Ecological Drivers of the Brown Macroalgal Genus Lobophora on Pacific Coral Reefs

Ecological drivers of the brown macroalgal genus Lobophora on Pacific coral reefs

Laura Dorothea Puk B. Sc., M. Sc.

A thesis submitted for the degree of Doctor of Philosophy at The University of Queensland in 2019 School of Biological Sciences

Abstract Coral reefs are threatened by increasingly frequent large-scale coral bleaching events that often lead to mass coral mortality. In the wake of coral mortality, macroalgae cover can increase drastically. A common alga contributing to these blooms is the brown macroalgal genus Lobophora. Lobophora has been shown to have many negative effects on corals, including severely reduced coral larval settlement, reduced coral fecundity and growth, and direct competition with adult corals. However, it is still unclear what drives the establishment of Lobophora on coral reefs. Therefore, the overall aim of this thesis is to study drivers of Lobophora assemblages and the control of the alga by herbivorous fishes.

In Chapter 2 I examine the local cryptic diversity of Lobophora in Palau and study how wave exposure, herbivorous fish biomass and depth influence Lobophora diversity and species assemblages. I collected Lobophora specimens from 12 reefs around Palau along a wave exposure and herbivorous fish biomass gradient at 3 m and 10 m depth. The specimens were identified to species level using two chloroplast and one mitochondrial marker. I found 15 species of which ten are currently undescribed. The results further suggest that neither of the studied environmental factors influences the diversity of Lobophora. Species assemblages are shaped only by wave exposure on an island-wide geographic scale but by wave exposure as well as herbivore biomass and depth on a more local geographic scale. I also identify generalist and specialist species. These findings have important implications as Lobophora species may have different competitive strengths and the coral-algal interactions may, therefore, vary between Lobophora species. Understanding the drivers of Lobophora assemblages and which species may have a particularly large potential to spread to other reefs (i.e., generalists) may guide future studies of Lobophora-coral interactions.

In Chapter 3 I study the life-stage specific control of Lobophora by herbivorous fishes. Using field observations and caging experiments, I found that Lobophora is more readily controlled at the recruit stage than the adult stage. I also observed a temporal trend in Lobophora growth and recruitment, with growth overpowering herbivory over the course of a few weeks. In a controlled tank experiment and camera-based field observations, I identify certain fish species as particularly important for the control of Lobophora at the recruit stage. Out of seven observed fish species, only three (the parrotfish Scarus niger, Chlorurus spilurus and the surgeonfish Acanthurus nigrofuscus) removed Lobophora recruits. These findings are important as they show that fish identity plays a role in the control of Lobophora – even at the recruit stage. Further, my results suggest that if a reduction in herbivory was to occur at a time of high Lobophora growth and recruitment, the alga may be able to overpower herbivory and increase its cover on a reef.

In Chapter 4 I study the role refuges play for the establishment of Lobophora on coral reefs. Using a combination of field surveys and manipulative field experiments, I show that concealed microhabitats (i.e. crevices) provide a refuge for Lobophora recruits and that Lobophora abundance in the field is negatively associated with crevice size. My results also suggest that parrotfish are the main control of Lobophora abundance and that fish size is a good predictor of grazing rates in microhabitats. These findings show that Lobophora gains protection from herbivory in crevices and may be able to spread from these microhabitats when herbivory is decreased due to fishing or mass coral mortality to establish dominance on a coral reef.

In Chapter 5 I developed an individual-based model using Lobophora recruitment rates, recruit and adult mortality associated with specific fish species and microhabitat types collected in Chapters 3 and 4 to simulate Lobophora dynamics. Some additional information from existing literature was incorporated and some minor adjustments allowed the model to successfully simulate Lobophora cover within different microhabitat types. The model predictions show the impact of fish species- specific overfishing on Lobophora cover and highlight the importance of concealed microhabitats in facilitating Lobophora.

Overall, this thesis increases our knowledge of Lobophora diversity and ecology and its control by herbivorous fishes. I have shown that a strong shift in grazing susceptibility exists between early life- stage and adult Lobophora and that microhabitats are important in facilitating Lobophora. Further, this thesis has revealed fish species-specific abilities to control Lobophora and has identified parrotfish and Acanthurus nigrofuscus as the herbivorous fish species to have the strongest control over the alga’s recruits. However, a diverse suite of herbivorous fish species may be more effective at controlling Lobophora in a variety of microhabitats. These findings have implications for fisheries management and managers should consider protecting herbivorous fish species diversity, possibly with particular attention to Acanthurus spp. and parrotfish. The strong ontogenetic shift in grazing susceptibility further indicates that fisheries restrictions will likely be more effective if implemented before a shift to Lobophora dominance occurs.

ii Declaration by author This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, financial support and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my higher degree by research candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis and have sought permission from co-authors for any jointly authored works included in the thesis.

iii Publications included in this thesis Incorporated as Chapter 3: Puk LD, Cernohorsky N, Marshell A, Dwyer J, Kennedy Wolfe, Mumby PJ (2020). Species-specific effects of herbivorous fishes on the establishment of the macroalga Lobophora on coral reefs. Marine Ecology Progress Series 637:1-14.

Contributor Statement of contribution Laura D Puk Conception (60%) Data collection (70%) Data analysis and interpretation (70%) Drafting and production (75%) Nicole Cernohorsky Data collection (20%) Drafting and production (5%) Alyssa Marshell Conception (20%) Drafting and production (5%) John Dwyer Data analysis and interpretation (20%) Drafting and production (5%) Kennedy Wolfe Data collection (10%) Drafting and production (5%) Peter J Mumby Conception (20%) Data analysis and interpretation (10%) Drafting and production (5%)

Incorporated as Chapter 4: Puk LD, Marshell A, Dwyer J, Evensen NR, Mumby PJ (2020). Refuge dependent herbivory controls a key macroalga on coral reefs. Coral Reefs. Doi: https://doi.org/10.1007/s00338-020-01915-9.

Contributor Statement of contribution Laura D Puk Conception (50%) Data collection (90%) Data analysis and interpretation (60%) Drafting and production (80%) Alyssa Marshell Conception (10%)

iv Drafting and production (5%) John Dwyer Data analysis and interpretation (20%) Drafting and production (5%) Nicolas R Evensen Conception (5%) Data collection (10%) Drafting and production (5%) Peter J Mumby Conception (35%) Data analysis and interpretation (20%) Drafting and production (5%)

v Submitted manuscripts included in this thesis No manuscripts submitted for publication.

Other publications during candidature Book chapters Gordon IJ, Prins HHT, Mallon J, Puk LD, Miranda EBP, Starling-Manne C, van der Wal R, Moore B, Foley W, Lush L, Maestri R, Matsuda I, Clauss M (2019) The ecology of browsing and grazing in other vertebrate taxa. In: Gordon I, Prins H (eds) The ecology of browsing and grazing II. Ecological Studies (Analysis and Synthesis), vol 239. Springer, Cham.

Conference abstracts Puk LD, Marshell A, Dwyer J, Mumby PJ. The role of herbivorous fish diversity and algal life-stage specific grazing vulnerability in keeping reefs free of a key macroalga. Annual Australian Marine Sciences Association AMSA Conference. Fremantle, Western Australia, Australia. Oral presentation.

Puk LD, Vieira C, Roff G, Marshell A, Dwyer J, Mumby PJ. DNA barcoding reveals high cryptic diversity in a common coral reef macroalga (Lobophora, Phaeopyhceae). The 92nd ACRS Conference. Moreton Island, Queensland, Australia. Oral presentation.

Puk LD, Marshell A, Dwyer J, Mumby PJ. Where macroalgae hide – Reduced herbivore access to microhabitats provides refuge for the brown macroalga Lobophora. The 91st ACRS Conference. Exmouth, Western Australia, Australia. Oral presentation.

Contributions by others to the thesis Chapter 1: J Dwyer and A Marshell provided feedback on the manuscript.

Chapter 2: PJ Mumby and G Roff provided advice on the study design. NR Evensen, C Wiseman, and A Kavanagh helped with the data collection. C Vieira conducted the genetic analysis and provided advice on the histological analysis of the Lobophora specimen. PJ Mumby, C Vieira, G Roff provided feedback and comments on the manuscript. LD Puk designed the sampling, collected and analysed the data, conducted the histological analysis of the samples and wrote the manuscript.

Chapter 3: PJ Mumby, A Marshell and G Roff provided advice on the experimental design. N Cernohorsky and N Evensen helped collect the data. A Marshell, J Dwyer, PJ Mumby, N Cernohorsky

vi provided feedback and comments on the manuscript. J Dwyer provided statistical advice. LD Puk designed the experiments, collected and analysed the data, and wrote the manuscript.

Chapter 4: PJ Mumby and A Marshell provided advice on the experimental design. NR Evensen helped collect field work data. J Dwyer provided statistical advice. PJ Mumby, NR Evensen, A Marshell, J Dwyer provided feedback and comments on the manuscript. LD Puk designed the experiments, collected and analysed the data, and wrote the manuscript.

Chapter 5: PJ Mumby assisted with the design and implementation of the model. LD Puk collected the data, designed the model, analysed the results and wrote the manuscript. PJ Mumby and A Marshell provided feedback on the manuscript.

Chapter 6: J Dwyer and A Marshell provided feedback on the manuscript.

vii Statement of parts of the thesis submitted to qualify for the award of another degree No works submitted towards another degree have been included in this thesis.

Research Involving Human or Animal Subjects The Animal Ethics Committee has approved this research under the title “Population dynamics of the brown macroalga Lobophora sp. under fish herbivory” under the AEC Approval Number SBS/396/16/PALAU (Approval Letter in Thesis Appendix).

viii Acknowledgements Firstly, I would like to thank my supervisor, Peter Mumby, for giving me the opportunity to be part of the Marine Spatial Ecology Lab after a brief meeting in the Philippines. Pete, I am grateful for your guidance and support throughout this process; you taught me a lot.

I would also like to thank my co-supervisors John Dwyer and Alyssa Marshell for their ongoing support and advice. Thank you, Alyssa, for letting me benefit from your previous experience conducting similar experiments. And John, I would like to thank you especially for your help with the statistical analysis and trying to understand the weird experiments we marine ecologists do. I learnt a lot.

I am grateful to all the staff at PICRC for enabling me to do my field work in such a beautiful country. I also want to recognise the incredible amount of work all my volunteers have put into this project: Thank you, Brooke Brown, Nicole Cernohorsky, Alan Kavanagh, Kelsea Miller, Friederike Pfeiffer, Shannen Smith, Alex Tredinnick, Charlie Wiseman, and Kelly Wong. I couldn’t have done this without you!

Everyone at the Marine Spatial Ecology Lab has been helpful along the way. However, I would like to especially thank Nicolas Evensen, George Roff and Nils Krueck. Nick, thanks for all your help during field work. Long days in the field can be exhausting, but it’s been a lot of fun and it was great being involved in some coral spawning experiments as well! Jez, I want to thank you for the continued advice throughout this journey and long hours figuring out how to use R for spatial mapping. You are an incredible scientist and teacher and I’m very happy I got to learn from you. Nils, thanks so much for your advice in all things PhD, I can’t say how much I appreciated your help! Thank you also to Tania Kenyon for going through this PhD experience with me and all the coffees and lunches spent talking through our experiences. Kay Critchell, thank you for being the generous and caring person you are and always offering your help and support.

I would also like to thank the Australian Research Council for making this all possible by providing funding to my supervisor. Thank you to the Winifred Violet Scott Trust Fund for their generous support of my research as well. I am also grateful to the University of Queensland for my scholarship and the Australian Coral Reef Society and the Centre of Excellence for Coral Reef Studies for funding my attendance at conferences.

ix A special thanks goes to my family, who have put up with me being away for so long but have stayed supportive throughout the whole process. Thank you to my Mum, Birgit, for always being supportive of my journey even if it takes me to the other end of the world. You have had such a big role to play in making my childhood dream become reality! Thank you to my Dad, Christian, for all your help and support and always making me feel like I can do whatever I set my mind on. To my sisters, Cora and Rosa Li: Thank you so much for always being there, you are so important to me! Thank you also to Jol, Deven, Martin, Hanane, Detlef and everyone I might have forgotten here right now!

I also want to thank my friends for their support. A special thanks goes to Carrie Sims. It was great to share this experience with you. My time here wouldn’t have been the same without you! Thank you also to my housemates here in Brisbane, who were always there to listen to the ups and downs – you make me feel at home here. Lastly, I want to thank Dan for always being there for me, tolerating the bad days and celebrating the little victories. I am very happy to have you in my life!

x Financial support This research was supported by the University of Queensland International Scholarship, partly funded by a grant by the Australian Research Council to PJ Mumby. This research was further supported by the Winifred Violet Scott Trust Fund.

Keywords macroalgae, fish ecology, herbivory, structural complexity, cryptic diversity, species interactions

Australian and New Zealand Standard Research Classifications (ANZSRC) ANZSRC code: 060205, Marine and Estuarine Ecology, 60 % ANZSRC code: 060207, Population Ecology, 20 % ANZSRC code: 050102, Ecosystem Function, 20 %

Fields of Research (FoR) Classification FoR code: 0602, Ecology, 80% FoR Code: 0501, Ecological Applications, 20%

xii Table of Contents

1.1 Biodiversity and the degradation of ecosystems ...... 2

1.2 Herbivory and phase shifts on coral reefs ...... 2

1.3 The brown macroalga Lobophora ...... 5

1.4 Thesis aims ...... 8

1.5 Thesis structure ...... 9

2.1 Abstract ...... 12

2.2 Introduction ...... 12

2.3 Materials & Methods ...... 15

2.4 Results ...... 18

2.5 Discussion ...... 27

2.6 Acknowledgements ...... 33

xiii 3.1 Abstract ...... 35

3.2 Introduction ...... 35

3.3 Material & Methods ...... 38

3.4 Results ...... 43

3.5 Discussion ...... 49

3.6 Acknowledgements ...... 52

3.7 Supplementary ...... 52

4.1 Abstract ...... 59

4.2 Introduction ...... 59

4.3 Material & Methods ...... 61

4.4 Results ...... 68

4.5 Discussion ...... 76

4.6 Acknowledgements ...... 80

4.7 Supplementary ...... 80

5.1 Abstract ...... 94

5.2 Introduction ...... 94

5.3 Methods ...... 97

5.4 Results ...... 106

5.5 Discussion ...... 113

6.1 Significance of this thesis ...... 119

6.2 The importance of Lobophora bottlenecks ...... 120

6.3 Role of individual fish species in controlling Lobophora blooms ...... 120 xiv 6.4 Lobophora species identity ...... 122

6.5 Challenges ...... 123

6.6 Conclusions & Future studies ...... 123

xv List of Figures Figure 2.1: Map and location of sites used in Lobophora sampling around Palau. Wave exposure (W m-2) gradient is shown on the border of the reef and the twelve sites are marked (S1 – S12)...... 16 Figure 2.2: Phylogenetic tree of identified Lobophora species. The tree was created using one mitochondrial (cox3) and two chloroplast (psbA, rbcL) markers. Species found in Palau are highlighted in yellow, all other species have previously been collected and added to a Lobophora database...... 20 Figure 2.3: Transversal cut and in-situ pictures of the 12 Lobophora species with enough material to allow histological analysis. A) L. sp. 10, B) L. sp. 48, C) L. sp. 109, D) L. asiatica, E) L. boudeuseae, F) L. gibbera, G) L. providenceae, H) L. sp. 108, I) L. sp. 106, J) L. sp. 111, K) L. sp. 105, L) L. sp. 112...... 21 Figure 2.4: DbRDA plot of the Lobophora species assemblage showing wave exposure categories (low, medium, high) and clusters based on 45% similarity; the sites are identified as S1 – S12 following the numbering system on the small map...... 22 Figure 2.5: DbRDA plot of the Lobophora species assemblages on the East coast of Palau only, showing A) wave exposure categories (low, medium, high) and B) herbivore biomass categories (low, medium, high) as used to classify the Eastern sites only. Symbol shape represents depth. Clusters are based on 50% similarity; the sites are identified as S1 – S9 following the numbering system on the small map...... 24 Figure 2.6: Lobophora species clustering as determined by morphology, microhabitat, thallus thickness, height and width. Morphology and microhabitat were predictors in the analysis and are presented besides spatial refugia in exposure, herbivory and depth, which were not used as predictors. Additionally, how many sites the species was found in is shown. All Lobophora species are categorized as local, specialist, generalist - less common or generalist - common. The categorization was as follows: Local = only at one site; specialist ≥2 sites, restricted to one level in at least one category (further details as per the nature of specialization in graph); generalist - uncommon = at less than half of all sites, at least two levels out of each category; generalist - common = at over half of all sites, at least two levels out of each category. Species underneath dashed line were excluded from cluster analysis due to missing morphological data...... 27 Figure 3.1: Percentage Lobophora cover observed over nine weeks. Data is obtained through 50 random cells which were followed throughout the experiment. Individual observations and polynomial regressions are displayed. Error margins show the 95 % confidence interval...... 44 Figure 3.2: Kaplan-Meier survival curve using 50 random cells within each plot type of which cells occupied by Lobophora were followed throughout a nine-week period. + symbols show that some

xvi cells did not ‘die’ at that time-point, but disappeared from observations, which happens when a cell was still alive at the end of the experimental period and its fate is therefore unknown...... 45 Figure 3.3: Percentage of bites taken by surgeonfish and parrotfish on Lobophora in a monitored plot of 50 cm × 50 cm graphed over the percentage Lobophora cover within that plot (as percentage of available area). Dots are individual observations, colours show species identity and the line shows a ratio of 1:1. Dots falling above the line indicated more than proportionate amounts of bites have been taken, dots below the line indicate fewer bites than expected...... 46 Figure 3.4: Comparison of Lobophora recruits (per cm2) on easily accessible crowns of tiles in March 2018 and flat tiles deployed in September 2017. Letters symbolize significantly different results. Mean area and standard error are displayed...... 47 Figure 3.5: A) Mean proportion of Lobophora recruits removed by different fish species during the controlled tank experiment. Letters symbolize significantly different results. Errorbars show standard error. B) Probability of a Lobophora recruit being removed when a fish species visited a tile during in-situ feeding observations and took bites on locations with Lobophora recruits present. Locations were included when only one species took bites on them to avoid confounding feeding by multiple species on the same recruit location. Letters symbolize significantly different results. ‘Max. bites’ are the maximum number of bites which were taken on one location...... 48 Figure 4.1: Feeding angles used to classify the ability of fish to take bites in certain microhabitats...... 63 Figure 4.2: Experimental design to test the influence of crevice width on Lobophora recruitment. Three different crevice widths were used within three different herbivory treatments. Coloured microhabitats mark experimental crevices (sizes in graph) and crowns (12 mm × 12 mm). Recruits were counted in each of these coloured microhabitats...... 65 Figure 4.3: A – C) Densities of observed and measured microhabitats in the field. All densities are proportionate within their category (overall or with Lobophora). The colour bar scale represents the density of each distribution, with reds indicating highest densities, whites representing zeros and greens indicating negative densities as can be seen in C. A) Densities of all microhabitat sizes measured; B) Densities of microhabitat sizes occupied by Lobophora; C) Difference between B) and A). Red tinged values indicate higher densities of microhabitats (A and B) and a higher density of microhabitats supporting Lobophora than overall microhabitats (C). D) The microhabitats which exclude most individuals within a fish species are circled. Colours are according to fish species. The threshold for the lines marking microhabitats excluding fish was set to a density of 0.001 to get the strongest exclusion observed per fish...... 69 Figure 4.4: Lobophora likelihood as a function of grazing pressure. Grazing pressure was calculated by scaling the access by each fish figure with the biomass of that fish species on the reef. Data

xvii displayed is model output data. Points are jittered along y-axis (x-axis differences are true variation) and are raw data. Blue ribbon marks 95% confidence interval...... 70 Figure 4.5: A) The density of Lobophora recruits found in different microhabitat types as estimated by a generalized linear mixed effects model, with an offset included for microhabitat area. Each tile had one crevice type (large crevice, medium crevice or small crevice) and crowns (see Methods: Microhabitat types). Lower case letters indicate significantly different results within herbivory treatments. Upper case letters indicate significant differences of a microhabitat type (according to the number and colouration) between different herbivory treatments. Errorbars are standard error as calculated by the model. B) Mean sediment load estimated. Small letters indicate significantly different results across all panes. Errorbars are standard error...... 72 Figure 4.6: A) Mean size-adjusted bites for all species combined within each microhabitat type observed in video surveys. Error bars symbolise standard error, letters symbolise significantly different results. B) Distribution of bites cm-2 min-1 within crevices by all fish species observed in video surveys. Small crevices, medium crevices and large crevices were each on different tiles. Colours indicate crevice types (and crowns) and images indicate fish species...... 74 Figure 4.7: A) Probability of a fish visiting the tiles taking a bite in different microhabitat types as a response to increasing fish size. Binomial model prediction of raw data. Ribbon signifies the 95% confidence interval. B) The graph shows the mean bite rate compared to body length of each species, including standard deviation...... 75 Figure S4.1: Abundance of blennies, snails, and hermit crabs found on microhabitat tiles within herbivory treatments. Note the different scales. Errorbars show standard error...... 83 Figure 5.1: Conceptual overview of the model showing the main parameters. + indicates an increase in the following parameter, - indicates a decrease. Colours represent different processes………….98 Figure 5.2: Model validation plots. Mean Proportion of microhabitats with Lobophora in surveys (straight line) and 95% confidence intervals (shaded area) of survey data are plotted in red. Burn-in time (month 0 – 20) is cut off. 20 model runs are plotted in black for each microhabitat type...... 106 Figure 5.3: Sensitivity analysis shows deviation of overall Lobophora cover when the input parameter is varied by +20 % (red bars) and by -20% (green bars). Averages from 20 model runs...... 107 Figure 5.4: Lobophora cover over 60 months under different microhabitat distributions. The burn-in time (month 0 – 20) is cut-off. Results are presented for 20 model runs. The red line shows the expected cover if the area additionally assigned to crevices was filled with Lobophora and added to the cover of the in-situ observed microhabitat distribution (first pane). Values below the red line indicate that Lobophora did not occupy all the added crevice surface area. Text inlay shows the overall distribution of all four microhabitat types in each scenario...... 108

xviii Figure 5.5: Changes in % Lobophora cover driven by reductions in herbivorous fish biomass. “All species combined” means that the fish biomass of all species was reduced simultaneously. Dots show simulated values and linear regression lines and 95% confidence intervals are added. Note the different scales. Colours mark different fish species...... 110 Figure 5.6: Changes in % Lobophora cover with decreasing overall herbivorous fish biomass as driven by reductions in either individual groups (Acanthurus spp., Ctenochaetus spp., Zebrasoma spp. or parrotfish) or for all species simultaneously. Biomass was reduced to 100%, 80%, 50%, 20% and 5% of original values for each fish group. Some species had low biomass on the reef, causing the data to be concentrated at the upper end of herbivorous fish biomass on the x-axis...... 111 Figure 5.7: Change in % Lobophora cover in response to changes in parrotfish biomass and microhabitat distribution. Red lines (linear regression) and error margins (95% confidence interval) show results with manipulated microhabitat distribution. Grey shows results for normal microhabitat distribution. Grey dotted lines and open black circles show normal distribution next to manipulated microhabitat distribution to allow better comparison...... 112

xix List of Tables Table 2.1: Lobophora species abundance (total number of samples genetically identified as the respective species) and the number of sites each species was recorded at. Classification into generalist, specialist and local/rare species is shown. This classification is based on the environments they inhabit...... 25 Table 2.2: SIMPER output showing the similarity within exposure groups up to an average similarity of 90 %. G=Generalist; S=Specialist...... 26 Table S3.1: Polynomial generalized linear mixed effects model (glmm) output analyzing Lobophora cover over time. Bold print under p-value symbolizes significant results…………………………..54 Table S3.2: Coxme survival model output analyzing Lobophora adult survival. Bold print under p- value symbolizes significant results...... 54 Table S3.3: Negative binomial generalized linear model (glm) and post-hoc Tukey output analyzing temporal Lobophora recruitment. Bold print under p-value symbolizes significant results...... 55 Table S3.4: Binomial generalized linear mixed effects model (glmm) output analyzing Lobophora recruit removal facilitated by different fish species in the tank experiment. Bold print under p-value symbolizes significant results...... 56 Table S3.5: Binomial generalized linear mixed effects model (glmm) output analyzing Lobophora recruit removal facilitated by different fish species in-situ. Bold print under p-value symbolizes significant results...... 57 Table S4.1: List of fish species and measurements used to build fish shapes for microhabitat survey. These fish shapes were later used to discern the accessibility of microhabitats to the respective fish species. Morphological measurements for these shapes were obtained from Brandl et al. (2015) and missing species were measured using pictures provided by Sonia Bejarano or obtained online (S. vulpinus: www.australienmuseum.net.au) following the same measurement protocol as used in Brandl et al. (2015)...... 80 Table S4.2: Influence of microhabitat width and depth on access by fish groups. Generalized linear mixed effects models with microhabitat width and depth as additive predictors were run. a-level set at 0.05...... 81 Table S4.3: Binomial generalized linear mixed effects model (glmm) output analyzing Lobophora likelihood as a function of potential grazing pressure. Bold print under p-value symbolizes significant results...... 84 Table S4.4: Poisson generalized linear mixed effects model (glmm) output analyzing Lobophora recruit settlement and/or post-settlement survival in microhabitat. Bold print under p-value symbolizes significant results...... 84

xx Table S4.5: Post-hoc Tukey test output analyzing Lobophora recruit settlement and/or post- settlement survival in microhabitat. Bold print under p-value symbolizes significant results...... 86 Table S4.6: Poisson generalized linear mixed effects model (glmm) output analyzing possible confounding factors for Lobophora recruit settlement and/or post-settlement survival in microhabitat. Bold print under p-value symbolizes significant results...... 89 Table S4.7: Binomial generalized linear mixed effects model (glmm) output analyzing influence of fish body size on the ability to access different microhabitat types. Bold print under p-value symbolizes significant results. * symbolizes interactions of fixed factors...... 91 Table 5.1: Parameters used in the model. Details explain how parameters were calculated and mean and standard deviation (if applicable) are displayed...... 100 Table 5.2: Seven scenarios run to explore the influence of changing small scale structural complexity. The scenario name is given to compare the input to the results graph and the % area occupied by each microhabitat type is reported...... 105

xxi List of Abbreviations used in the thesis

AIC: Akaike Information Criteria

ARC: Australian Research Council

CCA: Crustose coralline algae cm: Centimetre

Co.: Company dbRDA: Distance-based redundancy analysis distLM: Distance-based linear model

DNA: Deoxyribonucleic acid

Glmm: generalized linear mixed effects model

GPS: Global positioning system h: Hour

Inc.: Incorporation

IP: Initial phase kg: Kilogram

LTD: Limited company m2: Square meter

µm: Micrometer ml: Maximum likelihood mm: Millimetre msb: Mass-standardised bites

xxii MUSCLE: Multiple sequence comparison by log-expectation

PCoA: Principal coordinate analysis

PERMANOVA: Permutational multivariate analysis of variance

S: Site sd: Standard deviation se: Standard error

SIMPER: Similarity percentage analysis sp.: Species (singular) spp.: Species (plural)

TP: Terminal phase

UQ: The University of Queensland

USA: United States of America vs.: versus

W: Watt

WI: Wisconsin

xxiii

______

CHAPTER 1: GENERAL INTRODUCTION

______

1 General Introduction 1.1 Biodiversity and the degradation of ecosystems Ecosystems are increasingly under pressure from anthropogenic activity, such as deforestation, exploitation of resources, pollution and human-driven climate change (Möllmann et al. 2009; Woodward et al. 2012; Barlow et al. 2016). The degradation of many ecosystems around the world has caused a decrease in habitat and in turn has caused the rapid loss of biodiversity (Tilman et al. 1994). Apart from the loss of whole groups of species, diversity loss can have consequences for the functioning of ecosystems (Symstad and Tilman 2001).

High diversity has long been suggested to buffer ecosystems against degradation due to the higher probability of redundancy amongst species (multiple species perform the same function; Lawton and Brown 1994) but many studies point to high complementarity among species (Tilman et al. 1997; Hector et al. 1999; Loreau et al. 2001). Accordingly, the impact of species loss could have fundamental consequences for whole ecosystems.

1.2 Herbivory and phase shifts on coral reefs Coral reefs are some of the most diverse ecosystems in the world and provide habitat to almost a million species (Fisher et al. 2015) and deliver social, economic and environmental benefits (‘ecosystem services’) to over 500 million people (Bryant et al. 1998). A healthy coral reef provides food, income, and coastal protection to the communities living close to it (Hicks et al. 2015). However, coral reef condition is declining worldwide due to local and global stressors (Hughes 1994; Hoegh-Guldberg 1999; Hoegh-Guldberg et al. 2007; Hughes et al. 2017). Local-scale stressors include overfishing, eutrophication by land run-off, disease outbreaks, and pollution (Hughes et al. 2003; Hoegh-Guldberg et al. 2007). On a global scale, climate change is driving coral reef decline through rising temperatures, ocean acidification and increasing frequency and intensity of natural disturbances, such as storms (Hughes 1994; Hoegh-Guldberg 1999; Hoegh-Guldberg et al. 2007). Local and global stressors can interact and decrease the resilience of coral reefs against acute disturbances, such as temperature-induced bleaching or storms (Carilli et al. 2010; Melbourne- Thomas et al. 2011). As a consequence of more frequent disturbances, mass coral mortality has become increasingly common in recent decades (Hughes et al. 2018).

As benthic marine ecosystems, coral reefs are space-limited (Roughgarden et al. 1985; Muko et al. 2001), making settlement space a valuable resource. Mass coral mortality frees up settlement space, which is quickly colonized by algae (Hughes 1994; Mumby et al. 2005). In contrast to corals, which are long-lived and slow-growing, most algae are short-lived, fast colonizers but their competitiveness 2 is controlled by herbivory (Lewis 1986; Carpenter 1986). Broadly, algae can be grouped into turf algae and macroalgae. Turf algae are small algae, less than 2 cm high and they are composed of fast- growing filamentous species and juvenile macroalgae (Littler and Littler 2011). Macroalgae are larger algae (> 2 cm), which are less readily consumed by herbivores than turf algae (Bennett et al. 2010). Turf and macroalgae are a normal component of coral reef ecosystems but if herbivory by invertebrates, such as sea urchins, or herbivorous fishes is low, algae spread and increasing numbers of mature macroalgae develop out of the turf mats (Bellwood et al. 2006; Poore et al. 2012; Mumby et al. 2016). Macroalgae delay or even prevent coral reef recovery, as they inhibit coral larval settlement (Kuffner et al. 2006; Evensen et al. 2019), hindering coral reef regeneration (Done 1992; Hughes 1994; Connell et al. 1997; McCook 1999). Therefore, reefs subjected to prolonged and/or intense disturbances (e.g., wide-spread bleaching events) can enter an ecosystem state dominated by algae (Hughes 1994; Hoegh-Guldberg et al. 2007; Hughes et al. 2007). Following definitions by Connell and Sousa (1983), we will define a disturbance as an event that can cause a perturbation, which in turn is a change in a variable of interest (e.g. macroalgal dominance). If this perturbation (e.g. the algal-dominated state) persists after the disturbance has receded, a coral reef could have shifted into an alternative stable state of macroalgal dominance. While disturbances can shift an ecosystem into an alternative state, positive feedback mechanisms, such as the inhibition of coral recovery through macroalgal characteristics are needed to maintain a regime shift (Petraitis and Latham 1999). Coral reefs that have undergone such a phase shift (whether stable or transient) provide less habitat (Graham et al. 2007; Alvarez-Filip et al. 2009) and deliver fewer, altered ecosystem services to the local communities compared to healthy reefs (Bryant et al. 1998; Rogers et al. 2015).

If a reef has shifted to a low-coral, algal-dominated state it is difficult to return the reef to a healthy coral-dominated state (Williams et al. 2001; Mumby et al. 2007). This difficulty is due to shifting thresholds (Mumby et al. 2007) and positive feedback loops (Hoey and Bellwood 2011). For example, a given herbivorous fish biomass is sufficient to control algae on a coral-dominated reef. However, once a large number of corals die, more space becomes available and the same biomass of fish now has to distribute its grazing activity over a much larger area available to algae (Williams et al. 2001; Mumby et al. 2007). This causes a decrease in grazing intensity per unit area (Mumby et al. 2007), allowing algae to escape intense grazing. As a consequence, a higher threshold of herbivorous fish biomass is now needed to remove the algae and free up space for coral larvae to settle and grow into adult corals and shift the reef back to coral dominance (Mumby et al. 2007). Feedback loops may further stabilise a macroalgal-dominated ecosystem state by influencing the intensity of herbivory (Hoey and Bellwood 2011). Once macroalgae have established high densities, they are avoided by herbivorous fishes able to remove the algae (Hoey and Bellwood 2011). Because of the ability of

3 macroalgae to deter coral larval recruitment and settlement, these feedbacks can delay or even inhibit reef recovery (Kuffner et al. 2006; Evensen et al. 2019).

The role of herbivorous fishes in regulating the competition between algae and corals makes them vital to coral reef resilience (Obura and Grimsditch 2009). Besides sea urchins (de Ruyter van Steveninck and Breeman 1987a), herbivorous fishes are the main consumers of algae on coral reefs and remove up to 50 – 100 % of algal production (Carpenter 1986). While the exclusion of herbivorous fishes with cages often leads to a dominance of brown macroalgae (Thacker et al. 2001; Bellwood et al. 2006; Burkepile and Hay 2008), the targeted algae vary between fish species (Burkepile and Hay 2008, 2010). For example, herbivorous fishes able to remove turf algae often do not target macroalgae (Green and Bellwood 2009). This is due to the algae’s different defences and the fishes’ varying abilities to deal with these defence mechanisms (Lewis 1985).

Macroalgae utilize a variety of morphological and chemical defence strategies (Paul and Fenical 1983; Paul and Hay 1986). Typical structural deterrents of brown macroalgae – the most abundant macroalgae on reefs – are tough, leathery thalli or calcification (Ogden 1976; Paul and Hay 1986). Chemical defences include the secondary metabolites polyphenolics, such as phlorotannins, and non- polar secondary compounds (Paul and Hay 1986; Steinberg and Paul 1990). These compounds are believed to deter feeding by associating with other molecules, for example through hydrogen bonding (Appel 1993), making the digestion of plant or algae material difficult (Zucker 1983). However, all of these defences seem to vary in their ability to deter herbivores and many algae with these defences are still heavily grazed upon (Lewis 1985). One reason for this may lie in the rather complex interactions between polyphenolics and other molecules (Zucker 1983). These interactions depend on the polyphenolics structure (Zucker 1983), their respective concentrations (Haslam 1974), their molecular size (Boettcher and Targett 1993) and the surrounding pH level (Goldstein and Swain 1965; Stern et al. 1996). Further, the nutritional benefit of macroalgal feeding is limited to fish that can digest the algae’s storage polysaccharides, which are difficult to digest for vertebrates (Saunders and Wiggins 1981).

Herbivorous fishes able to achieve nutrition from brown macroalgae utilise microorganisms to break down their storage polysaccharides mannitol and laminarian into short-chained fatty acids (SCFAs), which are digestible by fishes (Horn 1989; Clements et al. 1994; Seeto et al. 1996). Apart from being able to achieve nutrition from brown macroalgae, fish need to be able to deal with the algae’s structural and chemical defences. For example, parrotfish (Scarinae) readily consume macroalgae with strong structural defences, whereas surgeonfish (Acanthuridae) avoided the same species of alga

4 completely (Lewis 1985). This difference was attributed to the parrotfish’s ability to structurally break down the tough thalli with their strong jaw structure (Lewis 1985). Additionally, the fish’s gut physiology plays a role in their ability to consume macroalgae. A basic gut environment, such as found in some parrotfish species may enable the fishes to be unaffected by high polyphenolic contents (Targett et al. 1995) because a basic pH inhibits hydrogen bonding of the polyphenolics (Appel 1993). However, parrotfish are microphages that achieve their nutrition from microorganisms, such as cyanobacteria (Clements et al. 2016) while consuming varying substrates. Any removal of macroalgae by parrotfish can, therefore, be considered an incidental by-product of their foraging activity.

Feedback mechanisms, which cause the reduction of grazing pressure per unit area and lead to an avoidance of dense macroalgal stands by fishes, and the limited ability of fish to consume mature macroalgae, highlight the importance of management strategies aimed at avoiding shifts to macroalgal-dominance. Restoration of an ecosystem with a trajectory towards a macroalgal dominated state will be much more difficult than management measures initiated at an earlier point in time (Mumby 2009). Therefore, it is important to understand the ecology of macroalgae, how and where they become established, and how they are controlled by herbivores.

1.3 The brown macroalga Lobophora One macroalgal genus particularly harmful to corals (Jompa and McCook 2002a, b; Kuffner et al. 2006; Mumby et al. 2016) is the brown macroalga Lobophora (Dictyotales, Phaeophyceae). Lobophora has a variety of morphologies, ranging from firmly attached (crustose) to thalli only attached at the base (stipitate; Vieira et al. 2014). Lobophora is common on Caribbean and Indo- Pacific coral reefs (van den Hoek et al. 1978; Bennett et al. 2010). It dominates many Caribbean reefs after the sudden mass mortality of the sea urchin Diadema sp. (de Ruyter van Steveninck and Breeman 1987b; McClanahan and Muthiga 1998; Nugues and Bak 2008; Fricke et al. 2011), one of the most important herbivores in the region. Although Lobophora is not as dominant on Indo-Pacific reefs as it is on Caribbean reefs, it has been increasingly observed in phase shifts on Indo-Pacific reefs as well (Diaz-Pulido et al. 2009; Cheal et al. 2010; Roff et al. 2015b).

An increase of Lobophora on coral reefs may contribute to reef degradation. Lobophora is a strong inhibitor of coral larval settlement (Kuffner et al. 2006; Evensen et al. 2019), decreases growth of corals (Box and Mumby 2007), reduces coral fecundity (Foster et al. 2008), and damages and overgrows live corals (Jompa and McCook 2002b; Vieira et al. 2015). However, the interaction between corals and Lobophora is not unidirectional. Corals decrease Lobophora growth rates when 5 in close proximity (de Ruyter van Steveninck et al. 1988). The ability of Lobophora to damage live adult corals also seems to be limited to certain coral species and to be facilitated by reduced herbivore access (Rasher and Hay 2010; Vieira et al. 2015). Nevertheless, overall Lobophora seems to be one of the strongest competitors of corals.

Until recently, the type species Lobophora variegata (Lamouroux) Womersely was considered to be pantropical. However, studies have found Lobophora to be much more diverse (Sun et al. 2012; Vieira et al. 2014). Using mitochondrial DNA, the genus is estimated to contain over 100 species and species richness estimators project the diversity to reach up to around 200 species (Vieira et al. 2017). Lobophora also displays high regional diversity, with 29 species reported from New Caledonia alone (Vieira et al. 2014). Contemporary Lobophora species seem to be quite limited in their geographical range (Vieira et al. 2017) and the alga’s dispersal potential is assumed to be limited to short distances (de Ruyter van Steveninck and Breeman 1987b). However, at least historically Lobophora must have been able to disperse over long distances to spread to its current global distribution (Vieira et al. 2017).

Depending on the alga’s geographical location and other unknown factors, seasonality in Lobophora seems to vary. In tropical locations, seasonal influences on abundance were observed both in the Indo-Pacific and in the Caribbean by some authors (Diaz-Pulido et al. 2009; Ferrari et al. 2012b) whereas no seasonal influence on reproduction, growth, size, and abundance was observed by de Ruyter van Steveninck and Breeman (1987). Contrary to tropical Lobophora populations, strong seasonal and year-to-year variation in abundance was observed among populations on its northern distribution limits (Peckol and Searles 1984). Observations of shorter-term population dynamics showed that turn-over rates of individual Lobophora thalli are high, with blades exhibiting half-lives of 15 – 39 days (de Ruyter van Steveninck and Breeman 1987a). Lobophora was able to recolonize cleared patches within 5 –7 months (de Ruyter van Steveninck and Breeman 1987b) and 12– 60% of patches were turned-over per year (Mumby et al. 2005).

Lobophora grows in a variety of different habitats, from deeper reefs (de Ruyter van Steveninck and Breeman 1987b), where they form distinct belts (van den Hoek et al. 1978), up to exposed reef flats (Diaz-Pulido et al. 2009; Bennett et al. 2010). However, the alga is typically associated with concealed microhabitats (Bennett et al. 2010). The importance of microhabitats in shaping terrestrial and aquatic communities is well established (Bazzaz 1975; Menge and Lubchenco 1981; Tews et al. 2004; Poray and Carpenter 2014; Brandl and Bellwood 2016; Doropoulos et al. 2016a) and is related to structural protection from grazing thus facilitating the survival of juvenile plants (Baraza et al.

6 2006) and algae (Dudley and D’Antonio 1991; Bergey 2005; Loffler and Hoey 2019). On coral reefs, for example, coral larvae preferentially recruit to the underside of settlement tiles or in crevices (Edmunds et al. 2014; Doropoulos et al. 2016a). This preference has been linked to corallivory and grazing avoidance (Menge and Lubchenco 1981; Lubchenco 1983; Doropoulos et al. 2016a), which can either target or incidentally ingest coral recruits, respectively (Box and Mumby 2007; Doropoulos et al. 2012). Contrary to corals, macroalgae settle on exposed and protected surfaces equally (Poray and Carpenter 2014), but their distribution is linked to crevices (Menge and Lubchenco 1981; Lubchenco 1983) because of high predation on exposed algal recruits (Lubchenco and Gaines 1981; Poray and Carpenter 2014). The protection from herbivory decreases grazing pressure in microhabitats at the base of branching corals by 1 – 2 orders of magnitude (Bennett et al. 2010). As a consequence, distinct benthic communities exist in concealed microhabitats vs. open surfaces on coral reefs (Bennett et al. 2010; Brandl and Bellwood 2016).

Herbivory is considered one of the most important controlling factors of Lobophora abundance (de Ruyter van Steveninck and Breeman 1987a; Jompa and McCook 2002a; Ferrari et al. 2012b) and herbivores have been observed to consume Lobophora thalli (Lewis 1985; Pillans et al. 2004; Slattery and Lesser 2014). However, limited grazing on Lobophora either in general (Hay 1981), compared to feeding rates on other macroalgal species (Steinberg and Paul 1990; Bennett et al. 2010) or between herbivore groups (only grazing by urchins is important; Morrison 1988) has been observed as well. Further, the identity of fish species controlling Lobophora is largely unclear. The controversy about susceptibility to herbivory extends to chemical defences detected in Lobophora. Anti-herbivore substances have been identified by some authors (Paul and Hay 1986; Arnold et al. 1995), whereas others report no chemical defences (Lewis 1985; Slattery and Lesser 2014). Additionally, polyphenolic content varies even between individual thalli (Steinberg and Paul 1990). These contradictions may be explained by the high taxonomic diversity of Lobophora (Vieira et al. 2017), induced chemical resistance as a response to direct herbivore damage (Weidner et al. 2004), or variation of polyphenolic content with nitrogen availability (Arnold et al. 1995). As described earlier, the deterring effects of polyphenolics also depend on characteristics of the phenolics and the gut physiology of the respective herbivore (Zucker 1983; Boettcher and Targett 1993; Targett et al. 1995). Even though some thalli may be readily consumed in certain regions, broad evidence suggests a high level of resistance of Lobophora against herbivory (Hay 1981; Morrison 1988; Steinberg and Paul 1990; Bennett et al. 2010). Observations of adult thalli removal are rare, especially in the Indo-Pacific and it has been suggested that Lobophora is mainly controlled during the propagule stage, whereas herbivory on dense patches of adult thalli may be limited (Diaz-Pulido and McCook 2003). If

7 Lobophora is indeed controlled more easily during the propagule life-stage, management strategies applied after a Lobophora outbreak may have limited effects once the alga reaches adulthood.

Overall, studies investigating the ecology of Lobophora are rare and the existing records are exclusively on Caribbean reefs, which may well represent species not found in the Indo-Pacific (Vieira et al. 2016a). In contrast, research on Indo-Pacific Lobophora species is scarce but multiple more recent records of Lobophora dominance on Indo-Pacific coral reefs highlight the need to increase research efforts in the region. Further, due to the relatively recent investigations of taxonomic diversity of Lobophora, there is a lack of knowledge about species identities and distributions. Most studies do not report a species name or refer to the study species as Lobophora variegata. The problem could go as far as having used different species in the same study without knowledge thereof since species can look morphologically similar to the naked eye (Sun et al. 2012). This could at least partly explain the many contradicting reports on Lobophora palatability (Hay 1981; de Ruyter van Steveninck and Breeman 1987a; Jompa and McCook 2002a) and defences (Lewis 1985; Paul and Hay 1986; Arnold et al. 1995; Slattery and Lesser 2014). Accordingly, more research is needed to avoid unexpected shifts to Lobophora dominance on coral reefs.

1.4 Thesis aims Coral mass mortality is increasingly frequent (Hughes et al. 2018) and phase shifts to macroalgae dominance can occur (Graham et al. 2015). However, returning a reef to coral-dominance after such a shift is much more difficult (Mumby 2009) than building the resistance of reefs against disturbances which cause coral mass mortality. Therefore, it is important to understand macroalgal ecology. The known negative impact of Lobophora on coral recovery (Kuffner et al. 2006; Box and Mumby 2007; Evensen et al. 2019) makes it essential to understand what could drive Lobophora outbreaks if we are to predict and manage reefs with regard to future disturbance events. In the Indo-Pacific especially, we lack fundamental knowledge about the drivers of Lobophora species assemblages and the control of Lobophora by herbivores. This thesis aims to address these issues by investigating taxonomic diversity, life-stage specific control by herbivorous fishes, and the importance of microhabitats for Lobophora outbreaks in Palau. Specifically, I aim to determine:

1) The influence of wave exposure and herbivory on the taxonomic diversity of Lobophora and assemblages. 2) The importance of life-stage specific control driven by herbivorous fishes. 3) The role of different microhabitats in providing a refuge to Lobophora from grazing.

8 4) The impact of herbivore fisheries on a reef with a varying number and size of crevices on Lobophora trajectories using an individual-based model.

This research will increase our knowledge of the ecology and diversity of a pervasive macroalga and identify important environmental drivers of Lobophora distribution. Further, I will study whether there is an ontogenetic shift in the alga’s grazing susceptibility and identify fish species that are important to control it. Investigating the influence of microhabitats will increase our understanding of whether Lobophora needs structural protection to become established and why some reefs may be more prone to shifts to Lobophora dominance than others. These findings could help inform fisheries management as to which fish are particularly important to control macroalgal outbreaks and whether management interventions are more effective if implemented before or after a phase shift to macroalgal dominance.

1.5 Thesis structure Chapter 2: Cryptic diversity in the macroalgal genus Lobophora reveals environmental drivers of algal assemblages The macroalgal genus Lobophora has long been assumed to be made up of few species, with one species, Lobophora variegata, being particularly common. However, studies have found the influence of Lobophora on coral larvae to differ. This may be due to a high diversity within this genus. This chapter investigates the cryptic diversity of Lobophora in Palau and the influence of wave exposure, herbivorous fish biomass and depth on Lobophora species assemblages. Lobophora thalli were sampled at 12 sites along a wave exposure and herbivorous fish biomass gradient at 3 m and 10 m for 35 minutes each. Understanding the taxonomic diversity of Lobophora and whether there are differences between species is important as different species may have different competitive abilities with corals.

Chapter 3: Species-specific effects of herbivorous fishes on the establishment of the macroalga Lobophora on coral reefs Once macroalgae are established, they can inhibit coral reef recovery and the algae’s removal is limited to few fish species. The elimination of macroalgal recruits is therefore likely important in preventing macroalgal blooms. This chapter quantified the ability of four fish species, Acanthurus nigrofuscus, Ctenochaetus striatus, Zebrasoma scopas, and Chlorurus spilurus to remove Lobophora recruits using a tank experiment. These results were compared to in-situ observations of fish feeding and recruit mortality. Further, the survival of adult Lobophora was tracked for nine weeks on the reef

9 slope. The results highlight the role of nominally grazing fishes in preventing Lobophora blooms through their removal of macroalgal recruits.

Chapter 4: Refuge dependent herbivory controls a key macroalga on coral reefs Structural refuges shape consumer-producer interactions and can thus influence the ability of herbivorous fishes to control macroalgae on coral reefs. This chapter combined surveys and manipulative field experiments to test the influence of crevices on the survival of Lobophora recruits. A microhabitat survey recorded the distribution of Lobophora within microhabitats and the accessibility of these microhabitats to different fish species. Manipulated microhabitat tiles were used to investigate the effect of crevice sizes on the survival of Lobophora recruits. This chapter emphasises the importance of structural refuges for the establishment of macroalgae.

Chapter 5: Influence of overfishing and structural complexity on Lobophora cover: an individual- based modelling approach In Chapter 5, an individual-based model was developed to investigate Lobophora trajectories using data from chapters 3 and 4 to parameterise the model. The model uses life-stage specific grazing vulnerability, fish species-specific mortality and microhabitat influence on grazing pressure to predict Lobophora cover. Herbivorous fish biomass and microhabitat distribution were varied to analyse the impact of these two factors on Lobophora cover. This chapter increases our knowledge about which fish species are particularly important to control Lobophora and how microhabitat distribution influences macroalgal cover trajectories.

Chapter 6: General Discussion In Chapter 6 the major findings of this thesis are integrated and discussed.

______

CHAPTER 2: CRYPTIC DIVERSITY IN THE MACROALGAL GENUS LOBOPHORA REVEALS ENVIRONMENTAL DRIVERS OF ALGAL ASSEMBLAGES ______

11 Cryptic diversity in the macroalgal genus Lobophora reveals environmental drivers of algal assemblages

2.1 Abstract Coral reefs around the world are threatened by local stressors and mass mortality following thermal stress events. Increasingly, coral reefs are shifting towards macroalgal dominated states, which has implications for ecosystem function. In the Pacific, phase shifts are often dominated by the common macroalgal genus Lobophora. While many ecological studies report a single species (L. variegata), recent genetic advances have documented a global cryptic diversity of over 100 species of Lobophora. Here, we sampled Lobophora thalli from twelve reefs around Palau at varying depths, wave exposure and herbivore biomass. Using one mitochondrial marker (cox3) and two chloroplast markers (psbA, rbcL), we uncovered a striking cryptic diversity of 15 species, including 10 previously undescribed taxa. On an island-wide geographic scale, Lobophora assemblages were not related to depth or herbivore biomass but were significantly associated with the degree of wave exposure. On a smaller geographic scale on the Palauan East coast, all these factors influenced Lobophora assemblages. Widespread sampling indicates the presence of generalist species, which were found in almost all sampled habitats, and specialists, which occupy specific microhabitats. Such high levels of cryptic diversity have important implications for ecological studies, as algae commonly identified as L. variegata using field identification may be multiple species. Differences in microhabitats inhabited, dominant growth form (crustose – decumbent) and a 3-fold difference in thalli thickness suggest functional differences among cryptic species. If these differences manifest in competitive strengths, these results have broad implications for our understanding of coral reef recovery following perturbation.

2.2 Introduction Traditionally, species delimitation has heavily relied on morphological differences among species. While this is an important tool for the identification of species, recent advances in molecular methods have brought to light a large cryptic species diversity, which had previously been overlooked (reviewed by Bickford et al. 2007; Leliaert et al. 2014; Leray and Knowlton 2016). Importantly, there are often biological differences among cryptic species and their identification can reveal critical differences in the ecology of entire groups (e.g., Wellborn and Cothran 2004; Meleg et al. 2013; Fišer et al. 2018). For example, many insects previously assumed to be generalist plant herbivores are now known to be cryptic species complexes which are highly specialised (Hebert et al. 2004; Blair et al. 2005; Stireman III et al. 2005). Delineating cryptic species could thus have important consequences

12 for conservation planning and biological control (Bickford et al. 2007). Traditionally, most research investigating cryptic diversity has focused on animals in the less diverse temperate regions (Bickford et al. 2007). However, more recent research has started to address marine cryptic diversity (Vieira et al. 2014, 2017; Camacho et al. 2019).

On coral reefs, for example, DNA sequencing has already revealed cryptic diversity within corals (Ladner and Palumbi 2012; Warner et al. 2015) and macroalgae (Sun et al. 2012; Vieira et al. 2014, 2016a). As sessile benthic organisms, corals – the primary reef-builders – and macroalgae often compete for space (Birrell et al. 2008a; Rasher and Hay 2010). Reefs are increasingly under pressure due to cumulative impacts of climate change and local anthropogenic pressures, such as fishing and nutrient run-off (Fabricius 2005; Hoegh-Guldberg et al. 2007; Hughes et al. 2017). Reductions in grazing pressure, as caused by mass coral mortality or removal of herbivorous fishes can lead to macroalgal dominance (Thacker et al. 2001; Williams et al. 2001; Burkepile and Hay 2008; Mumby 2009). Once established, some macroalgae can inhibit coral reef recovery because they reduce coral larval settlement and increase post-settlement mortality (Kuffner et al. 2006; Mumby et al. 2016). If algae are not controlled by herbivores, a reef might get locked in a macroalgal dominated stable state (Mumby et al. 2013). Reefs that have undergone such a phase shift to macroalgal dominance are likely to provide different and fewer ecosystem services than healthy, coral-dominated reefs (Bryant et al. 1998; Rogers et al. 2015; Williams et al. 2019).

A common macroalga involved in phase shifts is the brown macroalgal genus Lobophora (Dictyotales, Phaeophyceae; Done 1992; Hughes 1994; McCook et al. 2001; Vieira 2019). Lobophora has an encrusting to foliose morphology (Vieira et al. 2014) and can be found globally. Many ecological studies report Lobophora variegata as the species studied and find varying effects of L. variegata on corals (Birrell et al. 2008b; Morrow et al. 2016). Most often, Lobophora is a strong negative competitor of corals (e.g., Jompa and McCook 2002; Foster et al. 2008; Mumby et al. 2016) and sponges (González-Rivero et al. 2012, 2016). Indeed, Lobophora sp. has surprising strong effects on coral recruitment, where even small amounts of the alga reduce coral larval settlement considerably (Evensen et al. 2019), making it an important interactor in coral reef recovery dynamics. In contrast, some studies observing L. variegata have observed positive effects on coral settlement or have found only small effects on live corals in-situ (Birrell et al. 2008b; Vieira et al. 2016c). These contradictory observations of Lobophora-coral interactions have been shown within a geographic region (Birrell et al. 2008b; Morrow et al. 2016; Evensen et al. 2019). Observations of L. variegata grazing susceptibility also vary considerably (e.g., Hay 1981; de Ruyter van Steveninck and Breeman 1987a; Steinberg and Paul 1990). These variations have been explained by differences in secondary

13 metabolites (Coen and Tanner 1989), location (Hay 1981), nutrient status (Pillans et al. 2004), previous grazing (Weidner et al. 2004) and morphology (Coen and Tanner 1989). Recent studies, however, have highlighted a large global cryptic diversity of over 100 species (Sun et al. 2012; Vieira et al. 2014, 2017; Schultz et al. 2015), which could well account for differences among study results. A large functional cryptic diversity of Lobophora has broad implications for ecological studies aiming to investigate coral reef recovery in the presence of this alga.

Ecological and environmental drivers of Lobophora species assemblages may be important predictors of Lobophora-coral interactions and understanding them could be helpful to guide future studies. Wave exposure and grazing pressure by herbivorous fishes have been identified as drivers of algal productivity and abundance (e.g., Sammarco 1983; Sangil et al. 2011; Pedersen et al. 2012; Marshell and Mumby 2015; Roff et al. 2019). Wave exposure has been shown to have a humpbacked relationship with algal diversity (Kautsky and Kautsky 1989) and herbivore pressure could influence algal diversity in a variety of ways (Kautsky and Kautsky 1989). Both wave exposure and herbivore biomass could thus be expected to influence algal assemblages. Lobophora also occurs from exposed reef flats down to the lower limits of the photic zone at 120 m depth (de Ruyter van Steveninck and Breeman 1987b; Diaz-Pulido et al. 2009; Bennett et al. 2010). Given the high diversity of Lobophora, it is likely that depth also plays a role for taxonomic diversity and composition. Further, identifying common/generalist species may point to species that pose a larger risk of spreading to other habitats. Generalists could thus be more likely to cause shifts from coral to Lobophora dominance following disturbances on reefs (Vieira 2019).

Here, we aim to determine the local diversity of the brown macroalgal genus Lobophora in a high diversity ecosystem of the Tropical West Pacific: Palau, Micronesia. Following catastrophic disturbance related to super-typhoon Bopha in 2012, the eastern fore-reefs of Palau underwent a widespread macroalgal phase shift that transitioned to Lobophora (Roff et al. 2015a, b) with strong negative effects on coral recovery (Doropoulos et al. 2014). Prior to disturbance, Lobophora was apparently absent in visual surveys, but the experimental exclusion of herbivores allowed marked increases in Lobophora abundance (Doropoulos et al. 2016b). Despite its widespread nature and the potential impacts of Lobophora on Pacific reefs (Mumby et al. 2016), remarkably little is known about the drivers and the ecological implications of the alga’s diversity, with ecological studies usually treating it as a single species (e.g., Coen and Tanner 1989; Eich et al. 2019). Through field surveys and molecular analysis, we identify species assemblages of Lobophora within forereefs throughout the Palau archipelago and investigate the role of wave exposure, herbivore biomass and depth in driving assemblage distribution. We also identify several putative new species.

2.3 Materials & Methods 2.3.1 Study site The study was conducted in the Pacific island nation of Palau. Twelve sites were chosen to represent a wave exposure and herbivory gradient (Figure 2.1). Each site was sampled on the forereef at depths of 10 m and 3 m. Sites were grouped according to wave exposure ranging from low (0.25 – 0.36 W m-2, n = 5), to medium (0.65 – 0.75 W m-2, n = 4) to high (1.1 – 1.59 W m-2, n = 4). Herbivore biomass was also categorized with four sites having low biomass (0.9 – 1.6 kg × 120 m-2), five with medium biomass (1.8 – 2.7 kg × 120 m-2) and three with high biomass (8 – 11 kg × 120 m-2). Additionally, two samples were taken in the backreef of Site 2 and an adjacent lagoon, which were not included within the main community analysis but were identified and categorized in the same way.

The nine Eastern sites (S1 – S9, Figure 2.1) were also separately categorized as follows: two low exposure Eastern sites (0.25 – 0.33 W m-2), three medium exposure Eastern sites (0.65 – 0.75 W m- 2) and four high exposure Eastern sites (1.1 – 1.59 W m-2). Herbivore biomass of the eastern sites was categorized as low herbivore biomass (0.9 – 1.4 kg × 120 m-2, n=3), medium herbivore biomass (1.6 – 1.9 kg × 120 m-2, n=3) and high herbivore biomass (2.4 – 2.7 kg × 120 m-2, n=3).

Exposure measures were obtained using a wave-theory approach in a Geographic Information System (GIS) at a spatial resolution of 100 m, integrating information on wind vectors and the configuration of the coastline and reef crest (Chollett and Mumby 2012). Here, exposure is calculated using the distance the wind blows over open sea to generate waves (fetch) while taking into account the direction and strength of the winds. Accordingly, exposure excludes the effects of swell or tide. Daily wind speed and direction used in this approach were obtained from Darksky (www.darksky.org) for Koror, Palau, between 17th March 2016 – 7th April 2018, i.e. the two years before the specimen collection using the R package darksky (https://CRAN.R-project.org/package=darksky). Wind weighted fetch (using wind direction, wind speed and fetch length) was calculated using the package haidawave (https://github.com/sebdalgarno/ haidawave/) in R to map the exposure around Palau. The two-year period was compared to an eight-year period to confirm that these values are representative. Further, wave energy flux (W m-2) was calculated for each of the twelve sampling sites using the package waver on R (https://CRAN.R-project.org/package=waver).

Herbivore biomass was surveyed by P.J. Mumby using six 30 m × 4 m transects per site. The surveys included roving herbivorous fishes, including Siganidae, Acanthuridae and Scarinae, which were identified to species level and length was estimated. Fish species, such as parrotfish, were considered 15 herbivores because they cause disturbance to algae (Roff et al. 2015b; sensu Steneck et al. 2017), even if they do not target the algae for nutrition (Clements et al. 2016). Biomass was calculated using published length-weight data on fishbase (www.fishbase.org). Herbivore biomass data were available for each site, with some surveys only conducted up to 2012 while others were conducted until 2018. We used the most recent survey result available for each site. The variation among years at a specific site was not larger than the variation between replicates at a site within one year.

Figure 2.1: Map and location of sites used in Lobophora sampling around Palau. Wave exposure (W m-2) gradient is shown on the border of the reef and the twelve sites are marked (S1 – S12).

2.3.2 Lobophora sampling design All sites were sampled between 18th March and 7th April 2018. At each depth, the substrate was searched for Lobophora thalli in exposed (open reef areas accessible from most directions) and cryptic (under ledges and within crevices accessible from a maximum of two directions) habitats. After sampling up to two thalli at one place, the divers moved three fin kicks along the reef outline before sampling more thalli to avoid re-sampling a big patch of one species. Each depth was sampled for 35 minutes. Notes were taken describing the habitat of each Lobophora sampled (exposed or concealed), the surface shape the alga was attached to (concave, convex or flat) and a description of the morphology of the alga. Morphologies include crustose, procumbent, conk-like, decumbent, anastomising, fasciculate and stipitate (following Vieira et al. 2014).

Samples were transported back to the research station and processed within 24 h. Until processing, they were kept in a flow-through water system. Two samples were dried using silica gel for later DNA extraction. One additional sample of each algal specimen was dried to make a herbarium specimen, which was also used for histological analysis of each species.

2.3.3 Genetic and histological analysis Cetyltrimethyl ammonium bromide-extraction methods were used to extract total genomic DNA (De Clerck et al. 2006). Afterwards, genomic DNA was purified with a Wizard DNA Clean-Up System

(Promega Inc., Madison, WI, USA) following instructions provided by the manufacturer. Two chloroplast genes (psbA, rbcL) and one mitochondrial gene (cox3) were used to generate sequences following Vieira et al. (2014). The generated sequences were added to a dataset, for each marker, composed of representative sequences of all known Lobophora species downloaded from GenBank. Sequences were aligned using MUSCLE (Edgar 2004) implemented in eBioX 1.5.1 (www.ebioinformatics.org). Phylogenetic trees were reconstructed for each marker separately and concatenated using Bayesian inference and Maximum Likelihood analyses under MrBayes and RAxML, respectively, following Vieira et al. (2016a).

External morphology was categorized as described above. Internal morphological analysis was conducted by manually cutting longitudinal and transversal sections of the thallus using a razor blade. A digital camera (Olympus Camedia C-5050 5.0 Megapixel, Olympus Optical Co., LTD, Tokyo, Japan) attached to a compound microscope (Olympus BH-2, Olympus Optical Co., LTD, Tokyo, Japan) was used to take photographs. Dorsal and ventral cortical cells, as well as medulla cells, were counted and measured. No histological analysis was conducted on the two species collected from a close-by lagoon and the backreef of Site 2 and on another species (L. sp. 107) because there was insufficient material.

2.3.4 Statistical analyses A first analysis took a coarse-scale view of the number of species among sites and in this case pooled data from both depths per site. A univariate PERMANOVA was run using a Euclidian distance dissimilarity matrix because diagnostics plots showed a bad model fit for parametric models. Lobophora species number was set as the response variable and wave exposure (low vs. medium vs. high) and herbivore biomass (low vs. medium vs. high) as categorical predictors.

17 A second, more detailed analysis explored the multivariate structure of Lobophora assemblages using the Jaccard dissimilarity matrix. A distance-based linear model (distLM) was run to analyse similarities in the Lobophora species assemblage (recorded as presence/absence of each species), using wave exposure, herbivore biomass and depth as numerical predictors. Significant predictors were then examined further using Similarity Percentage analysis (SIMPER) to compare groups. To understand the separate influence of the two most common herbivore groups, scarids and acanthurids, on Lobophora assemblages, a further distLM was run using scarid biomass, acanthurid biomass, exposure and depth (3 m and 10 m) as numerical predictors.

The distance-based linear model was repeated including only the nine Eastern sites (excluding the three Western sites) to analyse smaller scale geographic patterns. Jaccard dissimilarity matrix was created and the similarities of Lobophora species assemblages were analysed using wave exposure, herbivore biomass and depth as numerical predictors. An additional distLM was run on the Eastern sites to analyse the separate influence of scarids and acanthurids on Lobophora assemblages as described above.

Lobophora species were categorized based on their distribution and their presence at all 12 sites with varying herbivore biomass, wave exposure and depths. Species were considered as local (only at one site), specialist (≥2 sites, restricted to one level in at least one of the three categories) or generalist (at least at two levels out of each category). Generalists were further split into groups of common generalists (at over half of all sites) and less common generalists (at less than half of all sites).

To test for morphological and habitat differences among species and create clusters based on algal characteristics, a Principal Co-ordinate Analysis (PCoA) based on gower distance (which allows using both categorical and numerical values) was run, using morphology, microhabitat, thallus thickness, thallus length and thallus width as predictors. These predictors were fitted using ‘envfit’ in the R package vegan (Oksanen et al. 2019, version 2.5). A hierarchical cluster analysis using ‘hclust’ in the R package ‘stats’ with ‘ward.D’ agglomeration was used to analyse similarities among species.

2.4 Results 2.4.1 Taxonomy A total of 172 specimens were sampled across environments and successfully sequenced. Field observations indicate that specimens were visually similar, with small thalli, brown pigmentation and either tightly encrusting the reef substrate (crustose) or attached at the basal part only but growing close to the substratum (procumbent). Despite apparent visual similarity among samples, genetic 18 analyses revealed a high genetic diversity of Lobophora comprising 15 species (Figure 2.2). Five of these species (L. boudeuseae, L. providenceae, L. gibbera, L. asiatica, L. endeavoriae) have been previously described, while ten species were undescribed and assigned the following codes: Lobophora sp. 10, Lobophora sp. 48, Lobophora sp. 82, Lobophora sp. 105, Lobophora sp. 106, Lobophora sp. 107, Lobophora sp. 108, Lobophora sp. 109, Lobophora sp. 111, Lobophora sp. 112. Placing these species in a phylogenetic tree with currently identified Lobophora spp. indicates phylogenetic clustering among some species (Figure 2.2). Histological analysis revealed similarities among species at a cellular level (Figure 2.3). In general, cell layers for all species were between 3 - 5 cell layers and thallus thickness ranged from 49.0 to 84.8 µm, except for L. gibbera and L. sp. 48, which had 7 – 9 cell layers and were two to three times thicker (158.4 and 154.0 µm respectively).

Figure 2.2: Phylogenetic tree of identified Lobophora species. The tree was created using one mitochondrial (cox3) and two chloroplast (psbA, rbcL) markers. Species found in Palau are highlighted in yellow, all other species have previously been collected and added to a Lobophora database.

Figure 2.3: Transversal cut and in-situ pictures of the 12 Lobophora species with enough material to allow histological analysis. A) L. sp. 10, B) L. sp. 48, C) L. sp. 109, D) L. asiatica, E) L. boudeuseae, F) L. gibbera, G) L. providenceae, H) L. sp. 108, I) L. sp. 106, J) L. sp. 111, K) L. sp. 105, L) L. sp. 112.

2.4.2 Environmental drivers and distribution Wave exposure and herbivore biomass varied across the study sites, with wave exposure ranging between 0.25 – 1.59 W m-2, and herbivore biomass varying between 0.9 - 11 kg per 120 m2. Despite such strong environmental gradients, Lobophora species richness was not influenced by either wave exposure or herbivore biomass (PERMANOVA, df = 2, F = 2.7878, p = 0.13 and df = 2, F = 0.58657, p = 0.59, respectively). Out of the three environmental predictors, Lobophora assemblages were influenced only by wave exposure (distLM, SS(trace) = 5265.9, F = 2.3682, p = 0.03, Figure 2.4) on an island-wide geographic scale. The assemblages did not differ between 3 m and 10 m depth 21 (distLM, SS(trace) = 4116, F = 1.8085, p = 0.09), nor with herbivore biomass (distLM, SS(trace) = 2392.7, F = 1.0164, p = 0.42). Correspondingly, the largest variation was explained by wave exposure (10%), followed by depth (8%), whereas herbivore biomass explained the least variation (4%). Distance-based redundancy analysis (dbRDA) ordination indicated two overlapping clusters among exposure categories at the 45% level (Figure 2.4), whereas dbRDA vectors indicate an increasing wave exposure on the first axis (explaining 15.1% of the variation). Sites 10-12 on the western side clustered more closely with higher wave exposure than other similar lower exposure sites (Figure 2.4). Sites clustered by depth on the second axis (explaining 8.1% of the variation), with the western sites again forming an exception, but depth was not a significant driver of assemblages (distLM, SS(trace) = 4116, F = 1.8085, p = 0.09; Figure 2.4). Herbivore biomass varied across sites, with scarids representing on average three times higher biomass than acanthurids. However, separating herbivorous fish biomass into these two groups did not lead to any significant influence on Lobophora assemblages by either on an island-wide scale (distLM, parrotfish: SS(trace) = 1817.4, F = 1.1693, p = 0.35 and surgeonfish: SS(trace) = 492.73, F = 0.30518, p = 0.85).

Figure 2.4: DbRDA plot of the Lobophora species assemblage showing wave exposure categories (low, medium, high) and clusters based on 45% similarity; the sites are identified as S1 – S12 following the numbering system on the small map.

As the three western sites showed differences to all other sites, a second distLM analysis exclusively on the Eastern sites revealed that on this smaller geographic region, wave exposure (distLM, SS(trace) = 7696.5, F = 3.7771, p < 0.01), herbivore biomass (distLM, SS(trace) = 6841.6, F = 3.2718, p < 0.01) and depth (distLM, SS(trace) = 5731.4, F = 2.6528, p < 0.05) all explained some of the variation within Lobophora assemblages (Figure 2.5). This analysis also explained more of the overall variation (43%) than the larger scale analysis including the western sites (29%). Here, wave exposure explained 19.1%, herbivore biomass explained 17% and depth explained 14.2% of the Lobophora assemblage structure. On this smaller scale, separating parrotfish from surgeonfish did lead to a significant influence of parrotfish biomass only (distLM, SS(trace) = 6307, F = 2.9687, p < 0.01), which explained 15% of the variation observed. Surgeonfish biomass did not help explain Lobophora assemblages (distLM, SS(trace) = 2360.1, F = 0.99533, p = 0.44).

Figure 2.5: DbRDA plot of the Lobophora species assemblages on the East coast of Palau only, showing A) wave exposure categories (low, medium, high) and B) herbivore biomass categories (low, medium, high) as used to classify the Eastern sites only. Symbol shape represents depth. Clusters are based on 50% similarity; the sites are identified as S1 – S9 following the numbering system on the small map.

2.4.3 Generalists vs. specialists The distribution of Lobophora species varied across our 12 study sites. Some species were observed across a wide range of wave exposure and herbivore biomass, while others were restricted in distribution and found either at a single site or within a specific wave exposure regime. Based on 24 these distributions, the species were assigned to four groups: (1) common generalists that were found in more than half of all study sites and in most environments (e.g., low to high wave exposure and herbivore biomass, both depths); (2) less common generalists, which were also present in most environments, but in less than half of all sites; (3) specialists were restricted in the environments they occupied and (4) local species were found at only one site (Figure 2.6). Generalists were more abundant than specialist species with 4 – 85 genetically identified specimens compared to 3 – 11 specimens of specialist species. Similarly, the number of sites species were recorded in ranged from 3 – 12 for generalists and from 2 – 4 for specialists (Table 2.1). SIMPER analysis showed that generalist species (L. gibbera, L. providenceae, L. sp. 108, L. sp. 109) together explained at least 85% of the similarity among sites within a wave exposure group. The highest average similarity explained by a specialist species was 6% (L. sp. 105, Table 2.2). In contrast, generalist species individually explained up to 89% of similarity among sites (Table 2.2).

Table 2.1: Lobophora species abundance (total number of samples genetically identified as the respective species) and the number of sites each species was recorded at. Classification into generalist, specialist and local/rare species is shown. This classification is based on the environments they inhabit.

Lobophora species Abundance Number of Classification (based on habitat and sites found number of sites found) L. asiatica 1 1 Local / rare L. boudeuseae 3 3 Specialist L. endeavoriae 1 1 Local / rare L. gibbera 85 12 Generalist L. providenceae 14 4 Generalist L. sp. 10 7 4 Specialist L. sp. 48 5 3 Generalist L. sp. 82 2 1 Local / rare L. sp. 105 11 2 Specialist L. sp. 106 3 1 Local / rare L. sp. 107 4 3 Generalist L. sp. 108 17 9 Generalist L. sp. 109 20 6 Generalist L. sp. 111 1 1 Local / rare L. sp. 112 1 1 Local / rare

Table 2.2: SIMPER output showing the similarity within exposure groups up to an average similarity of 90 %. G=Generalist; S=Specialist.

Wave Species Avg. Avg. Sim/SD Contrib% exposure probability similarity category Low L. gibberaG 1.0 44.5 3.8 88.8 L. providenceaeG 0.3 2.6 0.3 5.2 Medium L. gibberaG 0.8 19.8 1.4 51.7 L. providenceaeG 0.5 6.7 0.5 17.4 L. sp. 108G 0.5 6.0 0.5 15.8 L. sp. 105S 0.3 2.2 0.3 5.8 High L. gibbera G 1.0 32.9 5.6 55.5 L. sp. 109 G 0.6 12.1 0.7 20.4 L. sp. 108 G 0.6 10.0 0.7 16.8

2.4.4 Morphological differences within species Hierarchical cluster analysis of morphological traits (thallus thickness, height, width, morphology) and microhabitat revealed three clear groupings of species (Figure 2.6). Thallus thickness (R2 = 0.6058, p < 0.05), height (R2 = 0.6338, p < 0.05), and width (R2 = 0.6338, p < 0.05) as well as morphology (R2 = 0.6088, p < 0.01) contributed to the clustering of Lobophora species, whereas microhabitat did not (R2 = 0.4939, p = 0.19). When comparing these microhabitat- and morphology- based clusters to the categorization into specialist, generalists and local species it became apparent that generalist species did not conform to these clusters. Still, their common traits, such as high frequency and low spatial refugia in depth, herbivore biomass and exposure (which were not included in the clustering analysis) are apparent (Figure 2.6). In contrast, local species (L. sp. 111, L. sp. 112, L. asiatica) showed the clearest clustering based on their morphology, thallus thickness, height and width, which corresponded to their high spatial refugia values and low abundance (Figure 2.6).

Figure 2.6: Lobophora species clustering as determined by morphology, microhabitat, thallus thickness, height and width. Morphology and microhabitat were predictors in the analysis and are presented besides spatial refugia in exposure, herbivory and depth, which were not used as predictors. Additionally, how many sites the species was found in is shown. All Lobophora species are categorized as local, specialist, generalist - less common or generalist - common. The categorization was as follows: Local = only at one site; specialist ≥2 sites, restricted to one level in at least one category (further details as per the nature of specialization in graph); generalist - uncommon = at less than half of all sites, at least two levels out of each category; generalist - common = at over half of all sites, at least two levels out of each category. Species underneath dashed line were excluded from cluster analysis due to missing morphological data.

2.5 Discussion Lobophora is a common pantropical coral reef macroalgal genus, yet recent genetic studies indicate that it is composed of multiple species (Sun et al. 2012; Vieira et al. 2014, 2017; Schultz et al. 2015; Camacho et al. 2019). Sampling of Lobophora sp. across diverse environmental gradients in the Palau

27 archipelago revealed a striking cryptic diversity in this common macroalga. Of 15 Lobophora species identified, 10 species are undescribed. Our results indicate that species richness was not influenced by either depth, wave exposure, or herbivore biomass, yet wave exposure was associated with Lobophora spp. assemblage structure island-wide. On a smaller geographic scale (Eastern sites of Palau only), wave exposure, herbivore biomass and depth were all associated with Lobophora spp. assemblage structure. Morphological analyses and habitat preferences indicated clear differences among species in their microhabitat preferences, dominant growth form, and thallus thickness. However, species clustering based on morphology and microhabitat did not accord with the grouping of species as generalists, specialists, and local species based on the environmental conditions they inhabit, indicating other factors influence the ability of species to tolerate environmental conditions. Generalists (or common species) inhabit many sites and most environmental conditions studied, whereas specialists and local species were either restricted in the environmental conditions they inhabit – depth, level of wave exposure or herbivore biomass – or were only found at one site, respectively. Given their ubiquitous nature, Lobophora species identified as generalists may be particularly likely to spread to multiple reefs and may, therefore, pose a higher risk of causing large- scale phase shifts if they experience favourable conditions.

Many Lobophora species found can be described as cryptic species, because they look morphologically similar and could only be identified genetically. Only two out of the 15 species identified were easy to distinguish in the field: L. sp. 82, which was rust-coloured and found exclusively at the base of Porites cylindrica colonies and L. gibbera, which has a greyish appearance due to dense mats of rhizoids on its dorsal side. At a microscopic level, L. gibbera and L. sp. 48 were two - three times thicker than all other species. Interestingly, both were generalists and L. gibbera was the by far most commonly sampled species around Palau. Other differences found among species, such as different morphologies and microhabitats may also point to differences among species that could influence their ecological function, for example their competitive abilities. Both morphologies and microhabitats (i.e., associational refugia) have been suggested as important regulators of Lobophora grazing (Coen and Tanner 1989; Bennett et al. 2010; Vieira et al. 2019) and species growing outside of cryptic microhabitats may be more resistant to grazing, for example because of their morphology. This large cryptic diversity, which is still below that of other regions (Vieira et al. 2014, 2016a), may explain why previous studies have found differing influences of Lobophora on corals (Kuffner et al. 2006; Nugues and Bak 2006; Mumby et al. 2016). This is especially true if the differences found among species translate into functional differences in the alga’s competitive ability or its influence on coral larval settlement. For example, Birrell et al. (2008) observed an increase in Acropora millepora settlement in response to L. variegata, whereas Morrow et al. (2016) found

28 reduced metamorphoses and increased mortality among A. millepora larvae when exposed to an unidentified Lobophora species and its aqueous and organic crude extracts. These findings suggest a functional difference in competitive strengths among cryptic Lobophora species, which could have broad implications for future studies of Lobophora-coral competition experiments and our understanding of coral reef recovery in general. Depending on which Lobophora species is present on a reef – or can recruit to it – may determine whether a coral reef recovers after a disturbance or whether a macroalgal phase shift will take place.

Out of the three predictors studied, wave exposure was the only one to show some association with Lobophora assemblage structure on a broad, island-wide geographic scale, but it did not influence Lobophora species richness. Wave exposure has previously been shown to influence algal productivity and distribution (Leigh et al. 1987; Gaylord et al. 2002; Renken et al. 2010; Pedersen et al. 2012). However, in a subtropical oceanic archipelago wave exposure explained only a low amount of variation within macroalgal assemblages (Sangil et al. 2011). The productivity of many foliose algae is increased because wave action decreases self-shading (Leigh et al. 1987; Pedersen et al. 2012) and breaks up the benthic boundary layer (Wheeler 1980; Leigh et al. 1987; Hurd et al. 1996). While the latter would certainly apply to Lobophora, the mostly encrusting alga is unlikely to experience reductions in self-shading even in high wave exposure. Further, while macroalgal dispersal is often locally restricted (de Ruyter van Steveninck and Breeman 1987a; Vieira et al. 2017), intense wave energy, for example as generated by storms, can cause dispersal over longer distances (Reed et al. 1988). Accordingly, exposed coasts may be more conducive for long-distance dispersal than protected shorelines, thus influencing algal species assemblages. However, while wave exposure did significantly influence Lobophora assemblages, clustering among exposure groups is not immediately clear and requires a more detailed examination. Interestingly, the western sites clustered with high exposure sites, even though they had low wave exposure. When only the Eastern sites were analysed wave exposure explained a higher percentage of the variation (19.1%) than when including western sites (10%). It is thus possible that other effects of larger-scale biogeography overpower the effects of wave exposure.

Similar to wave exposure, depth did not influence the species richness of Lobophora. Depth did have a marginally significant influence on the island-wide Lobophora assemblages and the sites clustered mostly by depth. The only exception to a clear depth clustering were again the western sites, consolidating that larger-scale biogeography may have a larger influence than local habitat variability. Correspondingly, when only the Eastern sites were analysed, depth did explain Lobophora assemblage structure. In all 12 sites, six species were restricted to either 3 m or 10 m, highlighting

29 that depth can have a strong influence on the distribution of some species but not others. Light availability decreases with depth and UV-radiation and photoinhibition have been shown to impact macroalgal species differently (Bischof et al. 1998; Hanelt 1998; Wiencke and Bischof 2012). Therefore, it is possible that increasing the depth difference at our sites further would reveal a clearer distinction between assemblages on a larger geographic scale and including deeper sites may uncover additional Lobophora species and assemblages.

Herbivorous fish biomass did neither explain Lobophora species richness nor Lobophora assemblage structure among all 12 sites. However, similarly to depth, when only the Eastern sites were analysed, herbivore biomass did influence Lobophora assemblages, highlighting that herbivorous fish biomass can have an influence on a smaller geographic scale. An alga’s grazing susceptibility is influenced by a variety of factors, including but not limited to secondary metabolites (Paul and Hay 1986; Steinberg and Paul 1990), previous grazing damage (Weidner et al. 2004), and morphology (Coen and Tanner 1989). Recent research has shown similar grazing susceptibility among different Lobophora species highlighting the potentially small effect of secondary metabolites on species- specific grazing susceptibility (Vieira et al. 2019). However, the use of feeding assays to suspend Lobophora samples in this study ignores the effects of algal growth habit. Therefore, Lobophora species growing on the reef substratum may have variable grazing susceptibility to fish herbivory because of their different growth habits. In Palau, all but one Lobophora species were predominantly decumbent or tightly attached crusts and are thus likely relatively resistant to feeding by many herbivores (Littler et al. 1983). Therefore, it is possible that the similar morphology and potentially small effect of secondary metabolites on Lobophora grazing susceptibility (Vieira et al. 2019) led to a relatively small influence of herbivore biomass on Lobophora assemblages. Consequently, while herbivorous fishes do seem to explain some variation among Lobophora assemblages on small geographic scales, the effect of herbivorous fish biomass on Lobophora assemblages island-wide may be overpowered by other ecological processes on larger geographic scales, such as Lobophora dispersal patterns.

Traditionally, macroalgal morphology has been suggested to be an adaptation to locally high grazing pressure (Coen and Tanner 1989) and has even been observed to change in response to changes in grazing pressure (Lewis et al. 1987). However, in light of recently identified high cryptic Lobophora diversity (Sun et al. 2012; Vieira et al. 2014), morphological differences may reflect different species and adaptations of Lobophora would thus be on an evolutionary time-scale. In fact, this would indicate that in the case of Lobophora instead of species changing their morphology in response to grazing, the morphology of a species would determine where it can become established. Encrusting

30 forms such as present in Palau have been observed in areas of particularly high grazing pressure (Coen and Tanner 1989) and may thus be particularly resistant to grazing. Some fish groups are more likely to remove these encrusting algae than other groups based on the fishes’ morphology. For example, scraping or excavating herbivorous fishes, such as parrotfish have a jaw morphology, which allows them to scrape off carbonate reef substrate (Bellwood and Choat 1990). When analysing all sites island-wide, we did not find an influence of parrotfish biomass on Lobophora assemblages. However, when analysing only the Eastern sites, Lobophora assemblage structure was explained by parrotfish biomass. This indicates that parrotfish may structure Lobophora assemblages, albeit only on smaller geographic scales. Parrotfish have been identified as microphages, which target microorganisms (Clements et al. 2016). They remove Lobophora (Roff et al. 2015b) while foraging for microorganisms, but this removal is unlikely to be targeted specifically towards Lobophora unless a species-specific association between the alga and microbial assemblages (Vieira et al. 2016b) causes parrotfish to preferentially consume some algal species over others (as yet undescribed). In contrast, we did not find any influence of surgeonfish biomass on Lobophora assemblages, both on a small and a larger geographic scale. This finding is not surprising as they can be expected to have limited impact on crustose algae because of their adaptation to remove turf algae and detritus (Green and Bellwood 2009; Cheal et al. 2012).

Overall, the influence of wave exposure, herbivore biomass, and depth seems to be confounded by other drivers on a larger geographic scale, but all three influence Lobophora assemblages locally. The lack of influence on larger geographic scales may be driven by local dispersal of Lobophora, which is common in macroalgae (de Ruyter van Steveninck and Breeman 1987a; Vieira et al. 2017) or seascape variation, including microhabitat distribution, which may provide different amounts of suitable habitat to Lobophora species. Further study of these important drivers would increase our understanding of algal distribution patterns immensely.

The Lobophora assemblages consisted of species that can be described as generalists, specialists and local species. Generalists were on average more abundant than specialists, which means we could have sampled them more frequently than specialists. An alternative theory is, therefore, that species described here as generalists are the most common species. However, specialist species were found at up to four sites and some generalists were found at as few as three sites. Therefore, some difference in their environmental niche may exist nonetheless. Three common generalists (differing combinations of L. gibbera, L. providenceae, L. sp. 108, L. sp. 109) explained at least 85% of assemblage similarity within wave exposure groups, highlighting their ubiquitous distribution. However, we did not find any similarities in morphology, size or microhabitat for algae categorized

31 as generalists. Other factors not measured here, such as the species’ dispersal potential or chemical profile may play a role in determining their ability to colonize many reef environments. While specialists did not cluster either, local species formed a cluster and all species in this group were crustose with low thallus thickness, highlighting potential common properties within this group.

Generalist species may have a higher potential to cause phase shifts on multiple reefs because of their ability to inhabit a variety of habitats. These species may be able to make use of space availability following coral mortality and spread to neighbouring reefs. However, generalists may also face trade- offs, such as reduced competitive ability or reduced growth rate because they are unlikely to be perfectly adapted to all environments (Gilchrist 1995; Bernays and Funk 1999). While able to spread to many reefs, generalists may thus have to encounter favourable conditions within the newly occupied reefs, such as radically reduced grazing pressure or reduced competition with corals to establish dominance. In contrast, specialists are often expected to have an advantage over generalists because of their good adaptations to the local environment (Jermy 1984). This may allow specialist Lobophora species to form more frequent local blooms. However, these phase shifts will likely be restricted to very few reefs as their ability to establish in neighbouring reefs with different environmental conditions seems limited. Accordingly, the species contributing to a phase shift may vary depending on the local conditions. Still, given that both grazing pressure and reduced competition with corals are a result of coral mass mortality, generalists may find suitable conditions in these impacted reefs to develop blooms.

In previous studies, Lobophora species used in coral-Lobophora interactions were often collected without a species identification, leading to differing outcomes (Kuffner et al. 2006; Birrell et al. 2008b; Morrow et al. 2016), which may be explained by the cryptic diversity of this genus. The presence of cryptic species has implications for understanding the dynamics of macroalgal phase shifts. In Palau, a phase shift following typhoon Bopha in 2012 resulted in a persistent Lobophora bloom of up to 40% cover (Roff et al. 2015b). Given the limited dispersal distance of macroalgae, one of the potential sources of this rapid phase shift was the presence of Lobophora in adjacent channels and backreef habitats (Bozec et al. 2018). Lobophora had been seemingly absent from the site for 20 years prior, while it dominated nearby channels since at least the 1970s (Birkeland et al. 1976). However, our analysis indicates that backreef habitats are an unlikely source of propagules to forereef habitats. Six species were present on the phase-shifted reef (L. gibbera, L. sp. 105, L. boudeuseae, L. sp. 108, L. providenceae, L. sp 107) but the species found in the backreef (L. endeavouriae) and the channel (L. sp. 18) were not found in any of the 12 outer reefs sampled. These results indicate lagoonal and back-reef habitats have distinct Lobophora assemblages compared to

32 outer reefs and this has implications for understanding spatial patterns of macroalgal dispersal on coral reefs. Further, potential functional differences between cryptic species can lead to different ecological conclusions in studies unknowingly using different species. While we did not test functional differences in the competitive abilities of different Lobophora species, differences in the thallus thickness, associational and spatial refugia of highly cryptic species combined with previous divergent results of the outcome of coral-Lobophora interactions indicate that Lobophora species may indeed differ in their ability to outcompete corals. These findings have broad implications for our understanding of coral reef recovery following disturbances. There may thus be limited predictive power of Lobophora-coral competition experiments on coral reef recovery, if rare species found in marginal environments, such as non-coral dominated back reefs, are used. We propose that future competition experiments choose common Lobophora types from areas of interest and, if possible, identify them to species level to allow better comparisons among studies and therefore further our understanding of coral-Lobophora phase shifts.

2.6 Acknowledgements I would like to thank Charlie Wiseman and Alan Kavanagh for their huge amount of help collecting the data for this chapter. I am also grateful to the Palau International Coral Reef Center for providing the necessary space and equipment to conduct our studies. This research was funded by the Winnifred Violet Scott Trust and by ARC grants awarded to PJ Mumby.

______

CHAPTER 3: SPECIES-SPECIFIC EFFECTS OF HERBIVOROUS FISHES ON THE ESTABLISHMENT OF THE MACROALGA LOBOPHORA ON CORAL REEFS ______

Puk LD, Cernohorsky N, Marshell A, Dwyer J, Kennedy Wolfe, Mumby PJ (2020). Species-specific effects of herbivorous fishes on the establishment of the macroalga Lobophora on coral reefs. Marine Ecology Progress Series 637:1-14.

34 Species-specific effects of herbivorous fishes on the establishment of the macroalga Lobophora on coral reefs

3.1 Abstract Herbivory is a key ecosystem function that influences ecosystem trajectories. However, interactions between plants and herbivores are species-specific and change throughout the plant’s lifetime. On coral reefs, herbivorous fishes reduce competition between corals and macroalgae through their grazing activity, thereby regulating the ecosystem state. Grazing vulnerability of marine algae generally decreases with increasing algal size. Therefore, the removal of newly-settled spores by herbivorous fish is likely important in preventing macroalgal blooms and reducing competition with corals. We studied the grazing susceptibility of recruits of the brown macroalga Lobophora to multiple fish species through a combination of feeding observations and manipulative in-situ and ex- situ experiments. Further, we recorded short-term Lobophora growth patterns and adult survival over nine weeks and observed two separate recruitment events. Lobophora recruits were more susceptible to herbivory than adults, likely owing to their smaller size. However, recruit mortality was driven by only three of the studied species: Acanthurus nigrofuscus, Scarus niger and Chlorurus spilurus, whereas other common herbivores did not remove any Lobophora recruits. Our data also suggests variable growth and recruitment among months, which can overwhelm the ability of “herbivorous” fish to control Lobophora during times of increased algal growth. These findings highlight an ontogenetic shift in grazing susceptibility of Lobophora towards higher resistance with age. As such, a decrease in grazing pressure by key fish species controlling Lobophora recruits could permit Lobophora to establish more grazing-tolerant adult populations.

3.2 Introduction Herbivory can exert strong impacts on plant community structure but the effect varies depending on the plant and herbivore species involved (Gruner et al. 2008; Barton and Koricheva 2010). Plants regulate their grazing susceptibility through a variety of defence mechanisms, such as physical defences, for example through the development of tough leaves (Gilbert 1971; Kearsley and Whitham 1989; Loney et al. 2006), or chemical defences such as secondary metabolites that deter herbivores (Feeny 1970; Whittaker and Feeny 1971; McConnell et al. 1982). Grazing susceptibility can change throughout ontogeny (e.g., Cipollini and Redman 1999; Fritz et al. 2001; Goodger et al. 2006), often decreasing with age, as predicted by the ‘growth-differentiation balance’ (Herms and Mattson 1992). In contrast, the ‘plant-age hypothesis’ predicts higher levels of defence in the most vulnerable stages, such as juvenile plants (Bryant et al. 1992; sensu Spiegel and Price 1996). Numerous studies have

35 provided supporting evidence for both hypotheses (e.g., Cipollini and Redman 1999; Fritz et al. 2001; Goodger et al. 2006; Lubchenco 1983), but a meta-analysis of these patterns found little support for decreasing defences in plants as they mature (Barton and Koricheva 2010). Instead, increased consumption of older plants was driven by herbivore species-specific preferences without an associated decrease in plant defences (Barton and Koricheva 2010), highlighting that ontogenetic changes in grazing susceptibility of plants depend on the herbivore species involved.

In tropical marine systems, herbivory is intense and herbivorous fishes play an important role in regulating algae (Carpenter 1986; Polunin and Klumpp 1992). On coral reefs, macroalgae constantly compete with reef-building corals for space, a major limiting resource (Connell et al. 1997; Miller et al. 1999). Reefs dominated by corals provide habitat to numerous species and deliver ecosystem services to millions of people living close to tropical coastlines (Moberg and Folke 1999). However, increasing anthropogenic impacts have shifted the balance in favour of algae, which has caused an increase in macroalgae on many of the world’s reefs (Hughes 1994; McClanahan and Muthiga 1998). It is therefore important to understand potential bottlenecks in the macroalgal life cycle and the role herbivorous coral reef fishes play in preventing macroalgal proliferation. While some notionally herbivorous fishes achieve their nutrition from other organisms, here, we will refer to ‘herbivores’ from an ecological perspective, i.e. if they cause biologically-mediated disturbance to algae (sensu Steneck et al. 2017).

The removal and digestion of adult brown macroalgae, the most common macroalgal group on coral reefs, is difficult and therefore confined to specialised herbivorous fish species ('browsers'; Green and Bellwood 2009). Brown macroalgae store their energy in mannitol and laminarian, two polysaccharides that are difficult to digest by vertebrates (Saunders and Wiggins 1981; Painter T J 1983). There is no evidence of fish being able to produce proteins that would enable them to break down these macroalgal polysaccharides (Clements and Choat 1997). To derive nutrition from brown macroalgae, common ‘browsing’ fish species, such as unicornfishes (nasids) and rudderfishes (kyphosids; Puk et al. 2016), have a hindgut which harbours microorganisms (Horn 1989; Seeto et al. 1996). These microorganisms are capable of fermenting mannitol and laminarin and convert them into short-chained fatty acids (SCFAs), which are digestible by fish (Clements et al. 1994; Seeto et al. 1996). Herbivores, however, need first to deal with the physical removal of macroalgae. Several algal species are readily removed (Mantyka and Bellwood 2007; Fox and Bellwood 2008), while others may present a more challenging resource because of their calcified or encrusting morphology (Paul and Hay 1986; Coen and Tanner 1989). Some fish species, such as parrotfish, are better equipped than others to remove tough or encrusting algae because of their strong jaws and scraping

36 or excavating feeding habit, which allows them to remove parts of the reef substrate, including the endolithic organisms growing within it (Bellwood and Choat 1990). Many brown macroalgae also have high levels of polyphenolics, which can hinder the herbivore’s protein assimilation by forming hydrogen bonds (Stern et al. 1996). Still, some fish species, such as parrotfish, may be unaffected by these polyphenolics because they have a basic gut environment that inhibits hydrogen bonding (Horn 1989; Appel 1993). However, parrotfish have been identified as microphages, which target microorganisms, including cyanobacteria (Clements et al. 2016). Therefore, while parrotfish may incidentally remove macroalgae when foraging for epiphytic microorganisms, they are not expected to target macroalgae (although the genus Sparisoma in the Atlantic is an exception; Targett et al. 1995). Other groups, such as nasids and kyphosids, which can achieve nutrition from macroalgae with the help of microorganisms (Horn 1989), will target macroalgae as their main source of nutrition but may be restricted in their ability to remove encrusting species.

It is commonly assumed that a large number of grazing fishes, which target smaller turf algae ingest macroalgal recruits while foraging (Green and Bellwood 2009). While macroalgal recruits are readily removed in-situ (Diaz-Pulido and McCook 2003; Loffler and Hoey 2019), limited empirical evidence exists identifying the fish species involved in this removal. Recruits of the brown macroalga Sargassum sp. were consumed by all herbivore species tested (Marshell 2014), indicating that some macroalgal species are readily consumed by most herbivorous fishes on coral reefs. However, other macroalgae may differ in their defences against herbivorous fishes.

A common macroalga, which has been shown to have multiple detrimental effects on corals is the genus Lobophora (e.g., Jompa and McCook 2002; Nyström et al. 2008; Rasher and Hay 2010). Lobophora impacts multiple life-history stages of corals because it decreases coral fecundity (Foster et al. 2008), inhibits coral larval settlement (Kuffner et al. 2006; Evensen et al. 2019) and growth (Box and Mumby 2007), and can even overgrow some live corals (Ferrari et al. 2012a; Vieira et al. 2015). Lobophora is a brown alga with an encrusting to foliose morphology (Vieira et al. 2014). Reports of effective herbivory on the alga vary, likely driven by different morphologies (Coen and Tanner 1989) and a large cryptic diversity which has only recently been revealed (Sun et al. 2012; Vieira et al. 2014). Herbivory on Lobophora is often low compared to other macroalgal species (Hay 1981; Steinberg and Paul 1990; Bennett et al. 2010), which may be driven by a larger proportion of large-sized polyphenolics, which have stronger negative impacts than smaller size classes (Boettcher and Targett 1993). For example, Lobophora was found to have 90% large-sized polyphenolics compared to only 50% of large-sized polyphenolics in Sargassum sp. (Boettcher and Targett 1993). Like many other macroalgae, Lobophora may exhibit a strong ontogenetic shift in its susceptibility

37 to grazing, as recruits were removed readily whereas adult algal control was limited (Diaz-Pulido and McCook 2003). However, which herbivores are responsible for this shift is unknown.

Here, we examine the role of several common fish herbivores in the control of the common brown macroalgal genus Lobophora. We use a series of field and tank experiments to examine whether a bottleneck exists in the establishment of Lobophora, driven by an ontogenetic shift in grazing susceptibility to fishes. We also aim to identify the fish species able to remove Lobophora recruits, which may inform efforts to protect key species that help prevent algal blooms.

3.3 Material & Methods 3.3.1 Study site The field experiments were conducted on Lighthouse Reef in the Pacific island nation of Palau (07°16’27.9’’N, 134°27’31.0’’E). Lighthouse Reef lies on the eastern coast of Palau and experiences medium wave exposure. This reef used to have high coral cover (77%) until typhoon Bopha destroyed nearly all corals in December 2012 (Roff et al. 2015b). Consequently, the reef experienced a bloom of the red alga Liagora sp., which had disappeared 6 months later (Roff et al. 2015b) but left in its wake a Lobophora sp. bloom which persisted for over two years (Bozec et al. 2018).

3.3.2 Experimental Design Adult Lobophora dynamics and impact of fish herbivory Three sites, >100 m apart, were chosen on Lighthouse Reef (Site 1: 07°16'30.3''N, 134°27'32.6'' E; Site 2: 07°16'27.9''N, 134°27'31.0''E; Site 3: 07°16'26.3''N, 134°27'29.0''E). At each site, three full cages, three partial cages, and three open plots without cages (all 50 cm long, 50 cm wide, 20 cm high) were set-up at a depth of 4 – 6 m, yielding a total of 27 plots. Plots were installed between 20th – 25th February 2017 and were left for nine weeks. Pictures were taken weekly throughout the nine weeks on the following dates: 6th March or 8th March 2017, 12th March 2017, 19th March 2017, 27th March 2017, 4th April 2017, 10th April 2017 and on the 27th April 2017. Fifty cells of 1 cm diameter were placed randomly on each plot and their Lobophora occupancy tracked throughout the study period to analyse Lobophora mortality.

GoPro cameras (GoPro Hero 3+, GoPro Inc, San Mateo, California USA) were deployed on the nine open (uncaged) plots used in the 9-week observational experiment described above and feeding behaviour recorded for 4 h each. Cameras were replaced after about 2 h due to battery and storage limitations. The cameras were deployed on three consecutive days (10th March, 11th March, 12th

38 March 2017). All cameras were deployed around noon and recorded throughout the afternoon to coincide with the highest grazing rates of herbivorous fishes.

Lobophora recruitment and influence of fish herbivory To investigate recruitment in the presence and absence of fish herbivory, two sets of caged and uncaged tiles were deployed. The first set comprised ten caged and three uncaged flat cement tiles that were deployed at each of the three sites on Lighthouse Reef between 27th September and 3rd October 2017. The second set comprised 27 caged and 27 uncaged ‘microhabitat’ cement tiles that were installed on the reef on 23rd March 2018. Both sets of tiles were removed after three weeks when macroalgal recruits became visible and the number of recruits was counted under a microscope (12× magnification). While there were differences between sets in tile morphology (the microhabitat tiles consisted of easily accessible crowns and concealed crevices; see Doropoulos et al. 2016 for details), in both cases Lobophora recruits were only counted on surfaces easily accessible to herbivores, i.e. the whole surface on the flat tiles and the crowns on the microhabitat tiles. The recruits in the concealed crevices of the microhabitat tiles were not included as herbivorous fish may have limited access to these areas and they, therefore, could not be compared to easily accessible areas. The number of recruits counted on each tile was standardized by unit area to permit comparison among sets of tiles.

Species-specific removal of Lobophora recruits Although we documented the feeding behaviour of reef fishes on Lobophora as part of the experimental study (described above), such observations do not indicate which species (if any) were responsible for algal mortality. We, therefore, ran two additional experiments to examine the species- specific removal of Lobophora recruits. A controlled tank experiment and an in-situ experiment to verify that the findings from the controlled experiment are applicable on a reef.

Tank experiment A tank experiment was conducted to evaluate the ability of different fish species to remove Lobophora recruits. Forty flat cement tiles (100 cm2) were deployed on 22nd September 2018 in cages at 4 - 6 m depth on Lighthouse Reef and retrieved once Lobophora recruits became visible on 13th October 2018, after ~3 weeks. Tiles were transported back to the research station and kept in a tank with flow-through seawater until they were used in the experiment (15th – 27th October 2018). Tiles were mapped under a microscope at 12x magnification. A grid of 2.5 cm x 2.5 cm was used to map the location of Lobophora recruits on each tile. Ten individuals of Acanthurus nigrofuscus, twelve individuals of Ctenochaetus striatus, ten initial phase Chlorurus spilurus (IP) and ten Zebrasoma

39 scopas were caught on the inshore reefs of Palau. These fish species are common on Palau’s coral reefs and were chosen for the following reasons: A. nigrofuscus is a common grazing surgeonfish with intermediate SCFA levels (Clements and Choat 1995), meaning it may be able to achieve some nutrition from macroalgae. Ct. striatus is not expected to have a strong impact on macroalgal recruits due to its feeding ecology and low SCFA levels (Clements and Choat 1995) but is the most common fish feeding on the benthos in Palau. C. spilurus can be expected to remove Lobophora because of its scraping feeding habit, even though it is unlikely to achieve nutrition directly from macroalgae. Z. scopas may also be able to achieve some nutrition from macroalgal recruits, as it has high SCFA levels and its SCFA profiles are more similar to browsing fishes than most other grazing surgeonfishes (Clements and Choat 1995; Clements et al. 2016). The fishes were transported to the station where two individuals of the same species were moved into a tank of size 235 cm x 95 cm x 70 cm (l × w × d) and left to acclimatise for two days before the experiments were run. One of the Z. scopas pairs displayed highly aggressive behaviour towards one another, so they had to be separated and only nine Z. scopas individuals participated in the experiment. Each fish was measured before being released into the tank. The average size of A. nigrofuscus was 10.2 cm, Ct. striatus 13.6 cm, C. spilurus IP 13.9 cm and Z. scopas 10.3 cm. While A. nigrofuscus can grow up to a maximum of 21 cm (fishbase.org), few large individuals were observed and the average size of A. nigrofuscus on Palauan reefs is closer to those used in the study.

The fish had access to turf algae and detritus associated with rubble except during the experimental trials. A tile with turf algae was left in the tank so fish could get used to the presence of a tile in their tank as we previously observed avoidance behaviour of the fish towards new tiles. During the experiment, the tile was replaced with one experimental tile in each tank on which feeding behaviour was recorded (GoPro Hero 3+, GoPro Inc, San Mateo, California USA) for 2.5 h. At the beginning of each set-up, the grid used for mapping the tiles was held into the frame on top of the tile to allow mapping of bites later (i.e. the grid was removed before the experiment). After the experiment, tiles were remapped as described earlier by counting Lobophora recruits. A total of five controls were run using the same procedure except that fish were unable to feed on the tiles because a cage was added.

In-situ experiment A similar experiment was conducted in-situ on the reef using the microhabitat tiles deployed in March 2018 (see description above). For the following analyses, only the flat crowns were used because they were readily accessible to all species. To relate fish feeding behaviour to recruit removal, the Lobophora recruits on 15 of these tiles, which had previously been caged, were mapped. Nine of these tiles were deployed on the reef at site 3 on 8 April 2018 in sets of three. GoPro cameras (GoPro

40 Hero 3+, GoPro Inc, San Mateo, California USA) were set up to observe feeding behaviour of herbivorous fish on the tiles for 4h in the absence of divers. Cameras were replaced after about 2h due to battery and storage limitations. The other six tiles were caged to act as controls. Tiles were retrieved after 4h and remapped. All observed bites taken on the flat crowns were counted, the species taking the bite identified and the fish length recorded. Data was only included in the following analysis if only a single species fed on a location.

3.3.3 Statistical analysis All statistical analysis was conducted using the program R (version 3.5.0, http://www.r-project.org), all generalized linear mixed effects models (glmms) were computed using the ‘lme4’ (Bates et al. 2015) package or the ‘glmmTMB’ package (Brooks et al. 2017).

Adult Lobophora dynamics and impact of fish herbivory on adult Lobophora A second-order polynomial generalized linear mixed-effects model with binomial error distribution was fitted to the 50 random cells after visual examination of the cover over time: The response variable was presence/absence of Lobophora within a cell, Treatment and Days were set as interacting fixed factors and Cell was nested within Plot, which was in turn nested within Site as random factors.

To analyse differences in adult Lobophora survival in response to whether they were caged or exposed to fish herbivory, a Kaplan-Meier curve was created and a survival coxme model was fitted. Coxme survival models allow observations to be censored if the observation period ends before an individual died, i.e. the fate of the individual is unknown after the observational period. Treatment was set as the predictor variable and Plot was nested within Site as a random effect. Kaplan Meier curves are a good tool to visualise the mortality of populations over time.

Whether fish avoid or target Lobophora was investigated visually by plotting % of bites on Lobophora over % Lobophora cover. ChiSquare was calculated after averaging % Lobophora cover, the sum of bites observed on Lobophora and the expected bites on Lobophora for each of the following fish groups: Acanthurus spp., Chlorurus spp., Ctenochaetus spp., Naso spp., Scarus spp. and Zebrasoma spp.. To confirm that the bites were correctly identified as being taken on Lobophora or another substrate type in areas further away from the camera, bites taken in the closer half of the plot to the camera were compared to bites taken in the half further away using separate linear models (lm) for parrotfishes and surgeonfishes. Percent bites taken on Lobophora was log + 1 transformed to meet model assumptions and set as the response variable, whereas Lobophora cover and Plot position (front vs. back) were set as interacting predictor variables. 41

Lobophora recruitment and influence of fish herbivory The number of Lobophora recruits observed on tiles in September/October 2017 was compared to the recruits counted on the crowns of tiles in March/April 2018. We fitted a negative binomial generalized linear model, with time and treatment as fixed factors and included an offset of the tile area to account for the different areas included. Post-hoc multiple comparisons using Tukey tests were conducted with the ‘multcomp’ package in R (Hothorn et al. 2008).

Species-specific removal of Lobophora recruits Tank experiment For the following analyses, only locations where feeding activity occurred were included. Z. scopas consumed some Lobophora recruits only partially. Since partially removed recruits may be able to recover, they were regarded as having ‘survived’.

To determine if a fish species was able to remove significantly more recruits than were lost on a control tile, we used a binomial model (eq. 1 and eq. 2, see supplementary) following Harborne et al. (2009), since quasi-complete separation prohibited the use of a generalized linear mixed effects model. This was followed-up with a Bonferroni Holm correction of the alpha level for each fish group to account for the use of multiple comparisons (Holm 1979).

A generalized linear mixed-effects model with binomial distribution was fitted to compare the species’ abilities to remove Lobophora recruits among each other. The number of successes and the number of failures of Lobophora removal were bound and set as the response variable, Species was set as fixed factor and Tile nested within Set-up was set as random factor. Post-hoc multiple comparisons using Tukey tests were conducted with the ‘multcomp’ package in R (Hothorn et al. 2008).

In-situ experiment To investigate which species can remove Lobophora recruits in-situ we used the video observations and recruit removal data obtained on the reef. Only locations that originally had Lobophora sp. recruits were included in the analysis. Additionally, locations were only included if they were bitten by a single fish species to avoid confounding results due to multiple species taking bites. This limited the bite data per location and we thus fitted a generalized linear mixed-effects model with binomial error distribution, setting the probability of a Lobophora mortality event (1 vs. 0) as the response

42 variable. Species (including control) was set as the predictor variable, with a species being recorded if any bites were taken on that location. Post-hoc multiple comparisons using Tukey tests were conducted with the ‘multcomp’ package in R (Hothorn et al. 2008).

3.4 Results 3.4.1 Adult Lobophora dynamics and impact of fish herbivory on adult Lobophora When herbivorous fishes were excluded in caged plots, Lobophora cover increased throughout the experiment with a slight levelling off after ~ 40 days (Figure 3.1). Both the first- and second-order polynomial terms were significant (glmm, p < 0.001 both). In contrast, when herbivores had access in partially caged and open plots, Lobophora cover increased in March but started to decrease again in April (Figure 3.1). Correspondingly, there was evidence for a second-order polynomial relationship (not different to second-order polynomial relationship in caged plots; glmm, p > 0.05, Table S3.1), but no evidence for a first-order polynomial relationship (significantly different to the linear increase in caged plots; glmm, p < 0.001; Table S3.1).

Figure 3.1: Percentage Lobophora cover observed over nine weeks. Data is obtained through 50 random cells which were followed throughout the experiment. Individual observations and polynomial regressions are displayed. Error margins show the 95 % confidence interval.

Lobophora survival did not differ among treatments (coxme, open: z = -1.075, p > 0.05, partially caged: z = -0.896, p > 0.05, Table S3.2, Figure 3.2). The mean age was 2.5 ± 3.0 (sd) weeks in caged treatments, 2.6 ± 3.8 weeks in partially caged treatments and 2.9 ± 3.3 weeks in open treatments.

Figure 3.2: Kaplan-Meier survival curve using 50 random cells within each plot type of which cells occupied by Lobophora were followed throughout a nine-week period. + symbols show that some cells did not ‘die’ at that time-point, but disappeared from observations, which happens when a cell was still alive at the end of the experimental period and its fate is therefore unknown.

The percentage of bites taken on Lobophora by both surgeonfishes and parrotfishes increased positively with the cover of Lobophora (Figure 3.3). The number of bites taken by parrotfishes on Lobophora was proportionate to the alga’s cover (ChiSquare: 0.4 and 1.1 for Scarus spp. and Chlorurus spp., respectively). All surgeonfishes species, including Acanthurus spp., Ctenochaetus spp., Zebrasoma spp., and Naso spp., took more bites on Lobophora than expected (Figure 3.3; ChiSquare: 37.7, 83.2, 47.1 and 7.2, respectively). Identification of the substrate bitten was consistent between the front and the back half of the plot for both surgeonfishes and parrotfishes (lm, surgeonfishes: t = -0.558, p > 0.05; parrotfish: t = 0.201, p > 0.05), implying that bites taken further away from the camera were recorded correctly.

Figure 3.3: Percentage of bites taken by surgeonfish and parrotfish on Lobophora in a monitored plot of 50 cm × 50 cm graphed over the percentage Lobophora cover within that plot (as percentage of available area). Dots are individual observations, colours show species identity and the line shows a ratio of 1:1. Dots falling above the line indicated more than proportionate amounts of bites have been taken, dots below the line indicate fewer bites than expected.

3.4.2 Lobophora recruitment and fish species-specific recruit mortality Recruitment dynamics When herbivorous fishes were excluded from tiles in caged treatments, more Lobophora recruits established compared to tiles which allowed access by fish to the algal recruits (Figure 3.4; glmm, z = -5.259, p < 0.001, Table S3.3). This held true in March/April 2018 and in September/October 2017 (glmm, March/April 2017: z = -5.259, p < 0.001, September/October 2017: z = -107.403, p < 0.001, Table S3.3).

More Lobophora recruits were observed in March/April 2018 compared to September/October 2017. This was the case for caged treatments (Tukey post-hoc, z = -250.705, p < 0.05, Table S3.3 Figure 3.4) and for uncaged treatments (Tukey post-hoc, z = -107.403, p < 0.05, Table S3.3).

Figure 3.4: Comparison of Lobophora recruits (per cm2) on easily accessible crowns of tiles in March 2018 and flat tiles deployed in September 2017. Letters symbolize significantly different results. Mean area and standard error are displayed.

Species-specific removal of Lobophora recruits In the tank experiment, only two fish species showed a clear impact on recruit mortality. Acanthurus nigrofuscus (binomial model, p < 0.05) and Chlorurus spilurus (binomial model, p < 0.01) removed more Lobophora recruits than were lost in control treatments, whereas Ctenochaetus striatus (binomial model, p > 0.05) and Zebrasoma scopas (binomial model, p > 0.05) did not. However, after post-hoc corrections, A. nigrofuscus (p = 0.021) was marginally insignificant, as the alpha level was adjusted to 0.017 following the Bonferroni-Holm method (Holm 1979). A. nigrofuscus and C. spilurus did not differ in their ability to remove Lobophora recruits (glmm, z = -0.731, p > 0.05, Table S3.4). Ct. striatus was not significantly different from A. nigrofuscus or C. spilurus either, but this observation was based on a single recruit removal. Only A. nigrofuscus and Z. scopas differed significantly, with A. nigrofuscus removing more Lobophora recruits than Z. scopas (glmm, z = - 2.112, p < 0.05, Table S3.4; Figure 3.5A).

During the in-situ experiment where observations of feeding could be linked to a single species at a time, seven nominally herbivorous fish species visited the tiles. Only A. nigrofuscus and Scarus niger caused significantly higher Lobophora recruit mortality compared to controls (Figure 3.5B, glmm, A. nigrofuscus: z = 2.931, p < 0.01 and Scarus niger: z = 2.291, p < 0.05, Table S3.5). Z. scopas was marginally significant (glmm, z = 1.901, p = 0.06, Table S3.5) and neither C. spilurus, nor Ctenochaetus binotatus, Ct. striatus or Naso lituratus caused Lobophora recruit mortality (glmm, all p > 0.05, Table S3.5). The maximum number of bites per location, and therefore the likelihood that a recruit may have been removed during non-targeted feeding, varied among species. Acanthurus nigrofuscus took a maximum of two bites per location, C. spilurus and S. niger took a maximum of three bites, Ct. binotatus took a maximum of four bites, and Z. scopas and N. lituratus each took a maximum of five bites. Ct. striatus took the most bites with a maximum of 12 bites per location.

Figure 3.5: A) Mean proportion of Lobophora recruits removed by different fish species during the controlled tank experiment. Letters symbolize significantly different results. Errorbars show standard error. B) Probability of a Lobophora recruit being removed when a fish species visited a tile during in-situ feeding observations and took bites on locations with Lobophora recruits present. Locations were included when only one species took bites on them to avoid confounding feeding by multiple species on the same recruit location. Letters symbolize significantly different results. ‘Max. bites’ are the maximum number of bites which were taken on one location.

3.5 Discussion There is a strong ontogenetic shift in the susceptibility of the macroalga Lobophora to fish grazing. While grazing activity severely reduced recruit establishment, control of adult algae was limited. However, the ability of fish species to remove recruits varied considerably. Out of seven species observed, only A. nigrofuscus, C. spilurus and S. niger removed Lobophora recruits. These ontogenetic shifts in grazing susceptibility and species-specific algal–fish interactions highlight the need to consider species-specific behaviours driving or preventing shifts to macroalgal dominance.

Fish herbivory had a much stronger influence on Lobophora recruit establishment and mortality compared to adult Lobophora mortality. During the 2.5 h tank experiment, up to 40% of recruits were removed depending on the fish species involved. During the 4h in-situ observations fish caused a mortality event with up to 50% probability. Additionally, over three weeks 78% – 95% fewer recruits became established when they were exposed to herbivores. These values are similar to the herbivore- driven spore and growth reductions of other algal species (Lotze et al. 1999). In contrast, we found much less impact of herbivory on adult Lobophora. The trajectory of Lobophora cover differed between caged vs. open plots and partially caged treatments, which implies that herbivores did influence adult Lobophora in some way. While Lobophora cover showed a linear increase in caged plots, the cover in open and partial cages increased for about four weeks before decreasing again to a similar Lobophora cover as recorded at the beginning of the experiment. Indeed, when herbivores had access to the adult algal patches, overall Lobophora cover did not change from the first timepoint to nine weeks later (Figure 3.1). However, cover did increase by 19.3% in cages. This indicates that there is an equilibrium between herbivory and Lobophora growth when herbivores have access to the alga but when herbivory is reduced the alga can increase its cover. Interestingly, adult Lobophora survival was independent of herbivore access, indicating that the observed change in cover may be driven by the establishment of new algae rather than the consumption of established Lobophora thalli. Given the low number of fish species that can be expected to derive nutrition from adult Lobophora (Horn 1989; Choat et al. 2002), this pattern is not surprising. These findings indicate that adult Lobophora may be less susceptible to grazing by herbivorous fishes than recruits.

While we did not test whether the increased grazing resistance with age is driven by chemical (e.g., secondary metabolites) or morphological (e.g., size) changes in the alga, there is some support for both concepts. Apart from differences in their sizes, recruits are relatively flush against the substrate (which also sets them apart from other macroalgal recruits), but are only attached by one holdfast, whereas adult encrusting or decumbent Lobophora of the same species is well attached to large areas 49 of the substrate (LD Puk, pers obs). These morphological differences along with larger sizes would make the removal of adult Lobophora much harder than the removal of recruits. Phlorotannins, a secondary metabolite of brown macroalgae, polymerize and increase in size as they age (Targett and Arnold 2001), which makes them more likely to interfere with the digestion of other macromolecules (Boettcher and Targett 1993). However, the ageing process is quite rapid and has been observed over a few hours (Targett and Arnold 2001). It is therefore unclear whether this mechanism would act over weeks or months as would be required to explain differences between adult and recruit chemical defences. Whichever mechanism dominates, our findings suggest a strong ontogenetic shift in grazing susceptibility as predicted by the growth-differentiation balance (Herms and Mattson 1992). This ontogenetic shift in grazing susceptibility may be more pronounced in Lobophora than in other macroalgae, because recruit removal rates were similar to other algae (Lotze et al. 1999) whereas Lobophora adult removal was low compared to other macroalgal species (Hay 1981; Steinberg and Paul 1990; Bennett et al. 2010).

The ability of herbivorous fishes to control Lobophora recruits is species-specific. The grazing surgeonfish A. nigrofuscus removed Lobophora recruits significantly in-situ while the removal of recruits in the tank experiment was marginally significant, possibly due to low replication. In contrast, the grazing/detritivorous surgeonfish Ct. striatus did not remove recruits in either of the two experiments. While Ct. striatus was not significantly different to A. nigrofuscus and C. spilurus in tank experiments the ability of this species to control Lobophora seems very limited. The lack of recruit removal in-situ where Ct. striatus took by far the highest number of bites without causing Lobophora mortality corroborates its likely minimal impact on Lobophora. This is an important difference between A. nigrofuscus and Ct. striatus, as Ct. striatus is one of the most abundant herbivorous fishes on Indo-Pacific reefs (Russ 1984; Cheal et al. 2012), but our study indicates that it is unable to control outbreaks of macroalgae like Lobophora.

The grazing tang Zebrasoma scopas did not remove Lobophora recruits in either the tank experiment or the in-situ experiment. However, we found multiple Lobophora recruits that had been partially removed by Z. scopas. Therefore, it seems that Z. scopas does feed on Lobophora recruits but whether its feeding activity has any impact on the survival of Lobophora recruits is unknown and would require longer-term monitoring of recruit growth and survival when exposed to Z. scopas.

The fourth species observed both in-situ and in the tank is the parrotfish C. spilurus, which removed recruits in the tank experiment but did not remove any recruits in-situ. It is possible that C. spilurus avoided Lobophora recruits in-situ because parrotfish have been identified as microphages, which

50 target microorganisms such as cyanobacteria (Clements et al. 2016). However, C. spilurus morphological adaptations to excavating reef substrate while feeding means they remove more substrate than the scraping parrotfish S. niger (Bellwood and Choat 1990), which removed recruits in-situ. It is therefore likely that C. spilurus does remove Lobophora recruits during its foraging and the lack of evidence from the field is due to the low number of bites observed in-situ. While Naso lituratus, a common browser adapted to removing macroalgae (Choat et al. 2004; Rasher et al. 2013; Plass-Johnson et al. 2015), did not remove any recruits in our in-situ study, we have only limited data on N. lituratus (two forays) and cannot draw conclusions about its ability to remove Lobophora recruits on a reef scale. Overall, our experiments show important differences in the ability of fish species to remove Lobophora recruits.

While Lobophora recruits are readily removed by multiple species, adult Lobophora is often persistent through time (van den Hoek et al. 1978; de Ruyter van Steveninck and Breeman 1987b; Roff et al. 2015b) and only a few herbivore species are expected to control the alga (Horn 1989; Choat and Clements 1998). Two groups of herbivorous fishes reported to remove adult Lobophora are rabbitfish (Pillans et al. 2004; Bennett et al. 2010), which were rare at our study site, and parrotfish (Roff et al. 2015b). Parrotfish took a proportionate number of bites on Lobophora during our study, indicating that they neither target nor avoid the alga, which is in line with their feeding ecology (Clements et al. 2016). In contrast to parrotfish, unicornfish (Naso spp.), which can digest macroalgae (Choat et al. 2004), consistently took disproportionately many bites on the alga, possibly indicating a targeted consumption of Lobophora. Other surgeonfish species are highly unlikely to remove adult Lobophora, especially the encrusting morphology dominating in Palau (LD Puk, pers obs) because of their jaw morphology (Purcell and Bellwood 1993) and gut physiology, which is unsuitable for macroalgal digestion (Horn 1989; Choat et al. 2004). It is more probable that surgeonfish target either the turf algae, epiphytes or detritus growing on the surface of Lobophora (Fricke et al. 2011; Eich et al. 2019) and therefore have little direct influence on Lobophora. Generally, it seems the removal of adult Lobophora is limited to few fish species.

Our study found variable Lobophora growth and recruitment at different times of the year. Independent of whether herbivores had access to Lobophora, the alga’s cover increased over four weeks indicating that herbivory did not control its growth sufficiently. After four weeks, Lobophora in caged treatments without herbivore access kept growing until it plateaued after around six weeks, which could have been driven by a reduction in growth during the second half of the experiment. In contrast, Lobophora in partial and open treatments, which allowed herbivore access, declined again after week four until the end of the experiment. This decline in Lobophora cover shows that herbivory

51 during the second half of the experiment was able to reverse Lobophora growth, possibly because of reduced growth rates. Similar to Lobophora cover, Lobophora recruitment showed variability throughout time, with four times higher recruitment rates in March 2018 compared to September 2017. While our data does not allow us to conclude that this variability is driven by seasonality, Lobophora has previously been found to be highly seasonal in the Indo-Pacific (Diaz-Pulido et al. 2009) and maximum Lobophora cover was observed in April in Palau (Roff et al. 2015b), a month after the highest recruitment in this study. Repeated multi-year observations would be necessary to assess seasonality. Temporally variable growth has implications at the scale of entire reefs. If reductions in grazing pressure, for example through fishing or increases in the grazable area after mass coral mortality, co-occur with peak growth and recruitment, it may have a much more substantial ecological impact than grazing reductions during low growth and recruitment times. The timing of disturbances may, therefore, play a role in the formation of Lobophora blooms, which can subsequently persist for several years (Roff et al. 2015b; Bozec et al. 2018).

We found an ontogenetic shift in grazing susceptibility paired with species-specific removal. Therefore, effective management strategies aimed at increasing the resilience of coral reefs should pay attention to species able to remove macroalgal recruits and not solely focus on fish species removing adult algae. Our findings suggest that management strategies aimed at controlling macroalgal growth on coral reefs are more effective when put into place before algae escape their vulnerable phase and develop into grazing-resistant adults. In a world of increasing macroalgal proliferation on coral reefs, fish species able to remove macroalgal recruits and thereby prohibit the establishment of more resistant adult populations, such as A. nigrofuscus and parrotfishes, are important for the resilience of coral reefs. The abundance of these key species should thus be monitored and fishing regulations should be considered by managers.

3.6 Acknowledgements We would like to thank Shannen Smith, Kelly Wong and Alex Tredinnick for their help during field work. We are also grateful to the Palau International Coral Reef Center for providing us with the space necessary to conduct our studies. This research was funded by the Winnifred Violet Scott Trust and by ARC grants awarded to PJ Mumby.

3.7 Supplementary Description of binomial model used to calculate Lobophora removal in the tank experiment The null hypothesis stated that the probability of a Lobophora recruit surviving during the experiment in a treatment was the same as the probability of one surviving within the control treatment. We did 52 not pool all recruits exposed to fish species k, but rather used the probability of a Lobophora recruit being removed on a tile level within a treatment (exposed to fish species k). These probabilities were then used to calculate a mean probability of finding the observed number of Lobophora recruits removed on the control tiles (Eq. 1).

P(nj)= 1 ∑a B(nj|Nj, pk) (1) a i=1

where nj is the number of Lobophora recruits removed on a control tile j, a is the number of tiles used in the treatment (number of tiles presented to fish species k), Nj is the overall number of recruits that were present on the control tile j, and pk is the probability of a Lobophora recruit being removed on a tile exposed to species k. The total probability of recording the observed number of recruits removed on the control tiles was calculated using the product of the five probabilities (P(nj)) calculated for each of the control tiles (see eq. 2) described below. If the total probability of finding the observed number of recruits removed on the five control tiles was < 0.05, the null hypothesis was rejected.

The individual probability of finding the number of Lobophora recruits removed from a treatment tile (exposed to fish species k) on a control tile was calculated with the following binomial equation (Eq. 2).

P(X)= n! *pX*(1-p)n-X (2) (n-X)!*X! where n is the number of recruits on a control tile, X is the number of recruits removed from this control tile, p is the probability of a recruit being removed from a treatment tile presented to species k. Since there were five control tiles, the five individual probabilities calculated were multiplied to achieve the overall probability, which was used to accept or reject the null hypothesis.

53 Statistical output tables

Table S3.1: Polynomial generalized linear mixed effects model (glmm) output analyzing Lobophora cover over time. Bold print under p-value symbolizes significant results. Analysis Test used Fixed effect Estimate Std. Z- P-value Error value Lobophora Polynomial Caged Plot -0.22 0.446 -0.494 0.622 cover over glmm (reference level) Time Open Plot -1.019 0.629 1.62 0.105 Partial Cage 0.111 0.630 0.176 0.861 Poly(Days, 2)1 38.938 5.568 6.993 < 0.001 Poly(Days, 2)2 -28.316 5.101 -5.551 < 0.001 OpenPlot: -43.601 7.650 -5.700 < 0.001 poly(Days, 2)1 PartialCage: -38.904 8.469 -4.594 < 0.001 poly(Days, 2)1 OpenPlot: 4.5104 7.568 0.596 0.551 poly(Days, 2)2 PartialCage: 3.347 7.151 0.468 0.640 poly(Days, 2)2

Table S3.2: Coxme survival model output analyzing Lobophora adult survival. Bold print under p- value symbolizes significant results.

Analysis Test used Fixed effect Value Std. Error Z -alue P-value Lobophora Coxme Cage 2.230 0.069 32.234 < 0.001 adult model (reference survival level) Open Plot -0.108 0.100 -1.075 0.283 Partial Cage -0.086 0.096 -0.896 0.370

54 Table S3.3: Negative binomial generalized linear model (glm) and post-hoc Tukey output analyzing temporal Lobophora recruitment. Bold print under p-value symbolizes significant results.

Analysis Test used Fixed effect Estimate Std. Z-value P-value Error Temporal Negative Caged -31.945 0.189 -169.370 < 0.001 Lobophora binomial March/April recruitment glm 2018 (reference level) Open Plot -1.4920 0.284 -5.259 < 0.001 March/April 2018 Caged Sep/Oct -65.943 0.263 -250.705 < 0.001 2017 Open Sep/Oct -67.374 0.627 -107.403 < 0.001 2017 Post-hoc Caged Sep/Oct -65.943 0.263 -250.705 < 0.001 Tukey test 2017 vs. Caged Mar/Apr 2018 Open Mar/Apr -1.4920 0.284 -5.259 < 0.001 2018 vs. Caged Mar/Apr 2018 Open Sep/Oct -2.923 0.618 -4.728 < 0.001 2017 vs. Caged Sep/Oct 2017 Open Sep/Oct -67.374 0.627 -107.403 < 0.001 2017 vs. Open Mar/Apr 2018

55 Table S3.4: Binomial generalized linear mixed effects model (glmm) output analyzing Lobophora recruit removal facilitated by different fish species in the tank experiment. Bold print under p-value symbolizes significant results.

Analysis Test used Fixed effect Estimate Std. Z-value P-value Error Tank Binomial Acanthurus 0.428 0.586 -0.731 0.465 experiment glmm nigrofuscus – Lobophora (reference recruit level) removal Chlorurus -0.585 0.870 -0.672 0.5012 – spilurus Comparison Ctenochaetus -1.220 1.333 -0.916 0.360 between fish striatus species Zebrasoma -2.706 1.281 -2.112 0.035 scopas

56 Table S3.5: Binomial generalized linear mixed effects model (glmm) output analyzing Lobophora recruit removal facilitated by different fish species in-situ. Bold print under p-value symbolizes significant results.

Analysis Test used Fixed effect Estimate Std. Error Z-value P-value In-situ Binomial Control -3.401 0.480 -7.079 < 0.001 – Lobophora glmm (reference level) recruit Acanthurus 3.401 1.161 2.931 0.004 removal nigrofuscus – Chlorurus -17.91 1.898e+04 -0.001 1.000 Comparison spilurus control vs. Ctenochaetus -16.04 9.618e+03 -0.002 1.000 species binotatus Ctenochaetus -20.09 3.062e+04 -0.001 1.000 striatus Naso lituratus -21.02 1.004e+05 0.000 1.000 Scarus niger 2.370 1.035 2.291 0.022 Zebrasoma 2.673 1.406 1.901 0.057 scopas

______

CHAPTER 4: REFUGE DEPENDENT HERBIVORY CONTROLS A KEY MACROALGA ON CORAL REEFS ______

Puk LD, Marshell A, Dwyer J, Evensen NR, Mumby PJ (2020). Refuge dependent herbivory controls a key macroalga on coral reefs. Coral Reefs. Doi: https://doi.org/10.1007/s00338-020-01915-9.

Contributor Statement of contribution Laura D Puk Conception (50%) Data collection (90%) Data analysis and interpretation (60%) Drafting and production (80%) Alyssa Marshell Conception (10%) Drafting and production (5%) John Dwyer Data analysis and interpretation (20%) Drafting and production (5%) Nicolas R Evensen Conception (5%) Data collection (10%) Drafting and production (5%) Peter J Mumby Conception (35%) Data analysis and interpretation (20%) Drafting and production (5%)

58 Refuge dependent herbivory controls a key macroalga on coral reefs

4.1 Abstract Small-scale structural complexity shapes how consumers and primary producers interact, which can influence ecosystem trajectories. Coral reefs are some of the most structurally complex ecosystems, though their complexity is threatened owing to anthropogenic influences. Some reefs shift towards macroalgal dominance following mass coral mortality, which can hinder the recovery of corals because they compete with the faster-growing macroalgae for space. Using video observations, surveys and in-situ experiments on a forereef in eastern Palau, we investigated the role different microhabitats play in facilitating the persistence of the macroalga Lobophora, which is one of the strongest negative interactors with corals. Collectively, our observational and experimental data show that small crevices provide a refuge to Lobophora recruits by excluding most adult herbivorous fishes. Consequently, Lobophora is disproportionately more abundant within these concealed microhabitats on the reef, which highlights the important role of microhabitats as macroalgal spore sources from which macroalgae can spread following mass coral mortality. While a large proportion of our current understanding of grazer – algae interactions is based on research using flat surfaces, our findings demonstrate that the interactions between herbivorous fishes and benthic organisms are strongly mediated by microhabitats. It is thus important to consider the influence of small-scale structural complexity in order to understand the nuances that govern benthic regimes.

4.2 Introduction Structural complexity is a key ecosystem feature that has been positively linked to species diversity (Bazzaz 1975; Graham and Nash 2013; Rogers et al. 2014). Structural complexity creates microhabitats, which act as spatial niches (Bazzaz 1975), and thus shapes communities both in terrestrial (Bazzaz 1975; Tews et al. 2004) and aquatic environments by creating microhabitats (Menge and Lubchenco 1981; Poray and Carpenter 2014; Brandl and Bellwood 2016), which act as spatial niches (Bazzaz 1975). Concealed microhabitats (i.e., crevices) shape consumer-producer interactions because they effectively reduce grazing pressure (Brandl and Bellwood 2016). By mediating these interactions, microhabitats can have a profound influence on the overall community structure.

Coral reefs are arguably some of the most structurally complex marine ecosystem which, not surprisingly, support highly diverse assemblages (Luckhurst and Luckhurst 1978), including large numbers of vertebrate and invertebrate species. On reefs, concealed microhabitats are the preferred settlement habitat for corals and post-settlement survival tends to be higher in such habitats (Nozawa 59 2008; Doropoulos et al. 2016a). However, concealed microhabitats both in coral reefs and in other ecosystems can also provide protection to competing species, including macroalgae (Menge and Lubchenco 1981; Lubchenco 1983; Poray and Carpenter 2014; Franco et al. 2015). Therefore, these microhabitats, which are ubiquitous on coral reefs worldwide may have a profound impact on critical ecological processes (e.g. primary production, herbivory) that determine reef functioning (Brandl et al. 2019).

Corals are deleteriously impacted by several algal taxa (Rasher and Hay 2010), but one of the strongest interactors is the macroalgal genus Lobophora (e.g., Jompa and McCook 2002; Rasher and Hay 2010; Mumby et al. 2016). Lobophora is a brown macroalga with an encrusting to foliose morphology (Vieira et al. 2014), which can impact multiple coral life history stages as it decreases coral fecundity (Foster et al. 2008), inhibits coral larval settlement (Kuffner et al. 2006; Johns et al. 2018; Evensen et al. 2019) and growth (Box and Mumby 2007), and can even overgrow some live corals (Ferrari et al. 2012; Vieira et al. 2015). On healthy coral reefs, Lobophora is often associated with concealed microhabitats, such as those at the base of branching corals (Bennett et al. 2010). This association is likely due to reduced herbivory within concealed microhabitats (Dudley and D’Antonio 1991; Bergey 2005; Bennett et al. 2010), which can be one to two orders of magnitude lower than on planar surfaces (Bennett et al. 2010; Brandl and Bellwood 2016; Doropoulos et al. 2016a).

While there are varied reports of the palatability of Lobophora (e.g., Hay 1981; Lewis 1985; Targett et al. 1995), herbivory is generally considered to constrain its abundance (de Ruyter van Steveninck and Breeman 1987a; Jompa and McCook 2002a). In the context of this study, we use the term ‘herbivore’ for species that contribute to algal disturbance (sensu Steneck, 1988) rather than species retrieving nutrition from algae (Clements et al. 2016). When herbivore pressure is reduced, Lobophora can spread rapidly and dominate previously coral-dominated reefs (Roff et al. 2015b). Ecosystem shifts to Lobophora dominance have been reported in the Caribbean (de Ruyter van Steveninck and Breeman 1987b; McClanahan and Muthiga 1998) and the Indo-Pacific (Diaz-Pulido et al. 2009; Cheal et al. 2010; Roff et al. 2015b). Coral reefs that have undergone such a ‘phase shift’ can be slow to recover and might even develop into stable states (Mumby et al. 2013). An algal- dominated reef provides less habitat to a variety of species (Graham et al. 2007; Alvarez-Filip et al. 2009) and a different – often reduced – set of ecosystem services (Bryant et al. 1998; Rogers et al., 2015).

Other examples of ecosystem shifts related to structural complexity have been observed in terrestrial ecosystems. For example, in grazed grassland systems concealed microhabitats can protect grazing-

60 sensitive plant species, providing a refuge from which they can recolonise space after grazing is reduced (Shitzer et al. 2008). Similarly, concealed microhabitats could act as a refuge to macroalgae by excluding herbivorous fishes (Brandl and Bellwood, 2016), thus providing a source of algal propagules to spread if herbivory declines. However, it is not yet clear whether concealed microhabitats provide such refuges for macroalgal recruits (Brandl et al. 2014, but see Diaz-Pulido and McCook, 2004). We refer to recruits as early life-stage Lobophora visible to the naked eye and thus including recruitment and post-settlement survival, whereas ‘spore’ will be used to refer to a pre- settlement reproductive body (following Gaylord et al. 2006). Macroalgal spore sources could be created through a reduction in grazing pressure and/or through excluding key fish species able to remove macroalgae. While reductions in grazing pressure are often observed (e.g., Menge and Lubchenco, 1981; Dudley and D’Antonio, 1991; Brandl and Bellwood, 2016), the role of microhabitats in shaping the herbivore species assemblage accessing them is less well understood (but see Fox and Bellwood 2013; Brandl and Bellwood 2014). It is therefore unclear how individual herbivorous fish species interact with these potential refuge microhabitats (but see Brandl and Bellwood 2016). Here, we investigate the importance of crevices as Lobophora refuges by addressing the following hypotheses:

H1: Adult Lobophora occurrence is negatively associated with crevice size and grazing pressure. Smaller crevices are expected to exclude more grazers than larger crevices or open areas and would thus provide more protection to Lobophora.

H2: Small crevices provide a refuge for Lobophora by increasing recruit survival. Recruit survival is expected to be a big driver of Lobophora abundance and due to reduced herbivory, more recruits are expected to survive in smaller crevices.

H3: The accessibility of microhabitats partitions fish herbivory, leading to changes in grazing pressure and in the identity of fish species feeding within different microhabitats. Different fish species are expected to have varying access to microhabitats, either because of their morphology or because of the fishes’ feeding behaviour.

4.3 Material & Methods 4.3.1 Study site This study was conducted at two forereef sites, at 5 m – 9 m depth, in the West Pacific island nation of Palau between March 2017 and May 2018. Both Lighthouse Reef (07°16’27.9’’N, 134°27’31.0’’E) and East Sheltered Reef (07°17’05.0’’N, 134°31’38.7’’E) are situated on the central-eastern coast of 61 Palau and are ~ 8 km apart. Lighthouse Reef had high coral cover (77%) until typhoon Bopha destroyed nearly all corals in December 2012 (Roff et al. 2015b). Consequently, the reef experienced a bloom of the red alga Liagora sp., which disappeared 6 months later but left in its wake a persistent Lobophora bloom which continued to dominate the reef for over two years (Roff et al. 2015b) and has only started to recede towards the end of the study period (LD Puk, pers. obs.). East Sheltered Reef was unaffected by the typhoon with coral cover increasing between 2012 and 2014 from 46.0% to 65.2% and the reef did not experience a shift to Lobophora dominance (Roff et al. 2015b).

4.3.2 Experimental design To determine the accessibility of microhabitats to herbivorous fish species and investigate the impact of this accessibility on Lobophora, this study consisted of three stages. The first stage included the construction of fish shapes based on morphological measurements of target fish groups and an analysis of feeding behaviour from videos to extract feeding angles utilised by these fish groups. The second stage was a microhabitat survey, which used the fish shapes and data collected during stage one to test the accessibility of microhabitats to fish species on the reef and relate this to the probability of Lobophora occurrence. The third stage consisted of a manipulative experiment using experimental tiles to study the influence of different microhabitat types on Lobophora recruits.

Creation of fish shapes and feeding behaviour Seven fish groups were chosen to represent a variety of feeding modes: Ctenochaetus striatus, Acanthurus nigrofuscus, Naso lituratus, Zebrasoma scopas, ‘general’ parrotfish (including Scarus spp., Hipposcarus spp. and Chlorurus spilurus), Chlorurus microrhinos and Siganus vulpinus. These species were included because of their possible ability to act as an agent of disturbance to macroalgae, including recruit removal, even though most of the species do not possess the enzymes or bacterial fermentation processes needed to derive nutrients from Lobophora (Clements and Choat 1995; Choat et al. 2002; Clements et al. 2016). Fish models were based on measurements from Ct. striatus, N. lituratus, Z. scopas, Scarus ghobban (to represent ‘general’ parrotfish), C. microrhinos and S. vulpinus. The shapes were built using armature (a mouldable metal wire) to make a flexible skeleton which was covered with steel mesh (i.e., ‘chicken fence’, 1.3 cm mesh). These shapes allow for some flexibility of the structure but keep their general shape. 15 models were built, using typical size classes of the chosen species (Table S4.1).

Feeding behaviour of herbivorous fishes was analysed from videos recorded by cameras placed randomly on the reef in the absence of divers. Three types of feeding angles were recorded for each bite taken: the rotation around the anterior-posterior axis, the absolute position of the fish in the water 62 column and the relative position to the substrate (Figure 4.1). The number of bites observed in the video recordings were counted for each species separately. In all cases, feeding angles were categorised in steps of mostly 45° (Figure 4.1). A total of 35 h of video material was analysed for all species except for Ct. striatus and A. nigrofuscus, which had taken over 3000 bites after just 13.5h and 3000 bites were deemed representative of their feeding behaviour. N. lituratus and S. vulpinus were rare in the videos and additional in-situ surveys were conducted to reach ~600 bites for each species. Individuals were followed at a distance of > 2 m to avoid disturbing them, their feeding behaviour recorded and later analysed following the described protocol.

Figure 4.1: A) - C) Feeding angles used to classify the ability of fish to take bites in certain microhabitats. D) Example of fish models used.

Microhabitat accessibility to fish – a survey using fish models Microhabitat characteristics were surveyed between 5th and 9th April 2017 at East Sheltered Reef by examining ten randomly placed quadrats of 30 cm × 30 cm (length × width). The quadrats were divided into nine sub-quadrats (10 cm × 10 cm) and within each sub-quadrat five microhabitats were

63 chosen representatively of the available microhabitats within the sub-quadrat, yielding a total of 450 microhabitats. The following parameters were measured for each microhabitat: surface shape (flat vs. concave vs. convex), surface orientation (horizontal vs. vertical) and crevice depth, length and width if not flat. These measurements extended beyond the size of the sub-quadrat if necessary to measure large microhabitats. Additionally, the type of alga in each microhabitat was recorded (Lobophora, turf algae, crustose coralline algae (CCA), other macroalgal genera). We then tested the accessibility of each microhabitat with the 15 fish shapes by inserting them into a microhabitat using each combination of the feeding angles utilised by the respective fish species (Figure 4.1). The combinations included the ‘rotation around the anterior-posterior axis’ and the ‘relative position to the substrate’ because a combination of ‘relative position to the substrate’ and ‘absolute position in the water column’ would have been virtually impossible to test.

Influence of microhabitats on Lobophora recruit establishment – a manipulative field experiment

Fish access in-situ (using the fish models) was mainly restricted by microhabitat width (see Supplementary information, Table S4.2). Based on this result, we created microhabitat cement tiles with microhabitat types of varying width to test whether fish species would be able to access recruits in experimental microhabitats. The cement tiles used were originally designed by Doropoulos et al. (2016b) but were modified to yield the necessary microhabitat types: crowns (uprising structures) and one of three crevice sizes (8 mm, 15 mm or 40 mm; Figure 4.2). The sizes were chosen based on their expected ability to exclude herbivorous fishes and the results from the microhabitat survey. Plotting the data from the microhabitat survey showed that the probability of Lobophora occurrence decreased steeply with microhabitat width until around 15 – 20 mm before levelling off. The smallest crevice category of 8 mm width was thus chosen to represent small crevices, which were expected to exclude most species. The 15 mm wide crevices were expected to exclude some but not all species as it was around the threshold observed. The largest crevices of 40 mm were chosen as they were expected to allow access by all fish species.

While microhabitat can refer to any type of small-scale habitat, crevices were defined as concealed microhabitats that are only accessible from one direction (following Brandl et al. 2014). We will use the term ‘crevice’ for such concealed microhabitats, including microhabitats that are accessible only from the top and ‘microhabitat’ for all microhabitats, including concealed and exposed ones. For the purpose of this experiment, microhabitat thus includes crowns and crevices, whereas crevice refers to the three differently sized concealed microhabitats.

Figure 4.2: Experimental design to test the influence of crevice width on Lobophora recruitment. Three different crevice widths were used within three different herbivory treatments. Coloured microhabitats mark experimental crevices (sizes in graph) and crowns (12 mm × 12 mm). Recruits were counted in each of these coloured microhabitats.

A total of 81 microhabitat tiles were created, 27 of each crevice size. These tiles were deployed on Lighthouse Reef on 23rd March 2018. The tiles were distributed equally within cages, partial cages and open plots, with three tiles, one of each crevice type, per plot (Figure 4.2). After two weeks, six cameras (GoPro Hero 3+, GoPro Inc, San Mateo, California USA) were deployed on open plots for two hours on a single day to observe feeding behaviour of herbivorous fishes on the tiles. For each discrete feeding activity, we recorded the fish species, the size of the individual and the number of

65 bites taken within each microhabitat type. This observation was conducted on six sets of tiles (where one set includes one large crevice tile, one medium crevice tile and one small crevice tile). The tiles were retrieved after three weeks, Lobophora recruits were counted under the microscope and the number of recruits in the different microhabitat types was recorded. Random recruits identified as Lobophora were sent for genetic analysis and their identity was confirmed.

Potential confounding factors in this manipulative experiment were grazing by mobile invertebrates (i.e. hermit crabs and snails), and small resident fish (i.e. blennies) as well as sedimentation. To account for these, we deployed 5 replicate sets of each herbivory treatment (caged, partially caged, and open plots) with a large, medium and small crevice tile in each set-up (see Figure 4.2) on 6th May 2018. The plots were left on the reef for 11 days to allow conditioning and turf algal growth. The abundance and location of fishes and invertebrates were assessed after 11, 14, 17, 24, 28, and 32 days. A diver first counted the resident fish, before a second diver counted the invertebrates. Additionally, sediment load was visually assessed by the same diver at each time point and rated on an increasing scale of 1 to 5, with 1 being almost no visible layer of sediment (i.e. a ‘fine dusting’) and 5 being a thick layer. As these estimates were subjective, the same person assessed the sediment loading throughout the experiment to ensure consistency.

Further information on the influence of blennies and invertebrates can be found in the Supplementary Information.

4.3.3 Statistical analysis Statistical analyses were performed using the program R (version 3.5.0, http://www.r-project.org) unless otherwise stated. All generalized linear mixed effects models were computed using ‘lme4’ (Bates et al. 2014) or ‘glmmTMB’ (Brooks et al. 2017).

Using the data acquired during the microhabitat survey, a chi-squared analysis was performed using microhabitat categories as incremental steps of 20 mm × 20 mm (width × length) to investigate whether the distribution of Lobophora within surveyed microhabitats was significantly different from random. The microhabitat size categories were chosen because using smaller microhabitat categories would have only increased the number of observed microhabitats by 30% but would have increased the number of categories without replication (only one microhabitat per category) by ca. 230%. Bigger microhabitat categories would not allow a discrimination between species, as microhabitats would be too large to exclude most fish species.

66 Additionally, a binomial generalized linear mixed effects model was run on this data set to determine whether the probability of Lobophora occurrence is related to an estimate of potential fish grazing pressure. Potential grazing pressure was calculated by multiplying the access by a fish group (1 for access, 0 for no access) with the biomass of the fish group observed on the reef (g m-2) in 2014 (PJ Mumby, unpublished data). To test whether a fish group helped explain the probability of Lobophora occurrence, each fish group was successively removed from the complete grazing pressure model. If the reduced model excluding a certain fish group had a higher AIC than the full model (difference > 2), the fish group was considered to contribute to the removal of Lobophora as the full model including the respective fish group had a smaller AIC and thus a better fit than the reduced model.

To analyse the data collected on Lobophora recruit establishment, a generalized linear mixed effects model with Poisson error distribution was used to predict Lobophora recruit numbers using the interacting fixed factors microhabitat type and herbivory treatment (cages, partial cages or open plot). The microhabitat area was standardized by adding an offset for area in the model and tile ID was nested within Plot ID as random effects. An observation-level random effect was included to account for overdispersion (Harrison 2014). Post-hoc multiple comparisons using Tukey tests were conducted with the ‘lsmeans’ package in R (Lenth 2016) and comparisons were carried out between herbivory treatments (cage vs. partial cage vs. open) for each microhabitat type and between microhabitat types (crown vs. large crevice vs. medium crevice vs. small crevice) within each herbivory treatment.

The video observations of fish feeding behaviour on tiles showed overall grazing pressure and species-specific access to microhabitat type. Size-adjusted bites were calculated (total bites × body mass in kilogram) to account for the bite size of fish using published length-weight relationships (Kulbicki et al. 2005) for all fish species observed. The size-adjusted bites were then compared between the microhabitat types using a univariate PERMANOVA based on a Euclidian distance matrix. Plotting fish bites over microhabitat type (Figure 4.6B) indicated that fish size may play a role in the fish’s ability to access a microhabitat. Therefore, a binomial generalized linear mixed effects model was fitted to estimate the probability of a fish taking a bite within a microhabitat, using microhabitat type and body length as predictors with an interaction term. Fish individual was nested within species and tile number nested within set-up ID as random effects. Using the data collected on confounding factors, the sediment load was analysed with a generalized linear mixed effects model with poisson distribution. Herbivory treatment and crevice type were set as interacting factors and plot ID (unique ID given to each set-up of three tiles with one crevice type each; see Figure 4.2) as random effect. Post-hoc multiple comparisons using Tukey tests across

67 treatments and crevice types were performed with the ‘emmeans’ package in R (https://github.com/rvlenth/emmeans).

4.4 Results 4.4.1 Physical factors driving microhabitat accessibility by fish and consequences for Lobophora abundance In the survey, Lobophora occurred most frequently in small crevices (~ 20 mm wide and ~ 20 mm deep Figure 4.3B). This was supported by ChiSquare analysis which revealed that compared to the overall distribution of measured microhabitats, Lobophora was found in a higher proportion than expected in small crevices (c2 = 16.2), whereas its abundance was found lower than expected in large crevices and open microhabitats (c2 = 7.7) (Figure 4.3). Here, subtracting the densities of all microhabitats (Figure 4.3A) from the densities of the microhabitats with Lobophora (Figure 4.3B) showed positive values where Lobophora was disproportionately abundant and negative values where it was disproportionately rare (Figure 4.3C). Density plots also showed that the smallest crevices excluded all herbivorous fishes models to varying degrees (Figure 4.3D).

Figure 4.3: A – C) Densities of observed and measured microhabitats in the field. All densities are proportionate within their category (overall or with Lobophora). The colour bar scale represents the density of each distribution, with reds indicating highest densities, whites representing zeros and greens indicating negative densities as can be seen in C. A) Densities of all microhabitat sizes measured; B) Densities of microhabitat sizes occupied by Lobophora; C) Difference between B) and A). Red tinged values indicate higher densities of microhabitats (A and B) and a higher density of microhabitats supporting Lobophora than overall microhabitats (C). D) The microhabitats which exclude individuals of a fish species are circled. The model size of each fish species causing the largest exclusion area is shown.

69 Correspondingly, Lobophora likelihood within surveyed microhabitats declined with increasing potential grazing pressure (glmm, z = -4.548, p < 0.001, Table S4.3; Figure 4.4). Comparing model fits using AIC while systematically removing fish groups from the grazing pressure model revealed that only parrotfish access was associated with Lobophora likelihood because adding parrotfish access to the model increased the model fit (lower AIC).

Figure 4.4: Lobophora likelihood as a function of grazing pressure. Grazing pressure was calculated by scaling the access by each fish figure with the biomass of that fish species on the reef. Data displayed is model output data. Points are jittered along y-axis (x-axis differences are true variation) and are raw data. Blue ribbon marks 95% confidence interval.

70 4.4.2 Influence of microhabitat type and crevice size on Lobophora recruitment and survival The manipulative field experiment revealed that in the absence of grazing (cages), Lobophora recruit density was highest on crowns with 0.38 recruits per cm2, followed by 0.2 recruits per cm2 established within large crevices (glmm, z = -0.736, p < 0.01, Table S4.4) and thereafter 0.05 recruits per cm2 in medium (glmm, z = -7.168, p < 0.01, Table S4.4) and 0.02 recruits per cm2 in small crevices (glmm, z = -2.556, p < 0.01, Table S4.4; Figure 4.5). Post-hoc multiple comparisons of recruit numbers in microhabitat types within cages revealed that medium and small crevices were not significantly different from each other (Tukey test, z = 1.704, p = 0.32, Table S4.5). In contrast, when fish had access to the tiles, recruit numbers were the same in all microhabitat types (Tukey test, p > 0.05 all, except the medium crevice in partial cages, Table S4.5).

Microhabitats had a different effect on Lobophora recruitment and post-settlement survival depending on whether fish had access to the tiles or not, as Lobophora recruit density in microhabitats was influenced by the interacting effects of microhabitat type and herbivory treatment (glmm, p < 0.05 for all microhabitat types between caged treatments vs. open treatments and between caged treatments vs. partial treatments, Table S4.4). The only exception was the effect of large crevices in open treatments, which was marginally significant (glmm, z = 1.955, p = 0.051, Table S4.4). There was no caging effect as open and partially caged plots were not significantly different from each other (crowns: glmm, z = 0.737, p = 0.74; large crevices: glmm, z = -0.388, p = 0.92; medium crevices: glmm, z = 2.231, p = 0.07; small crevices: glmm, z = 1.013, p = 0.57, Table S4.5).

On crevice tiles, sedimentation showed an increasing trend with decreasing crevice size (large – medium and small – medium glmm and Tukey test, p < 0.001 for all herbivory treatments, Table S4.6), which was consistent among caged and uncaged treatments (Figure 4.5B). However, small and medium crevices were not significantly different from each other (Caged: Tukey test, z = -0.592, p > 0.05; Partially caged: Tukey test, z = -0.529, p > 0.05; Open: Tukey test, z = -0.529, p > 0.05, Table S4.6). Sediment load in crevice types differed between open and caged (Tukey test, p < 0.05 for all crevice types, Table S4.6) as well as partially caged and caged treatments (Tukey test, p < 0.05 for all crevice types, Table S4.6). In contrast, open and partially caged treatments showed similar sediment loads within large (Tukey test, z = -0.004, p > 0.05), medium (Tukey test, z = -0.004, p > 0.05), and small crevices (Tukey test, z = -0.004, p > 0.05, Table S4.6).

Figure 4.5: A) The density of Lobophora recruits found in different microhabitat types as estimated by a generalized linear mixed effects model, with an offset included for microhabitat area. Each tile had one crevice type (large crevice, medium crevice or small crevice) and crowns (see Methods: Microhabitat types). Lower case letters indicate significantly different results within herbivory treatments. Upper case letters indicate significant differences of a microhabitat type (according to the number and colouration) between different herbivory treatments. Errorbars are standard error as calculated by the model. B) Mean sediment load estimated. Small letters indicate significantly different results across all panes. Errorbars are standard error.

On microhabitat tiles, size-adjusted grazing pressure was highest on crowns, followed by large crevices (univariate PERMANOVA, t = 4.705, p ≤ 0.001), which experienced higher grazing pressure than small (univariate PERMANOVA, t = 5.186, p ≤ 0.001) crevices. Grazing pressure in large crevices was only marginally higher than in medium crevices (univariate PERMANOVA, t = 1.866, p = 0.054), whereas small and medium crevices did not differ (univariate PERMANOVA, t = 1.354, p > 0.05; Figure 4.6A, ). Most bites were taken by Blenniidae on open tiles. Juvenile surgeonfish Ct. binotatus (mean size 3.7 cm) and Blenniidae (mean size 2.4 cm) took most bites in the crevices, followed by A. nigrofuscus and Z. scopas, which both took some bites in crevices. All other observed fish species took few - if any - bites in small or medium crevices on the experimental tiles (Figure 4.6B).

Figure 4.6: A) Mean size-adjusted bites for all species combined within each microhabitat type observed in video surveys. Error bars symbolise standard error, letters symbolise significantly different results. B) Distribution of bites cm-2 min-1 within crevices by all fish species observed in video surveys. Small crevices, medium crevices and large crevices were each on different tiles. Colours indicate crevice types (and crowns) and images indicate fish species. 74

Fish size influenced the feeding behaviour (i.e., probability of a bite being taken) among tile microhabitats. Feeding on crowns was positively associated with fish body length (glmm, z = 6.531, p < 0.001, Table S4.7) and this relationship differed to that in large (glmm, z = -6.637, p < 0.001, Table S4.7), medium (glmm, z = -9.675, p < 0.001, Table S4.7) and small crevices (glmm, z = -8.095, p < 0.001, Table S4.7; Figure 4.7). In contrast, feeding in the small and medium crevices was less likely by larger fishes and the same relationship was observed for both crevice classes (no difference between small and medium crevices, glmm, z = -0.222, p = 0.824). A third relationship was observed for large crevices by which feeding was insensitive to fish body size (large crevices vs. small: glmm, z = -4.263, p < 0.001, large crevices vs. medium crevices: glmm, z = -5.188, p < 0.001).

Figure 4.7: A) Probability of a fish visiting the tiles taking a bite in different microhabitat types as a response to increasing fish size. Binomial model prediction of raw data. Ribbon signifies the 95% confidence interval. B) The graph shows the mean bite rate compared to body length of each species, including standard deviation.

75 4.5 Discussion Our results show that the macroalga Lobophora, which is often involved in phase shifts (Cheal et al. 2010), is more abundant in concealed microhabitats where there is refuge from fish herbivory. These crevices protect Lobophora recruits by reducing grazing pressure and excluding some fish species. Further, we found that smaller fish are more likely to access crevices compared to larger fish. The reduction of grazing pressure and the exclusion of some fish species and sizes from concealed microhabitats highlights the possible role of crevices as macroalgal refuges and spore sources.

Concealed microhabitats on experimental tiles experienced lower grazing pressure compared to open microhabitats (Figure 4.5, Figure 4.6), similarly observed by Loffler and Hoey (2019), and thus provided a refuge to Lobophora recruits. A careful examination of the herbivore-accessible and herbivore-inaccessible recruitment rates is necessary to interpret the patterns among treatments in our manipulative field experiment. When herbivorous fishes were excluded (in caged treatments), Lobophora recruit density was highest on exposed crowns, and recruit numbers decreased with decreasing crevice size when microhabitat area was accounted for. This settlement pattern is in contrast to previous findings which found that settlement was highest within crevices (Loffler and Hoey 2019). This may be due to the different physical structure of the settlement tiles used between the two studies: 3 mm deep, 4 mm wide, 110 mm long crevices (Loffler and Hoey 2019) vs. 10 mm deep crevices, which varied in width and length between 8 mm and 40 mm (this study). These physical differences may have yielded disparities in micro-scale flow environments, which could be sufficient to influence macroalgal spore settlement.

The observed settlement pattern in this study has multiple possible causes. First, in the absence of grazing, visual estimates of sediment loads increased with decreasing crevice size, which corresponds with the decreased number of settled Lobophora recruits in smaller crevices. Sedimentation is detrimental to algal spore settlement (Mshigeni 1978) and thus could explain the observed settlement pattern in our study. Second, physical influences, such as eddies, which can occur around upright structures (Gaylord et al. 2002) could have influenced the settlement of spores. A third possible explanation for the observed patterns could be preferential settlement. While brown macroalgae often have motile spores, which can detect and choose substrate toughness, chemistry and are phototactic (e.g., Christie and Shaw 1968; Müller et al. 1979; Watanuki and Yamamoto 1990), the order Dictyotales, which Lobophora belongs to does not have motile spores (Fletcher and Callow 1992). It is thus unlikely that Lobophora spores can choose settlement substrate. Since the sedimentation trend is identical between cages, partial cages and open plots, we infer that undisturbed Lobophora recruit settlement and/or post-settlement survival should have been the same among the different plot types 76 as well. Thus, in partial cages and open plots, Lobophora recruitment and/or post-settlement survival on crowns would have been higher than in large crevices. Large crevices, in turn, would have had higher recruit densities than small and medium crevices, whereas the latter would have had the same recruit density. Comparing this expected recruit settlement and/or post-settlement survival to the observed recruit densities in open plots and partial cages indicates that recruit survival is negatively associated with crevice size. Below a width of 15 mm there does not seem to be any further protection from herbivory, which is in accordance with the in-situ observation of highest Lobophora likelihood in the smallest microhabitat category of 20 mm depth and 20 mm width.

Correspondingly, on the reef, Lobophora was found disproportionately more often in small concealed microhabitats, which excluded most fish species tested and disproportionately less in large crevices and open microhabitats which allowed access to all fish groups. Concealed microhabitats experience less grazing pressure (e.g., Bergey 2005; Shitzer et al. 2008; Bennett et al. 2010) and thus provide refuges for benthic species. Extremely high Lobophora recruit mortality in open areas (Diaz-Pulido and McCook 2003) may prohibit the establishment of extensive Lobophora cover on easily accessible substrate under high grazing pressure and the alga is thus often restricted to concealed microhabitats (Bennett et al. 2010). However, when conditions change and grazing pressure is reduced, the alga can likely spread to open areas from these concealed microhabitats instead of having to recruit from other reefs, which could enable it to form rapid blooms (Roff et al. 2015b).

While overall grazing pressure was lower in smaller crevices, species-specific functional differences are apparent (Figure 4.6). While most species took limited bites within small- and medium-sized crevices, Blenniidae and the surgeonfish Ct. binotatus (the two smallest species), took a similar number of bites in all microhabitats. Indeed, body size proved to be an important predictor for the number of bites taken within crevices, with smaller fish accessing concealed microhabitats more frequently, indicating the possibly important role of small bodied-fish to control macroalgae (Kuempel and Altieri 2017). Ensuring a diverse range of fish sizes on a reef may thus be important for effective algal control.

The ability of fish species to remove Lobophora likely vary. We expected neither Ct. binotatus nor Blenniidae to be able to remove adult thalli, with contradictory evidence of the ability of Blenniidae to consume macroalgal recruits (Marshell 2014; Loffler and Hoey 2019). While Salarius fasciatus was unable to remove Sargassum recruits in tank experiments (Marshell 2014), individuals of the genus Ecsenius are thought to have contributed to the mortality of Sargassum recruits in-situ (Loffler and Hoey 2019), highlighting the need for further research into the role of Blenniidae as grazers of

77 macroalgal recruits. The surgeonfishes A. nigrofuscus, Ct. striatus and Z. scopas took a large majority of their bites on exposed reef areas, but all three species did access crevices to some extent. Especially grazing species like A. nigrofuscus can be expected to remove Lobophora recruits and both A. nigrofuscus and Ct. striatus have been shown to remove recruits of a different macroalga (Sargassum sp.; Marshell 2014). These species could thus be able to reduce Lobophora recruit densities within concealed microhabitats. Still, our overall results highlight, that the reduction in grazing pressure does result in significantly reduced recruit mortality within concealed habitats even though these five species access crevices. Their ability to efficiently control Lobophora within crevices may thus be limited.

Parrotfishes on the other hand, such as S. niger and C. spilurus, did not access small or medium crevices at all. Interestingly, our analysis of the surveyed microhabitat accessibility on the reef infers that Lobophora abundance is likely to be controlled by parrotfish rather than other reef fish species. Indeed, parrotfish have been shown to remove adult Lobophora at the study site (Roff et al. 2015). However, parrotfish are unlikely to target Lobophora as recent research has identified them as microphages (Clements et al. 2016), which may consume various species of algae in the process of acquiring their nutritional targets: microorganisms, including cyanobacteria. Therefore, Lobophora removal may be driven by incidental ingestion. Given the lack of feeding in small and medium crevices, parrotfish mediated control of Lobophora, while important, may be restricted to easily accessible reef substrate.

Given the short conditioning time of tiles (~ 2 weeks), the benthic communities on tiles were likely in a very early successional state and this may have influenced the bite rates of some fish species. However, we do not expect the ability of fish to access crevices to be influenced by the successional state of the benthic community as all crevices were deployed on the reef at the same time. Therefore, while we do not assume our species-specific bite rates to necessarily be representative of feeding rates on the reef, we conclude that the relative bite rates within microhabitats should hold true. Overall, our findings thus highlight that a possibly wide range of fish species and sizes is necessary to efficiently control Lobophora and likely other macroalgae on a reef scale.

While Lobophora recruits on experimental tiles experienced increased protection from fish herbivory with decreasing crevice size, recruit density was similar in all microhabitat types on herbivore- accessible experimental tiles. In contrast, adult Lobophora on the reef was more commonly found in smaller crevices. There appears, therefore, to be a discrepancy between patterns of recruitment and subsequent adult distribution. It is possible that the experimental treatment to isolate herbivory – i.e.,

78 the use of cages – led to an unrealistically low level of recruitment in small crevices, which we attribute to the collection of sediment (Mshigeni 1978), perhaps because of reduced flushing in the cage or experimental crevices compared to reef crevices. Indeed, recruit densities were greater in small crevices open to full grazing (without cages). Yet, even if recruitment is consistent across crevice sizes under natural grazing, the skew of adult Lobophora distribution towards smaller crevices implies that post-recruitment mortality must be higher in larger crevices. The greater accessibility of such microhabitats to fish may well account for this. These findings imply that sedimentation on the reef may be highly sensitive to even minor flow impairment, such as created by cages or the crevices on the microhabitat tiles. Alternatively, the short experimental time of three weeks possibly caused reduced competition for Lobophora recruits compared to the conditions experienced on the reef. Here, competition effects will likely influence the abundance and distribution of Lobophora as competition in crevices can be intense (Day 1983). These differences in competition may thus lead to the observed differences in Lobophora distribution between experimental tiles and reef microhabitats. Another possibility is that the short timescale of recruitment measured here might have only been sufficient to reduce recruit densities to be even among experimental microhabitats but prolonged exposure to herbivores may cause further mortality in larger crevices and open areas, whereas recruits in smaller crevices would stay protected. This is more likely if Lobophora recruitment is episodic and not continually resupplied.

The refuge crevices provide shape consumer-producer interactions and is an important nuance of coral reef functioning because they influence the benthic community structure (Crowder and Cooper 1982; Brandl and Bellwood 2016). Our study shows that the macroalga Lobophora, which can be harmful to corals, thrives in concealed microhabitats due to reduced herbivory. A main feature of this exclusion is that the fish groups most likely to remove adult Lobophora are excluded from smaller concealed microhabitats. Further, fish size seems to be an important predictor of the fish’s ability to access crevices. Because of overall reduced herbivory and the restricted access of important nominal herbivores, crevices may act as refuges for macroalgae in which they can develop into adults. When grazing pressure on a reef is reduced, for example due to mass coral mortality or intensified herbivore fishery, the alga may be able to spread from concealed microhabitats to open areas, forming a bloom. The control of Lobophora thus depends on the interactions between fish species and their environment. Protecting a diverse suite of fish species and sizes able to access a variety of microhabitats may thus be critical to prevent the rise of a potentially potent macroalga to levels that inhibit coral recovery.

79 4.6 Acknowledgements We are grateful to Kelly Wong, Shannen Smith, Brooke Brown, and Nicole Cernohorsky for their help during field work. We would also like to thank the Palau International Coral Reef Center for hosting us. This research was funded by the Winnifred Violet Scott Trust and by ARC grants awarded to P. Mumby.

4.7 Supplementary

Development and usage of fish shapes for microhabitat survey Table S4.1: List of fish species and measurements used to build fish shapes for microhabitat survey. These fish shapes were later used to discern the accessibility of microhabitats to the respective fish species. Morphological measurements for these shapes were obtained from Brandl et al. (2015) and missing species were measured using pictures provided by Sonia Bejarano or obtained online (S. vulpinus: www.australienmuseum.net.au) following the same measurement protocol as used in Brandl et al. (2015).

Species Size Body Head Snout Head Snout Head (cm) depth length length depth angle angle (°) (cm) (cm) (cm) (cm) (°) Ctenocaetus striatus 5 2.6 1.3 0.7 2.5 0.3 173.3 10 5.1 2.6 1.5 4.9 28.6 173.3 15 7.7 3.9 2.2 7.4 0.6 173.3 Naso lituratus 10 4.6 2.6 1.5 4.4 84.4 189.0 15 6.9 3.9 2.3 6.6 84.4 189.0 20 9.2 5.2 3.1 8.9 84.4 189.0 Zebrasoma scopas 5 3.1 1.4 0.9 2.8 71.5 200.4 10 6.2 2.9 1.7 5.5 71.5 200.4 "general parrotfish shape" 10 3.7 2.9 0.9 3.4 121.6 148.3 20 7.3 5.9 1.8 6.8 114.4 148.3 30 11.0 8.8 2.7 10.1 107.2 148.3 Chlorurus microrhinos 20 7.6 5.4 1.7 7.5 109.2 158.7 30 11.7 9.7 3.7 10.9 118.6 121.6 Siganus vulpinus 10 4.6 2.7 1.7 3.9 56.6 185.7 15 6.8 4.0 2.6 5.8 56.6 185.7

Assessing the importance of microhabitat depth vs. width for excluding fish species Methods – Statistics To determine whether microhabitat depth or width was more important in restricting fish access, we used the data from the microhabitat survey and ran separate binomial generalized linear mixed effects models with access by each fish species as response and width and depth as additive fixed effects. Subquadrat was nested within quadrat as a random effect.

Results Microhabitat width restricted fish access more often than microhabitat depth (glmms, output see Table S4.2).

Table S4.2: Influence of microhabitat width and depth on access by fish groups. Generalized linear mixed effects models with microhabitat width and depth as additive predictors were run. a-level set at 0.05.

Fish group Size (cm) Access significantly Access significantly restricted by depth? restricted by width? Chlorurus 30 No (z =-1.885, p = 0.06) Yes (z = 7.742, p < 0.001) microrhinos 20 No (z = -0.468, p = 0.64) Yes (z = 7.260, p < 0.001) General parrotfish 10 No (z = 0.781, p = 0.44) Yes (z = 4.947, p < 0.001) 20 No (z = -0.163, p = 0.87) Yes (z = 8.479, p < 0.001) 30 No (z= -0.107, p = 0.91) Yes (z = 9.454, p < 0.001) Naso lituratus 15 Yes (z = -2.647, p < 0.01) Yes (z = 3.337, p < 0.001) 20 No ( z = -1.484, p = 0.14) Yes (z = 4.768, p < 0.001) 25 Yes (z = -2.823, p < 0.01) Yes (z = 5.645, p < 0.001) Siganus vulpinus 10 No (z = -0.480, p = 0.63) No (z = 0.366, p = 0.71) 15 No (z = -1.665, p = 0.10) No (z = 0.757, p = 0.45) Ctenochaetus 10 No (z =0.373, p = 0.71) Yes (z = 4.150, p < 0.001) striatus/Acanthurus 15 No (z = 0.663, p = 0.51) Yes (z = 6.391, p < 0.001) nigrofuscus 20 No (p = -0.607, p = 0.54) Yes (z = 7.716, p < 0.001) Zebrasoma scopas 5 No (z = 1.252, p = 0.21) No (z = 1.530, p = 0.13) 10 No (z = -0.389, p = 0.70) No (z = 1.083, p = 0.28)

81 Grazing pressure within herbivory treatments in the manipulative experiment Methods To compare overall grazing rates among herbivory treatments, cameras were deployed to observe feeding rates on each plot set up for testing the abundance of blennies and invertebrates after 32 days. Five GoPros were deployed with each camera recording fish feeding behaviour on three plots at a time, including one caged plot, one partially caged plot and one open plot. The grazing pressure in plots of different herbivory treatments was predicted using a generalized linear mixed-effects model with negative binomial error distribution. The number of bites taken on tiles in a plot was the response variable and herbivory treatment (caged, partially caged, open plots) the predictor. Plot was included as random effect.

Results Grazing pressure within caged plots was lower (2.24 Bites h-1 ± 1.11 (se)) compared to partially caged (336.0 Bites h-1 ± 103.0 (se); lm, t = 7.603, p < 0.001) and open plots (169.0 Bites h-1 ± 21.9 (se); lm, t = 6.057, p < 0.001). Partially caged and open plots were not significantly different from each other (Tukey posthoc, z = -1.546, p = 0.13).

Alternative hypotheses to explain observed recruit establishment patterns in the manipulative experiment Methods Apart from sediment load (see main text), other potential confounding factors in the manipulative field experiment were grazing by mobile invertebrates and small resident fishes. The analysis was conducted on count data of blennies, hermit crabs and snails. Results were averaged over time as we were interested in the average numbers and not whether any increase or decrease was observed. Generalized linear mixed effects models with Poisson distribution were run for blennies, hermit crabs and snails using herbivory treatment (caged, partially caged, open) and crevice type (large, medium, small) as predictors and plot ID as random effect.

Results Snails and hermit crabs did not differ among herbivory treatments (Wald test, snails: ChiSq = 2.3292, df = 2, p = 0.31, hermit crabs: ChiSq = 3.7386 , df = 2, p = 0.15) nor did they differ between microhabitat types (Wald test, snails: ChiSq = 0.1247 , df = 2, p = 0.94, hermit crabs: ChiSq = 4.2172, df = 2, p = 0.12; Figure S4.1). Blennies were less abundant in open treatments compared to partially caged ones (glmm, z = -2.253, p = 0.02) and marginally more abundant than in caged treatments (p = 0.07). Blennies were also more abundant on medium crevice tiles compared to large (glmm, z =

82 2.046, p = 0.04) but they were as common on small crevice tiles as in medium ones (Tukey test, z = 1.361, p = 0.32). Video observations showed that blennies moved regularly between microhabitat types within a herbivory treatment and they took even numbers of bites in all microhabitats (Figure S4.1).

Figure S4.1: Abundance of blennies, snails, and hermit crabs found on microhabitat tiles within herbivory treatments. Note the different scales. Errorbars show standard error.

83 Discussion Neither snails nor hermit crabs showed any significant abundance differences among treatments and were therefore unlikely to have played any role in the recruitment of Lobophora. Blennies were more common in cages, then they were on open treatments, likely because they gained protection from predators. If blennies were able to remove Lobophora recruits, they may have been responsible for the low recruitment in small crevices of caged plots, which was below the recruitment observed in small crevices of open plots. However, given our results that Lobophora was more abundant in small crevices, which were fed on equally to other microhabitats by blennies, it is unlikely that blennies influenced differences between Lobophora recruit abundance in our experiment. Based on our observations, neither of these possible alternative hypotheses can explain the difference in recruitment pattern in caged treatments compared to partially caged and open treatments. We are therefore certain that the observed differences are due to roving herbivorous fish feeding activity.

Statistical analysis tables Table S4.3: Binomial generalized linear mixed effects model (glmm) output analyzing Lobophora likelihood as a function of potential grazing pressure. Bold print under p-value symbolizes significant results.

Analysis Test used Fixed effect Estimate Std. Error Z-value P-value Lobophora Binomial Intercept -0.127 0.287 -0.443 0.658 likelihood glmm Grazing -0.013 0.003 -4.548 < 0.001 pressure

Table S4.4: Poisson generalized linear mixed effects model (glmm) output analyzing Lobophora recruit settlement and/or post-settlement survival in microhabitat. Bold print under p-value symbolizes significant results.

Analysis Test used Fixed effect Estimate Std. Z- P-value Error value Lobophora Poisson Open -1.666 0.255 6.535 < 0.001 recruit glmm microhabitats in settlement with cages and/or post- dummy (Reference settlement variable level)

84 survival in Open -1.005 0.368 -2.73 0.006 microhabitats microhabitats in open plots Open -1.287 0.374 -3.442 < 0.001 microhabitats in partial cages Large crevice in -0.736 0.227 -3.241 0.001 cages Medium crevice -1.816 0.253 -7.168 < 0.001 in cages Small crevice in -2.556 0.367 -6.959 < 0.001 cages Large crevices 0.711 0.363 1.955 0.051 in open plots Large crevices 1.178 0.339 3.479 < 0.001 in partial cages Medium 1.769 0.340 5.204 < 0.001 crevices in open plots Medium 0.975 0.407 2.396 0.017 crevices in partial cages Small crevices 2.512 0.435 5.774 < 0.001 in open plots Small crevices 2.318 0.467 4.960 7< 0.001 in partial cages

85 Table S4.5: Post-hoc Tukey test output analyzing Lobophora recruit settlement and/or post- settlement survival in microhabitat. Bold print under p-value symbolizes significant results.

Analysis Test used Contrast Estimate Std. Z- P-value Error value Lobophora Post-hoc Caged plots 0.736 0.227 3.241 0.007 recruit Tukey test Crown vs. large settlement crevice and/or post- Caged plots 1.816 0.253 7.168 <0.001 settlement Crown vs. survival in medium crevice microhabitats Caged plots 2.556 0.367 6.959 <0.001 crown vs. small crevice Caged plots 1.080 0.328 3.290 0.006 Large crevice vs. medium crevice Caged plots 1.819 0.410 4.435 < 0.001 large crevice vs. small crevice Caged plots 0.739 0.433 1.704 0.321 medium crevice vs. small crevice Open plots 0.026 0.276 0.094 0.999 Crown vs. large crevice Open plots 0.047 0.226 0.210 0.997 Crown vs. medium crevice Open plots 0.0427 0.242 0.176 0.998 crown vs. small crevice Open plots 0.022 0.328 0.066 0.999 Large crevice vs. medium crevice

86 Open plots 0.017 0.340 0.050 1.000 Large crevice vs. small crevice Open plots -0.005 0.298 -0.016 1.000 Medium crevice vs. small crevice Partial cages -0.442 0.262 -1.684 0.332 Crown vs. large crevice Partial cages 0.841 0.319 2.634 0.042 Crown vs. medium crevice Partial cages 0.238 0.289 0.822 0.844 Crown vs. small crevice Partial cages 1.283 0.381 3.370 0.004 Large crevice vs. medium crevice Partial cages 0.679 0.353 1.922 0.219 Large crevice vs. small crevice Partial cages -0.603 0.395 -1.528 0.421 Medium crevice vs. small crevice Crowns 1.005 0.368 2.730 0.017 Caged plots vs. open plots Crowns 1.287 0.374 3.442 0.002 Caged plots vs. Partial cages Crowns 0.282 0.383 0.737 0.742 open plots vs. partial cages Large crevices 0.294 0.474 0.621 0.809

87 Caged plots vs. open plots Large crevices 0.109 0.452 0.241 0.969 Caged plots vs. partial cages Large crevices -0.1852 0.478 -0.388 0.921 Open plots vs. partial cages Medium crevices -0.764 0.457 -1.674 0.215 Caged plots vs. open plots Medium crevices 0.312 0.503 0.620 0.809 Caged plots vs. partial cages Medium crevices 1.076 0.482 2.231 0.066 Open plots vs. partial cages Small crevices -1.508 0.532 -2.837 0.013 Caged plots vs. open plots Small crevices -1.031 0.554 -1.862 0.150 Caged plots vs. partial cages Small crevices 0.477 0.471 1.013 0.569 Open plots vs. partial cages

88 Table S4.6: Poisson generalized linear mixed effects model (glmm) output analyzing possible confounding factors for Lobophora recruit settlement and/or post-settlement survival in microhabitat. Bold print under p-value symbolizes significant results.

Analysis Test Fixed effect / contrast Estimate Std. Z-value P-value used Error Possible Poisson Caged large crevices 0.732 0.111 6.587 < 0.001 confounding glmm (reference level) factors - Open plots large -0.383 0.133 -2.874 0.004 Sediment crevices Partial cages large -0.382 0.133 -2.872 0.004 crevices Caged medium crevices 0.370 0.098 3.763 < 0.001 Caged small crevices 0.417 0.098 4.275 < 0.001 Posthoc Large crevices 0.383 0.133 2.874 0.011 Tukey Caged plots vs. open test plots Large crevices 0.382 0.133 2.872 0.011 Caged plots vs. partial cages Large crevices -0.001 0.139 -0.004 1.000 Open plots vs. partial cages Medium crevices 0.383 0.133 2.874 0.011 Caged plots vs. open plots Medium crevices 0.382 0.133 2.872 0.011 Caged plots vs. partial cages Medium crevices -0.001 0.139 -0.004 1.000 Open plots vs. partial cages Small crevices 0.383 0.133 2.874 0.011 Caged plots vs. open plots

89 Small crevices 0.382 0.133 2.872 0.011 Caged plots vs. partial cages Small crevices -0.001 0.139 -0.004 1.000 Open plots vs. partial cages Cages -0.370 0.098 3.763 < 0.001 Large crevices vs. medium crevices Cages -0.417 0.098 -4.275 < 0.001 Large crevices vs. small crevices Cages -0.047 0.088 -0.529 0.857 medium crevices vs. small crevices Open plots -0.370 0.098 3.763 < 0.001 Large crevices vs. medium crevices Open plots -0.417 0.098 -4.275 < 0.001 Large crevices vs. small crevices Open plots -0.047 0.088 -0.529 0.857 Medium crevices vs. small crevices Partial cages -0.370 0.098 3.763 < 0.001 Large crevices vs. medium crevices Partial cages -0.417 0.098 -4.275 < 0.001 Large crevices vs. small crevices Partial cages -0.047 0.088 -0.529 0.857 Medium crevices vs. small crevices

90 Table S4.7: Binomial generalized linear mixed effects model (glmm) output analyzing the influence of fish body size on the ability to access different microhabitat types. Bold print under p-value symbolizes significant results. * symbolizes interactions of fixed factors.

Analysis Test used Fixed effect Estimate Std. Error Z-value P-value Influence of Binomial Bites on crown -4.170 0.252 -16.540 < 0.001 fish size for glmm – (reference level) microhabitat crown as Bites in large -1.335 0.239 -5.596 < 0.001 accessibility reference crevices level Bites in medium -0.107 0.237 -0.452 0.651 crevices Bites in small -0.043 0.273 -0.158 0.874 crevices Body length 0.194 0.030 6.531 < 0.001 Bites in large -0.198 0.030 -6.637 < 0.001 crevices * Body length Bites in medium -0.448 0.046 -9.675 < 0.001 crevices * Body length Bites in small -0.434 0.054 -8.095 < 0.001 crevices * Body length Binomial Bites in small -4.207 0.336 -12.515 < 0.001 glmm – crevices small (reference level) crevice as Body length -0.241 0.056 -4.277 < 0.001 reference Bites on crown * 0.435 0.054 8.104 < 0.001 level Body length Bites in large 0.237 0.055 4.281 < 0.001 crevice * Body length Bites in medium -0.014 0.064 -0.222 0.824 crevice * Body length

91 Binomial Bites in large -5.491 0.296 -18.559 < 0.001 glmm – crevices large (reference level) crevice as Body length -0.006 0.032 -0.187 0.852 reference Bites in medium -0.249 0.048 -5.188 < 0.001 level crevices * Body length Bites in small -0.235 0.055 -4.263 < 0.001 crevices * Body length Bites on crowns 0.198 0.030 6.629 < 0.001 * Body length

______

CHAPTER 5: INFLUENCE OF OVERFISHING AND STRUCTURAL COMPLEXITY ON LOBOPHORA COVER: AN INDIVIDUAL-BASED MODELLING APPROACH ______

93 Influence of overfishing and structural complexity on Lobophora cover: an individual-based modelling approach

5.1 Abstract Coral reefs are increasingly shifting towards macroalgal-dominated states, because of global and local anthropogenic pressures, such as large-scale bleaching, storms and overfishing. Once established, macroalgae interact with corals and can reduce coral larval settlement, thus interfering with coral reef recovery. Whether macroalgae can dominate a reef depends on interactions between the algae and their consumers, most often herbivorous fishes. Reductions in herbivorous fish grazing pressure often lead to macroalgal dominance on reefs. However, the ability of herbivorous fishes to control macroalgae depends on fish species-specific abilities to remove and digest algae. Whether a fish can remove a macroalgal thallus is further influenced by the microhabitat the alga inhabits. Concealed microhabitats exclude some herbivorous fish species and reduce overall grazing pressure, thereby providing a refuge to algae. These refuges may allow algae to spread onto open reef areas when grazing pressure is reduced, for example through overfishing. We developed an individual-based model using the common brown macroalga Lobophora as a study organism to test the effect of overfishing and structural complexity on macroalgal trajectories. To parameterize the model, we combined field and experimental data of Lobophora recruitment, growth and life-stage specific mortality. Additionally, the model accounted for fish species-specific abilities to cause Lobophora mortality and access microhabitats. The model was tested against survey data of Lobophora distribution in different microhabitat types. Model simulations showed that reductions in fish biomass and increases in small-scale structural complexity lead to increases in macroalgal cover, with the strongest algal cover increases observed at 5% of the original herbivorous fish biomass. Further, overfishing of key herbivores seemed to have a slightly stronger effect on reefs with many grazing refugia. Finally, we showed that while parrotfish biomass exerted the strongest control on Lobophora cover, a diverse suite of fish species was more efficient at controlling the alga than any individual species alone. These results highlight the importance of fish species diversity to control macroalgae on a reef-wide scale and point to small-scale structural complexity (i.e., crevices) as an important agent of macroalgal cover trajectories.

5.2 Introduction Coral mass mortality is increasing in frequency because of climate change driven bleaching events as well as local stressors, such as eutrophication and overfishing (e.g., Hughes 1994; Hoegh-Guldberg et al. 2007; Hughes et al. 2018). As a consequence, macroalgae often spread and can dominate entire

94 reefs (Hughes 1994; McClanahan and Muthiga 1998). Macroalgae compete with corals and can radically reduce coral larval settlement (Kuffner et al. 2006; Evensen et al. 2019). Therefore, once a reef has shifted to macroalgal-dominance these negative effects of macroalgae on corals can inhibit the recovery of coral reefs (Hughes 1994; McClanahan and Muthiga 1998; Hoegh-Guldberg 1999).

These phase shifts to macroalgae dominance can happen if grazing pressure is reduced, either directly through overfishing of herbivorous fishes (Thacker et al. 2001; Burkepile and Hay 2008; herbivorous fishes defined here as causing disturbance to algae; sensu Steneck et al. 2017) or as a consequence of coral mass mortality (Mumby 2009; Graham et al. 2015). Additionally, increases in nutrient concentrations can be associated with increased macroalgal abundance (Fabricius et al. 2005). However, the interaction between nutrient enrichment and grazing reductions are complex (Littler et al. 2006) and the effects of nutrients on macroalgal proliferation may be stronger when herbivory is reduced (Diaz-Pulido and McCook 2003). Overall, while nutrient enrichment can have a profound influence on coral reef health (Fabricius 2005; Fabricius et al. 2005), top-down effects, such as grazing by herbivorous fishes, rather than bottom-up effects, seem to be the stronger driver of macroalgal proliferation on coral reefs (Szmant 2002; Rasher et al. 2012). After corals die, settlement space becomes available and is quickly utilized by algae. As a consequence, herbivorous fish spread out over a larger area to feed on the newly established algae, thereby effectively reducing the grazing pressure per unit area (Williams et al. 2001; Mumby 2009). Feeding intensity on individual algae is thus alleviated and macroalgal recruits are more likely to develop into adults. Therefore, a healthy herbivorous fish assemblage is considered crucial for coral reef resilience by providing sufficient grazing pressure to control macroalgae (Burkepile and Hay 2008).

The ability of herbivorous fishes to control algae is influenced by microhabitats, which shape benthic communities by influencing consumer-producer interactions (Brandl and Bellwood 2016). For example, concealed microhabitats provide protection to coral recruits, which preferentially settle here (Doropoulos et al. 2016a). Similarly, macroalgae are often associated with concealed microhabitats, which experience lower grazing pressure than exposed areas (Menge and Lubchenco 1981; Lubchenco 1983; Bennett et al. 2010; Poray and Carpenter 2014). However, the ability of fish to access concealed microhabitats is species-specific and size-specific (Brandl and Bellwood 2014, Chapter 4). Concealed microhabitats may play an important role in macroalgal trajectories as they can provide grazing refugia to macroalgae from which they can spread if conditions are favourable (Chapter 4).

95 Using a common brown macroalga, Lobophora, as a model organism, we evaluate the importance of individual fish species to prevent a shift to macroalgal dominance. Lobophora was chosen as a model organism because it is a particularly strong competitor of corals (e.g., Jompa and McCook 2002; Rasher and Hay 2010; Mumby et al. 2016) and can form rapid blooms (Roff et al. 2015b). The alga’s morphology varies between crustose to erect, ruffled thalli (Vieira et al. 2014). Tightly encrusting forms are likely difficult to remove for a majority of herbivorous fish species. Parrotfish can remove crustose forms of Lobophora (Roff et al. 2015b), but they are unlikely to specifically target them because parrotfish achieve their nutrition from microorganisms (Clements et al. 2016). However, Lobophora has a variety of epiphytes (Vieira et al. 2016b), which vary depending on the Lobophora species. These epiphytes include cyanobacteria (Fricke et al. 2011), one of the main food sources of parrotfish (Clements et al. 2016). Therefore, parrotfish probably remove adult Lobophora as well as recruits incidentally while targeting epiphytic microorganisms. In contrast, the grazing surgeonfish genera Acanthurus, Ctenochaetus and Zebrasoma are not likely to contribute to the removal of adult Lobophora because of their morphological and physiological adaptations to other resources (Horn 1989; Purcell and Bellwood 1993; Choat et al. 2004). However, at least Acanthurus spp. have been shown to control Lobophora recruits and thus the establishment of the alga (Chapter 3).

To understand the likelihood of a Lobophora bloom if herbivorous fishes were overfished, it is important to study the species-specific effects of herbivorous fishes on Lobophora, taking into account their ability to access different microhabitats. However, it is challenging to assess this in the field because of difficulties identifying algal recruits in-situ and controlling species-specific fish biomass. Here, we use an ecological modelling approach to reveal the effect of overfishing phylogenetic fish groups and to understand the role of concealed microhabitats for the establishment of Lobophora. We developed an individual-based model that simulates monthly time-steps of Lobophora recruitment, growth, and mortality taking into account fish species-specific effects. The model was parameterized using field data to reproduce patterns of Lobophora cover in three differently sized concealed and one exposed microhabitat. Using this model, we posit three research questions:

1) What is the effect of overfishing herbivorous fishes – both individual genera and all fish species combined – on Lobophora cover? 2) How do increases in spatial grazing refugia shape the trajectory of Lobophora cover? 3) Do reefs with higher structural complexity develop more Lobophora cover when the biomass of a key herbivore is reduced compared to less structurally complex reefs?

96 To answer these questions, we simulated Lobophora cover while varying the biomass of herbivorous fish genera between 5% and 100% of their original biomasses and changed the proportionate distribution of concealed and open microhabitats.

5.3 Methods 5.3.1 Model overview An individual-based model was developed to study the effects of microhabitat distribution and species-specific herbivorous fish grazing on Lobophora cover. The model simulates algal cover in four microhabitat types: three crevices and one planar ‘exposed’ microhabitat. The microhabitats are referred to as small crevices (8 mm × 15 mm, 1.25 cm2 area), medium crevices (15 mm × 15 mm, 2.25 cm2 area), large crevices (40 mm × 40 mm, 16 cm2 area) and open microhabitats (planar area, 1.0 cm2 area). The distribution of microhabitats followed the natural distribution observed on a study reef on the East coast of Palau (07°17’05.0’’N, 134°31’38.7’’E), a Pacific island nation, which resulted in small crevices occupying 5% of the simulated reef area, medium crevices occupied 14%, large crevices occupied 37% and open microhabitats occupied 44% of the reef area. The model includes Lobophora recruitment rates (Chapter 3), species-specific mortality of Lobophora recruits by herbivorous fishes (Chapter 3), Lobophora growth rate (van der Zande, Master thesis), microhabitat specific grazing rates of four different fish groups (Chapter 4), Lobophora adult natural mortality (Chapter 3) and adult herbivore facilitated mortality (Roff et al. 2015; see Table 5.1 for detailed parameters). The fish groups included are Acanthurus spp., Ctenochaetus spp., Chlorurus and Scarus spp. (in the following referred to as ‘parrotfish’) and Zebrasoma spp. The data were collected through observational surveys, manipulative field experiments and controlled tank experiments (Chapters 3 and 4). A conceptual overview of how these parameters influence Lobophora cover is presented in Figure 5.1.

The model simulates an area of 50 cm × 50 cm, which is distributed between the microhabitats as described above. All space is originally free of any organisms. The model simulates 60 monthly time steps (i.e. 5 years) of Lobophora recruitment, recruit mortality, recruit growth, adult mortality, and adult growth. During the first time-step, a single Lobophora recruitment event takes place. For each following time-step (month 2 – 60), the model simulates the following five stages in the described order: First, a grazing driven recruit mortality is applied to the existing recruits. Second, if recruits survive, they then grow using a relative growth rate (van der Zande, unpublished Master thesis). Third, a mortality is applied to adult Lobophora, both natural (Chapter 3) and herbivore driven (Roff et al. 2015b), before allowing the surviving algae to grow with the same relative growth rate up to a maximum thallus size (Vieira et al. 2014) and/or until the microhabitat specific area is filled up (4th 97 stage). If a crevice fills up (i.e., an alga reaches a size bigger than the crevice size) algal cover is allowed to grow into planar areas, where it will then assume grazing and mortality rates of the open microhabitat. Lastly, a new recruitment event is simulated if space is available. The overall area per microhabitat is summed and presented as Lobophora cover over time. Each simulation detailed below was run 20 times.

Figure 5.1: Conceptual overview of the model showing the main parameters. + indicates an increase in the following parameter, - indicates a decrease. Colours represent different processes.

5.3.2 Model parameterization 1) Recruitment: Recruitment was simulated once per month (pers conv with C Vieira) with 0.084 rec cm-2 (± 0.098 sd) using data obtained from caging of settlement tiles for three weeks at two time points (Chapter 3). A spore was assigned the size of 0.25 cm diameter, as this size would allow the recruit to grow to ~ 0.5 cm diameter within a month, which is equivalent to sizes observed under the microscope. Recruitment was assumed independent of adult population because no empirical data is available on density-dependent Lobophora recruitment. This is due to the 98 cryptic nature of Lobophora, which makes a species identification in-situ nearly impossible. The recruitment was simulated by sampling from the mean ± sd using a negative binomial distribution.

2) Growth: Growth was applied to recruits and adults every month, using a relative growth rate of 0.0217 cm2 cm-2 day-1 (± 0.0163 sd) and sampling from a normal distribution. This growth was measured by van der Zande (unpublished Master thesis) in Bonaire, with values being comparable to average values measured by Birrell (2014) on the Great Barrier Reef, Australia. When Lobophora individuals reached a size equal to the crevice size they were in, they were allowed to grow into open microhabitats. The parts of the alga in the open microhabitat then experienced grazing pressure from the respective microhabitat, whereas the parts that were smaller than the crevice size still experienced grazing pressure as observed in the crevices.

3) Recruit mortality: Mortality of recruits and adults was calculated separately based on a size threshold of 0.2 cm2. Recruit mortality was applied based on results of observed mortality over 2.5h in a tank experiment (Chapter 3) and was specific to four different fish species. Acanthurus nigrofuscus caused a proportionate decrease of recruits of 0.396, Ctenochaetus striatus of 0.25, Zebrasoma scopas of 0.119 and Chlorurus spilurus of 0.261. Recruit mortality was generated by sampling randomly from a binomial distribution using the species-specific mean probability of a mortality event. These mortalities were scaled to grazing pressure on a reef by comparing grazing rates on tiles in the tank experiment and on the reef. The mortalities were then weighted by the biomass of the respective fish group on the reef. Further, the access each fish species had to the four different microhabitat types (see above for details) was taken into account, by dividing the grazing rate in the respective microhabitat by the grazing rate on open areas. This was done as the mortality in the experiment was measured on flat areas, which are equivalent to open microhabitats. Taking into account the species-specific microhabitat access led to a reduction of grazing pressure and thus mortality in concealed microhabitats.

4) Adult mortality: Adult mortality was based on both natural mortality and herbivorous fish driven mortality. Natural proportionate mortality was 0.146 (± 0.0058) as observed at discrete timepoints four weeks apart in caged plots (Chapter 3) and herbivorous fish driven mortality was based on

99 data collected by Roff et al. (2015), which shows a 0.69 proportionate mortality of Lobophora based on parrotfish bites on the reef and scaled to one month. Adult mortality was weighted by parrotfish biomass only, as AIC comparisons of microhabitat surveys showed that only parrotfish biomass presence improved the model fit of Lobophora cover in response to grazing pressure (Chapter 4). Further, mortality was scaled by microhabitat access as explained above for recruit mortality. Further values and parameters used can be found in Table 5.1.

Table 5.1: Parameters used in the model. Details explain how parameters were calculated and mean and standard deviation (if applicable) are displayed.

Parameters Details Mean (± standard deviation) Spore size Spore size was estimated based on the 0.25 cm diameter model reaching 0.5 cm diameter after 1 month, which is equivalent to observed values at 3 weeks age. Maximum thallus Maximum thallus size based on L. 6 cm2 size gibbera, the most common Lobophora species in Palau (Vieira et al. 2014) Mortality threshold The threshold at which mortality switched 0.2 cm2 recruit to adult from recruit mortality to adult mortality was estimated based on at which size fishes are unlikely to accidentally swallow a recruit with one bite. This value is reached after about one month. Recruitment rate Recruitment rate was estimated based on 0.084 rec cm-2 mean and standard deviation (sd) sampled (± 0.098 rec cm-2) from a negative binomial distribution. Mean and sd were obtained during field experiments (Chapter 3) where the number of recruits was counted on caged settlement tiles at two different times in the year.

100 Growth rate Growth rate was estimated based on 0.0217 cm2 cm-2 day-1 values from van der Zande (unpublished (± 0.0163 cm2 cm-2 day-1) Master thesis). Birrell (2014) obtained similar values on the Great Barrier reef, but they were measured by seasonality. Since there was no empirical data on seasonal growth in Palau one average growth rate for the whole year was used for this model. Growth rate was sampled from a normal distribution of mean and standard deviation. Recruit mortality in Proportion of recruits removed in 2.5h on 0.261 the tank experiment, a 100 cm2 tile exposed to feeding by driven by C. spilurus exclusively C. spilurus (Chapter 3), randomly sampled from a binomial distribution Recruit mortality in Proportion of recruits removed in 2.5h on 0.25 the tank experiment, a 100 cm2 tile exposed to feeding by driven by Ct. striatus exclusively Ct. striatus (Chapter 3), randomly sampled from a binomial distribution Recruit mortality in Proportion of recruits removed in 2.5h on 0.396 the tank experiment, a 100 cm2 tile exposed to feeding by driven by A. exclusively A. nigrofuscus (Chapter 3), nigrofuscus randomly sampled from a binomial distribution Recruit mortality in Proportion of recruits removed in 2.5h on 0.119 the tank experiment, a 100 cm2 tile exposed to feeding by driven by Z. scopas exclusively Z. scopas (Chapter 3), randomly sampled from a binomial distribution Multiplier grazing in Grazing rates by C. spilurus within a tank 0.59 tank experiment were compared to grazing on the reef compared to on the (Chapter 3) to scale mortality from tank to reef for C. spilurus reef conditions based on grazing intensity 101 Multiplier grazing in Grazing rates by Ct. striatus within a tank 2.95 tank experiment were compared to grazing on the reef compared to on the (Chapter 3) to scale mortality from tank to reef for Ct. striatus reef conditions based on grazing intensity Multiplier grazing in Grazing rates by A. nigrofuscus within a 2.01 tank experiment tank were compared to grazing on the reef compared to on the (Chapter 3) to scale mortality from tank to reef for A. reef conditions based on grazing intensity nigrofuscus Multiplier grazing in Grazing rates by Z. scopas within a tank 0.49 tank experiment were compared to grazing on the reef compared to on the (Chapter 3) to scale mortality from tank to reef for Z. scopas reef conditions based on grazing intensity Adult Lobophora Proportion of adult Lobophora area 0.69 mortality driven by removed. Field data collected by Roff et herbivores al. (2015b). Lobophora patches without grazing and with grazing were compared, accounting for growth, biomass of parrotfish and available substrate during the study period to calculate mortality, which is based on parrotfish biomass. Adult Lobophora mortality is exclusively driven by parrotfish because data indicate that parrotfish are the only group enhancing model fit between Lobophora cover and herbivorous fish biomass (Chapter 4). Natural adult Calculated using field data on Lobophora 0.146 mortality cover and survival probability from caged (± 0.0058) plots (Chapter 3) Parrotfish biomass Average parrotfish biomass from surveys 6.3 g m-2 (± 4.45 g m-2) conducted by P Mumby in 2016 at Lighthouse Reef (used to parameterize this model) based on eight surveys of 4 m × 30 m. This was used to weigh the mortality of 102 Lobophora recruits and adults as caused by C. spilurus in the tank experiment (Chapter 3) and by parrotfish on the reef (Roff et al. 2015b), respectively. Acanthurus spp. Average Acanthurus spp. biomass from 0.52 g m-2 (± 0.43 g m-2 ) biomass surveys conducted by P Mumby in 2016 at Lighthouse Reef (used to parameterize this model) based on eight surveys of 4 m × 30 m. This was used to weigh the mortality of Lobophora recruits as caused by A. nigrofuscus in the tank experiment (Chapter 3). Ctenochaetus spp. Average Ctenochaetus spp. biomass from 0.27 g m-2 (± 0.23 g m-2) biomass surveys conducted by P Mumby in 2016 at Lighthouse Reef (used to parameterize this model) based on eight surveys of 4 m × 30 m. This was used to weigh the mortality of Lobophora recruits as caused by Ct. striatus in the tank experiment (Chapter 3). Zebrasoma spp. Average Zebrasoma spp. biomass from 1.98 g m-2 ± (0.31 g m-2) biomass surveys conducted by P Mumby in 2016 at Lighthouse Reef (used to parameterize this model) based on eight surveys of 4 m × 30 m. This was used to weigh the mortality of Lobophora recruits as caused by Z. scopas in the tank experiment (Chapter 3).

5.3.3 Model validation We compared the model outputs to results from a microhabitat survey (see Methods of Chapter 4 for details), which quantified the proportion of microhabitat types with Lobophora. The survey data was exclusively used to validate the model and did not contribute to the model parameterization in any way. For the comparison, survey microhabitats were only included if they fit the model specified microhabitats, i.e. small (8 mm × 15 mm), medium (15 mm × 15 mm), and large crevices (40 mm × 40 mm) as well as open microhabitats (planar area) with ± 50% variation around each size. Each 103 microhabitat type was plotted separately, including the survey results (mean ± 95% confidence interval) and 20 model runs. Each simulation was run for 60 times steps (i.e. 5 years). The model predicted the proportion of microhabitats with Lobophora well for small and large crevices but was overestimating the proportion of medium crevices with Lobophora. The proportion of open microhabitats harbouring Lobophora was slightly underestimated but still at the lower end of the 95% confidence interval.

5.3.4 Model adjustment A behavioural survey of fish feeding within microhabitats was conducted. An individual fish was followed at a distance of 2 m to avoid disturbing the individual until a bite was taken. The microhabitat in which the bite was taken was then measured. A total of 31 bites by parrotfish were measured in this way. While video observations used to parameterise the model showed no feeding of parrotfishes in medium microhabitats, the data collected in the behavioural survey showed that parrotfish took one bite each out of the 31 in medium-sized crevices and in large crevices (based on crevice sizes used in the model; medium: 15 mm × 15 mm ± 50%; large: 40 mm × 40 mm ± 50%). Therefore, we deemed it most realistic to apply an intermediate parrotfish grazing pressure between small and large crevices to medium crevices. Allowing intermediate access to medium microhabitats caused the model to predict the proportion of microhabitats with Lobophora well.

5.3.5 Sensitivity analysis The influence of the main model parameters growth rate, recruit mortality, recruitment rate, adult natural mortality, adult herbivore-facilitated mortality and microhabitat accessibility on Lobophora cover was quantified by varying each of the parameters by ± 20%. The effects were estimated for each parameter separately while keeping the other parameters fixed. The results were reported for the overall simulated reef area (Figure 5.3).

5.3.6 Model scenarios Influence of simulated herbivorous fish overfishing To understand the effect overfishing of key herbivorous fish groups could have on Lobophora cover, we varied the biomass of each of the four main fish groups in the model, Acanthurus spp., Ctenochaetus spp., Zebrasoma spp., and parrotfish (including Scarus spp., and Chlorurus spp.) to 5%, 20%, 50%, 80% and 100% of the original biomass. Additionally, we simulated the effect for all fish groups combined. The outcome for each microhabitat type was plotted separately as well as for the overall reef area.

104 Influence of concealed microhabitats on Lobophora cover trajectory Grazing pressure is reduced within concealed microhabitats. To analyse the influence this may have onto overall Lobophora cover, we investigated the influence of small, medium and large crevices onto Lobophora cover. The overall Lobophora cover was recorded while varying the proportionate area of small, medium and large crevices. The normal area distribution among microhabitat was 5% small crevices, 14% medium crevices, 37% large crevices and 44% open microhabitats. Small, medium and large crevices were set to 20% and 40% while changing the proportion of open microhabitats to reach 100% of the area. See Table 5.2 for the seven different microhabitat distribution scenarios.

Table 5.2: Seven scenarios run to explore the influence of changing small scale structural complexity. The scenario name is given to compare the input to the results graph and the % area occupied by each microhabitat type is reported.

Scenario # Scenario name % small % medium % large % open crevice crevice crevice microhabitat 1 Normal distribution 5 14 37 44 2 20% small 20 14 37 29 3 40% small 40 14 37 9 4 20% medium 5 20 37 38 5 40% medium 5 40 37 18 6 20% large 5 14 20 61 7 40% large 5 14 40 41

Parrotfish overfishing on reefs of varying small-scale complexity Finally, as parrotfish showed the largest influence on Lobophora cover, parrotfish biomass and microhabitat distribution were varied in different combinations. Each of the seven microhabitat distributions (Table 5.2) was run while setting parrotfish biomass to 100%, 50%, 20% and 5% of its original value.

105 5.4 Results 5.4.1 Model validation Comparing the model output with survey results showed that the model accurately reproduced the observed Lobophora distribution for small and large crevices within the 95% confidence intervals, and close to the observed mean percentage cover (Figure 5.2). Lobophora distribution in open microhabitats was predicted on the lower edge of the survey’s 95% confidence interval. After adjusting parrotfish access in medium crevices to half-way between large and small crevices the model reproduced the observed Lobophora distribution in medium crevices within the 95% confidence interval.

Figure 5.2: Model validation plots. Mean Proportion of microhabitats with Lobophora in surveys (straight line) and 95% confidence intervals (shaded area) of survey data are plotted in red. Burn-in time (month 0 - 20) is not shown. 20 model runs are plotted in black for each microhabitat type.

106 5.4.2 Sensitivity analysis Predictions of Lobophora cover were most sensitive to microhabitat accessibility by herbivorous fishes with up to 40% change in Lobophora when microhabitat accessibility was decreased by 20%. The alga’s rates of natural mortality, growth, and recruitment caused an intermediate change of 20% - 30% Lobophora cover. In contrast, the influence of grazing-driven rates of adult or recruit mortality on Lobophora was at least five-fold weaker than the influence of microhabitat accessibility and caused only small changes in Lobophora cover of up to ~ 8% (Figure 5.3).

Figure 5.3: Sensitivity analysis shows deviation of overall Lobophora cover when the input parameter is varied by +20% (red bars) and by -20% (blue bars). Averages from 20 model runs.

5.4.3 Influence of concealed microhabitats on Lobophora cover trajectory Increases in the proportion of concealed microhabitats (while keeping the total surface area constant) caused an increase in overall Lobophora cover (Figure 5.4). The higher the proportion of crevices, the higher the Lobophora cover. However, this cover was below what would have been expected if all additional microhabitats were inhabited by the alga except when the proportion of large crevices was reduced. In this case, the expected cover would have been below what was observed.

107

Figure 5.4: Lobophora cover over 60 months under different microhabitat distributions. The burn-in time (month 0 – 20) is not shown. Results are presented for 20 model runs. The red line shows the expected cover if the area additionally assigned to crevices was filled with Lobophora and added to the cover of the in-situ observed microhabitat distribution (first pane). Values below the red line indicate that Lobophora did not occupy all the added crevice surface area. Text inlay shows the overall distribution of all four microhabitat types in each scenario.

5.4.4 Influence of fishing herbivorous fishes Lobophora cover increased with decreasing herbivorous fish biomass (Figure 5.5). Reductions of herbivorous fish biomass of all species had the strongest impact on Lobophora within large crevices and open microhabitats, whereas cover increases in small and medium crevices were two- to three- fold lower. Out of the four fish groups studied, parrotfish biomass reductions caused the strongest increase in Lobophora in large crevices (Figure 5.5D) and in all microhabitats combined (Figure 5.5A). Reductions in the biomass of parrotfish and Acanthurus spp. led to similar increases of Lobophora within medium-sized crevices with parrotfish showing a slightly stronger influence (Figure 5.5C). In small crevices, Z. scopas had the strongest control over Lobophora, and parrotfish abundance had no impact in this microhabitat (Figure 5.5B). No cover increase was observed in open microhabitats when the biomass of each fish species was reduced individually while leaving the others unfished (Figure 5.5E). In contrast, when the biomass of all fish was reduced, Lobophora cover increased slowly before rising rapidly to 100% when all species were reduced to 5% of their original biomass (Figure 5.5E). Importantly, parrotfish biomass at the study site was 6.3 g m-2 and thus three times as high as Ctenochaetus spp. biomass, which was 1.98 g m-2. Parrotfish biomass was also about 12 times higher than Acanthurus spp. biomass (0.52 g m-2) and 24 times higher than Zebrasoma spp. biomass (0.27 g m-2). Accordingly, when the overall herbivorous fish biomass was taken into account 108 while reducing individual fish species biomass the pattern changed. Here, the increase in Lobophora per gram herbivorous fish biomass was strongest when only Zebrasoma spp. biomass was reduced. The second strongest increase in Lobophora was observed when the biomass of all fish species was reduced simultaneously or when Acanthurus spp. biomass was reduced individually. The lowest increase in Lobophora cover per gram herbivorous fish biomass was now observed when parrotfish or Ctenochaetus spp. biomass was reduced individually (Figure 5.6).

109

Figure 5.5: Changes in % Lobophora cover driven by reductions in herbivorous fish biomass. “All species combined” means that the fish biomass of all species was reduced simultaneously. Dots show simulated values and linear regression lines and 95% confidence intervals are added. Note the different scales. Colours mark different fish groups.

110

Figure 5.6: Changes in % Lobophora cover with decreasing overall herbivorous fish biomass as driven by reductions in either individual groups (Acanthurus spp., Ctenochaetus spp., Zebrasoma spp. or parrotfish) or for all species simultaneously (“all species combined”). Dots show simulated values when biomass was reduced to 100%, 80%, 50%, 20% and 5% of original values for each fish group. Some species had low biomass on the reef, causing the data to be concentrated at the upper end of herbivorous fish biomass on the x-axis even when a fish species was reduced to 5% of its in- situ biomass. Linear regression lines and 95% confidence intervals were added. Note the different scales. Colours mark different fish groups.

5.4.5 Interactive effects of parrotfish overfishing and microhabitat complexity on Lobophora populations As parrotfish had the highest biomass on the reef, parrotfish biomass reductions were combined with manipulations of small-scale complexity. The proportion of each crevice type was set to its original value observed in-situ, to 20% or 40% while correspondingly changing the proportion of exposed microhabitats to achieve a distribution of microhabitats that covered the overall area. Using these proportions allowed two of the three crevice types to stay unmanipulated while only varying the proportions of one crevice type and exposed microhabitats. When parrotfish biomass was at 100% of its original value, Lobophora increased with the proportion of microhabitats in the following order: Lobophora cover was highest when 40% of microhabitats were small crevices, followed by 40%

111 medium crevices, 20% small crevices, 20% medium crevices and 40% large crevices, and lastly by 20% large crevices (Figure 5.7).

When parrotfish biomass was decreased to 50%, 20% and 5%, Lobophora increased, irrespective of the microhabitat distribution. However, the positive response of Lobophora to a decline in herbivory was greatest where small and medium crevices dominated the substratum. Similarly, a reduction of large crevices showed a reduced effect of parrotfish biomass on Lobophora. The incline of the relationship of Lobophora cover with parrotfish biomass was similar, yet slightly steeper, on reefs with increased small and medium crevice distributions compared to the microhabitat distribution observed in-situ. Correspondingly, cover increased less steeply when the number of large crevices was reduced (Figure 5.7).

Figure 5.7: Change in % Lobophora cover in response to changes in parrotfish biomass and microhabitat distribution. Red lines (linear regression) and error margins (95% confidence interval) show results with manipulated microhabitat distribution. Grey shows results for in-situ microhabitat distribution. Grey dotted lines and open black circles show normal distribution next to manipulated microhabitat distribution to allow better comparison.

112 5.5 Discussion Investigating the influence of species-specific fish biomass reductions and their interactions with varying levels of structural refugia is difficult to accomplish in-situ, which led us to develop a model – based on empirical data – to explore these interactions. Model predictions imply that Lobophora cover increases when more concealed microhabitats are available and – not surprisingly – when herbivorous fish biomass decreases. Of four fish groups, parrotfish exerted the strongest control over Lobophora except within small crevices, where parrotfish had no impact. However, this was largely driven by the large proportion of parrotfish biomass on the reef. When overall herbivorous fish biomass was taken into account, the increase in Lobophora was stronger when all fish species were overfished simultaneously than when parrotfish were overfished. Reductions in parrotfish biomass resulted in more Lobophora on reefs with a higher fine-scale structural complexity compared to less complex reefs. The relationship between Lobophora cover and parrotfish biomass was similar at all simulated levels of structural complexity but Lobophora increased slightly faster on more complex reefs. This indicates that overfishing of parrotfish may cause a stronger increase of Lobophora on reefs with high small-scale structural complexity compared to less complex reef systems.

Predicted Lobophora cover was higher on reefs with a greater proportion of crevices compared to less structurally complex reefs. This indicates that the more abundant crevices are, the more they promote Lobophora cover. Larger-scale rugosity as measured using chains or by evaluating structural complexity visually is beneficial for coral reef recovery (Graham et al. 2015) but our research indicates that smaller-scale complexity on the scale of centimetres may facilitate macroalgal cover. Future research could examine the role of fine-scale complexity on the scale of fish feeding access against the broader-scale complexity often correlated with fish biomass. Interestingly, a sensitivity analysis revealed that the model was about four to six times more sensitive to differences in microhabitat accessibility than to changes in recruit or adult grazing mortality. This highlights the magnitude with which small-scale structural complexity influences the ability of fish to remove algae on a reef.

Lobophora is often found associated with concealed microhabitats (Bennett et al. 2010) and may experience increases in its cover on reefs with higher structural complexity because crevices reduce grazing pressure by fishes considerably (Bennett et al. 2010; Brandl and Bellwood 2016). However, not all added crevices became occupied by Lobophora because the model reached an equilibrium between recruitment rate, growth and mortality rates before all crevices were filled. Our results show that while Lobophora cover is positively associated with increased crevice abundance, the alga seems unable to spread to planar reef areas as long as grazing pressure remains unchanged. Consequently, 113 a shift to Lobophora dominance on a reef with a healthy herbivorous fish biomass is unlikely unless considerable reductions in grazing pressure occur (Vieira 2019).

Reductions in grazing pressure can be caused directly through reductions in herbivorous fish biomass (e.g., Bellwood et al. 2006) or indirectly by increases in available settlement space, for example after mass coral mortality (Williams et al. 2001; Mumby et al. 2007), which causes fish to distribute their feeding activity over larger areas. Reducing herbivorous fish biomass in the model resulted in increased Lobophora cover, which was most pronounced when fish biomass was reduced for all species simultaneously. Individually, parrotfish exerted the strongest control over algal cover when the percentage reduction in biomass among species was compared. This was likely driven by the high proportion of parrotfish biomass compared to other herbivorous fishes. When the overall herbivorous fish biomass was taken into account while reducing fish biomass of individual species, a different pattern emerged. Out of the four fish groups, the two rarest groups, Zebrasoma spp. and Acanthurus spp. had the strongest influence on Lobophora cover per gram biomass. Decreases in the biomass of Zebrasoma spp. caused the steepest increase in Lobophora, followed by Acanthurus spp. and only afterwards by parrotfish and Ctenochaetus spp. This may be driven by the ability of Z. scopas and A. nigrofuscus to access all microhabitats (Brandl et al. 2015, Chapter 4), whereas parrotfish cannot access all microhabitats (Chapter 4) and prefer to feed on convex surfaces (Bellwood et al. 2003). Only reductions in Zebrasoma spp. caused a stronger increase in Lobophora than reductions of all herbivorous fish biomass combined. While this may imply that Zebrasoma spp. is better at controlling Lobophora than all fish species combined, the biomass of Zebrasoma spp. on reefs in Palau is low, possibly reducing its overall impact on the reef. The low biomass of Zebrasoma spp. may be caused by its highly territorial behaviour and large territory size compared to other surgeonfishes (Robertson et al. 1979). The low biomass of Zebrasoma spp. and Acanthurus spp. on the reef used to parameterize the model may limit our ability to test whether there may be a saturation of the impact these fish groups have on Lobophora cover. Further studies on reefs with higher biomass of these species would be valuable. Importantly, we show that a combination of fish species has stronger control over Lobophora than the most common herbivores (parrotfish and Ctenochaetus spp.) individually. The varying ability of fish species to access microhabitats combined with species-specific Lobophora grazing mortality may give rise to this pattern. These findings highlight the role of small-scale reef topography in shaping the ability of fish species to control Lobophora. Importantly, crevices can exclude whole functional groups, which creates distinct benthic communities within crevices (Brandl and Bellwood 2016).

114 Lobophora cover in easily accessible reef areas did not change when individual fish species biomass was reduced. Instead, only reductions of the combined herbivorous fish biomass caused a slow increase in Lobophora cover before a sharp increase to almost 100% cover occurred at 5% of the original herbivorous fish biomass, likely driven by the particularly strong reduction in grazing pressure when the biomass of all herbivorous fish species was reduced. While this effect was strongest in easily accessible reef areas, the effect was still present when all microhabitats were evaluated simultaneously. These results reveal two processes. First, when herbivorous fish biomass falls below 20% (but is still above 5%) of its original value, herbivores are unable to control Lobophora in open reef areas effectively. This implies there may be a threshold between 5% - 20% at which grazing is too low to control Lobophora. Thresholds at which a reef has a trajectory towards coral or algal dominance have been identified before but these are usually towards the upper end of the naturally occurring fish biomass (Mumby et al. 2007). The present model did not explore the impact of increasing Lobophora cover on corals, which may have shown negative effects on coral trajectories long before herbivorous fish biomass was reduced to 20%. Second, the limited effect of overfishing a single fish group in open reef areas indicates a possible redundancy in the control of Lobophora among fish groups. However, redundancy is concentrated within few functional groups whereas many functions are performed by single species (Mouillot et al. 2014). Here, the varying ability of fish species to access different microhabitat types adds a layer of complexity to the consumer-producer interaction that decreases the apparent redundancy observed within open microhabitats. Accordingly, increases in Lobophora cover on the overall reef area were much higher when all fish species were overfished compared to reductions in individual species biomass. Therefore, while there seems to be some redundancy in macroalgal recruit control, species diversity is likely important to control macroalgae on a reef scale. This is especially true considering that a variety of macroalgal species exists on a reef and different herbivorous fish species control different macroalgal species (Burkepile and Hay 2011). Correspondingly, recent research suggests that herbivorous fish diversity is a better predictor of macroalgal consumption than fish biomass (Topor et al. 2019).

In addition to species diversity, ecosystem functioning also relies on key species (Hoey and Bellwood 2009; Topor et al. 2019). Our simulations have shown parrotfish to be the most important group investigated to control overall Lobophora cover (see also Roff et al. 2015b), even though this is at least partly due to parrotfish making up a large proportion of the overall herbivorous fish biomass. Reduced parrotfish biomass on reefs with differing levels of small-scale structural complexity led to higher Lobophora cover on reefs with many crevices. Interestingly, while Lobophora cover increased with decreasing parrotfish biomass in all simulated reef environments, a slightly stronger increase in structurally more complex reefs was observed (Figure 5.7). This may be caused by the ability of the

115 alga to increasingly spread from crevices to open areas when grazing pressure is reduced. Our model uses a fixed recruitment rate as data on density-dependent recruitment of Lobophora is currently unavailable. Including density-dependent recruitment would likely strengthen this effect as recruitment accelerates with increasing Lobophora cover and could thus eventually exceed Lobophora mortality.

On an unfished reef, Lobophora may be slightly more abundant if the reef has high small-scale structural complexity. Neither on a complex nor a less complex reef is the alga expected to form a bloom unless grazing pressure is reduced considerably (Roff et al. 2015b; Vieira 2019). However, when herbivorous fishes, especially parrotfish, are overfished, all reefs will experience an increase in Lobophora cover. In this case, structurally more complex reefs seem more likely to form a bloom because they may have a higher starting biomass of Lobophora within crevices and more crevices from which the alga can start growing into open spaces. Overfishing may thus have a stronger impact on macroalgal cover in reefs with higher small-scale complexity.

There are currently some limitations to the model. While the model predicts Lobophora distribution in crevices well, Lobophora in exposed microhabitats is predicted at the low margin of the 95% confidence interval, indicating that we may be overestimating mortality on planar reef surfaces. Spatial patterns in feeding activity could allow Lobophora to escape grazing in some reef areas, a mechanism not represented in the model. Parrotfish, for example, have been shown to focus their feeding in core areas with higher complexity (Welsh and Bellwood 2011; Nash et al. 2012; Adam et al. 2015), a behaviour linked to their diet (Clements et al. 2016). This feeding behaviour could result in spatial grazing refuges that allow Lobophora to become established. Further, the model uses a recruitment rate which is independent of adult Lobophora cover. This assumption allows the model to reproduce the local Lobophora distribution, but it is likely not realistic as more adult Lobophora can be expected to produce more propagules. However, empirical data of a stock-recruitment relationship is currently unavailable for Lobophora and will be complicated by the high cryptic species diversity in this genus (Sun et al. 2012; Vieira et al. 2014). Incorporating density-dependent recruitment would allow me to explore further research questions. For example, the model could be used to explore differences between phase shift prevention and phase shift reversal. This could include investigating whether more herbivore biomass is necessary to return a reef to low algal cover compared to the herbivore biomass required to keep a reef free from Lobophora in the first place. Additionally, incorporating species-specific recruitment and growth rates may enable us to explore Lobophora species-specific abilities to form blooms.

116 Overall, our model allowed us to investigate the influence and interaction of structural complexity and changes in herbivorous fish biomass on the trajectory of the common brown macroalga Lobophora. We show that there may be a small interactive effect where Lobophora biomass increases more steeply on structurally complex reefs than on less complex reefs. Even small increases in Lobophora cover could disrupt reef recovery after disturbances, as the alga increases coral larval mortality (Morrow et al. 2016) and reduces coral larval settlement (Evensen et al. 2019). The implications of our findings for the future of coral reefs are difficult to disentangle. The decline of reef structural complexity (Alvarez-Filip et al. 2009) is likely going to increase as coral mass mortality becomes more frequent (Hughes et al. 2018). This reduction in grazing refugia may make it harder for algae, such as Lobophora, to become established. On the other hand, coral mortality causes an increase in settlement substrate and therefore decreases grazing intensity (Williams et al. 2001; Mumby et al. 2007), which in turn would benefit Lobophora. More research is needed to disentangle these two opposing trends but given the observations of Lobophora blooms on reefs which have recently experienced mass coral mortality (Mumby et al. 2005; Roff et al. 2015b), the latter seems to prevail. Finally, our model shows that parrotfishes are the most important herbivorous fish group to control Lobophora on the overall simulated reef but Zebrasoma spp. and Acanthurus spp. have a stronger impact on Lobophora per gram biomass, likely because of their ability to access all microhabitat types. Indeed, reductions in herbivorous fish biomass as caused by overfishing all species simultaneously had a stronger impact on Lobophora cover than overfishing of the most common herbivore species. Therefore, the protection of a diverse suite of herbivorous fish species might prove to be more successful to avoid macroalgal blooms on coral reefs than species-specific fishing bans.

117

______

CHAPTER 6: GENERAL DISCUSSION

______

118 General Discussion

6.1 Significance of this thesis Coral reefs around the world are threatened by rising temperatures and local stressors, such as eutrophication and overfishing (Hughes 1994; Hoegh-Guldberg 1999; Hoegh-Guldberg et al. 2007; Hughes et al. 2017). Consequently, mass coral mortality events are becoming more frequent (Hughes et al. 2018). Following mass coral mortality, reefs can become dominated by other organisms, most frequently macroalgae (Done 1992; Hughes 1994; Rogers and Miller 2006; Ledlie et al. 2007). Yet, whether a reef shifts towards macroalgal dominance or not is still hard to predict. The macroalgal genus Lobophora is often a major component of these shifts to macroalgal dominance (Done 1992; Hughes 1994; McCook et al. 2001). Lobophora is one of the strongest competitors of corals and even small amounts of the alga can severely impair coral larval settlement (Evensen et al. 2019). Understanding what contributes to blooms of this alga and which species can control it is thus important. However, research into the ecology of Lobophora and its control by herbivores is still limited and many contradicting reports on its interactions with corals and herbivorous fishes exist (Kuffner et al. 2006; Nugues and Bak 2006; Birrell et al. 2008b; Vieira et al. 2015, 2016c; Mumby et al. 2016). In this thesis, I explored the cryptic diversity of Lobophora and the environmental driver of this diversity, which may explain contrasting results from previous studies. I further investigated the abilities of several common herbivorous fish species to remove Lobophora at different life-stages and studied the role of microhabitats as Lobophora refuges and spore sources. Subsequently, I combined these findings in an individual-based model to simulate overfishing of different herbivorous fish species and subsequent impacts on Lobophora cover.

My research shows that there is a large cryptic diversity of Lobophora and that algal assemblages are influenced by wave exposure on larger geographic scales (island-wide) and by wave exposure, herbivore biomass and depth on smaller geographic scales. I found a strong decrease in grazing susceptibility in adult compared to early life-stage Lobophora, indicating that management of herbivorous fish will be more effective when implemented before grazing-resistant adult populations form. Further, I identified Acanthurus nigrofuscus and especially parrotfish as important agents of Lobophora control. The findings of this thesis also highlight the importance of crevices for protecting Lobophora recruits from fish grazing, in particular from parrotfish. Overall, this thesis enhances our understanding of the ecology of the pervasive macroalgal genus Lobophora and the factors contributing to shifts to Lobophora dominance on coral reefs. My findings may aid fisheries management by identifying fish species that are most effective at controlling Lobophora and point

119 towards small-scale structural complexity as an important contributor to blooms of this pervasive macroalga.

6.2 The importance of Lobophora bottlenecks We found a strong bottleneck in the grazing-mediated mortality of Lobophora (Chapter 3). While recruits were readily removed, adult mortality was minimal (Chapter 3). Grazing-mediated recruit mortality is dependent on the fish species and the microhabitat the alga has settled in (Chapter 3, 4, 5). Lobophora recruits experience low herbivore-driven mortality in small crevices, which is also reflected in the abundance of adult Lobophora (Chapter 4, 5). Correspondingly, simulated grazing pressure on Lobophora recruits is high on open reef substrate and does not allow the establishment of adult Lobophora except when grazing pressure is severely reduced (Chapter 5). In contrast to recruits, adult survival was not influenced by herbivorous fishes (Chapter 3), indicating that once Lobophora has become established it is hard to control. Previous research has found considerable variation in the mortality of adult Lobophora, ranging from barely any biomass consumed to the almost complete removal of assays by herbivores (e.g., Hay 1981; de Ruyter van Steveninck and Breeman 1987a; Steinberg and Paul 1990). There are multiple explanations for this variation. For example, the large diversity of Lobophora species (Sun et al. 2012; Vieira et al. 2014) may cause differences in chemical defences between the species, which in turn could influence grazing pressure. However, recent research shows a negligible effect of secondary metabolites (Vieira et al. 2019). Alternatively, morphology may play a more important role (Coen and Tanner 1989). Since most Lobophora species in Palau are crustose to decumbent, their grazing susceptibility may be particularly low. A strong ontogenetic shift in grazing susceptibility can affect coral reef recovery negatively because once adult Lobophora thalli become established their harmful interactions with corals may persist for extended periods of time (Jompa and McCook 2002a, b; Kuffner et al. 2006; Mumby et al. 2016). This may have caused the persistence of a phase shift to Lobophora dominance at one of the study reefs in Palau that occurred when grazing was severely reduced but persisted through higher grazing pressure for ~ 5 years (Roff et al. 2015b).

6.3 Role of individual fish species in controlling Lobophora blooms Effective control of Lobophora seems to be driven by a small number of fish species (Chapter 3,4). Lobophora recruits were only removed by three out of the seven species studied. The surgeonfish Acanthurus nigrofuscus and the parrotfishes Chlorurus sordidus and Scarus niger removed the alga’s recruits, whereas the impact of the surgeonfishes Ctenochaetus striatus, Ctenochaetus binotatus, Zebrasoma scopas, and the unicornfish Naso lituratus was at best limited (Chapter 3). In contrast,

120 adult Lobophora abundance in surveys was associated with reductions in parrotfish biomass only, indicating that parrotfish may be the main control of encrusting adult Lobophora (Chapter 4).

Whether a fish can access concealed microhabitats likely plays a role in its ability to control Lobophora. A. nigrofuscus, which removed most Lobophora recruits can access crevices but does so infrequently (Chapter 3, 4), possibly limiting its ability to control the alga in crevices. Further, A. nigrofuscus is unlikely to remove adult Lobophora because it is adapted to removing turf algae (Randall 1955; Purcell and Bellwood 1993; Bellwood et al. 2014). Therefore, A. nigrofuscus is important in controlling Lobophora but only at the early life-stages of the alga. In contrast, parrotfish remove both Lobophora recruits and adults (Roff et al. 2015, Chapter 3, 4), but they do not access small to medium-sized crevices (Chapter 4). While certain fish species are more capable of controlling Lobophora recruits than others (Chapter 3), the ability of fishes to access concealed microhabitats adds another layer of complexity to the consumer-producer relationship. Therefore, a suite of herbivorous fish species able to remove Lobophora and access a variety of microhabitats may be more efficient at controlling Lobophora compared to any individual species. Similarly, the individual-based model revealed that while reductions in parrotfish biomass led to the largest increases in Lobophora cover, reductions in the biomass of all fish species combined showed a stronger effect on Lobophora (Chapter 5).

Interestingly, Z. scopas and Ct. striatus did not significantly remove recruits in the tank and in-situ experiments (Chapter 3), but their observed removal rates were sufficient to exert some control over Lobophora on a simulated reef environment, especially within small crevices (Chapter 5). Z. scopas did partially remove recruits in the tank experiment (pers obs, Chapter 3), indicating that over longer time periods this species may contribute to the removal of Lobophora. Notably, it is also the species most likely to access small crevices (Chapter 4) and could thus fill an important ecological role in reducing the number of Lobophora recruits in sheltered microhabitats and help prevent shifts to Lobophora dominance. Ct. striatus on the other hand only removed one recruit during the tank experiments and none in field observations. They may thus have some limited capability to remove Lobophora recruits but are most likely unable to exert any control over adults (Randall 1955; Purcell and Bellwood 1993). The lack of evidence in this thesis for removal of Lobophora recruits by a typical browsing species, N. lituratus, is likely due to low abundances of this species and a low number of feeding observations, as N. lituratus has been observed repeatedly to remove macroalgae (Rasher et al. 2013; Plass-Johnson et al. 2015). It is also possible that N. lituratus mainly feeds on other macroalgal species, such as Sargassum (Fox and Bellwood 2008; Hoey and Bellwood 2009), or that

121 it specifically targets adult macroalgae. In line with this evidence, N. lituratus took a disproportionately large number of bites on adult Lobophora patches (Chapter 3).

Interestingly, while neither surgeonfish nor parrotfish biomass was an important driver in structuring Lobophora species assemblages on Palau’s forereefs on larger geographic scales, parrotfish biomass did drive assemblages on the Eastern reefs of Palau (Chapter 2). This may be due to the very crustose morphology of many Lobophora species, which probably makes its removal difficult for species other than parrotfishes. While parrotfish achieve their nutrition from consuming microorganisms (Clements et al. 2016), Lobophora does have species-specific bacterial communities (Vieira et al. 2016b) and parrotfish may thus target Lobophora species based on these assemblages – a theory yet to be explored. Future studies would also benefit from attempting to sample taxonomic diversity quantitatively, as this is more likely to reveal patterns within Lobophora species assemblages on larger geographic scales than presence/absence data as used in this thesis. Here, a quantitative approach was not possible as many samples could not be identified with molecular methods and the relative abundances may thus have been skewed.

6.4 Lobophora species identity This thesis found a large cryptic diversity of 15 Lobophora species in Palau, which is somewhat lower than in other regions (Vieira et al. 2014, 2016a). Lobophora species showed some differences in thickness, associational refuge and morphology. If these differences translate into different growth and recruitment rates or chemical profiles the identity of Lobophora species may have a functional impact on coral-algal interactions and thus on the trajectory of coral reefs. For example, the individual-based model showed a strong sensitivity to variability within Lobophora recruitment and growth on Lobophora cover (Chapter 5), indicating that variation within these characteristics could influence the ability of a Lobophora species to colonise reefs quickly. Many studies on coral-algal interactions do not identify the Lobophora species involved or assume it is L. variegata, a species previously thought to have a worldwide distribution, which may contribute to the variation between results (Kuffner et al. 2006; Birrell et al. 2008b; Nugues and Bak 2008; Harborne et al. 2016; Morrow et al. 2016).

This thesis has attempted to identify Lobophora species used in experiments whenever possible. The dominant species in the study reefs are known and thus likely to be the species observed in in-situ observations. We identified Lobophora recruits genetically during the first field trip to verify that recruits were classified correctly as Lobophora sp. in the following experiments. However, we cannot be sure that the grazing experiments used the same Lobophora species due to being conducted at a 122 different time and subsequent attempted genetic analyses did not provide clear results. Still, our findings should be consistent among Lobophora species, as secondary metabolites have been shown to have limited impact on grazing susceptibility of fishes between Lobophora species (Vieira et al. 2019). Morphology plays a role in the alga’s grazing susceptibility (Coen and Tanner 1989). However, morphological differences likely develop later in the alga’s life as recruits of many brown macroalgae look similar. Further, most Lobophora species found in Palau are crustose with some thalli turning procumbent or decumbent and our results should thus be consistent between these species. The results of this thesis may or may not apply to more foliose Lobophora species.

6.5 Challenges Working with an alga with high cryptic diversity caused some difficulties that were not anticipated before the start of this thesis. As a result, one intended experiment on density-dependent recruitment could not be used as identifying adult Lobophora species on the reef proofed to be unreliable. Further, while the alga seems resilient on the reef, it started dying in two intended experiments, which led to a lack of data on growth rates. This had consequences for the model parameterization in Chapter 5 and data had to be obtained from the literature. While the microhabitat tiles created in Chapter 4 worked as intended, no parrotfish were observed feeding in medium crevices even though other surveys and video observations had shown them to be able to access similar-sized crevices. It is likely that because of the low grazing pressure within medium-sized crevices we were unable to observe bites within the recording time, highlighting the need for repeat recordings to quantify rare behaviour.

6.6 Conclusions & Future studies Given that political will to aggressively curb greenhouse gas emissions remains weak globally, mass coral mortality is likely to become more frequent (Hughes et al. 2018). Declines in coral cover are often followed by increases in macroalgae biomass (Done 1992; Hughes 1994; Rogers and Miller 2006; Ledlie et al. 2007; Graham et al. 2015) with Lobophora forming a noteworthy component of the macroalgal assemblages (Diaz-Pulido et al. 2009; Cheal et al. 2010; Roff et al. 2015b). This thesis highlights some of the intricate algal-herbivore interactions that will influence whether a reef shifts towards a macroalgal-dominated state. I show that the identity of the Lobophora and fish species is important and that grazing pressure on Lobophora is structured by habitat complexity. While high habitat complexity is a good predictor of coral reef recovery following disturbance (Graham et al. 2015), small-scale complexity on the scale of centimetres also seems to create a refuge for macroalgae to grow in. This thesis reveals that Lobophora is mainly controlled in the recruit stage and that there is limited functional redundancy among herbivorous fishes concerning their ability to control Lobophora. Interestingly, there seems to be a higher functional redundancy in the control of other 123 macroalgae, such as Sargassum sp. (Marshell 2014), with more fish species removing Sargassum recruits. Protecting key species, such as parrotfish and Acanthurus nigrofuscus (and possibly other Acanthurus species) may prove a good tool to control Lobophora populations. However, a more diverse suite of herbivorous fish species able to access a variety of microhabitats is likely required to avoid shifts to Lobophora dominance on a reef scale. Once Lobophora has become established, a return to coral dominance could be difficult as adult populations are relatively resistant to grazing by herbivorous fishes (Chapter 3).

In this thesis, I also show that concealed microhabitats may have an important role in the formation of Lobophora blooms. Lobophora can survive easily within concealed microhabitats (Chapter 3), even if grazing pressure is high on the reef and it is unable to establish on open reef substrate (Chapter 5). However, when grazing pressure is reduced, as is the case after coral bleaching and mortality, it may be able to spread rapidly (Roff et al. 2015b). If temporal peaks in Lobophora recruitment and growth (Chapter 3), which have been linked to increased algal abundance (Rogers 1997; Diaz-Pulido et al. 2009; Birrell 2014), coincide with reductions of grazing pressure, the alga may be able to escape grazing control and establish grazing-resistant adult populations on open reef substrate. Therefore, the timing of disturbances may influence whether a shift towards Lobophora dominance occurs. Future research should test this hypothesis by excluding herbivores from reefs at different times of the year and across different years and investigate whether the amount of small-scale structural complexity influences the likelihood of macroalgal phase shifts. Additionally, investigating ecological drivers of the quantitative distribution of Lobophora may reveal further insights into what structures the abundance of this pervasive macroalga. Finally, further studies should also investigate whether there is a functional difference between Lobophora species in terms of coral-algal interactions. If Lobophora species identity does influence the outcome of these interactions, the identity of the alga establishing dominance may determine coral reef recovery.

Beyond increasing our knowledge about Lobophora ecology, this thesis highlights that both the diversity of producers and consumers may play an important role for ecosystem functioning. This thesis shows the complexity of consumer-producer interactions, which decreases the redundancy amongst species. Therefore, a diverse suite of species is likely necessary for coral reef ecosystem functioning. The protection of biodiversity in light of increasing anthropogenic pressures is thus of utmost importance.

124 List of References Adam TC, Kelley M, Ruttenberg BI, Burkepile DE (2015) Resource partitioning along multiple niche axes drives functional diversity in parrotfishes on Caribbean coral reefs. Oecologia 179:1173– 1185. doi: 10.1007/s00442-015-3406-3 Alvarez-Filip L, Dulvy NK, Gill J a, et al (2009) Flattening of Caribbean coral reefs: region-wide declines in architectural complexity. Proc Biol Sci 276:3019–25. doi: 10.1098/rspb.2009.0339 Appel HM (1993) Phenolics in ecological interactions: The importance of oxidation. J Chem Ecol 19:1521–1552. doi: 10.1007/BF00984895 Arnold TM, Tanner CE, Hatch WI (1995) Phenotypic variation in polyphenolic content of the tropical brown alga Lobophora variegata as a function of nitrogen availability. Mar Ecol Prog Ser 123:177–184. doi: 10.3354/meps123177 Baraza E, Zamora R, Hódar JA (2006) Conditional outcomes in plant-herbivore interactions: neighbours matter. Oikos 113:148–156. doi: 10.1111/j.0030-1299.2006.14265.x Barlow J, Lennox GD, Ferreira J, et al (2016) Anthropogenic disturbance in tropical forests can double biodiversity loss from deforestation. Nature 1–16. doi: 10.1038/nature18326 Barton KE, Koricheva J (2010) The Ontogeny of plant defense and herbivory: characterizing general patterns using meta-analysis. Am Nat 175:481–493. doi: 10.1086/650722 Bates D, Mächler M, Bolker B, Walker S (2015) Fitting Linear Mixed-Effects Models using lme4. J Stat Softw. doi: 10.18637/jss.v067.i01 Bazzaz FA (1975) Plant Species Diversity in Old-Field Successional Ecosystems in Southern Illinois. Ecology 56:485–488. doi: 10.2307/1934981 Bellwood DR, Choat JH (1990) A functional analysis of grazing in parrotfishes (family Scaridae): the ecological implications. Environ Biol Fishes 28:189–214. doi: 10.1007/BF00751035 Bellwood DR, Hoey AS, Bellwood O, Goatley CHR (2014) Evolution of long-toothed fishes and the changing nature of fish-benthos interactions on coral reefs. Nat Commun 5:1–6. doi: 10.1038/ncomms4144 Bellwood DR, Hoey AS, Choat JH (2003) Limited functional redundancy in high diversity systems: Resilience and ecosystem function on coral reefs. Ecol Lett 6:281–285. doi: 10.1046/j.1461- 0248.2003.00432.x Bellwood DR, Hughes TP, Hoey AS (2006) Sleeping functional group drives coral-reef recovery. Curr Biol 16:2434–9. doi: 10.1016/j.cub.2006.10.030 Bennett S, Vergés A, Bellwood DR (2010) Branching coral as a macroalgal refuge in a marginal coral reef system. Coral Reefs 29:471–480. doi: 10.1007/s00338-010-0594-5 Bergey EA (2005) How protective are refuges? Quantifying algal protection in rock crevices. Freshw Biol 50:1163–1177. doi: 10.1111/j.1365-2427.2005.01393.x 125 Bernays EA, Funk DJ (1999) Specialists make faster decisions than generalists: experiments with aphids. Proc R Soc B 266:151–156. Bickford D, Lohman DJ, Sodhi NS, et al (2007) Cryptic species as a window on diversity and conservation. Trends Ecol Evol 22:148–155. doi: 10.1016/j.tree.2006.11.004 Birkeland C, Tsuda RT, Randall RH, et al (1976) Limited current and underwater biological surveys of a proposed sewer outfall sites on Malakal Island, Palau. Birrell CL (2014) Algal dynamics on coral reefs: a study of macroalgal traits and growth strategies of Lobophora J. Agardh (Dictyotales, Phaeophyceae). PhD thesis, The Universisty of Queensland, Australia Birrell CL, McCook LJ, Willis BL, Diaz-Pulido GA (2008a) Effects Of Benthic Algae On The Replenishment Of Corals And The Implications For The Resilience Of Coral Reefs. Oceanogr Mar Biol An Annu Rev 46:25–63. doi: 10.1201/9781420065756.ch2 Birrell CL, McCook LJ, Willis BL, Harrington L (2008b) Chemical effects of macroalgae on larval settlement of the broadcast spawning coral Acropora millepora. Mar Ecol Prog Ser 362:129– 137. doi: 10.3354/meps07524 Bischof K, Hanelt D, Wiencke C (1998) UV-radiation can affect depth-zonation of Antarctic macroalgae. Mar Biol 131:597–605. Blair CP, Abrahamson WG, Jackman JA, Tyrrell L (2005) Cryptic speciation and host-race formation in a purportedly generalist tumbling flower beetle. Evolution (N Y) 59:304–316. Boettcher AA, Targett NM (1993) Role of Polyphenolic Molecular Size in Reduction of Assimilation Efficiency in Xiphister mucosus. Ecology 74:891–903. Box SJ, Mumby PJ (2007) Effect of macroalgal competition on growth and survival of juvenile Caribbean corals. Mar Ecol Prog Ser 342:139–149. doi: 10.3354/meps342139 Bozec Y, Doropoulos C, Roff G, Mumby PJ (2018) Transient Grazing and the Dynamics of an Unanticipated Coral–Algal Phase Shift. Ecosystems. doi: 10.1007/s10021-018-0271-z Brandl SJ, Bellwood DR (2016) Microtopographic refuges shape consumer-producer dynamics by mediating consumer functional diversity. Oecologia. doi: 10.1007/s00442-016-3643-0 Brandl SJ, Bellwood DR (2014) Individual-based analyses reveal limited functional overlap in a coral reef fish community. J Anim Ecol 83:661–670. doi: 10.1111/1365-2656.12171 Brandl SJ, Hoey AS, Bellwood DR (2014) Micro-topography mediates interactions between corals, algae, and herbivorous fishes on coral reefs. Coral Reefs 33:421–430. doi: 10.1007/s00338-013- 1110-5 Brandl SJ, Rasher DB, Côté IM, et al (2019) Coral reef ecosystem functioning: eight core processes and the role of biodiversity. Front Ecol Environ 17:445–454. doi: 10.1002/fee.2088 Brandl SJ, Robbins WD, Bellwood DR (2015) Exploring the nature of ecological specialization in a

126 coral reef fish community: morphology, diet and foraging microhabitat use. Proc R Soc B Biol Sci 282:20151147. doi: 10.1098/rspb.2015.1147 Brooks ME, Kristensen K, Benthem KJ Van, et al (2017) glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. R J 9:378–400. Bryant D, Burke L, McManus J, Spalding M (1998) Reefs at risk: a map-based indicator of threats to the world’s coral reefs. Bryant JP, Reichardt PB, Clausen TP, et al (1992) Woody plant-mammal interactions. In: Rosenthal GA, Berenbaum MR (eds) Herbivores: their interactions with plant secondary metabolites. Academic Press, San Diego, CA, pp 344–371 Burkepile DE, Hay ME (2011) Feeding complementarity versus redundancy among herbivorous fishes on a Caribbean reef. Coral Reefs 30:351–362. doi: 10.1007/s00338-011-0726-6 Burkepile DE, Hay ME (2010) Impact of herbivore identity on algal succession and coral growth on a Caribbean reef. PLoS One. doi: 10.1371/journal.pone.0008963 Burkepile DE, Hay ME (2008) Herbivore species richness and feeding complementarity affect community structure and function on a coral reef. Proc Natl Acad Sci U S A 105:16201–6. doi: 10.1073/pnas.0801946105 Camacho O, Fernández-García C, Vieira C, et al (2019) The systematics of Lobophora (Dictyotales, Phaeophyceae) in the western Atlantic and eastern Pacific oceans: eight new species. J Phycol 624:611–624. doi: 10.1111/jpy.12850 Carilli JE, Norris RD, Black B, et al (2010) Century-scale records of coral growth rates indicate that local stressors reduce coral thermal tolerance threshold. Glob Chang Biol 16:1247–1257. doi: 10.1111/j.1365-2486.2009.02043.x Carpenter RC (1986) Partitioning Herbivory and Its Effects on Coral Reef Algal Communities. Ecol Monogr 56:345–363. doi: 10.2307/1942551 Cheal A, Emslie M, Miller I, Sweatman H (2012) The distribution of herbivorous fishes on the Great Barrier Reef. Mar Biol 159:1143–1154. doi: 10.1007/s00227-012-1893-x Cheal AJ, MacNeil MA, Cripps E, et al (2010) Coral–macroalgal phase shifts or reef resilience: links with diversity and functional roles of herbivorous fishes on the Great Barrier Reef. Coral Reefs 29:1005–1015. doi: 10.1007/s00338-010-0661-y Choat JH, Clements KD (1998) Vertebrate herbivore in marine and terrestrial environments : A Nutritional Ecology Perspective. Annu Rev Ecol Syst 29:375–403. Choat JH, Clements KD, Robbins WD (2002) The trophic status of herbivorous fishes on coral reefs. Mar Biol 140:613–623. doi: 10.1007/s00227-001-0715-3 Choat JH, Robbins WD, Clements KD (2004) The trophic status of herbivorous fishes on coral reefs. Mar Biol 145:445–454. doi: 10.1007/s00227-004-1341-7

127 Chollett I, Mumby PJ (2012) Predicting the distribution of Montastraea reefs using wave exposure. Coral Reefs 31:493–503. doi: 10.1007/s00338-011-0867-7 Christie AO, Shaw M (1968) Settlement experiments with zoospores of Enteromorpha intestinalis (L.) link. Br Phycol Bull 3:529–534. doi: 10.1080/00071616800650141 Cipollini DF, Redman AM (1999) Age-dependent effects of jasmonic acid treatment and wind exposure on foliar oxidase activity and insect resistance in tomato. J Chem Ecol 25:271–281. doi: 10.1007/s00484-012-0580-4 Clements KD, Choat JH (1997) Comparison of herbivory in the closely-related marine fish genera Girella and Kyphosus. Mar Biol 127:579–586. doi: 10.1007/s002270050048 Clements KD, Choat JH (1995) Fermentation in tropical marine herbivorous fishes. Physiol Zool 68:355–378. doi: 10.1086/physzool.68.3.30163774 Clements KD, German DP, Piché J, et al (2016) Integrating ecological roles and trophic diversification on coral reefs: multiple lines of evidence identify parrotfishes as microphages. Biol J Linn Soc 00:000–000. doi: 10.1111/bij.12914 Clements KD, Gleeson VP, Slaytor M (1994) Short-chain fatty acid metabolism in temperate herbivorous fish. J Comp Physiol B 164:372–377. Coen LD, Tanner CE (1989) Morphological variation and differential susceptibility to herbivory in the tropical brown alga Lobophora variegata. Mar Ecol Prog Ser 54:287–298. Connell JH, Hughes TP, Wallace CC (1997) A 30-year study of coral abundance, recruitment, and disturbance at several scales in space and time. Ecol Monogr 67:461–488. doi: 10.1890/0012- 9615(1997)067[0461:AYSOCA]2.0.CO;2 Connell JH, Sousa WP (1983) On the evidence needed to judge ecological stability or persistence. Am Nat 121:789–824. Crowder LB, Cooper WE (1982) Habitat Structural Complexity and the Interaction Between Bluegills and Their Prey. Ecology 63:1802–1813. doi: 10.2307/1940122 Day RW (1983) Effects of benthic algae on sessile animals: observational evidence from coral reef habitats. Bull Mar Sci 33:597–605. De Clerck O, Leliaert F, Verbruggen H, et al (2006) A revised classification of the Dictyoteae (Dictyotales, Phaeophyceae) based on rbcL and 26S ribosomal DNA sequence analyses. J Phycol 42:1271–1288. doi: 10.1111/j.1529-8817.2006.00279.x de Ruyter van Steveninck ED, Breeman AM (1987a) Deep water populations of Lobophora variegata (Phaeophyceae) on the coral reef of Curacao: influence of grazing and dispersal on distribution patterns. Mar Ecol Prog Ser 38:241–250. doi: 10.3354/meps036081 de Ruyter van Steveninck ED, Breeman AM (1987b) Deep water vegetations of Lobophora variegata (Phaeophyceae) in the coral reef of Curaçao: population dynamics in relation to mass mortality

128 of the sea urchin Diadema antillarum. Mar Ecol Prog Ser 36:81–90. doi: 10.3354/meps036081 de Ruyter van Steveninck ED, Van Mulekom LL, Breeman AM (1988) Growth inhibition of Lobophora variegata (Lamouroux) Womersley by scleractinian corals. J Exp Mar Bio Ecol 115:169–178. Diaz-Pulido G, McCook L (2004) Effects of live coral, epilithic algal communities and substrate type on algal recruitment. Coral Reefs 23:225–233. doi: 10.1007/s00338-004-0370-5 Diaz-Pulido G, McCook LJ (2003) Relative Roles of Herbivory and Nutrients in the Recruitment of Coral-Reef Seaweeds. Ecology 84:2026–2033. Diaz-Pulido G, McCook LJ, Dove S, et al (2009) Doom and boom on a resilient reef: climate change, algal overgrowth and coral recovery. PLoS One 4:e5239. doi: 10.1371/journal.pone.0005239 Done TJ (1992) Phase shifts in coral reef communities and their ecological significance. Hydrobiologia 247:121–132. doi: 10.1007/BF00008211 Doropoulos C, Roff G, Bozec Y-M, et al (2016a) Characterizing the ecological trade-offs throughout the early ontogeny of coral recruitment. Ecol Monogr 86:20–44. doi: 10.1890/15-0668.1 Doropoulos C, Roff G, Visser MS, Mumby PJ (2016b) Sensitivity of coral recruitment to subtle shifts in early community succession. Ecology 98:304–314. doi: 10.1002/ecy.1663 Doropoulos C, Roff G, Zupan M, et al (2014) Reef-scale failure of coral settlement following typhoon disturbance and macroalgal bloom in Palau, Western Pacific. Coral Reefs 33:613–623. doi: 10.1007/s00338-014-1149-y Doropoulos C, Ward S, Marshell A, et al (2012) Interactions among chronic and acute impacts on coral recruits: The importance of size-escape thresholds. Ecology 93:2131–2138. doi: 10.1890/12-0495.1 Dudley TL, D’Antonio CM (1991) The Effects of Substrate Texture , Grazing, and Disturbance on Macroalgal Establishment in Streams. Ecology 72:297–309. doi: https://doi.org/10.2307/1938923 Edgar RC (2004) MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res 32:1792–7. doi: 10.1093/nar/gkh340 Edmunds PJ, Nozawa Y, Villanueva RD (2014) Refuges modulate coral recruitment in the Caribbean and the Pacific. J Exp Mar Bio Ecol 454:78–84. doi: 10.1016/j.jembe.2014.02.009 Eich A, Ford AK, Nugues MM, et al (2019) Positive association between epiphytes and competitiveness of the brown algal genus Lobophora against corals. PeerJ 7:e6380:1–13. doi: 10.7717/peerj.6380 Evensen NR, Doropoulos C, Wong KJ, Mumby PJ (2019) Stage-specific effects of Lobophora on the recruitment success of a reef-building coral. Coral Reefs 38:489–498. doi: 10.1007/s00338-019- 01804-w

129 Fabricius K, De’ath G, McCook L, et al (2005) Changes in algal, coral and fish assemblages along water quality gradients on the inshore Great Barrier Reef. Mar Pollut Bull 51:384–98. doi: 10.1016/j.marpolbul.2004.10.041 Fabricius KE (2005) Effects of terrestrial runoff on the ecology of corals and coral reefs: Review and synthesis. Mar Pollut Bull 50:125–146. doi: 10.1016/j.marpolbul.2004.11.028 Feeny P (1970) Seasonal Changes in Oak Leaf Tannins and Nutrients as a Cause of Spring Feeding by Winter Moth Caterpillars. Ecology 51:565–581. Ferrari R, Gonzalez-Rivero M, Mumby PJ (2012a) Size matters in competition between corals and macroalgae. Mar Ecol Prog Ser 467:77–88. doi: 10.3354/meps09953 Ferrari R, Gonzalez-Rivero M, Ortiz JC, Mumby PJ (2012b) Interaction of herbivory and seasonality on the dynamics of Caribbean macroalgae. Coral Reefs 31:683–692. doi: 10.1007/s00338-012- 0889-9 Fišer C, Robinson CT, Malard F (2018) Cryptic species as a window into the paradigm shift of the species concept. Mol Ecol 27:613–635. doi: 10.1111/mec.14486 Fisher R, O’Leary RA, Low-Choy S, et al (2015) Species richness on coral reefs and the pursuit of convergent global estimates. Curr Biol 25:500–505. doi: 10.1016/j.cub.2014.12.022 Fletcher RL, Callow ME (1992) The settlement, attachment and establishment of marine algal spores. Br Phycol J 27:303–329. doi: 10.1080/00071619200650281 Foster NL, Box SJ, Mumby PJ (2008) Competitive effects of macroalgae on the fecundity of the reef- building coral Montastraea annularis. Mar Ecol Prog Ser 367:143–152. doi: 10.3354/meps07594 Fox RJ, Bellwood DR (2008) Remote video bioassays reveal the potential feeding impact of the rabbitfish Siganus canaliculatus (f: Siganidae) on an inner-shelf reef of the Great Barrier Reef. Coral Reefs 27:605–615. doi: 10.1007/s00338-008-0359-6 Fox RJ, Bellwood DR (2013) Niche partitioning of feeding microhabitats produces a unique function for herbivorous rabbitfishes (Perciformes, Siganidae) on coral reefs. Coral Reefs 32:13–23. doi: 10.1007/s00338-012-0945-5 Franco JN, Wernberg T, Bertocci I, et al (2015) Herbivory drives kelp recruits into ‘hiding’ in a warm ocean climate. Mar Ecol Prog Ser 536:1–9. doi: 10.3354/meps11445 Fricke A, Titlyanova T V., Nugues MM, Bischof K (2011) Depth-related variation in epiphytic communities growing on the brown alga Lobophora variegata in a Caribbean coral reef. Coral Reefs 30:967–973. doi: 10.1007/s00338-011-0772-0 Fritz RS, Hochwender CG, Lewkiewicz DA, et al (2001) Seedling herbivory by slugs in a willow hybrid system: Developmental changes in damage, chemical defense, and plant performance. Oecologia 129:87–97. doi: 10.1007/s004420100703

130 Gaylord B, Reed DC, Raimondi PT, et al (2002) A Physically Based Model of Macroalgal Spore Dispersal in the Wave and Current-Dominated Nearshore. Ecology 83:1239–1251. doi: 10.2307/3071939 Gaylord B, Reed DC, Raimondi PT, Washburn L (2006) Macroalgal Spore Dispersal in Coastal Environments : Mechanistic Insights Revealed by Theory and Experiment. Ecol Monogr 76:481–502. Gilbert LE (1971) Butterfly-Plant Coevolution: Has Passiflora adenopoda Won the Selectional Race with Heliconiine Butterflies? Science (80- ) 172:585–586. Gilchrist GW (1995) Specialists and Generalists in Changing Environments. Am Nat 146:252. doi: 10.1086/285797 Goldstein JL, Swain T (1965) The inhibition of enzymes by tannins. Phytochemistry 4:185–192. doi: 10.1016/S0031-9422(00)86162-2 González-Rivero M, Bozec Y-M, Chollett I, et al (2016) Asymmetric competition prevents the outbreak of an opportunistic species after coral reef degradation. Oecologia 181:161–173. doi: 10.1007/s00442-015-3541-x González-Rivero M, Ferrari R, Schönberg CHL, Mumby PJ (2012) Impacts of macroalgal competition and parrotfish predation on the growth of a common bioeroding sponge. Mar Ecol Prog Ser 444:133–142. doi: 10.3354/meps09424 Goodger JQD, Gleadow RM, Woodrow IE (2006) Growth cost and ontogenetic expression patterns of defence in cyanogenic Eucalyptus spp. Trees - Struct Funct 20:757–765. doi: 10.1007/s00468-006-0090-2 Graham NAJ, Jennings S, MacNeil MA, et al (2015) Predicting climate-driven regime shifts versus rebound potential in coral reefs. Nature 518:94–97. doi: 10.1038/nature14140 Graham NAJ, Nash KL (2013) The importance of structural complexity in coral reef ecosystems. Coral Reefs 32:315–326. doi: 10.1007/s00338-012-0984-y Graham NAJ, Wilson SK, Jennings S, et al (2007) Lag effects in the impacts of mass coral bleaching on coral reef fish, fisheries, and ecosystems. Conserv Biol 21:1291–300. doi: 10.1111/j.1523- 1739.2007.00754.x Green AL, Bellwood DR (2009) Monitoring Functional Groups of Herbivorous Reef Fishes as Indicators of Coral Reef Resilience: A practical guide for coral reef managers in the Asia Pacific Region. IUCN Work. Gr. Clim. Chang. Coral Reefs. IUCN, Gland. Switz. 6–70. Gruner DS, Smith JE, Seabloom EW, et al (2008) A cross-system synthesis of consumer and nutrient resource control on producer biomass. Ecol Lett 11:740–755. doi: 10.1111/j.1461- 0248.2008.01192.x Hanelt D (1998) Capability of dynamic photoinhibition in Arctic macroalgae is related to their depth

131 distribution. Mar Biol 131:361–369. doi: 10.1007/s002270050329 Harborne A, Renaud P, Tyler E, Mumby P (2009) Reduced density of the herbivorous urchin Diadema antillarum inside a Caribbean marine reserve linked to increased predation pressure by fishes. Coral Reefs 28:783–791. doi: 10.1007/s00338-009-0516-6 Harborne AR, Nagelkerken I, Wolff NH, et al (2016) Direct and indirect effects of nursery habitats on coral-reef fish assemblages, grazing pressure and benthic dynamics. Oikos 125:957–967. doi: 10.1111/oik.02602 Harrison XA (2014) Using observation-level random effects to model overdispersion in count data in ecology and evolution. PeerJ 2:e616. doi: 10.7717/peerj.616 Haslam E (1974) Polyphenol-protein interactions. Biochem J 139:285–288. Hay ME (1981) Spatial patterns of agrazing intensity on a caribbean barrier reef: Herbivory and algal distribution. Aquat Bot 11:97–109. doi: 10.1016/0304-3770(81)90051-6 Hebert PDN, Penton EH, Burns JM, et al (2004) Ten species in one: DNA barcoding reveals cryptic species in the neotropical skipper butterfly Astraptes fulgerator. Proc Natl Acad Sci 101:14812– 14817. doi: 10.1073/pnas.0406166101 Hector A, Schmid B, Beierkuhnlein C, et al (1999) Plant diversity and productivity experiments in European grasslands. Science (80) 286:1123–1127. doi: 10.1126/science.286.5442.1123 Herms DA, Mattson WJ (1992) The Dilemma of Plants: To Grow or Defend. Wuarterly Rev Biol 67:283–335. Hicks CC, Cinner JE, Stoeckl N, Mcclanahan TR (2015) Linking ecosystem services and human- values theory. Conserv Biol 29:1471–1480. doi: 10.1111/cobi.12550 Hoegh-Guldberg O (1999) Climate change, coral bleaching and the future of the world’s coral reefs. Mar Freshw Res 50:839–66. doi: 10.1071/MF99078 Hoegh-Guldberg O, Mumby PJ, Hooten a J, et al (2007) Coral reefs under rapid climate change and ocean acidification. Science 318:1737–1742. doi: 10.1126/science.1152509 Hoey AS, Bellwood DR (2009) Limited Functional Redundancy in a High Diversity System: Single Species Dominates Key Ecological Process on Coral Reefs. Ecosystems 12:1316–1328. doi: 10.1007/s10021-009-9291-z Hoey AS, Bellwood DR (2011) Suppression of herbivory by macroalgal density: a critical feedback on coral reefs? Ecol Lett 14:267–73. doi: 10.1111/j.1461-0248.2010.01581.x Holm S (1979) A Simple Sequentially Rejective Multiple Test Procedure. Scand J Stat 6:65–70. Horn MH (1989) Biology of marine herbivorous fishes. Oceanogr Mar Biol An Annu Rev 27:167– 272. Hothorn T, Bretz F, Westfall P (2008) Simultaneous Inference in General Parametric Models. Biometrical J 50:346–363.

132 Hughes TP (1994) Catastrophes, Phase Shifts, and Large-Scale Degradation of a Caribbean Coral Reef. Science (80) 265:1547–1551. doi: 10.1126/science.265.5178.1547 Hughes TP, Anderson KD, Connolly SR, et al (2018) Spatial and temporal patterns of mass bleaching of corals in the Anthropocene. Science (80- ) 359:80–83. Hughes TP, Baird AH, Bellwood DR, et al (2003) Climate change, human impacts, and the resilience of coral reefs. Science 301:929–33. doi: 10.1126/science.1085046 Hughes TP, Kerry JT, Álvarez-Noriega M, et al (2017) Global warming and recurrent mass bleaching of corals. Nature 543:373–377. doi: 10.1038/nature21707 Hughes TP, Rodrigues MJ, Bellwood DR, et al (2007) Phase shifts, herbivory, and the resilience of coral reefs to climate change. Curr Biol 17:360–365. doi: 10.1016/j.cub.2006.12.049 Hurd CL, Harrison PJ, Druehl LD (1996) Effect of seawater velocity on inorganic nitrogen uptake by morphologically distinct forms of Macrocystis integrifolia from wave-sheltered and exposed sites. Mar Biol 126:205–214. Jermy T (1984) Evolution of insect/host plant relationships. Am Nat 124:609–630. Johns KA, Emslie MJ, Hoey AS, et al (2018) Macroalgal feedbacks and substrate properties maintain a coral reef regime shift. Ecosphere. doi: 10.1002/ecs2.2349 Jompa J, McCook LJ (2002a) The effects of nutrients and herbivory on competition between a hard coral (Porites cylindrica) and a brown alga (Lobophora variegata). Limnol Oceanogr 47:527– 534. doi: 10.4319/lo.2002.47.2.0527 Jompa J, McCook LJ (2002b) Effects of competition and herbivory on interactions between a hard coral and a brown alga. J Exp Mar Bio Ecol 271:25–39. doi: 10.1016/S0022-0981(02)00040-0 Kautsky L, Kautsky H (1989) Algal species diversity and dominance along gradients of stress and disturbance in marine environments. Vegetatio 83:259–267. Kearsley MJC, Whitham TG (1989) Developmental Changes in Resistance to Herbivory: Implications for Individuals and Populations. Ecology 70:422–434. Kuempel CD, Altieri AH (2017) The emergent role of small-bodied herbivores in pre-empting phase shifts on degraded coral reefs. Sci Rep 7:39670. doi: 10.1038/srep39670 Kuffner IB, Walters LJ, Becerro MA, et al (2006) Inhibition of coral recruitment by macroalgae and cyanobacteria. Mar Ecol Prog Ser 323:107–117. doi: 10.3354/meps323107 Kulbicki M, Guillemont N, Amand M (2005) A general approach to length-weight relationships for New Caledonian lagoon fishes. Cybium 29:235–252. Ladner JT, Palumbi SR (2012) Extensive sympatry, cryptic diversity and introgression throughout the geographic distribution of two coral species complexes. Mol Ecol 21:2224–2238. doi: 10.1111/j.1365-294X.2012.05528.x Lawton JH, Brown VK (1994) Redundancy in Ecosystems. In: Biodiversity and Ecosystem Function.

133 pp 255–270 Ledlie MH, Graham NAJ, Bythell JC, et al (2007) Phase shifts and the role of herbivory in the resilience of coral reefs. Coral Reefs 26:641–653. doi: 10.1007/s00338-007-0230-1 Leigh EGJ, Paine RT, Quinn JF, Suchanek TH (1987) Wave energy and intertidal productivity. Proc Natl Acad Sci U S A 84:1314–1318. doi: 10.1073/pnas.84.5.1314 Leliaert F, Verbruggen H, Vanormelingen P, et al (2014) DNA-based species delimitation in algae. Eur J Phycol 49:179–196. doi: 10.1080/09670262.2014.904524 Lenth R V. (2016) Least-Squares Means: The R Package lsmeans. J Stat Softw. doi: 10.18637/jss.v069.i01 Leray M, Knowlton N (2016) Censusing marine eukaryotic diversity in the twenty-first century. Philos Trans R Soc B Biol Sci. doi: 10.1098/rstb.2015.0331 Lewis SM (1986) The Role of Herbivorous Fishes in the Organization of a Caribbean Reef Community. Ecol Monogr 56:183–200. doi: 10.2307/2937073 Lewis SM (1985) Herbivory on coral reefs: algal susceptibility to herbivorous fishes. Oecologia 65:370–375. doi: 10.1007/BF00378911 Lewis SM, Norris JN, Searles RB (1987) The Regulation of Morphological Plasticity in Tropical Reef Algae by Herbivory. Ecology 68:636–641. Littler M, Taylor P, Littler D (1983) Algal resistance to herbivory on a Caribbean barrier reef. Coral Reefs 2:111–118. doi: 10.1007/bf02395281 Littler MM, Littler DS (2011) Algae, Turf. In: D H (ed) Ecyclopedia of Modern Coral Reefs. Encyclopedia of Earth Sciences Series. Springer, Dordrecht, Littler MM, Littler DS, Brooks BL (2006) Harmful algae on tropical coral reefs: Bottom-up eutrophication and top-down herbivory. Harmful Algae 5:565–585. Loffler Z, Hoey AS (2019) Microtopographic refuges enhance recruitment and survival but inhibit growth of propagules of the tropical macroalga Sargassum swartzii. Mar Ecol Prog Ser 627:61– 70. doi: 10.3354/meps13057 Loney PE, McArthur C, Potts BM, Jordan GJ (2006) How does ontogeny in a Eucalyptus species affect patterns of herbivory by Brushtail Possums? Funct Ecol 20:982–988. doi: 10.1111/j.1365- 2435.2006.01193.x Loreau M, Naeem S, Inchausti P, et al (2001) Biodiversity and ecosystem functioning: Current knowledge and future challenges. Science 294:804–8. doi: 10.1126/science.1064088 Lotze HK, Schramm W, Schories D, Worm B (1999) Control of macroalgal blooms at early developmental stages: Pilayella littoralis versus Enteromorpha spp. Oecologia 119:46–54. doi: 10.1007/s004420050759 Lubchenco J (1983) Littorina and Fucus: Effects of Herbivores, Substratum Heterogeneity, and Plant

134 Escapes During Succession. Ecology 64:1116–1123. doi: https://doi.org/10.2307/1937822 Lubchenco J, Gaines SD (1981) A Unified Approach to Marine Plant-Herbivore Interactions. I. Populations and Communities. Annu Rev Ecol Syst 12:405–437. Luckhurst E, Luckhurst K (1978) Analysis of the Influence of Substrate Variables on Coral Reef Fish Communities. Mar Biol 323:317–323. doi: 10.1007/BF00455026 Mantyka CS, Bellwood DR (2007) Direct evaluation of macroalgal removal by herbivorous coral reef fishes. Coral Reefs 26:435–442. doi: 10.1007/s00338-007-0214-1 Marshell A, Mumby PJ (2015) The role of surgeonfish (Acanthuridae) in maintaining algal turf biomass on coral reefs. J Exp Mar Bio Ecol 473:152–160. doi: 10.1016/j.jembe.2015.09.002 Marshell AL (2014) The grazing impact of common surgeonfish on algal dynamics of the Great Barrier Reef. PhD thesis, The Universisty of Queensland, Australia McClanahan TR, Muthiga NA (1998) An ecological shift in a remote coral atoll of Belize over 25 years. Environ Conserv 25:122–130. doi: 10.1017/S0376892998000174 McConnell OJ, Hughes PA, Targett NM, Daley J (1982) Effects of secondary metabolites from marine algae on feeding by the sea urchin, Lytechinus variegatus. J Chem Ecol 1437–1453. McCook L, Jompa J, Diaz-Pulido G (2001) Competition between corals and algae on coral reefs: a review of evidence and mechanisms. Coral Reefs 19:400–417. doi: 10.1007/s003380000129 McCook LJ (1999) Macroalgae, nutrients and phase shifts on coral reefs: Scientific issues and management consequences for the Great Barrier Reef. Coral Reefs 18:357–367. Melbourne-Thomas J, Johnson CR, Fulton EA (2011) Regional-scale scenario analysis for the Meso- American Reef system: Modelling coral reef futures under multiple stressors. Ecol Modell 222:1756–1770. doi: 10.1016/j.ecolmodel.2011.03.008 Meleg IN, Zakšek V, Fišer C, et al (2013) Can Environment Predict Cryptic Diversity? The Case of Niphargus Inhabiting Western Carpathian Groundwater. PLoS One 8:e76760. doi:10.1371/journal.pone.0076760. doi: 10.1371/journal.pone.0076760 Menge BA, Lubchenco J (1981) Community Organization in Temperate and Tropical Rocky Intertidal Habitat: Prey Refuges in Relation to Consumer Pressure Gradients. Ecol Monogr 51:429–450. doi: https://doi.org/10.2307/2937323 Miller MW, Hay ME, Miller SL, et al (1999) Effects of nutrients versus herbivores on reef algae: A new method for manipulating nutrients on coral reefs. Limnol Oceanogr 44:1847–1861. doi: 10.4319/lo.1999.44.8.1847 Moberg F, Folke C (1999) Ecological goods and services of coral reef ecosystems. Ecol Econ 29:215– 233. doi: 10.1016/S0921-8009(99)00009-9 Möllmann C, Diekmann R, Müller-Karulis B, et al (2009) Reorganization of a large marine ecosystem due to atmospheric and anthropogenic pressure: A discontinuous regime shift in the Central

135 Baltic Sea. Glob Chang Biol 15:1377–1393. doi: 10.1111/j.1365-2486.2008.01814.x Morrison D (1988) Comparing Fish and Urchin Grazing in Shallow and Deeper Coral Reef Algal Communities. Ecology 69:1367–1382. Morrow KM, Bromhall K, Motti CA, et al (2016) Allelochemicals produced by Brown Macroalgae of the genus Lobophora are Active against Coral Larvae and Associated Bacteria, Supporting Pathogenic Shifts to Vibrio Dominance. Appl Environ Microbiol 83:AEM.02391-16. doi: 10.1128/AEM.02391-16 Mouillot D, Villéger S, Parravicini V, et al (2014) Functional over-redundancy and high functional vulnerability in global fish faunas of tropical reefs. Proc Natl Acad Sci 111:13757–13762. doi: 10.1073/pnas.1317625111 Mshigeni KE (1978) Field Observations on the Colonization of New Substrata and Denuded Intertidal Surfaces by Benthic Macrophytic Algae. Bot Mar 21:49–58. doi: 10.1515/botm.1978.21.1.49 Muko S, Sakai K, Iwasa Y (2001) Size distribution dynamics for a marine sessile organism with space-limitation in growth and recruitment: Application to a coral population. J Anim Ecol 70:579–589. doi: 10.1046/j.1365-2656.2001.00513.x Müller DG, Gassmann G, Lüning K (1979) Isolation of a spermatozoid-releasing and -attracting substance from female gametophytes of Laminaria digitata [18]. Nature 279:430–431. Mumby PJ (2009) Phase shifts and the stability of macroalgal communities on Caribbean coral reefs. Coral Reefs 28:761–773. doi: 10.1007/s00338-009-0506-8 Mumby PJ, Foster NL, Fahy EAG (2005) Patch dynamics of coral reef macroalgae under chronic and acute disturbance. Coral Reefs 24:681–692. doi: 10.1007/s00338-005-0058-5 Mumby PJ, Hastings A, Edwards HJ (2007) Thresholds and the resilience of Caribbean coral reefs. Nature 450:98–101. doi: 10.1038/nature06252 Mumby PJ, Steneck RS, Adjeroud M, Arnold SN (2016) High resilience masks underlying sensitivity to algal phase shifts of Pacific coral reefs. Oikos 125:644–655. doi: 10.1111/oik.02673 Mumby PJ, Steneck RS, Hastings A (2013) Evidence for and against the existence of alternate attractors on coral reefs. Oikos 122:481–491. doi: 10.1111/j.1600-0706.2012.00262.x Nash KL, Graham NAJ, Januchowski-Hartley FA, Bellwood DR (2012) Influence of habitat condition and competition on foraging behaviour of parrotfishes. Mar Ecol Prog Ser 457:113– 124. doi: 10.3354/meps09742 Nozawa Y (2008) Micro-crevice structure enhances coral spat survivorship. J Exp Mar Bio Ecol 367:127–130. doi: 10.1016/j.jembe.2008.09.004 Nugues MM, Bak RPM (2008) Long-term dynamics of the brown macroalga Lobophora variegata on deep reefs in Curaçao. Coral Reefs 27:389–393. doi: 10.1007/s00338-007-0346-3 Nugues MM, Bak RPM (2006) Differential competitive abilities between Caribbean coral species

136 and a brown alga: a year of experiments and a long-term perspective. Mar Ecol Prog Ser 315:75– 86. doi: 10.3354/meps315075 Nyström M, Graham NAJ, Lokrantz J, Norström A V. (2008) Capturing the cornerstones of coral reef resilience: linking theory to practice. Coral Reefs 27:795–809. doi: 10.1007/s00338-008- 0426-z Obura D, Grimsditch G (2009) Resilience Assessment of Coral Reefs bleaching and thermal stress. Ogden JC (1976) Some aspects of herbivore-plant relationships on Caribbean reefs and seagrass beds. Aquat Bot 2:103–116. doi: 10.1016/0304-3770(76)90013-9 Oksanen AJ, Blanchet FG, Friendly M, et al (2019) The Vegan Package, R package version 2.5-5. Painter T J (1983) Algal polysaccharides. In: Aspinall GO (ed) The polysaccharides. Vol. 2. Academic Press, New York. pp 195–285 Paul VJ, Fenical W (1983) Isolation of Halimedatrial: Chemical Defense Adaptation in the Calcareous Reef-Building Alga Halimeda. Science (80- ) 221:747–749. doi: 10.1126/science.221.4612.747 Paul VJ, Hay ME (1986) Seaweed susceptibility to herbivory: chemical and morphological correlates. Mar Ecol Prog Ser 33:255–264. doi: 10.3354/meps033255 Peckol P, Searles RB (1984) Temporal and Spatial Patterns of Growth and Survival of Invertebrate and Algal Populations of a North Carolina Continental Shelf Community. Estuar Coast Shelf Sci 18:133–143. Pedersen MF, Nejrup LB, Fredriksen S, et al (2012) Effects of wave exposure on population structure, demography, biomass and productivity of the kelp Laminaria hyperborea. Mar Ecol Prog Ser 451:45–60. doi: 10.3354/meps09594 Petraitis PS, Latham RE (1999) The importance of scale in testing the origins of alternative community states. Ecology 80:429–442. Pillans RD, Franklin CE, Tibbetts IR (2004) Food choice in Siganus fuscescens: influence of macrophyte nutrient content and availability. J Fish Biol 64:297–309. doi: 10.1111/j.0022- 1112.2004.00261.x Plass-Johnson JG, Ferse SC, Jompa J, et al (2015) Fish herbivory as key ecological function in a heavily degraded coral reef system. Limnol Oceanogr 1–33. doi: 10.1002/lno.10105 Polunin NV., Klumpp D. (1992) Algal food supply and grazer demand in a very productive coral- reef zone. J Exp Mar Bio Ecol 164:1–15. doi: 10.1016/0022-0981(92)90132-T Poore AGB, Campbell AH, Coleman RA, et al (2012) Global patterns in the impact of marine herbivores on benthic primary producers. Ecol Lett 15:912–922. doi: 10.1111/j.1461- 0248.2012.01804.x Poray AK, Carpenter RC (2014) Distributions of coral reef macroalgae in a back reef habitat in

137 Moorea, French Polynesia. Coral Reefs 33:67–76. doi: 10.1007/s00338-013-1104-3 Puk LD, Ferse SCA, Wild C (2016) Patterns and trends in coral reef macroalgae browsing: a review of browsing herbivorous fishes of the Indo-Pacific. Rev Fish Biol Fish 26:53–70. doi: 10.1007/s11160-015-9412-z Purcell SW, Bellwood DR (1993) A functional analysis of food procurement in two surgeonfish species, Acanthurus nigrofuscus and Ctenochaetus striatus (Acanthuridae). Environ. Biol. Fishes 37:139–159. Randall JE (1955) A contribution to the biology of the Acanthuridae (Surgeon Fishes). Rasher DB, Engel S, Bonito V, et al (2012) Effects of herbivory, nutrients, and reef protection on algal proliferation and coral growth on a tropical reef. Oecologia 169:187–198. doi: 10.1007/s00442-011-2174-y Rasher DB, Hay ME (2010) Chemically rich seaweeds poison corals when not controlled by herbivores. Proc Natl Acad Sci U S A 107:9683–8. doi: 10.1073/pnas.0912095107 Rasher DB, Hoey AS, Hay ME (2013) Consumer diversity interacts with prey defenses to drive ecosystem function. Ecology 94:1347–1358. doi: 10.1890/12-0389.1 Reed DC, Laur DR, Ebeling AW (1988) Variation in Algal Dispersal and Recruitment : The Importance of Episodic Events. Ecol Monogr 58:321–335. Renken H, Mumby PJ, Matsikis I, Edwards HJ (2010) Effects of physical environmental conditions on the patch dynamics of Dictyota pulchella and Lobophora variegata on Caribbean coral reefs. Mar Ecol Prog Ser 403:63–74. doi: 10.3354/meps08441 Robertson DR, Polunin NVC, Leighton K (1979) The behavioral ecology of three Indian Ocean surgeonfishes (Acanthurus lineatus, A. leucosternon and Zebrasoma scopas): their feeding strategies, and social and mating systems. Environ Biol Fishes 4:125–170. doi: 10.1007/BF00005448 Roff G, Bejarano S, Priest M, et al (2019) Seascapes as drivers of herbivore assemblages in coral reef ecosystems. Ecol Monogr. doi: 10.1002/ecm.1336 Roff G, Chollett I, Doropoulos C, et al (2015a) Exposure-driven phase shift following catastrophic disturbance on coral reefs. Coral Reefs 34:715–725. doi: 10.1007/s00338-015-1305-z Roff G, Doropoulos C, Zupan M, et al (2015b) Phase shift facilitation following cyclone disturbance on coral reefs. Oecologia 178:1193–1203. doi: 10.1007/s00442-015-3282-x Rogers A, Blanchard JL, Mumby PJ (2014) Vulnerability of coral reef fisheries to a loss of structural complexity. Curr Biol 24:1000–1005. doi: 10.1016/j.cub.2014.03.026 Rogers A, Harborne AR, Brown CJ, et al (2015) Anticipative management for coral reef ecosystem services in the 21st century. Glob Chang Biol 21:504–514. doi: 10.1111/gcb.12725 Rogers CS, Miller J (2006) Permanent “phase shifts” or reversible declines in coral cover? Lack of

138 recovery of two coral reefs in St. John, US Virgin Islands. Mar Ecol Prog Ser 306:103–114. Rogers RW (1997) Brown Algae on Heron Reef Flat, Great Barrier Reef, Australia: Spatial, Seasonal and Secular Variation in Cover. Bot Mar 40:113–117. Roughgarden J, Iwasa YOH, Baxter C (1985) Demographic Theory for an Open Marine Population with Space-Limited Recruitment. Ecology 66:54–67. Russ GR (1984) Distribution and abundance of herbivorous grazing fishes in the central Great Barrier Reef. II: Patterns of zonation of mid-shelf and outershelf reefs. Mar Ecol Prog Ser 20:35–44. Sammarco PW (1983) Effects of fish grazing and damselfish territoriality on coral reef algae. I.Algal community structure. Mar Ecol Prog Ser 13:1–14. doi: Doi 10.3354/Meps013001 Sangil C, Sansón M, Afonso-Carrillo J (2011) Spatial variation patterns of subtidal seaweed assemblages along a subtropical oceanic archipelago: Thermal gradient vs herbivore pressure. Estuar Coast Shelf Sci 94:322–333. doi: 10.1016/j.ecss.2011.07.004 Saunders DR, Wiggins HS (1981) Conservation and raffinose by the human colon. Am J Physiol 241:397–402. Schultz NE, Lane CE, Gall L Le, et al (2015) A barcode analysis of the genus Lobophora (Dictyotales, Phaeophyceae) in the western Atlantic Ocean with four novel species and the epitypification of L. variegata (J.V. Lamouroux) E.C. Oliveira. Eur J Phycol 50:481–500. doi: 10.1080/09670262.2015.1078500 Seeto GS, Veivers PC, Clements KD, Slaytor M (1996) Carbohydrate utilisation by microbial symbionts in the marine herbivorous fishes Odax cyanomelas and Crinodus lophodon. J Comp Physiol B 165:571–579. doi: 10.1007/BF00387519 Shitzer D, Noy-Meir I, Milchunas DG (2008) The role of geologic grazing refuges in structuring Mediterranean grassland plant communities. Plant Ecol 198:135–147. doi: 10.1007/s11258-007- 9391-9 Slattery M, Lesser MP (2014) Allelopathy in the tropical alga Lobophora variegata (Phaeophyceae): mechanistic basis for a phase shift on mesophotic coral reefs? J Phycol 50:493–505. doi: 10.1111/jpy.12160 Spiegel LH, Price PW (1996) Plant aging and the distribution of Rhyacionia neomexicana (Lepidoptera: Tortricidae). Environ Entomol 25:359–365. Steinberg PD, Paul VJ (1990) Fish feeding and chemical defenses of tropical brown algae in Western Australia. Mar Ecol Prog Ser 58:253–259. doi: 10.3354/meps058253 Steneck RS (1988) Herbivory on coral reefs: a synthesis. Proc 6th Int Coral Reef Symp 1:37–49. Steneck RS, Bellwood DR, Hay ME (2017) Herbivory in the marine realm. Curr Biol 27:R484–R489. doi: 10.1016/j.cub.2017.04.021 Stern JL, Hagerman AE, Steinberg PD, Mason PK (1996) Phlorotannin-protein interactions. J Chem

139 Ecol 22:1877–1899. Stireman III JO, Nason JD, Heard SB (2005) Host-Associated Genetic Differentiation in Phytophagous Insects: General Phenomenon or Isolated Exceptions? Evidence From a Goldenrod-Insect Community. Evolution (N Y) 59:2573–2587. doi: 10.1554/05-222.1 Sun Z, Hanyuda T, Lim P-E, et al (2012) Taxonomic revision of the genus Lobophora (Dictyotales, Phaeophyceae) based on morphological evidence and analyses rbcL and cox3 gene sequences. Phycologia 51:500–512. doi: 10.2216/11-85.1 Symstad AJ, Tilman D (2001) Diversity loss, recruitment limitation, and ecosystem functioning: lessons learned from a removal experiment. Oikos 92:424–435. Szmant AM (2002) Nutrient Enrichment on Coral Reefs : Is It a Major Cause of Coral Reef Decline? Estuaries 25:743–766. Targett NM, Arnold TM (2001) Effects of Secondary Metabolites on Digestion in Marine Herbivores. In: Marine Chemical ecology. Targett NM, Boettcher AA, Targett TE, Vrolijk NH (1995) Tropical marine herbivore assimilation of phenolic-rich plants. Oecologia 103:170–179. Tews J, Brose U, Grimm V, et al (2004) Animal species diversity driven by habitat heterogeneity/diversity: the importance of keystone structures. J Biogeogr 31:79–92. doi: 10.1046/j.0305-0270.2003.00994.x Thacker RW, Ginsburg DW, Paul VJ (2001) Effects of herbivore exclusion and nutrient enrichment on coral reef macroalgae and cyanobacteria. Coral Reefs 19:318–329. doi: 10.1007/s003380000122 Tilman D, Knops J, Wedin D, et al (1997) The influence of functional diversity and composition on ecosystem processes. Science (80- ) 277:1300–1302. doi: 10.1126/science.277.5330.1300 Tilman D, May RM, Lehman CL, Nowak MA (1994) Habitat destruction and the extinction debt. Nature 371:65–66. doi: 10.2307/2269483 Topor ZM, Rasher DB, Duffy JE, Brandl SJ (2019) Marine protected areas enhance coral reef functioning by promoting fish biodiversity. Conserv Lett 1–9. doi: 10.1111/conl.12638 van den Hoek C, Breeman AM, Bak RPM, Vanbuurt G (1978) Distribution of Algae, Corals and Gorgonians in Relation To Depth, Light Attenuation, Water-Movement and Grazing Pressure in Fringing Coral-Reef of Curacao, Netherlands Antilles. Aquat Bot 5:1–46. doi: 10.1016/0304- 3770(78)90045-1 Vieira C (2019) Lobophora-coral interactions and phase shifts: summary of current knowledge and future directions. Vieira C, Camacho O, Sun Z, et al (2017) Historical biogeography of the highly diverse brown seaweed Lobophora (Dictyotales, Phaeophyceae). Mol Phylogenet Evol 110:81–92. doi:

140 10.1016/j.ympev.2017.03.007 Vieira C, Camacho O, Wynne MJ, et al (2016a) Shedding new light on old algae: Matching names and sequences in the brown algal genus Lobophora (Dictyotales, Phaeophyceae). Taxon 65:689– 707. doi: 10.12705/654.1 Vieira C, D’hondt S, De Clerck O, Payri CE (2014) Toward an inordinate fondness for stars, beetles and Lobophora? Species diversity of the genus Lobophora (Dictyotales, Phaeophyceae) in New Caledonia. J Phycol 50:1101–1119. doi: 10.1111/jpy.12243 Vieira C, Engelen AH, Guentas L, et al (2016b) Species specificity of bacteria associated to the brown seaweeds Lobophora (Dictyotales, Phaeophyceae) and their potential for induction of rapid coral bleaching in Acropora muricata. Front Microbiol 7:1–13. doi: 10.3389/fmicb.2016.00316 Vieira C, Payri C, De Clerck O (2015) Overgrowth and killing of corals by the brown alga Lobophora hederacea (Dictyotales, Phaeophyceae) on healthy reefs in New Caledonia: A new case of the epizoism syndrome. Phycol Res 63:152–153. doi: 10.1111/pre.12082 Vieira C, Stenger P-L, Moleana T, et al (2019) Limited interspecific variation in grazing susceptibility of the brown alga Lobophora to herbivory. J Exp Mar Bio Ecol 518:151175. doi: 10.1016/j.jembe.2019.151175 Vieira C, Thomas OP, Culioli G, et al (2016c) Allelopathic interactions between the brown algal genus Lobophora (Dictyotales, Phaeophyceae) and scleractinian corals. Sci Rep 6:18637. doi: 10.1038/srep18637 Warner PA, Van Oppen MJH, Willis BL (2015) Unexpected cryptic species diversity in the widespread coral Seriatopora hystrix masks spatial-genetic patterns of connectivity. Mol Ecol 24:2993–3008. doi: 10.1111/mec.13225 Watanuki A, Yamamoto A (1990) Settlement of seaweeds on coastal structures. Hydrobiologia 204– 205:275–280. doi: 10.1007/BF00040245 Weidner K, Lages BG, Da Gama BAP, et al (2004) Effect of mesograzers and nutrient levels on induction of defenses in several Brazilian macroalgae. Mar Ecol Prog Ser 283:113–125. doi: 10.3354/meps283113 Wellborn GA, Cothran RD (2004) Phenotypic similarity and differentiation among sympatric cryptic species in a freshwater amphipod species complex. Freshw Biol 49:1–13. doi: 10.1046/j.1365- 2426.2003.01160.x Welsh JQ, Bellwood DR (2011) Spatial ecology of the steephead parrotfish (Chlorurus microrhinos): an evaluation using acoustic telemetry. Coral Reefs 31:55–65. doi: 10.1007/s00338-011-0813- 8 Wheeler WN (1980) Effect of boundary layer transport on the fixation of carbon by the giant kelp Macrocystis pyrifera. Mar Biol 56:103–110. doi: 10.1007/BF00397128

141 Whittaker RH, Feeny PP (1971) Chemical Interactions between Species. Science (80- ) 280:564–567. Wiencke C, Bischof K (2012) Seaweed Biology: Novel Insights into Ecophysiology, Ecology and Utilization. Williams GJ, Graham NAJ, Jouffray JB, et al (2019) Coral reef ecology in the Anthropocene. Funct Ecol 33:1014–1022. doi: 10.1111/1365-2435.13290 Williams ID, Polunin NVC, Hendrick VJ (2001) Limits to grazing by herbivorous fishes and the impact of low coral cover on macroalgal abundance on a coral reef in Belize. Mar Ecol Prog Ser 222:187–196. doi: 10.3354/meps222187 Woodward G, Gessner MO, Giller PS, et al (2012) Continental-scale effects of nutrient pollution on stream ecosystem functioning. Science (80) 336:1438–1440. doi: 10.1017/cbo9780511781216.005 Zucker W V (1983) Tannins: Does Structure Determine Function? An Ecological Perspective. Am Nat 121:335–365.

142 Appendices Animal Ethics Approval

UQ Research and Innovation Director, Research Management Office Nicole Thompson Animal Ethics Approval Certificate 21-Oct-2016 Please check all details below and inform the Animal Welfare Unit within 10 working days if anything is incorrect.

Activity Details Chief Investigator: Professor Peter Mumby, Biological Sciences Title: Population dynamics of the brown macroalga Lobophora sp. under fish herbivory AEC Approval Number: SBS/396/16/PALAU Previous AEC Number: Approval Duration: 01-Nov-2016 to 01-Nov-2019 Funding Body: ARC Group: Native and exotic wildlife and marine animals Other Staff/Students: Laura Puk, Nicolas Evensen

Location(s): Other International Location

Summary Subspecies Strain Class Gender Source Approved Remaining Fish Adults Unknown 572 572

Permits

Provisos

Approval Details

Description Amount Balance

Fish (Unknown, Adults, ) 11 Oct 2016 Initial approval 572 572

Please note the animal numbers supplied on this certificate are the total allocated for the approval duration

Please use this Approval Number: 1. When ordering animals from Animal Breeding Houses 2. For labelling of all animal cages or holding areas. In addition please include on the label, Chief Investigator's name and contact phone number. 3. When you need to communicate with this office about the project.

It is a condition of this approval that all project animal details be made available to Animal House OIC. (UAEC Ruling 14/12/2001)

The Chief Investigator takes responsibility for ensuring all legislative, regulatory and compliance objectives are satisfied for this project. This certificate supercedes all preceeding certificates for this project (i.e. those certificates dated before 21-Oct-2016)

Animal Welfare Unit Cumbrae-Stewart Building +61 7 336 52925 (Enquiries) [email protected] UQ Research and Innovation Research Road +61 7 334 68710 (Enquiries) uq.edu.au/research The University of Queensland Brisbane Qld 4072 Australia +61 7 336 52713 (Coordinator) Page 1 of 1

143