS) of reef fish, including benthic estimate data of the U.S. Pacific Reefs since 2007. NOAA National Center. (2016). 183. Rao, C. R. Diversity and Dissimilarity Coefficients - A Unified Approach. Theor. Popul. Biol. 21, 24–43 (1982). 184. Ricotta, C. & Szeidl, L. Diversity partitioning of Rao’s quadratic entropy. Theor. Popul. Biol. 76, 299–302 (2009). 185. Debastiani V., P. V. SYNCSA - R tool for analysis of metacommunities based on functional traits and phylogeny of the community components. Bioinformatics 28, 2067– 2068 (2012). 186. Halpern, B. S. et al. Spatial and temporal changes in cumulative human impacts on the world’s ocean. Nat. Commun. 6, 7615 (2015). 187. McDole Somera, T. et al. Energetic differences between bacterioplankton trophic groups and coral reef resistance. Proc. R. Soc. B Biol. Sci. 283, 20160467 (2016). 188. Lefcheck, J. S. piecewiseSEM: Piecewise structural equation modelling in R for ecology,

128 evolution, and systematics. Methods Ecol. Evol. 7, 573–579 (2016). 189. Barton, K. MuMin: Multi-Model Inference. R package version 1.40.0 http://CRAN.R- project.org/package=MuMin. (2017). 190. Williams, G. J., Gove, J. M., Eynaud, Y., Zgliczynski, B. J. & Sandin, S. A. Local human impacts decouple natural biophysical relationships on Pacific coral reefs. Ecography (Cop.). 38, 751–761 (2015). 191. Maire, E. et al. Community-wide scan identifies fish species associated with coral reef services across the Indo-Pacific. Proc. R. Soc. B Biol. Sci. 285, (2018). 192. Rahel, F. J. Homogenization of fish fauna across the United States. Science (80-. ). 288, 854–856 (2000). 193. Halpern, B. S. et al. An index to assess the health and benefits of the global ocean. Nature (2012). doi:10.1038/nature11397 194. McDole, T. et al. Assessing coral reefs on a Pacific-wide scale using the microbialization score. PLoS One 7, e43233 (2012). 195. Athens, J. S., Ward, J. V. & Blinn, D. W. Vegetation history of Laysan Island, Northwestern Hawaiian Islands. Pacific Sci. 61, 17–37 (2007). 196. Boyle, S. et al. Bridging science and management in the Pacific Remote Islands Marine National Monument. (2017). 197. Papahānaumokuākea Marine National Monument. Maritime heritage research, education, and management plan: Papahānaumokuākea Marine National Monument. Honolulu, Hawai‘i. 97 (2011). 198. Friedlander, A. M. et al. Human-induced gradients of reef fish declines in the Hawaiian Archipelago viewed through the lens of traditional management boundaries. Aquat. Conserv. Mar. Freshw. Ecosyst. 28, 146–157 (2018). 199. Mora, C. et al. Global human footprint on the linkage between biodiversity and ecosystem functioning in reef fishes. PLoS Biol. 9, e1000606 (2011). 200. Brewer, T. D., Cinner, J. E., Fisher, R., Green, A. & Wilson, S. K. Market access, population density, and socioeconomic development explain diversity and functional group biomass of coral reef fish assemblages. Glob. Environ. Chang. 22, 399–406 (2012). 201. DeMartini, E. E., Friedlander, A. M., Sandin, S. A. & Sala, E. Differences in fish-

129 assemblage structure between fished and unfished atolls in the northern Line Islands, central Pacific. Mar. Ecol. Prog. Ser. 365, 199–215 (2008). 202. Williams, I. D. et al. Human, oceanographic and habitat drivers of Central and Western Pacific coral reef fish assemblages. PLoS One 10, e0120516 (2015). 203. Heenan, A., Hoey, A. s., Williams, G. J. & Williams, I. D. Natural bounds on herbivorous coral reef fishes. Proc. R. Soc. B 283, (2016). 204. Helyer, J. & Samhouri, J. F. Fishing and environmental influences on estimates of unfished herbivorous fish biomass across the Hawaiian Archipelago. Mar. Ecol. Prog. Ser. (2017). doi:10.3354/meps12235 205. Lefévre, C. D. & Bellwood, D. R. Temporal variation in coral reef ecosystem processes: Herbivory of macroalgae by fishes. Mar. Ecol. Prog. Ser. 422, 239–251 (2011). 206. Loffler, Z., Bellwood, D. R. & Hoey, A. S. Associations among coral reef macroalgae influence feeding by herbivorous fishes. Coral Reefs 34, 51–55 (2015). 207. McClanahan, T. R., Hendrick, V., Rodrigues, M. J. & Polunin, N. V. C. Varying responses of herbivorous and invertebrate-feeding fishes to macroalgal reduction on a coral reef. Coral Reefs (1999). doi:10.1007/s003380050181 208. Hoey, A. S. & Bellwood, D. R. Suppression of herbivory by macroalgal density: A critical feedback on coral reefs? Ecol. Lett. 14, 267–273 (2011). 209. Chong-Seng, K. M., Mannering, T. D., Pratchett, M. S., Bellwood, D. R. & Graham, N. A. J. J. The influence of coral reef benthic condition on associated fish assemblages. PLoS One 7, e42167 (2012). 210. Bauman, A. G. et al. Macroalgal browsing on a heavily degraded, urbanized equatorial reef system. Sci. Rep. 7, 1–8 (2017). 211. Selkoe, K. A. et al. The DNA of coral reef biodiversity: predicting and protecting genetic diversity of reef assemblages. Proc. R. Soc. B 283, (2016). 212. Alvarez-Filip, L., Dulvy, N. K., Gill, J. A., Watkinson, A. R. & Co, I. M. Flattening of Caribbean coral reefs: region-wide declines in architectural complexity. Proc. R. Soc. B 276, 3019–3025 (2009). 213. Worm, B. et al. Impacts of biodiversity loss on ocean ecosystem services. Science (80-. ). 314, 787–790 (2006).

130 214. Meyer, C. G., Papastamatiou, Y. P. & Clark, T. B. Differential movement patterns and site fidelity among trophic groups of reef fishes in a Hawaiian marine protected area. Mar. Biol. 157, 1499–1511 (2010). 215. Liem, F. Adaptive significance of intra- and interspecific differences in the feeding repertoires of cichlid fishes. Am. Zool. 20, 295–314 (1980). 216. Robinson, B. W. & Wilson, D. S. Optimal foraging, specialization, and a solution to Liem’s paradox. Am. Nat. 151, 223–235 (1998). 217. Rasher, D. B., Hoey, A. S. & Hay, M. E. Consumer diversity interacts with prey defenses to drive ecosystem function. Ecology 94, 1347–58 (2013). 218. Cheal, A., Emslie, M., Miller, I. & Sweatman, H. The distribution of herbivorous fishes on the Great Barrier Reef. Mar. Biol. 159, 1143–1154 (2012). 219. Tyler, E., Speight, M., Henderson, P. & Manica, a. Evidence for a depth refuge effect in artisanal coral reef fisheries. Biol. Conserv. 142, 652–667 (2009). 220. Bejarano, S., Golbuu, Y., Sapolu, T. & Mumby, P. Ecological risk and the exploitation of herbivorous reef fish across Micronesia. Mar. Ecol. Prog. Ser. 482, 197–215 (2013). 221. State of Hawaiʻi Department of Land and Natural Resources. Report on the findings and recommendations of effectiveness of the West Hawai’i Regional Fishery Management Area. (2014). 222. Eagle, J. V & Jones, G. P. Mimicry in coral reef fishes: ecological and behavioural responses of a mimic to its model. J. Zool. 264, 33–43 (2004). 223. Rainey, M. M. Evidence of a geographically variable competitive mimicry relationship in coral reef fishes. J. Zool. 279, 78–85 (2009). 224. Choat, J. H., Robbins, W. D. & Clements, K. D. The trophic status of herbivorous fishes on coral reefs II . Food processing modes and trophodynamics. Mar. Biol. 145, 445–454 (2004). 225. Jones, G. P., McCormick, M. I., Srinivasan, M. & Eagle, J. V. Coral decline threatens fish biodiversity in marine reserves. Proc. Natl. Acad. Sci. (2004). doi:10.1073/pnas.0401277101 226. Randall, J. E. A contribution to the biology of the convict surgeonfish of the Hawaiian Islands, Acanthurus triostegus sandoicensis. Pacific Sci. 15, 215–272 (1961).

131 227. Huisman, J. M., Abbott, I. A. & Smith, C. M. Hawaiian Reef Plants. (University of Hawaiʻi Sea Grant College Program, 2007). 228. Hobbs, J.-P. a., Jones, G. P. & Munday, P. L. Rarity and extinction risk in coral reef angelfishes on isolated islands: interrelationships among abundance, geographic range size and specialisation. Coral Reefs 29, 1–11 (2009). 229. Clua, E. & Legendre, P. Shifting dominance among Scarid species on reefs representing. Aquat. Living Resour. 21, 339–348 (2008). 230. Choat, J. H., Axe, L. M. & Lou, D. C. Growth and longevity in fishes of the family Scaridae. Mar. Ecol. Prog. Ser. 145, 33–41 (1996). 231. Lobban, C. S., Schefter, M. & Donaldson, T. J. Cluster dynamics in Maristentor dinoferus, a gregarious benthic ciliate with zooxanthellae and a hypericin-like pigment, in relation to biofilm grazing by the fish Ctenochaetus striatus. Symbiosis 63, 137–147 (2014). 232. Hobson, S. Feeding relationship of teleostean fishes on coral reefs in Kona, Hawaii. Fish. Bull. 72, 915–1031 (1974). 233. Fitzpatrick, B. M., Harvey, E. S., Heyward, A. J., Twiggs, E. J. & Colquhoun, J. Habitat specialization in tropical continental shelf demersal fish assemblages. PLoS One 7, e39634 (2012). 234. Clausing, R. J. et al. Experimental evidence of dietary ciguatoxin accumulation in an herbivorous coral reef fish. Aquat. Toxicol. 200, 257–265 (2018). 235. Gaboriau, M., Ponton, D., Darius, H. T. & Chinain, M. Ciguatera fish toxicity in French Polynesia: Size does not always matter. Toxicon 84, 41–50 (2014). 236. Pennings, S., Puglisi, M., Pitlik, T., Himaya, A. & Paul, V. Effects of secondary metabolites and CaCO3 on feeding by surgeonfishes and parrotfishes: within-plant comparisons. Mar. Ecol. Prog. Ser. 134, 49–58 (1996). 237. Tissot, B. N. Adaptive management of aquarium fish collecting in Hawaii. SPC Live Reef Fish Inf. Bull. 6, 16–19 (1999). 238. Tupper, M. H. Spillover of commercially valuable reef fishes from marine protected areas in Guam, Micronesia. Fish. Bull. 105, 527–537 (2007). 239. Lindfield, S. J., McIlwain, J. L. & Harvey, E. S. Depth refuge and the impacts of SCUBA spearfishing on coral reef fishes. PLoS One 9, e92628 (2014).

132 240. Nadon, M. O., Ault, J. S., Williams, I. D., Smith, S. G. & DiNardo, G. T. Length-based assessment of coral reef fish populations in the Main and Northwestern Hawaiian Islands. PLoS One 10, e0133960 (2015). 241. Sabater, M. G. & Tofaeono, S. P. Scale and benthic composition effects on biomass and trophic group distribution of reef fishes in American Samoa. Pacific Sci. 61, 503–520 (2007). 242. Lumbang, W. A. & Paul, V. J. Chemical defenses of the tropical green seaweed Neomeris annulata Dickie: effects of multiple compounds on feeding by herbivores. J. Exp. Mar. Bio. Ecol. 201, 185–195 (1996). 243. Paul, V., Nelson, S. & Sanger, H. Feeding preferences of adult and juvenile rabbitfish Siganus argenteus in relation to chemical defenses of tropical seaweeds. Mar. Ecol. Prog. Ser. 60, 23–34 (1990). 244. Bryan, P. G. Food habits, functional digestive morphology, and assimilation efficiency of the rabbitfish Siganus spinus (Pisces, Siganidae) on Guam. Pacific Sci. 29, 269–277 (1975). 245. Paul, V. J., Meyer, K. D., Nelson, S. G. & Sanger, H. R. Deterrent effects of seaweed extracts and secondary metabolites on feeding by the rabbitfish siganus spinus. Proc. Seventh Int. Coral Reef Symp. 2, 1–8 (1992). 246. Guiasu, R. C. & Winterbottom, R. Yellow juvenile color pattern, diet switching and the phylogeny of the surgeonfish genus Zebrasoma (Percomorpha, Acanthuridae). Bull. Mar. Sci. 63, 277–294 (1998).

133