Species-Habitat Relationships and Community Structure of Reef Fishes Associated with Temperate Hardbottom Reefs of North Carolina, Usa

SPECIES-HABITAT RELATIONSHIPS AND COMMUNITY STRUCTURE OF REEF FISHES ASSOCIATED WITH TEMPERATE HARDBOTTOM REEFS OF NORTH CAROLINA, USA

Avery Byrd Paxton

A dissertation submitted to the faculty at the University of North Carolina at Chapel Hill in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Biology in the College of Arts and Sciences.

Chapel Hill 2018

Approved by:

Charles H. Peterson

J. Christopher Taylor

Stephen R. Fegley

Johanna H. Rosman

F. Joel Fodrie

ii ABSTRACT

Avery Byrd Paxton: Species-Habitat Relationships and Community Structure of Reef Fishes Associated with Temperate Hardbottom Reefs of North Carolina, USA (Under the direction of Charles H. Peterson)

Numbers of human-made, artificial structures in coastal oceans are increasing.

Humans deploy these artificial structures for a variety of purposes, including to protect shorelines, harness energy resources, create and restore habitat, and foster tourism, fishing, and diving opportunities. Because the introduction of artificial structures to coastal oceans has the potential to drive ecological change, understanding how species use these artificial structures as habitat is important. Along the southeastern USA continental shelf, shipwrecks and intentionally sunk artificial reefs provide an opportunity to determine how reef fish communities rely on artificial structures, especially in comparison to naturally-occurring rocky reefs. Here, I investigated five applied research questions on artificial reefs, shipwrecks, and rocky reefs of North Carolina, USA using methods ranging from diver- conducted surveys and fisheries acoustics to time-lapse videography and audio recordings.

First, I tested how reefs that vary in topographic complexity function to support reef fishes

(Chapter 1). I discovered that flat reefs, which are often difficult to detect, provide similar support for reef fishes as more easily detectable complex, high-relief reefs. Second, I examined how reefs support fishes with different thermal affinities and determined that because tropical and subtropical fishes occurred in higher abundances on artificial structures

iii than rocky reefs that deploying additional human-made reefs may help warm-water fishes move poleward from the tropics (Chapter 2). Third, I examined spatial relationships between planktivorous fishes, zooplankton (prey), and piscivorous fishes (predators) around artificial structures (Chapter 3). I found that aggregations of planktivorous fishes around artificial structures related to spatial patterns across adjacent trophic levels, suggesting that artificial structures influence multiple trophic levels. Fourth, I documented how reef fishes reacted to underwater noises typically emitted when searching for oil and gas beneath the seafloor, finding that fish abundance decreased 78% during evening hours when exposed to these loud noises, raising conservation concerns (Chapter 4). Fifth, I assessed how quickly newly established artificial reefs create fish habitat and discovered that newly deployed human- made reefs can provide fish habitat comparable to 20-year old artificial reefs within five months (Chapter 5). Taken together, these results demonstrate that differences exist between fish communities residing on artificial structures versus natural reefs (Chapters 1 & 2). These differences are driven largely by higher abundances of planktivorous fishes (Chapter 3), as well as tropical and subtropical highly-reef associated species, on artificial structures than on natural reefs (Chapter 2). When considering future installation of human-made structures, artificial reefs can quickly establish fish habitat comparable to previously established artificial structures (Chapter 5), yet effects from exploration prior to installation of these structures, such as underwater noises, can disrupt how fish use reefs (Chapter 4). Artificial structures clearly serve as fish habitat, and as artificial structures continue to be deployed, conservation and management efforts must recognize that species-habitat relationships differ between artificial structures and naturally occurring rocky reefs and that artificial structures can have both ecological benefits and impacts.

iv To my family and friends who inspired and fostered my love for the ocean.

ACKNOWLEDGEMENTS

Six years ago, I was told that I was crazy to want to study reefs off the coast of North

Carolina (NC), USA as a graduate student. I was told that these reefs were too expensive to study and too difficult to access, so this line of research was too ambitious and likely to fail.

At present, I have generated a body of research on fish communities of offshore reefs of NC, like the one that was reportedly unachievable. Countless individuals helped make this research a reality. Thank you all for the diversity of support that you have provided, from conceptualization through data analyses and interpretation. It has been a team effort, and together we have made significant advancements in our understanding of the unique reefs and associated fish communities that exist off the coast of NC.

My dissertation advisor, Pete Peterson, has provided unwavering support for my research. Instead of dissuading me from pursuing offshore reef research, Pete encouraged me to try harder. He learned about NC reefs along with me, as he taught me to develop testable research questions, how to write scientifically, and so much more. Pete has been one of my biggest advocates these past six years, and I am grateful for everything that he has taught me.

To my other committee members – thank you for challenging me to improve my research, develop testable questions, and to write and speak succinctly. Chris Taylor has been an unflagging source of encouragement, teaching me the minutia of fisheries acoustics and other remote sensing methods. Most importantly, Chris has grounded my research by helping me learn how to conduct and then communicate results from applied science. His patience with

vi my endless questions has been unparalleled. Thank you for being a mentor and friend, providing insightful advice, and endless opportunities. Steve Fegley has provided countless statistical consultations, complete with laughter about ‘squirrel-syndrome.’ Steve’s attention to detail yet ability to see the bigger picture has been a guiding light. Johanna Rosman has provided thoughtful encouragement and a grounding in reality. She has taught me about physical oceanography and the intersection of physical oceanography and ecology. Joel

Fodrie has provided enthusiastic support since the day that I accepted to UNC. Thank you,

Joel, for your guidance and advice. Allen Hurlbert welcomed me to his lab meetings on main campus and since then, has helped me to grow as an ecologist, challenging me to frame my applied research into the broader field of ecology. Thank you, Allen, for the multiple remote meetings and valuable broader advice.

To Emily Pickering, Alyssa Adler, Hayley Lemoine, Claire Rosemond, and Rebecca

Gaesser – I am so thankful to have had the opportunity to work with you. Each of you has been instrumental in this body of research. Your constant optimism, endless energy, and friendship have been an inspiration, and I look forward to continuing to collaborate with each of you in the future.

To boat captains J. Purifoy, K. Johns, G. Compeau, S. Hall, D. Wells, J. Styron, R.

Purifoy, B. Wilde, T. Leonard, and their crew - thank you for safely transporting our team to and from reefs. To those who provided boating support, including C. Lewis, S. Davis, E.

Kromka, P. Herbst, W. Fluellen, crew and staff from Olympus Dive Center, and crew and staff from Discovery Diving, thank you for facilitating our safety and our research. To diving safety officers G. Safrit, K. Johns, and B. Degan – thank you for your oversight, advice, and attention to safety. Thank you to the officers and crew of the NOAA ship Nancy Foster, as

vii well for their attention to safety, and to J. McCord, D. Sybert, and M.L. Parker who provided underwater and topside videos and photos, helping us share our science.

This research would not have been possible without support from scientific divers including A. Adler, E. Pickering, H. Lemoine, C. Rosemond, R. Gaesser, L. Revels, G.

Safrit, K. Johns, B. Degan, G. Sorg, J. Fleming, T. Courtney, M. Kenworthy, A. Poray, D.

Keller, I. Kroll, C. Hamilton, J. Hughes, J. Boulton, T. Dodson, E. Ebert, J. Vaner Pluym, B.

Teer, J. Hackney, R. Munoz, R. Mays, D.W. Freshwater, M. Dionesotes, C. Marino, I. Conti-

Jerpe, E. Weston, M. Wooster, L. Bullock. A. Pickett, J. Geyer, A. Rok, T. Dodson, J.

Styron, D. Wells, S. Hall, J. McCord, and D. Sybert. To each of you, thank you for your contributions. You were the essence of our team.

To undergraduate researchers – K. McCormick, C. Peters, R. Granzotti, T. Oruganti,

S. Richardson, P. Oliveira, Y. Azevedo, K. Wiedbusch, L. Revels, D. Rouse, A. Requarth, and R. Snider – thank you for your enthusiasm and dedication. I enjoyed having you on our team and learning along with you. Thank you also to high school students R. Condra and T.

Buck and to volunteers B. Langdale and O. Newton.

To E. Ebert, C. Taylor, F. Campanella, T. Jarvis, and B. Scoulding – thank you for teaching me fisheries acoustics and for assistance with data processing. To B. Degan and E.

Ebert – thank you for assistance with video processing. To T. Casserly, E. Ebert, and officers and crew of NOAA ship Nancy Foster – thank you for assistance with fisheries acoustics fieldwork. To C. Buckel – thank you for teaching me how to identify invertebrates and macroalgae, how to construct a database, and much more. To Wilson Freshwater for his enthusiasm, teaching me how to identify macroalgae, for diving support, and fun brainstorming sessions – thank you! To P. Whitfield – thank you for teaching me about

viii natural rocky reefs of NC. To A. Poray – thank you for teaching me how to identify macroalage. To S. Fegley, J. Weiss, D. Urban, L. Yeager, J. Byrnes, J. Hench, and S.

Viehman – thank you for statistical guidance. To J. Francesconi, G. Bodnar, J. Peters, C.

Jensen, C. Weychert, and others at DMF – thank you for your assistance and guidance.

Special thank you to K. Irish for teaching me to communicate my science, being my constant cheerleader, and for providing such wonderful advice, mentorship, and friendship.

Thank you to R. Leuttich for support in securing boating resources and advice in tricky situations. Thank you to J. Stack, R. Smith, M. Connor, D. Napier, and K. Wood for administrative support and for your patience in helping me with budgeting. To the rest of the

IMS ‘family,’ thank you for making my time at IMS so enjoyable. Special thanks to M.

Broduer, K. Lauer, C. White, D. Keller, K. Onorevole, K. Augustine, C. Payne, and so many more for friendship. Thank you to C. Smith, R. Gittman, J. Morton, and others in the

Peterson lab past and present for their support. Thank you to C. Voss for encouragement and for helping me learn the ropes. To B. McDermott, R.D. Price, and P. Murphy – thank you for inspiring my love for NC shipwrecks and my attention to safety. To D. Smith and T. Dwyer – thank you for your encouragement.

Thank you to coauthors of manuscripts resulting from my dissertation research, including E. Pickering, A. Adler, J.C. Taylor, C.H. Peterson, C.M. Voss, D. Nowacek, E.

Cole, J. Dale, S. Fegley, J. Rosman, L. Revels, H. Lemoine, and R. Rosemond. Thank you to reviewers, including G. Kellison, C. Schobernd, S. Brandl, H. Patterson, R. Munoz, N.

Bacheler, D. Gruccio, committee members, and anonymous reviewers for thoughtful feedback on manuscripts resulting from this research.

ix Funding was provided by BOEM under Cooperative Agreement M13AC00006, NC

Coastal Recreational Fishing License Grants (#5115 and #6446), a NSF Graduate Research

Fellowship awarded to A.B. Paxton under Grant No. DGE-1144081, a P.E.O. Scholar Award to A.B. Paxton, a Carol and Edward Smithwick Dissertation Fellowship awarded to A.B.

Paxton through the UNC Royster Society of Fellows, NOAA National Ocean Service and

National Centers for Coastal Ocean Science. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the opinions or policies of the US Government, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.

And saving the best for last - to my family and closest friends, thank you for encouraging me at every step of the way. Your love and support mean the world to me.

x TABLE OF CONTENTS

LIST OF TABLES ...... xv

LIST OF FIGURES ...... xvi

LIST OF VIDEOS ...... xix

LIST OF AUDIO ...... xx

LIST OF SUPPORTING TEXT ...... xxi

CHAPTER 1: FLAT AND COMPLEX TEMPERAE REEFS PROVIDE SIMILAR SUPPORT FOR FISH: EVIDENCE FOR A UNIMODAL SPECIES-HABITAT RELATIONSHIP ...... 1

Summary ...... 1

Introduction ...... 2

Materials and Methods ...... 6

Survey Sites ...... 6

Fish Community Assessments ...... 7

Structural Complexity ...... 8

Water Temperature ...... 10

Sediment Cover ...... 11

Statistical Analyses ...... 11

Results ...... 15

Discussion ...... 19

Acknowledgements ...... 26

Figures...... 28

Tables ...... 34

xi CHAPTER 2: ARTIFICIAL STRUCUTRES HOST HIGHER ABUNDNACES OF TROPICAL AND SUBTROPICAL FISHES BUT LOWER ABUNDANCES OF TEMPERATE FISHES THAN ROCKY REEFS OF NORTH CAROLINA, USA ...... 36

Summary ...... 36

Introduction ...... 37

Materials and Methods ...... 42

Survey Sites ...... 42

Fish Community...... 43

Benthic Community ...... 44

Reef Topography ...... 45

Water Temperature ...... 46

Statistical Analyses ...... 46

Results ...... 50

Discussion ...... 55

Acknowledgements ...... 61

Figures...... 63

Tables ...... 68

CHAPTER 3: AGGREGATIONS OF PLANKTIVOROUS FISHES AROUND SHIPWRECKS RELATE TO SPATIAL PATTERNS IN ADJACENT TROPHIC LEVELS ...... 70

Summary ...... 70

Introduction ...... 71

Materials and Methods ...... 75

Site Selection ...... 75

Surveys Conducted ...... 76

Mapping Shipwrecks ...... 76

Quantifying Fishes & Zooplankton ...... 77

xii Quantifying Water Currents ...... 79

Data Analyses ...... 80

Results ...... 83

Spatial Location ...... 83

Spatial Relationships ...... 86

Water Current...... 87

Discussion ...... 88

Acknowledgements ...... 94

Figures...... 95

Tables ...... 100

CHAPTER 4: SEISMIC SURVEY NOISE DISRUPTED FISH USE OF A TEMPERATE REEF ...... 102

Summary ...... 102

Introduction ...... 103

Materials and Methods ...... 105

Results and Discussion ...... 107

Conclusion ...... 110

Acknowledgements ...... 111

Figures...... 112

CHAPTER 5: CONVERGENCE OF FISH COMMUNITY STRUCTURE BETWEEN A NEWLY DEPLOYED AND AN ESTABLISHED ARTIFICIAL REEF ALONG A FIVE-MONTH TRAJECTORY ...... 115

Summary ...... 115

Introduction ...... 116

Materials and Methods ...... 118

Results ...... 124

xiii Discussion ...... 130

Acknowledgements ...... 134

Figures...... 135

APPENDIX 1: SUPPORTING INFORMATION FOR CHAPTER 1 ...... 139

APPENDIX 2: SUPPORTING INFORMATION FOR CHAPTER 2 ...... 151

APPENDIX 3: SUPPORTING INFORMATION FOR CHAPTER 3 ...... 168

APPENDIX 4: SUPPORTING INFORMATION FOR CHAPTER 4 ...... 191

APPENDIX 5: SUPPORTING INFORMATION FOR CHAPTER 5 ...... 195

REFERENCES ...... 198

xiv LIST OF TABLES

Table 1.1. GLM results for the relationship between fish community metrics (abundance, biomass, richness) and environmental predictor variables by reef type...... 34 Table 2.1. Linear mixed effects model results for fish abundance by climate range...... 68 Table 2.2. Linear mixed effects model results for percent cover of the benthic community by climate zone...... 69 Table 3.1. Definitions and equations for indicators to quantify spatial distributions of individual groups of organisms and spatial relationships between organism pairs...... 100 Table 3.2. Spatial indicators (mean ± standard error) for distributions of zooplankton and fishes around shipwrecks...... 101 Table S1.1. Descriptions of thirty reefs surveyed...... 141 Table S1.2. Species list from 246 fish belt-transects conducted on warm-temperate reefs of the NC continental shelf...... 144 Table S1.3. GLM results for the relationship between fish abundance and environmental predictor variables by reef type and fish size class...... 149 Table S2.1. Descriptions of thirty reefs surveyed...... 153 Table S2.2. Fish, shark, and turtle species list from 226 fish belt-transects conducted on warm-temperate reefs of the NC continental shelf...... 155 Table S2.3. Benthic invertebrate and macroalgae species list from 226 photoquadrat benthic photoquadrats collected on warm-temperate reefs of the NC continental shelf...... 161 Table S2.4. Linear mixed effects model results for fish biomass by climate zone...... 165 Table S2.5. Fishes not commonly reported as far north as surveyed reefs...... 166 Table S3.1. Descriptions of fifteen shipwrecks surveyed...... 188 Table S3.2. Descriptions of twenty surveys across fifteen shipwrecks...... 190 Table S4.1. Fish species list from 140 videos recorded on the natural rocky reef three days before and one day during seismic surveying...... 193 Table S5.1. Species list for fishes observed on the new and established reef...... 196

xv LIST OF FIGURES

Figure 1.1. Thirty temperate reefs, including natural (blue circles) and artificial (red triangles) reefs, surveyed on the continental shelf of NC. Point size is proportional to mean digital reef rugosity (DRR) from transects on the particular reef...... 28 Figure 1.2. Habitat complexity of temperate reefs. a-d) Representative images of temperate reef morphologies. e-h) Representative depth contours of each reef morphology along the surveyed transect length. i-l) Representative semivariograms of each reef for half the distance of the surveyed transect length...... 29 Figure 1.3. Relationship between digital reef rugosity (DRR) and fish community metrics on natural (blue) and artificial (red) temperate reefs...... 31 Figure 1.4. Fish community metrics by morphological category for natural reefs (blue; N pavement&rubble = 38, N ledge = 29) and artificial reefs (red; N concrete = 17, N ship = 39)...... 32 Figure 1.5. Biplot of nonmetric multidimensional scaling (nMDS) ordination for fish community at the family level overlaid with indicators of reef morphologies...... 33 Figure 2.1. Thirty warm-temperate reefs surveyed on the inner continental shelf...... 63 Figure 2.2. Abundance of demersal fishes (a-c) and pelagic fishes (d-e) (per 120 m 2) on artificial (dark colored) versus natural (light colored) reefs by fish climate range: (a) temperate (blue), (b) subtropical (green), (c) tropical (red)...... 64 Figure 2.3. Nonmetric multidimensional scaling ordination of demersal fish community by climate range...... 65 Figure 2.4. Mean log (abundance + 0.01) (per 120 m 2) of fishes not commonly documented as far north as surveyed reefs for artificial and natural reefs...... 66 Figure 2.5. Multigroup structural equation model (SEM) for fish response to abiotic and benthic variables...... 67 Figure 3.1. Locations of fifteen shipwrecks surveyed on the continental shelf of NC. Gray lines and corresponding text indicate water depth in 10 m increments...... 95 Figure 3.2. Distances between the mean center of (a) small fishes, (b) medium fishes, and (c) large fishes and the nearest shipwreck edge for each survey...... 96 Figure 3.3. Spatial clusters of (a) zooplankton, (b) small fishes, (c) medium fishes, and (d) large fishes around the U-352 shipwreck...... 97

xvi Figure 3.4. Box plots describing spatial relationships between pairwise groupings of zooplankton, small fishes, medium fishes, and large fishes...... 98 Figure 3.5: a-c) Relationships between current magnitude and distances of a) small fishes, b) medium fishes, and c) large fishes from shipwreck, d) Relationship between number of predators on shipwrecks and within 5 m outward of shipwrecks...... 99 Figure 4.1. Track of seismic survey vessel (black line) relative to three monitoring reefs on the inner continental shelf of NC: two outfitted with hydrophones (blue triangles) and one with video camera (orange square)...... 112 Figure 4.2. Acoustic signatures of A) ambient noise and B-D) noise from seismic airgun shots on reef 0.7 km from closest approach of seismic surveying vessel: B) 22.2 km from reef before closest approach; C) 0.7 km from reef showing the seismic shots just prior to shots that overloaded our instruments; D) 19.6 km from reef following closest approach. Insets depict 10 Hz – 5 kHz range of low frequency...... 113 Figure 4.3. Hourly fish abundance on the reef 7.9 km from the closest approach of the seismic survey ship during three days before (solid black line) and on one day during the height of seismic activity near the reef (red line)...... 114 Figure 5.1. Experimental design to quantify fish community change over time on a newly deployed and established artificial reef...... 135 Figure 5.2. Mean fish community metrics on newly deployed (gray) versus nearby established (black) artificial reef by sampling period for a) abundance, b) species richness, and c) Pielou’s evenness...... 136 Figure 5.3. Nonmetric multidimensional scaling ordination of fish community composition on newly deployed reef (gray) and established reef (black) during daytime hours. a) sampling period 1 (May 17-26), b) sampling period 2 (July 21 – August 5), and c) sampling period 3 (September 13-19)...... 137 Figure 5.4. Fish species abundance (mean ± 1 SE) on new reef (gray bars) and on established reef (black bars) during three sampling periods (May, July, September) for: a) Archosargus probatocephalus , b) Centropristis striata , c) Decapterus sp., and d) Haemulon aurolineatum ...... 138 Figure S1.1. Response of fish abundance to digital reef rugosity (DRR) by reef type and fish size class...... 139 Figure S2.1. Biomass of demersal tropical fishes on artificial (dark colored) versus natural (light colored) reefs by reef depth and season...... 151

xvii Figure S2.2. Percent cover of a-c) overall benthos, d-f) benthos comprised of benthic invertebrates, g-i) benthos comprised of macroalgae on artificial (dark colored) versus natural (light colored) reefs by organism climate range a, d, g) temperate (blue), b, e, h) subtropical (green), c, f, i) tropical (red)...... 152 Figure S3.1. Bathymetric maps of four surveyed shipwrecks: a) U-352 , b) USS Schurz, c) USS Tarpon, and d) W.E. Hutton ...... 184 Figure S3.2. Spatial location of small fishes (blue; planktivorous fishes), medium fishes (orange; piscivorous fishes), and large fishes (red; piscivorous fishes) relative to four shipwrecks (black polygon): a) U-352 , b) HMT Bedfordshire , c) W.E. Hutton , and d) USS Schurz ...... 185 Figure S3.3. Spatial clusters of zooplankton, small fishes, medium fishes, and large fishes relative to each of four shipwrecks (black): a) USS Schurz, b) USS Tarpon, c) Proteus , and d) Merak ...... 186 Figure S3.4. Bootstrapped metrics describing spatial relationships between pairs of organisms...... 187 Figure S4.1. Hourly time series of fish abundance on natural rocky reef on four separate days: A) September 17, 2014; B) September 18, 2014; C) September 19, 2014; D) September 20, 2014...... 191 Figure S4.2. Test of equality of variance in fish counts on three days pre-seismic surveying and one day during seismic surveying, based on analysis of means for variance (ANOMV) with Levene transformation...... 192 Figure S5.1. Nonmetric multidimensional scaling ordination of fish community on the established reef (USS Indra ) prior to the deployment of the new artificial reef nearby...... 195

xviii LIST OF VIDEOS

Video S4.1 Video recording from reef located 7.9 km from closest approach of the seismic surveying vessel during the evening one day prior to seismic surveying on the inner continental shelf...... 194 Video S4.2 Video recording from reef located 7.9 km from closest approach of the seismic surveying vessel during active seismic surveying on the inner continental shelf...... 194 Video S5.1. Underwater video recording from the new artificial reef (James J. Francesconi ) during the first sampling period in May 2016...... 197 Video S5.2. Underwater video recording from the established artificial reef (USS Indra ) during the first sampling period in May 2016...... 197 Video S5.3. Underwater video recording from the new artificial reef (James J. Francesconi ) during the third sampling period in September 2016...... 197 Video S5.4. Underwater video recording from the established reef (USS Indra ) during the third sampling period in September 2016...... 197

xix LIST OF AUDIO

Audio S4.1 Audio recording from reef located 0.7 km from the closest approach of the seismic surveying vessel prior to the seismic surveying on the inner continental shelf...... 194 Audio S4.2 Audio recording from reef located 0.7 km from the closest approach of the seismic surveying vessel during active seismic surveying on the inner continental shelf...... 194

xx LIST OF SUPPORTING TEXT

Text S3.1: Detailed methods for quantifying fishes and zooplankton...... 168 Text S3.2: Detailed methods for quantifying water currents...... 174 Text S3.3: Detailed methods for data analyses...... 176

xxi CHAPTER 1: FLAT AND COMPLEX TEMPERAE REEFS PROVIDE SIMILAR SUPPORT FOR FISH: EVIDENCE FOR A UNIMODAL SPECIES-HABITAT RELATIONSHIP 1

Summary

Structural complexity, a form of habitat heterogeneity, influences the structure and function of ecological communities, generally supporting increased species density, richness, and diversity. Recent research, however, suggests the most complex habitats may not harbor the highest density of individuals and number of species, especially in areas with elevated human influence. Understanding nuances in relationships between habitat heterogeneity and ecological communities is warranted to guide habitat-focused conservation and management efforts. We conducted fish and structural habitat surveys of thirty warm-temperate reefs on the southeastern US continental shelf to quantify how structural complexity influences fish communities. We found that intermediate complexity maximizes fish abundance on natural and artificial reefs, as well as species richness on natural reefs, challenging the current paradigm that abundance and other fish community metrics increase with increasing complexity. Naturally occurring rocky reefs of flat and complex morphologies supported equivalent abundance, biomass, species richness, and community composition of fishes. For flat and complex morphologies of rocky reefs to receive equal consideration as essential fish habitat (EFH), special attention should be given to detecting pavement type rocky reefs

1 A version of this chapter is in published in PLOS ONE as: Paxton, A.B., E.A. Pickering, A.M. Adler, J.C. Taylor, and C.H. Peterson. 2017. Flat and complex temperate reefs provide similar support for fish: evidence for a unimodal species-habitat relationship. PLOS ONE 12(9): e0183906. DOI: 10.1371/journal.pone.0183906.

because their ephemeral nature makes them difficult to detect with typical seafloor mapping methods. Artificial reefs of intermediate complexity also maximized fish abundance, but human-made structures composed of low-lying concrete and metal ships differed in community types, with less complex, concrete structures supporting lower numbers of fishes classified largely as demersal species and metal ships protruding into the water column harboring higher numbers of fishes, including more pelagic species. Results of this study are essential to the process of evaluating habitat function provided by different types and shapes of reefs on the seafloor so that all EFH across a wide range of habitat complexity may be accurately identified and properly managed.

Introduction

Habitat heterogeneity plays an important role in structuring ecological communities, as heterogeneous habitats generally support increased species density, richness, and diversity across terrestrial (MacArthur and MacArthur 1961, Jung et al. 2012, Khanaposhtani et al.

2012, Kovalenko et al. 2012), freshwater (Gorman and Karr 1978, Schneider and Winemiller

2008), and marine (McCormick 1994, Dustan et al. 2013) ecosystems. Habitat heterogeneity, also referred to as structural complexity, habitat diversity, spatial heterogeneity, architectural complexity, and other variations of these key words (Tews et al. 2004), influences fundamental processes that organize communities, including species coexistence (Holt 1984), dispersal (Huffaker 1958), recruitment success and mortality (Connell and Jones 1991,

Almany 2004), predation risk (Gilinsky 1984, Gotceitas and Colgan 1989, Beukers and Jones

1997), resource acquisition (Crowder and Cooper 1982, Gotceitas and Colgan 1989, Diehl

1992), and the strength of trophic cascades (Grabowski 2004).

Despite the well-documented role of structural complexity in supporting more abundant, more diverse, and richer communities, recent findings challenge the notion that as complexity increases so does the magnitude of community metrics (abundance, diversity, richness), suggesting that under certain scenarios, the relationship between habitat complexity and community metrics is negative or unimodal, rather than positive (Tews et al.

2004, Gazol et al. 2013). The ‘area-heterogeneity tradeoff’ combines the conceptual frameworks of niche theory (Hutchinson 1957) and island biogeography (MacArthur and

Wilson 1963, MacArthur 1967, Simberloff and Wilson 1970) to explain why the shape of the relationship between heterogeneity and community metrics may be context dependent

(Kadmon and Allouche 2007, Allouche et al. 2012). The tradeoff hypothesis posits that complex habitats have more fundamental niches and can support more species, yet as heterogeneity increases, the area suitable for each species decreases to the point where the population size decreases and the probability of stochastic extinction increases (Kadmon and

Allouche 2007, Allouche et al. 2012). The applicability of the area-heterogeneity tradeoff, however, has been questioned (Carnicer et al. 2013, Hortal et al. 2013), especially as anthropogenic impacts may influence the nature of this relationship (Seiferling et al. 2014).

In the marine environment, management decisions to alleviate anthropogenic pressures, such as fishing (Pauly et al. 1998, Jackson et al. 2001), coastal development

(Martínez et al. 2007), and tourism (Arkema et al. 2015), often limit human uses of and provide legal protection for habitats characterized by high biodiversity and ecosystem stability (Cronk 1997, McCann 2000, Lawler et al. 2006, Worm et al. 2006). Under the assumption that habitats with highest complexity support the most abundant, rich, and diverse concentrations of marine life, habitat-protection decisions commonly prioritize

conservation of the most complex habitats as opposed to the least complex habitats (National

Oceanic and Atmospheric Administration 1996, Bohnsack 1997). This paradigm ignores recent findings and the accompanying conceptual framework (i.e., area-heterogeneity tradeoff), suggesting the most complex habitats, potentially including marine habitats, may not harbor the highest density of individuals and number of species, especially in areas with elevated human influence. Understanding the structure of marine communities as a function of habitat complexity is warranted to ensure that habitat-focused conservation and management efforts encompass appropriate habitat morphologies.

Temperate reefs of the continental shelf of the southeastern United States (US) vary in structural complexity, providing a suitable system to empirically test how to guide habitat- focused management of marine habitats based on structural complexity. These reefs include naturally occurring rocky reefs ranging from flat pavements and rubble fields to substantial ledge systems with up to several meters of vertical relief (Riggs et al. 1996, 1998). The continental shelf also forms the resting place for shipwrecks (Stick 1989), as well as architecturally unique human-made structures, ranging from concrete pipes to large ships intentionally sunk to enhance fisheries (NC DMF 1988, Stick 1989, Gregg and Murphey

1994). While these natural and artificial reefs vary in morphology, they also experience dramatic state changes due to sedimentary, biological, and physical processes that alter the degree of sediment cover by alternately burying and exposing the flattest reefs (Riggs et al.

1996, 1998, Renaud et al. 1996, 1997, 1999).

Temperate reefs, including flat-to-complex rocky reefs and artificial reefs, of the southeastern US are federally-designated essential fish habitat (EFH) under the Magnuson-

Stevens Fishery Conservation and Management Act (2007) because they function as

nurseries, refugia, foraging sites, and spawning grounds. Unlike rocky reefs and artificial reefs, shipwrecks are not designated as EFH, despite forming important habitat for fishes.

Rocky reefs, artificial reefs, and shipwrecks provide habitat for a diversity of fishes, ranging from tropical and subtropical to warm-temperate reef fishes and coastal pelagics. Temperate reefs also support fishes in the federally-managed snapper-grouper complex (South Atlantic

Fishery Management Council 1983, 2016) whose status is of particular concern because of their recreational and commercial value and their frequently depressed numbers (Stephan and

Lindquist 1989, Parker Jr. and Dixon 1998, Deaton et al. 2010). These reefs are valuable for the coastal economy and culture because they create and sustain commercially and recreationally important fisheries and recreational diving opportunities. Aside from risks of overexploitation through fisheries, emerging risks on the continental shelf from offshore renewable and conventional energy development makes understanding the habitat function of these reefs pressing.

The objectives of this study were to: 1) Quantify how structural complexity of temperate reefs, measured as reef rugosity, influences fish communities; and 2) Provide management and conservation recommendations based on habitat complexity to achieve goals of fisheries and ecosystem management. This study is essential to the process of evaluating habitat function provided by different types and shapes of hard structures on the seafloor so that EFH may be accurately identified and effectively managed.

Materials and Methods

Survey Sites

We conducted scuba-diver surveys of thirty reefs off the coast of North Carolina

(NC) along the southeastern US continental shelf (Figure 1.1; Table S1.1). We selected these thirty reefs, including sites representative of different topographic complexities. Twenty- three of these warm-temperate reefs occur within Onslow Bay, NC, whereas the remaining seven sites lie in northeastern Long Bay, NC within an area designated for potential offshore wind energy development. Sites in Onslow Bay were selected a priori based on a design that was stratified by water depth, which is correlated with distance from shore. Sites in Long

Bay were selected from side-scan sonar and multibeam bathymetry datasets acquired during a seafloor mapping cruise in June 2013 (Taylor et al. 2016). Sixteen of the thirty sites are natural reefs, ranging from flat pavements to ledges, and fourteen are artificial, human-made reefs include historic shipwrecks, as well as concrete pipes and ships purposely sunk as part of the NC Artificial Reef Program.

Sites were sampled seasonally during 2013 – 2015 (Table S1.1). Most sites were sampled during each season (e.g., summer, fall, etc.), but due to rough sea conditions, several sites were sampled during only one season (Table S1.1). At each site, two 30-m long transects were established along prominent reef features. When no prominent feature existed, the transect direction was selected from a list of randomly generated compass headings. The transect location at each site varied among seasons. Diver surveys to quantify fishes and structural complexity were conducted along each transect. No specific permissions were required to survey the selected thirty reefs.

Fish Community Assessments

To quantify fish community metrics, such as abundance and composition, divers sampled along a 30-m x 4-m (120-m2) belt transect (Brock 1954, 1982, Samoilys and Carlos

2000), while recording the species and abundance of all fishes present throughout the water column, including both conspicuous and cryptic categories of reef fishes, to the lowest taxonomic level possible. Fish length was estimated to the nearest cm. Biomass was calculated with the length-weight power function as:

= where L is length (cm) recorded during the fish transect and W is weight (g). When there was a school of fish that spanned different sizes, L was calculated as the midpoint of the recorded size range. Species-specific morphometric values for a and b were obtained from Fishbase

(Froese and Pauly 2016). For species that were identifiable only to the family level, the average morphometric values for other known species in the family observed on the reefs were used. Weight was converted to kg. When two belt transects were conducted at a reef during a single sampling season, fish abundances and biomasses from each transect were averaged as a single sample to calculate respective abundance and biomass metrics. We computed species richness (S) at the finest taxonomic resolution possible (e.g., species), as well as for families. In addition to overall metrics that were inclusive of all sizes of fishes, size-class specific metrics were calculated for small fishes 1-10 cm, medium fishes 11-29 cm, large fishes 30-49 cm, and extra-large fishes ≥ 50 cm. Ethics approval was not required, as this was an observational study where fishes were visually counted and identified in situ by scientific divers.

Structural Complexity

To document how structural complexity affects fish use of temperate reefs, we collected measurements of the contour of each reef using an Onset HOBO U20 Titanium

Water Level Logger (U20-001-02-Ti) containing a pressure-transducer that records pressure at 1 Hz, from which bottom elevations are inferred. As per methods in Dustan et al. (Dustan et al. 2013), a diver swam over the reef with the logger suspended from a line and positioned as close to the substrate as possible. If benthic organisms, such as sponges, coral, and dense meadows of macroalgae, rose above the substrate preventing divers from positioning the logger close to the substrate, then divers moved the logger above these habitat-forming animals and plants to avoid damaging them and to account for the contributions of these organisms to reef complexity. The logger was moved at ~ 10 cm per second over the length of each 30-m transect. The logger was raised 1 m above and rapidly lowered back down to the substrate surface in a spike motion five times at the start of each transect, three times every 5 m thereafter, and five times at the end of each transect. Since the logger records continuously, these spikes were used to identify each transect within the data record and convert sample time to distance along transects. During post-dive processing, the distance calibration spikes were removed from each file, and raw pressures recorded by the pressure- transducer were converted from units of PSI to m, assuming an atmospheric pressure of 1 atm. If the diver swim-speed differed from the target rate of ~ 10 cm per second, then the actual swim speed was computed from the transect length and time between calibration spikes and used to determine distance along the 30-m transect.

For each transect, reef shape was visualized by plotting water depth against along- transect distance. Mean, minimum, and maximum depths were determined for each transect.

Vertical relief of each transect was computed as the difference between the minimum and maximum depth. Digital reef rugosity (DRR) (Dustan et al. 2013) was calculated as the standard deviation of depths along each transect (m). An alternative measure of rugosity was calculated as the ratio of the actual surface contour distance to the linear transect distance as:

= where C = rugosity, L = linear distance of transect (m), and D = distance of transect following the natural reef surface contour (m) (Risk 1972, McCormick 1994). Distance of the natural surface contour ( D) was computed as the sum of the hypotenuses between every two successive depth measurements recorded by the water level logger. We compared the two values for rugosity, DRR and C, and the one value for vertical relief, to ensure that these metrics were correlated across transects, and upon confirmation, DRR was selected as the metric of choice because of its precision and previously documented positive correlation with fish diversity on coral reefs (Dustan et al. 2013). To visualize the distribution of complexity values across reefs, kernel density (Sheather and Jones 1991) was estimated using the ‘stats’ package in R (R Development Core Team 2015).

Spatial variability of each structural complexity transect was visualized with variograms. Variograms decompose the spatial variability in a transect among distance classes (Legendre and Fortin 1989, Legendre and Legendre 2012). In the case of the structural complexity transects, distance classes corresponded to every measurement of depth

(m) separated by 10 cm through to 300 cm (30 m), or the entire transect distance (e.g, 10 cm,

20 cm, 30 cm… 280 cm, 290 cm, 300 cm). The variance attributed to each of these distance classes is called the semivariance. Semivariance was calculated as:

1 = − 2 where γ(d) is the semivariance at distance class d, N(d) is the number of pairs for separation of distance class d, yi is the depth at location i and yi+d is the depth at location i plus the distance class value d, and W(d) is the final location of the transect that corresponds to distance class d (Isaaks and Srivastava 1989, Legendre and Legendre 2012). Semivariance was plotted against distance classes. We plotted semivariance up to distance classes that were half the transect length to ensure that we plotted the spatially structured component of each transect (Legendre and Legendre 2012). Resulting variograms depicted the spatial scale over which the complexity of each reef varied.

Water Temperature

We measured temperature on each transect using the same Onset HOBO U20

Titanium Water Level Logger (U20-001-02-Ti) that we used to measure structural complexity. The water level logger recorded temperature every second over the duration of each transect. These raw temperature values were used to calculate mean temperature ( oC) over each transect. When multiple transects were conducted in the same sampling season, water temperatures were averaged as a single replicate.

Sediment Cover

We measured sediment depth using a hollow 2 cm diameter PVC rod containing graduated markings to the nearest cm. The rod was shaped as a ‘T,’ so that divers could apply pressure on the top, horizontal component of the ‘T’ to press the rod into the sediment.

Sediment depth measurements were obtained every three meters along the same transect that fishes and structural complexity were sampled. Sediment data were also averaged over multiple transects when a reef was surveyed more than once in a sampling season so that these data could be compared to fish and complexity data. Standard deviation of sediment depth (cm) was calculated to indicate how permanent (low standard deviation) or ephemeral

(high standard deviation) sediment cover was on reefs.

Statistical Analyses

Statistical analyses were conducted in R version 3.2.0 (R Development Core Team

2015) . We examined environmental variables for collinearity, and variables that were not collinear were retained for analyses. For example, water temperature and reef depth had a low correlation coefficient (0.04), so both were retained for subsequent analyses.

We used generalized linear models (GLMs) to determine the relationships between fish community metrics (abundance, biomass, and richness) and environmental variables and to specifically investigate how reef complexity influenced reef fishes. For fish abundance, we conducted GLMs with a negative-binomial error distribution and a log-link function using the ‘MASS’ package (Venables and Ripley 2002). Fish abundance values from each reef were initially integers. Because we conducted two transects per reef during each sampling

season, however, we later averaged the abundances from replicate transects to avoid pseudoreplication. Averaging resulted in non-integer abundances, so prior to performing

GLMs, we rounded the mean abundance data to the nearest integer since we did not encounter fractions of fish and since the negative-binomial distribution requires integers. For fish biomass data, which are inherently continuous, we utilized a gamma distribution with a log link. For species richness data, which are integers, we used a Poisson distribution.

For each response variable (e.g., abundance, biomass, richness), we fit the most complex GLM first and then compared the most complex model to candidate models of reduced complexity until reaching the most parsimonious model. The most complex models regressed fish community metrics against a linear term (DRR) and squared term (DRR 2) for complexity, as per methods in Allouche et al. (Allouche et al. 2012) to determine whether fish community metrics and complexity exhibited a unimodal relationship. These complex models also included two environmental variables, depth and water temperature, to determine whether these additional abiotic factors helped explain variance in fish community metrics. We included an additional environmental variable, sediment standard deviation, exclusively for natural reefs.

Model selection from among our most complex and more parsimonious candidate models was conducted using Akaike information criterion (AIC) values based on minimum

AIC. We conducted graphical and analytical assessments of fit to compare the predicted values from the model to the observed values. For graphical assessments of fit, we plotted the estimated probability distribution with the observed fish community metric values superimposed. This graphical method allowed us to determine whether the observed values appear typical of the estimated distribution. For analytical assessments of fit, we calculated

P-values where the observed value of fish community metrics was treated as the test statistic and the predicted probability distribution was treated as the null model.

The magnitude of the coefficients for predictor variables and the associated P-values for the best model, as determined by AIC comparisons and both graphical and analytical assessments of fit, determined the type of relationship between fish community metrics and

DRR: linear, quadratic, unimodal, or no relationship. If only the linear term (DRR) was significant, then a linear model was assumed. If only the quadratic term (DRR 2) was significant, then a quadratic relationship was assumed. If both linear (DRR) and quadratic

(DRR 2) model terms were significant, then the relationship was categorized as unimodal

(Allouche et al. 2012). If no term was significant, then this indicated no effect of DRR on fish community metrics. Models were evaluated separately by reef type: natural reefs and artificial reefs for total fishes and for each individual size class of fishes; small (1-10 cm), medium (11-29 cm), large (30-49 cm), extra large (≥ 50 cm).

To evaluate whether fish community metrics varied by category of reef morphology, we used rugosity and in situ observations to classify natural reefs as either pavements-and- rubble or extensive ledges and artificial reefs as either low-relief concrete structures or complex ships. We calculated average fish abundance, biomass, and species richness for these four reef morphologies for all fishes. We tested for differences in fish community metrics by reef morphology using analysis of variance (ANOVA) followed by post-hoc pairwise t-tests. Abundance and biomass data were both log transformed to meet homogeneity of variance assumptions.

To determine whether fish community composition varied by reef morphology, we used permutational analysis of variance (PERMANOVA), nonmetric multidimensional

scaling (nMDS) analysis, and indicator species analysis. These tests were applied to the square-root transformed fish abundance matrix at the family taxonomic level.

PERMANOVA, a permutation-based technique that uses variance partitioning (Anderson

2001), explicitly tested whether fish community composition differed by morphological categories. PERMANOVA used Bray-Curtis distances and 1,000 permutations and accounted for reef morphology (pavement-and-rubble, ledge, concrete, ships) and was run using the ‘vegan’ package (Oksanen et al. 2015). nMDS, an ordination method that summarizes patterns in the structure of multivariate datasets (Shepard 1962, Kruskal 1964,

Legendre and Legendre 2012), was performed on the fish community data using the ‘vegan’ package (Oksanen et al. 2015). Samples were mapped into ordination space using the ecological distances between samples ordered by rank terms. Bray-Curtis distances summarized pairwise distances among samples and helped overcome the problem of joint absences in species data (Oksanen et al. 2015). A Shepard diagram ensured linearity between the ordination distance and Bray-Curtis distance. Biplots with samples colored by reef morphology and superimposed ellipses indicating 50% confidence intervals allowed visualization of the relationships among samples in ordination space. Indicator species analysis determined which species were indicators of the four classes of reef morphology and was performed with the ‘indicspecies’ package (De Cáceres and Legendre 2009). Weighted averages of the indicator families were projected on top of the sample space on the nMDS biplot to visualize community patterns by reef morphology.

Results

We sampled a total of 246 transects on 30 temperate reefs located on the continental shelf of NC. Across the transects, 336,774 individual fishes belonging to 141 species and 47 families were observed (Table S1.2). Total biomass of fishes was 43,570 kg. When two transects were conducted at a reef in a single season, the transects were averaged as a single replicate; results reported below correspond to these average values.

Sampled reefs included both naturally occurring rocky reefs and human-made structures that varied in habitat complexity (Figure 1.2). Natural reefs ranged from flat pavements to distinct ledges (Figure 1.2a-b). Flat pavement-and-rubble reefs displayed relatively uniform contours (Figure 1.2e), with low variability in reef structure over the length of the transect along which fishes were surveyed (Figure 1.2i). Ledges, in contrast, contained either sharp or gradual drops and rises in reef height and exhibited higher spatial variability compared to the pavement-type reefs (Figure 1.2b, f, j). Artificial structures represented architecturally diverse habitats ranging from concrete pipes to shipwrecks and purposely scuttled vessels (Figure 1.2c-d). Structures nearly flush with the natural sandy seafloor, such as concrete pipes, displayed a relatively uniform contour map, where slight peaks in elevation coincided with the occurrence of human-made reef materials (Figure

1.2g), as well as low variability in structural complexity over transects (Figure 1.2k).

Shipwrecks and purposely sunk vessels protruded into the water column forming pronounced peaks and valleys in their contours, characterized by greater variability than lower relief structures, such as concrete pipes (Figure 1.2h, l).

Complexity of both natural and artificial reefs was calculated with a DRR metric, such that low rugosity reflects low structural complexity and high rugosity coincides with

high structural complexity. The distribution of rugosity for all reefs ranged from flat (0.1 m

DRR minimum) to highly rugose (3.3 m DRR maximum; Figure 1.3a). The distribution of natural reefs centered on flatter rugosity values (0.1 – 1.0 m DRR ) than those of artificial, which had a wider range (0.2 – 3.3 m DRR ) weighted towards the more complex part of the rugosity spectrum. Temperate reefs on the continental shelf encompassed a wide variety of shapes and sizes but natural reefs occurred over the lower third of the range in complexity exhibited by artificial, human-made structures. Likewise, vertical relief, which was highly correlated with DRR (correlation 0.98), was greater for artificial reefs (1.0 – 8.7 m vertical relief) than for natural reefs (0.5 – 3.6 m vertical relief).

Intermediate levels of reef complexity maximized fish abundance for both natural and artificial reefs (Figure 1.3b-c). As complexity increased, fish abundance increased until reaching an inflection point at intermediate levels of reef complexity; when reef complexity surpassed intermediate levels (inflection point), fish abundance decreased. For naturally occurring reefs, in the GLM with a negative-binomial error distribution containing linear and quadratic terms for natural reef DRR and a linear term accounting for reef depths, all terms were significant (Table 1.1; DRR P < 0.0001; DRR 2 P < 0.001; depth P < 0.0001), and the inflection point was within the range of the data, suggesting a unimodal curve (Figure 1.3b).

The relationship between habitat complexity and fish abundance on artificial reefs was marginally unimodal and was most significantly influenced by reef depth (Table 1.1; DRR P

< 0.04; DRR 2 P < 0.08; depth P > 0.0001; Figure 1.3c). For biomass, neither the linear nor quadratic terms for DRR described the relationship with complexity across reef types (Table

2.1; biomass: DRR P > 0.05; DRR 2 P > 0.05; Figure 1.3d-e). The model that contained DRR and DRR 2, however, fit better than models excluding DRR terms, indicating that DRR did

explain a small amount of variation in fish biomass. Regardless, reef depth explained the greatest amount of variation in fish biomass on both natural (Table 2.1; depth for natural reefs P = 0.05) and artificial reefs (Table 2.1; depth for artificial reefs P < 0.0001). On natural reefs, species richness displayed a unimodal relationship with reef complexity, when accounting for reef depth (Table 1; DRR P < 0.01; DRR 2 P < 0.01; depth P < 0.0001; Figure

1.3f), whereas species richness was unrelated to DRR on artificial reefs where reef depth and water temperature positively influenced richness (Table 2.1; depth: P < 0.001; temperature: P

< 0.01; Figure 1.3g).

The relationship between fish abundance and reef structure differed by fish size class for each type (natural versus artificial) of temperate reef (Figure S1.1; Table S1.3). The unimodal relationship between complexity and abundance for natural reefs that occurred for total fishes (Figure 1.3) was replicated for just small fishes (1-10 cm) and also influenced by reef depth, water temperature, and sediment dynamics (Figure S1.1a), whereas abundances of just medium (11-29 cm) and just large fishes (30-49 cm) were unrelated to complexity but were related to depth and sediment, respectively (Figure S1.1b-c). A marginally significant linear, positive relationship described the abundance of extra-large (≥ 50 cm) fishes as a function of complexity (Figure S1.1d; DRR P = 0.06), when accounting for reef depth where deeper reefs supported more extra-large fishes. For artificial reefs, the pattern of fish abundance having a unimodal relationship with habitat complexity for total fishes (Figure

1.3e) was preserved for just extra-large (( ≥ 50 cm) fishes, yet the inflection point occurred at lower measures of reef complexity for this size class of fishes compared to the curve for total fishes (Figure S1.1h). Abundance of large (30-49 cm) fishes was marginally and linearly related to complexity, when accounting for reef depth (Figure S1.1g). Small fish abundance

was greater on deeper reefs, and medium fish abundance was greater on deeper and warmer temperature reefs, but abundance of both of these size classes was unrelated to complexity

(Figure S1.1e-f).

Because the unimodal relationship indicated that the least complex and most complex of each reef type were similar in numbers of fishes across a range of complexity values, we categorically compared the least versus most complex reefs by morphologies: pavement-and- rubble (natural); ledge (natural); concrete (artificial); ships (artificial). Abundance and biomass of fishes using flat and highly complex natural habitats and low complexity concrete habitats did not differ from each other but were substantially less than the number and biomass of fishes using ships for habitat (Figure 1.4a-b; ANOVA: abundance F 3,119 = 11.13,

P < 0.0001; biomass F 3,119 = 9.12, P < 0.0001). Richness differed by reef morphology

(Figure 1.4c; ANOVA, F 3,119 = 4.33, P = 0.006). Flat and complex natural reefs supported equivalent numbers of species; however, complex artificial reefs (ships) supported more species than low complexity artificial reefs (concrete). Pavement-and-rubble reefs hosted higher species richness than concrete reefs.

Community compositions of fishes on pavement and ledge morphologies were similar

(Figure 1.5), while the communities of fishes on low-lying concrete structures diverged from those of structurally unique ships (Figure 1.5; PERMANOVA: F 3,122 = 4.00, P < 0.001).

Balistidae (triggerfish; indicator value = 0.49; P = 0.018) occurred on both pavements and ledges, whereas Muraenidae (eels; indicator value = 0.42; P = 0.023) and Ptereleotidae (blue dartfish; indicator value = 0.43; P = 0.007) indicated pavements. There were no indicators exclusive of ledges. Diodontidae (porcupinefish; indicator value = 0.33; P = 0.043) characterized concrete artificial reefs, whereas pelagic Scombridae (mackerel; indicator value

= 0.41; P = 0.023) and Lutjanidae (snapper; indicator value = 0.61; P = 0.001) signified submerged vessels. Sphyraenidae (barracuda; indicator value = 0.56; P = 0.015),

Odontaspididae (sandtiger; indicator value = 0.45; P = 0.002), and Dasyatidae (whiptail stingray; indicator value = 0.30; P = 0.045) denoted artificial reefs regardless of topography.

Discussion

We provide evidence that intermediate levels of warm-temperate reef complexity maximized fish abundance on natural and artificial reefs, as well as species richness on natural reefs, challenging the current paradigm that reefs of highest complexity support the most fishes and the most species of fishes. For naturally occurring rocky reefs, we discovered that flat pavement-and-rubble fields supported similar abundance, biomass, species richness, and community composition of fishes as pronounced ledges. Although low- and high- complexity artificial reefs supported equivalent numbers of fishes, artificial reefs composed of low-lying concrete structures hosted lower abundance, biomass, and species richness, as well as different community composition, than submerged metal vessels protruding high into the water-column. Our results suggest that habitat-focused management efforts should include reefs representative of a wide-variety of structural complexities, including both the most topographically complex reefs and those that are low-lying and often ephemeral EFH on the continental shelf.

Our finding that intermediate levels of reef complexity, as measured by reef rugosity, maximized fish abundance of warm-temperate reefs disagrees with the notion that the most complex reefs support the most abundant communities of fishes. We suggest several explanations for why we found a unimodal shape for the curve portraying the relationship

between structural complexity and fish abundance. First, our study focused on a fine spatial scale (30 m transects). Many studies use broader spatial scales to examine landscape- or ecosystem- scale species-habitat patterns and find that broad-scale topographic complexity positively correlates with species abundance (Tews et al. 2004, Dunn and Halpin 2009). On temperate reefs on the continental shelf similar to those we evaluated, for example, fish abundance increases with increasing complexity values derived from multibeam bathymetry

(Kracker et al. 2008a, Taylor et al. 2016). However, how fishes use their habitat changes across spatial scale (Arias-González et al. 2006). Here, perhaps our choice in spatial scale illuminated a novel relationship between fish abundance and rugosity at a local spatial scale.

Second, our study decoupled substrate type for natural reefs from an inherent productivity gradient. In most studies of habitat heterogeneity, substrate is not held constant, meaning that substrate type is coupled with an intrinsic productivity gradient (Allouche et al. 2012). In a terrestrial heterogeneity gradient, for example, substrate type could hypothetically stem from two distinct habitat types, deserts and high grass prairies, each with different substrates.

Prairies have higher densities of plants by nature of their substrate than deserts, forming an inherent productivity gradient. In our study, however, because the substrate type remained consistent - rock substrate - for natural reefs, the raw substrate (excluding benthic community) for each reef type was decoupled from a respective productivity gradient. This decoupling allowed us to examine structural complexity independently from the habitat type, perhaps resulting in a unimodal curve for natural reefs for both abundance and species richness, whereas other studies found a positive relationship when coupling productivity and substrate type. For artificial reefs, however, the substrate type varied from concrete pipes to metal ships, perhaps explaining why we found a unimodal relationship for only one fish

response metric, abundance. Third, we hypothesize that fishing pressure contributed to these relationships. Fishing pressure applied to complex reefs likely exceeds pressure on flat reefs because complex reefs create striking, permanent features on the seafloor that are easy for fishers to repeatedly locate on their bottom-finders, especially when using GPS units. In contrast, flat reefs, which are covered and exposed by sediment over time, form transient features difficult to locate with bottom-finders. Fishers can readily return to the same complex reefs, potentially reducing numbers of fishes and/or numbers of targeted species, such as apex predators. Reducing numbers of fishes and/or targeted species could drive decreased fish abundance on complex reefs. However, since we did not measure fishing pressure on the reefs, we were unable to quantitatively test the hypothesis that fishing pressure may shape the relationship between structural complexity and fish by decreasing fish abundance on the most complex reefs.

Fishes of different size classes responded differently to structural complexity. On natural reefs, abundance of total fishes, as well as just small fishes, had a unimodal relationship with complexity. Medium and large fishes were unrelated to structural complexity. Numbers of extra-large fishes, often including apex predators, such as

Mycteroperca microlepsis (gag), Mycteroperca phenax (scamp), and Seriola dumerili

(greater amberjack), increased linearly with rugosity on natural reefs, albeit with marginal statistical significance, concurring with previous temperate reef research (Kracker et al.

2008b, Kendall et al. 2009, Taylor et al. 2016). On artificial reefs, total fishes and extra-large fishes only both exhibited a unimodal relationship with habitat complexity, while large fish were linearly related to complexity, and other size classes were unrelated. These size-class specific responses may be explained by inherent associations of fishes with habitat that

change through ontogeny. For example, fish, such as M. microlepsis , move from shallow, nearshore habitats of typically lower complexity to offshore, deeper habitats of generally higher complexity as they mature (Ross and Moser 1995, Stallings et al. 2010). Our data support this notion, as abundance of extra-large fishes on both natural and artificial reef types increased with reef depth. This change in habitat preference through ontogeny may explain differences in how fishes respond to complexity by size class.

The strength of the unimodal relationship between habitat complexity and fish abundance was stronger for natural than artificial reefs, and intermediate complexity maximized fish species richness on natural but not artificial reefs; we pose three explanations. First, artificial reefs harbored three-to-four times as many fishes as natural reefs, represented most prominently by schooling fishes. Schooling fishes, such as Haemulon aurolineatum (tomtate) , Rhomboplites aurorubens (vermillion snapper) , and Decapterus sp.

(scad), drove the pattern of elevated fish abundance on artificial versus natural reefs. This supports previous findings that schooling fishes, including those that are partially planktivorous ( H. aurolineatum and R. aurorubens ; e.g., consume plankton and other prey items) and those that are strictly planktivores ( Decapterus sp.), are more abundant on artificial than natural reefs (Arena et al. 2007, Simon et al. 2013). Because presence of schooling fishes on reefs is more ephemeral than the presence of demersal fishes, variability introduced by schooling fishes may have allowed the expression of a less pronounced unimodal shape to the abundance-complexity curve on artificial as compared to natural reefs.

Second, natural reef complexity clustered in the lower third of the value range of artificial reefs. Lower complexity of natural temperate reefs makes them susceptible to burial and exposure by sediment movement (Renaud et al. 1997). Low-lying artificial reefs, such as

concrete pipes, face sediment movement similar to those experienced by natural reefs, however, vertically-extensive artificial reefs, such as metal ships, do not experience the same levels of sediment burial and exposure as their lower-relief counterparts. This discrepancy where low- and high- complexity artificial reefs face differing levels of physical disturbance could explain the weaker species-complexity relationship on artificial versus natural reefs.

Third, although we surveyed an identical area on each reef, artificial reefs often occupy a smaller benthic footprint than natural reefs. Natural reefs often form extensive, branching networks, whereas artificial reefs act as discrete islands. The island-like nature of artificial reefs where habitat occupies discrete patches may contribute to a less pronounced species- complexity relationship for artificial reefs than for their natural reef counterparts. Assessing the relationship of fish community metrics and natural reef rugosity in the context of the arrangement of habitats on a larger scale represents a compelling avenue for future research.

Our finding that intermediate levels of complexity on natural reefs maximized fish species richness provides evidence that the area-heterogeneity tradeoff (Kadmon and

Allouche 2007, Allouche et al. 2012) operates on warm-temperate reefs. Ours is the first study, to our knowledge, suggesting that complexity of rocky, warm-temperate reefs reaches a threshold, above which species richness decreases. For abundance, the area-heterogeneity tradeoff predicts a negative relationship with increasing heterogeneity, depending on the system (Allouche et al. 2012). In our temperate system, we found a unimodal relationship between fish abundance and heterogeneity. We posit two explanations for why our results for abundance differ from theoretical expectations. First, the area-heterogeneity tradeoff is typically envisioned in a landcover-diversity context with heterogeneity indicating different habitat types. In our study, we measured heterogeneity of different reef types as DRR, a

relevant metric for fishes across multiple scales (Dunn and Halpin 2009, Dustan et al. 2013), yet rugosity may not be as relevant at a landscape scale for the tradeoff hypothesis. Second, besides tradeoffs between the number of fishes and fish species that can be supported by different levels of habitat complexity, other tradeoffs likely occur on temperate reefs that are not area related. For example, we found that environmental variables, such as reef depth and water temperature, influenced fish communities, agreeing with previous temperate-reef research on the southeast US Atlantic continental shelf (Whitfield et al. 2014).

Naturally occurring pavement-and-rubble reefs harbored similar communities

(abundance, biomass, richness, community composition) of fishes as did rocky ledges.

Similar community types across natural reef morphologies is particularly interesting from a management perspective. This is because flat, pavement morphologies, often covered with a veneer of sediment, although federally designated as EFH prove difficult to detect (Lucieer et al. 2013), as they are frequently buried by sediment. Commonly employed seafloor mapping methods include side-scan sonar and multibeam bathymetry, which use sound waves to ensonify the seafloor topography (Simmonds and MacLennan 2005). Side-scan sonar cannot adequately detect pavements covered in a veneer of sand unless mapping is conducted at fine resolution to detect invertebrates, such as soft coral, that create a texture distinguishable from sand [82; fine-resolution

reefs covered in a veneer of sediment may be responsible for the elevated numbers and/or biomass of fishes. Other instruments, such as video cameras and sub-bottom profilers, as well as in situ diver-based visual surveys, can easily detect pavements (Kendall et al. 2005,

Walker et al. 2008). However, the sampling area of these instruments is so small that for surveys of large geographic areas (> 10-100’s of m), these methods prove inefficient. Novel methods to detect flat pavements should be developed given that these low-lying habitats support similar numbers and types of fishes as ledges.

Low- and high- complexity artificial reefs harbored similar numbers of fishes as a function of the continuous predictor DRR, but when artificial reefs were separated into morphological categories of low-lying concrete structures and metal ships, the pattern differed. Concrete pipes hosted fewer fish and species of fish, as well as distinct community types, compared to metal ships. Concrete structures nearly flush with the sandy seafloor formed prime habitat for demersal fishes, such as Diodontidae (porcupinefish), that prey on animals growing on reefs and living within the sediment and also use the reef structure to seek refuge from predators. Pelagic species often found in the water column above reefs, including Scombridae (mackerel) and Lutjanidae (snapper), however, preferred ships with vertically-extensive topography. Three families of top predators, Odontaspididae (sand-tiger sharks), Sphyraenidae (barracuda), and Dasyatidae (whiptail stingrays) indicated generic artificial reefs. Distinguishable communities of fishes relying on low- versus high- complexity artificial reefs suggest that managers should deploy human-made reefs of varying topographic complexity based on particular fisheries they aim to enhance. Renewable energy infrastructure, such as wind turbine monopiles that extend throughout the water column

(high-complexity) and associated anti-scour aprons of rocks and concrete (low-complexity),

may combine attributes from reefs across a range of complexities, providing habitat for both demersal and pelagic reef-associated communities. Additionally, given similarities in fish community composition between low-lying concrete pipes and natural reefs, concrete pipes may serve as refugia for fishes commonly occupying natural reefs in the future.

Marine conservation and management initiatives commonly target the most structurally complex and diverse reefs. Our results, however, suggest that less complex habitats require as much consideration for these initiatives as more complex morphologies.

This is a pressing issue as human uses of the coastal ocean increase and marine-spatial planning becomes more commonplace along the continental shelf. Management efforts should afford equal consideration to a diversity of reef types, including both low- and high- complexity reefs. Given current difficulties in detecting naturally occurring rock pavements covered with a veneer of sediment, these flat reefs, even though already designated as EFH, warrant extra attention when obtaining data used for spatially-explicit planning during seafloor mapping so that they can be delineated. Submerged, human-made structures across a range of topographies support different communities of fishes, and this information will prove useful when designing and deploying additional unnatural structures.

Acknowledgements

We thank G. Safrit, G. Sorg, J. Fleming, T. Courtney, M. Kenworthy, A. Poray, D.

Keller, I. Kroll, C. Hamilton, J. Hughes, J. Boulton, T. Dodson, J. Purifoy, S. Davis, C.

Lewis, E. Kromka, E. Ebert, J. Vander Pluym, B. Teer, B. Degan, J. Hackney, R. Muñoz,

D.W. Freshwater, K. Johns, G. Compeau, J. Styron, D. Wells, S. Hall, M. Dionesotes, C.

Marino, I. Conti-Jerpe, E. Weston, M. Wooster, J. McCord, D. Sybert, R. Purifoy and crew

from Olympus Dive Center, T. Leonard and crew from Discovery Diving for diving and boating assistance. We thank J. Fodrie, J. Rosman, A. Hurlbert, S. Fegley, G. Kellison and C.

Schobernd for reviews and guidance. We thank P. Whitfield for advice on site selection and survey methods. We also thank S.J. Brandl, H.M. Patterson, and an anonymous reviewer for their thoughtful reviews. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the opinions or policies of the US

Government, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.

Figures

Figure 1.2. Habitat complexity of temperate reefs. a-d) Representative images of temperate reef morphologies. e-h) Representative depth contours of each reef morphology along the surveyed transect length. i-l) Representative semivariograms of each reef for half the distance of the surveyed transect length. Columns refer to different reef morphologies as follows, from left to right: naturally occurring pavement-and-rubble reef, naturally occurring ledge outcrop, artificial reef composed of concrete pipes, and a ship representative of historic shipwrecks and vessels intentionally sunk to enhance fish habitat.

Figure 1.3. Relationship between digital reef rugosity (DRR) and fish community metrics on natural (blue) and artificial (red) temperate reefs. A) Kernel density of digital reef rugosity

(DRR) by reef type (N natural = 67, N artificial = 56). B-G) Three-dimensional surface plot of

GLM between fish community metrics and environmental predictor variables for natural reefs (left column) and artificial reefs (right column). Perspective grid surface represents

GLM predictions. Points are raw data. Perpendicular segments attached to points depict whether the raw data are above (positive, dark color) or below (negative, light color) the surface predicted by GLM. Abundance (fishes / 120 m2) was modeled with a negative- binomial error distribution (b-c), biomass (kg / 120 m 2) with a gamma distribution (d-e), and species richness with a Poisson distribution (f-g).

Figure 1.4. Fish community metrics by morphological category for natural reefs (blue;

Npavement&rubble = 38, N ledge = 29) and artificial reefs (red; N concrete = 17, N ship = 39). A) Fish abundance (fishes per 120 m 2). B) Fish biomass (kg / 120 m 2). C) Fish species richness. Data displayed are untransformed, whereas ANOVAs were conducted on log-transformed data for abundance and biomass to meet assumptions of homogeneity of variance.

Figure 1.5. Biplot of nonmetric multidimensional scaling (nMDS) ordination for fish community at the family level overlaid with indicators of reef morphologies. Ellipses are

50% confidence intervals for samples classified by each reef morphology. Family names correspond to weighted averages of indicator families, colored according to morphology or reef type (artificial or natural).

Tables

Table 1.1. GLM results for the relationship between fish community metrics (abundance, biomass, richness) and environmental predictor variables by reef type. Environmental variables include digital reef rugosity (DRR (m)), squared digital reef rugosity (DRR 2 (m)), average reef depth (m), average water temperature (oC), and standard deviation of sediment cover (m) approximating sediment dynamics. Coefficients, standard error (SE), Z-values and

P-values are provided for each environmental parameter. Bold values indicate significance or marginal significance. Interpretation of the pattern (unimodal or non-significant (NS)) between rugosity and the fish community metric are displayed for each model. Model results displayed here were from the best models that we evaluated.

Response Error Predictor Rugosity variable Reef type distribution variable Coefficient SE Z-value P-value pattern Negative Abundance Natural binomial Intercept 2.26 0.79 2.84 <0.01 Unimodal

DRR 9.72 2.38 4.08 <0.0001 DRR 2 -9.81 2.58 -3.81 <0.001 Depth 0.11 0.02 4.76 <0.0001 Negative Unimodal Abundance Artificial binomial Intercept 4.37 0.45 9.72 <0.0001 (marginally) DRR 1.31 0.63 2.07 0.04 DRR 2 -0.37 0.21 -1.75 0.08 Depth 0.12 0.02 5.82 <0.0001 Biomass Natural Gamma Intercept 2.38 0.95 2.50 0.02 NS

DRR -0.17 2.86 -0.06 0.95 DRR 2 2.24 3.10 0.72 0.47 Depth 0.06 0.02 2.00 0.05 Biomass Artificial Gamma Intercept 1.89 0.57 3.34 <0.01 NS DRR -0.55 0.80 -0.68 0.49 DRR 2 0.20 0.27 0.77 0.44 Depth 0.18 0.02 7.12 <0.0001 Species richness Natural Poisson Intercept 1.77 0.20 8.79 <0.0001 Unimodal

DRR 2.22 0.60 3.69 <0.001 DRR 2 -2.33 0.67 -3.49 <0.001 Depth 0.04 0.01 6.16 <0.0001

Species richness Artificial Poisson Intercept 1.28 0.28 4.54 <0.0001 NS Depth 0.03 0.01 6.37 <0.0001 Temperature 0.04 0.01 3.78 <0.001

35 CHAPTER 2: ARTIFICIAL STRUCUTRES HOST HIGHER ABUNDNACES OF TROPICAL AND SUBTROPICAL FISHES BUT LOWER ABUNDANCES OF TEMPERATE FISHES THAN ROCKY REEFS OF NORTH CAROLINA, USA 2

Summary

Increasing numbers of human-made structures are deployed in marine waters to enhance fish habitat and mimic natural reefs. Whether ecological functions of purposely sunk artificial reefs are similar to natural reefs is debated. Understanding whether these reef types provide equivalent habitat for fishes and benthos with different climate ranges is important for predicting effects of future marine urbanization and for guiding future installation of human-made structures . We conducted diver-surveys of thirty temperate reefs in North

Carolina, USA to contrast ecological functions of artificial and natural reefs . First, we tested if and how each reef type provided habitat for fishes and benthos with different climate ranges: temperate, subtropical, tropical. Second, we determined whether bottom-up factors, including environmental variables and benthic cover, affected tropical, subtropical, and temperate fish communities differently on each reef type. We found that reef-associated, demersal fishes with different climate ranges exhibited numerical affinities for reef types.

Temperate fishes occurred in higher abundances on natural reefs, whereas subtropical and tropical fishes at deep depths (25-35 m) resided in higher numbers on artificial structures.

2 A version of this chapter is in revision at OCEAN AND COASTAL MANAGEMENT as: Paxton, A.B., C.H. Peterson, J.C. Taylor, A.M. Adler, and E.A. Pickering. Marine urban structures host higher abundances of tropical and subtropical fishes but lower abundances of temperate fishes than rocky reefs of North Carolina, USA.

36 Subtropical and tropical species that rarely occur as far north as sites studied contributed to these patterns, with species inhabiting artificial reefs consuming zooplankton and nekton and those occupying natural reefs consuming macroalgae. For the benthos, temperate and tropical species exhibited higher cover on natural than artificial reefs; subtropical species displayed the opposite trend. Bottom-up forcing of the fish community acted through abiotic and biotic factors on artificial reefs but largely through abiotic factors on natural reefs. Our findings suggest that as marine urbanization continues world-wide, introduction of human-made structures will likely affect temperate, subtropical, and tropical species differently.

Specifically, artificial reefs may facilitate movement of subtropical and tropical fishes to functionally temperate waters by providing habitat corridors for poleward movement.

Introduction

Numbers of human-made structures in coastal oceans are increasing. Worldwide, an estimated three million or more shipwrecks rest on the seafloor (UNESCO 2017) along with

7,500 oil rigs that extend throughout the water column (Parente et al. 2006, Macreadie et al.

2011). Humans deploy additional non-natural structures for a variety of purposes, including to protect shorelines, harness energy resources, create and restore habitat, and foster tourism, fishing, and diving opportunities (Baine 2001, Bulleri and Chapman 2010, Dugan et al. 2012,

Dafforn et al. 2015). For example, artificial reefs, ranging from concrete materials to decommissioned vessels, continue to be deployed to enhance fish habitat and provide fishing and diving sites (Seaman 2000, Baine 2001, NOAA 2007, UNEP 2009). These deliberately sunk artificial reefs are intended to mimic natural reefs and restore habitats by alleviating pressures to natural habitats from degrading human activities, such as fishing and tourism.

Introduction of these human-made, artificial structures to coastal oceans qualifies as “marine urbanization” and has the potential to drive ecological change (Bulleri and Chapman 2010,

Dafforn et al. 2015).

While managers deploy artificial reefs as surrogates for natural reefs, physical attributes, such as material, topography, and isolation, consistently differ between artificial and natural reefs (Carr and Hixon 1997, Perkol-Finkel et al. 2006). Naturally occurring reefs form from reef-building organisms and/or geological features, such as rocky outcrops. Non- natural materials, including metal and concrete, comprise artificial reefs (Baine 2001).

Artificial reefs project into the water column above the seafloor and are often more structurally complex and have higher vertical relief than natural reefs (Perkol-Finkel et al.

2006, Granneman and Steele 2015, Taylor et al. 2016, Paxton et al. 2017). From a landscape perspective, artificial reefs form discrete islands on the seafloor, often isolated from other reefs (Carr and Hixon 1997), whereas natural reefs form branching, often contiguous networks.

Differences in reef-specific physical attributes likely determine the extent to which artificial structures drive ecological change and affect ecological functions of these reefs for fish and benthic communities. Questions remain, however, of whether artificial reefs function identically to natural reefs that they are intended to mimic ecologically. Some artificial reefs support higher fish community metrics, including abundance, richness, and diversity, than natural reefs (Arena et al. 2007, Folpp et al. 2013, Koeck et al. 2014, Granneman and Steele

2015), whereas other artificial reefs host lower values of these community metrics

(Burchmore et al. 1985, Carr and Hixon 1997, Rooker et al. 1997, Burt et al. 2009, Carvalho et al. 2013) or equivalent values (Fowler and Booth 2012) compared to natural reefs.

Artificial reefs can provide habitat for higher abundances of planktivorous and piscivorous fishes than natural reefs, whereas natural reefs can support greater abundances of herbivores and generalist feeders (Arena et al. 2007, Simon et al. 2013). Within the benthic community, some artificial reefs support greater coral cover (Burt et al. 2009) but others support lower benthic diversity (Carvalho et al. 2013) than natural reefs; specifically, carnivorous species often prevail on artificial reefs, whereas suspension feeders dominate natural reefs (Carvalho et al. 2013). These community level differences in fishes and benthos on artificial and natural reefs influence ecological processes, such as the strength of predation (Ferrario et al. 2016) and trophic interactions (Simon et al. 2013, Simonsen et al. 2015)

Artificial reefs can provide habitat for species with different climate ranges, such as temperate, subtropical, and tropical species. If the function of artificial reefs differs from natural reefs, it has the potential to affect how species with dissimilar climate ranges react to marine urbanization. For instance, artificial reefs are hypothesized to form ‘stepping-stones’ or connectivity corridors that facilitate organism movement among habitats (Bulleri and

Airoldi 2005, Macreadie et al. 2011, Airoldi et al. 2015, Glasby and Connell 2017). Artificial reefs can support a higher proportion of transient fish species than neighboring natural reefs

(Burchmore et al. 1985) and can facilitate the spread of invasive species, such as some tunicates, to nonnative habitats at greater rates than their natural reef counterparts (Dafforn et al. 2012, Airoldi et al. 2015). It is unknown whether human-made structures also facilitate the movement of subtropical and tropical species poleward to track suitable water temperatures, a phenomenon known as ‘tropicalization’ (Horta e Costa et al. 2014, Barceló et al. 2016, Vergés et al. 2016), at a greater rate than natural reefs and/or whether species with different climate ranges prefer one reef type over the other.

Although artificial and natural reefs can both provide habitat for fish species with different climate ranges, it remains unknown whether each reef type is driven by similar physical factors. Naturally occurring temperate reefs in the coastal ocean experience physical forcing from processes that transport and modify sediment cover, thereby alternately burying and exposing reefs and sedentary organisms (Renaud et al. 1997, Riggs et al. 1998). Water temperatures also drive fish communities on natural reefs, with winter water temperatures determining fish community composition the following spring and summer, and warmer, deeper reefs supporting high abundances of tropical fishes (Whitfield et al. 2014). On artificial reefs, abiotic factors also likely exert ecological control. For example, water currents (Baynes and Szmant 1989, Lindquist and Pietrafesa 1989), topography (Carr and

Hixon 1997, Hunter and Sayer 2009), and water temperature (Lindquist et al. 1985) affect fish and benthic community structure. Evidence exists, however, that biotic factors may trump abiotic factors in determining fish community structure on artificial reefs (Ferrario et al. 2016). With environmental conditions, such as water temperature and currents, forecast to change, understanding pathways of bottom-up forcing on each reef type is timely.

Reefs on the inner continental shelf of North Carolina (NC), United States (US) present an opportunity to test effects of marine urbanization on temperate reefs by examining how ecological function and bottom-up drivers differ between human-made reefs and their natural reef counterparts. Human-made structures in NC include artificial reefs and shipwrecks (NC DMF 1988, Stick 1989). Naturally occurring rocky reefs range from flat pavements to pronounced ledges. Both artificial and natural reefs provide habitat for fishes with diverse climate ranges, including temperate, subtropical, and tropical species (Lindquist and Pietrafesa 1989, Parker Jr. and Dixon 1998, Whitfield et al. 2014). These reefs face

dynamic oceanographic conditions, as they occur at the confluence of warm Gulf Stream waters flowing northward and colder waters associated with the Labrador Current flowing southward (Churchill and Berger 1998, Churchill and Gawarkiewicz 2009). These dynamic conditions create changes in water temperature over hourly, daily, and monthly timescales, and Cape Hatteras, NC is recognized as a biogeographic transition zone (Cerame-Vivas and

Gray 1966, Karlson and Osman 2012). Despite proximity to a biogeographic transition zone, mean annual seawater temperatures have not increased in NC coastal waters over ten years

(Whitfield et al. 2014). This dynamic oceanographic environment decoupled from annual increases in mean sea temperatures provides an opportunity to largely isolate effects of marine urbanization from those of global climate change and directly test how human-made submerged structures influence ecological functions.

The objective of this study was to contrast ecological functions of artificial and natural reefs. First, we tested if and how each reef type provided habitat for species with different climate ranges: temperate, subtropical, and tropical. We tested this question for fishes and benthos separately. Second, we determined whether bottom-up factors, including environmental variables and benthic cover, affected temperate, subtropical, and tropical fish communities differently on artificial versus natural reefs. Answering these questions is important for predicting effects of future marine urbanization and for guiding future installation of human-made structures. We answered these questions by conducting diver- based assessments of thirty temperate reefs.

Materials and Methods

Survey Sites

We conducted SCUBA-diver surveys of thirty reefs off the coast of NC along the southeastern US continental shelf (Figure 2.1; Table S2.1). Fourteen of the thirty sites are artificial, human-made reefs including historic shipwrecks, as well as concrete pipes and ships purposely sunk as part of the NC Artificial Reef Program (NC DMF 1988). Sixteen are natural reefs, representing flat- to ledge-type morphologies. Reefs span 1.3 degrees of latitude, ranging from 33.4N to 34.7N. Twenty-three of these reefs occur within Onslow Bay,

NC, whereas the remaining seven sites lie farther south in northeastern Long Bay, NC within an area designated for potential offshore wind energy development. Sites in Onslow Bay were selected a priori based on a design that was stratified by water depth, which is correlated with distance from shore. Sites in Long Bay were selected from side-scan sonar and multibeam bathymetry datasets acquired during a seafloor mapping cruise in June 2013

(Taylor et al. 2016).

Sites were sampled seasonally during 2013 – 2015 (Table S3.1). Seasons were classified as: winter (January - March), spring (April - June), summer (July - September), and fall (October – December). Most sites were sampled during each season, but several were sampled fewer times because of rough sea conditions. At each site, two 30-m long transects were established along prominent reef features. If the reef lacked a prominent feature, we used a list of randomly generated compass headings to select the transect direction. Transect location at each site varied among seasons. Diver surveys to quantify fishes, benthos, reef topography, and water temperature were conducted along each transect.

Fish Community

To quantify fish community metrics, such as abundance and biomass, divers sampled along a 30-m x 4-m (120-m2) belt transect (Brock 1954, 1982, Samoilys and Carlos 2000), while recording species and abundance of all fishes present throughout the water column.

Each belt-transect included both conspicuous and cryptic categories of reef fishes that were identified to the lowest taxonomic level possible. Fish length was estimated visually to the nearest cm. Biomass was calculated with the length-weight power function as:

= where L is length (cm) recorded on the fish transect, and W is weight (g). When a school of fishes spanned multiple sizes, L was calculated as the midpoint of the recorded size range.

Species-specific morphometric values for a and b were obtained from Fishbase (Froese and

Pauly 2016). For species that were identifiable only to the family level, the average morphometric values for other known species in the family present on the reefs were used.

Weight was converted to kg. Fish habitat zones were assigned as demersal for reef-associated fishes or pelagic for fishes associated with the water-column. Fish climate ranges were assigned as temperate, subtropical, or tropical using published classifications from Fishbase

(Froese and Pauly 2016) and Whitfield et al. (2014), which designated species as ‘tropical’ if their northern distribution limit occurred at the northern Atlantic coast of FL, ‘subtropical’ if their range extended poleward to NC, and ‘temperate’ if their range encompassed the northeastern US. For fishes identified to the family level, the predominant climate range of other species in that family also present on the reef was assigned. Species whose northernmost distributional latitude in Fishbase (Froese and Pauly 2016) fell south of studied

reefs were also identified, but we caution that latitudinal ranges reported in Fishbase do not always include rare sightings at higher latitudes. When two belt transects were conducted at a reef during a single sampling season, the fish abundances and biomasses from each transect were each averaged as a single replicate to characterize abundance and biomass metrics.

Benthic Community

To quantify cover of benthic invertebrates and macroalgae, divers conducted photoquadrat surveys (Bohnsack 1979). Eleven 25-cm x 25-cm photoquadrats were obtained every 3 m starting at 0 m and ending at 30 m along each 30-m transect with an Olympus E-

PM1 twelve-megapixel digital camera, Olympus PT-EP06 underwater housing, and Ikelite

DS161 Substrobe. An underwater framer was attached to the camera housing to ensure fixed photoquadrat size (25 cm 2) and fixed distance between the camera lens and the substrate.

Two images of each quadrat were taken to provide duplicate versions if needed (e.g., if one photo was not in focus). Because transects were randomly established during each sampling season, specific photoquadrat plots were not revisited.

CoralNet software (Beijbom et al. 2012) overlaid 100 stratified random points on each photoquadrat image. These points were stratified by dividing the image into five rows and five columns of cells. In each cell, four points were randomly generated. The organism present at each point was identified to the lowest taxonomic level possible by CoralNet using a machine-learning algorithm. Each point identified by the algorithm was then verified or corrected by trained human experts. If multiple layers of epibiota were present, only the topmost layer was identified. Percent cover for each photoquadrat was calculated at the lowest taxonomic level possible; we also calculated percent cover by phylum and functional

group. Benthos climate ranges were assigned using AlgaeBase (Guiry and Guiry 2007),

World Register of Marine Species (WoRMS Editorial Board 2016), Gray’s Reef National

Marine Sanctuary species identification guide (Gleason et al. 2015), Marine Species

Identification Portal (“Marine Species Identification Portal” n.d.), and Schneider and Searles

(1991). Species were classified as tropical, subtropical, and temperate as described for fishes above. Points on each quadrat identified as transect hardware, fish, or unclear were removed, and the total number of points was scaled to 100 points per quadrat so that all quadrats were weighted equally. Scaled cover values from photoquadrats along a transect were averaged as a single replicate when more than one transect was conducted on a reef during a sampling season.

Reef Topography

We collected measurements of the contour of each reef using an Onset HOBO U20

Titanium Water Level Logger (U20-001-02-Ti) containing a pressure-transducer that records pressure at 1 Hz, from which bottom elevations were inferred. As per methods developed by

Dustan et al. (2013) and implemented by Paxton et al. ( In review) , a diver swam over the reef with the logger suspended from a line and positioned as close to the substrate as possible. If benthic organisms, such as sponges, coral, and dense meadows of macroalgae, rose above the substrate preventing us from positioning the logger close to the substrate, then we moved the logger above these habitat-forming animals and plants to avoid damaging them and to account for them in our complexity measurement. The logger was moved at ~ 10 cm per second over the length of each 30-m transect. The logger was raised 1 m above and rapidly lowered back down to the substrate surface in a spike motion five times at the start of each

transect, three times every 5 m thereafter, and five times at the end of each transect. Since the logger records continuously, these spikes were used to identify each transect within the data record and convert sample time to distance along transects. During post-dive processing, distance calibration spikes were removed from each file, and raw pressures recorded by the pressure-transducer were converted from units of psi to m, assuming an atmospheric pressure of 1 atm. If the diver swim-speed differed from the target rate of ~ 10 cm per second, then the actual swim speed was computed from the transect length and time between calibration spikes and used to determine distance along the 30-m transect. For each transect, mean, minimum, and maximum depths were determined. Digital reef rugosity (DRR) (Dustan et al.

2013) was calculated as the standard deviation of depths along each transect (m). Mean depth was used to categorically classify reef depth as: shallow: 5-18 m, intermediate: 18-25 m, deep: 25-35 m.

Water Temperature

We measured temperature on each transect using the same Onset HOBO U20

Statistical Analyses

Statistical analyses were conducted in R version 3.3.2 (R Development Core Team

2016). Unless otherwise noted, for fish analyses, we examined demersal fishes and then

separately considered pelagic fishes because pelagic fishes occur in dramatically higher numbers, often in large schools, and are more transient inhabitants of both reef types.

We used linear mixed effects models to test whether fixed effects of reef type

(artificial, natural), depth (shallow, intermediate, deep), and season (fall, winter, spring, summer) affected abundance and biomass of fishes and cover of benthos by individual climate ranges. Site was included as a random effect. We first ensured that data met assumptions of homogeneity of variance. We then fit each model by maximizing the restricted log-likelihood (REML) using the ‘nlme’ package (Pinheiro et al. 2013) with all fixed effects predictor variables and their associated interactions; nonsignificant predictor variables and/or interactions were then removed and more parsimonious models were fit until we reached a final model that was selected based on the minimum AIC value. Models were run individually for abundance and biomass of all fishes, demersal fishes, and pelagic fishes for each of the three climate ranges. Separate models were also run for overall benthic cover, as well as invertebrate and macroalgal cover, by each of the three climate ranges. Model fits were examined by plotting fitted values against standardized residuals to ensure that residuals closely resembled fitted values. We compiled model results into tables and then visualized results of linear mixed effects models by plotting predicted values of the means obtained from the variance-covariance matrix, as well as their associated standard error.

We used permutational analysis of variance (PERMANOVA), nonmetric multidimensional scaling (nMDS), and similarity percentage analysis (SIMPER) to understand how fish community composition, as classified by climate range, varied by reef type. The input data matrix for PERMANOVA and nMDS was composed of square-root transformed abundances of tropical, subtropical, and temperate reef fishes. PERMANOVA

explicitly tested whether community composition differed by reef type and was conducted with 1,000 permutations and Bray-Curtis distances. nMDS summarized patterns in the fish community by mapping samples into ordination space using ecological distances between samples ordered by rank terms. We used a Shepard diagram to ensure linearity between the ordination distance and Bray-Curtis distances. SIMPER identified how temperate, subtropical, and tropical fishes contributed to differences in community composition. Both

PERMANOVA and nMDS were performed using the ‘vegan’ package (Oksanen et al. 2015), whereas SIMPER was conducted with the ‘ecodist’ package (Goslee and Urban 2007).

For fishes observed on the reefs whose normal range limits were documented in

Fishbase (Froese and Pauly 2016) as south of the studied reefs, we calculated their average abundances on artificial and natural reefs. We visualized whether these fishes were more abundant on artificial or natural reefs by plotting the mean abundance of each selected fish by reef type. Because many species exhibited zero- or low-abundance on one or both reef types, we log-transformed the abundance values as follows:

log + 0.01 to better visualize the plotted data. We used two-sample t-tests to quantify whether the mean abundances of these fishes statistically differed by reef type.

To determine whether fish abundance on artificial and natural reefs was driven by similar bottom-up factors, we used a multigroup structural equation model (SEM) because it enabled us to investigate direct and indirect effects of potential bottom-up drivers for each reef type (Grace 2008). We fit multiple multigroup-SEMs using the ‘lavaan’ package

(Rosseeel 2012) and following SEM fitting recommendations (Grace 2008, Byrnes et al.

2011, Duffy et al. 2015). Exogenous predictor variables included both abiotic and biotic

variables: digital reef rugosity (DRR, m), vertical relief (m), mean reef depth (m), mean water temperature ( oC), reef latitude (decimal degrees), temperate benthic cover (%), subtropical benthic cover (%), and tropical benthic cover (%). Endogenous response variables included abundance of temperate, subtropical, and tropical fishes. Initial models included latent variables termed complexity (comprised of DRR and vertical relief; high, positive correlation accounted for) and reef location (comprised of mean depth, mean water temperature, and latitude). With the latent variables, our model did not fit because the model was not identified, so we separately fixed coefficients within each group of latent variables; however, this model did not satisfy goodness of fit tests. Next, we removed the latent variables and instead included DRR, mean depth, mean water temperature, and latitude in the

SEM as individual exogenous predictor variables. Vertical relief was excluded because it was highly correlated with DRR (r = 0.98). We used ecological knowledge of the system to develop and test additional models related to bottom-up factors. For example, instead of including total benthos cover for each climate range, we included benthic cover of each climate range as separate groups for macroalgal and invertebrate cover because we hypothesized that macroalgae and invertebrates may have separate effects on fishes given trophic ecology. We also tested models where bottom-up forcing on fishes from abiotic variables was first mediated by the benthos, as well as models where abiotic drivers influenced fishes directly and also through an indirect benthic pathway. We chose our final

SEM based on four measures of model fit: 1) Chi-square ( p > 0.05 indicates good fit), 2)

Root Mean Square Error of Approximation (RMSEA with lower 90% confidence interval <

0.05 indicates good fit), 3) Comparative fit index (CFI > 0.90 indicates good fit), and 4)

Standardized Root Mean Square Residual (SRMR < 0.10 indicates good fit) (Kline 2016).

We visualized model results with path diagrams and reported standardized path coefficients because the standardized coefficients permit comparison of indirect and direct pathways of bottom-up drivers between each reef type.

Results

We sampled a total of 226 transects on thirty warm-temperate reefs on the inner continental shelf of NC. Across the belt transects, 334,579 individual fishes belonging to 134 species and 43 families were observed (Table S3. 2). Of the 136 fish species, seven (six demersal; one pelagic) were classified as temperate, 80 (62 demersal; 17 pelagic) as subtropical, and 48 (47 demersal; 2 pelagic) as tropical species. Two species of sharks, belonging to two different families, and two species of turtles were also observed. Of the total number of fishes, 90,640 were demersal, reef-associated fishes, and 243,939 were pelagic, water-column-associated fishes. Total biomass of fishes, sharks, and turtles across the study was 42,551 kg. Demersal fish biomass was 9,004, and pelagic fish biomass was

33,493 kg. Sharks and turtles represented a combined biomass of 54 kg. Sharks and turtles were included in total fish metrics but not in demersal or pelagic values because of their inconsistent use of the reef versus water column; also, turtles are not fish. Across the photoquadrat surveys, a total of 2,486 individual photoquadrats, 87 benthic species or groups of species (e.g., Gracilaria sp. / Rhodoymenia sp. represents a group of species, rather than an individual species because they are indistinguishable in photographs) were observed

(Table S3.3). These benthic species belonged to 12 phyla. For the benthos overall, 20 species or groups of species were classified as temperate, 53 as subtropical, and 13 as tropical. A rarely observed black-crust algae was unidentifiable to genus level and was not assigned a

range because there was insufficient information. Invertebrate cover was 1.5 times greater than macroalgae cover. Raw substrate, including shells, shell hash, sediment, bare artificial reef, bare natural reef, and unclassified biota (fuzz, likely filamentous algae or hydroids) represented the remaining cover. For invertebrates, 15 species were classified as temperate,

29 as subtropical, and 11 as tropical. For macroalgae, 5 species were classified as temperate,

24 as subtropical, and 2 as tropical. When two belt or photoquadrat transects were conducted at a reef in a single season, the transects of the same type (e.g., belt or photoquadrat) were averaged as a single replicate for fishes and a single replicate for benthos; unless otherwise noted, results reported below correspond to these average values, and fish results represent demersal, reef-associated fishes and do not include pelagic fishes because of their transient, schooling nature and high abundances.

Abundance of reef-associated, demersal fishes of temperate, subtropical, and tropical ranges differed by reef type (Figure 2.2; Table 2.1). Temperate demersal fishes were 2.8 times more abundant on natural than artificial reefs (Figure 2.2a; Table 2.1b). Subtropical demersal fishes displayed higher abundances on artificial than natural reefs, but this relationship was also influenced by reef depth (Figure 2.2b; Table 2.1a). Shallow and intermediate depth artificial and natural reefs supported similar numbers of subtropical, demersal fishes, whereas deep artificial reefs supported 3.3 times more subtropical fishes than their natural reef counterparts. Similarly, demersal tropical fishes exhibited equal numbers on artificial and natural reefs, but when incorporating reef depth, which was a significant predictor of tropical fish abundance, a pattern emerged (Figure 2.2c; Table 2.1a).

Tropical fish abundance increased with depth on both artificial and natural reefs, and tropical fish abundance was similar by reef type for shallow and intermediate depths. On deep reefs,

however, there were 5.9 times as many tropical fishes on artificial than natural reefs. Fish biomass displayed similar patterns as for fish abundance for demersal fish of certain climate ranges but were more nuanced (Table S3.4). Biomass of demersal temperate fishes was higher on natural reefs than on artificial reefs, but this effect was not statistically significant.

Demersal subtropical fish biomass increased with depth on artificial reefs, yet depth was not a significant predictor of fish biomass. Season did predict fish biomass with highest subtropical fish biomass occurring on artificial reefs during the spring. Interactions among three factors: reef type, depth, and season, influenced tropical fish biomass (Figure S2.1).

Artificial reefs located in deep waters during the spring supported the highest biomass of tropical fishes. Biomass on deep reefs was highest during the spring, followed by summer and then fall. Fall displayed the most even distribution of biomass among shallow, intermediate, and deep reefs.

Abundances of pelagic fishes depended less on reef type and characteristics, such as reef depth, than did demersal fishes. When accounting for site as a random factor within linear mixed effects models, pelagic temperate fish abundances did not differ by reef type

(Figure 2.2d, g; Table 2.1b). In contrast, pelagic subtropical fishes displayed higher abundances on artificial than natural reefs (Figure 2.2e, h; Table 2.1b). This pattern for subtropical pelagic fishes matched the pattern for demersal fishes, with reef depth and the interaction between depth and reef type also influencing subtropical abundance. Shallow and intermediate depths of artificial and natural reefs supported similar numbers of subtropical fishes; deep artificial reefs supported more subtropical fishes than natural reefs. Pelagic tropical fishes exhibited equivalent abundance on each reef type (Figure 2.2f; Table 2.1b).

Fish biomass displayed the same patterns as fish abundance for pelagic fishes (Table S2.4).

Elevated abundances of pelagic fish schools drove patterns in total fish abundance, obscuring patterns of tropical and temperate, but not subtropical, demersal fish abundance (Table 2.1c).

Demersal fish community composition differed between artificial reefs and natural reefs (PERMANOVA F 1,116 = 6.56, P <0.001; Figure 2.3). Abundances of temperate, subtropical, and tropical fishes drove these between-reef differences. Subtropical and tropical fishes contributed to 30% and 23%, respectively, of the differences with elevated abundances on artificial compared to natural reefs (SIMPER analysis). Temperate fishes displayed the opposite trend, occurring more prevalently on natural than artificial reefs, accounting for 6% of the differences between reef types (SIMPER analysis).

Subtropical and tropical demersal fishes not normally found as far north as our study sites and which are documented to consume zooplankton or nekton occurred in higher numbers on artificial reefs, whereas fishes that consume macroalgae exhibited higher abundances on natural reefs (Figure 2.4; Table S2.5). For example, fish species that exhibited higher abundances on artificial reefs included planktivores ( Chromis cyanea (blue chromis) and Thalassoma bifasciatum (bluehead wrasse) and piscivores ( Mycteroperca interstitialis

(yellowmouth grouper), as well as more generalist feeders. Of the generalist feeders, several are known to consume plankton and nekton as part of their diet or are associated with predatory fishes. For example, Anisotremus surinamensis (black margate), although an invertivore, consumes zooplankton and small fishes, and Bodianus rufus (Spanish hogfish) consumes parasites off of large, predatory fishes. In contrast, fish species that exhibited a numerical preference for natural reefs included herbivores ( Stegastes variabilis (cocoa damselfsih) and Sparisoma atomarium (green blotch parrotfish), as well as generalist feeders.

One generalist feeder, Abudefduf taurus (night sergeant) , also consumes macroalgae as part

of its diet and displayed higher numbers on natural reefs. Two-sample t-tests indicated that all of these differences were not significant.

Similar to how fish abundances differed between artificial and natural reefs by climate range, cover of benthos on artificial and natural reefs also differed by species climate range (Figure S2.2, Table 2.2). For the whole benthic community, pooling both invertebrates and macroalgae, cover of temperate and tropical species was 1.8 and 3.2 times higher, respectively, on natural than artificial reefs (Figure S2.2a; Figure S2.2c; Table 2.2). In contrast, subtropical benthic species exhibited elevated (1.7 times) cover on artificial as compared to natural reefs (Figure S2.2b; Table 2.2). When examining just invertebrates and just macroalgae separately, species preferences for reef type by climate range were more nuanced (Figure S2.2d-i; Table 2.2). Temperate and subtropical invertebrate cover was higher on artificial than natural reefs, 3.2 and 2.2 times, respectively, (Figure S2.2d-e; Table

2.2), whereas tropical invertebrate cover was 3.2 times higher on natural reefs. Macroalgal cover was higher on natural than artificial reefs regardless of macroalgal climate range, although only temperate macroalgal cover was significantly higher on natural than artificial reefs (Figure S2.2g-i; Table 2.2).

Bottom-up factors influenced the communities differently on artificial and natural reefs (Figure 2.5). We fit multiple structural equation models; here we report results from the

SEM with the best fit. The final SEM indicated that benthic cover was correlated with three variables (rugosity, temperature, latitude) on artificial reefs and three variables (rugosity, depth, latitude) on natural reefs. Temperate benthos cover increased with rugosity values on artificial reefs (Figure 2.5a), whereas temperate benthos cover increased at poleward latitudes on natural reefs (Figure 2.5b). Subtropical benthic cover increased with depth and latitude on

natural reefs but was unrelated to abiotic variables on artificial reefs. Tropical benthic cover decreased with temperature and latitude on tropical reefs but increased with rugosity on natural reefs. Demersal fish abundance was directly influenced by four variables

(temperature, depth, temperate benthos, latitude) on artificial reefs and one (temperature) on natural reefs. Temperate fish abundance was higher when exposed to colder water temperatures on both artificial and natural reefs. Subtropical fish abundance increased with depth and latitude on artificial reefs, but their abundance remained unrelated to abiotic or biotic factors on natural reefs. Tropical fish abundance increased with cover of temperate benthos, depth, and latitude on artificial reefs. On natural reefs, tropical fish abundance was not driven by abiotic or benthic variables.

Discussion

We provide evidence that reef-associated, demersal fishes with different climate ranges exhibited distinct preferences for either natural or artificial reefs. Temperate fishes occurred in higher abundances on natural, rocky reefs, whereas subtropical fishes and tropical fishes at deep depths (25-35 m) resided in higher numbers on human-made, artificial structures. These differences repeated at the community composition level, with artificial reefs occupied by subtropical and tropical fishes and natural reefs supporting temperate fishes. Of the subtropical and tropical fishes with ranges rarely encompassing reefs as far north as the studied reefs, those fishes inhabiting artificial reefs are known to consume zooplankton and fishes, whereas those occupying natural reefs are documented to consume macroalgae. The benthic community also indicated numerical affinities for particular reef types, as temperate and tropical species exhibited higher cover on natural reefs than artificial

reefs, whereas subtropical species displayed the opposite trend. Bottom-up drivers of the fish community acted differently on artificial versus natural reefs, as fish communities were driven by both abiotic factors and benthic prey on artificial reefs but largely by abiotic factors on natural reefs. These ecological differences suggest that as managers are presented with increasing numbers of human-made structures in temperate marine environments, artificial reefs operate to support fishes from differing climate ranges than natural reefs.

Additionally, artificial reefs may facilitate the movement of subtropical and tropical fishes to temperate waters by providing suitable habitat corridors, where temperature regimes occur within thermal tolerances of the species.

Based on our observed patterns, we consider three explanations for why temperate demersal fishes may prefer naturally occurring rocky reefs over human-made reefs and also for why subtropical and tropical reef-associated fishes on deep reefs exhibited higher abundance on human-made reefs. First, we hypothesize that differential fishing pressure on artificial versus natural reefs contributes to our observations of fishes with particular climate ranges occurring in higher numbers on certain reef types. For instance, artificial reefs are often deployed as fishing sites for recreational fishers, so if fishing pressure is higher on artificial than natural reefs, then elevated fishing pressure on artificial reefs may remove piscivorous, predatory fishes. If tropical fish recruit to these artificial reefs, they face reduced predation thereby facilitating survival. In contrast, since fishing pressure can be lower on natural reefs, numbers of tropical fish could be lower on natural reefs because piscivorous fishes not removed from fishing would actively consume these smaller tropical fishes.

Second, natural reefs represent the ‘historic’ or ‘baseline’ habitat condition of the study area.

Natural rocky reefs formed from exposed bedrock occur in high concentrations south of Cape

Lookout, NC (Riggs et al. 1996). Although some shipwrecks date back to the 1600’s or earlier (Stick 1989), wide-scale marine urbanization of the southeast inner continental shelf is a relatively recent phenomenon. Temperate fishes may occur in higher numbers on natural rocky reefs, given that these were the original types of habitat available to them prior to increased marine urbanization (Floeter et al. 2008). This may confer habitat biases in how species shift in response to marine urbanization (e.g., similar to as positied for climate change in Stuart-Smith et al. , 2015). Third, our findings may be temporally-dependent.

Previous research documented that tropical species occur in higher abundance on natural reefs in Australia (Burchmore et al. 1985). This finding directly contrasts with our finding in

NC that subtropical and, on deep reefs, tropical fishes were more abundant on artificial reefs.

Perhaps this discrepancy relates to the date of our study, more than three decades after

Burchmore et al. (1985). In the two decades between these studies, marine urbanization has become more wide-spread (Bulleri and Chapman 2010, Dafforn et al. 2015), resulting in high habitat availability of novel, artificial structures.

Demersal and pelagic fishes of subtropical ranges exhibited the same pattern of reef- type use. In contrast, demersal and pelagic fishes did not display consistent patterns in how they used each reef type for fishes classified as temperate and tropical. For temperate fishes, demersal fishes preferred natural reefs, whereas pelagic fishes demonstrated no differences in abundance by reef type. Likewise, tropical pelagic fishes occurred in similar numbers on artificial and natural reefs, whereas demersal tropical fishes were more prevalent on deep artificial than natural reefs. We assume that these differences in reef preference between demersal and pelagic fishes relate to life-history traits. Pelagic fishes move transiently among habitats, whereas demersal fishes have smaller home ranges closely linked to particular reef

habitats and surrounding sand flats (Topping and Szedlmayer 2011). Transience of pelagic fishes may contribute to their lack of demonstrated preference for reef type exhibited by both temperate and tropical pelagic fishes. Additionally, these pelagic fishes may not exhibit a marked preference for reef type because factors that are not habitat-associated drive their distribution. For example, pelagic fishes respond to biological productivity in the water column and zooplankton prey to a higher degree than demersal fishes (Champion et al.

2015). Trophic ecology of the one temperate pelagic fish, Menidia menidia (silversides), and the two tropical pelagic species, Scomberomorus cavalla (king mackerel) and Euthynnus alletteratus (little tunny), that resided on the studied reefs support this notion. Menidia menidia consumes plankton, whereas both S. cavalla and E. alletteratus prey on fishes. In contrast, 17 species of pelagic subtropical fishes occurred on the reefs, and some of these, such as tomtate, consume reef-based organisms, as well as those from surrounding sand flats, as part of their diets (Lindquist et al. 1994).

Select subtropical and tropical fishes occurred poleward of their generally characterized, ranges. These fishes have likely been present on the surveyed reefs for years in low and/or variable abundances, avoiding detection and subsequent reporting, especially as annual fluctuations in winter temperatures control fish communities (Whitfield et al. 2014).

Because of their uncommon occurrences as far north as the reefs we studied, however, these fishes provide a case study for understanding how fishes moving poleward, while tracking suitable water temperatures (e.g., tropicalization), may utilize artificial and natural reefs. Our results suggest that prey preferences may drive fish habitat choice in waters outside of their normal geographic ranges but within their thermal tolerances. We posit that when expanding to waters beyond their normal ranges, fishes that consume zooplankton and nekton would

preferentially utilize artificial reef habitats, whereas fishes that consume macroalgae would preferentially occupy natural reef habitats. Because artificial reefs form optimal habitat for planktivorous and piscivorous tropical fishes, artificial reefs have the potential to facilitate tropicalization. Facilitation of tropicalization is plausible, especially as artificial reefs facilitate a similar process, the introduction of invasive species, such as some macroalgae and invertebrates (Bulleri and Airoldi 2005, Dafforn et al. 2012). While subtropical and tropical fish are not invasive species, the habitat-related mechanisms of invasion and tropicalization are likely comparable.

Temperate and tropical benthic species exhibited higher cover on natural reefs, whereas subtropical benthos displayed higher cover on artificial reefs. There were also differences in cover of invertebrates and macroalgae when considering these two broad groups of organisms separately from the overall benthos. Macroalgae of temperate, subtropical, and tropical climate ranges all exhibited higher cover on natural reefs, probably because natural rock substrate is more suitable than human-made substrates. High occupation of macroalgae on natural reefs was mirrored in the occurrence of herbivorous fishes not commonly documented as far north as our study sites for natural over nearby artificial reefs because their food source, macroalgae, existed largely on natural reefs. Temperate and subtropical invertebrates, perhaps because of larval dispersal and more free space on artificial reefs displayed higher cover on these human-made reefs. Tropical invertebrates, however, preferred natural reefs, which was unexpected and assumed to be because physical features of the rocky, temperate reefs are more similar to physical features found in rocky reefs located closer to the equator, including coral reefs.

Community metrics can be driven by bottom-up and top-down mechanisms (Hunter and Price 1992, Power 1992, Pace et al. 1999, Vinueza et al. 2014). Here, we examined bottom-up forcing on fish communities. We propose three explanations for why bottom-up drivers on artificial reefs included biotic and abiotic factors, whereas only one abiotic factor drove the fish community on natural reefs. First, connections between the benthic community and fish community were not as strong as expected given previous findings that benthic community structure affects fishes on each reef type (Lindquist et al. 1994). We hypothesize that our structural equation model (SEM) did not illuminate strong associations between fish and benthic communities because the benthic community accounted for in photoquadrat surveys omits planktonic prey sources, such as zooplankton and phytoplankton, that occur in the water column. Additionally, our models did not include prey resources on sand flats because we did not sample sand flats. Some reef fishes forage on sand flats during the evening (Lindquist et al. 1994), meaning links between benthos and fishes may have been difficult to discern with SEM using variables only measured on reefs. Second, differing physical features between artificial and natural reefs make natural reefs more susceptible to physical forcing. Natural reefs are typically lower relief than artificial reefs, meaning that changes in sedimentation and near bottom currents bury and expose these reefs, especially during large storm events (Renaud et al. 1996, 1997). Artificial reefs, on the other hand, are less susceptible to geological disturbance because these structures occur high enough off the seafloor so as to rarely be covered by sediment. With forecast changes in water temperature, although not in the region studied here, it remains debated how this particular abiotic factor, water temperature, would exert bottom-up forcing on fish communities (Habary et al. 2017).

From our results, we posit that changes in water temperatures would directly drive fish

communities on natural reefs, whereas changing water temperatures on artificial reefs would be mediated first by the benthic community, adding an extra step to influence the fishes. If this hypothesis is supported, then artificial reefs may be less susceptible to changes in the ecological community that would be expected with changing water temperatures.

While artificial reefs are intended to mimic natural reefs, there are known differences in physical features, trophic ecology, and habitat value. Our finding that artificial reefs do not mimic natural reefs in how they provide habitat to species with different climate ranges adds another dimension to our understanding of how artificial and natural reefs differ, primarily that temperate reef-associated fishes occur in higher abundances on natural reefs, whereas subtropical and tropical fishes on deep reefs reside in higher abundances on human-made, artificial reefs. This dichotomy has potential ecological advantages and disadvantages that must be considered when managers deploy additional human-made structures in the environment. For example, installing additional human-made infrastructure, such as offshore wind turbines, may facilitate the spread of species to new ranges, disrupting ecological processes. Regardless, as marine urbanization continues world-wide, the introduction of human-made structures will likely affect temperate, subtropical, and tropical species differently.

Acknowledgements

We thank G. Safrit, G. Sorg, J. Fleming, T. Courtney, M. Kenworthy, A. Poray, D.

Keller, I. Kroll, C. Hamilton, J. Hughes, J. Boulton, T. Dodson, J. Purifoy, S. Davis, C.

Lewis, E. Kromka, E. Ebert, J. Vander Pluym, B. Teer, B. Degan, J. Hackney, R. Muñoz,

D.W. Freshwater, K. Johns, G. Compeau, J. Styron, D. Wells, S. Hall, M. Dionesotes, C.

Marino, I. Conti-Jerpe, E. Weston, M. Wooster, J. McCord, D. Sybert, R. Purifoy and crew from Olympus Dive Center, T. Leonard and crew from Discovery Diving for diving and boating assistance. We thank J. Fodrie, J. Rosman, A. Hurlbert, and S. Fegley for thoughtful reviews and guidance. We thank P. Whitfield for advice on site selection and survey methods. Funding was provided by BOEM under Cooperative Agreement M13AC00006, NC

Coastal Recreational Fishing License Grant (#5115), a NSF Graduate Research Fellowship awarded to A.B. Paxton under Grant No. DGE-1144081, and a P.E.O. Scholar Award to A.B.

Paxton. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the opinions or policies of the US Government, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.

Figures

Figure 2.1. Thirty warm-temperate reefs surveyed on the inner continental shelf. Dark triangles represent artificial reefs, and light circles represent natural reefs. Inset depicts reefs relative to eastern US. Gray lines indicate bathymetric contours, beginning with 10 m deep closest to shore.

Figure 2.2. Abundance of demersal fishes (a-c) and pelagic fishes (d-e) (per 120 m 2) on artificial (dark colored) versus natural (light colored) reefs by fish climate range: (a) temperate (blue), (b) subtropical (green), (c) tropical (red). Reef depth zones are: shallow: 5-

18 m, intermediate: 18-25 m, deep: 25-35 m. Lines and associated error bars (± 1 SE) represent fitted linear mixed effects model with site as a random effect. Lines and error bars are jittered to prevent overlap. Table 2.1 contains model results by for each climate range.

Biomass models appear in Table S2.4.

Figure 2.3. Nonmetric multidimensional scaling ordination of demersal fish community by climate range. Dark triangles represent artificial reefs and light circles indicate natural reefs.

Superimposed ellipses are 50% confidence intervals for artificial reefs (dark color) and natural reefs (light color). Colored text indicates the weighted average of demersal fishes associated with each climate range: tropical, subtropical, and temperate.

Figure 2.4. Mean log (abundance + 0.01) (per 120 m 2) of fishes not commonly documented as far north as surveyed reefs for artificial and natural reefs. Each fish is indicated by a point.

Point color corresponds to climate range: subtropical (green) or tropical (red). Text labels indicate common name. Fishes falling in the gray region of the graph, below the 1:1 line, occur in higher numbers on artificial reefs, whereas fishes in the white region, above the 1:1 line, display higher abundance on natural reefs. Metadata for these fishes are displayed in

Table S2.5.

Figure 2.5. Multigroup structural equation model (SEM) for fish response to abiotic and benthic variables. (a) Fit model. Significant model paths for (b) artificial reefs and (c) natural reefs. Purple and orange indicate positive and negative path coefficients, respectively.

Numbers are standardized path coefficients. Arrow widths are proportional to standardized path coefficients. Fish variables represent demersal fish abundance, whereas benthic variables represent cover. Goodness of fit indices for the multigroup SEM appear at the bottom of the figure.

Tables

Table 2.1. Linear mixed effects model results for fish abundance by climate range. Reef type, depth, season, and the interaction between reef type and depth (Reef type : Depth) were included as fixed effects. Site was included as a random effect. F-values and p-values for each climate range (temperate, subtropical, tropical) are provided. Best models are reported here for (a) demersal, reef-associated fishes, (b) pelagic, water-column associated fishes, and

(c) entire fish community. Bold values indicate significance with significance level displayed as: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Parameters excluded from the model are indicated by ----. Biomass models appear in Table S2.4.

(a) Temperate demersal Subtropical demersal Tropical demersal

F-value p-value F-value p-value F-value p-value Intercept 20.45 <0.0001**** 16.78 0.0001**** 7.89 0.006** Reef type 4.38 0.045* 7.49 0.01* 1.41 0.25 Depth ------2.84 0.06 2.43 0.09 Reef type : depth ------4.72 0.01* 3.13 0.049*

(b) Temperate pelagic Subtropical pelagic Tropical pelagic

F-value p-value F-value p-value F-value p-value Intercept 4.47 0.04* 64.61 <0.0001**** 2.09 0.15 Reef type 0.11 0.75 18.69 <0.001*** 2.11 0.16 Depth ------18.28 <0.0001**** ------Reef type : depth ------10.4 <0.001*** ------

F-value p-value F-value p-value F-value

Intercept 6.38 0.01* 78.32 <0.0001**** Intercept 6.38

Reef type 0.04 0.84 28.79 <0.0001**** Reef type 0.04 Depth ------18.41 <0.0001**** Depth ---- Reef type : Reef type : depth ------17.23 <0.0001**** depth ----

Table 2.2. Linear mixed effects model results for percent cover of the benthic community by climate zone. Reef type, depth, season, and the interaction between reef type and depth (Reef type : Depth) were included as fixed effects. Site was included as a random effect. F-values and p-values for each climate range (temperate, subtropical, tropical) are provided. Best models are reported here for (a) entire benthic community, (b) benthic invertebrates, and (c) macroalgae. Bold values indicate significance with significance level displayed as: * p <

0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Parameters excluded from the model are indicated by ----. Other tested models appear bin Table S2.4.

(a) Temperate benthos Subtropical benthos Tropical benthos

F-value p-value F-value p-value F-value p-value Intercept 51.69 <0.0001**** 238.16 <0.0001**** 16.93 0.0001*** Reef type 3.52 0.07 17.34 0.0003*** 4.00 0.055

Season 9.84 <0.0001*** ------Depth ------0.68 0.51 ------

Reef type : Season 3.41 0.03* ------

Reef type : Depth ------5.57 0.005** ------

(b) Temperate invertebrates Subtropical invertebrates Tropical invertebrates

F-value p-value F-value p-value F-value p-value Intercept 10.37 0.002** 106.97 <0.0001**** 16.40 0.0001***

Reef type 3.46 0.07 16.86 0.0003*** 3.80 0.06 Season 3.68 0.03* ------Depth ------0.68 0.51 ------Reef type : Season 2.6 0.08 ------

Reef type : Depth ------5.57 0.005** ------

F-value p-value F-value p-value F-value p-value Intercept 42.70 <0.0001**** 37.9 <0.0001**** 1.01 0.32 Reef type 7.59 0.01* 0.28 0.60 0.87 0.36 Season 7.13 0.001** 8.39 0.0005** ------Depth ------

Reef type : Season 4.04 0.02* 3.87 0.02* ------Reef type : Depth ------

69 CHAPTER 3: AGGREGATIONS OF PLANKTIVOROUS FISHES AROUND SHIPWRECKS RELATE TO SPATIAL PATTERNS IN ADJACENT TROPHIC LEVELS 3

Summary

Aggregations of planktivorous fishes on naturally-occurring structured marine habitats, such as coral reefs and rocky reefs, often induce spatial patterns in their prey and predators. Whether similar spatial patterns occur around submerged artificial structures, often with more abrupt topographies than natural habitats, remains less well understood. We conducted acoustic surveys of fifteen shipwrecks to test whether consistent spatial patterns in planktivorous fishes, their prey (zooplankton), and their predators (piscivorous fishes) were present on artificial structures. We addressed three specific questions: 1) where are planktivorous fishes and both their prey and predators relative to shipwrecks, 2) how are distributions of planktivorous fishes and their prey and predators related around shipwrecks, and 3) does the spatial distribution of fishes on shipwrecks change with water current? We found that aggregations of planktivorous fishes were patchy, and their mean center occurred an average of 39 m away from shipwreck edges. Whereas zooplankton occupied nearly 25% of surveyed areas around shipwrecks, aggregations of planktivorous fishes were spatially distinct from high concentrations of zooplankton prey. Aggregations of planktivorous fishes also did not co-occur with piscivorous fishes, as piscivorous fishes concentrated closer to

3 A version of this chapter is in preparation for submission to a peer-reviewed journal as: Paxton, A.B., J.C. Taylor, C.H. Peterson, S.R. Fegley, and J.H. Rosman. Aggregations of planktivorous fishes around shipwrecks relate to spatial patterns in adjacent trophic levels.

70 shipwrecks. Locations of planktivorous fishes were unrelated to current magnitude but occurred upstream of shipwrecks over half of the time. Their predators, piscivorous fishes, were further from shipwrecks when currents were weak and remained closer when currents were strong. Because spatial patterns in planktivorous fishes related to distributions of prey

(zooplankton) and predators (piscivorous fishes), our findings indicate that installing artificial structures may influence spatial patterns across adjacent trophic levels.

Introduction

Global rates of urbanization are increasing (Seto et al. 2012), and coastlines experience particularly concentrated development since one-third of humans reside within

100 km of coasts (Dugan et al. 2012). Urbanization, however, is no longer restricted to terrestrial and shoreline environments and now extends beneath the surface of the ocean, a phenomenon known as ‘marine urbanization’ or ‘ocean sprawl’ (Bulleri and Chapman 2010,

Dafforn et al. 2015), characterized by the introduction of artificial structures to coastal waters. These artificial structures range from wind turbines and oil- and gas- rigs to artificial reefs and shipwrecks. The introduction of artificial structures in marine environments can drive ecological changes ranging from altered ecosystem connectivity (Bishop et al. 2017) and degradation of natural habitats (Bulleri and Chapman 2010) to community-level shifts

(Ferrario et al. 2016) and the spread of invasive species (Airoldi et al. 2015). Understanding ecological functions of artificial structures, especially in comparison to natural habitats, is important.

Many artificial structures emerge from the seafloor well into or extend throughout the water column and can be considered ‘abrupt topographies’ (Genin 2004). On naturally abrupt

topographies, such as coral reefs (Hamner et al. 1988, 2007, Hanson 2011, Leichter et al.

2013) and rocky reefs (Bray 1980, Gaines and Roughgarden 1987), aggregations of planktivorous fishes consistently occur. Here, we use the term ‘aggregation’ to refer to groups of animals formed by any mechanism, whether passive or active (Simard et al. 1986,

Genin 2004). Aggregations of planktivorous fishes spatially associate with zooplankton prey

(Hamner et al. 1988, Genin 2004). For example, on coral reefs, planktivorous fishes can form a ‘wall of mouths’ by concentrating in large schools near up-current sections of reefs, consuming zooplankton transported towards reefs (Hamner et al. 1988, 2007). Similarly, on rocky topographies with kelp forests, planktivorous fishes, such as juvenile rockfishes, consume zooplankton as they are transported towards kelp forests (Gaines and Roughgarden

1987). In both coral reefs and rocky reefs, zooplankton become depleted upon reaching dense schools of planktivorous fishes that consume zooplankton (Gaines and Roughgarden 1987,

Hamner et al. 1988); this depletion can lead to ‘holes’ in zooplankton biomass, creating increased patchiness (Genin et al. 1988). Similar spatial patterns occur near deeper abrupt topographies, such as seamounts, shelf breaks, and canyons, where aggregations of planktivorous fishes consume zooplankton, often generating pronounced spatial gradients in zooplankton distribution (Genin 2004). Aggregations of planktivorous fishes near both shallow and deep naturally-occurring, structured habitats also have effects that propagate up food webs because piscivorous fishes and other common predators, such as marine mammals, are attracted to and actively prey on planktivorous fishes (reviewed in: Genin

2004, Prairie et al. 2012).

Multiple mechanisms, such as predator-prey dynamics and physical forcing (e.g., water currents) have been proposed to explain why elevated concentrations of planktivorous

fishes occur around habitats classified as naturally abrupt topographies (Genin 2004, Genin et al. 2005). In shallow natural habitats, where zooplankton vertical migration is often less pronounced than in deep habitats, the most widely accepted explanation for elevated densities of planktivorous fishes is that as water currents transport zooplankton towards both unstructured and structured habitats, planktivorous fishes are attracted to structured habitats because they can both feed on incoming zooplankton and use structure for refuge (Genin

2004). While it can be competitively advantageous for planktivorous fishes to feed as far into the oncoming stream of zooplankton as possible, there is a tradeoff between acquiring prey and seeking refuge from predators, with planktivorous fishes experiencing higher foraging success but also reduced protection greater distances from structured habitats (Bray 1980). In addition to the food-refuge tradeoff, planktivorous fishes also experience a tradeoff as a function of water current strength. When water currents are too strong, planktivorous fishes are unable to hold their position within the water column and often retreat to find refuge in holes and crevices located closer to the seafloor (Webb 1998, Liao 2007, Johansen et al.

2008).

Because artificial structures offer similarly abrupt, or in some cases more abrupt, topographies to coral reefs, rocky reefs, and other types of emergent natural habitats, the expectation would be that planktivorous fishes would likewise aggregate near artificial structures. Observations on artificial reefs support the hypothesis that these planktivorous fishes aggregate near artificial structures (Rilov and Benayahu 1998, Arena et al. 2007,

Simon et al. 2013, Champion et al. 2015). Comparisons of numbers of planktivorous fishes between artificial reefs and nearby natural reefs suggest that sizes and likely effects of aggregations on artificial structures may exceed those around naturally-occurring reefs. For

example, planktivorous fishes accounted for 54% of total fish abundance on artificial reefs but only 27% on natural reefs in Florida, USA (Arena et al. 2007). Similarly, planktivorous fishes had higher biomass on artificial than natural reefs in Brazil (Simon et al. 2013). When present on artificial structures, planktivorous fishes are often among the most abundant fishes, as they account for the top ten most abundant fish species on artificial reefs in the Red

Sea (Rilov and Benayahu 1998), the top two on artificial reefs in North Carolina (NC), USA

(Lindquist and Pietrafesa 1989), and seven of the ten most abundant species on reefs in

Florida, USA (Arena et al. 2007). Recent studies from Australia estimate that zooplanktivory makes artificial reefs among the habitats with the highest fish production in the world

(Champion et al. 2015, Smith et al. 2016). Despite the documented prevalence of planktivorous fishes on artificial structures, it is unknown whether spatial patterns in planktivorous fishes occur consistently on artificial structures and how aggregations of planktivorous fishes relate to spatial distributions of other trophic groups.

The goals of this study were to document spatial distributions of planktivorous fishes around shipwrecks, a type of artificial structure, and to determine whether spatial patterns in planktivorous fishes relate to distributions of their prey (zooplankton) and predators

(piscivorous fishes). We specifically addressed the following questions:

1) Spatial Location: Where are planktivorous fishes, their zooplankton prey and their

piscivorous predators relative to shipwrecks? How much area does each of these

groups occupy, where is its spatial center, what is its variation in spatial distribution,

does it occur in spatial patches, and, if so, where are clusters of high- and low-

organism concentrations?

2) Spatial Relationships: How are spatial distributions of planktivorous fishes,

zooplankton, and piscivorous fishes related around shipwrecks?

3) Water Current: Does the spatial distribution of planktivorous fishes on shipwrecks

change with the magnitude and direction of water current?

Answering these questions will determine whether patterns in planktivorous fish distributions and distributions of their associated prey and predators occur consistently on artificial structures, such that effects of artificial structures influence spatial patterns across multiple trophic levels.

Materials and Methods

Site Selection

The artificial structures for this study are fifteen shipwrecks that rest on the continental shelf of North Carolina (NC), USA (Figure 3.1, Table S3.1), an area known as the ‘Graveyard of the Atlantic’ (Hoyt et al. 2014). The shipwrecks were selected because each rests on the continental shelf of NC in shallow water (< 100 m) and has been submerged for over 50 yrs. The shipwrecks range from 18.9 m to 98.6 m deep and sank between 1862

(USS Monitor ) and 1957 (USS Tarpon ). Because the selected shipwrecks are distributed haphazardly across the continental shelf, have been submerged for multiple decades, and also form reefs that support a diversity of fishes (Paxton et al. 2017), these shipwrecks provide an opportunity to quantify planktivorous fish, zooplankton, and piscivorous fish distributions.

Surveys Conducted

The fifteen selected shipwrecks were surveyed onboard the NOAA ship Nancy Foster from 1-6 November 2016 and 5-14 July 2017 using multiple hull-mounted instruments to measure bathymetry, zooplankton, several size-classes of fish, and currents (Table S3.2).

Each of the fifteen wrecks was surveyed once and several were surveyed up to three times.

During 1-6 November 2016, twelve of the fifteen shipwrecks were each surveyed once using splitbeam echosounder (SBES) to measure zooplankton concentration and fish densities and multibeam echosounder (MBES) to measure bathymetry. From 5-14 July 2017, five shipwrecks, including some of those previously surveyed in November 2016, were surveyed using not only SBES and MBES but also an acoustic Doppler current profiler (ADCP) to measure currents. During the July 2017 surveys, several of the five shipwrecks were surveyed multiple times to examine whether shipwreck-specific patterns occurred repeatedly.

Twenty surveys were conducted in total, all during the daytime between sunrise and sunset.

The survey vessel travelled at a speed of about 7 kts during the surveys. Position was logged using differential GPS and Applanix POS M/V motion sensor. All MBES, SBES, and ADCP data were spatially referenced to North American Datum 1983 Universal Transverse

Mercator Zone 18 North.

Mapping Shipwrecks

The MBES (Reson 7125 SV2, 400 kHz) collected multibeam bathymetry of each shipwreck at fine resolution (< 1 m). The resolution was determined for each shipwreck, based on the size and depth of the shipwreck, to ensure optimal coverage; this resulted in resolutions ranging from 0.15 m to 0.40 m (Table S3.2). The MBES operated with a 130º

beam angle, emitting 512 beams downward into the water column. MBES data were corrected for changes in the speed of sound throughout the water, as well as tidal influence, static draft, latency, roll, pitch, yaw, and sensor offsets.

Bathymetry data for each shipwreck were processed within ArcMAP version 10.5

(ESRI 2016) using functions from the Spatial Analyst Extension (ESRI 2016). Each shipwreck was manually delineated from the surrounding sand as a polygon, and then the centroid of each delineated shipwreck was identified. The resulting polygon and centroid were each exported as shapefiles used to determine distances of organisms to shipwreck edges and the shipwreck center, respectively.

Bathymetry maps of each shipwreck were also used onboard the survey vessel to create acoustic survey transect lines along which zooplankton, fishes, and water currents would be later measured. Transect lines were created overlying the bathymetry maps in

HYPACK version 16.0 (Xylem Inc. 2016) so that multiple transect lines extended parallel to the major axis of each shipwreck. Perpendicular to these major-axis lines, we established cross lines along the minor axis of each shipwreck. Spacing among lines and the lengths of the major-axis and minor-axis lines were determined based on the size of the shipwreck to ensure high spatial coverage necessary for quantifying zooplankton and fishes using the

SBES with a narrow beam angle (Figure S3.1).

Quantifying Fishes & Zooplankton

Fishes and zooplankton were measured along the transect lines determined from multibeam bathymetry using the SBES (SIMRAD EK60; 7º beam angle). The SBES emitted sound pulses downwards into the water at three frequencies: 38 kHz, 120 kHz, and 200 kHz

to detect zooplankton, fishes, the shipwreck structure, and the surrounding seafloor. For the

38 kHz frequency, the pulse length was 0.256 µs, whereas for the 120 kHz and 200 kHz frequencies, the pulse length was 0.128 µs. To reduce acoustic interference among sonar transducers, the SBES emitted pings when triggered by the MBES. As such, the ping rate for the SBES was determined by settings of the MBES. The SBES was calibrated using a tungsten carbide sphere to enable accurate measurements of fish size (Foote et al. 1987).

SBES data, displayed as echograms, were processed with Echoview version 8.0 (Myriax

Software Pty. Ltd 2017) to quantify the spatial distributions of zooplankton and fishes around each shipwreck (Text S1).

Within Echoview, individual fishes and schools of fishes were identified from the 120 kHz SBES data. Individual fishes were identified with a fish tracking algorithm that classified sequential acoustic targets as discrete fishes (Echoview 2017). Schools of fishes were delineated using a school detection algorithm (Barange 1994). Single acoustic targets corresponding to fishes in school perimeters, where fishes were loosely aggregated and visible as discrete fishes, were also identified using an acoustic threshold, which permitted calculation of fish size for schools. Data for individual fishes and fish schools were exported from Echoview in cells measuring 5 m (horizontal) x 1 m (vertical, depth) using a threshold of -65 dB. These exported data were processed within R, version 3.3.2 (R Core Team 2016) to calculate individual and schooling fish density per cell on a linear scale. The resulting individual fish and schooling fish densities were summed to obtain total density of fishes, as well as fish densities corresponding to three size classes: small (< 11 cm fish length), medium (11-29 cm fish length), and large (>29 cm fish length). Size class were assigned

using the general logarithmic equation for the relationship between mean target strength

(TS mean ) and fish total length ( Lfish ) based on Love (1977):

.⁄ . = 10 1 The resulting data included a per cell value corresponding to the density of both total fishes and fishes separated by size classes (fish m -3). See Text S1 for additional details on delineating and quantifying fishes.

Zooplankton were identified from the SBES data using decibel differencing by comparing the mean volume backscattering strength (Sv) detected by echosounders operating at multiple discrete frequencies (Higginbottom et al. 2000, Korneliussen and Ona 2003).

Decibel differencing was conducted in Echoview using data from the 38 kHz and 120 kHz transducers. Prior to decibel differencing, the 120 kHz data were resampled so that their pulse length matched the pulse length of the 38 kHz transducer data. Zooplankton were delineated from the 38 kHz data as areas where acoustic signatures from 38 kHz data were less than the corresponding acoustic signatures from 120 kHz data. Following decibel differencing, zooplankton delineated in the 38 kHz data were exported from Echoview in 5 m

(horizontal) x 1 m (vertical, depth) cells using a -65 dB threshold. Within R, zooplankton data were converted from logarithmic form (dB re 1 m -1) to linear form (m -1) to calculate the volume backscattering coefficient ( sv) of zooplankton per cell, which we refer to as the zooplankton concentration. See Text S1 for additional details on quantifying zooplankton.

Quantifying Water Currents

The hull-mounted ADCP (Teledyne RDI Ocean Surveyor 150 kHz) recorded the magnitude and direction of water currents in 2-m vertical bins above and around shipwrecks.

ADCP data were processed in Matlab (The Mathworks Inc. 2017) by separately averaging the east-west and north-south velocity components throughout the entire water column and across the spatial extent of the ADCP survey to obtain the mean velocity components for each survey. These depth-averaged mean velocity components were used to calculate the current magnitude and direction for each survey. See Text S2 for additional details on quantifying water currents.

Data Analyses

Fish, zooplankton, and current data were analyzed in R. Per cell values of zooplankton relative concentration (m -1), small fish density (fish m -3), medium fish density (fish m -3), and large fish density (fish m -3) were each summed vertically across the water column. These resulting-depth-collapsed values for each (zooplankton area concentration (m 2) and fish area density (m 2) per size class) were used for spatial analyses. Here, we call ‘small fishes’ planktivorous fishes, and we call ‘medium fishes’ and ‘large fishes’ piscivorous fishes. We report medium fishes separately from large fishes because the two different size classes help distinguish between ecological roles of these fishes. All analyses were conducted for each of the twenty surveys, unless otherwise noted. Many analyses included calculation of spatial indicators developed specifically for geostatistical data like our fisheries acoustics data

(Woillez et al. 2007, 2009). These spatial indicators are applicable for data like ours that inherently contain autocorrelation (Woillez et al. 2007, 2009). See Text S3 for additional details on data analyses.

Spatial Location: To describe the general spatial location of zooplankton, small fishes, medium fishes, and large fishes around shipwrecks, we calculated six spatial indicators for

each organism: 1) positive area (PA), 2) spreading area (SA), 3) equivalent area (EA), 4) microstructure (MI), 5) center of gravity (CG), and 6) inertia (I) (Table 3.1). The first three indicators, PA, SA, and EA, quantify the area an organism occupies accounting for presence only, density variations, and equivalent density, respectively (Table 3.1). We scaled PA, SA, and EA to account for differing total areas among shipwreck surveys. The fourth indicator, microstructure, describes the irregularity in organism distribution (Table 3.1). The last two indicators, CG and I, describe an organism’s mean center and its dispersion around the mean center. These indicators were calculated using equations and corresponding functions from

Woillez et al. (2007, 2009; Table 3.1). We tested differences in indicators among organisms using one-way analyses of variance (ANOVAs) followed by post-hoc Tukey HSD tests.

In addition to calculating six spatial indicators, we quantified the patchiness of each organism by identifying and counting spatial patches where organism density exceeded 10% of the sum of organism density at all sample locations in a survey. Patches were detected using the ‘SI.patches’ function from the ‘RGeostats’ package (Renard et al. 2017). We tested differences in patchiness among organisms using one-way analyses of variance (ANOVAs) followed by post-hoc Tukey HSD tests. To further describe organismal distribution, we identified clusters of high density (hot-spots) and low density (cold-spots). Hot-spots and cold-spots were detected using the Getis-Ord Gi* statistic (Getis and Ord 1992, Ord and

Getis 1995), calculated based on the k-nearest neighbors using the ‘localG’ function within the ‘spdep’ package (Bivand and Piras 2015). The resulting z-values indicated whether and where statistically significant hot-spots ( z-value > 1.96) or cold-spots ( z-value < -1.96) occurred.

Spatial Relationships: We quantified spatial relationships between organism pairs (e.g., zooplankton and small fishes, small fishes and medium fishes, etc.) at both global and local scales (Table 3.1). At the global scale of each survey, we calculated the global index of collocation (GIC; Table 3.1). To complement the global metrics, we calculated two metrics to describe spatial overlap at the local scale of sampling points: 1) the local index of collocation (LIC, Table 3.1) and 2) the co-occurrence (CO, Table 3.1). While the GIC compared organism distribution over the entire extent of each survey, the local metrics assessed overlap at the level of each sample (e.g., 5 m x 1 m depth-collapsed sample).

Together, the GIC, LIC, and CO values provide complementary means to interpret relationships between two organisms that were appropriate for fisheries acoustics data like ours that inherently contain autocorrelation (Bez and Rivoirard 2000, Woillez et al. 2009,

Saraux et al. 2014). To ensure that the calculated values were not extreme, we bootstrapped the GIC, LIC, and CO metrics to verify that the mean of each across the twenty surveys approximated the mean obtained from bootstrapping with 1,000 samples.

Water Current: We used linear regressions to investigate relationships between fish location and current magnitude. First, we regressed distances between the mean center of each fish size class and the nearest shipwreck edge against current magnitude. Second, because small fishes were unrelated to water current and given prior research suggesting that predator-prey dynamics influence planktivorous fish aggregations (Holzman et al. 2005), we fit additional linear regressions investigating relationships between the distance of small fishes from wreck edges and predictor variables representing apparent predation risk. The predictor variables for apparent predation risk were the number of medium, large, and both medium and large fishes present within the shipwreck, designated as the manually delineated

polygon corresponding to the shipwreck structure, and 5-m outward of the delineated shipwreck structure. We conducted assessments of fit by comparing observed distances of small fishes from edges to those of the estimated distribution.

To determine how water current direction influenced small fish density, we visualized the location of the mean center of small fishes relative to upstream and downstream portions of the shipwreck. We also calculated the total density of small fishes located upstream and downstream on each survey and fit a beta regression using the ‘betareg’ package (Cribari-

Neto and Zeileis 2010) to examine whether the proportion of fishes located upstream or downstream related to current magnitude. Since the upstream and downstream fishes were classified based on their orientation relative to the current direction, these tests incorporated both current direction and current magnitude.

Results

Spatial Location

Zooplankton occurred throughout nearly 25% of each survey, as indicated by high spatial area of zooplankton presence (PA; Table 3.2). When accounting for variations in zooplankton concentration (SA, Table 3.2) or when assuming a uniform distribution for present zooplankton (EA, Table 3.2), zooplankton occupied smaller areas than the areas of their presence alone, indicating that finer-scale variations in zooplankton concentration occurred. Even with zooplankton present at nearly 1 in 4 sampling locations, zooplankton concentrations were variable and irregularly structured across the survey (MI, Table 3.2), again indicative of underlying finer-scale spatial patterns in their distribution. Because of the high area occupied by zooplankton, the mean center and variation in zooplankton spatial

distribution were influenced by survey extent. Given the influence of survey extent on zooplankton distribution, we do not report further on zooplankton here but instead revisit their finer-scale spatial patterns below.

Fishes of all size classes were present across 2 – 3% of surveyed areas (PA; Table

3.2). When incorporating density variations, rather than just fish presence, small fishes occupied more area than medium or large fishes (SA; Table 3.2). If each fish size class was assumed to have uniform density, then this pattern persisted (EA; Table 3.2). For all size classes of fishes, spatial structures of their distributions were irregular and poorly structured, suggesting that fine-scale variability occurred (MI; Table 3.2).

Because spatial indicators demonstrated that fishes, in comparison to zooplankton, occurred over less of the surveyed areas, two additional spatial indicators, the mean center

(CG) and associated dispersion around the mean center (I), could be meaningful. Distances between the mean center and shipwreck edges, as well as the dispersion, differed among small fishes, medium fishes, and large fishes (ANOVAs p < 0.01; Figure 3.2; Figure S3.2;

Table 3.2). Small fishes occurred further from shipwreck edges than either medium or large fishes (Figure 3.2a; CG = 39.1 ± 8.2 m; Tukey HSD p < 0.05). Medium fishes were located more than twice as close to reef edges as small fishes (Figure 3.2b; CG = 15.1 ± 7.5 m).

Large fishes occurred more than five times closer to structure edges than small fishes and nearly twice as close as medium fishes (Figure 3.2c; CG = 7.5 ± 3.8 m). Small fishes were highly dispersed around their mean center (I =14,210 ± 2,670 m). Medium and large fishes were less dispersed around their mean centers than small fishes (I medium = 6,600 ± 2,680 m,

I large = 3,540 ± 920 m).

Patchiness differed among organisms (ANOVA p = 0.04; Table 3.2). Patchiness of small and medium fishes was characterized by similarly high numbers of mean patches

(small fishes: 27 ± 8 patches, medium fishes: 20 ± 8 patches; Tukey HSD p > 0.05). The mean number of patches for small and medium fishes exceeded that of large fishes, (Tukey

HSD p < 0.05); large fishes occurred in fewer patches than small or medium fishes (8 ± 3 patches). Zooplankton patchiness (87 ± 39 patches) exceeded the patchiness of all size classes of fishes (Tukey HSD p < 0.05).

When we further quantified organism patchiness by detecting statistically significant clusters of high concentrations (hot-spots) and low concentrations (cold-spots), general patterns emerged (Figure 3.3; Figure S3.3). For fishes of all size classes, hot-spots were present, but cold-spots were absent. Because the Getis-Ord Gi* is calculated as the ratio of the local average to the global average of density, the lack of cold-spots suggests that whereas fish density is globally low, it is high in localized areas, which means that areas without fishes are no different than global lows so are not cold-spots (Fortin and Dale 2005).

Hot-spots corresponding to small fishes were located further from shipwrecks than hot-spots of either medium fishes or large fishes (Figure 3.3b-d).

Even though zooplankton were present over nearly 25% of each surveyed area, the

Getis-Ord Gi* method for cluster detection was appropriate because it accounted for variations in low and high zooplankton concentrations. As for fishes, the Getis-Ord Gi* values indicated that zooplankton occurred in hot-spots but not cold-spots (Figure 3.3a;

Figure S3.3). Zooplankton hot-spots were usually located on one side of shipwrecks (19 out of 20 surveys) either on the opposite side from small fish hot-spots (9 out of 20 surveys) or between small-fish hot-spots and shipwrecks (10 out of 20 surveys; Figure 3.3a-b; Figure

S3.3). If zooplankton and small fishes occurred in the exact same location within a sampling cell, however, we would be able to detect fishes but not zooplankton, which could affect our interpretation of these results.

Spatial Relationships

At the scale of each survey, known as the global scale, where the spatial distribution was characterized entirely by an organism’s center of gravity and inertia, spatial distributions of small fishes highly overlapped with zooplankton (Figure 3.4a; Figure S3.2; GIC 0.84 ±

0.04). Small fishes also overlapped globally with both medium fishes (GIC 0.75 ± 0.06) and large fishes (GIC 0.77 ± 0.05) but to a lesser extent than the global overlap between small fishes and zooplankton. Medium and large fishes were globally collocated (GIC 0.83 ± 0.04), and both also demonstrated a high degree of global overlap with zooplankton (GIC 0.90 ±

0.03 medium, GIC 0.89 ± 0.03 large). Given that zooplankton occurred throughout nearly

25% of each survey, GIC values between zooplankton and other organisms were high.

At the scale of sampling points within surveys, known as the local scale, high densities of small fishes minimally overlapped with high zooplankton concentrations (Figure

3.4b; LIC 0.11 ± 0.05 mean). Similarly, fishes regardless of size class did not occur at similar density levels at the same sampling locations. For example, small fishes and medium fishes

(LIC 0.09 ± 0.02), as well as medium and large fishes (LIC 0.09 ± 0.04), exhibited near zero levels of local collocation, as did small and large fishes (LIC 0.05 ± 0.02). When removing the influence of organism density and instead examining presence versus absence of organisms, small fishes and zooplankton co-occurred locally at 9.2 ± < 0.1 % of the sampling points where at least one of the two organisms was present (Figure 3.4c). Fish co-occurrence

was lower than the co-occurrence between zooplankton and small fishes, with small and medium fish presence co-occurring locally at a rate of 6.1 ± < 0.1 %, and medium and large fishes at 8.9 ± < 0.1 %. Means for organism overlap metrics at both global and local scales matched the center of distributions from the bootstrapping procedure, indicating that the values presented here were representative (Figure S3.4).

Water Current

Distances from mean center of small fishes to shipwreck edges displayed no pattern with current magnitude ( p > 0.05; Figure 3.5a). Similarly, distances of medium fishes from structure edges were unrelated to current magnitude ( p > 0.05; Figure 3.5b). Distances of large fishes from shipwreck edges, however, related to current magnitude ( p = 0.04; Figure

3.5c). Large fishes resided closer to shipwreck edges, and in some cases, directly above the shipwreck, with stronger currents. With weaker currents, large fishes were located further from shipwreck edges. Distances of small fishes from structure edges showed no pattern relative to densities of predators (medium fishes, large fishes, both medium and large fishes) residing on structure and 5 m outward of structure ( p > 0.05; Figure 3.5d). We recognize that there is high leverage in these linear regressions because of extreme values drive three of the regressions (Figure 3.5 a-b, d).

When we examined relationships between small fishes and current direction, we found that the mean center of small fishes generally occurred upstream of shipwrecks (6 out of 9 surveys, one survey without small fish) (Figure S3.2). There was no relationship between the proportion of fishes located upstream or downstream of the shipwreck and current magnitude (beta-regressions p > 0.05).

Discussion

We found that planktivorous fishes consistently aggregated around the shipwrecks surveyed in this study. Aggregations of planktivorous fishes were patchy and their mean center occurred an average of 39 m from structure edges. Despite zooplankton occurring throughout nearly 25% of surveyed areas around shipwrecks, aggregations of planktivorous fishes were spatially distinct from high concentrations of zooplankton prey. Instead, planktivorous fishes occurred either on the opposite side of shipwrecks from high zooplankton concentrations or on the same side but further from shipwrecks than zooplankton. Aggregations of planktivorous fishes did not co-occur with piscivorous fishes, which concentrated closer to shipwrecks, especially with stronger water currents. More than half of the time, planktivorous fish aggregations resided on upstream sides of shipwrecks, but their position was unrelated to current magnitude. Because spatial distributions of planktivorous fishes were related to distributions of prey (zooplankton) and predators

(piscivorous fishes), our findings suggest that effects from added structure from shipwrecks, and by extension, other types of artificial structures, influence spatial patterns in adjacent trophic levels.

Our finding that planktivorous fishes occupied localized patches positioned an average of 39 m (mean center) from shipwreck edges is consistent with the expectation that artificial structures aggregate planktivorous fishes. The distance of 39 m between planktivorous fishes and shipwrecks approximates distances found on other reefs in the same geographic area. For example, on artificial reefs near our study sites, the fish community, of which the most abundant species was a partially planktivorous species ( Haemulon

aurolineatum ), resided close (77% within 30 m) to artificial reefs (Rosemond et al. 2018). In

Australia, planktivorous fishes exhibit similarly dense aggregations, although the dominant species ( Atypichthys strigatus ) forages a maximum of 4 m from the reef structure (Champion et al. 2015). This discrepancy in the distance of planktivorous fishes from artificial structures is likely a product of differing species-specific foraging ranges in distant geographic locations, as well as landscape level differences in habitat availability and habitat types between the southeastern USA and eastern Australia. Regardless of the precise distance away from artificial structures, the repeated high densities of planktivorous fishes that we documented adds to a growing body of evidence that artificial reefs and other artificial structures function to support elevated densities of planktivores (Arena et al. 2007, Champion et al. 2015).

Aggregations of planktivorous fishes require zooplankton for prey resources. Our finding that at local scales, zooplankton concentrated either between planktivorous fishes and shipwrecks or on the opposite side of shipwrecks from planktivorous fishes has at least two implications. First, this finding indicates that low to normal background concentrations of zooplankton co-occur with planktivorous fishes. Second, it demonstrates that zooplankton concentrations depend on spatial scale, as evidenced by marked patches of high- zooplankton concentration emerging at fine spatial resolution but being obscured at broader resolutions where zooplankton appeared more ubiquitous. Previous models have demonstrated and empirical data have supported the notion that artificial reefs in Australia can support high levels of zooplanktivory (Champion et al. 2015). Considering we sampled multiple shipwrecks, which are similar to artificial reefs, our study suggests that shipwrecks and

artificial reefs located in NC, USA, with differing fish communities than those in Australia, host patchy distributions of zooplankton that planktivorous fishes require.

Dense schools of planktivorous fishes on shipwrecks have the potential to be prey resources for piscivorous fishes. In our study, piscivorous fish predators clustered near shipwrecks. Similar activity where predators remain close to edges has been observed on piers, likely because the predators can hide in the shade provided by the pier structure and then attack prey (Able et al. 2013). A similar phenomenon may occur on shipwrecks because predatory fishes located closer to structures could hide in shaded areas of structures, such as overhangs and interiors. Similarly, piscivores could physically hide behind emergent areas of the shipwreck structure, rather than solely shaded areas, especially when light is dim. On several artificial reefs near the shipwrecks that we studied, we have visually observed that predators remain close to the edges and center of structures (A.B. Paxton, personal observation). This supports the hypothesis that piscivorous fishes may optimize foraging success by lurking near structures to more effectively hide from and surprise potential prey fishes.

Elevated densities of planktivorous fishes occurred at the same locations as low concentrations of zooplankton, suggesting that patterns in planktivorous fishes may propagate down the trophic web to create spatial gradients in zooplankton. Based on similar studies on naturally occurring reefs with abrupt topographies (e.g., Hamner et al. 1988,

2007), we expected that the relationship between zooplankton and planktivorous fishes would be more pronounced near shipwrecks, such that ‘holes’ in zooplankton concentration would be evident. While patterns existed at local scales, we posit that the magnitude of the relationship may have been obscured at broader scales because indicators of zooplankton

distributions, such as the mean center and dispersion, were dependent on sampling extent.

Extending transects and associated sampling efforts further away from artificial structures could provide more rigorous evaluations of spatial scales at which spatial relationships between zooplankton and planktivorous fishes are most pronounced. The localized trend that we documented between these organisms on twenty surveys across fifteen shipwrecks is consistent with stable isotope findings from four artificial reefs in the Mediterranean that artificial structures host pronounced pelagic pathways between zooplankton and planktivorous fishes (Cresson et al. 2014), as well as a model and field data from one artificial reef in Australia that zooplanktivory is a key process on artificial structures

(Champion et al. 2015).

Planktivorous fishes occurred in the same global vicinity as their piscivorous fish predators but locally these two trophic groups did not co-occur. Our classification of piscivorous fishes includes both medium fishes and large fishes. We caution that because fisheries acoustics data do not yet allow species discrimination, these size classes of fishes may include not only piscivores but also other larger fishes with different trophic ecologies.

We are confident, however, that the inclusion of other trophic groups as large fishes is unlikely given our knowledge of the reef fish communities in NC because fishes in the large size class (> 29 cm) are mainly piscivores (Whitfield et al. 2014, Taylor et al. 2016, Paxton et al. 2017). Fishes in the medium size (11 – 29 cm) class may include more species from other trophic groups: many of the medium sized fishes from other trophic groups actively forage on benthic invertebrates and other bottom-dwelling organisms. Because of their close association with the bottom, these fishes are difficult to differentiate from the surrounding structure and seafloor with SBES, which minimizes their detectability in our study and this

potentially confounding effect. Similarly, we are confident in our classification of small fishes as planktivorous fishes because other small fishes that are not planktivores would be undistinguishable from the shipwreck structure and surrounding seafloor, again reducing this potentially confounding effect.

Relationships between fish distributions and water current were complex. Here, we present our four main findings about relationships between fishes and current. First, on 67% of the surveys where we collected current measurements, the mean center of planktivorous fish aggregations occurred upstream of shipwrecks. This is consistent with findings on coral reefs that planktivorous fishes aggregate on leading edge of reef structures (Hamner et al.

1988). Second, the distance of planktivorous fishes from shipwreck edges was unrelated to current magnitude. Because most planktivorous fishes have streamlined bodies with forked caudal fins that enable them to hold position in currents without high energetic costs (Hobson

1991), the lack of relationship was unsurprising. We hypothesize that once current magnitude exceeds a particular threshold, it becomes energetically costly for planktivorous fishes to hold their position, so these fishes would retreat towards shipwreck structure to find refuge from current (Liao 2007). Since our observations were not indicative of this ‘flow refuging’ behavior, the magnitude of this current threshold likely exceeds the highest current magnitude (0.33 m/s) that we measured. Third, in contrast to planktivorous fishes, piscivorous fishes remained closer to shipwreck structures when currents were strong. This pattern may reflect high energetic costs incurred by large fishes trying to hold position in currents (Johansen et al. 2008). Fourth, we did not find a significant relationship between planktivorous fish location relative to shipwrecks and piscivorous fish density, suggesting that predation risk did not significantly influence planktivorous fish locations. These four

main results should be interpreted cautiously since they are based on measurements from the nine surveys where we collected current measurements and because some analyses indicated an effect of current on fishes whereas others did not. Relationships between fish locations relative to shipwrecks and current, and tradeoffs between energetic costs of swimming against currents and predation risk, are interesting and complex problems that require further investigation.

The spatial patterns in planktivorous fish, zooplankton, and piscivorous fish distributions that we documented on artificial structures are similar to phenomena documented on natural reefs that are also characterized as abrupt topographies. Despite the apparent similarities, functional effects of these processes occurring on artificial versus natural structured habitats remain to be tested. Artificial and natural habitats are known to differ in many characteristics, including structural complexity (Paxton et al. 2017), fish community composition (Paxton et al. 2017), and trophic structure (Simon et al. 2013), so investigating differences in the magnitude of the zooplankton-planktivore-piscivore relationship on both reef types is an important question to address. Regardless of the similarities and differences in zooplanktivory and associated processes on artificial versus natural habitats, artificial structures certainly provide habitat for aggregations of planktivores and their prey and predators. Effects from artificial structures could continue to propagate up and down the food web, respectively, to trophic levels higher than piscivorous fishes, such as marine mammals, seabirds, and humans, and lower than zooplankton, such as phytoplankton.

93 Acknowledgements

We thank T. Casserly, E. Ebert, and officers and crew of the NOAA Nancy Foster for assistance with fieldwork. We thank E. Ebert, F. Campanella, T. Jarvis, and B. Scoulding for assistance with fisheries acoustics data processing. We thank J. Hummon for assistance with

ADCP data processing. We thank T. Casserly and NOAA Monitor National Marine

Sanctuary staff for providing shipwreck metadata. We thank J. Fodrie, and A. Hurlbert for thoughtful reviews and guidance. Funding was provided by NOAA National Ocean Service and National Centers for Coastal Ocean Science, a NC Coastal Recreational Fishing License

Grant (#6446), a NSF Graduate Research Fellowship awarded to A.B. Paxton under Grant

No. DGE-1144081, a P.E.O. Scholar Award to A.B. Paxton, and a Carol and Edward

Smithwick Dissertation Fellowship awarded to A.B. Paxton through the UNC Royster

Society of Fellows.

Figures

Figure 3.1. Locations of fifteen shipwrecks surveyed on the continental shelf of NC. Gray lines and corresponding text indicate water depth in 10 m increments. Multibeam bathymetry of the shipwreck Proteus is in bottom right with warm colors indicating shallower depths and cooler colors indicating deeper depths.

Figure 3.2. Distances between the mean center of (a) small fishes, (b) medium fishes, and (c) large fishes and the nearest shipwreck edge for each survey. Positive distances are outwards of shipwreck structures; negative distances are within shipwreck structures. Horizontal lines

(black) represent individual surveys (N=20), and the sequence of surveys (Table S3.2) remains consistent among a, b, and c. Vertical lines are mean (solid red) ± standard error

(dashed red).

Figure 3.3. Spatial clusters of (a) zooplankton, (b) small fishes, (c) medium fishes, and (d) large fishes around the U-352 shipwreck. Colors correspond to the Getis-Ord Gi* z-value, where darker red represents more pronounced high density (hot-spots). Gray points are locations along the survey where data were collected but the Getis-Ord Gi* z-value was not significant so represent neither hot- nor cold- spots. The shipwreck is shown in black, and data are from 14 July 2017. North is up. Each scale bar is a total of 100 m; the scale division occurs at 50 m. Results from cluster analysis on other shipwrecks are in Figure S3.3.

Figure 3.4. Box plots describing spatial relationships between pairwise groupings of zooplankton, small fishes, medium fishes, and large fishes. a) Global index of collocation

(GIC) is a spatial statistic that ranges from 0 – 1, with 1 representing identical distributional centers. b) Local index of collocation (LIC) is based on organism density and ranges from 0

– 1, with 1 representing correlated densities at sampling points. c) Local index of co- occurrence (CO) is based on presence – absence and ranges from 0 – 1, with 1 representing co-occurrence.

Figure 3.5: a-c) Relationships between current magnitude and distances of a) small fishes, b) medium fishes, and c) large fishes from shipwreck, d) Relationship between number of predators on shipwrecks and within 5 m outward of shipwrecks. Linear regression displayed as solid line. Dashed lines are standard error. Raw data are points. Shaded gray areas represent fishes directly above the shipwreck structure, whereas white areas represent fishes located outwards of the shipwreck structure.

Tables

Table 3.1. Definitions and equations for indicators to quantify spatial distributions of individual groups of organisms and spatial relationships between organism pairs.

Spatial Description Equation Reference indicator Positive area Area organism Woillez et al. (PA) occupies based on = 1 2007, 2009 presence only

Spreading Area organism , Woillez et al. area occupies based on = 2 2007, 2009

(SA) density where: = # Equivalent Area organism would Woillez et al. area occupy if all sample ∑ 2007, 2009 = (EA) locations had the same ∑ density Microstructure Fine-scale variability Woillez et al. 0 − ℎ (MI) in organism = 0,1 2007, 2009 distribution 0 where: if MI = 0, well-structured distribution if MI = 1, poorly-structured distribution Center of Weighted mean center Woillez et al. ∑ gravity (CG) of organism 2007, 2009 =

Spatial ofdistribution groupone organism distribution ∑ Inertia Dispersion of organism Woillez et al. ∑ (I) around center of − 2007, 2009 = gravity ∑ Global index Index of collocation Woillez et al. ∆ of collocation between two organisms 2007, 2009 = 1 − 0,1 (GIC) at global scale of ∆ + + survey where: if GIC = 0, different CGs and 0 inertia if GIC = 1, coinciding CGs with positive inertia Local index of Index of collocation Bez and collocation between two organisms , = 0,1 Rivoirand (LIC) at local scale of 2000; Woillez sampling points, based where: et al. 2009 on density if LIC = 0, no overlap at sample locations if LIC = 1, overlap with same density at sample

organismgroups locations Co-occurrence Index of collocation Saraux et al. , (CO) between two organisms = 0,1 2014 at local scale of + Spatial relationships between pairs of sampling points, based where: on presence if CO = 0, no co-occurrence at sample locations if CO = 1, co-occurrence at all sample locations Term definitions: zi is fish density or zooplankton concentration at sample position xi si is area of influence of the sample at position xi, which is the area of points in space that are closer to this sample than to others Q is overall abundance Q(T) is the cumulative abundance in the cumulative area ( T) occupied by fish density or zooplankton concentration values g is the transitive covariogram (Bez and Rivoirard 2001) h0 is the mean lag between samples l is the number of samples where organism 1 ( l1), organism 2 ( l2), or both organisms ( l1,2 ) are present

100

Table 3.2. Spatial indicators (mean ± standard error) for distributions of zooplankton and fishes around shipwrecks. PA, SA, and EA values are percentages of the total area surveyed that each indicator occupies. For example, if the total area surveyed around a shipwreck was

100,000 m 2, and PA occurred in 25,000 m 2, then the positive area percentage would be 25%.

Inertia is not reported for zooplankton because their high positive area suggests that inertia is influenced by sampling extent. N = 20 surveys.

Mean Positive area Spreading area Equivalent area Microstructure Inertia number of Organism (PA) % (SA) % (EA) % (MI) [0 – 1] (I) [m 2] patches Zooplankton 23.9 ± 13.1 % 5.9 ± 1.4 % 2.1 ± 0.7 % 0.84 ± 0.04 NA 87 ± 39 Small fish 2.4 ± 0.7 % 1.7 ± 0.5 % 1.8 ± 0.5 % 0.95 ± 0.03 14,210 ± 2,670 27 ± 8 Medium fish 2.7 ± 0.3 % 1.1 ± 0.1 % 0.9 ± 0.1 % 0.82 ± 0.02 6,600 ± 2,680 20 ± 8 Large fish 2.2 ± 0.5 % 0.9 ± 0.2 % 0.8 ± 0.2 % 0.84 ± 0.03 3,540 ± 920 8 ± 3

101 CHAPTER 4: SEISMIC SURVEY NOISE DISRUPTED FISH USE OF A TEMPERATE REEF 4

Summary

Marine seismic surveying discerns subsurface seafloor geology, indicative of, for example, petroleum deposits, by emitting high-intensity, low-frequency impulsive sounds.

Impacts on fish are uncertain. Opportunistic monitoring of acoustic signatures from a seismic survey on the inner continental shelf of North Carolina, USA, revealed noise exceeding 170 dB re 1 m Pa peak on two temperate reefs federally designated as Essential Fish Habitat 0.7 and 6.5 km from the survey ship path. Videos recorded fish abundance and behavior on a nearby third reef 7.9 km from the seismic track. During seismic surveying, reef-fish abundance declined by 78% during evening hours when fish habitat use was highest on the previous three days without seismic noise. Despite absence of videos documenting fish returns after seismic surveying, the significant reduction in fish occupation of the reef represents disruption to daily pattern. This numerical response confirms that conservation concerns associated with seismic surveying are realistic.

4 A version of this chapter is published as: Paxton, A.B., J.C. Taylor, D.P. Nowacek, J. Dale, E. Cole, C.M. Voss, and C.H. Peterson. 2017. Seismic survey noise disrupted fish use of a temperate reef. Marine Policy 78: 68-73. DOI: 10.1016/j.marpol.2016.12.017.

102 Introduction

Marine seismic surveys emit high-intensity (up to 260 dB re 1 m Pa rms @ 1m), low- frequency (5-300 Hz peak spectral levels) sounds from airgun arrays downward into the water column (Hildebrand 2009). The resultant sound waves penetrate the seafloor to provide imagery of the underlying geology. These surveys can detect reservoirs of oil and natural gas, determine site-specific suitability for installation of offshore renewable energy infrastructure, evaluate sources of minerals for commercial extraction or sand for use in beach nourishment, and/or provide information on the continental substructure for geological research. Noise from seismic surveying can alter marine mammal vocalizations and foraging rates, and can lead to marine mammal displacement (Miller et al. 2009, Pirotta et al. 2014, Blackwell et al.

2015); however, there remain unanswered questions regarding how wild fish respond to seismic survey noise. Understanding whether fish are affected through alterations in behaviors associated with feeding, growth and survival has conservation and management implications.

Acute impacts to individual fish from seismic noise, including damage to sensory ear hair cells, can occur with close-range exposure to low-frequency, high-intensity sounds in laboratory settings (McCauley et al. 2003, Popper et al. 2005). Impulsive sounds similar to those from seismic surveys, such as noise made by pile driving, can cause mild to lethal injuries ranging from swim bladder rupture to hematoma and hemorrhaging (Popper and

Hastings 2009, Halvorsen et al. 2012, Popper et al. 2014). Behavioral responses of fish to impulsive noise are more difficult to quantify but may include changes in abundance in particular habitats (Slotte et al. 2004), changes in swimming patterns or feeding (Purser and

Radford 2011, Hawkins et al. 2014), as well as physiological stress even leading to mortality

103

(Popper and Hastings 2009). In contrast, in two studies that were specific to noise associated with seismic surveying, there were no marked changes in fish physiology or behavior

(Popper et al. 2005, Song et al. 2008). Reductions in fish catches can persist for up to five days after seismic activity (Skalski et al. 1992, Slotte et al. 2004, Løkkeborg et al. 2012).

Aside from those mentioned previously, most studies testing fish response to seismic noise occurred in laboratory settings; underwater observations of fish in their natural environment during seismic surveys are rare (Popper and Hastings 2009). Wardle et al. (2001) experimentally exposed fish in situ to noise from three synchronized airguns and observed startle responses in some fish but did not detect other changes in behavior or abundance.

Although fish in their natural environment may be expected to respond to seismic surveys based on laboratory experiments and reduction in fisheries catch (Slabbekoorn et al. 2010), no previous study has documented such an in situ behavioral response.

Opportunistic monitoring of a seismic survey offshore of North Carolina (NC) during

September 2014 determined whether reef-associated fishes in their natural environment respond to marine seismic surveying. The academic objective of the seismic survey was to study the formation and evolution of the Eastern North American Margin (Cruise Report:

Eastern North American Margin Community Seismic Experiment, Cruise MGL1408, R/V

Marcus G Langseth 2014), which involved use of an airgun array of similar volume to those used during oil and gas exploration. The majority of the survey occurred in deep (> 1000 m) waters off the continental shelf, although it continued across the shelf and into shallow (< 35 m) inner continental shelf waters of northeastern Onslow Bay, NC (Figure 4.1). This area supports hardbottom reefs that sustain an abundance of fish representing a diverse

104

community, including tropical, subtropical, and warm-temperate species (Parker Jr. 1990,

Parker Jr. and Dixon 1998, Whitfield et al. 2014). Fish use the temperate reefs for spawning and foraging, as well as for nurseries and refugia, qualifying them as Essential Fish Habitat under the Magnuson-Stevens Fishery Conservation and Management Act (2007).

Materials and Methods

As an empirical test of whether noise from seismic surveying can elicit a response from reef-associated fishes, such as a change in abundance, passive underwater monitoring stations were opportunistically established on three temperate reefs during September 2014

(Figure 4.1). The reefs, ranging from 25 to 33 m deep, were located 0.7, 6.5, and 7.9 km from the path of the vessel continuously conducting the seismic survey. The reefs were selected based on their proximity to the seismic survey track and because they have been the focus of various marine fisheries and ecological studies for several decades and have been documented to have notable abundances of fish in the federally-managed snapper-grouper complex and other commercially and recreationally important fishery species (Parker Jr. and

Dixon 1998, Whitfield et al. 2014).

The two reefs located closest to the survey track, a natural rocky reef and an artificial reef, were equipped with hydrophones (SoundTrap 202 recorders, Ocean Instruments, New

Zealand) that documented the acoustic signatures of the surveying noise (Audio S4.1-S4.2).

Hydrophones sampled continuously at 16-bit, 96 kHz. A video camera recorded fish abundance and behavior on the third reef, a naturally occurring rocky reef, farthest from the survey path (Video S4.1-S4.2). The video camera (GoPro, USA) was outfitted with an

105

intervalometer (cam-do, USA) to record 10-sec videos every 20 min. These monitoring instruments were mounted on conical metal frames (0.5 m high, 0.3 m base diameter), anchored with 60-80 kg of lead, and deployed on each reef on September 17, 2014 so that the instruments could record before and during seismic surveying. Video cameras deployed at the two reefs outfitted with hydrophones malfunctioned. Logistical constraints prevented collection of data following seismic surveying.

Acoustic data from the two hydrophones were processed and then five shots were aggregated for each of nine selected time points. Shots were processed in groups of five to obtain a ‘local average’ to smooth fine scale variation that occurs in the propagation conditions. The time points were chosen relative to the closest point of approach (CPA) on both the landward and seaward components of the survey path. The five shots closest to the

CPA that were not clipped were processed, and other locations were chosen to compare the received signals from the reefs, e.g., the more distant sampling locations gave similar propagation paths to the reefs, while the closer locations were subject to very different parts of the non-uniform source beam pattern (Tolstoy et al. 2009). On acoustic recordings from the reef located 0.7 km from the path of the seismic surveying vessel, the noise of the seismic shots overloaded the recorders when the ship was at its CPA. Using the known source sound level of the survey vessel’s airgun array (Tolstoy et al. 2009), the anticipated broadband level of received sound at the reef was calculated based on two models, spherical spreading and cylindrical spreading (Urick 1983). All acoustic values reported are in dB re 1 m Pa peak- peak.

106

Each 10-sec video recording from daylight hours was used to identify fish to the lowest taxonomic level possible, count the maximum number of fish in the frame by species, and document their behaviors as feeding, resting, schooling, or swimming. Noises from seismic surveying were audible as discrete airgun shots in video recordings, allowing us to associate any observed behavioral responses with timing of individual shots. To prevent observer bias, fish were first counted with video sound turned off; then sound was turned back on to detect whether shots were present.

Fish data obtained from video recordings were analyzed in R (R Development Core

Team 2015). The time series of hourly untransformed fish abundance was plotted for each of three days before and the following day during seismic surveying to visualize daily abundance patterns. The smoothed conditional mean of the hourly fish abundance for the combined three days before seismic activity and the accompanying standard error, as well as the smoothed conditional mean of hourly fish abundance on the day with seismic activity, was also calculated. The resulting two curves and the standard error were compared to determine whether the temporal pattern of fish abundance differed from before to during seismic surveying.

Results and Discussion

Noise levels on the two reefs designated as Essential Fish Habitat and located closest to the seismic survey track, 0.7 and 6.5 km away, exceeded 170 dB re 1 m Pa (Figure 4.2).

The peak levels that actually occurred at the sites are unknown because the noise overloaded the recorders. Using a sound source level of 258.6 dB re 1 m Pa (Tolstoy et al. 2009), the

107

received sound was estimated using two different models, spherical spreading and cylindrical spreading (Urick 1983). Based on a spherical spreading model, the corresponding received sound level on the closest reef would have been 202 dB re 1 m Pa, whereas based on the cylindrical spreading model, the received level would have been 230 dB re 1 m Pa. Realized peak sound levels likely fall between those predicted by spherical and cylindrical spreading models (Nowacek et al. 2013). The high intensity of this low-frequency sound is consistent with previous measurements (Guerra et al. 2011, Racca et al. 2015). The intensity of the noise is of significant concern because laboratory experiments indicate that fish experience recoverable injuries and/or potentially mortal injuries at noise levels > 207 dB re 1 m Pa peak

(Popper et al. 2014).

Ten-second videos were recorded every 20 min for three days before and through the day with seismic surveying on a 33-m-deep reef located 7.9 km from the closest approach of the seismic survey vessel. Although a hydrophone did not record sound on this reef, based on spherical spreading and a source sound of 258.6 dB re 1 m Pa the estimated noise experienced on this reef was 181 dB re 1 m Pa when the survey vessel was closest. Using a second model based on cylindrical spreading, the received sound level was 220 dB re 1 m Pa on the reef.

Realized peak sound levels probably lie between the predictions of these two spreading models (Nowacek et al. 2013). The resulting 140 videos from daylight hours were used to identify and count maximum number of fish in frame by species and document their behaviors. During the four days of monitoring this temperate reef, 32 species belonging to 17 families (Table S4.1), including many federally managed as part of the snapper-grouper complex, were observed.

108

On the reef monitored by video camera, fish occupation during three days prior to the seismic survey exhibited a daily pattern of increasing abundance during the evening, as compared to morning and afternoon (Figure 4.3). On the following day with airgun noise, this pattern in fish use did not emerge from observations across periods of the day. Fish abundance remained low for the entire day, with the exception of one outlying observation during evening (Figure S4.1). The outliers were predominately comprised of Haemulon aurolineatum (tomtate), a grunt that consumes benthic invertebrates and zooplankton, and

Decapterus spp. (scad), a forage fish that eats zooplankton. Reductions in fish abundances during seismic surveying proved statistically significant using two different statistical tests.

First, the mean variance in fish counts on each of the three days without seismic noise was greater than the corresponding mean variance on the day with seismic surveying (via analysis of means for variance (ANOMV) with Levene transformation, p = 0.047; Figure S4.2). The statistically significant differences in fish abundance between the single day with and the three days without seismic noise were driven by data from a four-hour evening period (1600-

2000 local time). Whether fish occupation of the reef differed during the evening across all days was further tested. The total number of fish occupying the reef during evening declined by 78% when exposed to seismic noise (ANOVA followed by post-hoc pairwise t-test with

Box-Cox transformation, F 3,36 = 4.74, p = 0.007).

In addition to counting fish, video recordings were examined to assess whether fish exhibited behaviors that could help understand the change in reef use. Noises from seismic surveying were audible as discrete airgun shots in video recordings, allowing association of any observed behavioral responses with timing of individual shots. Eight shots were audible

109

on video. The other shots occurred at 30 to 90-s intervals and did not coincide with the recording schedule. Only one observed fish, a H. aurolineatum , exhibited an apparent behavioral response to an airgun shot by swimming away from a ledge. From the lack of abundant fish observed during evening when repeatedly exposed to seismic noise, it is presumed that at least some reef-associated fishes left the reef.

Conclusion

Although working with limited data, this study provides evidence that during exposure to seismic noise, the prevailing pattern of heavy fish use of reefs during the evening was suppressed. The finding is notable because it goes well beyond detection of a startle response from individual fish (Wardle et al. 2001), instead suggesting a multi-species response to airgun noise and/or particle acceleration, validating expectations (Slabbekoorn et al. 2010) that fish respond to seismic surveying in their natural environments. The

Magnuson-Stevens Fishery Conservation and Management Act (2007) mandates protection of reefs, including those studied here, as Essential Fish Habitat. Reducing opportunities for fish to aggregate causes concern as this could reduce options for foraging, mating, or other important life history functions. Though there are no observations to indicate the duration of the observed effect, these research results augment and confirm issues raised by marine mammal experts (Nowacek et al. 2015) and suggest that concerns associated with marine seismic surveys appear to be realistic and well-founded.

110

Acknowledgements

We thank A. Adler, E. Pickering, J. Vander Pluym, R. Mays, R. Purifoy, and crew from Olympus Dive Center for field assistance, B. Degan and E. Ebert for help with video processing, and S.R. Fegley for assistance with data analysis. We thank R.C. Muñoz, N.M.

Bacheler, and D. Gruccio for thoughtful reviews. Funding was provided by BOEM under

Cooperative Agreement M13AC00006, a NSF Graduate Research Fellowship awarded to

A.B.P. under Grant No. (DGE-1144081), and NOAA National Centers for Coastal Ocean

Science. We thank scientists from LDEO and the ENAM project for scientific transparency and cooperation that facilitated this opportunistic monitoring. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the opinions or policies of the US Government, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.

111

Figures

Figure 4.1. Track of seismic survey vessel (black line) relative to three monitoring reefs on the inner continental shelf of NC: two outfitted with hydrophones (blue triangles) and one with video camera (orange square).

112

Figure 4.2. Acoustic signatures of A) ambient noise and B-D) noise from seismic airgun shots on reef 0.7 km from closest approach of seismic surveying vessel: B) 22.2 km from reef before closest approach; C) 0.7 km from reef showing the seismic shots just prior to shots that overloaded our instruments; D) 19.6 km from reef following closest approach. Insets depict 10 Hz – 5 kHz range of low frequency.

113

Figure 4.3. Hourly fish abundance on the reef 7.9 km from the closest approach of the seismic survey ship during three days before (solid black line) and on one day during the height of seismic activity near the reef (red line). The solid black line is the smoothed conditional mean and the black dotted lines are standard error of the hourly fish abundance for three days before seismic surveying. The red line is the smoothed conditional mean of hourly fish abundance on the day with seismic activity.

114 CHAPTER 5: CONVERGENCE OF FISH COMMUNITY STRUCTURE BETWEEN A NEWLY DEPLOYED AND AN ESTABLISHED ARTIFICIAL REEF ALONG A FIVE-MONTH TRAJECTORY 5

Summary

Numbers of human-made reefs in the world’s oceans are increasing, yet questions remain about patterns and speed of fish colonization of these artificial reefs. Here, we tested

1) whether the fish community on a newly deployed artificial reef converged with the fish community on an adjacent, established artificial reef over time and 2) whether fish colonization of the new artificial reef occurred sequentially. To answer these questions, we simultaneously collected time-lapse videos of fishes colonizing a new (2 wks old) artificial reef and those inhabiting a nearby (438 m away) established (> 20 yrs old) artificial reef. We found that fish community composition on the new artificial reef converged with the fish community composition on the established artificial reef over five months. Community development on the new reef followed a trajectory: schooling, planktivorous fishes initially colonized the reef in high numbers, whereas demersal fishes exhibited delayed colonization.

These findings suggest that fishes may colonize human-made reefs along a specific trajectory of pelagic fishes followed by demersal fishes and that community convergence between reefs can occur over relatively short temporal scales given similar environmental conditions. When deploying additional structures, including human-engineered habitats, in the marine

5 A version of this chapter is in review at ECOLOGICAL ENGINEERING as: Paxton, A.B., L.W. Revels, R.C. Rosemond, R. Gaesser, H.R. Lemoine, J.C. Taylor, and C.H. Peterson. Convergence of fish community structure between a newly deployed and an established artificial reef along a five-month trajectory.

115 environment, our findings on fish colonization of artificial reefs are important to consider because they provide new insight into how artificial structures can be utilized to enhance particular fishes over different temporal scales.

Introduction

Numbers of constructed submerged habitats are increasing worldwide (Dafforn et al.

2015, Gittman et al. 2015). These artificial habitats often provide ecological functions different from those of their natural counterparts (Bulleri and Chapman 2010). For example, naturally occurring marshes can deliver more protection from shoreline erosion than do human-constructed bulkheads (Gittman et al. 2014), and subtidal artificial reefs can facilitate the spread of invasive species to a higher degree than natural reefs (Dafforn et al. 2012).

Because the function of natural habitats can differ from the function of introduced, artificial habitats, these artificial structures can drive ecological changes in communities (Bulleri and

Chapman 2010), warranting an understanding of how species utilize human-made structures and what processes regulate the associated community structure.

In general, communities exposed to similar environmental conditions often develop similar community structures over time (Samuels and Drake 1997). Yet, whether communities exposed to similar environmental conditions but occupying individual human- constructed habitats converge with or diverge from each other over time is debated. Some human-engineered habitats, such as reservoirs, can become more similar to each other over time (Gido et al. 2009), likely due to similarities in local scale habitat characteristics and species interactions among the unnatural habitats. Others, including artificial reefs, can diverge from each other given high frequency of physical disturbance (Cummings 1994)

116

and/or differing physical features (Thanner et al. 2006). Despite growing numbers of unnatural habitats, there remain unanswered questions about how colonization proceeds on these novel structures.

Submerged human-made structures intentionally deployed in the marine environment as artificial reefs present an opportunity to test how colonization proceeds on unnatural habitats. These marine artificial reefs are often part of government-managed artificial reef programs where structures, ranging from ships and concrete pipes to boxcars and bridges, are purposely deployed on the ocean floor to enhance fisheries and provide fishing and diving opportunities (NOAA 2007). Marine artificial reefs often lie in close proximity (m’s to km’s) to one another, forming dense arrays of hard-bottom habitat. Networks of marine artificial reefs, with varied proximity to one another and different dates of deployment, allow empirical tests of how habitat provision afforded by artificial reefs changes over time.

Artificial reefs provide habitat for a diversity of species, including macroalgae, benthic invertebrates, and fishes. Macroalgae and benthic invertebrates often exhibit distinct colonization sequences on artificial reefs (Fitzhardinge and Bailey-Brock 1989, Perkol-

Finkel and Benayahu 2005, Thanner et al. 2006). For example, fouling organisms, including bryozoans and hydrozoans, colonize hard surfaces of artificial reefs first, often followed by more persistent invertebrates, such as octocorals (Carter and Prekel 2008). Whether mobile species, such as fishes, colonize artificial habitats in sequential patterns is not well understood. Most studies examine whether fish communities on artificial reefs converge with fish communities on natural reefs that they are intended to mimic (Koeck et al. 2014, Becker et al. 2017). Few examine whether adjacent artificial reefs support similar fish communities

117

to each other and how fish community development progresses with time since reef deployment. Those studies that examined fish community change on artificial structures discovered that communities diverged because of differing environmental conditions

(Oricchio et al. 2016), and likely because of physical differences in reef structures (Thanner et al. 2006). Fish colonization on new structures occurs rapidly, often within hours of deployment (Cummings 1994, Clark and Edwards 1999), but the colonization sequence varies. In some systems, pelagic species colonize first (Golani and Diamant 1999, Dance et al. 2011), whereas in others demersal fishes or a combination of demersal and pelagic species initially dominate reefs (Alevizon and Gorham 1989, Dance et al. 2011).

The objective of this study was to document colonization of fish communities on a newly deployed artificial reef at several time intervals since reef creation. We explicitly tested hypotheses that: 1) the fish community on a newly deployed artificial reef (2 wks old) will converge with the community on an established (>20 yrs old) artificial reef, and 2) fishes that rely on the physical structure of artificial reefs will initially colonize the new reef, whereas fishes that rely on benthic biota associated with the artificial reef structure will exhibit delayed colonization. This research is important for understanding dynamics of how human-made structures may enhance fisheries over time.

Materials and Methods

We selected two artificial reefs of similar structure: 1) a newly deployed artificial reef and 2) an established artificial reef (Figure 5.1a). The new artificial reef, a 33-m long U.S.

Army tugboat renamed the James J. Francesconi , was scuttled on 7 May 2016 (34.5634 N x

118

76.8552 W). The established artificial reef, a 100-m long landing craft repair vessel USS

Indra, was deployed in 1992 (34.5623 N x 76.8515 W). Given its deployment over two decades ago, the USS Indra was assumed to have reached an established biological community (climax community within context of ocean reef succession). Both vessels are made of steel and located within a state-designated artificial reef site (AR-330) within

Onslow Bay, North Carolina (NC). These two intentionally sunk vessels lie 438 m from each other on a sandy seafloor at a depth of 20 m, so both are exposed to similar environmental conditions. We documented fish communities on the new and established artificial reefs during three sampling periods over five months in 2016. The three sampling periods were: 1)

May 17 – 26 (10 days); 2) July 21 – August 5 (16 days); 3) September 13 – 19 (7 days). As the new artificial reef was deployed on 7 May 2016, these sampling periods occurred one, three, and five months, after the sinking of the James J. Francesconi as a new artificial reef.

During each of the three sampling periods, we used time-lapse videography to simultaneously document fish communities on the new and established artificial reefs

(Videos S5.1-S5.4). On each reef, divers deployed a GoPro Hero 3+ Black video camera

(GoPro, USA) housed inside of a cylindrical aluminum housing (Sexton Co., USA), which included an acrylic dome port with lens optically centered, an internal battery and programmable intervalometer, and an external light (Figure 5.1b). We programmed the intervalometer to interface with the video camera and light to collect videos on a set schedule during both the day and night. Video cameras recorded 20-s long videos every 20 min. The external white LED light remained off for the first 10 s of every video and then turned on during the final 10 s. Each video-camera unit was mounted on a secured, conical metal frame

119

(0.5 m high with 0.3 m diameter base). Each frame was positioned by scientific divers on the horizontally-oriented deck of the artificial reefs with the lens of each video camera facing areas of superstructure. These units were deployed in identical locations during all three sampling periods. At the end of each sampling period, divers retrieved the video cameras and offloaded videos.

Each video was processed by an analyst who identified and counted the maximum number (maxN) of fishes in a given frame to the lowest taxonomic level possible. For each reef during each of the three sampling periods, 150 videos were randomly selected and processed. We chose to process 150 videos per reef and sampling period based on species accumulation curves. Only the last 10 s of each video were processed because this was the period during both day and night videos when the LED light was on. The area illuminated in night videos was much smaller than the visible area from the daytime videos. Because of differences in field of view between night and day videos, we chose not to analyze results from night videos. Rather, we analyzed videos from only the daytime hours of 0700 to 1900 local time.

To determine whether the trajectory of fish community development on the new artificial reef converged with the fish community on the established artificial reef and whether a pattern of fish colonization occurred, we analyzed data from processed videos using R version 3.2.0 (R Development Core Team 2015). Because we utilized a repeated observation design where we sampled the same reefs during three time periods, we employed linear mixed effects models fit with a random intercept and random slope to determine whether community metrics (abundance, species richness, Pielou’s evenness) differed among

120

the repeated observations (sampling periods) on each of the two reefs using the ‘nlme’ package (Pinheiro et al. 2013). The random intercept allowed for the possibility that the two different reefs initially supported different numbers of fishes, whereas the random slope component allowed for the possibility that the two reefs had a different temporal response over which changes in community metrics could occur. We found that there was no effect of the repeated observations design (factor sampling period P > 0.05) on community metrics with this random slope and random intercept model. Since the repeated observations component (factor sampling period) was not significant, we proceeded to a two-way analysis of variance (ANOVA) that tested for differences in mean community metrics (abundance, species richness, Pielou’s evenness) between reefs, between sampling periods, and the interaction of these two factors and did not explicitly incorporate effects of repeated observations. We followed the two-way ANOVAs with post-hoc Tukey HSD tests. The assumption of normality was violated; however, the P-values indicated that observed differences in mean community metrics were significant. Given the very low P-values, the amount of error introduced to calculated P-values from the violation of normality would not have been sufficient to make the results nonsignificant, so we decided to run our models on untransformed data rather than on transformed data. Models were visualized with effects plots.

We tested how fish community composition varied by site and season with permutational analysis of variance (PERMANOVA), homogeneity of multivariate dispersions (PERMDISP), and non-metric multidimensional scaling (nMDS). All multivariate tests were performed on square-root transformed data using the ‘vegan’ package

121

(Oksanen et al. 2015). PERMANOVA, a permutation-based technique that uses variance partitioning (Anderson 2001), tested whether fish community composition differed by artificial reef, sampling period, and their interaction. The PERMANOVA procedure used

Bray-Curtis distances and 1,000 permutations. To interpret results from PERMANOVA, we used PERMDISP, which is a distance-based test for homogeneity of multivariate dispersion

(Anderson 2006). PERMDISP determined whether community composition differed between reefs, sampling periods, and combinations of reefs and sampling periods. Together,

PERMANOVA and PERMDISP allowed us to evaluate whether fish community composition converged over time. Next, NMS, an ordination method that visually summarizes patterns in the structure of multivariate datasets (Shepard 1962, Kruskal 1964,

Legendre and Legendre 2012), was performed on the fish community data. Samples were mapped into ordination space using the ecological distances between samples ordered by rank. Bray-Curtis distances summarized pairwise distances among samples and helped overcome the problem of joint absences in species data (Oksanen et al. 2015). A Shepard diagram ensured linearity between the ordination distance and Bray-Curtis distance. Biplots with samples and superimposed ellipses indicating 50% confidence intervals allowed visualization of the relationships among samples in ordination space.

To determine which species contributed to differences in fish community composition, we conducted similarity percentage (SIMPER) and indicator species analyses on the square-root transformed abundance data. Fishes contributing to differences in community composition between reefs and sampling periods were determined by SIMPER

(Clarke 1993) with 1,000 permutations using the ‘vegan’ package (Oksanen et al. 2015).

122

Indicator species analysis determined which fishes were indicators of each reef during each sampling period and was performed using the ‘indicspecies’ package (De Cáceres and

Legendre 2009).

To confirm that patterns from the video data were not confounded by seasonality, we compared the video data collected during the three sampling periods on the established reef, the USS Indra, to data obtained previously on the USS Indra. These previously-obtained data were collected on the USS Indra before the new artificial reef was deployed nearby and were collected by divers along 30-m x 4-m belt transects (120 m 2) (Brock 1954, 1982, Samoilys and Carlos 2000) during 2013-2016 (Table S5.1). Because these older data were collected using belt-transects and the newer data using time-lapse videography, fish counts are not directly comparable. Instead, we visualized seasonal trends in the diver-collected transect data using NMS and compared these trends to the video data to ensure that seasonality did not confound our interpretation of patterns in video data over time.

To determine whether the colonization of the newly deployed artificial reef followed a particular sequence, we tested differences in fish abundance over time for demersal and pelagic species. We selected two demersal species, Archosargus probatocephalus

(sheepshead) and Centropristis striata (black sea bass) , as well as two pelagic species ,

Decapterus sp. (scad species) and Haemulon aurolineatum (tomtate) . These species were selected for analysis because SIMPER analyses indicated that they contributed to differences in community composition between the reefs. Two of these species, C. striata and H. aurolineatum, are also federally managed as part of the snapper-grouper complex (South

Atlantic Fishery Management Council 1983, 2016).

123

Results

In the daytime videos, we observed 22 identifiable fish species (Table S5.1; Videos

S5.1-S5.4). Some fishes were unidentifiable to the species level because they were far from the video camera: these were categorized at the family level and represented 11 families. If fishes could not be identified to the family level, then they were classified as either unknown pelagic fishes or unknown demersal fishes. In total, videos recorded fishes belonging to 16 families. On the new reef, we observed 53,098 individual fishes belonging to 10 families, whereas on the established reef we observed 37,040 individual fishes belonging to 14 families. There were four species on the new reef that were absent from the established reef:

Apoginidae sp. (unknown cardinalfish), Carangoides bartholomaei (yellow jack),

Rhomboplites aurorubens (vermillion snapper), and Rachycentron canadum (cobia). Eight fish species, including one shark species, occurred exclusively on the established reef:

Alopias vulpinus (thresher shark), Remora remora (remora), Chaetodipterus faber (Atlantic spadefish), Stephanolepsis hispidus (planehead filefish), Pomacentridae sp. (unknown damselfish), Stegastes sp. (beaugregory or cocoa damselfish), Rypticus maculatus (white spotted soapfish), and Serranus subligarius (belted sandfish).

Fish community metrics (abundance, richness, evenness) converged between the new and established artificial reef over the three sampling periods (Figure 5.2; Videos S5.1-S5.4).

Fish abundance initially differed between the new and established artificial reefs, but these metrics grew more similar with time (Figure 5.2a; ANOVA F reef, 1, 447 = 6.3, Preef = 0.01,

Fsampling period, 2, 447 = 32.7, Psampling period < 0.0001, F reef x sampling period, 2, 447 = 5.9, Preef x sampling period

= 0.003). Fish abundance on the new reef was 5.7 times higher than on the established reef

124

during the first sampling period, which began two weeks following the intentional sinking of the new artificial reef. During the second sampling period, abundance of fishes on the new reef was 1.3 times the abundance on the established reef. During the third sampling period, however, abundances on the two reefs were statistically indistinguishable. The new reef was initially dominated by a smaller number of species than the established reef (Figure 5.2b;

ANOVA F reef, 1, 447 = 173.2, Preef < 0.0001, F sampling period, 2, 447 = 13.5, Psampling period < 0.0001,

Freef x sampling period, 2, 447 = 6.9, Preef x sampling period = 0.001). As with fish abundance, initial differences in species richness between the two reefs decreased in the second sampling period. Richness on the new reef was lower than on the established reef during the third sampling period. Species evenness differed between reefs within sampling periods (Figure

5.2c; ANOVA F reef, 1, 447 = 0.01, Preef = 0.9, F sampling period, 2, 447 = 20.4, Psampling period < 0.0001,

Freef x sampling period, 2, 447 = 5.0, Preef x sampling period = 0.007). Evenness was higher on the established reef than on the new reef during the initial two sampling periods. During the last sampling period, however, evenness on the new reef exceeded evenness on the nearby established reef.

Fish community composition initially differed between the established artificial reef and the newly deployed reef, but community composition converged over five months

(Figure 5.3; Videos S5.1-S5.4). Specifically, fish community composition differed between the two artificial reefs (PERMANOVA F 1,66 = 21.28, P > 0.001) and among the three sampling periods (PERMANOVA F 2,66 = 21.08, P > 0.001). The interaction between reefs and sampling period also affected fish community composition (PERMANOVA F 2,138 =

14.76, P > 0.001). To interpret differences between reefs and sampling periods indicated by

125

PERMANOVA, we used PERMDISP to test for homogeneity of variances by reef and sampling period. While the dispersion in fish community composition between the two reefs was not different (PERMDISP F 1,70 = 0.58, P = 0.70), there were differences in dispersion among the three sampling periods (PERMDISP F 2,69 = 25.51, P > 0.0001). Dispersion of fish community composition during the first and second sampling periods, as well as during the first and third, were different from each other (post-hoc Tukey HSD P > 0.0001 and P >

0.001, respectively). However, dispersion between the last two sampling periods was not different, indicating convergence of community composition over time (post-hoc Tukey HSD

P = 0.82). When we combined reef site and sampling into a single factor (factor levels: new reef during sampling period 1, new reef during sampling period 2, etc.) to examine differences between the new and old reefs over time (PERMDISP cannot test two factors simultaneously), a pattern emerged. During the first sampling period, initially high dispersion among fish communities on the new and established reefs decreased over time (Tukey HSD between new and established reefs for: first sampling period P = 0.02; second sampling period P = 0.43; third sampling period P = 0.99). Taken together, results from

PERMANOVA and PERMDISP indicate that fish community composition on the two reefs converged over time, such that the community compositions on the two reefs were equivalent during the third sampling period, which was five months after deployment of the new artificial reef.

Initial differences in fish community composition between the two reefs were driven by higher abundances of schooling fishes on the new reef (Video S5.1) and higher abundances of demersal fishes on the established reef (Video S5.2). During the first sampling

126

period, schooling fishes that we were unable to identify to family or species level from video footage contributed to 22.6% of differences between the new and established reefs; abundance on the new reef averaged 10.4 fishes per frame, whereas on the established reef, there were none. These reported values of square-root transformed numbers of fishes per frame, as well as those below, are based on SIMPER analyses conducted with square-root transformed abundance data. Similarly, schooling fishes Haemulon aurolineatum (tomtate;

0.4 fish per frame on established vs. 6.9 fish per frame on new reef) and Decapterus sp. (scad species; 0 fishes per frame on established vs. 4.3 fishes per frame on new reef) were both more prevalent on the new reef than the old reef and accounted for 13.6% and 8.6%, respectively, of differences between reefs. While schooling fish abundance was higher on the new reef than the established reef, abundances of demersal, reef-associated fishes were higher on the established reef than the newly deployed reef. For example, Diplodus holbrooki

(spottail pinfish), an herbivorous fish, contributed to 6.7% of differences in fish community composition between reefs, as it exhibited abundances of 4.4 fish per frame on the established but 1.4 fish per frame on the new reef. Higher abundances of Serranidae

(seabasses), including Centropristis striata (black sea bass) and S. subligarius , C. faber , and

Archosargus probatocephalus (sheepshead) on the established reef compared to the new reef each contributed to < 2% of differences in fish communities between reefs.

During the second sampling period, differences in schooling fishes between the two reefs declined. While abundance of one schooling fish, R. aurorubens (0 fish per frame established vs. 8.3 fish per frame new reef; 10.4% contribution), was markedly greater on the new reef, abundances of other schooling fishes, including Decapterus sp. (4.3 fishes per

127

frame established vs. 4.7 fishes per frame new reef, 8.0% contribution) and H. aurolineatum

(6.9 fish per frame established vs. 7.8 fish per frame new reef, 6.2% contribution), were only slightly higher on the new than on the old reef. In contrast, unknown schooling fishes (10.4 fishes per frame established vs. 9.8 fishes per frame new reef, 6.6% contribution) exhibited higher abundance on the established than new reef. Demersal fishes, including Mycteroperca microlepsis (gag grouper), S. subligarius , and D. holbrooki , occurred in higher numbers on the established reefs, and each contributed to < 1% of differences between reefs.

Five months following the sinking of the new artificial reef, differences in fish communities diminished (new reef: Video S5.3; old reef: Video S5.4). Whereas formerly, schooling fishes were more abundant on the new reef, schooling fish abundance was nearly equal on the two reefs during the third sampling period ( Decapterus sp. 4.1 fishes per frame established vs. 1.8 fishes per frame new reef, 9.2% contribution; unknown schooling fishes

4.3 fishes per frame established vs. 3.4 fishes per frame new reef, 7.8% contribution; H. aurolineatum 5.9 fish per frame established vs. 5.4 fish per frame new reef, 4.4% contribution).

Select fish species served as indicators of the colonization trajectory of the new reef compared to the established nearby reef. Indicators on the established reef during all sampling periods included Serranidae (indicator value = 0.89, P = 0.001), S. subligarius

(indicator value = 0.88, P = 0.001), M. microlepsis (indicator value = 0.84, P = 0.001) and A. probatocephalus (indicator value = 0.67, P = 0.001). As time passed since the sinking of the new reef, joint indicators of the established reef and of the developing reef community emerged. Halichoeres bivittatus (slippery dick, indicator value = 0.68, P = 0.004) occurred

128

on the established reef through all sampling periods and also on the new reef during the second sampling period. Diplodus holbrooki (indicator value = 0.988, P =0.001) represented the established reef during the experiment and the new reef during the last two sampling rounds.

An ordination of diver-collected fish data from the old reef prior to deployment of the new reef indicated that i nter-annual variation in fish community composition occurred for several years prior to scuttling the new reef nearby (Figure S5.1). However, there were no repeatable patterns in seasonal variation of the fish community composition from one year to the next on the established reef. The lack of repeatable seasonal patterns in the fish community composition on the established reef during years prior to sinking the new reef suggest that seasonal shifts in community composition did not confound our interpretation of community convergence between the new and the old reefs over the five-month period from which we obtained video data.

Following multivariate analyses which indicated that community composition differed between the new and established reefs because of higher initial concentrations of pelagic, schooling fishes on the new reef compared to the established reef, we tested for differences in counts of select species of pelagic and demersal fishes between reefs over time . Generally, demersal fishes colonized the new reef slowly, as indicated by low abundances during the early sampling period(s) and higher abundances in later sampling period(s) (Figure 5.4a-b). For example, A. probatocephalus was absent on the new reef in the first two sampling periods, but was present in the third sampling period in abundances similar to those on the established reef (Figure 5.4a). Centropristis striata abundance

129

gradually increased during the first two sampling periods to reach the abundance observed on the established reef (Figure 5.4b). Pelagic fishes colonized the new reef quickly (Figure 5.4c- d). Decapterus sp. displayed initially high numbers on the new reef which tapered off over time to converge with numbers of Decapterus sp. on the established reef (Figure 5.4c).

Haemulon aurolineatum , although present in low numbers on the new reef during the first sampling period, was present in highest numbers during the second sampling period and then in similar numbers to the established reef in the third sampling period (Figure 5.4d).

Discussion

We provide empirical evidence that fish community composition on a newly deployed (2 wks old) artificial reef converged with the community composition of a nearby established (>20 yrs old) artificial reef within five months. Community development on the new artificial reef occurred sequentially. Schooling, planktivorous fishes initially colonized the new reef in high numbers, whereas demersal fishes arrived later. These findings suggest that fishes colonize human-made reefs along a sequence and that community convergence between new and established reefs can occur over relatively short (5 mo) temporal scales, given spatial proximity and similar environmental conditions between reefs. These results on fish colonization of artificial reefs illuminate colonization dynamics that are important to consider when deploying additional human-made structures to enhance fish habitat in the marine environment.

Fish community composition on the new artificial reef became more similar to the community composition on the established reef over time. Previous studies regarding how

130

fish species colonize artificial reefs focus largely on species-specific abundances and do not examine community composition (Solonsky 1985, Alevizon and Gorham 1989, Cummings

1994, Golani and Diamant 1999). Our study provides new insights to understanding marine colonization dynamics because we not only examined univariate metrics, such as abundance, richness, and evenness, but we also examined the community structure. Our conclusion that fish communities converged was based largely on the increase in similarity of fish communities occupying the two reefs over the three sampling periods. The increase in similarity was driven by decreased multivariate dispersion in fish community composition over time, which was linked to reduced differences between abundances of pelagic fishes and demersal fishes on the two reefs.

We found that community convergence between fishes on these new and old reefs occurred within five months. This temporal scale for convergence differs from findings of previous studies. In California, for example, colonization of two constructed artificial reefs by only three-quarters of the regional species pool occurred within six months (Solonsky

1985). Additionally, even though species diversity plateaued on modular artificial reefs in

Florida within 2 – 2.5 months, community composition did not converge among reefs over time (Cummings 1994). Our finding also contrasts with results from another study in Florida where fish communities inhabiting different types of artificial reefs (boulders and modules) did not converge over five years (Thanner et al. 2006). Our study documented rapid community-wide convergence of fishes between two artificial reefs within a short time scale, likely because the reefs we studied were located close together (438 m) and environmental conditions remained similar. The reefs may have also converged quickly as a function of

131

colonization patterns associated with high concentrations of artificial reefs, shipwrecks, and naturally occurring rocky reefs within Onslow Bay, NC surrounding the two studied artificial reefs.

Our study illuminated a colonization sequence for human-made reefs, where schooling fishes initially occupied the new structure in high numbers but then decreased in numbers over time as reef-associated demersal fishes slowly increased in abundance over time. This trajectory is consistent with previous findings that initial colonization of fishes occurs quickly and in high numbers. Schooling fishes that consume zooplankton during part of their life history often constitute the early stages of fish communities on artificial reefs

(Golani and Diamant 1999, Thanner et al. 2006, Becker et al. 2017). These schooling planktivorous fishes attract high numbers of piscivores, likely inhibiting colonization of cryptic demersal fishes (Dance et al. 2011). When schooling planktivorous fish numbers decrease on reefs over time, as we observed in this study, numbers of piscivores may also decline, removing predation pressure and allowing colonization of cryptic demersal fishes.

This could explain why we saw delayed colonization of demersal fishes on the new artificial reef but high numbers of demersal fishes on the nearby established artificial reef.

Additionally, the time it takes for epibiota to colonize the new reef could explain the slow arrival of demersal fishes on the new reef. For example, demersal fishes that consume epibiota growing on reef structures may not have an adequate food supply soon after sinking of a new reef because the epibiota community is not well established. After the epibiota community becomes established, however, demersal fishes would have a suitable food source. The colonization sequence of pelagic fishes colonizing new reefs first, followed by

132

demersal fishes, likely shapes the development of the fish community through trophic interactions. In coral reefs, predation pressure from herbivorous fishes causes different colonization trajectories (Hixon and Brostoff 1996). Perhaps, predation from piscivorous fishes and/or the food supply for demersal fishes determine how quickly the community of fishes on artificial temperate reefs develops through similar mechanisms.

The community change that we documented was not confounded by seasonality.

When we examined seasonality using three years of diver-collected fish transect data from before the new vessel was deployed, there was no pattern in fish community composition attributable to seasonal changes. We were not able to discern whether fishes occupying the newly deployed reef recruited from the established reef and/or other nearby reefs. This question is important to consider, especially given classic theory that colonization of artificial structures consists of a redistribution phase, where fishes are attracted to new structure from other nearby reefs, followed by an equilibrium phase (Alevizon and Gorham 1989). Based on the community composition convergence, we posit that the new reef had an initial redistribution phase with high numbers of schooling fishes and after five-months post deployment had entered an equilibrium phase characterized by demersal fishes actively using the reef structure.

Our finding that the fish community on a newly deployed artificial reef converged with the fish community on a nearby reef sunk more than two decades earlier suggests that artificial reefs deployed in close geographical proximity to other reefs can provide additional habitat to support complementary communities of fishes. The rapid rate of convergence suggests that nascent artificial reefs function similarly to established artificial reefs after just

133

a few months. Additionally, the sequence of schooling fishes initially frequenting the new reef followed by demersal fishes slowly occupying the reef, suggests that reefs become more productive over time as their habitat provision changes. Managers responsible for artificial reef deployment and installation of other submerged, human-engineered structures may use relative abundances of pelagic, schooling fishes versus demersal fishes as metrics to assess colonization trajectories on unnatural habitats. In disturbance prone areas, such as those where natural or unnatural types of disturbance occur, these metrics may be used to evaluate community development following disturbance events .

Acknowledgements

We thank E. Pickering, A. Adler, G. Safrit, E. Ebert, L. Bullock, A. Pickett, D.W.

Freshwater, J. Fleming, J. Hughes, M. Kenworthy, G. Sorg, A. Poray, J. Geyer, A. Rok, T.

Dodson, J. Hackney, J. Purifoy, S. Davis, C. Lewis, E. Kromka, JR. Purifoy and crew from

Olympus Dive Center, and T. Leonard and crew from Discovery Diving for diving and boating assistance. We thank C. Buckel for assistance with database construction. We thank

S. Fegley, J. Fodrie, A. Hurlbert, and J. Rosman for thoughtful reviews and guidance. This research was supported by funding from NC Coastal Recreational Fishing License Grants

(#5115 and #6446), a National Science Foundation Graduate Research Fellowship awarded to A.B. Paxton under Grant No. DGE-1144081, and a P.E.O. Scholar Award to A.B. Paxton.

134

Figures

Figure 5.1. Experimental design to quantify fish community change over time on a newly deployed and established artificial reef. a) Location of newly deployed artificial reef ( James

J. Francesconi ; gray triangle) and established artificial reef (USS Indra ; black circle) within the state-designated artificial reef site (black square (AR-330)) located in Onslow Bay, North

Carolina. b) Video camera unit deployed facing superstructure on each artificial reef with labeled key parts: 1) subsurface marker, 2) cylindrical housing containing intervalometer, battery, and video camera, 3) external light, 4) dome port component of the housing, and 5) weighted frame.

135

Figure 5.2. Mean fish community metrics on newly deployed (gray) versus nearby established (black) artificial reef by sampling period for a) abundance, b) species richness, and c) Pielou’s evenness. Error bars represent ± 1 SE. Jitter along both the x and y axes is included within sampling periods to prevent the extent of error bars from being obscured in cases where they overlap.

136

Figure 5.3. Nonmetric multidimensional scaling ordination of fish community composition on newly deployed reef (gray) and established reef (black) during daytime hours. a) sampling period 1 (May 17-26), b) sampling period 2 (July 21 – August 5), and c) sampling period 3

(September 13-19). The ordination was conducted on square-root transformed fish abundance values; each plot displays a subset of the overall ordination corresponding to each sampling period. Points represent mean hourly (0700 – 1900 local time) fish community composition on newly deployed reef (gray triangles) and established reef (black circles), and ellipses, colored identically to points, indicate 50% confidence intervals.

137

Figure 5.4. Fish species abundance (mean ± 1 SE) on new reef (gray bars) and on established reef (black bars) during three sampling periods (May, July, September) for: a) Archosargus probatocephalus , b) Centropristis striata , c) Decapterus sp., and d) Haemulon aurolineatum .

138 APPENDIX 1: SUPPORTING INFORMATION FOR CHAPTER 1

Figure S1.1. Response of fish abundance to digital reef rugosity (DRR) by reef type and fish size class. Color denotes reef type: natural reefs (blue; a-d) and artificial reefs (red; e-h). Row

139

indicates fish size classes: a and e) small (1-10 cm) fishes; b and f) medium (11-29 cm) fishes; c and g) large (30-49 cm) fishes; d and h) extra-large fishes ( ≥ 50 cm). Solid lines represent unimodal relationships between DRR and fish abundance (DRR: P < 0.05, DRR 2:

P < 0.05), whereas absence of a line indicates a non-significant relationship between DRR and fish abundance and a dashed line indicates a marginally-significant relationship.

140

Table S1.1. Descriptions of thirty reefs surveyed. Mean environmental variables include digital reef rugosity (DRR), vertical relief (relief), depth, water temperature (temp), sediment standard deviation (sed; natural reefs only), and location. Date indicates month and year

(month/year) of replicate transects.

reef_name reef_type morphology DRR relief depth temp sed location date (m) (m) (m) (oC) (cm) 12/2013, Pavement & Long 6/2014, Bumpy Ledge Natural 0.1 0.6 29.3 22.5 4.1 rubble Bay 9/2014, 12/2014 12/2013, Hammerhead Pavement & Long 6/2014, Natural 0.1 0.6 25.4 21.8 3.7 Ledge rubble Bay 9/2014, 12/2014 12/2013, Lightning Pavement & Long 6/2014, Natural 0.2 0.8 28.5 21.8 8.8 Bolt Ledge rubble Bay 9/2014, 12/2014 9/2013, Pavement & Long Thumb Ledge Natural 0.2 0.8 26.5 26.5 3.9 6/2014, rubble Bay 9/2014 9/2013, Pavement & Onslow 210 Rock Natural 0.2 0.9 30.2 25.2 8.2 6/2014, rubble Bay 10/2014 9/2013, 10/2013, Pavement & Onslow 6/2014, Station Rock Natural 0.3 1.1 15.6 23.4 10.1 rubble Bay 8/2014, 10/2014, 5/2015 9/2013, Pavement & Onslow 6/2014, West Rock Natural 0.3 1.1 24.9 25.2 8.9 rubble Bay 10/2014, 5/2015 8/2013, Southwest of 10/2013, Pavement & Onslow Knuckle Natural 0.3 1.4 14.3 23.9 7.0 8/2014, rubble Bay Buoy 10/2014, 5/2015 9/2013, Pavement & Onslow 7/2014, Dallas Rocks Natural 0.3 1.5 16.5 24.7 1.7 rubble Bay 9/2014, 11/2014 23 Mile Onslow 6/2014, Natural Ledge 0.4 1.6 28.7 25.5 3.3 Ledge Bay 9/2014 9/2013, 200 / 200 Onslow Natural Ledge 0.4 1.5 25.3 25.0 1.6 9/2014, Ledge Bay 11/2014 9/2013, Onslow 7/2014, Keypost Rock Natural Ledge 0.4 1.5 15.0 25.0 8.1 Bay 10/2014, 10/2014 Onslow 9/2013, 5 Mile Ledge Natural Ledge 0.4 1.7 15.8 24.7 0.7 Bay 7/2014,

141

9/2014, 11/2014 8/2013, Northwest Onslow 6/2014, Natural Ledge 0.6 2.3 20.9 23.9 3.8 Reef Bay 10/2014, 5/2015 8/2013, Onslow 7/2014, Barge Rock Natural Ledge 0.6 2.2 16.1 24.8 6.6 Bay 9/2014, 10/2014 8/2013, 10/2013, Onslow 6/2014, 10 Fathom Natural Ledge 0.6 2.4 20.9 23.7 3.8 Bay 8/2014, 10/2014, 5/2015 7/2013, 10/2013, Concrete Onslow 6/2014, Pipes, 2006 Artificial Concrete 0.2 1.2 18.7 22.9 --- Bay 10/2014, (AR-345) 10/2014, 5/2015 9/2013, Concrete 6/2014, Onslow Pipes, 2007 Artificial Concrete 0.4 1.6 16.1 24.2 --- 8/2014, Bay (AR-342) 10/2014, 5/2015 7/2013, 10/2013, Atlantic Onslow 7/2014, Beach Bridge Artificial Concrete 0.7 2.9 15.2 23.3 --- Bay 9/2014, (AR-320) 10/2014, 5/2015 9/2013, City of Long Artificial Ship 0.4 1.8 28.0 27.7 --- 6/2014, Houston Bay 9/2014 12/20143, Unknown Long Artificial Ship 0.5 2.2 29.0 24.1 --- 6/2014, Wreck Bay 9/2014 7/2013, Theodore 10/2013, Onslow Parker (AR- Artificial Ship 0.5 2.4 10.2 24.3 --- 7/2014, Bay 315) 10/2014, 10/2014 9/2013, Onslow 6/2014, John D. Gill Artificial Ship 0.7 3.1 25.1 25.0 --- Bay 8/2014, 11/2014 Onslow 9/2014, Cassimir Artificial Ship 0.8 3.3 32.6 25.4 --- Bay 11/2014 7/2013, 10/2013, Titan (AR- Onslow 6/2014, Artificial Ship 0.8 3.5 16.6 23.0 --- 345) Bay 10/2014, 10/2014, 5/2015 6/2014, USS Indra Onslow Artificial Ship 1.1 3.8 15.0 24.0 --- 8/2014, (AR-342) Bay 5/2015 Yard Oiler Onslow FS-26 (AR- Artificial Ship 1.4 5.0 26.4 24.6 --- 6/2014 Bay 300)

142

8/2013, Spar (AR- Onslow Artificial Ship 1.4 5.1 28.3 25.1 --- 6/2014, 305) Bay 10/2014 9/2013, Long 6/2014, Raritan Artificial Ship 1.7 5.7 21.5 23.3 --- Bay 9/2014, 12/2014 9/2013, Alexander Onslow 7/2014, Ramsey (AR- Artificial Ship 2.3 6.6 12.0 24.7 --- Bay 9/2014, 370) 11/2014

143

Table S1.2. Species list from 246 fish belt-transects conducted on warm-temperate reefs of the NC continental shelf. Bold text indicates fishes in the federally managed snapper-grouper complex. Abundance values indicate the total numbers of each species observed across the

246 transects, as well as the number of each species observed on natural and artificial reefs.

Common Total Natural Artificial Family Genus species Name Abundance Abundance Abundance

Acanthuridae Acanthurus chirurgus Doctorfish 31 23 8

Acanthuridae Acanthurus coeruleus Blue Tang 1 0 1

Anguillidae Anguilla rostrata American Eel 2 2 0 Two Spot Apogonidae Apogon pseudomaculatus Cardinalfish 85 75 10 Pale Apogonidae Apogon planifrons Cardinalfish 2 2 0

Atherinopsidae Menidia menidia Silversides 24,775 10,520 14,255 Grey Balistidae Balistes capriscus Triggerfish 62 53 9 Oyster Batrachoididae Opsanus tau Toadfish 48 31 17 Flat Belonidae Ablennes hians Needlefish 3 3 0 Seaweed Blenniidae Parablennius marmoreus Blenny 461 284 177 Unknown Blenniidae Blenniidae sp. Blenny 110 22 88 Crested Blenniidae Hypleurochilus geminatus Blenny 18 0 18 Blenniidae Ophioblennius macclurei Redlip Blenny 1 1 0 Mackeral Carangidae Decapterus macarellus Scad 56,904 20,038 36,866 Carangidae Decapterus punctatus Round Scad 30,139 5,883 24,256

Carangidae Selar crumenophthalmus Big Eye Scad 9,312 52 9,260 Decapterus Carangidae Decapterus sp. Species 1,100 0 1,100 Greater Carangidae Seriola dumerili Amberjack 763 123 640 Carangidae Caranx crysos Blue Runner 257 225 32 Carangidae Carangoides bartholomaei Yellow Jack 237 72 165

Carangidae Seriola rivoliana Almaco Jack 97 13 84

Carangidae Caranx ruber Bar Jack 62 7 55

Carcharhinidae Carcharhinus plumbeus Sandbar Shark 1 0 1 Spotfin Chaetodontidae Chaetodon ocellatus Butterflyfish 23 15 8 Reef Chaetodontidae Chaetodon sedentarius Butterflyfish 11 8 3 Loggerhead Cheloniidae Caretta caretta Turtle 1 1 0 Unknown Cheloniidae Cheloniidae sp. Turtle 1 1 0 Southern Dasyatidae Dasyatis americana Stingray 8 0 8

144

Striped Diodontidae Chilomycterus schoepfi Burrfish 6 3 3 Bridled Diodontidae Chilomycterus antennatus Burrfish 2 0 2

Echeneidae Remora remora Remora 8 3 5 Atlantic Ephippidae Chaetodipterus faber Spadefish 832 338 494

Gobiidae Coryphopterus eidolon Pallid Goby 72 47 25 Gobiidae Coryphopterus glaucofraenum Bridled Goby 63 36 27 Unknown Gobiidae Gobiidae sp. Goby 54 24 30 Goldspot Gobiidae Gnatholepsis thompsoni Goby 2 2 0

Grammatidae Gramma loreto Fairy Basslet 2 2 0

Haemulidae Haemulon aurolineatum Tomtate 118,724 26,161 92,563 Unknown Juvenile Haemulidae Haemulidae sp. Grunt 37,773 14,465 23,308 Haemulidae Orthopristis chrysoptera Pigfish 1,477 1,340 137

Haemulidae Haemulon plumieri White Grunt 600 476 124 Black Haemulidae Anisotremus surinamensis Margate 28 9 19 White Haemulidae Haemulon album Margate 1 1 0 Bermuda Kyphosidae Kyphosus sectatrix Chub 3 2 1

Labridae Halichoeres bivittatus Slippery Dick 2,901 2,217 684 Painted Labridae Halichoeres caudalis Wrasse 172 154 18 Labridae Tautoga onitis Tautog 97 70 27 Bluehead Labridae Thalassoma bifasciatum Wrasse 54 15 39 Spanish Labridae Bodianus rufus Hogfish 27 4 23 Labridae Halichoeres radiatus Pudding Wife 21 10 11 Unknown Labridae Labridae sp. Wrasse 15 0 15

Labridae Lachnolaimus maximus Hogfish 5 0 5 Labridae Doratonotus megalepis Dwarf Wrasse 2 2 0

Lotidae Brosme brosme Cusk 1 0 1 Vermillion Lutjanidae Rhomboplites aurorubens Snapper 18,764 193 18,571

Lutjanidae Lutjanus campechanus Red Snapper 30 22 8 Mahogany Lutjanidae Lutjanus mahogoni Snapper 27 17 10 Lane Lutjanidae Lutjanus synagris Snapper 6 2 4 Gray Lutjanidae Lutjanus griseus Snapper 5 0 5 Yellowtail Lutjanidae Ocyurus chrysurus Snapper 1 1 0 Planehead Monacanthidae Stephanolepis hispidus Filefish 55 27 28

Mullidae Mullus auratus Red Goatfish 107 2 105 Spotted Mullidae Pseudupeneus maculatus Goatfish 56 49 7 Dwarf Mullidae Upeneus parvus Goatfish 42 8 34

145

Yellow Mullidae Mulloidichthys martinicus Goatfish 37 7 30 Reticulate Muraenidae Muraena retifera Moray 12 10 2 Goldentail Muraenidae Gymnothorax milaris Moray 4 3 1 Muraenidae Anguilliformes sp. Unknown Eel 1 1 0 Sandtiger Odontaspididae Carcharias taurus Shark 49 1 48 Goldspotted Ophichthidae Myrichthys ocellatus Eel 2 0 2 Unknown Osteichthyes Osteichthyes sp. Fish Species 22 18 4 Honeycomb Ostraciidae Acanthostracion polygonius Cowfish 1 1 0 Scrawled Ostraciidae Acanthostracion quadricornis Cowfish 1 0 1 Smooth Ostraciidae Rhinesomus triqueter Trunkfish 1 0 1

Paralichthyidae Paralichthys albigutta Gulf Flounder 36 25 11 Summer Paralichthyidae Paralichthys dentatus Flounder 22 9 13 Southern Paralichthyidae Paralichthys lethostigma Flounder 5 1 4

Phycidae Urophycis earllii Carolina Hake 22 17 5 Blue Pomacanthidae Holacanthus bermudensis Angelfish 138 73 65 Queen Pomacanthidae Holacanthus ciliaris Angelfish 9 4 5 Unknown Juvenile Pomacanthidae Pomacanthidae sp. Angelfish 3 3 0 Purple Reef Pomacentridae Chromis scotti Fish 1,022 60 962 Unknown Juvenile Pomacentridae Pomacentridae sp. Damselfish 127 101 26 Beaugregory Pomacentridae Stegastes leucostictus Damselfish 105 26 78 Cocoa Pomacentridae Stegastes variabilis Damselfish 82 69 13 Bicolor Pomacentridae Stegastes partitus Damselfish 71 42 29 Pomacentridae Chromis cyanea Blue Chromis 23 0 23 Longfin Pomacentridae Stegastes diencaeus Damselfish 9 0 9 Dusky Pomacentridae Stegastes adustus Damselfish 3 0 3 Sergeant Pomacentridae Abudefduf saxatilis major 2 1 1 Yellowtail Pomacentridae Chromis enchrysura Reef Fish 2 2 0 Night Pomacentridae Abudefduf taurus Sergeant 1 1 0

Ptereleotridae Ptereleotris calliura Blue Dartfish 27 27 0

Rachycentridae Rachycentron canadum Cobia 1 0 1 Barndoor Rajidae Dipturus laevis Skate 1 1 0

Rhincodontidae Ginglymostoma cirratum Nurse Shark 1 0 1 Green Blotch Scaridae Sparisoma atomarium Parrotfish 10 8 2 Striped Scaridae Scarus iseri Parrotfish 3 3 0

146

Sciaenidae Pareques umbrosus Cubbyu 1,332 715 617 Sciaenidae Pareques acuminatus High Hat 6 3 3 Spanish Scombridae Scomberomorus maculatus Mackerel 594 2 592 King Scombridae Scomberomorus cavalla Mackerel 250 0 250 Scombridae Euthynnus alletteratus Little Tunny 154 8 146

Scorpaenidae Pterois volitans Lionfish 47 12 35 Spotted Scorpaenidae Scorpaena plumieri Scorpionfish 3 2 1 Black Sea Serranidae Centropristis striata Bass 3,208 2,283 925 Belted Serranidae Serranus subligarius Sandfish 1,228 731 497

Serranidae Mycteroperca microlepis Gag 390 197 193 Bank Sea Serranidae Centropristis ocyurus Bass 245 200 45 Serranidae Mycteroperca phenax Scamp 199 108 91 White Spotted Serranidae Rypticus maculatus Soapfish 118 58 60 Serranidae Diplectrum formosum Sand Perch 39 34 5

Serranidae Epinephelus guttatus Red Hind 2 0 2 Barred Serranidae Hypoplectrus puella Hamlet 2 2 0 Serranidae Serranus baldwini Lantern Bass 2 2 0 Serranidae Serranus phoebe Tattler Bass 2 2 0 Harlequin Serranidae Serranus tigrinus Bass 2 2 0 Wrasse Serranidae Liopropoma eukrines Basslet 1 1 0 Yellowmouth Serranidae Mycteroperca interstitialis Grouper 1 0 1 Greater Serranidae Rypticus saponaceus Soapfish 1 0 1 Spottail Sparidae Diplodus holbrookii Pinfish 16,207 6,661 9,546 Sparidae Stenotomus chrysops Scup 1010 475 535 Sparidae Diplodus argenteus Silver Porgy 590 9 581

Sparidae Archosargus probatocephalus Sheepshead 518 146 372 Sparidae Lagodon rhomboides Pinfish 275 83 192 Longspine Sparidae Stenotomus caprinus Porgy 245 189 56 Sheepshead Sparidae Calamus penna Porgy 240 164 76 Littlehead Sparidae Calamus proridens Porgy 126 124 2 Saucereye Sparidae Calamus calamus Porgy 117 71 46 Unknown Juvenile Sparidae Sparidae sp. Porgy 100 30 70 Knobbed Sparidae Calamus nodosus Porgy 28 26 2 Jolthead Sparidae Calamus bajonado Porgy 15 10 5 Unknown Sparidae Sparidae sp. Porgy 10 6 4 Sparidae Archosargus rhomboidalis Sea Bream 4 0 4

147

Sparidae Pagrus pagrus Red Porgy 3 0 3

Sphyraenidae Sphyraena guachancho Guaguanche 790 200 590

Sphyraenidae Sphyraena barracuda Barracuda 80 7 73 Northern Sphyraenidae Sphyraena borealis Sennet 1 1 0 Inshore Synodontidae Synodus foetens Lizardfish 4 3 1 Sharpnose Tetraodontidae Canthigaster rostrata Puffer 56 23 33 Bandtail Tetraodontidae Sphoeroides spengleri Puffer 21 11 10

148

Table S1.3. GLM results for the relationship between fish abundance and environmental predictor variables by reef type and fish size class. Environmental variables include digital reef rugosity (DRR (m)), squared digital reef rugosity (DRR 2 (m)), average reef depth (m), average water temperature ( oC), and standard deviation of sediment cover (m) approximating sediment dynamics. Coefficients, standard error (SE), Z-values and P-values are provided for each environmental parameter. Bold values indicate significance. Interpretation of the pattern

(unimodal, linear, or non-significant (NS)) between rugosity and the fish abundance are displayed for each model. Model results displayed here were from the best models that we evaluated.

Predictor Standard Reef type Size class variable Coefficient error z-value P-value Natural Small Intercept -1.84 1.32 -1.39 0.16

DRR 10.29 2.79 3.68 0.0002 DRR 2 -11.08 3.06 -3.62 0.0003 Depth 0.12 0.03 4.34 <0.0001 Temperature 0.15 0.05 3.22 0.001 Sediment -0.08 0.03 -2.60 0.01 Artificial Small Intercept 4.01 0.54 7.38 <0.0001

DRR 1.21 0.78 1.55 0.12 DRR 2 -0.34 0.27 -1.27 0.20 Depth 0.11 0.25 4.38 <0.0001 Natural Medium Intercept 4.15 0.88 4.71 <0.0001

DRR 2.04 2.67 0.76 0.45 DRR 2 -1.99 2.94 -0.68 0.50 Depth 0.05 0.03 1.94 0.05 Artificial Medium Intercept 5.39 1.25 4.31 <0.0001

DRR 1.09 0.72 1.51 0.13 DRR 2 -0.37 0.24 -1.50 0.13 Depth 0.16 0.02 6.90 <0.0001 Temperature -0.11 0.05 -2.28 0.02 Natural Large Intercept 1.75 0.28 6.17 <0.0001

DRR 0.34 0.56 0.61 0.54 Sediment 0.07 0.03 2.48 0.01 Artificial Large Intercept 0.59 0.62 0.95 0.34

149

DRR 0.80 0.23 3.55 0.0003 Depth 0.08 0.03 2.71 0.007 Natural Extra large Intercept -1.65 0.75 -2.20 0.03

DRR 1.30 0.68 1.92 0.06 Depth 0.09 0.03 3.44 0.001 Artificial Extra large Intercept -0.49 0.74 -0.67 0.51

DRR 2.34 1.09 2.15 0.03 DRR 2 -0.82 0.38 -2.17 0.03 Depth 0.10 0.03 3.09 0.002

150 APPENDIX 2: SUPPORTING INFORMATION FOR CHAPTER 2

Figure S2.1. Biomass of demersal tropical fishes on artificial (dark colored) versus natural

(light colored) reefs by reef depth and season. Reef depth zones are: shallow: 5-18 m, intermediate: 18-25 m, deep: 25-35 m. Lines and associated error bars (± 1 SE) represent fitted linear mixed effects model with site as a random effect. Points indicate observed abundances. Lines, error bars, and points are jittered to prevent overlap. Table S2.5 contains model results

151

Figure S2.2. Percent cover of a-c) overall benthos, d-f) benthos comprised of benthic invertebrates, g-i) benthos comprised of macroalgae on artificial (dark colored) versus natural

(light colored) reefs by organism climate range a, d, g) temperate (blue), b, e, h) subtropical

(green), c, f, i) tropical (red). Lines and associated error bars (± 1 SE) represent fitted linear mixed effects model with site as a random effect. Lines and error bars are jittered to prevent overlap. Table 2.2 contains model results for each climate range.

152

Table S2.1. Descriptions of thirty reefs surveyed. Coordinates are provided to the nearest tenth of a degree for latitude (Lat) and longitude (Lon). Mean environmental variables include depth (Dep), digital reef rugosity (DRR), water temperature (Tem), and sediment standard deviation (Sed; natural reefs only). Date indicates month and year (month/year) of replicate transects. Coordinates of a previously unknown wreck are purposely excluded.

Reef Reef Lat Lon Dep DRR Tem Sed name type Structure (dd) (dd) (m) (m) (oC) (cm) Date 7/8/2013, Theodore Parker - 10/29/13, 7/29/14, (AR-315) Artificial Ship 34.7 76.7 10.2 0.5 24.3 ----- 10/2/14, 10/20/14 10/31/2013, Atlantic Beach - 7/29/14, 9/29/14, Bridge (AR-320) Artificial Concrete 34.7 76.8 15.2 0.8 23.0 ----- 10/20/14, 5/19/15 9/6/2013, 6/3/14, Concrete Pipes, - 8/20/14, 10/21/14, 2007 (AR-342) Artificial Concrete 34.6 77.0 16.1 0.4 24.2 ----- 5/19/15 USS Indra (AR------6/9/2014, 8/21/14, 330) Artificial Ship 34.6 76.9 15.0 1.1 24.0 5/18/15 ----- 7/18/2013, 10/28/13, 6/2/14, - 10/2/14, 10/21/14, Titan (AR-345) Artificial Ship 34.5 77.0 16.6 0.8 23.0 5/19/15 ----- 7/18/2013, 10/28/13, 6/2/14, Concrete Pipes, - 10/2/14, 10/21/14, 2006 (AR-345) Artificial Concrete 34.5 77.0 18.7 0.2 22.9 5/19/15 Yard Oiler FS------26 (AR-300) Artificial Ship 34.3 76.4 26.4 1.4 24.6 6/17/14 ------8/30/2013, Spar (AR-305) Artificial Ship 34.3 76.6 28.3 1.4 25.1 6/16/14, 10/10/14 Alexander ------9/19/2013, 7/1/14, Ramsey Artificial Ship 34.2 77.8 12.0 2.3 24.7 9/5/14, 11/13/14 ------Cassimir Artificial Ship 34.0 77.0 32.6 0.8 25.4 9/4/2014, 11/5/14 ----- 9/24/2013, - 6/24/14, 8/7/14, John D. Gill Artificial Ship 33.9 77.5 25.1 0.7 25.0 11/13/14 ----- 9/25/2013, - 6/25/14, 9/11/14, Raritan Artificial Ship 33.5 77.9 21.5 1.7 23.3 12/18/14 ------9/25/2013, City of Houston Artificial Ship 33.4 77.7 28.0 0.4 27.7 6/25/14, 9/11/14 12/4/2013, Unknown Wreck Artificial Ship ------29.0 0.5 24.1 ----- 6/25/14, 9/11/14 - Keypost Rock Natural Ledge 34.6 77.0 15.0 0.4 25.0 8.1 10/24/14 8/19/2013, - 7/30/14, 7/29/14, Barge Rock Natural Ledge 34.6 76.6 16.1 0.6 24.8 6.6 10/20/14

153

9/6/2013, 10/30/13, 6/2/14, Pavement - 8/20/14, 10/21/14, Station Rock Natural & rubble 34.6 77.1 15.6 0.3 23.4 10.1 5/19/15 8/12/2013, - 6/17/14, 10/10/14, Northwest Reef Natural Ledge 34.4 76.6 21.4 0.7 23.8 4.5 10/20/14, 5/18/15 8/20/2013, Southwest of Pavement - 10/28/13, Knuckle Buoy Natural & rubble 34.4 76.5 14.2 0.2 23.2 6.7 10/20/14, 5/18/15 8/20/2013, 10/29/13, 6/17/14, - 8/22/14, 10/20/14, 10 Fathom Natural Ledge 34.4 76.6 20.9 0.6 23.7 3.8 5/18/15 Pavement - 9/4/2013, 6/16/14, West Rock Natural & rubble 34.3 76.6 24.9 0.3 25.2 8.9 10/10/14, 5/27/15 Pavement - 9/4/2013, 6/16/14, 210 Rock Natural & rubble 34.2 76.6 30.2 0.2 25.2 8.2 10/10/14 Pavement - 9/20/2013, 7/1/14, Dallas Rocks Natural & rubble 34.2 77.6 16.5 0.3 24.7 1.7 9/4/14, 11/5/14 - 200 / 200 Ledge Natural Ledge 34.1 77.4 25.1 0.3 23.8 0.9 9/20/2013, 11/5/14 - 9/19/2013, 7/1/14, 5 Mile Ledge Natural Ledge 34.1 77.8 15.8 0.4 24.7 0.7 9/5/14, 11/13/14 - 23 Mile Ledge Natural Ledge 34.0 77.4 28.7 0.4 25.5 3.3 6/24/2014, 9/5/14 12/13/2013, Hammerhead Pavement - 6/30/14, 9/11/14, Ledge Natural & rubble 33.5 77.9 25.4 0.1 21.8 3.7 12/19/14 Pavement - 9/25/2013, Thumb Ledge Natural & rubble 33.5 77.9 26.5 0.2 26.5 3.9 6/25/14, 9/11/14 12/13/2013, Lightning Bolt Pavement - 6/25/14, 9/11/14, Ledge Natural & rubble 33.5 77.9 28.5 0.2 21.8 8.8 12/19/14 Pavement - 12/13/2013, Bumpy Ledge Natural & rubble 33.5 77.9 29.2 0.1 21.0 4.4 6/25/14, 12/15/14

154

Table S2.2. Fish, shark, and turtle species list from 226 fish belt-transects conducted on warm-temperate reefs of the NC continental shelf. Abundance values indicate the total number of each species observed across the 226 transects on artificial reefs (AR), natural reefs (NR), and in total (Total). Bold indicates members of the federally-managed snapper- grouper complex.

Family Genus Species Common Climate Functional Habitat AR NR Total name range group zone

Acanthuridae Acanthurus chirurgus Doctorfish Subtropical Herbivore Demersal 8 23 31

Acanthuridae Acanthurus coeruleus Blue tang surgeonfish Tropical Herbivore Demersal 1 0 1

Anguillidae Anguilla rostrata American eel Subtropical Invertivore Demersal 0 2 2

Apogonidae Apogon pseudomaculatus Twospot cardinalfish Tropical Invertivore Demersal 10 72 82

Apogonidae Apogon planifrons Pale cardinalfish Tropical Invertivore Demersal 0 2 2

Atherinopsidae Menidia menidia Silversides Temperate Planktivore Pelagic 14255 10320 24575

Balistidae Balistes capriscus Grey triggerfish Subtropical Invertivore Demersal 9 51 60

Batrachoididae Opsanus tau Oyster toadfish Subtropical Benthic Demersal 17 31 48 Carnivore

Belonidae Ablennes hians Flat needlefish Subtropical Piscivore Demersal 0 3 3

Blenniidae Parablennius marmoreus Seaweed blenny Subtropical Planktivore Demersal 177 271 448

Blenniidae Blenniidae sp. Unknown blenny Tropical Omnivore Demersal 88 22 110

Blenniidae Hypleurochilus geminatus Crested blenny Subtropical Omnivore Demersal 18 0 18

Blenniidae Ophioblennius macclurei Redlip blenny Tropical Omnivore Demersal 0 1 1

Carangidae Decapterus macarellus Mackerel scad Subtropical Planktivore Pelagic 36866 19636 56502

Carangidae Decapterus punctatus Round scad Subtropical Planktivore Pelagic 24256 5883 30139

Carangidae Selar crumenophthalmus Bigeye scad Subtropical Carnivore Pelagic 9260 52 9312

Carangidae Decapterus sp. Scad species Subtropical Planktivore Pelagic 1100 0 1100

Carangidae Seriola dumerili Greater amberjack Subtropical Piscivore Pelagic 640 114 754

Carangidae Caranx crysos Blue runner Subtropical Benthic Pelagic 32 225 257 Carnivore

Carangidae Carangoides bartholomaei Yellow jack Subtropical Piscivore Pelagic 165 72 237

Carangidae Seriola rivoliana Almaco jack Subtropical Piscivore Pelagic 84 13 97

155

Carangidae Caranx ruber Bar jack Subtropical Omnivore Pelagic 55 7 62

Carcharhinidae Carcharhinus plumbeus Sandbar shark Subtropical Carnivore Shark 1 0 1

Chaetodontidae Chaetodon ocellatus Spotfin butterflyfish Tropical Omnivore Demersal 8 15 23

Chaetodontidae Chaetodon sedentarius Reef butterflyfish Subtropical Omnivore Demersal 3 7 10

Cheloniidae Caretta caretta Loggerhead turtle Subtropical Herbivore Turtle 0 1 1

Cheloniidae Cheloniidae sp. Unknown turtle Subtropical Herbivore Turtle 0 1 1

Dasyatidae Dasyatis americana Southern stingray Subtropical Benthic Demersal 8 0 8 Carnivore

Diodontidae Chilomycterus schoepfi Striped burrfish Tropical Invertivore Demersal 3 3 6

Diodontidae Chilomycterus antennatus Bridled burrfish Tropical Invertivore Demersal 2 0 2

Echeneidae Remora remora Remora Subtropical Carnivore Demersal 5 3 8

Ephippidae Chaetodipterus faber Atlantic spadefish Subtropical Invertivore Pelagic 494 248 742

Gobiidae Coryphopterus eidolon Pallid goby Tropical Herbivore Demersal 25 47 72

Gobiidae Coryphopterus glaucofraenum Bridled goby Tropical Omnivore Demersal 25 34 59

Gobiidae Gobiidae sp. Unknown goby Tropical Omnivore Demersal 26 24 50

Gobiidae Gnatholepsis thompsoni Goldspot goby Tropical Omnivore Demersal 0 2 2

Grammatidae Gramma loreto Fairy basslet Tropical Invertivore Demersal 0 2 2

Ginglymostom- Ginglymostoma cirratum Nurse shark Subtropical Benthic Shark 1 0 1 atidae Carnivore

Haemulidae Haemulon aurolineatum Tomtate Subtropical Invertivore Pelagic 92563 25676 118239

Haemulidae Haemulidae sp. Unknown juvenile grunt Tropical Invertivore Demersal 23308 14461 37769

Haemulidae Orthopristis chrysoptera Pigfish Temperate Invertivore Demersal 137 1312 1449

Haemulidae Haemulon plumierii White grunt Subtropical Invertivore Demersal 124 464 588

Haemulidae Anisotremus surinamensis Black margate Subtropical Invertivore Demersal 19 9 28

Haemulidae Haemulon album White margate Tropical Invertivore Demersal 0 1 1

Kyphosidae Kyphosus sectatrix Bermuda chub Subtropical Omnivore Demersal 1 2 3

Labridae Halichoeres bivittatus Slippery dick Tropical Invertivore Demersal 683 2005 2688

Labridae Halichoeres caudalis Painted wrasse Subtropical Invertivore Demersal 18 151 169

Labridae Tautoga onitis Tautog Temperate Benthic Demersal 26 70 96 Carnivore

Labridae Thalassoma bifasciatum Bluehead wrasse Tropical Planktivore Demersal 39 15 54

156

Labridae Bodianus rufus Spanish hogfish Tropical Invertivore Demersal 23 4 27

Labridae Halichoeres radiatus Puddingwife Tropical Invertivore Demersal 11 10 21

Labridae Labridae sp. Unknown wrasse Tropical Invertivore Demersal 15 0 15

Labridae Lachnolaimus maximus Hogfish Subtropical Invertivore Demersal 5 0 5

Lotidae Brosme brosme Cusk Temperate Invertivore Demersal 1 0 1

Lutjanidae Rhomboplites aurorubens Vermilion snapper Subtropical Benthic Demersal 18571 193 18764 Carnivore

Lutjanidae Lutjanus campechanus Red snapper Subtropical Benthic Demersal 8 22 30 Carnivore

Lutjanidae Lutjanus mahogoni Mahogany snapper Subtropical Benthic Demersal 9 17 26 Carnivore

Lutjanidae Lutjanus synagris Lane snapper Subtropical Benthic Demersal 4 2 6 Carnivore

Lutjanidae Lutjanus griseus Gray snapper Subtropical Benthic Demersal 5 0 5 Carnivore

Lutjanidae Ocyurus chrysurus Yellowtail snapper Subtropical Invertivore Demersal 0 1 1

Monacanthidae Stephanolepis hispidus Planehead filefish Subtropical Benthic Demersal 28 24 52 Carnivore

Mullidae Mullus auratus Red goatfish Subtropical Benthic Demersal 105 2 107 Carnivore

Mullidae Pseudupeneus maculatus Spotted goatfish Subtropical Benthic Demersal 7 45 52 Carnivore

Mullidae Upeneus parvus Dwarf goatfish Tropical Benthic Demersal 34 8 42 Carnivore

Mullidae Mulloidichthys martinicus Yellow goatfish Subtropical Benthic Demersal 30 7 37 Carnivore

Muraenidae Muraena retifera Reticulate moray Subtropical Benthic Demersal 2 10 12 Carnivore

Muraenidae Gymnothorax miliaris Goldentail moray Subtropical Benthic Demersal 1 3 4 Carnivore

Muraenidae Muraenidae sp . Unknown moray eel Subtropical Piscivore Demersal 0 1 1

Odontaspididae Carcharias taurus Sand tiger shark Subtropical Piscivore Pelagic 48 1 49

Ophichthidae Myrichthys ocellatus Goldspotted eel Tropical Benthic Demersal 2 0 2 Carnivore

Osteichthyes Osteichthyes sp. Unknown fish species Subtropical Omnivore Demersal 3 12 15

Ostraciidae Acanthostracion polygonius Honeycomb cowfish Tropical Invertivore Demersal 0 1 1

Ostraciidae Acanthostracion quadricornis Scrawled cowfish Subtropical Invertivore Demersal 1 0 1

Ostraciidae Lactophrys triqueter Smooth trunkfish Subtropical Invertivore Demersal 1 0 1

Paralichthyidae Paralichthys albigutta Gulf flounder Subtropical Benthic Demersal 11 25 36 Carnivore

Paralichthyidae Paralichthys dentatus Summer flounder Temperate Benthic Demersal 13 8 21 Carnivore

Paralichthyidae Paralichthys lethostigma Southern flounder Subtropical Benthic Demersal 4 1 5 Carnivore

157

Phycidae Urophycis earllii Carolina hake Subtropical Benthic Demersal 5 17 22 Carnivore

Pomacanthidae Holacanthus bermudensis Blue angelfish Subtropical Invertivore Demersal 65 66 131

Pomacanthidae Holacanthus ciliaris Queen angelfish Subtropical Invertivore Demersal 5 4 9

Pomacanthidae Pomacanthidae sp. Unknown juvenile Subtropical Invertivore Demersal 0 3 3 angelfish

Pomacentridae Chromis scotti Purple reeffish Tropical Planktivore Demersal 962 60 1022

Pomacentridae Pomacentridae sp. Unknown juvenile Tropical Herbivore Demersal 26 85 111 damselfish

Pomacentridae Stegastes leucostictus Beaugregory Tropical Omnivore Demersal 78 26 104 damselfish

Pomacentridae Stegastes variabilis Cocoa damselfish Tropical Herbivore Demersal 13 69 82

Pomacentridae Stegastes partitus Bicolor damselfish Tropical Herbivore Demersal 29 42 71

Pomacentridae Chromis cyanea Blue chromis Tropical Planktivore Demersal 23 0 23

Pomacentridae Stegastes diencaeus Longfin damselfish Tropical Omnivore Demersal 9 0 9

Pomacentridae Stegastes adustus Dusky damselfish Tropical Herbivore Demersal 3 0 3

Pomacentridae Abudefduf saxatilis Sergeant-major Tropical Omnivore Demersal 1 1 2

Pomacentridae Chromis enchrysura Yellowtail reeffish Tropical Planktivore Demersal 0 2 2

Pomacentridae Abudefduf taurus Night sergeant Subtropical Omnivore Demersal 0 1 1

Ptereleotridae Ptereleotris calliura Blue dartfish Tropical Planktivore Demersal 0 27 27

Rachycentridae Rachycentron canadum Cobia Subtropical Carnivore Pelagic 1 0 1

Rajidae Dipturus laevis Barndoor skate Temperate Carnivore Demersal 0 1 1

Scaridae Sparisoma atomarium Greenblotch parrotfish Tropical Herbivore Demersal 2 7 9

Scaridae Scarus iseri Striped parrotfish Subtropical Herbivore Demersal 0 3 3

Sciaenidae Pareques umbrosus Cubbyu Subtropical Benthic Demersal 617 669 1286 Carnivore

Sciaenidae Pareques acuminatus High-hat Tropical Benthic Demersal 3 3 6 Carnivore

Scombridae Scomberomorus maculatus Spanish mackerel Subtropical Piscivore Pelagic 592 2 594

Scombridae Scomberomorus cavalla King mackerel Tropical Piscivore Pelagic 250 0 250

Scombridae Euthynnus alletteratus Little tunny Tropical Piscivore Pelagic 146 8 154

Scorpaenidae Pterois volitans Lionfish Tropical Piscivore Demersal 35 12 47

Scorpaenidae Scorpaena plumieri Spotted scorpionfish Subtropical Benthic Demersal 1 2 3 Carnivore

Serranidae Centropristis striata Black sea bass Temperate Benthic Demersal 919 2116 3035 Carnivore

158

Serranidae Serranus subligarius Belted sandfish Subtropical Invertivore Demersal 494 703 1197

Serranidae Mycteroperca microlepis Gag Subtropical Benthic Demersal 190 185 375 Carnivore

Serranidae Centropristis ocyurus Bank sea bass Subtropical Benthic Demersal 45 197 242 Carnivore

Serranidae Mycteroperca phenax Scamp Subtropical Piscivore Demersal 91 106 197

Serranidae Rypticus maculatus White spotted soapfish Subtropical Benthic Demersal 60 55 115 Carnivore

Serranidae Diplectrum formosum Sand perch Subtropical Benthic Demersal 5 34 39 Carnivore

Serranidae Epinephelus guttatus Red hind Tropical Invertivore Demersal 2 0 2

Serranidae Hypoplectrus puella Barred hamlet Tropical Invertivore Demersal 0 2 2

Serranidae Serranus baldwini Lantern bass Tropical Benthic Demersal 0 2 2 Carnivore

Serranidae Serranus phoebe Tattler bass Subtropical Invertivore Demersal 0 2 2

Serranidae Serranus tigrinus Harlequin bass Tropical Invertivore Demersal 0 2 2

Serranidae Liopropoma eukrines Wrasse bass Subtropical Benthic Demersal 0 1 1 Carnivore

Serranidae Mycteroperca interstitialis Yellowmouth grouper Subtropical Piscivore Demersal 1 0 1

Serranidae Rypticus saponaceus Greater soapfish Tropical Benthic Demersal 1 0 1 Carnivore

Sparidae Diplodus holbrookii Spottail pinfish Subtropical Omnivore Demersal 10053 6513 16566

Sparidae Stenotomus chrysops Scup Subtropical Benthic Demersal 533 474 1007 Carnivore

Sparidae Archosargus probatocephalus Sheepshead Subtropical Omnivore Demersal 372 135 507

Sparidae Lagodon rhomboides Pinfish Subtropical Omnivore Demersal 192 83 275

Sparidae Calamus penna Sheepshead porgy Tropical Invertivore Demersal 76 161 237

Sparidae Stenotomus caprinus Longspine porgy Subtropical Benthic Demersal 56 171 227 Carnivore

Sparidae Calamus proridens Littlehead porgy Subtropical Invertivore Demersal 2 122 124

Sparidae Calamus calamus Saucereye porgy Subtropical Invertivore Demersal 46 67 113

Sparidae Calamus nodosus Knobbed porgy Subtropical Invertivore Demersal 2 26 28

Sparidae Calamus bajonado Jolthead porgy Subtropical Invertivore Demersal 5 10 15

Sparidae Sparidae sp. Unknown porgy Tropical Invertivore Demersal 74 6 80

Sparidae Archosargus rhomboidalis Sea bream Subtropical Omnivore Demersal 4 0 4

Sparidae Pagrus pagrus Red porgy Subtropical Invertivore Demersal 3 0 3

Sphyraenidae Sphyraena guachancho Guaguanche Subtropical Piscivore Pelagic 590 200 790

159

Sphyraenidae Sphyraena barracuda Barracuda Subtropical Piscivore Pelagic 73 7 80

Sphyraenidae Sphyraena borealis Northern sennet Subtropical Piscivore Pelagic 0 1 1

Synodontidae Synodus foetens Inshore lizardfish Tropical Piscivore Demersal 1 3 4

Tetraodontidae Canthigaster rostrata Sharpnose puffer Tropical Omnivore Demersal 33 23 56

Tetraodontidae Sphoeroides spengleri Bandtail puffer Subtropical Benthic Demersal 10 11 21 Carnivore

160

Table S2.3. Benthic invertebrate and macroalgae species list from 226 photoquadrat benthic photoquadrats collected on warm-temperate reefs of the NC continental shelf. Cover values indicate the mean percent cover of each species observed across the 226 transects for artificial reefs (AR), natural reefs (NR), and in total (Total). For example, Occulina sp. exhibited 3.58% mean cover on artificial reefs, 0.57% mean cover on natural reefs, and

4.15% mean cover across the total 226 transects.

Broad Phyla Species Climate range AR NR Total category Invertebrate Annelida Polychaete / Tube Worm Subtropical 0.01 0.11 0.12

Invertebrate Annelida Filograna implexa Temperate 0.01 0.03 0.04

Invertebrate Annelida Serpulidae sp. Subtropical 0.01 0.00 <0.01

Invertebrate Arthropoda Anoplodactylus sp. Temperate 0.01 <0.01 0.01

Invertebrate Arthropoda Mithraculus forceps Subtropical <0.001 0.00 <0.001

Invertebrate Bryozoa Encrusting Bryozoan Subtropical 0.25 0.06 0.31

Invertebrate Bryozoa Amathia sp. Temperate 0.14 0.03 0.16

Invertebrate Bryozoa Schizoporella floridana Temperate <0.001 0.01 0.01

Invertebrate Bryozoa Unknown Bryozoan Subtropical 0.00 0.01 0.01

Invertebrate Cnidaria Oculina sp. Subtropical 3.58 0.57 4.15

Invertebrate Cnidaria Telesto sp. Subtropical 1.30 1.22 2.52

Invertebrate Cnidaria Titanideum frauenfeldii Tropical 0.02 0.96 0.98

Invertebrate Cnidaria Anemone / Zoanthid Temperate 0.49 0.04 0.53

Invertebrate Cnidaria Thesea nivea Tropical 0.23 0.21 0.45

Invertebrate Cnidaria Unknown Hydroid Subtropical 0.04 0.34 0.38

Invertebrate Cnidaria Aglaophenia trifida Subtropical <0.01 0.27 0.28

Invertebrate Cnidaria Unidentified Hard Coral Tropical 0.11 0.02 0.13

Invertebrate Cnidaria Paracyathus pulchellus / Tropical 0.06 0.02 0.08 Phyllangia americana Invertebrate Cnidaria Leptogorgia hebes Subtropical 0.08 0.00 0.08

Invertebrate Cnidaria Unidentified Octocoral Subtropical 0.01 0.04 0.06

161

Invertebrate Cnidaria Unknown Invasive Subtropical 0.04 0.00 0.04 Octocoral Invertebrate Cnidaria Leptogorgia punicea Subtropical 0.03 0.00 0.03

Invertebrate Cnidaria Leptogorgia virgulata Temperate 0.00 0.02 0.02

Invertebrate Cnidaria Muricea pendula Subtropical 0.00 0.01 0.01

Invertebrate Cnidaria Solenastrea hyades Subtropical <0.01 3.54 <0.01

Invertebrate Echinodermata Arbacia puntculata Temperate 0.08 0.14 0.22

Invertebrate Echinodermata Sea Stars Subtropical <0.01 0.01 0.01

Invertebrate Echinodermata Echinometra lucunter Tropical <0.01 3.54 <0.01

Invertebrate Echinodermata Lytechinus variegatus Subtropical <0.01 0.00 <0.01

Invertebrate Echinodermata Common Sea Star Temperate 0.00 <0.01 <0.001

Invertebrate Hydrozoa / Bryozoa Hydroid / Bryozoan Subtropical 6.94 3.54 10.48

Invertebrate Mollusca Nudibranch sp. Subtropical <0.001 0.00 <0.01

Invertebrate Porifera Cliona sp. Temperate 0.74 0.18 0.92

Invertebrate Porifera Chondrilla nucula Tropical 0.01 0.29 0.30

Invertebrate Porifera Upright Sponge Subtropical 0.07 0.05 0.12

Invertebrate Porifera Ircinia campana Tropical 0.00 0.12 0.12

Invertebrate Porifera Smenospongia Subtropical 0.02 0.00 0.02 cerebriformis Invertebrate Porifera Ircinia felix Temperate <0.01 0.00 <0.01

Invertebrate Porifera Desmapsamma anchorata Temperate 0.00 <0.01 <0.01

Invertebrate Porifera Aplysina fulva Tropical 0.00 <0.01 <0.01

Invertebrate Porifera Aplysilla longispina Subtropical <0.01 <0.001 <0.01

Invertebrate Porifera Chondrosia collectrix Tropical <0.01 0.00 <0.01

Invertebrate Porifera Dysidea fragilis Temperate <0.01 0.00 <0.01

Invertebrate Porifera Spirastrella sp. Subtropical <0.001 0.00 <0.001

Invertebrate Porifera / Chordata Encrusting Sponge / Subtropical 1.43 0.73 2.16605 Tunicate Invertebrate Porifera / Chordata Unkown Sponge / Subtropical 0.12 0.26 0.38244 Tunicate Invertebrate Porifera / Chordata Chondrosia sp. / Subtropical 0.07 0.01 0.08086 Didemnum sp. Invertebrate Chordata Symplegma sp. Subtropical 0.05 0.04 0.08931

Invertebrate Chordata Aplidium sp. Temperate <0.001 0.07 0.06615

162

Invertebrate Chordata Unknown Tunicate Subtropical 0.03 0.02 0.05367

Invertebrate Chordata Stomozoa roseola Tropical 0.02 <0.01 0.02476

Invertebrate Chordata Euherdmania gigantea Tropical <0.001 0.01 0.01531

Invertebrate Chordata Styela plicata / Molgula Temperate <0.01 <0.001 <0.01 occidentalis Invertebrate Chordata Distaplia bermudensis Subtropical 0.00 <0.001 <0.001

Invertebrate Chordata Botrylloides sp. Temperate <0.001 0.00 <0.001

Macroalgae Chlorophyta Codium fragile Temperate <0.01 0.18 0.18

Macroalgae Chlorophyta Unidentified Green Algae Subtropical 0.11 0.03 0.14

Macroalgae Chlorophyta Codium carolinianum Subtropical 0.00 0.01 0.01

Macroalgae Chlorophyta Cladophora prolifera Subtropical 0.00 <0.01 <0.01

Macroalgae Chlorophyta Udotea cyathiformis Subtropical 0.00 <0.01 <0.01

Macroalgae Chlorophyta Avrainvillea longicaulis Tropical 0.00 <0.001 <0.001

Macroalgae Phaeophyta Sargassum sp. Temperate 0.78 3.04 3.83

Macroalgae Phaeophyta Dictyopteris hoytii Subtropical <0.01 1.03 1.04

Macroalgae Phaeophyta Unidentified Brown Algae Subtropical 0.12 0.56 0.68

Macroalgae Phaeophyta Dictyota sp. Subtropical 0.19 0.35 0.54

Macroalgae Phaeophyta Lobophora variegata Subtropical 0.20 0.26 0.47

Macroalgae Phaeophyta Dictyopteris Subtropical <0.01 0.19 0.19 membranacea Macroalgae Phaeophyta Padina sp. Subtropical <0.01 0.03 0.03

Macroalgae Phaeophyta Colpomenia sinuosa Subtropical 0.00 <0.01 <0.01

Macroalgae Rhodophyta Solieria filiformis Temperate 0.20 3.19 3.39

Macroalgae Rhodophyta Crustose Coralline Algae Temperate 1.09 1.06 2.15

Macroalgae Rhodophyta Red Turf-Type Algae Subtropical 1.19 0.35 1.55

Macroalgae Rhodophyta Red Strap-Type / Red Subtropical 0.62 0.22 0.84 Filamentous-Type Algae Macroalgae Rhodophyta Peyssonnelia sp. Temperate 0.39 0.15 0.54

Macroalgae Rhodophyta Gracilaria sp. / Subtropical 0.02 0.33 0.34 Rhodymenia sp. Macroalgae Rhodophyta Amphiroa beauvoisii Subtropical 0.05 0.24 0.28

Macroalgae Rhodophyta Unidentified Red Algae Subtropical 0.04 0.24 0.28

Macroalgae Rhodophyta Botryocladia occidentalis Subtropical 0.00 0.11 0.11

163

Macroalgae Rhodophyta Red Gooey-Type Algae Subtropical <0.001 0.10 0.10

Macroalgae Rhodophyta Halymenia floridana Subtropical 0.00 0.05 0.05

Macroalgae Rhodophyta Palisada corallopsis Subtropical 0.00 0.03 0.03

Macroalgae Rhodophyta Halymenia floresia Subtropical 0.00 0.02 0.02

Macroalgae Rhodophyta Eucheuma isiforme Tropical 0.00 0.02 0.02

Macroalgae Rhodophyta Chondria sp. Subtropical 0.00 <0.01 <0.01

Macroalgae Rhodophyta Chrysemenia Subtropical 0.00 <0.01 <0.01 enteromorpha Macroalgae Rhodophyta Halymenia trigona Subtropical 0.00 <0.001 <0.001

Macroalgae Unknown Black Crust-Type Algae <0.01 0.00 <0.01

Substrate Substrate Sediment 8.88 19.63 28.51

Substrate Substrate Shell / Shell Hash 3.22 1.83 5.05

Substrate Substrate Wreck 1.03 <0.01 1.04

Substrate Substrate Rock / Rubble 0.16 0.18 0.35

Substrate Substrate Rock / Rubble 0.03 0.17 0.21

Fuzz Fuzz Fuzz (Unidentifiable Biological Matter) 11.49 9.52 21.02

164

Table S2.4. Linear mixed effects model results for fish biomass by climate zone. Reef type, depth, season, and associated interaction between reef type and depth (Reef type : Depth) were included as fixed effects. Site was included as a random effect. F-values and p-values for each climate range (temperate, subtropical, tropical) are provided. Best models are reported here for (a) demersal, reef-associated fishes, (b) pelagic, water-column associated fishes, and (c) entire fish community. Bold values indicate significance with significance level displayed as: * p < 0.05, ** p < 0.01, *** p < 0.001 , **** p < 0.0001 . Parameters excluded from the model are indicated by -- --.

(a) Demersal Temperate demersal Subtropical demersal Tropical demersal

F-value p-value F-value p-value F-value p-value Intercept 16.04 <0.0001**** 31.47 <0.0001**** 6.42 0.01* Reef type 2.68 0.11 4.62 0.04* 3.35 0.08 Depth ------1.66 0.2

Reef type : depth ------3.35 0.04* Season ------3 0.06 1.32 0.27

Reef type : season ------3.55 0.03* 0.66 0.52 Depth : season ------3.17 0.02* Reef type : depth : season ------3.27 0.02*

(b) Pelagic Temperate pelagic Subtropical pelagic Tropical pelagic

F-value p-value F-value p-value F-value p-value Intercept 2.96 0.09 7.6 0.007** 2.93 0.09 Reef type 0.38 0.54 6.31 0.02* 2.38 0.13

Depth ------3.14 0.049* ------Reef type : depth ------3.36 0.04* ------

F-value p-value F-value p-value F-value p-value

Intercept 18.6 <0.0001**** 10.18 0.002** 6.46 0.01* Reef type 2.41 0.13 7.21 0.01* 4.03 0.055 Depth ------3.22 0.045* ------

Reef type : depth ------3.32 0.04* ------

165

Table S2.5. Fishes not commonly reported as far north as surveyed reefs. Each species includes its scientific name, common name, climate range, and northern latitude (N_lat;

(Froese and Pauly 2016), as well as mean count ± standard error on artificial

(Artificial_Mean; N = 55) and natural (Natural_Mean; N = 63). Reef preference indicates whether fishes exclusively occurred on artificial (AR_exclusive) or natural (NR_exclusive) reefs or whether they generally resided on artificial (AR) or natural (NR) reefs. Prey consumed by each species is provided. Northern latitudes represent the commonly documented and preferred latitudes from Fishbase (Froese and Pauly 2016). Some may seem counterintuitive. For example, Mulloidichtys martinicus has a northern latitude of 32 °N and is subtropical species while Thalassoma bifasciatum has a northern latitude of 32 °N but is classified as tropical. These classifications are based on preferred and commonly documented ranges, as well as climate ranges assigned in Whitfield et al. 2014.

Scientific Common Climate N Artificial Natural Reef_ name name range lat (mean ± se) (mean ± se) preference Prey

Chromis Blue AR_ cyanea chromis Tropical 32 0.21 ± 0.20 0.00 ± 0.00 exclusive Planktivore

Myrichthys Goldspotted AR_ Benthic ocellatus eel Tropical 25 0.02 ± 0.02 0.00 ± 0.00 exclusive Carnivore

Mycteroperca Yellowmouth AR_ interstitialis grouper Subtropical 33 0.01 ± 0.01 0.00 ± 0.00 exclusive Piscivore

Thalassoma Bluehead bifasciatum wrasse Tropical 32 0.35 ± 0.24 0.12 ± 0.09 AR Planktivore

Canthigaster Sharpnose rostrata puffer Tropical 30 0.30 ± 0.12 0.18 ± 0.06 AR Omnivore

Mulloidichthys Yellow Benthic martinicus goatfish Subtropical 32 0.27 ± 0.26 0.06 ± 0.06 AR Carnivore Invertivore; Bodianus Spanish eat rufus hogfish Tropical 32 0.21 ± 0.09 0.03 ± 0.02 AR parasites Invertivore; Anisotremus Black also surinamensis margate Subtropical 30 0.17 ± 0.12 0.07 ± 0.05 AR zooplankto Holacanthus Queen ciliaris angelfish Subtropical 32 0.05 ± 0.03 0.03 ± 0.03 AR Invertivore

166

Pareques High Benthic acuminatus hat Tropical 32 0.03 ± 0.02 0.02 ± 0.02 AR Carnivore

Calamus Sheepshead penna porgy Tropical 30 0.69 ± 0.23 1.28 ± 0.50 NR Invertivore

Stegastes Cocoa variabilis damselfish Tropical 30 0.12 ± 0.05 0.55 ± 0.34 NR Herbivore

Sparisoma Green blotch atomarium parrotfish Tropical 32 0.02 ± 0.01 0.06 ± 0.04 NR Herbivore

*Abudefduf NR_ Omnivore; taurus Night sergeant Subtropical 24 0.00 ± 0.00 0.01 ± 0.01 exclusive also algae

Chromis Yellowtail NR_ enchrysura reeffish Tropical 32 0.00 ± 0.00 0.02 ± 0.02 exclusive Planktivore

Haemulon NR_ album White margate Tropical 33 0.00 ± 0.00 0.01 ± 0.01 exclusive Invertivore

Serranus Lantern NR_ Benthic baldwini bass Tropical 25 0.00 ± 0.00 0.02 ± 0.02 exclusive Carnivore

Tattler NR_ Serranus phoebe bass Subtropical 32 0.00 ± 0.00 0.02 ± 0.02 exclusive Invertivore

Serranus NR_ tigrinus Harlequin bass Tropical 32 0.00 ± 0.00 0.02 ± 0.02 exclusive Invertivore

167 APPENDIX 3: SUPPORTING INFORMATION FOR CHAPTER 3

Text S3.1: Detailed methods for quantifying fishes and zooplankton.

The SBES (SIMRAD EK60; 7º beam angle) emitted sound pulses downwards into the water at three frequencies: 38 kHz, 120 kHz, and 200 kHz to detect zooplankton, fishes, the shipwreck structure, and the surrounding seafloor on transect lines established around each shipwreck. For the 38 kHz frequency, the pulse length was 0.256 µs, whereas for the

120 kHz and 200 kHz, the pulse length was 0.128 µs. To reduce acoustic interference among the sonar transducers, the SBES emitted pings when triggered by the MBES. As such, the ping rate for the SBES was determined by parameters set for the MBES. The SBES was calibrated using a tungsten carbide sphere to enable accurate measurements of fish size

(Foote et al. 1987).

SBES data were processed with Echoview version 8.0 (Myriax Software Pty. Ltd

2017) to quantify the spatial distribution of zooplankton and fishes around each shipwreck.

Data were corrected to account for the location of the transducers and the motion of the ship

(pitch and roll), which was logged during data collection with the differential GPS and

Applanix POS M/V motion sensor. We verified that the echosounder-detected bottom corresponding to the seafloor and shipwrecks was accurate; if the bottom had been assigned inaccurately, then it was manually corrected. Data beneath the corrected bottom-line were excluded from further analyses. To remove acoustic interference at the water surface created by bubbles, we excluded from the surface (0 m depth) down to 7.5 m beneath the surface. If the bubbles extended deeper than 7.5 m into the water column, then they were manually removed up to their deepest extent. Impulse noise from the ADCP was removed from SBES data using an algorithm that detected samples containing noise and then replaced those

168

samples with the mean values from the surrounding samples (Ryan et al. 2015)(settings: vertical window = 5 samples, horizontal context window = 5 pings, sample threshold = 10 dB).

Individual fishes and schools of fishes were identified from the 120 kHz SBES data.

Individual fishes were detected using a 2D fish tracking algorithm (Echoview 2017). We visually inspected the resulting fish tracks to verify that sequential single targets were identified correctly as either fish tracks or not. Metrics describing each individual fish track were exported for further analysis. For fishes occurring in schools, we applied the SHAPES algorithm (Barange 1994) separately to each school to delineate the boundaries of the school.

We extracted two sets of data from the schools. First, we applied the algorithm using a minimum volume backscattering coefficient ( Sv) of -65 dB. The resulting schools were integrated to determine the volume backscattering strength of each school and position of the school, and these values were exported for further analysis. Second, we defined the perimeter of each school where fishes were visible as distinct targets. To select only the perimeter of the school, we applied the SHAPES algorithm with a minimum Sv of -40 dB corresponding to the strongest return signals in each school where fishes overlapped since they were densely aggregated. These densely aggregated fishes were masked from the school, leaving only those loosely-spaced and clearly-defined fishes in the perimeter of the schools. Fishes in the perimeter of the schools were then detected as single targets and exported for further analysis. To avoid double counting fishes in the perimeter of schools that were distinct individual fishes as both schooling fishes and as individual fishes, the loosely-aggregated fishes in the perimeter of the schools were included in schooling fish analyses but masked from individual fish analyses.

169

All fish data were exported from Echoview in cells measuring 5 m (horizontal) by 1 m (vertical, depth) and a threshold of -65 dB. For each cell, the approximate volume of water sampled (Vcell ) was calculated automatically within Echoview during data export in a two- step process (Kieser and Mulligan 1984). First, the volume sampled for one ping ( Vping ) was calculated as:

∅ = sin − 1 2 where l is the length of the cell along the cruise track [m], N is the number of pings in the cell to be analyzed ( N = 1 for Vping ), ϕ is the across track beam angle, δ is a parameter indicating data to be excluded from the calculation ( δ = 0; e.g., sample below the bottom line, sample above the surface line, or sample includes no data or bad data) or otherwise included in the calculation ( δ =1), and R represent the range [m], measured along the beam axis, from the start ( Ri) to the end ( Ri+1 ) of the analysis domain. Second, the volume sampled per ping

(Vping ) for the pings inside of each cell was summed to obtain the volume sampled per cell

(Vcell ) as:

= 2

The resulting volume sampled per cell ( Vcell ), known as the ‘wedge volume sampled,’ approximates the sampled volume of water based on several assumptions: 1) distance traveled over the analysis domain (e.g., each cell) is a straight line and 2) speed of the survey vessel and ping rate are uniform. This approach for calculating the volume sampled accounts for time-varied gain and is appropriate for vessel speeds and/or ping rates that result in overlapping or non-overlapping pings (Kieser and Mulligan 1984). Because these volumes

170

were used later to calculate the density of individual fishes, rather than schooling fish density or zooplankton backscatter, these volumes were calculated from the single target detection echogram that excluded portions of the cell categorized as wreck structure or seafloor, as well as portions of the cell classified as fish schools. This approach ensured that the calculated volume of the cell matched the actual volume in which individual fishes could be detected, providing correct density values.

Exported metrics for both individual fishes (e.g., outside of schools) and fish schools were processed with R, version 3.3.2 (R Core Team 2016) to calculate fish density per cell (5 m horizontal x 1 m vertical). For individual fishes, we divided the number of fish per cell

(Nfish ) obtained from the tracked fishes by the approximate volume of water sampled per cell

(Vcell ) obtained from the single target detection echogram to calculate the density of individual fishes (e.g., non-schooling) per cell ( ρvolume (individual) ) (Parker-Stetter et al. 2009):

ℎ = 3 If the track of an individual fish spanned more than one cell, the fish was assigned to the cell where the endpoint of its track occurred.

For the entirety of each fish school (e.g., extent of each school over multiple cells), the mean target strength ( TS ) of fish detected in the perimeter of the school was first converted from the logarithmic to linear form as the backscattering cross section ( σbs ):

⁄ = 10 4 Second, for each cell that included part of the school, the volume backscattering strength of each cell attributed to the school ( Sv) was converted from the logarithmic to linear form as the volume backscattering coefficient ( sv):

⁄ = 10 5

171

Third, the volume density of each school ( ρvolume (school) ) per cell was calculated by dividing the volume backscattering coefficient (sv) by the backscattering cross section (σbs ) corresponding to the particular school of fishes:

ℎ = 6 This approach assumed that when a school spanned multiple cells, all of the fishes in the school were the same species and thus had the same backscattering cross section (σbs ).

The volume density of each school and the volume density of the individual fishes were added together for each cell to calculate the total volume density of fishes per cell

(ρvolume (total) ) as:

ℎ = + 7 Total volume density of fishes was calculated for all fishes, as well as fishes corresponding to three size classes: small (< 11 cm fish length), medium (11-29 cm fish length), and large

(>29 cm fish length). Size class assignments were based on the general logarithmic equation for the relationship between mean target strength (TS mean ) and fish total length (Love 1977):

.⁄ . ℎ = 10 8

For individual fishes, these assignments were based on the TS mean of each individual fish. For schooling fishes, size class assignments were based on TS mean of fishes in the perimeter of the entire school, again assuming each school was composed of fishes of the same size.

Zooplankton were identified from the SBES data using decibel differencing, a technique capable of distinguishing zooplankton from fishes by comparing the mean volume backscattering strength ( Sv) detected by echosounders operating at multiple discrete frequencies (Higginbottom et al. 2000, Korneliussen and Ona 2003). Decibel differencing

172

was conducted in Echoview using data from the 38 kHz and 120 kHz transducers. Since the pulse lengths of the 38 kHz and 120 kHz differed (0.256 µs and 0.128 µs, respectively), the

120 kHz data were resampled by time using a mean function so that the resulting resampled

120 kHz data had a pulse length of 0.256 µs, matching the pulse length of the 38 kHz transducer data. The resampled 120 kHz data and the original 38 kHz data were smoothed using a median statistic applied to 3 samples (rows) and 3 pings (columns). The 38 kHz echogram was subtracted from the 120 kHz echogram to visualize the zooplankton. Fishes characterized as individuals and schools were masked from the zooplankton echogram, preserving only the zooplankton. These exclusively zooplankton data were exported by cells from the 38 kHz echogram matching the dimensions of cells for the fishes (5 m intervals x 1 m layers) using a -65 dB threshold. Within R (R Core Team 2016), the mean volume

-1 backscattering strength ( Sv) of zooplankton was converted from the logarithmic (dB re 1 m ) to linear form (m -1), as per fishes above, to generate the volume backscattering coefficient

(sv) of zooplankton per cell.

Fully processed SBES data included a per cell value corresponding to the volume density of both total fishes and fishes separated by size classes (fish m -3), as well as the volume backscattering coefficient of zooplankton (m-1). We refer to the volume backscattering coefficient of zooplankton as the zooplankton concentration

173

Text S3.2: Detailed methods for quantifying water currents.

The hull-mounted ADCP (Teledyne RDI Ocean Surveyor 150 kHz) recorded the magnitude and direction of water currents in 2-m vertical bins above and around the shipwrecks. ADCP data were processed with the University of Hawaii DAS system

(UHDAS) (Firing and Hummon 2010, Firing et al. 2012) to account for beam volume, geometry, transducer orientation, and heading using custom UHDAS Python (Python

Software Foundation 2017) scripts. These scripts produced Matlab (The Mathworks Inc.

2017) files containing the corrected five-minute averaged velocity profiles for each 2-m vertical bin. Data were also edited manually within the UHDAS interface to remove outliers and bad data. Data were further processed in Matlab using a custom UHDAS script that calculated the velocity of the currents from the velocity of water relative to the ship, measured by the ADCP, and the velocity of the survey vessel, determined from the differential GPS and Applanix POS M/V data. The component of current velocity in the east- west direction above and around the wreck ( u) was calculated for each 2-m vertical bin as the sum of the velocity of the water measured by the ADCP ( umeasured ) and the velocity of the survey ship ( uship ):

= + 1 The north-south component of current velocity ( v) was calculated in the same manner. Bad data values, such as those containing noise, bubbles, ping interference, or bottom interference, were also removed using the same custom UHDAS script within Matlab.

For each survey, the processed ADCP data corresponding to the duration of the SBES surveys conducted above and around each wreck were used to calculate the mean current magnitude and direction in Matlab with three steps. First, the east-west velocities and north-

174

south velocities were each averaged across the entire water column and across the spatial extent of the ADCP survey to obtain the mean velocity components ( U and V) for each survey. Second, the current magnitude ( ) was calculated from the east-west ( U) and north- south ( V) velocity components as:

= + 2 Third, the direction of the current ( θ) was calculated in radians as the inverse tangent of the east-west ( U) and north-south mean ( V) velocity components:

= tan 3 Direction values in radians were converted to degrees clockwise from north.

175

Text S3.3: Detailed methods for data analyses.

Fish, zooplankton, and current data were analyzed in R. The per cell values of zooplankton concentration (m -1), small fish density (fish m -3), medium fish density (fish m -3), and large fish density (fish m -3) were each summed across the entire water column vertically.

These resulting-depth-collapsed values for each organism (zooplankton area concentration

(m 2) and fish area density (m 2) per size class) were used for spatial analyses. Here, we call

‘small fishes’ planktivorous fishes, and we call ‘medium fishes’ and ‘large fishes’ piscivorous fishes. We report medium fishes separately from large fishes because the two different size classes help distinguish between the ecological roles of these fishes. Analyses were conducted for each of the twenty surveys, unless otherwise noted. Many of the analyses included calculation of spatial indicators developed specifically for geostatistical data like our fisheries acoustics data (Woillez et al. 2007, 2009). These spatial indicators are applicable for data like ours that inherently contain autocorrelation, even when other types of spatial analyses are inapplicable (Woillez et al. 2007, 2009).

Spatial Location: To describe the general spatial location of zooplankton, small fishes, medium fishes, and large fishes around artificial structures, we calculated six spatial indicators for each organism: 1) positive area, 2) spreading area, 3) equivalent area, 4) microstructure, 5) mean center, and 6) dispersion. We calculated these indicators by applying functions provided by Woillez et al. (2007, 2009). Here, we present the definition of each indicator, as well as the formula used to calculate it, as presented in Woillez et al. (2007,

2009), which are the appropriate citations for the below text on the six spatial indicators.

After calculating each indicator, we tested differences in indicators among organisms using one-way analyses of variance (ANOVAs) followed by post-hoc Tukey HSD tests.

176

Three different indicators, positive area (PA), spreading area (SA), and equivalent area

(EA) were calculated to describe the area occupied by organisms. PA, which is the area occupied by each organism not including locations with zero density, was calculated as the sum of the area of influence ( si) around where fish density ( zi) at sample i exceeds 0 as:

= 1 1 Because PA excludes locations with zero density, this spatial indicator represents the area of organism presence. SA is the area an organism occupies accounting for variations in density.

SA was calculated as:

− = 2 2 where T is the area occupied by fish density values and Q(T) is the cumulative abundance over this area. Q is the total abundance, calculated as:

= #ℎ 3

The difference between PA and SA is that because SA incorporates density variations, it is more sensitive to density than PA, which is based on presence. Equivalent area (EA) is the ratio of total abundance to the mean density per individual, so it is the area that each organism would occupy in a scenario where all individuals have the same density. EA was calculated as:

∑ = 4 ∑ using the same terminology as above. EA is especially useful when an organism has several large density values. The values of EA range from 0 to the positive area. To account for

177

differing total areas among shipwreck surveys, we divided PA, SA, and EA [m 2] by survey- specific total area [m 2]. This provided us with scaled values of PA, SA, and EA. These scaled values represent the percent of the total survey area occupied by PA, SA, and EA.

Microstructure index (MI) was calculated to describe fine-scale variability in organism distribution. MI ranges from 0 to 1, where 0 represents a well-structured and regular density surface, whereas 1 represents a highly irregular and poorly structured density surface

(Woillez et al. 2007, 2009). MI was calculated as the relative decrease in the transitive covariogram (g) constructed between distance 0 and distance h0 that is the mean lag between samples:

0 − ℎ = 0,1 5 0 Two indicators, center of gravity (CG) and inertia (I), were calculated to describe the center of each organism’s distribution and the spread around the center, respectively (Woillez et al. 2007, 2009). CG, which is defined as the weighted mean center of each organism’s distribution, was calculated as:

∑ = 6 ∑ where zi is the organism density at point xi, and si is the area of influence; i represents the sample number. I, which is the dispersion around the mean center, was calculated as:

∑ − = 7 ∑ To determine how the mean center of organisms was positioned relative to the shipwreck, we calculated the distance between the mean center and edge of the shipwreck with the

‘gDistance’ function in the ‘rgeos’ package (Bivand and Rundel 2017).

178

To quantify the patchiness of each organism, we identified spatial patches where organism density exceeded 10% of the organism’s density in the entire survey. The patches were detected using the ‘SI.patches’ function from the ‘RGeostats’ package (Renard et al.

2017). The function also counted the number of spatial patches for each organism, identified the location of the mean center for each patch, and calculated the percent area and percent abundance of each patch (Woillez et al. 2007, 2009). We tested differences in patchiness among organisms using one-way analyses of variance (ANOVAs) followed by post-hoc

Tukey HSD tests.

To further understand organism distribution, we identified clusters of high density (hot- spots) and low density (cold-spots). Hot-spots and cold-spots were detected using the Getis-

Ord Gi* statistic (Getis and Ord 1992, Ord and Getis 1995):

∗ ∑ , = ∑ for sample location ( i) based on neighbors ( j) within a search radius ( d) with corresponding weight matrix ( wi,j (d) ) based on k-nearest neighbors. Getis-Ord Gi* values were calculated for variables of interest ( x), fish density by size class and zooplankton concentration, using the ‘localG’ function within the ‘spdep’ package (Bivand and Piras 2015). The resulting z- values, representing standard deviation, indicated whether hot-spots (positive z-value) or cold-spots (negative z-value) were present. To determine whether the hot- and cold-spots were statistically significant, we used the criteria that a z-value between 1.96 and -1.96 corresponded to p-values > 0.05. Z-values < 1.96 and >1.96 corresponded to p-values < 0.05, indicating a statistically significant hot-spot or cold-spot. This relationship is based on the generalized one-to-one relationship between z- and p-values that is used within ArcGIS implementations of the function as an alternative to applying critical values for different

179

sample sizes (Ord and Getis 1995, ESRI 2016). This alternative implementation was appropriate for our data because each of our surveys had a different sampling extent and associated sample size.

2007, 2009). For two organisms with densities z 1(x) and z 2(x) at point x with respective CGs and inertias ( CG 1 and I1, and CG 2 and I2), the GIC is:

∆ = 1 − 8 ∆ + +

Where ΔCG is the difference between the CG 1 and CG 2. GIC values range from 0 to 1 with 1 representing coinciding CGs with positive inertia values and 0 representing the opposite, where each organism has a different CG and 0 inertia. Two organisms with identical CGs but different inertias would have a GIC of 1, so the GIC does not provide information on how organisms are arranged at a finer scale (e.g, concentric distributions, etc.) (Bez and Rivoirard

2000).

To complement global metrics, we calculated two metrics to describe spatial overlap at the local scale: 1) the local index of collocation (LIC) and 2) the co-occurrence. While the

GIC compared organism distribution over the entire extent of each survey, the local metrics assess overlap at the level of each sample (e.g., 5 m x 1 m depth collapsed sample). The first local metric, LIC, is a non-centered correlation value between the densities of two organisms.

LIC is calculated as:

, = 9

180

where the density of organism 1 is z1(x) and the density of organism 2 is z2(x) at point x. LIC ranges from 0 to 1. When no individuals from the two species co-occur at any sample location, LIC is 0. When the density of each organism at all sampling locations is the same,

LIC is 1. Because the LIC is based on a sum, it is robust against zeros and overcomes the associated problems of correlation coefficients. The second metric, co-occurrence (CO), is the proportion of samples ( l) where both species 1 and species 2 co-occur (Saraux et al.

2014):

, = 0,1 10 + CO includes points where either one or both species are present and excludes samples where both species are absent. CO differs from LIC because LIC is calculated based on densities, whereas co-occurrence is calculated based on presence absence. Together, the GIC, LIC, and

CO values provide complementary means to interpret relationships between two organisms that were appropriate for fisheries acoustics data like ours. We bootstrapped the GIC, LIC, and CO results for each pair of organisms to ensure that the mean overlap metrics across all twenty surveys approximated the mean distribution from 1,000 bootstrapped pulls conducted within the ‘bootstrap’ package (Tibshirani and Leisch 2017).

Water Current: To determine how water current magnitude affected the location of planktivorous fishes relative to each shipwreck, we first calculated the distance from the mean center of small fishes to the shipwreck edges. We also calculated distances from the mean center of medium and large fishes from shipwreck edges. The distances were calculated with the ‘gDistance’ function in the ‘rgeos’ package (Bivand and Rundel 2017).

We did not calculate the distance between the center of small fishes and zooplankton because zooplankton were present throughout nearly 25% of surveyed areas, meaning that their mean

181

center was biased by survey extent. We used linear regressions to investigate relationships between fish location and current magnitude. First, we regressed distances between the mean center of each fish size class and the nearest shipwreck edge against current magnitude.

Second, because small fishes were unrelated to water current and given prior research suggesting that predator-prey dynamics influence planktivorous fish aggregations (Holzman et al. 2005), we fit additional linear regressions investigating relationships between the distance of small fishes from wreck edges and predictor variables representing apparent predation risk. The predictor variables for apparent predation risk were the number of medium, large, and both medium and large fishes present within the shipwreck, designated as the manually delineated polygon corresponding to the shipwreck structure, and 5-m outward of the delineated shipwreck structure. We conducted assessments of fit by comparing observed distances of small fishes from edges to those of the estimated distribution.

To determine how water current direction influenced small fish density, we visualized the location of the mean center of small fishes relative to upstream and downstream portions of the shipwreck. We designated the upstream and downstream areas corresponding to each survey where current data were collected. To make the upstream versus downstream distinction, we drew the current vector associated with each survey so that it bisected the survey extent through the centroid of the wreck. We then drew a line perpendicular to the current vector. Surveyed areas on either side of the perpendicular line were then designated as ‘upstream’ or ‘downstream.’ For example, areas that the current hit first were ‘upstream,’ and those areas that the current would move to only after passing the wreck were

‘downstream.’ We also used the upstream versus downstream designations to determine whether each sampling cell that corresponded to a value of fish density was upstream or

182

downstream. To do this, we calculated the bearing between the centroid of the shipwreck and the sampling cells. We compared this bearing to the bearings required to be in the upstream or downstream sections of the sampling extent (on either side of the line perpendicular to the current vector). Fishes that occurred within the upstream extent were categorized as upstream, and those in the downstream extent as downstream. We calculated the total number of small fishes located upstream and downstream on each survey. We used these data to fit a beta regression using the ‘betareg’ package (Cribari-Neto and Zeileis 2010) to examine whether the proportion of fishes located upstream or downstream was affected by current magnitude.

183

Figure S3.1. Bathymetric maps of four surveyed shipwrecks: a) U-352 , b) USS Schurz, c)

USS Tarpon, and d) W.E. Hutton . Warm colors are shallower depths; deeper colors are cooler depths. Black dotted lines are survey paths. North is up. Each scale bar is a total of

100 m; the scale division occurs at 50 m. The date surveyed (yyyy-mm-dd) is displayed.

184

Figure S3.2. Spatial location of small fishes (blue; planktivorous fishes), medium fishes

(orange; piscivorous fishes), and large fishes (red; piscivorous fishes) relative to four shipwrecks (black polygon): a) U-352 , b) HMT Bedfordshire , c) W.E. Hutton , and d) USS

Schurz . Circles represent the weighted mean center of each organism’s distribution (center of gravity). Solid lines represent the dispersion around the center (inertia). Gray dotted lines indicate the sampling extent of each survey. Current vector (black arrow) displays current direction (orientation of vector) and current magnitude (length of vector). Black dashed line divides survey into upstream or downstream components. North is up. Each scale bar is a total of 100 m; the scale division occurs at 50 m. The date surveyed (yyyy-mm-dd) is displayed.

185

Figure S3.3. Spatial clusters of zooplankton, small fishes, medium fishes, and large fishes relative to each of four shipwrecks (black): a) USS Schurz, b) USS Tarpon, c) Proteus , and d) Merak . Colors correspond to the Getis-Ord Gi* z-value, where darker red represents more pronounced areas of high density (hot-spots). Gray points are locations along the survey where data were obtained but the the Getis-Ord Gi* z-value was not significant. North is up.

Each scale bar is a total of 100 m; the scale division occurs at 50 m. The date surveyed

(yyyy-mm-dd) is displayed.

186

Figure S3.4. Bootstrapped metrics describing spatial relationships between pairs of organisms. Black bars represent the distribution from the 20 surveys following bootstrapping with 1,000 samples. Red vertical lines represent the mean across all twenty surveys. Columns correspond to organism pairs: zooplankton and small fishes, small fishes and medium fishes, small fishes and large fishes, and medium fishes and large fishes. Rows correspond to overlap metrics: global index of collocation (GIC), local index of collocation (LIC), and co- occurrence (CO).

187 Tables

Table S3.1. Descriptions of fifteen shipwrecks surveyed. Shipwreck metrics include mean depth (mean dep), minimum depth (min

dep), maximum depth (max dep), vertical relief (relief), area on wreckage on the seafloor (area), and perimeter of wreckage on the

seafloor (perimeter). Width and length of the originally built ship and the sunken wreckage on the seafloor are provided. Bathymetry

and transect lines for survey for several wrecks are displayed in Figure S1. Historical data for each shipwreck, such as date of sinking

and original ship specifications, were provided by the NOAA Monitor National Marine Sanctuary.

Built Built Mean Min Max Sunk Sunk Ship Ship length width Date Sinking dep dep dep Relief Area Perim width length 188 name type Material [m] [m] sunk description [m] [m] [m] [m] [m 2] [m] [m] [m] sunk by Ashkhabad Tanker steel 122.2 15.9 1942 U-402 -17.7 -18.9 -13.4 5.4 2657.6 293.5 44.8 104.8 Empire sunk by Thrush Steam Merchant steel 120.5 16.8 1942 U-203 -20.6 -22.3 -19.3 3.1 32.6 27.0 5.9 24.0

F.W. Abrams Tanker steel 142.5 19.1 1942 sunk by mine -25.0 -26.0 -21.1 4.9 3261.2 407.4 39.5 158.6 HMS Senateur Trawler (converted collision with Duhamel to anti-submarine) steel 58.6 9.5 1942 USS Semmes -18.7 -19.8 -15.8 4.0 858.5 164.8 23.3 62.0 HMT Trawler (converted sunk by Bedfordshire to convoy escort) steel 49.5 8.1 1942 U-140 -29.1 -29.6 -27.4 2.2 385.7 116.9 27.1 34.0 sunk by Malchace Cargo Ship steel 101.7 14.6 1942 U-160 -59.6 -61.1 -56.1 5.0 2005.6 288.2 29.1 107.9 sunk by Merak Steam Merchant steel 99.4 14.3 1918 U-140 -38.0 -43.5 -33.0 10.5 1485.4 257.3 47.5 87.1 Passenger / collision with Proteus Freighter steel 118.9 14.6 1918 SS Cushing -36.0 -38.0 -31.4 6.6 2335.2 320.0 31.2 121.0 sunk by USS U-352 U-boat (VII C) steel 67.1 6.1 1942 Icarus -32.0 -34.1 -29.8 4.3 372.2 139.2 9.1 64.7 sunk by US Army Hudson U-701 U-boat (VII C) steel 67.1 6.1 1942 aircraft -33.1 -35.4 -30.8 4.6 283.2 141.7 7.4 68.2

USS Monitor Ironclad iron 52.7 12.6 1862 sunk in storm -66.9 -68.8 -64.6 4.2 558.0 115.7 16.2 45.7

composite hull wood collision with USS Schurz Gunboat / steel 68.6 9.8 1918 SS Florida -32.4 -33.3 -29.6 3.7 1111.1 226.3 22.1 81.9 intentionally USS Tarpon Submarine steel 90.8 7.6 1957 sunk -39.8 -42.2 -37.3 4.8 510.8 180.5 10.2 84.7 Trawler (converted sunk by USS YP-389 to patrol) steel 31.2 6.7 1942 U-701 -96.0 -98.6 -94.2 4.4 277.1 79.9 9.3 36.3 sunk by W.E. Hutton Tanker steel 132.6 17.1 1942 U-124 -33.2 -37.1 -26.6 10.5 2293.8 353.5 25.3 141.2 189

Table S3.2. Descriptions of twenty surveys across fifteen shipwrecks. Survey number corresponds to the order from top to bottom of surveys in Figure 2. MBES resolution (res) indicates the resolution of multibeam bathymetry data used to delinate shipwrecks and establish transect lines. SBES and ADCP indicate whether these two instruments were active

(X) or inactive (--).

Surve Date Shipwreck MBES res (m) SBES ADCP y 1 2016-11-01 HMT Bedfordshire 0.25 X -- 2 2016-11-01 Proteus 0.25 X -- 3 2016-11-01 USS Tarpon 0.25 X -- 4 2016-11-01 USS YP-389 0.35 X -- 5 2016-11-02 Empire Thrush 0.15 X -- 6 2016-11-02 U-701 0.25 X -- 7 2016-11-03 Malchace 0.35 X -- 8 2016-11-03 Merak 0.25 X -- HMS Senateur 9 2016-11-05 0.20 X -- Duhamel 10 2016-11-06 F.W. Abrams 0.25 X -- 11 2016-11-06 USS Monitor 0.40 X -- 12 2017-07-05 Ashkhabad 0.25 X X 13 2017-07-06 W.E. Hutton 0.35 X X 14 2017-07-06 USS Schurz 0.25 X X 15 2017-07-06 U-352 0.35 X X 16 2017-07-10 W.E. Hutton 0.35 X X 17 2017-07-11 W.E. Hutton 0.35 X X 18 2017-07-12 USS Schurz 0.25 X X 19 2017-07-13 HMT Bedfordshire 0.25 X X 20 2017-07-14 U-352 0.35 X X

190

APPENDIX 4: SUPPORTING INFORMATION FOR CHAPTER 4

Figure S4.1. Hourly time series of fish abundance on natural rocky reef on four separate days: A) September 17, 2014; B) September 18, 2014; C) September 19, 2014; D) September

20, 2014. Each point represents fish abundance in a single video clip. Although seismic surveying was active on September 20, seismic activity was not audible on all collected videos. The color and shape of each point corresponds to whether seismic activity was audible on the video (red triangles) or not audible (black circles). Black lines are smoothed conditional means.

80 p=0.954 p=0.602 p=0.767 p=0.047

UDL

60 ● ● ●

40 Variance

● LDL

n=29 n=35 n=39 n=37 0 Pre 1 Pre 2 Pre 3 During Day

Figure S4.2. Test of equality of variance in fish counts on three days pre-seismic surveying and one day during seismic surveying, based on analysis of means for variance (ANOMV) with Levene transformation. Daily means for variance (black points) are contrasted with grand mean for variance (solid black horizontal line). P-values indicate whether daily means of variance are significantly different from grand mean, as do horizontal dashed lines that represent 95% confidence limits. On the day with seismic surveying, variance in fish counts was significantly lower than on each of three days before, driven by reduced abundance. N is number of videos.

192

Table S4.1. Fish species list from 140 videos recorded on the natural rocky reef three days before and one day during seismic surveying. Several fish that were unidentifiable from video recordings, coded below as family ‘UKNOWN,’ were excluded from the count of 32 species representing 17 families. Fish with * after their common name are in the snapper- grouper management group.

Family Genus species Common_Name CARANGIDAE Caranx ruber Bar Jack* Decapterus spp. Scad Species Seriola dumerili Greater Amberjack* CARCHARHINIDAE Carcharhinus spp . Unknown Shark CHAETODONTIDAE Chaetodon ocellatus Spotfin Butterflyfish HAEMULIDAE Haemulon aurolineatum Tomtate* Haemulon plumieri White Grunt* LABRIDAE Halichoeres garnoti Yellowhead Wrasse Halichoeres spp. Unknown Wrasse Halichoeres bivittatus Slippery Dick LUTJANIDAE Lutjanus griseus Gray snapper* Rhomboplites aurorubens Vermillion Snapper* MONACANTHIDAE Stephanolepis hispidus Planehead Filefish UNKNOWN POMACANTHIDAE Holacanthus bermudensis Blue Angelfish POMACENTRIDAE Chromis scotti Purple Reef Fish Stegastes variabilis Cocoa Damselfish SCARIDAE Sparisoma atomarium SCIAENIDAE Pareques umbrosus Cubbyu SCORPAENIDAE Pterois volitans Lionfish SERRANIDAE Centropristis ocyurus Bank Sea Bass* Centropristis striata Black Sea Bass* Diplectrum formosum Sand Perch Mycteroperca microlepis Gag* Mycteroperca phenax Scamp* Rypticus maculatus SPARIDAE Archosargus probatocephalus Sheepshead Calamus spp. Unknown Porgy Diplodus holbrookii Spottail Pinfish Stenotomus caprinus Longspine Porgy* SYNODONTIDAE Synodus spp. Lizardfish TETRAODONTIDAE Sphoeroides spengleri Bandtail Puffer TRIGLIDAE Prionotus spp. Unknown Searobin

Video S4.1 Video recording from reef located 7.9 km from closest approach of the seismic surveying vessel during the evening one day prior to seismic surveying on the inner continental shelf.

Video S4.2 Video recording from reef located 7.9 km from closest approach of the seismic surveying vessel during active seismic surveying on the inner continental shelf. Noise from a discrete airgun shot is audible, and the surveying vessel was 8.0 km away from the reef at the time of this video.

Audio S4.1 Audio recording from reef located 0.7 km from the closest approach of the seismic surveying vessel prior to the seismic surveying on the inner continental shelf. Only ambient noise is audible, and the surveying vessel was 22.2 km away from the reef at the time of this audio recording.

Audio S4.2 Audio recording from reef located 0.7 km from the closest approach of the seismic surveying vessel during active seismic surveying on the inner continental shelf.

Noise from the discreet airgun shot is audible, and the surveying vessel was 0.7 km away from the reef at the time of this audio recording. This is just prior to shots that overloaded the instruments.

APPENDIX 5: SUPPORTING INFORMATION FOR CHAPTER 5

Figure S5.1. Nonmetric multidimensional scaling ordination of fish community on the established reef (USS Indra ) prior to the deployment of the new artificial reef nearby. Each sample represents fish community composition based on square-root transformed abundance data collected from diver-conducted belt-transect surveys. Labels indicate year and month

(year – month) of each diver-conducted fish survey. Inter-annual variation in fish community composition occurs. There is no repeatable pattern to seasonal variation, suggesting that seasonality on the established reef does not contribute to shifts in community composition.

195 Table S5.1. Species list for fishes observed on the new and established reef. Number of individuals observed on the new reef and the established reef are provided in respective columns.

Family Scientific name Common Name Habitat zone New reef Established reef Alopiidae Alopias vulpinus Thresher shark Pelagic 0 1

Apogonidae Apoginidae sp. Unknown cardinalfish Demersal 2 0 Carangidae Decapterus sp. Scad species Pelagic 14843 6540 Carangidae Carangidae sp. Unknown jack Pelagic 86 25

Carangidae Carangoides bartholomaei Yellow jack Pelagic 59 0 Carangidae Seriola dumerili Greater amberjack Pelagic 25 19 Carangidae Caranx crysos Blue runner Pelagic 21 8

Carangidae Caranx ruber Bar jack Pelagic 7 20 Echeneidae Remora remora Remora Demersal 0 1 Ephippidae Chaetodipterus faber Atlantic spadefish Pelagic 0 25

Haemulidae Haemulon aurolineatum Tomtate Pelagic 8904 8121 Haemulidae Haemulidae sp. Unknown grunt Demersal 1 6 Labridae Halichoeres bivittatus Slippery dick Demersal 6 26

Labridae Labridae sp. Unknown wrasse Demersal 6 16 Labridae Tautoga onitis Tautog Demersal 1 1 Lutjanidae Rhomboplites aurorubens Vermillion snapper Pelagic 9574 0

Lutjanidae Lutjanidae sp. Unknown snapper Demersal 5 124 Monacanthidae Stephanolepis hispidus Planehead filefish Demersal 0 1 Osteichthyes Osteichthyes sp. Unknown schooling fish Pelagic 13505 11738

Osteichthyes Osteichthyes sp. Unknown fish species Demersal 5786 7894 Pomacentridae Pomacentridae sp . Unknown damselfish Demersal 0 8 Beaugregory or cocoa Pomacentridae Stegastes sp. damselfish Demersal 0 3 Rachycentridae Rachycentron canadum Cobia Pelagic 1 0 Serranidae Centropristis ocyurus Bank sea bass Demersal 12 2 Serranidae Mycteroperca microlepis Gag Demersal 12 123

Serranidae Centropristis striata Black sea bass Demersal 11 23 Serranidae Serranidae sp. Unknown seabass Demersal 11 153 Serranidae Rypticus maculatus Whitespotted soapfish Demersal 0 1

Serranidae Serranus subligarius Belted sandfish Demersal 0 135 Sparidae Diplodus holbrookii Spottail pinfish Demersal 182 1942 Archosargus Sparidae probatocephalus Sheepshead Demersal 4 33 Sparidae Sparidae sp. Unknown porgy Demersal 1 20 Sphyraenidae Sphyraena barracuda Barracuda Pelagic 33 31

196

Video S5.1. Underwater video recording from the new artificial reef ( James J. Francesconi ) during the first sampling period in May 2016.

Video S5.2. Underwater video recording from the established artificial reef (USS Indra ) during the first sampling period in May 2016.

Video S5.1. Underwater video recording from the new artificial reef ( James J. Francesconi ) during the third sampling period in September 2016.

Video S5.2. Underwater video recording from the established reef (USS Indra ) during the third sampling period in September 2016.

REFERENCES

Able, K. W., T. M. Grothues, and I. M. Kemp. 2013. Fine-scale distribution of pelagic fishes relative to a large urban pier. Marine Ecology Progress Series 476:185–198.

Airoldi, L., X. Turon, S. Perkol-Finkel, and M. Rius. 2015. Corridors for aliens but not for natives: effects of marine urban sprawl at a regional scale. Diversity and Distributions 21:755–768.

Alevizon, W. S., and J. C. Gorham. 1989. Effects of artificial reef deployment on nearby resident fishes. Bulletin of Marine Sceince 44:646–661.

Allouche, O., M. Kalyuzhny, G. Moreno-Rueda, M. Pizarro, and R. Kadmon. 2012. Area- heterogeneity tradeoff and the diversity of ecological communities. Proceedings of the National Academy of Sciences 109:17495–17500.

Almany, G. R. 2004. Differential effects of habitat complexity, predators and competitors on abundance of juvenile and adult coral reef fishes. Oecologia 141:105–13.

Anderson, M. J. 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecology 26:32–46.

Anderson, M. J. 2006. Distance-based tests for homogeneity of multivariate dispersions. Biometrics 62:245–253.

Arena, P. T., L. K. B. Jordan, and R. E. Spieler. 2007. Fish assemblages on sunken vessels and natural reefs in southeast Florida, USA. Hydrobiologia 580:157–171.

Arias-González, J. E., T. J. Done, C. A. Page, A. J. Cheal, S. Kininmonth, and J. R. Garza- Pérez. 2006. Towards a reefscape ecology: relating biomass and trophic structure of fish assemblages to habitat at Davies Reef, Australia. Marine Ecology Progress Series 320:29–41.

Arkema, K. K., G. M. Verutes, S. A. Wood, C. Clarke-Samuels, S. Rosado, M. Canto, A. Rosenthal, M. Ruckelshaus, G. Guannel, J. Toft, J. Faries, J. M. Silver, R. Griffin, and A. D. Guerry. 2015. Embedding ecosystem services in coastal planning leads to better outcomes for people and nature. Proceedings of the National Academy of Sciences 112:7390–7395.

Baine, M. 2001. Artificial reefs: a review of their design, application, management and performance. Ocean & Coastal Management 44:241–259.

Barange, M. 1994. Acoustic identification, classification and structure of biological patchiness on the edge of the Agulhas Bank and its relation to frontal features. South African Journal of Marine Science 14:333–347.

Barceló, C., L. Ciannelli, E. M. Olsen, T. Johannessen, and H. Knutsen. 2016. Eight decades

198

of sampling reveal a contemporary novel fish assemblage in coastal nursery habitats. Global Change Biology 22:1155–1167.

Baynes, T. W., and A. M. Szmant. 1989. Effect of current on the sessile benthic community structure of an artificial reef. Bulletin of Marine Science 44:545–566.

Becker, A., M. D. Taylor, and M. B. Lowry. 2017. Monitoring of reef associated and pelagic fish communities on Australia’s first purpose built offshore artificial reef. ICES Journal of Marine Science 74:277–285.

Beijbom, O., P. J. Edmunds, D. I. Kline, B. G. Mitchell, and D. Kriegman. 2012. Automated annotation of coral reef survey images. 2012 IEEE Conference on Computer Vision and Pattern Recognition:1170–1177.

Beukers, J. S., and G. P. Jones. 1997. Habitat complexity modifies the impact of piscivores on a coral reef fish population. Oecologia 114:50–59.

Bez, N., and J. Rivoirard. 2000. Future applications of CUFES: indices of collocation between populations. Pages 48–52 in D. M. J. Checkley, J. R. Hunter, L. Motos, and C. D. van der Lingen, editors.Report of a workshop on the use of the Continuous Underway Fish Egg Sampler (CUFES) for mapping spawning habitats of pelagic fish. GLOBEC Report No. 14.

Bez, N., and J. Rivoirard. 2001. Transitive geostatistics to characterise spatial aggregations with diffuse limits: An application on mackerel ichtyoplankton. Fisheries Research 50:41–58.

Bishop, M. J., M. Mayer-Pinto, L. Airoldi, L. B. Firth, R. L. Morris, L. H. L. Loke, S. J. Hawkins, L. A. Naylor, R. A. Coleman, S. Y. Chee, and K. A. Dafforn. 2017. Effects of ocean sprawl on ecological connectivity: impacts and solutions. Journal of Experimental Marine Biology and Ecology 492:7–30.

Bivand, R., and G. Piras. 2015. Comparing implementations of estimation methods for spatial econometrics. Journal of Statistical Software 63:1–36.

Bivand, R., and C. Rundel. 2017. rgeos: Interface to geometry engine - open source (’GEOS’).

Blackwell, S. B., C. S. Nations, T. L. McDonald, A. M. Thode, D. Mathias, K. H. Kim, C. R. Greene, and A. M. Macrander. 2015. Effects of airgun sounds on bowhead whale calling rates: evidence for two behavioral thresholds. Plos ONE 10:e0125720.

Bohnsack, J. A. 1979. Photographic quantitative sampling of hard-bottom benthic communities. Bulletin of Marine Science 29:242–252.

Bohnsack, J. A. 1997. Consensus development and the use of marine reserves in the Florida Keys, U.S.A. Proceedings of the 8th International Coral Reef Symposium 2:1927–1930.

Bray, R. N. 1980. Daily foraging migrations of the blacksmith ( Chromis punctipinnis ), a planktivorous kelp-bed damselfish. Bulletin of Marine Sceince 30:325–326.

Brock, R. E. 1982. A critique of the visual census method for assessing coral reef fish populations. Bulletin of Marine Science 32:269–276.

Brock, V. E. 1954. A preliminary report on a method of estimating reef fish populations. Journal of Wildlife Management 18:297–308.

Bulleri, F., and L. Airoldi. 2005. Artificial marine structures facilitate the spread of a non- indigenous green alga, Codium fragile ssp. tomentosoides , in the north Adriatic Sea. Journal of Applied Ecology 42:1063–1072.

Bulleri, F., and M. G. Chapman. 2010. The introduction of coastal infrastructure as a driver of change in marine environments. Journal of Applied Ecology 47:26–35.

Burchmore, J. J., D. A. Pollard, J. D. Bell, M. J. Middleton, B. C. Pease, and J. Matthews. 1985. An ecological comparison of artificial and natural rocky reef fish communities in Botany Bay, New South Wales, Australia. Bulletin of Marine Science 37:70–85.

Burt, J., A. Bartholomew, P. Usseglio, A. Bauman, and P. F. Sale. 2009. Are artificial reefs surrogates of natural habitats for corals and fish in Dubai, United Arab Emirates? Coral Reefs 28:663–675.

Byrnes, J. E., D. C. Reed, B. J. Cardinale, K. C. Cavanaugh, S. J. Holbrook, and R. J. Schmitt. 2011. Climate-driven increases in storm frequency simplify kelp forest food webs. Global Change Biology 17:2513–2524.

De Cáceres, M., and P. Legendre. 2009. Associations between species and groups of sites: indices and statistical inference. Ecology 90:3566–3574.

Carnicer, J., L. Brotons, S. Herrando, and D. Sol. 2013. Improved empirical tests of area- heterogeneity tradeoffs. Proceedings of the National Academy of Sciences of the United States of America 110:E2858–E2860.

Carr, M. H., and M. A. Hixon. 1997. Artificial reefs: The importance of comparisons with natural reefs. Fisheries 22:28–33.

Carter, A., and S. Prekel. 2008. Benthic colonization and ecological successional patterns on a planned nearshore artificial reef system in Broward County, SE Florida. Proceedings of the 11th International Coral Reef Symposium:1209–1213.

Carvalho, S., A. Moura, J. Cúrdia, L. Cancela da Fonseca, and M. N. Santos. 2013. How complementary are epibenthic assemblages in artificial and nearby natural rocky reefs? Marine environmental research 92:170–177.

Cerame-Vivas, M. J., and I. E. Gray. 1966. The distributional pattern of benthic invertebrates of the continental shelf off North Carolina. Ecology 47:260–270.

Champion, C., I. M. Suthers, and J. A. Smith. 2015. Zooplanktivory is a key process for fish production on a coastal artificial reef. Marine Ecology Progress Series 541:1–14.

Churchill, J. H., and T. J. Berger. 1998. Transport of Middle Atlantic Bight shelf water to the Gulf Stream near Cape Hatteras. Journal of Geophysical Research 103:30605–30621.

Churchill, J. H., and G. Gawarkiewicz. 2009. Shelfbreak frontal eddies over the continental slope north of Cape Hatteras. Journal of Geophysical Research 114:1–15.

Clark, S., and A. J. Edwards. 1999. An evaluation of artificial reef structures as tools for marine habitat rehabilitation in the Maldives. Aquatic Conservation: Marine and Freshwater Ecosystems 9:5–21.

Clarke, K. R. 1993. Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18:117–143.

Connell, S. D., and G. P. Jones. 1991. The influence of habitat complexity on postrecruitment processes in a temperate reef fish population. Journal of Experimental Marine Biology and Ecology 151:271–294.

Cresson, P., S. Ruitton, and M. Harmelin-Vivien. 2014. Artificial reefs do increase secondary biomass production: mechanisms evidenced by stable isotopes. Marine Ecology Progress Series 509:15–26.

Cribari-Neto, F., and A. Zeileis. 2010. Beta regression in R. Journal of Statistical Software2 34:1–24.

Cronk, Q. C. B. 1997. Islands: stability, diversity, conservation. Biodiversity & Conservation 6:477–493.

Crowder, L. B., and W. E. Cooper. 1982. Habitat structural complexity and the interaction between bluegills and their prey. Ecology 63:1802–1813.

Cruise Report: Eastern North American Margin Community Seismic Experiment, Cruise MGL1408, R/V Marcus G Langseth. 2014. .

Cummings, S. L. 1994. Colonization of a nearshore artificial reef at Boca Raton (Palm Beach County), Florida. Bulletin of Marine Science 55:1193–1215.

Dafforn, K. A., T. M. Glasby, L. Airoldi, N. K. Rivero, M. Mayer-Pinto, and E. L. Johnston. 2015. Marine urbanization: An ecological framework for designing multifunctional artificial structures. Frontiers in Ecology and the Environment 13:82–90.

Dafforn, K. A., T. M. Glasby, and E. L. Johnston. 2012. Comparing the invasibility of experimental “reefs” with field observations of natural reefs and artificial structures. PLoS ONE 7:e38124.

Dance, M. A., W. F. Patterson III, and D. T. Addis. 2011. Fish community and trophic

structure at artificial reef sites in the northeastern Gulf of Mexico. Bulletin of Marine Science 87:301–324.

Deaton, A. S., W. S. Chappell, K. Hart, J. O’Neal, and B. Boutin. 2010. North Carolina Coastal Habitat Protection Plan. North Carolia Department of Environment and Natural Resources, Division of Marine Fisheries, Morehead City, NC.

Diehl, S. 1992. Fish predation and benthic community structure: the role of omnivory and habitat complexity. Ecology 73:1646–1661.

Duffy, J. E., P. L. Reynolds, C. Boström, J. A. Coyer, M. Cusson, S. Donadi, J. G. Douglass, J. S. Eklöf, A. H. Engelen, B. K. Eriksson, S. Fredriksen, L. Gamfeldt, C. Gustafsson, G. Hoarau, M. Hori, K. Hovel, K. Iken, J. S. Lefcheck, P. O. Moksnes, M. Nakaoka, M. I. O’Connor, J. L. Olsen, J. P. Richardson, J. L. Ruesink, E. E. Sotka, J. Thormar, M. A. Whalen, and J. J. Stachowicz. 2015. Biodiversity mediates top-down control in eelgrass ecosystems: A global comparative-experimental approach. Ecology Letters 18:696–705.

Dugan, J. E., L. Airoldi, M. G. Chapman, S. J. Walker, and T. Schlacher. 2012. Estuarine and coastal structures: Environmental effects, a focus on shore and Nnearshore structures. Page Treatise on Estuarine and Coastal Science. Waltham: Academic Press.

Dunn, D. C., and P. N. Halpin. 2009. Rugosity-based regional modeling of hard-bottom habitat. Marine Ecology Progress Series 377:1–11.

Dustan, P., O. Doherty, and S. Pardede. 2013. Digital reef rugosity estimates coral reef habitat complexity. PLoS ONE 8:e57386.

Echoview. 2017. Fish tracking algorithm: Echoview’s Alpha-Beta tracking algorithm. http://support.echoview.com/WebHelp/Reference/Algorithms/Fish_tracking_module/Fis h_tracking_algorithms.htm.

ESRI. 2016. ArcGIS. Redlands, California, USA. Environmental Systems Research Institute. Environmental Systems Research Institute, Redlands, California, USA.

Ferrario, F., L. Iveša, A. Jaklin, S. Perkol-Finkel, and L. Airoldi. 2016. The overlooked role of biotic factors in controlling the ecological performance of artificial marine habitats. Journal of Applied Ecology 53:16–24.

Firing, E., and J. M. Hummon. 2010. Shipboard ADCP measurements. The GO-SHIP repeat hydrography manual: a collection of expert reports and guidelines. IOCCP Report No. 14, ICPO Publication Series No. 134.

Firing, E., J. M. Hummon, and T. K. Chereskin. 2012. Improving the quality and accessibility of current profile measurements in the Southern Ocean. Oceanography 25:164–165.

Fitzhardinge, R. C., and J. H. Bailey-Brock. 1989. Colonization of artificial reef materials by corals and other sessile organisms. Bulletin of Marine Sceince 44:567–579.

Floeter, S. R., L. A. Rocha, D. R. Robertson, J. C. Joyeux, W. F. Smith-Vaniz, P. Wirtz, A. J. Edwards, J. P. Barreiros, C. E. L. Ferreira, J. L. Gasparini, A. Brito, J. M. Falcón, B. W. Bowen, and G. Bernardi. 2008. Atlantic reef fish biogeography and evolution. Journal of Biogeography 35:22–47.

Folpp, H., M. Lowry, M. Gregson, and I. M. Suthers. 2013. Fish assemblages on estuarine artificial reefs: natural rocky-reef mimics or discrete assemblages? PLOS ONE 8:e63505.

Foote, K. G., H. P. Knudsen, G. Vestnes, D. N. MacLennan, and E. J. Simmonds. 1987. Calibration of acoustic instruments for fish density estimation: a practical guide. ICES Cooperative Report 144:1–57.

Fortin, M.-J., and M. Dale. 2005. Spatial analysis: a guide for ecologists. Cambridge University Press, Cambridge, UK.

Fowler, A. M., and D. J. Booth. 2012. How well do sunken vessels approximate fish assemblages on coral reefs? Conservation implications of vessel-reef deployments. Marine Biology 159:2787–2796.

Froese, R., and D. Pauly. 2016. FishBase. www.fishbase.org.

Gaines, S. D., and J. Roughgarden. 1987. Fish in offshore kelp forests affect recruitment to intertidal barnacle populations. Science (New York, N.Y.) 235:479–481.

Gazol, A., R. Tamme, J. N. Price, I. Hiiesalu, L. Laanisto, and M. Pärtel. 2013. A negative heterogeneity-diversity relationship found in experimental grassland communities. Oecologia 173:545–555.

Genin, A. 2004. Bio-physical coupling in the formation of zooplankton and fish aggregations over abrupt topographies. Journal of Marine Systems 50:3–20.

Genin, A., L. Haury, and P. Greenblatt. 1988. Interactions of migrating zooplankton with shallow topography: predation by rockfishes and intensification of patchiness. Deep-Sea Research 35:151–175.

Genin, A., J. S. Jaffe, R. Reef, C. Richter, and P. J. S. Franks. 2005. Swimming against the flow: a mechanism of zooplankton aggregation. Science (New York, N.Y.) 308:860– 862.

Getis, A., and J. K. Ord. 1992. The analysis of spatial association by use of distance statistics. Geographical Analysis 24:189–206.

Gido, K. B., J. F. Schaefer, and J. A. Falke. 2009. Convergence of fish communities from the littoral zone of reservoirs. Freshwater Biology 54:1163–1177.

Gilinsky, E. 1984. The role of fish predation and spatial heterogeneity in determining benthic community structure. Ecology 65:455–468.

Gittman, R. K., F. J. Fodrie, A. M. Popowich, D. A. Keller, J. F. Bruno, C. A. Currin, C. H. Peterson, and M. F. Piehler. 2015. Engineering away our natural defenses: an analysis of shoreline hardening in the US. Frontiers in Ecology and the Environment 13:301–307.

Gittman, R. K., A. M. Popowich, J. F. Bruno, and C. H. Peterson. 2014. Marshes with and without sills protect estuarine shorelines from erosion better than bulkheads during a Category 1 hurricane. Ocean and Coastal Management 102:94–102.

Glasby, T. M., and S. D. Connell. 2017. Urban structures as marine habitats. Ambio 28:595– 598.

Gleason, D. F., A. W. Harvey, and S. P. Vives. 2015. A Guide to the Benthic Invertebrates and Cryptic Fishes of Gray’s Reef. http://bio.vtn2.com/gr-inverts/.

Golani, D., and A. Diamant. 1999. Fish colonization of an artificial reef in the Gulf of Elat, northern Red Sea. Environmental Biology of Fishes 54:275–282.

Gorman, O. T., and J. R. Karr. 1978. Habitat structure and stream fish communities. Ecology 59:507–515.

Goslee, S. C., and D. L. Urban. 2007. The ecodist package for dissimilarity-based analysis of ecological data. Journal of Statistical Software 22:1–19.

Gotceitas, V., and P. Colgan. 1989. Predator foraging success and habitat complexity: quantitative test of the threshold hypothesis. Oecologia 80:158–166.

Grabowski, J. H. 2004. Habitat complexity distrupts predator-prey interactions but not the trohpic cascade on oyster reefs. Ecology 85:995–1004.

Grace, J. B. 2008. Structural equation modeling for observational studies. The Journal of Wildlife Management 72:14–22.

Granneman, J. E., and M. A. Steele. 2015. Between Artificial and Natural Reefs 72:2385– 2397.

Gregg, K., and S. Murphey. 1994. The role of vessels as artificial reef material on the Atlantic and Gulf of Mexico coasts of the United States. Special Report No. 38 of the Atlantic States Marine Fisheries Commission. Morehead City, NC.

Guerra, M., A. M. Thode, S. B. Blackwell, and A. M. Macrander. 2011. Quantifying seismic survey reverberation off the Alaskan North Slope. The Journal of the Acoustical Society of America 130:3046–3058.

Guiry, M. D., and G. M. Guiry. 2007. AlgaeBase. http://www.algaebase.org.

Habary, A., J. L. Johansen, T. J. Nay, J. F. Steffensen, and J. L. Rummer. 2017. Adapt, move or die - how will tropical coral reef fishes cope with ocean warming? Global Change Biology 23:566–577.

Halvorsen, M. B., B. M. Casper, F. Matthews, T. J. Carlson, and A. N. Popper. 2012. Effects of exposure to pile-driving sounds on the lake sturgeon, Nile tilapia and hogchoker. Proceedings of the Royal Society B: Biological Sciences 279:4705–4714.

Hamner, W. M., P. L. Colin, and P. P. Hamner. 2007. Export-import dynamics of zooplankton on a coral reef in Palau. Marine Ecology Progress Series 334:83–92.

Hamner, W. M., M. S. Jones, J. H. Carleton, I. R. Hauri, and D. M. Willaims. 1988. Zooplankton, planktivorous fish, and water currents on a windward reef face: Great Barrier Reef, Australia. Bulletin of Marine Science 42:459–479.

Hanson, K. M. W. 2011. Planktivorous fish link coral reef and oceanic food webs: causes and consequences of landscape-scale patterns in fish behavior, diet and growth. University of California at San Diego.

Hawkins, A. D., L. Roberts, and S. Cheesman. 2014. Responses of free-living coastal pelagic fish to impulsive sounds. The Journal of the Acoustical Society of America 135:3101– 3116.

Higginbottom, I. R., T. J. Pauly, and D. C. Heatley. 2000. Virtual echograms for visualization and post-processing of multiple-frequency echosounder data. Pages 1497– 1502Proceedings of the Fifth European Conference on Underwater Acoustics.

Hildebrand, J. 2009. Anthropogenic and natural sources of ambient noise in the ocean. Marine Ecology Progress Series 395:5–20.

Hixon, M. A., and W. N. Brostoff. 1996. Succession and herbivory: effects of differential fish grazing on Hawaiian coral-reefs algae. Ecological Monographs 66:67–90.

Hobson, E. S. 1991. Trophic relationships of fishes specialized to feed on zooplankters above coral reefs. Pages 69–95 in P. F. Sale, editor.The ecology of fishes on coral reefs. Academic Press, London.

Holt, R. D. 1984. Spatial heterogeneity, indirect interactions, and the coexistence of prey species. The American Naturalist 124:377–406.

Holzman, R., M. A. Reidenbach, S. G. Monismith, J. R. Koseff, and A. Genin. 2005. Near- bottom depletion of zooplankton over a coral reef II: Relationships with zooplankton swimming ability. Coral Reefs 24:87–94.

Horta e Costa, B., J. Assis, G. Franco, K. Erzini, M. Henriques, E. J. Gonçalves, and J. E. Caselle. 2014. Tropicalization of fish assemblages in temperate biogeographic transition zones. Marine Ecology Progress Series 504:241–252.

Hortal, J., L. M. Carrascal, K. A. Triantis, E. Thébault, S. Meiri, and S. Sfenthourakis. 2013. Species richness can decrease with altitude but not with habitat diversity. Proceedings of the National Academy of Sciences of the United States of America 110:E2149–E2150.

Hoyt, J., J. P. Delgado, B. Barr, B. Terrell, and V. Grussing. 2014. “Graveyard of the Atlantic:” an overview of North Carolina’s maritime cultural landscape. Maritime Heritage Program Series: Number 4. NOAA Office of National Marine Sanctuaries Heritage Program. Page Maritime Heritage Program Series: Number 4.

Huffaker, C. B. 1958. Experimental studies on predation: dispersion factors and predator- prey oscillations. Hilgardia 27:343–383.

Hunter, M. D., and P. W. Price. 1992. Playing chutes and ladders: heterogeneity and the relative roles of bottom-up and top-down forces in natural communities. Ecology 73:724–732.

Hunter, W. R., and M. D. J. Sayer. 2009. The comparative effects of habitat complexity on faunal assemblages of northern temperate artificial and natural reefs. ICES Journal of Marine Science 66:691–698.

Hutchinson, G. E. 1957. Concluding remarks. Population Studies: Animal Ecology and Demography. Cold Spring Harbor Symposia on Quantitative Biology 22:415–427.

Isaaks, E. H., and R. M. Srivastava. 1989. Applied Geostatistics. Oxford University Press, Inc., New York, New York, USA.

Jackson, J. B. C., M. X. Kirby, W. H. Berger, K. A. Bjorndal, L. W. Botsford, B. J. Bourque, R. H. Bradbury, R. Cooke, J. Erlandson, J. A. Estes, T. P. Hughes, S. Kidwell, C. B. Lange, H. S. Lenihan, J. M. Pandolfi, C. H. Peterson, R. S. Steneck, M. J. Tegner, and R. R. Warner. 2001. Historical overfishing and the recent collapse of coastal ecosystems. Science 293:629–638.

Johansen, J. L., D. R. Bellwood, and C. J. Fulton. 2008. Coral reef fishes exploit flow refuges in high-flow habitats. Marine Ecology Progress Series 360:219–226.

Jung, K., S. Kaiser, S. Böhm, J. Nieschulze, and E. K. V. Kalko. 2012. Moving in three dimensions: effects of structural complexity on occurrence and activity of insectivorous bats in managed forest stands. Journal of Applied Ecology 49:523–531.

Kadmon, R., and O. Allouche. 2007. Integrating the effects of area, isolation, and habitat heterogeneity on species diversity: a unification of island biogeography and niche theory. The American Naturalist 170:443–454.

Karlson, R. H., and R. W. Osman. 2012. Species composition and geographic distribution of invertebrates in fouling communities along the east coast of the USA: a regional perspective. Marine Ecology Progress Series 458:255–268.

Kendall, M. S., L. J. Bauer, and C. F. G. Jeffrey. 2009. Influence of hard bottom morphology on fish assemblages of the continental shelf off Georgia, Southeastern USA. Bulletin of Marine Science 84:265–286.

Kendall, M. S., O. P. Jensen, C. Alexander, D. Field, G. McFall, R. Bohne, and M. E.

Monaco. 2005. Benthic mapping using sonar, video transects, and an innovative approach to accuracy assessment: a characterization of bottom features in the Georgia Bight. Journal of Coastal Research 21:1154–1165.

Khanaposhtani, M. G., M. Kaboli, M. Karami, and V. Etemad. 2012. Effect of habitat complexity on richness, abundance and distributional pattern of forest birds. Environmental Management 50:296–303.

Kieser, R., and T. J. Mulligan. 1984. Analysis of echo counting data: a model. Canadian Journal of Fisheries and Aquatic Sciences 41:451–458.

Kline, R. B. 2016. Principles and practice of structural equation modeling. The Guilford Press, New York, NY, USA.

Kloser, R. J., J. D. Penrose, and A. J. Butler. 2010. Multi-beam backscatter measurements used to infer seabed habitats. Continental Shelf Research 30:1772–1782.

Koeck, B., A. Tessier, A. Brind’Amour, J. Pastor, B. Bijaoui, N. Dalias, P. Astruch, G. Saragoni, and P. Lenfant. 2014. Functional differences between fish communities on artificial and natural reefs: a case study along the French Catalan coast. Aquatic Biology 20:219–234.

Korneliussen, R. J., and E. Ona. 2003. Synthetic echograms generated from the relative frequency response. ICES Journal of Marine Science 60:636–640.

Kovalenko, K. E., S. M. Thomaz, and D. M. Warfe. 2012. Habitat complexity: approaches and future directions. Hydrobiologia 685:1–17.

Kracker, L., M. Kendall, and G. McFall. 2008a. Benthic Features as a Determinant for Fish Biomass in Gray’s Reef National Marine Sanctuary. Marine Geodesy 31:267–280.

Kracker, L., M. Kendall, and G. McFall. 2008b. Benthic features as a determinant for fish biomass in Gray’s Reef National Marine Sanctuary. Marine Geodesy 31:267–280.

Kracker, L. M., J. C. Taylor, E. F. Ebert, T. A. Battista, and C. Menza. 2011. Integration of fisheries acoustics surveys and bathymetric mapping to characterize midwater-seafloor habitats of US Virgin Islands and Puerto Rico (2008 – 2010). NOAA Technical Memorandum NOS NCCOS 130. 44pp.

Kruskal, J. B. 1964. Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis. Psychometrika 29:1–27.

Lawler, J. J., J. E. Aukema, J. B. Grant, B. S. Halpern, P. Kareiva, C. R. Nelson, K. Ohleth, J. D. Olden, M. A. Schlaepfer, B. R. Silliman, and P. Zaradic. 2006. Conservation science: a 20-year report card. Frontiers in Ecology and the Environment 4:473–480.

Legendre, P., and M. J. Fortin. 1989. Spatial pattern and ecological analysis. Vegetatio 80:107–138.

Legendre, P., and L. Legendre. 2012. Numerical ecology. 3rd Ed. Elsevier, Amsterdam, the Netherlands.

Leichter, J. J., A. L. Alldredge, G. Bernardi, A. J. Brooks, C. A. Carlson, R. C. Carpenter, P. J. Edmunds, M. R. Fewings, K. M. Hanson, J. L. Hench, S. J. Holbrook, C. E. Nelson, R. J. Schmitt, R. J. Toonen, L. Washburn, and A. S. J. Wyatt. 2013. Biological and physical interactions on a tropical island coral reef: transport and retention processes on Moorea, French Polynesia. Oceanography 26:52–63.

Liao, J. C. 2007. A review of fish swimming mechanics and behaviour in altered flows. Philosophical transactions of the Royal Society of London. Series B, Biological sciences 362:1973–93.

Lindquist, D. G., L. B. Cahoon, I. E. Clavijo, M. H. Posey, S. K. Bolden, L. A. Pike, S. W. Burk, and P. A. Cardullo. 1994. Reef fish stomach contents and prey abundance on reef and sand substrata associated with adjacent artificial and natural reefs in Onslow Bay, North Carolina. Bulletin of Marine … 55:308–318.

Lindquist, D. G., M. V. Ogburn, W. B. Stanley, H. L. Troutman, and S. M. Pereira. 1985. Fish utilization patterns on temperate rubble-mound jetties in North Carolina. Bulletin of Marine Science 37:244–251.

Lindquist, D. G., and L. J. Pietrafesa. 1989. Current vortices and fish aggregations: the current field and associated fishes around a tugboat wreck in Onslow Bay, North Carolina. Bulletin of Marine Science 44:533–544.

Løkkeborg, S., E. Ona, A. Vold, and A. Salthaug. 2012. Sounds from seismic air guns: gear- and species-specific effects on catch rates and fish distribution. Canadian Journal of Fisheries and Aquatic Sciences 69:1278–1291.

Love, R. H. 1977. Target strength of an individual fish at any aspect. The Journal of the Acoustical Society of America 62:1397–1403.

Lucieer, V., N. A. Hill, N. S. Barrett, and S. Nichol. 2013. Do marine substrates “look” and “sound” the same? Supervised classification of multibeam acoustic data using autonomous underwater vehicle images. Estuarine, Coastal and Shelf Science 117:94– 106.

MacArthur, R. H. 1967. The theory of island biogeography. Volume 1. Princeton Univeristy Press.

MacArthur, R. H., and J. W. MacArthur. 1961. On bird species diversity. Ecology 42:594– 598.

MacArthur, R. H., and E. O. Wilson. 1963. An equilibrium theory of insular zoogeography. Evolution 17:373–387.

Macreadie, P. I., A. M. Fowler, and D. J. Booth. 2011. Rigs-to-reefs: Will the deep sea

benefit from artificial habitat? Frontiers in Ecology and the Environment 9:455–461.

Marine Species Identification Portal. (n.d.). . http://species-identification.org.

Martínez, M. L., A. Intralawan, G. Vázquez, O. Pérez-Maqueo, P. Sutton, and R. Landgrave. 2007. The coasts of our world: ecological, economic and social importance. Ecological Economics 63:254–272.

McCann, K. S. 2000. The diversity-stability debate. Nature 405:228–233.

McCauley, R. D., J. Fewtrell, and A. N. Popper. 2003. High intensity anthropogenic sound damages fish ears. The Journal of the Acoustical Society of America 113:638–642.

McCormick, M. I. 1994. Comparison of field methods for measuring surface topography and their associations with a tropical reef fish assemblage. Marine Ecology Progress Series 112:87–96.

Miller, P. J. O., M. P. Johnson, P. T. Madsen, N. Biassoni, M. Quero, and P. L. Tyack. 2009. Using at-sea experiments to study the effects of airguns on the foraging behavior of sperm whales in the Gulf of Mexico. Deep Sea Research I 56:1168–1181.

Myriax Software Pty. Ltd. 2017. Echoview, version 8.0. Myriax Software Pty. Ltd, Hobart, Australia.

National Oceanic and Atmospheric Administration. 1996. Florida Keys National Marine Sanctuary final management plan / environmental impact statement I:1–809.

NC DMF. 1988. North Carolina Artificial Reef Master Plan.

NOAA. 2007. National Artificial Reef Plan (as amended): guidelines for siting, construction, development, and assessment of artificial reefs.

Nowacek, D. P., K. Bröker, G. Donovan, G. Gailey, R. Racca, R. R. Reeves, A. I. Vedenev, D. W. Weller, and B. L. Southall. 2013. Responsible practices for minimizing and monitoring environmental impacts of marine seismic surveys with an emphasis on marine mammals. Aquatic Mammals 39:356–377.

Nowacek, D. P., C. W. Clark, D. Mann, P. J. O. Miller, H. C. Rosenbaum, J. S. Golden, M. Jasny, J. Kraska, and B. L. Southall. 2015. Marine seismic surveys and ocean noise: time for coordinated and prudent planning. Frontiers in Ecology and the Environment 13:378–386.

Oksanen, J., F. Guillaume Blanchet, R. Kindt, P. Legendre, P. R. Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, and H. Wagner. 2015. vegan: Community Ecology Package R package.

Ord, J. K., and A. Getis. 1995. Local spatial autocorrelation statistics: distributional issues and an application. Geographical Analysis 27:286–306.

Oricchio, F. T., G. Pastro, E. A. Vieira, A. A. V Flores, F. Z. Gibran, and G. M. Dias. 2016. Distinct community dynamics at two artificial habitats in a recreational marina. Marine Environmental Research 122:85–92.

Pace, M. L., J. J. Cole, S. R. Carpenter, and J. F. Kitchell. 1999. Trophic cascades revealed in diverse ecosystems. Trends in Ecology & Evolution 14:483–488.

Parente, V., D. Ferreira, E. Moutinho dos Santos, and E. Luczynski. 2006. Offshore decommissioning issues: deductibility and transferability. Energy Policy 34:1992–2001.

Parker-Stetter, S. L., L. G. Rudstam, P. J. Sullivan, and D. M. Warner. 2009. Standard operating procedures for fisheries acoustic surveys in the Great Lakes. Great Lakes Fish. Comm. Spec. Pub. 09-01.

Parker Jr., R. O. 1990. Tagging studies and diver observations of fish populations on live- bottom reefs of the U.S. southeastern coast. Bulletin of Marine Science 46:749–760.

Parker Jr., R. O., and R. L. Dixon. 1998. Changes in a North Carolina reef fish community after 15 years of intense fishing — global warming implications. Transactions of the American Fisheries Society 127:908–920.

Pauly, D., V. Christensen, J. Dalsgaard, R. Froese, and F. J. Torres. 1998. Fishing down marine food webs. Science 279:860–863.

Paxton, A. B., E. A. Pickering, A. M. Adler, J. C. Taylor, and C. H. Peterson. 2017. Flat and complex temperate reefs provide similar support for fish: evidence for a unimodal species-habitat relationship. PLOS ONE 12:e0183906.

Perkol-Finkel, S., and Y. Benayahu. 2005. Recruitment of benthic organisms onto a planned artificial reef: shifts in community structure one decade post-deployment. Marine Environmental Research 59:79–99.

Perkol-Finkel, S., N. Shashar, and Y. Benayahu. 2006. Can artificial reefs mimic natural reef communities? The roles of structural features and age. Marine environmental research 61:121–135.

Pinheiro, J., D. Bates, S. DebRoy, D. Sarkar, and R Development Core Team. 2013. nlme: Linear and Nonlinear Mixed Effect Models R package.

Pirotta, E., K. L. Brookes, I. M. Graham, and P. M. Thompson. 2014. Variation in harbour porpoise activity in response to seismic survey noise. Biology Letters 10:20131090.

Popper, A. N., and M. C. Hastings. 2009. The effects of human-generated sound on fish. Integrative Zoology 4:43–52.

Popper, A. N., A. D. Hawkins, R. R. Fay, D. A. Mann, S. Bartol, T. J. Carlson, S. Coombs, W. T. Ellison, R. L. Gentry, M. B. Halvorsen, S. Løkkeborg, P. H. Rogers, B. L. Southal, D. G. Zeddies, and W. N. Tavolga. 2014. Sound Exposure Guidelines for

Fishes and Sea Turtles: A Technical Report prepared by ANSI-Accredited Standards Committe S3/SC1 and registered with ANSI. Springer Briefs in Oceanography, ASA Press and Springer, London.

Popper, A. N., M. E. Smith, P. A. Cott, B. W. Hanna, A. O. MacGillivray, M. E. Austin, and D. A. Mann. 2005. Effects of exposure to seismic airgun use on hearing of three fish species. The Journal of the Acoustical Society of America 117:3958–3971.

Power, M. E. 1992. Top-down and bottom-up forces in food webs: do plants have primacy? Ecology 73:733–746.

Prada, M. C., R. S. Appeldoorn, and J. A. Rivera. 2008. The effects of minimum map unit in coral reefs maps generated from high resolution side scan sonar mosaics. Coral Reefs 27:297–310.

Prairie, J. C., K. R. Sutherland, K. J. Nickols, and A. M. Kaltenberg. 2012. Biophysical interactions in the plankton: a cross-scale review. Limnology & Oceanography: Fluids & Environments 2:121–145.

Purser, J., and A. N. Radford. 2011. Acoustic noise induces attention shifts and reduces foraging performance in three-spined sticklebacks (Gasterosteus aculeatus ). PLoS ONE 6:e17478.

Python Software Foundation. 2017. Python v. 2.7.12.

Racca, R., M. Austin, A. Rutenko, and K. Bröker. 2015. Monitoring the gray whale sound exposure mitigation zone and estimating acoustic transmission during a 4-D seismic survey, Sakhalin Island, Russia. Endangered Species Research 29:131–146.

R Core Team. 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

R Development Core Team. 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

R Development Core Team. 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Renard, D., N. Bez, N. Desassis, H. Beaucher, F. Ors, and X. Freulon. 2017. RGeostats: Geostatistical package.

Renaud, P. E., W. G. Ambrose Jr., S. R. Riggs, and D. A. Syster. 1996. Multi-level effects of severe storms on an offshore temperate reef system: benthic sediments, macroalgae, and implications for fisheries. Marine Ecology 17:383–398.

Renaud, P. E., S. R. Riggs, W. G. Ambrose Jr., K. Schmid, and S. W. Snyder. 1997. Biological-geological interactions: storm effects on macroalgal communities mediated by sediment characteristics and distribution. Continental Shelf Research 17:37–56.

Renaud, P. E., D. A. Syster, and W. G. Ambrose Jr. 1999. Recruitment patterns of continental shelf benthos off North Carolina, USA: effects of sediment enrichment and impact on community structure. Journal of Experimental Marine Biology and Ecology 237:89–106.

Riggs, S. R., W. G. Ambrose Jr., J. W. Cook, S. W. Snyder, and S. W. Snyder. 1998. Sediment production on sediment-starved continental margins: the interrelationship between hardbottoms, sedimentological and benthic community processes, and storm dynamics. Journal of Sedimentary Research 68:155–168.

Riggs, S. R., S. W. Snyder, A. C. Hine, and D. L. Mearns. 1996. Hardbottom morphology and relationship to the geologic framework: mid-Atlantic continental shelf. Journal of Sedimentary Research 66:830–846.

Rilov, G., and Y. Benayahu. 1998. Vertical artificial structures as an alternative habitat for coral reef fishes in disturbed environments. Marine Environmental Research 45:431– 451.

Risk, M. J. 1972. Fish diversity on a coral reef in the Virgin Islands. Atoll Research Bulletin:1–4.

Rooker, J. R., Q. R. Dokken, C. V. Pattengill, and G. J. Holt. 1997. Fish assemblages on artificial and natural reefs in the Flower Garden Banks National Marine Sanctuary, USA. Coral Reefs 16:83–92.

Rosemond, R. C., A. B. Paxton, H. R. Lemoine, S. R. Fegley, and C. H. Peterson. 2018. Fish use of reef structures and adjacent sand flats: implications for selecting minimum buffer zones between artificial reefs and existing reefs. Marine Ecology Progress Series In press.

Ross, S. W., and M. L. Moser. 1995. Life history of juvenile gag, Mycteroperca microlepsis , in North Carolina estuaries. Bulletin of Marine Science 56:222–237.

Rosseeel, Y. 2012. lavaan: an R package for structural equation modeling. Journal of Statistical Software 48:1–36.

Ryan, T. E., R. A. Downie, R. J. Kloser, and G. Keith. 2015. Reducing bias due to noise and attenuation in open-ocean echo integration data. ICES 72:2482–2493.

Samoilys, M. A., and G. Carlos. 2000. Determining methods of underwater visual census for estimating the abundance of coral reef fishes. Environmental Biology of Fishes 57:289– 304.

Samuels, C. L., and J. A. Drake. 1997. Divergent perspectives on community convergence. Trends in Ecology and Evolution 12:427–432.

Saraux, C., J.-M. Fromentin, J.-L. Bigot, J.-H. Bourdeix, M. Morfin, D. Roos, E. Van Beveren, and N. Bez. 2014. Spatial structure and distribution of small pelagic fish in the

northwestern Mediterranean Sea. PLoS ONE 9:e111211.

Schneider, C. W., and R. B. Searles. 1991. Seaweeds of the southeastern United States: Cape Hatteras to Cape Canaveral. Page 569. Duke University Press.

Schneider, K. N., and K. O. Winemiller. 2008. Structural complexity of woody debris patches influences fish and macroinvertebrate species richness in a temperate floodplain-river system. Hydrobiologia 610:235–244.

Seaman, W. J., editor. 2000. Artificial reef evaluation with application to natural marine habitats. CRC Press, Boca Raton, FL.

Seiferling, I., R. Proulx, and C. Wirth. 2014. Disentangling the environmental-heterogeneity- -species-diversity relationship along a gradient of human footprint. Ecology 95:2084– 2095.

Seto, K. C., B. Guneralp, and L. R. Hutyra. 2012. Global forecasts of urban expansion to 2030 and direct impacts on biodiversity and carbon pools. Proceedings of the National Academy of Sciences 109:16083–16088.

Sheather, S. J., and M. C. Jones. 1991. A reliable data-based bandwidth selection method for kernel density estimation. Journal of the Royal Statistical Society. Series B 53:683–690.

Shepard, R. N. 1962. The analysis of proximities: multidimensional scaling with an unknown distance function. I. Psychometrika 27:125–140.

Simard, Y., R. de Ladurantaye, and J.-C. Therriault. 1986. Aggregation of euphausiids along a coastal shelf in an upwelling environment. Marine Ecology Progress Series 32:203– 215.

Simberloff, D. S., and E. O. Wilson. 1970. Experimental zoogeography of islands. A two- year record of colonization. Ecology 51:934–937.

Simmonds, J., and D. N. MacLennan. 2005. Fisheries Acoustics: Theory and Practice. 2nd Ed. Blackwell Science, Oxford, UK.

Simon, T., J.-C. Joyeux, and H. T. Pinheiro. 2013. Fish assemblages on shipwrecks and natural rocky reefs strongly differ in trophic structure. Marine Environmental Research 90:55–65.

Simonsen, K. A., J. H. Cowan, and K. M. Boswell. 2015. Habitat differences in the feeding ecology of red snapper ( Lutjanus campechanus , Poey 1860): a comparison between artificial and natural reefs in the northern Gulf of Mexico. Environmental Biology of Fishes 98:811–824.

Skalski, J. R., W. H. Pearson, and C. I. Malme. 1992. Effects of sounds from a geophysical survey device on catch-per-unit-effort in a hook-and-line fishery for rockfish (Sebastes spp.). Canadian Journal of Fisheries and Aquatic Sciences 49:1357–1365.

Slabbekoorn, H., N. Bouton, I. van Opzeeland, A. Coers, C. ten Cate, and A. N. Popper. 2010. A noisy spring: the impact of globally rising underwater sound levels on fish. Trends in Ecology and Evolution 25:419–427.

Slotte, A., K. Hansen, J. Dalen, and E. Ona. 2004. Acoustic mapping of pelagic fish distribution and abundance in relation to a seismic shooting area off the Norwegian west coast. Fisheries Research 67:143–150.

Smith, J. A., M. B. Lowry, C. Champion, and I. M. Suthers. 2016. A designed artificial reef is among the most productive marine fish habitats: new metrics to address “production versus attraction.” Marine Biology September:163–188.

Solonsky, A. C. 1985. Fish colonization and the effect of fishing activities on two artificial reefs in Monterey Bay, California. Bulletin of Marine Science 37:336–347.

Song, J., D. A. Mann, P. A. Cott, B. W. Hanna, and A. N. Popper. 2008. The inner ears of Northern Canadian freshwater fishes following exposure to seismic air gun sounds. The Journal of the Acoustical Society of America 124:1360–1366.

South Atlantic Fishery Management Council. 1983. Fishery management plan, regulatory impact review, and final environmental impact statement for the snapper-grouper fishery of the South Atlantic region. Prepared by the South Atlantic Fishery Management Council in cooperation with National Marine Fisheries Service.

South Atlantic Fishery Management Council. 2016. Snapper Grouper Management Complex: Species Managed by the South Atlantic Fishery Management Council. http://safmc.net/wp- content/uploads/2016/06/SAFMC_SnapperGrouperManagedSpecies_091614.pdf.

Stallings, C. D., F. C. Coleman, C. C. Koenig, and D. A. Markiewicz. 2010. Energy allocation in juveniles of a warm-temperate reef fish. Environmental Biology of Fishes 88:389–398.

Stephan, C. D., and D. G. Lindquist. 1989. A comparative analysis of the fish assemblages associated with old and new shipwrecks and fish aggregating devices in Onslow Bay, North Carolina. Bulletin of Marine Science 44:698–717.

Stick, D. 1989. Graveyard of the Atlantic: Shipwrecks of the North Carolina Coast. Univeristy of North Carolina Press.

Stuart-Smith, R. D., G. J. Edgar, N. S. Barrett, S. J. Kininmonth, and A. E. Bates. 2015. Thermal biases and vulnerability to warming in the world’s marine fauna. Nature 528:88–92.

Taylor, J. C., and E. Ebert. 2012. Mapping coral reef fish schools and aggregations with high-frequency multibeam and split-beam sonars. Acoustical Society of America: Proceedings of Meetings on Acoustics:1–9.

Taylor, J. C., A. B. Paxton, C. M. Voss, B. Sumners, C. A. Buckel, J. Vander Pluym, E. F. Ebert, T. S. Viehman, S. R. Fegley, E. A. Pickering, A. M. Adler, C. Freeman, and C. H. Peterson. 2016. Benthic habitat mapping and assessment in the Wilmington-East wind energy call area. OCS Study BOEM 2016-003 and NOAA Technical Memorandum 196. Atlantic OCS Region, Sterling, VA.

Tews, J., U. Brose, V. Grimm, K. Tielbörger, M. C. Wichmann, M. Schwager, and F. Jeltsch. 2004. Animal species diversity driven by habitat heterogeneity/diversity: the importance of keystone structures. Journal of Biogeography 31:79–92.

Thanner, S. E., T. L. McIntosh, and S. M. Blair. 2006. Development of benthic and fish asemblages on artificial reef materials compared to adjacent natural reef assemblages in Miami-Dade County, Florida. Bulletin of Marine Science 78:57–70.

The Mathworks Inc. 2017. MATLAB v. 2017b. Natick, MA.

Tibshirani, R., and F. Leisch. 2017. bootstrap: functions for the book “An Introduction to the Bootstrap.”

Tolstoy, M., J. Diebold, L. Doermann, S. Nooner, S. C. Webb, D. R. Bohnenstiehl, T. J. Crone, and R. C. Holmes. 2009. Broadband calibration of the R/V Marcus G. Langseth four-string seismic sources. Geochemistry, Geophysics, Geosystems 10:1–15.

Topping, D. T., and S. T. Szedlmayer. 2011. Home range and movement patterns of red snapper ( Lutjanus campechanus ) on artificial reefs. Fisheries Research 112:77–84.

UNEP. 2009. London Convention and Protocol / UNEP Guidelines for the Placement of Artificial Reefs. London, UK.

UNESCO. 2017. Underwater Cultural Heritage: Wrecks. http://www.unesco.org/new/en/culture/themes/underwater-cultural-heritage/underwater- cultural-heritage/wrecks/.

Urick, R. J. 1983. Principles of Underwater Sound. 3rd Ed. Peninsula Publishing, Westport, CT.

Venables, M. N., and B. D. Ripley. 2002. Modern Applied Statistics with S. 4th Ed. Springer, New York.

Vergés, A., C. Doropoulos, H. A. Malcolm, M. Skye, M. Garcia-Pizá, E. M. Marzinelli, A. H. Campbell, E. Ballesteros, A. S. Hoey, A. Vila-Concejo, Y.-M. Bozec, and P. D. Steinberg. 2016. Long-term empirical evidence of ocean warming leading to tropicalization of fish communities, increased herbivory, and loss of kelp. Proceedings of the National Academy of Sciences of the United States of America 113:13791– 13796.

Vinueza, L. R., B. A. Menge, D. Ruiz, and D. M. Palacios. 2014. Oceanographic and climatic variation drive top-down / bottom-up coupling in the Galápagos intertidal meta-

ecosystem. Ecological Monographs 84:411–434.

Walker, B. K., B. Riegl, and R. E. Dodge. 2008. Mapping coral reef habitats in southeast Florida using a combined technique approach. Journal of Coastal Research 24:1138– 1150.

Wardle, C. S., T. J. Carter, G. G. Urquhart, A. D. F. Johnstone, A. M. Ziolkowski, G. Hampson, and D. Mackie. 2001. Effects of seismic air guns on marine fish. Continental Shelf Research 21:1005–1027.

Webb, P. W. 1998. Entrainment by river chub Nocomis micropogon and smallmouth bass Micropterus dolomieu on cylinders. The Journal of Experimental Biology 201:2403– 2412.

Whitfield, P. E., R. C. Muñoz, C. A. Buckel, B. P. Degan, D. W. Freshwater, and J. A. Hare. 2014. Native fish community structure and Indo-Pacific lionfish Pterois volitans densities along a depth-temperature gradient in Onslow Bay, North Carolina, USA. Marine Ecology Progress Series 509:241–254.

Woillez, M., J.-C. Poulard, J. Rivoirard, P. Petitgas, and N. Bez. 2007. Indices for capturing spatial patterns and their evolution in time, with application to European hake (Merluccius merluccius ) in the Bay of Biscay. ICES Journal of Marine Science 64:537– 550.

Woillez, M., J. Rivoirard, and P. Petitgas. 2009. Notes on survey-based spatial indicators for monitoring fish populations. Aquatic Living Resources 22:155–164.

Worm, B., E. B. Barbier, N. Beaumont, J. E. Duffy, C. Folke, B. S. Halpern, J. B. C. Jackson, H. K. Lotze, F. Micheli, S. R. Palumbi, E. Sala, K. A. Selkoe, J. J. Stachowicz, and R. Watson. 2006. Impacts of biodiversity loss on ocean ecosystem services. Science 314:787–790.

WoRMS Editorial Board. 2016. World Register of Marine Species.

Xylem Inc. 2016. HYPACK. Middletown, CT.