|UFDC Home||myUFDC Home | Help|
1 INDIRECT EFFECTS OF A MARINE ECOSYSTEM ENGINEER ALTER THE ABUNDANCE AND DISTRIBUTION OF FOUNDATION CORAL SPECIES By JADA SIMONE SHANTI WHITE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010
2 2010 JadaSimone Shanti White
3 To all my Mother and my Sammy
4 ACKNOWLEDGMENTS First and foremost, I thank my Mot her She was my provider and my inspiration: she taught me to value integrity and that contributing to knowledge and education was a noble career path. My Grandmother, Sammy (Eileen) Ringold, also provided a positive role model and broadened my horizons in various ways throughout my life. My Father, William B. White, and other Grandparents, Ross D. Reeves, and Elmer and Nancy White, offered support, inspiration, and encouragement throughout my life. My first employer, Dr. John B. Roberts, and my first mentor in Field Ecology, Albin Bills, taught me about graduate school and fostered my belief that, through science and education, I could have a positive influence on the world. My amazing brothers, Jaaron, Tory, and Tealson, as well as my dear chosen sisters, Cailyn, Kendee Rae, Jen Rose, and Erica Lena, offered unconditional support and love; despite my, often ridiculous, travel schedules. My wonderful nephews, Sage, Shilo, and Seth, and the rest of the beloved children of our community, especially Ayden, Laur en, and Thomas, reminded me through our snorkeling adventures that the youth have a yearning desire to know about the natural world and that a life in pursuit of this knowledge was valuable and inspiring. I am extremely grateful to Benjamin M. Bolker, Gust av Paulay, and members of my supervisory committee, Karen Bjorndal, Bill Lindberg, Brian R. Silliman, and Colette M. St. Mary, for their mentoring, support, and, especially, for providing exceptional models for academic and professional conduct for me to g lean from (in countless ways!). Craig Osenberg taught me to value critical exchange, reinforced the strong experimental design training I had gained at UCSB, and improved my writing significantly, for which I will always be grateful (and no, I do not belie ve that the error bars ever overlapped zero). Ben Bolker taught me to write using outlines, how to be
5 quantitative, in particular, he improved both my accuracy at, and understanding of, ecological inference, and, of course, he taught me how to program in R; while Gustav Paulay corrected my natural history knowledge and improved my precision in species identifications, writing, and scientific presentations. These skills were invaluable to my success as a graduate student and will be treasured and passed on with pride. Karen Bjorndal reinforced the importance of nutritional ecology and, through her work as Director of the Archie Carr Sea Turtle Center and a seminar in Sea Turtle Biology, exemplified the multi faceted approach that is necessary do effective c onservation science. Bill Lindberg broadened my critical thinking skills, as well as my historical knowledge of scientific philosophies. Brian Sillimans enthusiasm was contagious and he taught me to focus on the main story, first and foremost. Colette St. Mary exemplified the linking theory with data that I had learned to strive toward as an undergraduate and offered support and mentoring at critical times. The St. Mary Osenberg Bolker Lab fostered a love of critical exchange and offered the fine polishing that I needed to grow; while the Paulay lab supported my knowledge (and love) of natural history, including tried and true species identification, collection, and curation skills, as well as cutting edge approaches. Marta Wayne, was a powerful role model and offered crucial support as the Faculty Advisor of our grass roots organization, Women in Science and Engineering; she was also a dear friend. I am also grateful to the staff and members at the Department of Biology, in particular Karen Patterson and Cathy Moore, for their countless assistance through the years. Daniella Ceccarelli, Richard Aronson, and one anonymous reviewer offered helpful comments on my second chapter and improved the quality of the work immensely.
6 I give mauruuru roa (large thanks) t o Jimmy ODonnell, Shelby Boyer, Emily Vuxton, and Crystal Hartman. These students invested long hours in collecting both my, and their own data, and contributed to our collective understanding of the system. I also gained invaluable insight into the exper ience of advising through the process of mentoring their senior honors theses and I will always be grateful they chose to work with me. Shelby, in particular, was an exceptional research assistant (and dear friend) through several field seasons and nearly a year in the field (collectively). She was hard working, independent, honest, and caring, and I really appreciate her countless hours of effort and, especially, her big smile at the end of every long day. My roommates at Gump for many years, Nichole Price and Jennifer Lape, were my dearest friends in Moorea and I will always love and cherish them for their unconditional support. Erica Spotswood and David Hembry filled their void when they graduated and moved to greener pastures, and I really appreciate our wonderful nights on the hill. The other graduate students working at Gump contributed to my development, as a scientist and as a person, in countless ways: Tom Adam, Gerick Bergsma, Carol Chaffee, Adrien Delval, Anne Duplouy, Robin Elahi, Shane Geange, Da nny Green, Will Goldenheim, Nick Haring, Rebecca Habeeb, Emily Hornett, Joshua Idjadi, Jenny Kahn, Kate ( Hansen) Karakyl, Maireid Maheighn, Nancy Muehllehner, Jennie Oates, Abigail Poray, Hollie Putnam, Alice Rogers, Melissa Spitler, Hannah Stewart, Adri an Stier, Stephanie Talmage, and Annie Yau. As did our regular visitors from the R.V. Braveheart (they know who they are!). And, of course, my own graduate student mentors from UCSB, Katie Arkema, Sarah Lester, Scott Hamilton, Ben Ruttenberg, Andrew Rassweiler, Jamael Samhouri, Jeff Shima, Rick Wilder, Will White,
7 and, especially, Andrew Thompson, who showed me the way through their good work and offered support and encouragment, both prior to, and throughout my graduate studies. I will always cherish the m emories and I look forward to creating more at WSN, ESA, and ICRS, or elsewhere in the Pacific! Folks that cycled in and out of Gump sporadically were also instrumental, especially the following, who volunteered countless hours and helped collect the data that I needed: Dieldrich Bermudez, Lindsey Carr Joseph Chipperfield, Sarah Gravem, Amanda Roe, Melissa Schmitt, Will van Stippen Michael Way, and, especially, Wesley Davis. The East West course, and the other courses that came through (UCLA and Berkeley in particular) also offered many helping hands, diving buddies, fun distractions (fish prints!), and fantastic role models (Rick Vance, Peggy Fong, Mark Steele, and Claire Wormald, in particular), for which I am extremely grateful. Other folks were part of the Gump family and offered wonderful support through the years in many different ways. I offer my sincere mauruuru roa to the station Director, Neil Davies, Managers, Hinano and Frank Murphy (and their nehenehe family: Teheire, Tangaroa, Terava, Kohina, and Temakahu), and the amazing Gump staff: Valentine Brotherson, Tony YouSing, Irma and Jacques YouSing, and all of the wonderful nehenehe Gump children, especially, Melissa Schmitt, for always keeping the Gump family spirit alive and smiling. My undergr aduate mentors, Russell Schmitt and Sally Holbrook, and the rest of the LTER professors, offered a supportive and academically stimulating environment throughout my graduate career. Of these, Andrew Brooks, Peter Edmunds, Ruth Gates, and Hunter Lenihan (as well as his beautiful family: Olivette, Rea, Loana, and the rest of
8 the extended family), were particularly instrumental in my development as a scientist (and as a lover Tahitian food and dance, respectively). The LTER research technicians Keith Seydel, M ichael Murray, and Vincent Moriarty, were diving savvy, knowledgeable about all things technical, and perhaps most importantly, they contributed to the fun! Leslie and the rest of the McIlroy family (Tiare, Matt Jean, Prinz, Douglas, Ioane, Moava, and Manukura) were my second family. They offered me shelter, food, love, and affection (and, of course, fine instruments, keen problem solving minds, and power tools). Without their open house and hearts, I could not have stretched my research funds and complete d the magnitude of work undertaken. I only hope I can offer the same support to them one day. My experiences with C.R.I.O.B.E. and the University of Perpignan rounded my scientific and cultural experiences in French Polynesia. In particular, Ren Galzin, S erge Planes, Thierry Lison de Loma, and Yannick Chancerelle challenged me to think (and speak) like the French! They also offered me amazing opportunities for personal and professional growth through expeditions to France, the Marquesas, and the South Tuam otus, for which I am extremely grateful, and I look forward to continuing our productive collaboration as we publish the data we collected together. The CRIOBE staff and graduate students were the most friendly and supportive folks imaginable and offered t he familial environment I missed, just as Gump was expanding at record speeds. While, my French exchange companion, Adrien Delval (or mon petite monkey) challenged me culturally in countless other ways and, through it all, was a wonderful friend.
9 Last, but certainly not least, my wonderful friends and colleagues at the University of Florida provided a supportive and nurturing home for me, both during and between my long trips abroad, for which I will always be grateful. My chosen family here spoiled me rott en the first six year (Lisa, Donovan, Marina, James, Skye, D, Deena, Trace, Trav, Thea, Brandon, Lisa K., Francois, Adrien, Mary, Andy, Shane, Ellie, and the rest of the crew) and I ached for their companionship in the final year, after they had moved to g reener pastures. Luckily, I found Fay Ray, and fun friends old and new inspired me in new directions and kept the fire of adventure and discovery alive. This offered a wonderful reprieve from the norm of pretending to be locked in the Ivory Tower (like Rap unzel!), writing my dissertation at an accelerated pace. All in all, the Department of Biology was the ideal place for my graduate study and, although the years flew by, they will never be forgotten. Methods to remove Stegastes nigricans were approved by t he University of Florida (U.F.) Institutional Animal Care and Use Committee (Approval #: E 109: June 2005 July 2008). I gratefully acknowledge research funding from the following sources: NSF Graduate Research Fellowship; U.F. Alumni Fellowship ; French A merican Cultural Exchange Ocean Bridges Program ; U.F. H oward H ughes Medical I nstitute G roup A dvantaged T raining O f R esearch ( G.A T O R .) Mentorship Graduate Fellowship Program ; International Society for Reef Studies / The Ocean Conservancy Graduate Fellows hip in Coral Ecosystem Research; U.C. Berkeleys Gump Station Pearl Internship; and grants from the American Museum of Natural Historys Lerner Gray Fund for Marine Research ; the American Society of Ichthyologists and Herpetologists
10 Raney Fund; and Project AWARE (all to JSW) in addition to base funds from NSF (OCE 0242312 to C.W. Osenberg, B.M. Bolker, C.M. St. Mary, and J.S. Shima ).
11 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES .......................................................................................................... 15 LIST OF FIGURES ........................................................................................................ 18 LIST OF OBJECTS ....................................................................................................... 20 LIST OF ABBREVIATIONS ........................................................................................... 21 CHAPTER 1 INTRODUCTION .................................................................................................... 24 Farmerfish as Ecosystem Engineers ...................................................................... 25 Study System .......................................................................................................... 28 Research Goals and Approach ............................................................................... 29 2 INDIRECT EFFECTS OF A KEY ECOSYSTEM ENGINEER A LTER SURVIVAL AND GROWTH OF FOUNDATION CORAL SPECIES ........................................... 31 Introduction ............................................................................................................. 31 Methods .................................................................................................................. 33 Study System ................................................................................................... 33 Experimental Reefs .......................................................................................... 36 Farmerfish and Algal Removal Manipulations .................................................. 37 Coral Transplants and Caging SubTreatments ............................................... 3 7 Fish Surveys and Reef Composition ................................................................ 38 Termi nation of experiment ................................................................................ 39 Statistical Analyses .......................................................................................... 39 Results .................................................................................................................... 41 Treat ment Efficacy ........................................................................................... 41 Fish removals ............................................................................................. 41 Turf removals ............................................................................................. 42 Field Surveys .................................................................................................... 42 Chase frequency ........................................................................................ 42 Density ....................................................................................................... 43 Microhabitat use ......................................................................................... 43 Foraging frequency .................................................................................... 43 Changes in reef composition ...................................................................... 44 Experimental Responses.................................................................................. 45 Frequency of predation .............................................................................. 45 Temporal change in coral mass ................................................................. 45 Frequency of algal overgrowth ................................................................... 46
12 Discussion .............................................................................................................. 47 Conclusions and Implications ................................................................................. 51 3 INDIRECT EFFECTS OF STEGASTES NIGRICANS ON MASSIVE PORITES CORAL AND IMPLICATIONS FOR OTHER COMMUNITY MEMBERS ................. 56 Introduction ............................................................................................................. 56 Methods .................................................................................................................. 59 Paired Field Surveys ........................................................................................ 59 Transition Zone Field Experiment ..................................................................... 59 Treatment implementation ......................................................................... 60 Digital photographic analyses .................................................................... 61 Statistical Analyses .......................................................................................... 62 Paired field surveys .................................................................................... 62 Transition zone experiment ........................................................................ 62 Results .................................................................................................................... 63 Observational Results: Paired Surveys ............................................................ 63 Experimental Results: Porites Turf Transition Zone ......................................... 63 Discussion .............................................................................................................. 64 Conclusions and Future Directions ......................................................................... 66 4 EFFECTS OF FARMERFISH AND CORAL PREDATORS ON THE POPULATION DYNAMICS OF BRANCHING CORALS IN A DISTURBED LAGOON SYSTEM ................................................................................................. 75 Introduction ............................................................................................................. 75 Evidence Farmerfish Engineer the Coral Community ....................................... 76 Diverse Life History Strategies ......................................................................... 77 Demographic Complexity ................................................................................. 78 Sca ling Up ........................................................................................................ 79 Methods .................................................................................................................. 81 Study System ................................................................................................... 81 Experimental Reefs .......................................................................................... 82 Demographic Monitoring .................................................................................. 83 Life Table Response Experiment ..................................................................... 84 Statisti cal Analyses .......................................................................................... 84 Reef attributes ............................................................................................ 85 Mortality ..................................................................................................... 85 Change in size o f corals ............................................................................. 87 Matrix Models ................................................................................................... 88 Recruitment and Fragmentation ....................................................................... 88 Re sults .................................................................................................................... 89 Life Table Response Experiment ..................................................................... 89 Acropora mortality ...................................................................................... 89 Pocillopora mortality ................................................................................... 90 Fragmentation ............................................................................................ 91 Change in coral size .................................................................................. 92
13 Experimental change in Acropora size ....................................................... 92 Experimental change in Pocillopora size ................................................... 93 Matrix Models ................................................................................................... 94 Recruitment Patterns ........................................................................................ 94 Discussion .............................................................................................................. 94 Observed Community Response ...................................................................... 95 Differential Susceptibility to Predation .............................................................. 96 Recruitment Failure .......................................................................................... 99 Conclusions .................................................................................................... 100 5 CONCLUSIONS ................................................................................................... 112 Variable Outcomes of Farming Behaviors ............................................................ 113 Farme rfish Effect Varies with Community Structure ............................................. 114 Expanding the Grazing Continuum ....................................................................... 115 Evidence Based Upon Research in Moorea ......................................................... 117 APPENDIX A SAMPLING METHODS AND RESULTS CONCERNING EXPERIMENTAL REEF CHARACTERISTICS PRIOR TO AND AFTER MANIPULATIONS ............ 124 Methods ................................................................................................................ 124 Results .................................................................................................................. 124 B TABLES AND FIGURES SUPPORTING THE EFFICACY OF REMOVAL TREATMENTS ..................................................................................................... 133 C EXPERIMENTAL METHODS, DETAILS, AND RESULTS FOR DAILY CHANGE IN CORAL MASS .................................................................................................. 140 Methods ................................................................................................................ 140 Results .................................................................................................................. 140 D RESULTS OF FULL MODELS TESTING THE FREQUENCY OF PREDATION AND OVERGROWTH ........................................................................................... 144 E RESULTS OF FULL MODELS TESTING THE TERRITORIA L EFFORT EXPENDED BY STEGASTES NIGRICANS AND RESPONSES BY IMPORTANT MOBILE FISHES (DENSITY, FORAGING, REEF USE) ................. 148 F DETAILED METHODOLOGY FOR LAB AND FIELD OBSERVATIONS OF COMPETITION BETWEEN PORITES AND ALGAL TURF .................................. 157 G EXPERIMENTAL REEFS BEFORE AND AFTER FARMERFISH REMOVALS ... 163 H POPULATION MEASURES FOR ACROPORA AND POCILLOPORA IN THE PRESENCE AND REMOVAL OF FARMERFISH ................................................. 173
14 I TRANSITION MATRICES FOR ACROPORA AND POCILLOPORA IN THE PRESENCE AND REMOVAL OF FARMERFISH ................................................. 184 LIST OF REFERENCES ............................................................................................. 191 BIOGRAPHICAL SKETCH .......................................................................................... 206
15 LIST OF TABLES Table page 3 1 Transitionzone Experimental Treatments .......................................................... 68 4 1 General susceptibility of Acropora, Pocillopora, and Porites to natural disturbances. .................................................................................................... 101 4 2 Overall mortality patterns for all sizes of Acropora. .......................................... 102 4 3 Overall mortality patterns for all sizes of Pocillopora. ....................................... 103 4 4 Record of fragmentation of corals by period and treatment .............................. 104 4 5 Record of coral recruitment to reefs by period and treatment ........................... 105 4 6 Pocillopora and Acropora. ................................................. 106 5 1 Summary of differences between foraging and farming coral reef fishes. ........ 122 A 1 Physical factors of all reefs before manipulations (GLM) .................................. 124 A 2 Composition of Stegastes Reefs before removals (GLM) ................................. 125 A 3 Comparisons of Stegastes and Porites reefs before removals (t tests) ............ 126 A 4 Composition of Stegastes Reefs after removals (GLM) .................................... 127 B 1 Density of adult Stegastes nigricans (GLMM)................................................... 133 B 2 Contrasts for density of adult Stegastes nigricans (GLMM) .............................. 134 B 3 Density of adult Stegastes nigricans before removals 2). .............................. 135 B 4 Density of adult Stegastes nigricans after removals 2). ................................. 136 B 5 Mean mass of dried algal turf (GLM) ................................................................ 137 C 1 Daily coral growth (GLM) .................................................................................. 140 C 2 Tests of growth effect slices for each coral species (GLM). ............................. 141 C 3 Daily growth of caged individuals (GLM) .......................................................... 142 D 1 Predation frequency (Logisitic Regression). ..................................................... 144 D 2 Frequency of algal overgrowth (Logistic Regression) ....................................... 145
16 D 3 Turf abundance and coral overgrowth (Linear Regression) .............................. 146 E 1 Chase frequency (GLMM) ................................................................................ 148 E 2 Contrasts of chase frequency (GLMM) ............................................................. 149 E 3 List of species documented for each target fish fami ly ..................................... 150 E 4 Densities of target families of coral reef fishes (GLMM) ................................... 151 E 5 Contrasts of densities of target families of coral r eef fishes (GLMM) ................ 152 E 6 Reef use by target families of coral reef fishes (GLMM) ................................... 153 E 7 Contrasts of reef use by target famili es of coral reef fishes (GLMM) ................ 154 E 8 Foraging frequency of target families of coral reef fishes (GLMM) ................... 155 E 9 Contrasts of fo raging frequency of target families of reef fishes (GLMM) ......... 156 F 1 Test of average change in live Porites (LMM) .................................................. 157 F 2 Contrasts for change in live Porites (LMM) ....................................................... 158 F 3 Test of average change in bleaching front (LMM) ............................................ 159 G 1 Experimental reef characters B EFORE manipulations. .................................... 163 G 2 Differences in the average size of experimental reefs (LM) .............................. 164 G 3 Reef size model estimates (LM). ...................................................................... 165 G 4 Average density of Stegastes nigricans before removals (LMM). ..................... 166 G 5 Stegastes nigricans density estimates before remov als (LMM) ........................ 167 G 6 Average density of Stegastes nigricans after removals (LMM) ......................... 168 G 7 Stegastes nigricans density estimates after removals (LMM) ........................... 169 H 1 Mortality frequency for Acropora & Pocillopora (GLMM, LRT). ......................... 173 H 2 Mortality frequency model estimates (GLMM). ................................................. 174 H 3 Follow up mortality analyses for Acropora (GLMM, LRT) ................................. 175 H 4 Follow up mortality analyses for Pocillopora (GLMM, LRT) .............................. 176 H 5 Differences in final size of Acropora and Pocillopora (LMM). ........................... 177
17 H 6 Follow up change in size analyses for Acrop ora (LMM) ................................... 178 H 7 Follow up change in size analyses for Pocillopora (LMM) ................................ 179
18 LIST OF FIGURES Figure page 2 1 Mean frequency of coral predation ..................................................................... 53 2 2 Frequency of coral overgrowth ........................................................................... 54 2 3 Mean density, reef use, and foraging frequency of target fish families. .............. 55 3 1 Representative photos of reefs by type .............................................................. 69 3 2 Simplified schematic of tr ansition zone quadrat design ...................................... 70 3 3 Paired reef surveys: Mean number and size of Stegastes nigricans ................. 71 3 4 Paired reef surveys : Mean percent cover live massive Porites and algal turf ..... 72 3 5 Mean linear change in live massive coral tissue by treatment ............................ 73 3 6 Mean linear change in the amount of dead Porites tissue .................................. 74 4 1 Mortality rates of Acropora ............................................................................... 107 4 2 Mortality rates of Pocillopora ............................................................................ 108 4 3 Change in size (Delta) of Acropora as a function of previous maximum linear length ................................................................................................................ 109 4 4 Change in size (Delta) of Pocillopora as a function of previous maximum linear length ...................................................................................................... 110 4 5 Mean population decay rate of Acropora and Pocillopora ) ........................... 111 5 1 Schematic of Stegastes nigricans direct and indirect interactions .................... 123 A 1 Design schematic ............................................................................................. 128 A 2 Photographic exampl e of experimental reefs ................................................... 129 A 3 Photographic example of experimental coral nubbins ...................................... 130 A 4 Mean reef composition before removals ........................................................... 131 A 5 Mean reef composition after removals .............................................................. 132 B 1 Mean number of adult Stegastes nigricans ...................................................... 138 B 2 Mean mass of dried algal turf ........................................................................... 139
19 C 1 Mean daily coral growth rates ........................................................................... 143 D 1 Frequency of algal overgrowth for Pocillopora verrucosa ................................. 147 F 1 Paired reef surveys: Reef Size ......................................................................... 160 F 2 Example photo quadrats from a sample experimental reef .............................. 161 F 3 Schematic of line measurements for transition zone experiment. ..................... 162 G 1 Mean size of experimental reefs ....................................................................... 170 G 2 Mean density of Stegastes nigricans before removals ..................................... 171 G 3 Mean density of Stegastes nigricans at the termination of th e experiment ....... 172 H 1 Raw mortality values for Acropora .................................................................... 180 H 2 Raw mortality values for Pocillopora ................................................................. 181 H 3 Representative photos of Acropora predation before and after removals of Stegastes nigricans ......................................................................................... 182 H 4 Representative photos of Pocillopora predation before and after removals of Stegastes nigricans ......................................................................................... 183 I 1 Schematic of size classes and types of transitions for Acropora and Pocillopora. ....................................................................................................... 184 I 2 Size structure transition matrix for Acropora on all reefs prior to experimental manipulations ................................................................................................... 185 I 3 Size structure transition matrix for Acropora following application o f clove oil to experimental CONTROL reefs. ........................................................................ 186 I 4 Size structure transition matrix for Acropora following application of clove oil and removal of farmerfish. ................................................................................ 187 I 5 Size structure transition matrix for Pocillopora on all reefs prior to experimental manipulations. ............................................................................. 188 I 6 Size structure transition matrix for Pocillopo ra following application of clove oil to experimental CONTROL reefs. .................................................................... 189 I 7 Size structure transition matrix for Pocillopora following application of clove oil and removal of farmerfish. ........................................................................... 190
20 LIST OF OBJECTS Object page 3 1 Video of vermetid feeding (28.9 MB .avi) ............................................................ 64
21 LIST OF ABBREVIATION S GBR Great Barrier Reef GLMM Generalized Linear Mixed Model LM Linear Model LLM Linear Mixed Model LRT Likelihood Ratio Test PNG Papua New Guinea 2 Chi Square Test Statistic
22 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy INDIRECT EFFECTS OF A MARINE ECOSYSTEM ENGINEER ALTER T HE ABUNDANCE AND DISTRIBUTION OF FOUNDATION CORAL SPECIES By JadaSimone Shanti White May 2010 Chair: Benjamin M. Bolker Cochair: Gustav Paulay Major: Zoology Farmerfish engineer coral communities by facilitating algal turf and exerting resource control through territorial defense: Within territories, these behaviors indirectly (1) increase interactions between coral and farmed turf and (2) decrease interactions with mobile grazers. A small scale experiment indicated massive Porites were more vulnerable to competition with turf than were branching Acropora encrusting Montipora or lobed Pocillopora In contrast, the more delicate acroporids ( Acropora and Montipora), were more vulnerable to predation by mobile corallivores and grew and survived better in the presence of S. nigricans defense. I used a more realistic manipulation of turf, sediment, and sediment consuming vermetid snails to document the rate of massive Porites overgrowth by S. nigricans associated turf. I addressed the indirect effects of te rritoriality in a demographic context using size specific population monitoring of two of the four genera, Acropora and Pocillopora, in the presence and removal of S. nigricans It appears the disturbance history has played a pivotal role in the types of c ommunity changes observed: While S. nigricans usually colonizes Acropora thickets, a series of disturbances virtually eliminated these habitats and
23 farmerfish are found colonizing the dominant disturbance tolerant, but turf sensitive, massive Porites Taxa with a higher resistance to competition with turf can utilize dead portions of these massive corals. This increase in substrate availability, when coupled with lower mortality rates, has led to enhanced recovery of branching corals within farmerfish terri tories. Corymbose Acropora are relegated to high flow, or cracks and crevices, outside S. nigricans territories, suggesting grazing pressure is constraining recovery in the absence of this ecosystem engineer. Pocillopora also enjoys lower mortality inside S. nigricans territories, though susceptibility to losses by corallivorous fishes was lower than those attributed to a periodic outbreak of the crowno thorn seastar, Acanthaster planci This system supports the idea that disturbance can alter the engineer ing role: S. nigricans adversely affects branching Acropora in relatively undisturbed habitats, but can offer a less s tressful environment if predation pressure by corallivores is high. These results suggest that corallivory may sometimes function as an im portant regulatory process of coral community structure and warrants additional research. In particular, whether the observed high predation rates are driven by changes in topdown (e.g., trophic cascades induced by anthropogenic removal of top predators), or bottom up processes (e.g., changes in grazing pressure due to alterations in coral and algal community structure following natural disturbances).
24 CHAPTER 1 INTRODUCTION Our understanding of what factors are driving the community structure of complex systems, such as coral reefs, is often compromised by an inability to test factors constraining the distribution of multiple taxa at relevant scales. Frequent disturbance regimes (e.g., cyclones, bleaching events) prevent competitive exclusion and contribute to the maintenance of high diversity on coral reefs (Connell 1978). These complex communities are often considered to be nonequilibrium systems (Connell 1978) with species abundance constrained by recruitment limitation (i.e., whether larvae can arrive, Doherty 1981) and a competitive lottery (i.e., the first to arrive at a new habitat patch is expected to have an advantage, Sale 1977, 1982). In these stochastic environments, long lived individuals have a greater potential to store reproductive potent ial via recruitment storage effects (Warner and Chesson 1985). Nonetheless, top down processes can alter community composition, and herbivores, in particular, play a role in mediating competition between fast growing algae and slow growing coral (Huston 1985; Sluka and Miller 2001). The foraging rate of herbivores creates a gradient of competition between coral and algae from intense (e.g., in the absence of herbivory following Diadema die offs in overfished parts of the Caribbean and resulting coral declin es, e.g., Lessios et al. 1984) to absent (e.g., when grazing pressure is high enough that there is little contact between filamentous or macroalgae and corals). As predicted by theory (Connell 1978), the highest algal diversity has been reported at interme diate foraging rates, because lower standing algal biomass reduces competition for limited resources (Sammarco 1983; Hixon and Brostoff 1983,1996; Huston 1985).
25 The likelihood for a community to be tightly linked, in which a change in the abundance of one species affects others in a predictable way, is higher when a keystone species (Paine 1969; Mills et al. 1993; Power et al. 1996) or ecosystem engineer (Jones et al. 1994, 1997) is operating. Those species that create a predictable change in the environment due to nontrophic modifications (e.g., dam building by the beaver, Castor canadensis, Wright et al. 2002) have been termed keystone modifiers (Mills et al. 1993) or ecosystem engineers (Jones et al. 1994,1997). Indeed, indirect effects are gaining acceptance as an important driver of key demographic processes (e.g., recruitment, Webster and Almany 2002), as well as larger scale community change (e.g. Dulvy et al. 2004). In addition to biotic disturbances, such as competition and predation, natural physi cal disturbances also remove or kill organisms and alter resource availability (e.g., Sousa 1979a, 1979b). The high diversity and stochastic disturbance regimes characteristic of coral reefs (Connell 1978; Hughes 1989) offer an opportunity to explore how r esource control by key engineers (Boogert et al. 2006) changes when community composition is altered by larger scale natural disturbances. Farmerfish as Ecosystem Engineers Territorial damselfish occupy defined areas that they actively defend from competit ors and egg predators (reviewed in Ceccarelli et al. 2001). Within this diverse group, only territorial algal farmerfish influence the benthic structure of coral reefs. These farmerfish develop and maintain conspicuous filamentous algal mats (hereafter, turf) through a variety of behaviors, including: substrate preparation via killing coral (e.g., Kaufmann 1977; Wellington 1982); weeding by selectively removing unpalatable algae (e.g., Hata and Kato 2003, 2004); fertilization through defecation (e.g., K lumpp and Polunin 1989; but see Lison de Loma et al. 2000, Hata and Kato 2002); and
26 reducing herbivory through active defense (e.g., Hixon and Brostoff 1981). Species of damselfish that farm turf are more aggressive than nonfarming congeners and effectiv ely alter the spatial distribution of other (nonfarming) pomacentrids through territorial defense (Robertson 1996). Farmerfish aggressively defend against urchins (e.g., Sammarco and Williams 1982), large herbivorous and corallivorous fishes (e.g., Hixon and Brostoff 1981), and potential egg predators (e.g, Haley and Mller 2002), thereby reducing resource availability for those species. The direct effects of farmerfish farming on benthic algal community structure have been well documented, with respect to increases in algal biomass (Hata et al. 2002; Hata and Nishihira 2002), diversity (Hixon and Brostoff 1983, 1996), and productivity (Klumpp et al. 1987). Given high algal abundance within territories, it is not surprising that previous research has docu mented farmed turf can overgrow hermatypic corals and coral spat (Vine 1974; Potts 1977; Lobel 1980; Sammarco and Carleton 1981). Territorial behavior by farmerfish toward other species serves to directly reduce access to defended reefs (i.e., resource co ntrol) and can indirectly benefit species that use territories without eliciting this response. These territory residents may be indirectly afforded protection due to the reduced frequency and intensity of visits from grazers (e.g., Hixon and Brostoff 1986), competitors (e.g., Williams 1980, 1981) or predators, including corallivores (e.g., Wellington 1982; Glynn and Colgan 1988). Several authors have documented increased recruitment (Sammarco and Carleton 1981; Gleason 1996) and survivorship of corals wi thin farmerfish territories (Sammarco and Williams 1982, Wellington 1982, Glynn and Colgan 1988, Gleason 1994), resulting
27 in higher coral abundance (Wellington 1982; Glynn and Colgan 1988; Done et al 1991; Gleason 1994, 1996; Suefuji and van Woesik 2001). The underlying mechanism appears to be an indirect positive effect of territorial behavior creating a spatial refuge from mortality due to foraging by grazing herbivorous and corallivorous fishes (Gleason 1996, Wellington 1982) and invertebrates (urchins: Sammarco and Williams 1981, Done et al 1991, Gleason 1996; crownof thorns starfish: Glynn and Colgan 1988). Thus, within algal systems, farmerfish have been shown to maintain higher diversity by maintaining intermediate levels of (herbivorous) disturbance (Hixon and Brostoff 1983, 1996). With respect to corals, the evidence is more equivocal as farmerfish interact indirectly with corals along two opposing axes of already existing interaction chains. By reducing access of herbivores to territories, farmer fish facilitate algal abundance and increase competitive interactions between algae and corals. Simultaneously, the reduction of grazing pressure by herbivores and corallivores may indirectly benefit corals within territories. Therefore, farmerfish may indirectly affect coral in two ways: (1) increased competition with farmed algal turf; and (2) decreased predation by corallivores or incidental predation by grazing herbivores. Given the indirect nature of these interactions, the importance of one mechanism over the other could vary with community structure (i.e., the relative abundance of macroalgae, corals, or grazers). Previous research on the influence of farmerfish on corals has relied upon the spatially confounded approach of comparing territories with nonterritories. This approach, while strongly indicative of a trend, does not eliminate potential effects of natural variation in corallivore abundance or other physical factors, such as flow
28 (Osenberg et al. 2006). In contrast, I experimentally manipulat ed farmerfish and/ or turf abundance to test whether differences are due to indirect effects of competition with algae or protection via territorial defense. Study System In Moorea, French Polynesia, farmerfish territories account for approximately 5 45% of the back reef hard bottom environment (Done et al. 1991, Gleason 1994). The pattern of higher abundance of branching Acropora and Pocillopora coral abundance within dusky farmerfish, Stegastes nigricans, territories was first documented by Glynn and Colgan (1988) who hypothesized that farmerfish protect resident corals from the corallivorous echinoderm, Acanthaster planci Done et al. (1991) reported a higher recovery of some branching and rare corals inside farmerfish territories after a series of dist urbances. Gleason (1996) documented higher recruitment rates of pocilloporid corals inside territories in comparison to nearby grazed, as well as macroalgae ( Turbinaria ornata) dominated reefs. The proposed mechanism is an indirect effect of territorial behavior that results in decreased predation within territories. Further, at least some species of coral appear to be negatively impacted in farmerfish territories (Done et al. 1991). Massive Porites species dominate Moorea backreef (lagoon) coral assemblage and, Porites coral heads within territories have significantly lower coral cover than those outside territories (Glynn and Colgan 1988). This suggests that direct (e.g., killing coral) or indirect effects of farming turf (e.g., coral algal competition) may reduce the abundance of at least Porites within territories. Thus, the relative effects of farmerfish on coral assemblages remain poorly understood (reviewed in Ceccarelli et al 2001), and the net effect could be positive or negative, depending on coral species.
29 Research Goals and Approach For my dissertation research I sought to understand what processes were driving the observed coral community structure in farmerfish territories and to identify vulnerable demographic stages. I utilized a combination o f observational studies, small scale experiments, and a life table response experiment to evaluate the effects of this key ecosystem engineer on the dominant coral genera within Moorean lagoons. I conducted the first experimental test of the direct and indirect effects of a farmerfish on the growth and mortality of four coral taxa: Acropora striata, Montipora flowerii, Pocillopora verrucosa, and Porites australiensis (Chapter 2). In summer 2004 and spring 2005, I conducted paired reef surveys within Vaipahu lagoon to establish the current patterns of habitat and community within S. nigricans territories (Chapter 3). During the spring 2007, I evaluated how turf, sediment, and vermetid snails affect the rate of massive Porites overgrowth by algal turf (Chapter 3) I experimentally evaluated coral recruitment on reefs with S. nigricans naturally present, experimentally removed, or naturally absent (January 2006). I also began monitoring populations of branching Acropora and Pocillopora, as well as massive Porite s on reefs with S. nigricans to measure demographic parameters for these species in the presence of this farmerfish (July 2006) Simultaneously, my undergraduate assistant, Shelby E. Boyer, conducted an experiment on the effects of the fish anesthetic clove oil on growth and survival of representatives for the three target coral genera (Boyer et al. 2009). This work was instrumental in testing possible confounding effects and improved the methodology for subsequent removals of S. nigricans in July 2007, for a life table response experiment to test the effects of territoriality on Acropora and Pocillopora (Chapter 4). Finally, I
30 explore the ramifications of this, and related, work by placing the observed interactions within a larger theoretical context (Chapter 5).
31 CHAPTER 2 INDIRECT EFFECTS OF A KEY ECOSYSTEM ENGINEER ALTER SURVIVAL AND GROWTH OF FOUNDATION CORAL SPECIES Introduction Understanding how organisms modify communities via indirect interactions is particularly essential for strongly interactive species, i.e., species whose absence (or removal) leads to significant changes in their communities (Soul et al. 2003). E ffects of such keystone species on communities, often mediated through trophic interactions, are disproportionate to their abundance ( Paine 1966). Ecosystem engineers create, maintain, or modify the physical environment through behavioral control of resources (e.g., beavers) and also include foundation species (e.g., trees) that provide substrate or habitat for other species through thei r form (Jones et al. 1994, 1997). Like keystone species, key engineers exert resource control that is disproportionate to their abundance, but nontrophic in nature, thereby stimulating drastic changes in community structure (Boogert et al. 2006). While ec ologists have long used manipulative experiments to identify and understand the effects of keystone species, many studies of ecosystem engineers and foundation species have been observational (but see Bertness and Leonard 1997; Crain and Bertness 2005; Spo oner and Vaugn 2006, for examples of empirical studies). Farmerfish are territorial damselfish (Pomacentridae) that can alter vast expanses of the benthos via a combination of farming behaviors and territorial defense ( r eviewed in Ceccarelli et al. 2001): Farming behaviors include substrate preparation (including killing coral), weeding unpalatable algae, and fertilization through defecation. Territorial defense also reduces foraging rates by limiting access of fishes and
32 invertebrates (Ceccarelli et al. 2 001). Community change can be widespread and profound, and include altering the depth zonation of corals (Panama: Wellington 1982) and maximizing algal diversity (Hawaii: Hixon and Brostoff 1983, 1996). The effects of farmerfish on algal communities are well documented: algal abundance, productivity, and diversity are higher within territories (Ceccarelli et al. 2001). Algae are usually detrimental to coral health (Reviewed in McCook et al. 2001), and early reports indicated that effects of farmerfish on coral were negative due to enhanced algal overgrowth (Vine 1974, Potts 1977, Lobel 1980) and bioerosion (Risk and Sammarco 1982). Some farmerfish bite polyps to expand their territories (Kaufman 1977). In addition, filamentous turf can trap sediment (Nugues and Robert 2003) or indirectly promote harmful bacterial growth (Smith et al. 2006). However, other studies have found higher coral abundance, recruitment or survival within territories, relative to outside (Sammarco and Carleton 1981 Sammarco and William s 1982, Glynn and Colgan 1988, Done et al. 1991, Gleason 1994, Gleason 1996, Suefuji and van Woesik 2001). Thus, the impact s of farmerfish on algal turf are well documented and positive, but effects on coral are understudied and contradictory (Ceccarelli et al. 2001, see also Jones et al. 2006). Using patch reefs in Moorea, I transplanted corals and manipulated the presence of turf and S. nigricans to investigate the direct effects of this farmerfish on turf abundance and reef use by common herbivorous and corallivorous fishes, as well as the indirect effects on growth and survival of four species of coral ( Acropora striata, Montipora floweri, Pocillopora verrucosa, and Porites cf. australiensis ). Specifically, I tested the following hypotheses:
33 H1) Stegastes nigricans directly decrease coral survival and growth through predation. H2) S tegastes nigricans indirectly decrease coral survival and growth through increased competition with algae. H2A) Territorial defense by S tegastes nigricans decreases herbivore access. H2B) Decreased herbivore access increases algal turf abundance. H2C) Abundant turf decreases coral survival via overgrowth. H2D) Overgrowth by turf decreases coral growth. H3) S tegastes nigricans indirectly increase coral survival and growth, th rough decreased predation by other fish H3A) Territorial defense by S tegastes nigricans decreases corallivore access. H3B) Decreased corallivore access decreases coral predation. H3C) Lower predation allows higher coral growth. H4) Susceptibility to these direct and indirect effects will vary among coral species, due to variation in life history traits. Methods Study System My study was conducted June July 2005 within the northeastern lagoon of Moorea, French Polynesia (S 1728525645, W 149473355 35). Moorea is surrounded by shallow (usually 16 m deep) lagoons stretching 0.2 1.5 km offshore to a barrier reef (Galzin and Pointier 1985). Within the back reef, the lagoon floor is a mosaic of patch reefs and open substrate (e.g., sand, rubble, or pavement) with patch reefs increasing in abundance and coalescing toward the reef crest. Due to a series of disturbances in the 1980s (Done et al. 1991; Adjeroud et al. 2005), the underlying structure of these patch reefs is primarily formed by disturbance t olerant Porites corals
34 (primarily P. australiensis, P. lobata, and P. rus ) (Berumen and Pratchett 2006). Prior to these disturbances the lagoon was dominated by massive Porites, followed closely by branching Acropora, with low cover of Montipora and Pocillopora (Bouchon 1985). Berumen and Pratchett (2006) report that coral cover in the back reef of Moorea has returned to previous levels, however, there has been a notable change in community structure: Porites still dominate the reef flat, however, Pocillopo ra have increased significantly in cover, relative to predisturbance levels, while Acropora and Montipora have failed to recover except within Stegastes territories (Done et al. 1991, Mapstone et al. 2007) and in areas with high flow (e.g., near the reef crest, White et al., unpubl. data). Branching corals ( Acropora, Montipora and Pocillopora) are more susceptible to periodic, largescale, disturbances (e.g., storm damage, bleaching, crownof thorns starfish outbreaks) relative to massive Porites (Bouchon 1985, Hughes et al. 1989, Gleason 1993, Adjeroud et al. 2005, Berumen and Pratchett 2006, McClanahan et al. 2007). Branching corals are also more affected by small scale perturbations, such as corallivory by other echinoderms (e.g., sea urchins: Suefuji a nd van Woesik 2001; pincushion stars: Rotjan and Lewis 2008) and fishes (e.g., Wellington et al. 1982, this study), but they tend to grow rapidly relative to massive corals (Huston 1985) and are generally considered better space competitors (Connell 1978). Given the high frequency of disturbance, Porites continue to dominate the lagoons, despite their relatively slow growth rates (Berumen and Pratchett 2006) and lower live coral within abundant Dusky farmerfish ( Stegastes nigricans ) territories (Done et al. 1991, Chapter 3). When portions
35 of these massive corals die they provide substrate for other sessile organisms, such as crustose coralline algae, algal turf, macroalgae, and other corals. All four genera of coral tested are known to incur predation by co ral reef fishes (Reviewed by Rotjan and Lewis 2008). However, corallivorous fishes show distinct preferences for Acropora and Pocillopora in the IndoPacific, while Montastrea and Porites are the major prey species for corallivores in the Caribbean, due in part to differences in abundances between the two regions (Reviewed in Cole et al. 2008). In the eastern Pacific, benthic grazers with specialized dentition can ingest large portions of coral skeleton (e.g pufferfish (Tetradontidae), large triggerfishes ( Balistidae), and parrotfish (Scaridae)), often preferring growing tips (Cole et al. 2008). Indeed, pufferfish and parrotfishes reduced growth and survival of juvenile branching corals ( Pocillopora) through heavy grazing in Panama (Wellington 1982). In Moor ea, pufferfish are in lower abundance but the guineafowl puffer ( Arothron meleagris ) is observed to bite off large chunks of branching corals (especially Acropora spp.): In contrast, relatively infrequent observations of foraging on abundant massive corals (e.g., Porites australiensis ) by A. meleagris indicated damage is restricted to only the concave surfaces (White, pers. obs.), and reinforces that differences in predator preference, skeleton morphology, and skeletal density can alter susceptibility to pr edation. In Moorea, highly aggressive Stegastes nigricans communally defend large territories ranging from small isolated patch reefs with ~3 adults per m2 (this study) to very large continuous territories several meters wide (Done et al 1991). Communal defense yields lush 1.0 3.0 cm long algal turf of fine rhodophytes composed mostly of Polysiphonia spp. (Hata and Kato 2006). In this system, S. nigricans most commonly
36 colonizes dead portions of the dominant massive coral, Porites (Done et al. 1991), but also inhabits all of the few remaining thickets of staghorn Acropora with a larger congeneric farmerfish, S. lividus, at densities up to 30 damselfish (on average) per 5 m (Mapstone et al. 2007). On average, the abundance of farmerfish, and the extent of territories, is higher in the fringing reef and lower in the back reef ( Galzin et al. In preparation). Experimental Reefs I selected 32 Porites patch reefs colonized by S. nigricans within the back reef of Aroa Lagoon. Using a 2 x 2 factorial field desi gn, I removed S. nigricans and / or turf algae prior to introducing coral transplants. As a positive control, I also transplanted coral fragments to 8 (uncolonized) Porites reefs (naturally devoid of S. nigricans and turf) that were interspersed among experimental reefs. This yielded a total of five treatments (n=8 for all): (1) Stegastes present, Turf present, (2) Stegastes present, Turf (locally) removed (3) Stegastes removed, Turf present (4) Stegastes removed, Turf (locally) removed (5) Stegastes absent, Turf absent (control or Porites reefs) ( Figure 2 1 A; Photos in Appendix A: Figure A 2 ). Reefs were chosen based on similarity in size (mean 95% CI: 4.2 0.3 m2), depth (1.9 0.1 m), average (0.7 0.1 m) and maximum (1.1 0.2 m) reef height and relative isolation from other patch reefs (2.2 0.7 m; Appendix A: Table A 1). Prior to manipulations, the 32 Stegastes reefs did not differ significantly in gross community composition and were comprised of a mix of farmed algal turf (overlaying dead coral; 58.5 3.2%), reduced live massive Porites (26.2 5.0%), and other corals (Acropora, Montipora, Pocillopora; 8.96 0 .93%; Appendix A: Figure A 4 Table A 2). The 8 control reefs did not differ significantly from manipulated reefs in cover of macroalgae or
37 combined dead coral and crustose coralline algae; however, live Porites was significantly more abundant (89.0 5.9%), and turf (1.35 2.64%) and branching corals (0.50 0.64%) were less common (Appendix A: Figure A 4 Table A 3). Farmerfis h and Algal Removal Manipulations To test the hypotheses that S. nigricans reduces reef access for important herbivores ( H2A) and corallivores ( H3A), I removed S. nigricans prior to the experiment using microspears and the fish anesthetic clove oil (Eugenol, City Chemicals, USA). Because clove oil can increase coral bleaching and reduce growth (Boyer et al. 2009), S. nigricans removal treatments were not maintained, although densities remained low, likely due to isolation of the reefs. To test the hypothes is that turf decreases coral survival ( H2C) and growth ( H2D), I removed turf with steel brushes until bare (dead) coral substrate was exposed. I removed turf from twelve, 15cm x 15cm, similar patches distributed around the reef (random sites for coral transplants), rather than removing all turf from a given reef due to the large size of patch reefs and S. nigricans interdependence with turf. Throughout the experiment, I ensured there was no direct contact between coral transplants and turf in removal treat ments by scrubbing nubbin bases with stiff plastic brushes every five days; however, the rest of the reef maintained typical turf abundances. Coral Transplants and Caging SubTreatments To tease apart predation by S. nigricans ( H1), versus transient corall ivores ( H3B), I transplanted three coral transplants (hereafter, nubbins) of each species, with random assignment to varying levels of protection (cage, control, and open) to each reef ( Figure 2 1 B). For each species tested ( A. striata, M. floweri, P. verr ucosa, and P. australiensis ) I collected 40, approximately thumbsized, nubbins from three separate genets (large
38 colonies) to include, but not explicitly test, genetic variation (yielding 120 nubbins). Nubbins were transported to the lab, mounted on Vexar (plastic mesh) using Z spar compound (Splash zone A788, Kopcoat, Pittsburgh, Pennsylvania, USA), photographed, and weighed using the buoyant mass technique (Davies 1989). For each species, I randomly assigned one nubbin (from a randomly selected genet) t o a subtreatment (cage, control, or open), yielding 12 nubbins transplanted to each of the 40 experimental reefs (480 nubbins total). For the turf removal treatments, I outplanted nubbins to the middle of the scrubbed patch to avoid contact with turf. I constructed cages and cage controls from Vexar with a mesh size of 3cm to reduce effects on flow and light while still offering protection from most adult fishes, including S. nigricans Cages consisted of four sides and a top connected with plastic cableties, while cage controls lacked two (opposite) sides to allow access while controlling for the effects of reduced light or flow (Photos in Appendix A: Figure A 3 ). Cages and cage controls were cleaned every 5 days throughout the experiment. Fish Surveys and Reef Composition To document whether reef use by other fishes varies with S. nigricans presence, removal, and absence ( H2A and H3A), one diver (JSW) conducted 5 min counts of fishes occurring within 1 m of each experimental reef beginning 10 d before experimental implementation and continuing every 5 days until the termination of the experiment 32 days later. Species identity, microhabitat used (e.g., sand, water column, turf, coral, dead coral, etc.), and (binary) foraging status of each fish was recor ded. Simultaneously, the second diver (JOD) counted chases from S. nigricans and noted the species chased. To test whether herbivore access alters turf abundance at the reef
39 scale ( H2B), I re assessed reef composition at the end of the experiment (see Appendix A: Methods). Termination of experiment I collected and rephotographed 479 nubbins (1/480 was lost) 32 days after deployment. To test whether corallivores increase coral mortality ( H3B), photos were scored blindly (without knowledge of treatment orig in) using binary designations (present=1, absent=0) for evidence of destructive predation (i.e., CaCO3 skeleton removed). Simultaneously, to test whether turf increases coral mortality ( H2C), photos were scored (1) if nubbins were partially or fully overgr own by turf. After photographing, I removed sediment from the base of the nubbin using a seawater squirt bottle and scraped all turf into a separate beaker. To test whether herbivore access alters turf abundance at the nubbin scale ( H2B), I filtered the tu rf onto Millipore glass fiber filters, dried for 24 hours at 70 C, and weighed with an analytical balance. After removing turf, I scrubbed off all remaining encrusting organisms and buoyantly reweighed each nubbin to test the effects of turf ( H2D) and co rallivory ( H3C) on coral growth (see Appendix C). Statistical Analyses I used a general linear model to test whether reef characters (e.g., size, depth, isolation, benthic composition) varied among Treatment groupings prior to manipulations (SAS PROC GLM, Appendix A: Figure A 1 ). I tested the efficacy of S. nigricans removals by comparing adult abundances among treatments throughout the duration of the experiment using a generalized linear mixed model (GLMM) with a Poisson error distribution and log link and degrees of freedom estimated using the KenwardRoger approximation (Reviewed in Bolker et al. 2009): Treatment and Period (Before versus After) were used as fixed effects and Day as a random effect to account for the lack of
40 independence among observations through time (SAS PROC GLIMMIX, Appendix B: Table s B 1 and B 2). For mean response variables calculated once, i.e., algal mass ( H2B, nubbinscale) and change in daily mass of nubbins ( H2D), I used a linear mixed model to test the fixed factors of coral Species, Treatment, Subtreatment (cage, control, and open) and all twoand threeway interactions, with Reef as a random effect (SAS PROC MIXED, Appendix B: Tables B 3 and Appendix C: Table C 1 and C 2, respectively). I tested for Treatment effects on gr owth for caged individuals of each species ( H3C) using Reef as a random effect (SAS PROC MIXED, Appendix C: Table C 3). I used the same analysis, to determine whether there was a difference in the arcsine square root transformed reef composition (i.e., % t urf, etc.) prior to the start, and at the termination of, experimental manipulations ( H2B, reef scale) (Appendix A: Tables A 2, A 3, and A 4). I used logistic regression to compare the frequency of predation ( H3B) and overgrowth ( H2C) of corals in the pres ence and absence of S. nigricans and turf, and in relation to subtreatment (SAS PROC LOGISTIC, Appendix B: Tables B 3 and B 5). Because there were no significant differences between open and cage control subtreatments in any analyses, data are pooled for graphical presentation (only). As a follow up, I used linear regression to test for a linear relationship between (arc sine square root transformed) overgrowth frequency and the average algal abundance experienced by each treatment reef combination (Appen dix D: Table D3). I calculated differences in the frequency of territorial chases, reef use, and foraging for each family ( H2A, H3A) after experimental removals using GLMMs. I used Reef Type ( Stegastes nigricans present, removed, absent) as a fixed effect and Day as a random effect to account for lack of temporal independence (SAS PROC GLIMMIX, Binomial error
41 distribution, logit link function, Appendix E: Tables E 3 E 4, and E 5). For the density of fishes in a given family, I used the same model with a P oisson error distribution and log link (Appendix E: Tables E 2); however, because overdispersion was present (the average Pearson residuals were significantly > 1.0) I used a quasi binomial model (Bolker et al. 2009). When appropriate, I applied Tukeys HS D test or contrasts to make relevant comparisons. For all analyses, I evaluated whether I were using the appropriate distribution (e.g., binomial, normal, or Poisson) for each model using Q Q plots and Pearson residuals. For normal based tests, I also used Levenes test for equality of variances. All analyses were performed using SAS version 9.13 (SAS Institute Inc., Cary, NC, USA). I report mean 95% confidence intervals (1.96*S.E.) for all values parenthetically. Results Treatment Efficacy Fish r emovals Stegastes nigricans average daily abundance was nearly 7 times higher in presence, than in removal treatments, throughout the experiment. Both main effects (Treatment and Before After period) were significant (Appendix B: Table B 1) and recolonization rem ained at low levels in the removal treatments. Very few adult S. nigricans were sighted on control reefs either before (n=0) or after (n=1) experimental removals took place (Day 0) and absent reefs were omitted from analyses (Appendix B: Figure B 1). I use d Chi square contingency analyses to compare abundances among reefs colonized by S. nigricans, before and after removals: Prior to removals, S. nigricans were distributed approximately equally across treatments; as expected, there were significant differences among treatments following the pulse removals and ~88%
42 of all adult Stegastes were observed on Stegastes presence reefs (Appendix B: Table B 2). Turf r emovals At the end of the experiment, the manipulated turf environment (i.e., turf biomass around t he base of nubbins) differed significantly among treatments (ANOVA, F4,418=128.45, P<0.0001) but not among coral species (P = 0.50), nor was there a treatment x coral species interaction (P = 0.81). On average, algal biomass on nubbins within S. nigricans territories was 4.5 times higher than on reefs from which I removed turf and 2.6 times higher than outside territories, i.e., control (Porites) reefs. The smaller difference in the latter case is driven by a cage effect: algal abundance was 1.8 to 2.2 time s higher when nubbins were caged rather than exposed (control and open subtreatments) on all reefs in which turf was not manually removed (Treatment x SubTreatment: ANOVA, F2,418=29.31, P<0.0001). Thus, our removals were effective, but differences in grazing protection (i.e., S. nigricans presence and cages) resulted in more than the three intended levels of turf (Appendix B: Figure B 2). Field Surveys Chase f requency On average, I observed more fishes chased from reefs (e.g., butterflyfishes (Chaetodont idae), parrotfishes (Scaridae), and surgeonfishes (Acanthuridae)) when S. nigricans was present (11% 0.02%) than when removed (1% 0.01%) or naturally absent (<0.01% 0.0008%) (GLMM, Wald F2,597=5.15, P<0.006; P vs. R: t597= 2.15, P=0.03; P vs. A: t597= 2.62, P=0.009, Appendix E: Table E 1). The few S. nigricans that re colonized reefs were responsible for the chases observed on removal reefs; juvenile
43 S. nigricans, which occupy an array of habitats and are less aggressive (Letourneur 2000), were resp onsible for the three chases observed on naturally absent reefs. Density In general, the presence of S. nigricans did not influence the density of other focal fishes surrounding patch reefs (plus 1 m perimeter) (see Appendix E: Table E 2 for species lis t). However, surgeonfishes showed a significant difference in density among reef types, with insubstantially higher densities on absent and removal reefs than reefs with S. nigricans present (GLMM, Wald F2,193=3.81, P=0.02; A vs P: t193=2.13, P=0.03; R vs P: t193=1.21, P=0.01; Figure 24 Appendix E: Table E 3). Microhabitat u se Territorial behavior by S. nigricans influenced reef use of surgeonfishes, parrotfishes, and corallivores: I observed greater use of the substrate or water column surrounding the r eef rather than the actual reef when S. nigricans was present, and the opposite when removed ( H2A ,H3A: t197 Table E 4; Figure 24 ). In contrast, reef use did not differ between naturally absent (control) reefs and reefs with S. nigricans present, except for surgeonfishes, which were observed using control reefs more fr equently than present reefs ( Figure 24 ; A vs. P: T197=3.78, P=0.0002). Foraging f requency The frequency of individuals observed foraging varied significantly with reef type for butterflyfishes, surgeonfishes, and parrotfishes: For each of these families, individuals were observed foraging more frequently on S. nigricans removal reefs versus reefs with S. nigricans present and, sometimes, naturally absent reefs ( Figure 2
44 4 ). Roughly twice as many butterflyfishes, were observed foraging on reefs where S. ni gricans was removed, than when present ( H3A: 28 9% versus 10 6%; GLMM, Wald F2,197=6.01, P=0.002; t197=3.61, P<0.004). Similarly, I observed surgeonfishes foraging twice as often when S. nigricans was removed, relative to present ( H2A: 49 10% versus 24 9%; GLMM, Wald F2,197=99.48, P<0.0001; P vs. R: t197=5.16, P<0.0001). Parrotfishes exhibited a similar, slightly stronger, pattern, with roughly four times the number of individuals observed foraging on reefs with farmerfish removed (74 10%) relativ e to present (17 7%) ( H2A: GLMM, Wald F2,197=20.15, P<0.0001; t197=12.17, P<0.0001). There was no difference in foraging between reefs with S. nigricans naturally present or absent for all families ( Figure 24 THSD=1.13, P=0.40, Appendix E: Table E 5) a nd the observed foraging patterns among treatments suggest an interplay between reduced reef access (when S. nigricans is present) and differences in resource quality and abundance and, thus, attractiveness (when S. nigricans is removed vs. absent). Change s in reef composition At the end of the experiment, turf cover dropped by two thirds on reefs where S. nigricans was removed (16.6 7%) relative to reefs where S. nigricans remained (50.2 8%; H2B: ANOVA, F3,28=10.25, P<0.0001), thereby significantly increasing the amount of bare substrate and crustose coralline algae (55 7%; H2B: ANOVA, F3,28=40.25, P<0.0001). The level of macroalgae was also significantly lower; however, the cover of live Porites or branching corals did not differ among these treatments (Appendix A: Figure A 2, Table A 4).
45 Experimental Responses Frequency of p redation Corallivory of exposed nubbins occurred for the two acroporids, but not for the other two corals tested (H3B): Exposed Acropora were ~100 times (11.2 to 1000) more like ly to incur predation on reefs where S. nigricans was removed or absent (Wald P=0.0009; Figure 22 B). Cages provided an effective barrier to corall ivory (save for a few large individuals on removal and control reefs that were not completely protected by cages ( Figure 22 ) and there were no significant differences between control versus open subtreatments (Appendix D: Table D 1): Exposed Montipora and Acropora nubbins were ~200 (26.3 to 1000) and ~300 times (35.7 to 1000), respectively, more likely to suffer skeletal damage than caged nubbins (Logistic reg., Acropora: Wald Pocillopora or Porites coral nubbins on the same reefs suffered no obvious damage. Temporal c hange in c oral m ass I attribute losses in coral mass to predation because I did not observe storm surges, or other physical damage, during our short (32 d) study. Significant coral by S. nigricans treatment (ANOVA, F12,418=1.76, P=0.05) and coral by caging subtreatment interactions (ANOVA, F6,418=4.78, P<0.0001) suggest that corals v aried in their susceptibility to decreased growth due to predation (H3C). Acropora and Montipora had significantly reduced mass after 32 days when exposed (uncaged) on reefs where Stegastes was removed or naturally absent (control reefs), but not when Steg astes was present ( Figure 22 ). There were no significant differences in mass among treatments or
46 sub treatments for Porites, despite overgrowth, or for Pocillopora (Appendix C: Figure C 1, Tables C 1 & C 2). When I assessed growth in the absence of predat ion (i.e., caged subtreatment), there were no significant differences across treatments for any coral species (H2D) (Appendix C: Table C 3). I observed significantly lower losses for Acropora and Montipora in cages relative to control (THSD=3.65, P=0.0008) and exposed (THSD=4.90, P<0.0001) subtreatments ( Figure 22 A & B) indicating cages functioned as intended. The lack of significant differences between the control and open subtreatments (THSD=1.26, P=0.42) suggests that possible negative physical effects (e.g., reduced light and flow) of the cages can be disregarded. I estimated the magnitude of skeletal loss due to predation as (Exposed Caged) / Caged. This is a measure of corallivore consumption rates (exposed and control nubbins) relative to the i ntrinsic growth rate of coral (caged nubbins). I compared loss rates in the removal or absence, relative to the presence, of S. nigricans using ratios For Acropora predation resulted in 4.3 or 5.0 times higher skeletal loss when S. nigricans was removed or naturally absent, respectively, relative to when S. nigricans was present ( Figure 22 A). Whereas, for Montipora, skeletal losses were 2.3 or 2.4 higher on reefs where S. nigricans was naturally absent or experimentally removed, respectively, versus reef s with S. nigricans present. The low growth rate for Montipora when exposed on Stegastes reefs ( Figure 22 B) contributes to this reduced ratio and suggests that S. nigricans defense may be less effective for this species. Alternatively, S. nigricans may ac tually contribute to the predation (see below). Frequency of a lgal o vergrowth The frequency of algal overgrowth varied significantly among treatments for three of the four corals tested ( H2C) : Porites nubbins were 3.4 times (1.4 to 8.4) more likely to
47 be overgrown in the presence of turf ( Figure 23 2 = 7.41, P = 0.007) and Acropora nubbins were 7.1 times (1.4 to 36) more likely to incur overgrowth 2=5.47, P=0.02). However, the overall frequency of over growth was generally twice as high for Porites than for Acropora ( Figure 23 A). Montipora nubbins were 11.3 times (1.3 to 101.3) more likely to be overgrown in the presence of S. nigricans ( Figure 23 2=4.65, P=0.03), regardless of turf manipulation (Appendix D: Table D 2). Interestingly, exposed (open and control) Montipora nubbins were 3.4 times (1.0 to 11.4) more likely to be overgrown than were caged nubbins ( Figure 23 2=3.90, P=0.05), suggesting that Montipora may be bitten by S. nigri cans thus facilitating overgrowth, although I did not observe predation directly. Overall, variation in algal abundance predicted a large amount of the variation in overgrowth for Porites (Linear Reg., Adj. R2=0.53, P=0.001) and Acropora ( Figure 23 B, Adj. R2=0.46, P=0.003), but not Montipora ( Figure 23 B, Adj. R2=0.14, P=0.10). Pocillopora did not incur substantial overgrowth in any treatment (Appendix D: Figure D 1, Tables D 2 and D 3). Discussion On Moorea patch reefs, territorial behavior by S. nigrica ns effectively reduces reef access, and thus foraging opportunities, for common, facultative corallivores (butterflyfishes) and herbivores (surgeonfishes and parrotfishes) (Figure 3). This reduced corallivore access indirectly increased the survival and gr owth of Acropora and Montipora (Figure 1), but no destructive predation or differences in growth were observed for either Pocillopora or Porites. Similarly, reduced herbivore access in the presence of S. nigricans indirectly increased turf abundance at bot h the reef (Appendix
48 A: Figure A 5 ) and nubbin scales (Appendix B: Figure B 2), with detrimental consequences for three of the four coral species tested (Figure 2). The high predation and associated skeletal loss in the absence and removal of S. nigricans for Acropora striata and Montipora floweri (Figure 1), suggests that this key engineer may be capable of indirectly facilitating growth and survival of these disturbance sensitive corals. Indeed, it is this territorial behavior that likely underlies the o bserved higher recovery of some corals in the presence of Stegastes spp. following the series of natural disturbances in the 1980s (Done et al. 1991, Gleason 1994, Mapstone et al. 2007, Chapter 4, Galzin et al. In preparation). I was surprised not to obser ve a similar response for Pocillopora (Appendix C: Figure C 1), as Gleason (1994) previously documented reduced predation on juvenile fragments within S. nigricans territories in Moorea. This inconsistency may be due to spatiotemporal variation in coralliv ore distribution, as there are fundamental differences in the morphology, and thus likely the susceptibility, of these corals to predation (Cole et al. 2008). Acroporids are often thin and finely branching, with a fragile, spongy skeleton architecture, thus, when accessible, can be cropped substantially by corallivores (Figure 1). In contrast, Pocillopora forms thick lobelike branches and has a dense skeleton that makes destructive predation of large colonies only accessible for grazers with strong dentiti on and musculature like that of Tetradontiforms, e.g. puffers (Wellington 1982) and filefishes (Jayewardene et al. 2009). While these taxa were observed in the study system, they are numerically rare and shy, and were never sampled despite our intensive su rveys. Additional research on destructive corallivores is needed to tease apart variation in susceptibility, prey preferences, and frequency dependent predation.
49 Species also varied in their susceptibility to overgrowth by turf: Acroporids and poritids i ncurred significantly more overgrowth in the presence of farmerfish and farmed turf, but the overgrowth effect was 2 to 3 times larger for Porites, than for Acropora or Montipora respectively (Figure 2A). Thus, although these acroporids incur some overgro wth, the relative risk of (at least partial) mortality is likely higher outside of territories given the rate and magnitude of predation (Figure 1). For Porites, there is no apparent benefit of S. nigricans. Because Porites spp. are currently the dominant reef builders within the lagoons of Moorea (Done et al. 1991, Berumen and Pratchett 2006), overgrowth by Stegastes induced turf may significantly increase the amount of settlement space available for the other, more turf tolerant taxa, like Pocillopora, wh ich did not incur significant overgrowth. In Moorea, ambient rates of overgrowth of Porites by turf can be high, depending on biotic factors and season (White et al., unpubl. data), thereby opening space for any colonists that can tolerate the turf. Indeed, recruitment of Pocillopora has been shown to be higher within S. nigricans territories, relative to heavily grazed or macroalgae dominated areas (Gleason 1996). In Moorea, the reduction in post settlement mortality, due to decreased predation in the presence of territorial exclusion by S. nigricans may be particularly important for life stages and taxa that are vulnerable to grazing pressures (e.g., coral recruits and delicate acroporids). Partial predation by farmerfish may weaken, and thus predispose Montipora to overgrowth. I observed higher overgrowth for exposed (vs. caged) individuals in the presence of S. nigricans, regardless of turf treatment (Figure 2A). Given the lack of a relationship with turf abundance (Figure 2B), our results hint that i nitial bites by S. nigricans may predispose individuals to overgrowth, though additional manipulations,
50 with more frequent assessments of coral health, are necessary to tease apart these two factors. Within the lagoons of Moorea, large colonies of Montipor a and Acropora are most commonly found within Stegastes territories, except in areas of very high water flow ( Galzin et al. In preparation), suggesting that (even Stegastes facilitated) algal competition is less of a threat than coral predation in this sy stem. Given that S. nigricans territories can cover 40% or more of the reef area on a given Moorean back reef (Gleason 1996, Galzin et al. In preparation ), these differential responses can lead to marked changes in coral distributions. Pocillopora have in creased significantly in cover relative to predisturbance levels (Berumen and Pratchett 2006), which is consistent with our observed low mortality in the presence of both farmed turf and corallivores. In contrast, Acropora which incurred high mortality v ia predation in the removal and natural absence of S. nigricans, has failed to recover (Berumen and Pratchett 2006). Previous authors have suggested this lack of recovery is due to the frequency of disturbance (Berumen and Pratchett 2006). While this undoubtedly plays an important role, it does not explain the higher abundance (Done et al. 1991, Mapstone et al. 2007) and size of Acropora and other rare corals within Stegastes t erritories relative to outside (Chapter 4, Galzin et al. In preparation). Stega stes nigricans usually inhabit Acropora thickets (e.g., Indian Ocean: Lison de Loma and Ballesteros 2002; Indo West Pacific: Jones et al. 2006; East Pacific: Mapstone et al. 2007). The natural disturbances in the early 1980s led to a severe reduction of th ese habitats in Moorea lagoons (Beruman and Pratchett 2006, Mapstone et al. 2007) and mostly relegated farmerfish to dead portions of the remaining dominant coral, Porites (Done et al. 1991, Chapter 3). Thus, the manner in which farmerfish act
51 as a key eng ineer appears to vary across environmental stress regimes: In relatively pristine areas, farmerfish are a nuisance inhabitant of branching Acropora thickets because they bite living tissue and induce algal overgrowth (e.g., Kaufman 1977). In contrast, in areas with high grazing rates they can inhabit disturbancetolerant massive corals and territorially defend branching corals, thereby indirectly cultivating more turf tolerant corals that are sensitive to grazing outside territories (Wellington 1982, Suefuj i and van Woesik 2001, this study). Nubbins provide a convenient way to control the distribution of coral in an experimental setting; however, the degree to which their growth (Elahi and Edmunds 2006) and survival really reflect those of juvenile corals r emains to be tested for most taxa. I use nubbins as a relative test and acknowledge that a demographic understanding is crucial for understanding the observed community change, with the ultimate goal of predicting which life stages are most crucial for res toration due to ontogenetic variation in vulnerability (Werner and Gilliam 1984). Additional long term experimental removals, using in situ size distributions of corals paired across territories, are underway to scale up our measurements to the demographic level. Conclusions and Implications I suggest that Stegastes nigricans may act as a key engineer by indirectly facilitating recovery of coral species that are otherwise sensitive to disturbances. While the dominant massive Porites are less susceptible to disturbances than branching corals (Berumen and Pratchett 2006), they are more vulnerable to competition with farmed algal turf. Colonization of massive Porites ( P. lobata, P. lutea, and P. australiensis ) by S. nigricans decreases live coral cover (Done et al. 1991, White et al., In preparation, a) and can enhance substrate availability for many organisms, including
52 corals (Gleason 1996). Territorial defense can decrease predation for delicate acroporids ( Suefuji and van Woesik 2001, this study). The increased abundance (Done et al. 1991, Shima et al. 2008) and size of branching corals within territories ( Chapter 4, Galzin et al. In preparation) may enhance habitat diversity, as these taxa can serve as foundation species for crustaceans (e.g., Stewart et al. 200 6 ), other damselfishes (Holbrook et al. 2000), and early life stages of other reef fishes (Shima et al. 2008). Bioerosion appears to alter the morphology of Porites patch reefs, creating complex lobes of living coral surrounded by algal turf (Done et al. 1991). There is some evidence that farmerfish require structural complexity (Wellington 1982), thus the facilitation of branching corals may form an important feedback for farmerfish population growth and corresponding community dynamics.
53 Figure 2 1 Mean frequency of coral predation The average frequency of coral predation and corresponding average daily loss in skeletal mass ( 95% CI) was higher in the removal or absence of S. nigricans for A) Acropora striata and B) Montipora floweri Treatment labels indicate the experimental level (PRESENT or REMOVED ( )) of Stegastes nigricans ( Stegastes ) crossed with farmed turf (Turf), and naturally ABSENT control reefs (), yielding five unique treatment combinations. Bar color indicates whether nubbins wer e CAGED (black) or EXPOSED (cage control or open = gray). Pocillopora verrucosa and Porites australiensis did not incur any obvious predation (Appendix D: Table D1) or significant differences in daily growth among treatments or subtreatments during the course of the experiment (Appendix C: Figure C1).
Figure 2 2 Frequency of coral overgrowth. A) Three of the four species tested, Porites australiensis Montipora floweri, and Acropora striata, incurred significantly higher overgrowth when farmed turf was PRESENT, versus REMOVED or ABSENT (treatment designations match Figure 21, but order is altered to highlight the effect of interest) B) The average amount of algae present in a given treatment predicted 53% of the variation in overgrowth frequency for P australiensis and 46% of the variation in overgrowth frequency for A striata but offered no significant predictive power for M floweri (Linear Regression, algal mass was arc sine square root transformed to meet normality assumption, here I present raw data, Appendix D: Table D 3). Pocillopora verrucosa incurred very little overgrowth (~10% or less) in any treatment, regardless of turf abundance (Appendix D: Figure D1, Tables D2D3).
55 Figure 2 3 Mean density, reef use, and foraging frequency of target fish families. A) Mean density of surgeonfishes (Acanthuridae), parrotfishes (Scaridae) and butterflyfishes (Chaetodontidae) ( 95% CI) among reefs with Stegastes nigricans PRESENT (dark gray bars), REMOVED (gray bars), or ABSENT (light gray bars ). B) The mean proportion of fishes observed using the experimental reefs (vs. the water column and adjacent substrate; 95% CI), and C) the mean proportion of fishes foraging on experimental reefs ( 95% CI) was consistently higher on S. nigricans removal reefs for surgeonfishes, parrotfishes, and butterflyfishes (see Results, Appendix E: Tables E3E5).
56 CHAPTER 3 INDIRECT EFFECTS OF STEGASTES NIGRICANS ON MASSIVE PORITES CORAL AND IMPLICATIONS FOR OTHER COMMUNITY MEMB ERS Introduction Trees (Ellison et al. 2005), pl ants (Alteri et al. 2007), coral (Bruno et al. 2003), and algae (Eriksson et al. 2006) function as ecosystem engineers because they create, modify, or maintain habitats for other species through their physical structure (Jones et al. 1994). Consumers have been shown to mediate the distribution of habitat forming species across ecosystems (Grasslands: Chase et al. 2000; Oak Savanna: Ritchie et al. 1998; Rocky intertidal zone: Bertness et al. 2002, Bertness et al. 2004; Kelp forest ecosystems: Halpern et al. 2006; Coral Reefs: Burkepile and Hay 2006) thereby changing resource availability for other community members Simultaneous with foundational (e.g., Dayton 1972) and consumer driven habitat modification (e.g., Power 1997), ecosystem engineers can further m odify community structure through nontrophic interactions (i.e., behaviors), to create patches with assemblages of organisms that differ from unmodified surrounding habitats (e.g., dam building by beavers, Jones et al. 1994, 1997). Engineering can have bo th positive and negative consequences for species that inhabit the new or old habitat, respectively ( Jones et al 1997). Some territorial damselfishes (Pomacentridae), known as farmerfish, engineer the community structure of coral reefs by developing an d maintaining conspicuous filamentous algal mats (hereafter, turf) (reviewed by Ceccarelli et al. 2001). Farmerfish of the genus Stegastes tend to farm dense, filamentous red algal turf, and Polysiphonia species in particular, are the dominant algae within the territories of more than half the Stegastes species investigated (Ceccarelli 2007). Of the common mechanisms of competition between corals and algae, farmerfish associated turf is most likely to
57 preempt settlement space, or overgrow, rather shading or abrading, corals (reviewed by Mc Cook et al. 2001). Filamentous turf can also reduce flow and increase rates of sedimentation (Purcell 2000), and sediment trapped by turf can harm corals (Potts 1977, McCook et al. 2001, Nugues and Roberts 2003) or reduce r ecruitment within areas occupied by turf (Birrell et al. 2005). In addition, there is some experimental evidence that indirect contact between corals and turf may facilitate mortality via bacteriaassociated disease (Smith et al. 2006). Algal turf within t erritories can increase the rate of overgrowth on coral (Potts 1977, Lobel 1980) and coral spat (Sammarco and Carleton 1981). Farmerfish may also actively remove coral recruits or living coral tissue to expand the size of territories (Kaufman 1977, Roberts on et al. 1981) and susceptibility to this damage can vary depending on coral morphology (Wellington 1982). When territories are established on dead coral, through whatever mechanism, there is higher bioerosion due to increased densities of boring sponges and sipunculans, presumably as a result of reduced grazing by other fishes within territories (Risk and Sammarco 1982). There is evidence that coral species vary in their susceptibility to overgrowth within farmerfish territories (Sammarco and Williams 1982, Wellington 1982, Glynn and Colgan 1988, Done et al. 1991, White and ODonnell In press ). For example, in a recent experimental test, massive Porites australiensis incurred considerably higher frequency of overgrowth with turf present, than other corals tested ( Acropora striata, Montipora flowerii, and Pocillopora verrucosa ) (Chapter 2). Within Moorea lagoons today, the common Pacific farmerfish, Stegastes nigricans, typically colonizes these dominant massive Porites (Done et al. 1991, this study). When colonized, these reefs have an
58 altered morphology (Done et al. 1991, Figure 31), likely due to the eroding influence of abundant boring invertebrates within territories (e.g., Risk and Sammarco 1982, Sammarco and Risk 1990). Massive Porites colonies occupied by Stegastes nigricans maintain higher abundances of filamentous turf (especially Polysiphonia spp.) and sediment (Vuxton et al. In preparation). Dendropoma maxima, a conspicuous, tubedwelling vermetid snail (Vermetidae), commonly inhabits colonies of Porites in Moorea lagoons (Peyrot Clausade 1992). Vermetid snails cast a mucus web to filter feed (Hughes and Lewis 1974). These mucus webs remove sediment (Kappner et al. 2000) and can smother algae and reduce turf associated sedimentation within S. nigr icans territories (pers. obs.). They also have been shown to decrease coral growth, presumably because their mucus nets can also interfere with coral feeding and/or respiration (Colgan 1985). Thus, vermetid snails may function as cryptic consumers and medi ate turf Porites competition if they are near the zone of contact, by indirectly lowering the sediment accumulation and, possibly, the rate of overgrowth by turf. This study asked the following questions: 1. What is the abundance of turf and live coral on m assive Porites patch reefs with adult Stegastes nigricans present and absent? 2. On Porites reefs colonized by Stegastes nigricans, i s the rate of overg rowth by algal turf driven by the effects of turf, sediment, and/ or vermetid snails? Previous research i ndicates massive Porites are particularly susceptible to overgrowth by turf (Chapter 2). This design scales up from using coral fragments (nubbins) to measuring rates of turf overgrowth at the live Porites turf transition zone.
59 Methods Paired Field Surveys To answer the first question, I selected 24 pairs of massive Porites reefs (colonized versus uncolonized by Stegastes nigricans Figure 31), in Vaipahu lagoon on the north shore of Moorea in April 2005. Paired reefs were similar in size ( Appendix F: Figu re F1) and location (average distance between pairs, mean 95% C.I.: 7.9 1.9 m), but differed in that colonized reefs were defined as having adult Stegastes nigricans present and consequently had obvious turf cover, whereas uncolonized Porites reefs had no adult S. nigricans or associated turf. I measured the dimensions of each reef, including footprint (maximum length and perpendicular width), average and maximum height, and a rough estimate of reef volume (length x width x average height; m3). In a ddition, I measured the composition each reef using fixed point contact, at 10 cm intervals, of 3 transects laid perpendicular to the longest reef axis. This yielded an estimate of live massive coral ( Porites ) and farmed filamentous algal turf cover in the presence and absence of farmerfish. For each reef, I also counted the number, and estimated the size (within 5mm), of all S. nigricans Transition Zone Field Experiment To answer the second question, I selected 9 massive Porites reefs interspersed within Vaipahu lagoon that were colonized by Stegastes nigricans. Experimental reefs contained at least four areas selected for quadrats, approximately equally spaced around the reef, that met the following criteria: 1) vertical orientation; 2) relatively flat su rface to accommodate reference nails; 3) roughly equal cover of live coral and turf algae; and 4) 2 3 vermetid snails in algae near or below base of quadrat and not in contact with coral. This insured adequate coverage of mucus nets over algae within a
60 g iven quadrat, while eliminating impact of mucus nets on coral. On each of these experimental reefs, 4 permanent photo quadrats were established, using 3 or 4 permanent markers (stainless steel nails) (Figure 32). Nails provided visual reference for the 20 cm x 20 cm photoquadrat frame and allowed compensation for error associated with photoquadrat placement (see Appendix F: Figure F 2 for representative photographs). I used experimental manipulations to decouple and explore the effects of turf, sediment, and vermetid snails on the rate of turf overgrowth onto living massive Porites This yielded four treatments (one quadrat for each treatment) on each of 9 reefs: +turf/+sediment/ vermetid snails; turf/ sediment/ vermetid snails; +turf/ sediment/ vermetid snails; +turf/+sediment/+vermetid snails (Table 31). Some quadrats were lost because reefs crumbled and degrees of freedom were constrained accordingly (Appendix F: Table F 1, Figure F 2). Treatments were randomly assigned, to control for variation in flow around a given reef, and maintained daily. Treatment implementation For all but the vermetid snails present quadrat (Treatment 4, Table 31), vermetid tubes and snails were removed with large nails. Although this removal only occurred at the beginning, this can be considered a press treatment, because vermetid snails will not settle and grow to sufficient sizes within the short (6 week) duration of the experimental trial. For turf removal plots, all existing turf was initially removed with stainless st eel brushes to expose underlying dead coral substrata. Turf removal treatments were then maintained by daily scrubbing with firm toothbrushes, taking care not to abrade live coral. For sediment removal quadrats, sediment was removed daily by gentle (manual ) rubbing of turf to loosen sediment, combined with brief, directed flow
61 (with my hand). The other two treatments (ambient turf and sediment with and without vermetid snails) were not manipulated post removal of vermetid snails (see Appendix F: Figure F 2 for representative photographs of treatments after implementation). Digital photographic analyses The experiment was conducted during the austral summer (14 February 30 March 2007). To measure the response to manipulations, photos were taken immediately after treatment implementation (initial) and retaken after 6 weeks (final). Photos were analyzed using CPCe (Kohler and Gill 2006) and measurements made were blind with respect to period (initial vs. final) and treatment (except for the turf and sediment removal treatments, which were obvious), but not with regard to quadrat, as paired lines required placement verification. For each photo, I recorded linear measurements (n=4/ quadrat) from the edge of each quadrat to the boundary of turf, and a second measurement from the edge of the quadrat to a distinct coral boundary (i.e., a distinct lobe of living Porites ) (see Appendix F, Figure F 3, for schematic of line measurements). Thus, the zone of live Porites was defined by this latter measurement (live coral plus turf or dead substrata, depending on treatment), minus the first measurement (just turf or dead substrata). For each line, I also measured the size of obvious dead tissue / exposed skeleton at the transition zone (i.e., the zone of dead tissue, see inset in Figure 36, schematic in Appendix F: Table F 3). At the termination of the experiment, I calculated the average linear change (final initial) in the live coral boundary zone for each quadrat (schematic in Appendix F: Table F 3). In the case of change in live coral, a positive response indicates tissue growth, while a negative response indicates tissue loss (overgrowth) (Appendix F: Table F 3). For the zone of dead tissue, a positive change would indicate a larger zone of tissue damage (prior to
62 overgrowth), and a negative change would suggest this loss of tissue was ameliorated or that algae overgrew it. Statistical Analyses Paired field surveys I performed a paired t Test to verify that reefs did not vary in initial volume (length x width x aver age height, m3; measurements were missing for one reef, thus paired sample sizes were reduced accordingly). I also used paired t Tests to evaluate whether there was significantly higher abundance and size (in mm) of Stegastes nigricans on colonized versus uncolonized Porites reefs. I examined the average difference in Porites reef composition (i.e., % cover) in the presence and absence of adult S. nigricans by first summing the cover of each category across the three transects for each reef to reduce the degrees of freedom to the appropriate level (n=24). I then tested the mean (arc sine square root transformed) proportion cover of each variable as a function of Reef Type ( Porites reefs colonized by S. nigricans vs. uncolonized Porites reefs; t.test, optio n=paired, R, v.2.9.2). Transition zone experiment I performed a linear mixed model to evaluate the linear change in live Porites tissue by the fixed effect of Treatment (Table 31), with the random effects of quadrat (n=4) nested within reef (n=9) (lme, R, v. 2.9.2, Appendix F: Table F3). This took into account the unbalanced design and insured a conservative treatment denominator degree of freedom of 21 (Pinheiro and Bates 2000, p. 91). I examined the residuals and Q Q plot. The change in live Porites data set was not perfectly normal; however, when I tested the subset of the data set that satisfied this assumption the results did not change (Appendix F: Table F1), and I report results from the full data set. In addition, I
63 evaluated the effects of interest using contrasts defined a priori (Table 31; Appendix F: Table F2, Figure F 2 ). The same analysis was repeated for the change in the amount of dead tissue between live Porites and algal turf (i.e., the dead tissue zone) (lme (package: nlme), R, v. 2.9.2, Appendix F: Table F3). Results Observational Results: Paired Surveys As defined above, colonized reefs had adult Stegastes nigricans present, whereas uncolonized Porites reefs had no adult S. nigricans Paired reefs did not differ significantly in s ize (t=1.87, df=22, P >0.05; Appendix F: Figure F 1). Colonized Porites reefs had 10 times more S. nigricans (across size classes) than uncolonized Porites reefs (t=10.60, df=23, P <0.001, Figure 33). Of the 24 uncolonized Porites reefs examined, juvenile S nigricans were observed on 7 reefs. The sizes of S. nigricans were substantially larger on colonized reefs (Figure 33, t=5.55, df=6, P =0.001). With this difference in S. nigricans size and abundance (Figure 33), there was an associated change in reef c over (Figure 31 and 34). On average, uncolonized Porites reefs had nearly twice as much live Porites cover (80.5 8.9%) than colonized reefs (42.9 10.1%) (t test, t=6.71, df=23, P <0.001). In contrast, these colonized reefs had roughly 25 times more fi lamentous algal turf (42.0 8.2%) than uncolonized reefs (1.6 3.0%) (Figure 3 4; ttest, t=7.99, df=23, P <0.001). Each of the other cover categories (i.e., macroalgae, dead coral, crustose coralline algae, and other live corals) comprised on average <5% of the total reef cover and were not further analyzed. Experimental Results: Porites Turf Transition Zone The change in live Porites tissue varied significantly among treatments (Figure 35; lme, F3,21=3.33, P =0.04, Appendix F: Table F 1). A profound loss of live Porites
64 occurred under ambient abundances of turf and sediment (and removal of vermetid snails) (nearly 5 cm in 6 weeks, Figure 35). In contrast, loss of live Porites tissue was lowest with sediment removal (Figure 34). Loss rates were also relatively low with manual removal of turf or, as well as with ambient density of vermetid snails, which cast a mucus web and inadvertently remove sediment from turf (see video of D. maxima feeding, Object 31). As defined a priori (Table 31), there was not a significant Turf effect (t=1.04, df=21, P >0.05, Appendix F: Table F 2) or Vermetid effect (t=1.04, df=21, P >0.05, Appendix F: Table F 2); however, there was a significant effect of Sediment (t=2.27, df=21, P =0.03, Appendix F: Table F 2). There was no significant difference in the size of the dead zone among treatments (Figure 36; Appendix F: Table F 3). Object 3 1. Video of vermetid feeding (28.9 MB .avi) Discussion Stegastes nigricans farms filamentous algal turf through a series of behaviors, including weeding unpreferred algae (e.g., Hata and Kato 2002, 2006) and invertebrates (e.g., snails: Lobel 1980; urchins: Williams 1981), as well as territorial defense (reviewed by Ceccarelli et al. 2001). Massive Porite s are the dominant hermatypic coral in the Moorea lagoon today. Those occupied by S. nigricans have roughly half as much live coral cover as uncolonized Porites reefs, with a concomitant increase in filamentous algal turf (Figures 31 and 33). With ambi ent levels of turf and sediment (and removal of vermetid snails), the rate of overgrowth by algal turf was nearly 5 centimeters in a six week period (Figure 35). However, removal of sediment (only) countered this overgrowth and live coral tissue significa ntly increased, with a reciprocal loss of turf, within these quadrats (Figure 35). Among the competitive
65 mechanisms used by algae, recent evidence suggests that the alga is not always the direct agent of death, but rather the trigger. For example, the green, calcareous alga, Halimeda opuntia, hosts a bacterium (Aurantimonas coralicida) t hat stimulates disease (white plague type II) of the massive coral Montastraea faveolata, and thus facilitates its overgrowth ( Nugues et al. 2004). Similarly, Smith et al. (2006) suggested that the release of Dissolved Organic Carbon by algae could enhance microbial abundance and adversely affect corals. Potentially an increased availability of nutrients or food (such as Particulate Organic Carbon in sediment) may induce a bloom in the bacterial community that can either lead to pathogenic infection of corals or cause coral tissue loss due to physiological stress such as hypoxia as bacteria consume available oxygen (Kline et al. 2006; Kuntz et al. 2005). Indeed, recent analys es suggest there is a 60% reduction in Dissolved Oxygen at the interface between farmerfish turf and coral tissue, with associated loss of coral tissue (Barott et al. 2009) as was observed in the Moorea system (see inset in Figure 3.6). Farmerfish may fac ilitate settlement, growth, and survival of some mobile organisms through habitat modification (Ceccarelli et al. 2001). Dense algal turf on the lobes of massive corals may facilitate a rugose morphology, in the place of less complex mounds (Figure 31, Do ne et al. 1991), because there are higher densities of internal bioeroders within territories (reviewed in Ceccarelli et al. 2001). Increased bioerosion weakens the structure of a reef (Risk and Sammaro 1982; Huston 1985) and increases its structural compl exity (Figure 31, Done et al. 1991). The turf itself also adds structure, with some Polysiphonia (e.g., P. sertulosa) maintaining average lengths of 15 mm (pers. obs.). This added complexity facilitates the buildup of higher densities
66 of small invertebrat es than adjacent, more barren, areas (Klumpp et al. 1988), and, with the higher abundance of algal turf, may augment the food supply for other territory residents. For example, Green (1992, 1998) noted higher abundance of juvenile labrids and scarids withi n damselfish territories. Within S. nigricans territories in Moorea, it is common to find massive Porites colonies that have undergone partial mortality and are covered with turf, with live coral limited to the top of the lobes (Figures 31 and 32). These territories have disproportionately higher densities of juvenile fish than adjacent live coral habitat and juvenile wrasses, Thalassoma hardwicke, experience weaker density dependence (Shima et al. 2008), suggesting an indirect positive effect of the habi tat modification. Conclusions and Future Directions Massive Porites are the dominant hermatypic corals within the lagoons of Moorea (Beruman and Pratchett 2006). Their ability to settle to rubble on an otherwise sandy surface (Idjadi and Edmunds 2003) off ers a great resource to other taxa that require hard bottom for settlement. Vermetid snails are the most conspicuous sessile fauna on Porites (followed by the serpulid, Spirobranchus giganteus ODonnell and White, I n preparation, White et al. In preparati onb ). Sessile vermetid snails consume sediment and entrapped particles at a nearly constant rate (Kappner et al 2000) and this sediment consumption may indirectly reduce the rate of Porites overgrowth by turf within S. nigricans territories (Figure 3 5). Although, Coles and Strathmann (1973) reported no fish feeding on vermetid snail mucus, I have observed adult S. nigricans steal (and eat) the mucus webs from large vermetid snails ( Dendropoma maxima) within their territories. Indeed, S. nigricans is consi dered to be at least a partial detritivore (Wilson and Bellwood 1997), thus these sessile snails, which can arrest the
67 development of the farmed turf front, may compete with S. nigricans for food resources. Additional research is necessary to understand how physical drivers that influence sediment levels (e.g., flow), affect the rate of overgrowth of S. nigricans associated turf over massive Porites, as well as whether microbial drivers are responsible for the loss of live Porites in the presence of turf and sediment. Ongoing research seeks to quantify whether vermetid snails are depressed within S. nigricans territories (C.D. Hartman, senior honors thesis) by evaluating the change in abundance and size of vermetid snails through time using photographs of recruitment tiles outplanted on reefs with S. nigricans naturally absent, present, and experimentally removed (Hartman et al. In preparation).
68 Table 31. Transitionzone Experimental Treatments Treatment Vermetid snails Turf Sediment a priori Contrasts of Interest 1 (Control) + + 2 2 and 3: Turf effect 3 + 1 and 3: Sediment effect 4 + + + 1 and 4: Vermetid effect Notes: Summary of manipulations to lower (turf) portion of each randomly assigned experimental quadrat on a given reef (N=9). All quadrats original ly contained 23 vermetid snails within the turf portion. Vermetid snails were removed from all quadrats except treatment 4 (vermetid snails, turf, and sediment present).
69 Figure 31. Representative photos of reefs by type. Top panel: Healthy live un colonized Porites reef. Bottom panel: Porites reef colonized by the dusky farmerfish, Stegastes nigricans (note the silhouette of a S. nigricans in the top right corner, as well as several other visible individuals).
70 Figure 32. Simplified schematic of transition zone quadrat design. For each experimental reef, four quadrats were placed on a vertical relief, with live Porites occupying the top half, and farmed algal turf occupying the bottom half, of each quadrat. For each quadrat, stainless steel nails were used to mark the photoquadrats and verify quadrat placement during subsequent digital measurements using CPCe (Kohler and Gill 2006). For representative photos of actual quadrats and a schematic of linear measurements, see Appendix F: Figure F 2 and Figure F 3).
71 Figure 33. Paired reef surveys: Mean number and size of Stegastes nigricans Reef type refers to Porites reefs colonized by adult Stegastes nigricans versus uncolonized Porites reefs (n=24 pairs or reefs, mean 95% CI). Top panel: Mean number of all S. nigricans (across size classes) was higher on colonized, than uncolonized, Porites reefs (paired t test, t=10.60, df=23, P <0.001) Bottom panel: Only juvenile S. nigricans were encountered on 7 of the 24 uncolonized Porites reefs and, as expected, when compared to their paired colonized reef, the mean size of S. nigricans was smaller (paired t test, t=5.55, df=6, P =0.001).
72 Figure 34. Paired reef surveys: Mean percent cover live massive Porites and algal turf. For photos and descriptions of Reef Type see Figures 31 and Figure 33. Top panel: There was significantly higher cover of live massive Porites coral on uncolonized, relative to colonized, Porites reefs (n=24 pairs or reefs, mean 95% C; paired t test, t=6.71, df=23 P <0.001). Bottom panel: Percent cover of farmed algal turf was significantly higher on Porites reefs colonized by S. nigricans, versus uncolonized Porites reefs, (paired t test, t=7.99, df=23, P <0.001).
73 Figure 35. Mean linear change in live massive coral tissue by treatment. The average rate of linear change in live Porites tissue within Stegastes nigricans territories during a 6week period in the austral summer 2007(14 Feb 30 Mar 2007) varied according to treatment (mean s.e; LMM, F3,21=3.33, P =0.04; see Table 31 for description of treatments and contrasts). Of these, only sediment removal treatments were significantly less overgrown than the treatment with both turf and sediment present (i.e., a sediment effect; Table 31; ttest, t=2.27, df= 21, *P =0.03; Appendix F, Tables F1 and F2).
74 Figure 36. Mean linear change in the amount of dead Porites tissue. Inset illustrates a large dead tissue zone in which the live coral and algal turf are separated by approximately 1 cm. There was no signifi cant difference in the average rate of change of dead Porites tissue at the transition zone between healthy Porites and farmed turf within Stegastes nigricans territories according to treatment during a 6 week period in the austral summer 2007 (14 Feb 30 Mar 2007; mean s.e.; LMM, F3,21=0.26, P =0.86; Appendix F, Table F 3; see Table 31 for descriptions of treatments and contrasts).
75 CHAPTER 4 EFFECTS OF FARMERFIS H AND CORAL PREDATORS ON THE POPULATION DYNAMICS OF BRANCHING CORALS IN A DISTURBED LAGOON SYSTEM Introduction Coral reefs are declining worldwide due to natural and anthropogenic disturbances (Bruno and Selig 2007, Knowlton and Jackson 2008). Recovery varies both within and among systems, and ecologists are struggling to understand the mechanis ms that constrain resilience (i.e., recovery to the previous state) (e.g., Bellwood et al. 2004). Efforts to understand variable recovery are complicated because, as with most marine invertebrates, corals have complex life cycles (e.g., Hughes and Jackson 1980; Hughes 1984). Most ecologists use small scale experiments to draw inference on mechanisms limiting recovery (e.g., Burkepile and Hay 2006). Alternatively, monitoring programs seek to demonstrate the community wide importance of topdown and bottom up mechanisms on coral health (e.g., Mora 2008) or uncover largescale variation driving differential coral recovery (e.g., Connell et al. 1997; Hughes and Connell 1999). While both of these approaches are important and useful, neither considers the implicit complexity of coral population dynamics. One very successful strategy is the combination of small scale experiments with larger scale monitoring, and associated population modelling, to scaleup our understanding of processes constraining resilience (e.g. Edmunds 2005). These, however, are rare and, to my knowledge, this is the first experiment to measure population rates of interest (recruitment, growth, and survival) for in situ size classes of corals, both before and after experimental manipulations.
76 E vidence Farmerfish Engineer the Coral Community Farmerfish represent an important component of the coral reef fish community and can account for nearly half of the total fish abundance (e.g., La Reunion Island, Letourneur et al. 1997), with their territori es covering more than 50% of shallow reef tracts at some sites (Panama: Wellington 1982; Great Barrier Reef (GBR): Sammarco and Williams 1982; GBR/Papua New Guinea (PNG): Klumpp et al. 1987). These siteattached damselfish (Pomacentridae) can act as ecosys tem engineers ( sensu Jones et al. 1994) and alter the benthic environment using a combination of farming behaviors, which enhance algal turf, and territorial defense against herbivores and coral predators (corallivores) (reviewed in Ceccarelli et al. 2001; see also Chapter 2). Coral species can respond differently to the indirect effects of these algaefarmers (Wellington 1982, Chapters 2 and 3). Increased biomass (Hata and Nishihira 2002; Hata et al. 2002) and productivity (Klumpp et al. 1987) of farmed al gal turf within farmerfish territories can overgrow or increase sedimentation on corals (Vine 1974, Potts 1977) and increase (Risk and Sammarco 1982) or decrease bioerosion, depending on fish grazing (Sammarco et al. 1986). Territorial behavior can create spatial refuge from predation by foraging coral eating (corallivorous) fishes (Wellington 1982; Gleason 1996; White and ODonnell In press ) or echinoderms (urchins: Sammarco and Williams 1982, Suefuji and van Woesik 2001; crownof thorns seastar: Glynn and Colgan 1988). While it is wellaccepted that grazing herbivorous fishes (Bak and Engel 1979) and urchins (Sammarco and Williams 1982) can reduce survivorship of juvenile corals, few studies have examined whether corallivores can alter the distribution and abundance of larger size classes of corals (Randall 1974; Hixon 1997), and additional research is warranted (Rotjan and Lewis 2008).
77 In the lagoons of Moorea, vulnerable corals have recovered faster in the presence of the dusky farmerfish, Stegastes nigr icans, with greater abundance, size, and diversity of corals within territories (Done et al. 1991; Gleason 1994; Gleason 1996; Shima et al. 2006, 2008; Galzin et al., In preparation). Small scale manipulations utilizing coral fragments indicated that Acrop ora suffer considerably less predation in the presence (versus removal or natural absence) of S. nigricans (Chapter 2). However, this pattern did not hold for Pocillopora a now very common coral within the lagoons (Berumen and Pratchett 2006) that was previously reported to incur less predation within S. nigricans territories, relative to highly grazed or macroalgae dominated patches (Gleason 1994). Clearly, measuring interactions with coral fragments does not give us a clear indication of the true per cap ita effects on corals, nor whether sensitivity varies with size. Thus, a demographic framework is necessary to tease apart whether higher recruitment or reduced mortality for juvenile and/or adult stages (either independently or in combination) could be responsible for the observed differential abundances within territories. Diverse Life History Strategies A complication to understanding community resilience is incorporating the diverse life history strategies of various taxa (Hughes and Tanner 2000), as this variation drives whether a population declines, experiences local extinction, or recovers following disturbance (e.g., Hughes 1989). The skeletal morphology of a given taxon largely determines whether individuals are susceptible to breakage (fragmentati on), predation and associated loss of linear extension (shrinkage), or mortality due to natural disturbances (e.g., storms). In general, massive Porites are slow growing, but hearty, due to dense skeletal formation and boulder like colonies that are not am enable to
78 breakage (Hughes 1989). In contrast, fragile branching Acropora generally have less dense skeletal structure and are more susceptible to natural disturbances (Woodley et al. 1981, Hughes 1989, Knowlton et al. 1988), except competition, because th eir associated rapid growth enhances overgrowth and shading of other corals (Connell 1978, Connell et al. 2004). The coral reefs of Moorea have been subjected to a series of natural disturbances, since the 1980s, including cyclones, bleaching events, and outbreaks of predatory crown of thorn seastar, Acanthaster planci (Adjeroud et al. 2005). The dominant coral genera of Moorean reefs differ in their susceptibility to these disturbances (Table 41). By 2003, on the north side of Moorea, the mean coral cover in the back reef had returned to near predisturbance (1979) levels (~45% coral cover), but with altered community composition (Berumen and Pratchett 2006). Massive Porites spp. were little affected by the disturbances, but branching Pocillopora spp., whi ch were historically at low densities in this system (Bouchon 1985), have now surpassed once dominant Acropora spp. (Berumen and Pratchett 2006). Acropora are sensitive to disturbances (e.g., Woodley et al. 1981, Hughes 1989, Adjeroud et al. 2005), while Pocillopora can be considered to be weedy, in that they can recover more quickly following a perturbation (e.g., McClanahan et al. 2007). I monitored these two genera simultaneously to elucidate why Pocillopora have been observed to recover from disturbances better than historically dominant Acropora with in lagoons of Moorea, French Polynesia (Done et al. 1991, Berumen and Pratchett 2006). Demographic Complexity As long lived organisms, corals are more sensitive to increases in mortality rates for establ ished individuals, than changes due to fluctuation in recruitment, whereas
79 short lived taxa are generally recruitment limited (Hughes 1990). Indeed, large colonies may preserve abundances via the storage effect (Warner and Chesson 1985), though asexual reproduction (i.e., fragmentation) is likely only important for branching corals whose morphology is amenable to fragmentation and reattachment (or solitary fungiids, which bud) (Veron 2000). Asexual reproduction by fragmentation is common for dominant taxa in both the Caribbean and the IndoWest Pacific (Highsmith 1982). Further, the size and number of fragments depends on initial size and morphology of a given coral colony, and both factors contribute heavily to the expected contribution of fragments to po pulation growth (Highsmith 1982). Taxa that brood larvae (e.g., Pocillopora) also likely contribute to the next generation faster (i.e., their population structure is more closed) than the typical strategy of broadcast spawning (e.g., Acropora), especial ly at low densities following a disturbance (Knowlton 2001). Because corals are modular organisms, after recruitment to an adult population, they do not only progress sequentially through size classes. Corals can shrink, loop (i.e., stay in the same size class), fuse (with identical genets), or fragment (Hughes 1984), and these processes combine to decouple sizeage correlations (Hughes and Jackson 1980). Smaller corals usually have faster growth rates, but are more likely to be killed than larger corals (C onnell 1973) and, to further complicate matters, this relationship can vary with tissue age (i.e., fragments grow slower than true juveniles, Elahi and Edmunds 2007a, or younger tissue, Elahi and Edmunds 2007b). Scaling Up Most experimentation is limited to small coral fragments, due to both logistical and ethical concerns: This may give misleading results, because fragments may or may not behave like the life stage they are intended to represent (e.g., Elahi and Edmunds
80 2007a,b). More importantly, the ability to generalize within a population (across size classes) is limited. One possible approach is combining observed natural gradients in distribution (e.g., like those associated with keystone species or ecosystem engineers, Soul et al. 2003), with care ful experimentation to tease apart mechanisms driving the observed community structure. The main objective of this research was to scaleup our understanding of the demographic processes underlying the observed higher abundances of branching corals within S. nigricans territories, as well as possible drivers of the observed lower recovery of Acropora versus Pocillopora within the lagoon. Understanding the relative vulnerability of different size classes to mortality can inform management practices (e.g., Cr ouse et al. 1987), including restoration efforts, and identify data critical to refining our understanding of processes limiting coral recovery. In particular, I sought to answer the following questions: 1. Does the presence of Stegastes nigricans increase the survival, growth, or recruitment of Acropora or Pocillopora ? 2. Does the strength of this interaction vary with coral size? If so, what sizes of coral are most protected from predation by Stegastes nigricans territoriality? 3. Do these differences lead to a lower rate of p Acropora and/ or Pocillopora when Stegastes nigricans is present? I document ed the mortality and growth rates of in situ populations of Acropora and Pocillopora within Stegastes nigricans territories between 2006 and 2007. This pro vided a baseline of branching coral population dynamics (including the rate of in the presence of S. nigricans I then removed S. nigricans from half of the experimental reefs and documented the change in key demographic parameters for both control reefs, and reefs with S. nigricans removed between 2007 and 2008 (i.e., a life table response experiment) In addition, I measured fragmentation
81 and recruitment of Acropora and Pocillopora to experimental reefs throughout both monitoring periods to incorporate the dynamic aspect of these open populations. While following individuals through time a fecundity is essentially decoupled from the local population and a measure of input to an open population (i.e., recruitment and fragmentation) is necessary to complete our demographic understanding using matrix models (e.g., Hughes 1984, 1990). Methods Study System Moorea is a high volcanic island located in the Society Archipelago, approximately 14.5 kilometers from Tahiti. It has a well developed barrier reef along most of the island perimeter. Much of the lagoon is shallow (12 m). L agoon bottoms are dominated by sand and rubble inshore, with patch reefs ranging from relatively isolated in the backreef to nearly continuous near the reef crest on a foundation of hardened pavement. These patch reefs are predominately built by massive Porites corals, which have lower susceptibility to the natural disturbances observed in this system (Adjeroud et al. 2005, Penin et al. 2007, Berumen and Pratchett 2006). In Moorea, the dusky farmerfish, Stegastes nigricans, com munally defends large territories of thick filamentous turf (mostly Polysiphonia and Corallophila spp. Vuxton et al. In preparation) Around less disturbed islands, S. nigricans typically inhabit arborescent Acropora thickets (e.g., Indian Ocean (Reunion) : Lison de Loma and Ballesteros 2002; Indo West Pacific (PNG): Jones et al. 2006; Taiwan: Jan et al. 2003; French Polynesia: Mapstone et al. 2007). These habitats were virtually eliminated in Moorea (Berumen and Pratchett 2006, Mapstone et al. 2007), and S nigricans also colonizes the dominant massive Porites (Done et al. 1991) Porites reefs with adult S.
82 nigricans territories have significantly reduced live coral cover, than uncolonized reefs (Chapter 3), and altered coral morphology (Done et al. 1991). When portions of these massive corals die, they become substrate for other sessile organisms, including crustose coralline algae, filamentous and macroalgae, and various corals, including smaller massive, encrusting, branching, and plateforming species (Done et al. 1991, Gleason 1996, Galzin et al. In preparation). In Moorea lagoons, territory sizes can reach several meters across (Done et al. 1991), with fish densities of up to 80 per square meter (Gleason 1996). Experimental Reefs Matched pairs of co ntrol and experimental reefs were haphazardly distributed within northeastern Aroa lagoon. Prior to beginning the study, I chose experimental reefs using the following search image: large massive Porites patch reefs, colonized by Stegastes nigricans with ample branching Acropora (which are numerically rare in this area relative to Pocillopora ). Further, relatively isolated reefs were selected, to reduce the likelihood of recolonization following pulse removals of S. nigricans Whenever I found a suitable candidate reef, I searched the nearby area for another reef of similar size, isolation, coral community structure (e.g., size distibutions of Acropora and Pocillopora), and density of S. nigricans (Appendix G: Table G 1). For each reef, I measured the reef footprint (maximum length x perpendicular width), reef height (maximum and average), the water depth to lagoon floor, and distance to the three nearest patch reefs, as a rough estimate of isolation. In addition, I estimated the habitat composition of each reef by laying three transects perpendicular to the longest axis and documenting the dominant cover under each 0.10 cm interval. Bottom cover categories were combined as appropriate to form the following categories: live coral (identified to
83 genus), filam entous algal turf, macroalgae (identified to genus, except cyanobacteria), and dead coral and crustose coralline algae. Estimates of reef composition were made once, at the beginning of the study; however, I estimated the number and size of all S. nigricans throughout sampling (July 2006, April 2007, July 2007, April 2008). Between March and July 2007, strong storms collapsed or overturned several reefs, smashing delicate branching corals and reducing the sample size of large marked colonies, in particular In July, I added four additional reefs to compensate for large colonies lost during the storms. The search protocol for these reefs was the same as that described above; however, I chose reef pairs that were either very large, or that contained high dens ities of branching corals, in order to meet the goal of increasing the sample size of marked branching corals (Appendix G: Table G 1). Demographic Monitoring Between June 2006 and April 2008, I tagged and repeatedly measured and photographed almost 700 individual corals from the genera Acropora (N=221) and Pocillopora (N=458) during complete biannual censuses of 16 isolated S. nigricans territories (terminating in July 2006, April 2007, July 2007, and April 2008). Individuals were tagged using round aluminum Forestry tags affixed to the reef using nails and Z sparTM (a two part epoxy: Splash zone A788, Kopcoat, Pittsburgh, Pennsylvania, USA). During each census, each individual was measured (linear length, width, and height of average branch) and photographed. If an individual was partially overgrown by turf, I noted an estimate of the percent of the skeleton covered by turf, and the presence (=1) or absence (=0) of bleaching, obvious (skeletal) predation, or complete mortality.
84 Life Table Response Experiment Experimental reefs were selected in pairs based on reef size, location, abundance of adult S. nigricans, and branching coral abundance. One reef from each pair was randomly assigned to have S. nigricans removed (Appendix G: Table G 1). I removed Stegastes using clove oil, a common fish anesthetic, after the July 2007 census. A previous study indicated that high concentrations of clove oil could increase bleaching and reduce growth for both Acropora and Pocillopora (Boyer et al. 2009). Therefore, I used a l ow concentration of clove oil (7% in seawater), recorded the number of bottles applied to a given removal reef, and applied the same amount to the control (Present) reefs to eliminate potential confounding effects (but on control reefs S. nigricans indiv iduals were allowed to recover, rather than being removed). Throughout the experiment, any new Acopora or Pocillopora recruits were also marked, measured, and photographed during each census. Statistical Analyses For all analyses, Treatment refers to r eefs with Stegastes nigricans naturally PRESENT or experimentally REMOVED. Period distinguishes between the 200607 observational period BEFORE removals occurred, versus the 200708 EXPERIMENTAL period. We expect no differences between treatments in the Period before removals because there were no manipulations during that time; instead, this period controls for (and quantifies) possible preexisting differences among reefs. Reef size isolates differences due to differences in the size of experimental reefs (Appendix G: Table G 1). For coral responses, Previous Size is a continuous covariate testing whether the size of a coral colony at the previous time step influences the likelihood of mortality or the magnitude of change in size. In addition, I tested the interactions that were relevant for
85 each response variable / subset of the data, as reported below. In general, the main inference of interest is the Treatment x Period interaction, and I eliminated (only) nonsignificant higher order interact ions (Zuur et al. 2009, P. 127) to respect the marginality of lower order terms (Nelder 1998). For all tests that assume that the error is normally distributed, I assessed normality using Q Q plots and evaluated equality of variance using Levenes test. For mortality estimates, which assume a binomial error distribution, I assessed model fit graphically. Reef attributes I tested to be sure differences in the size of reefs (m2) were attributable to Reef Pair, rather than Treatment (i.e., that my blocking was effective), using a linear model (Linear Model (LM), lm, R v. 2.9.2, Appendix G: Tables G 2 and G 3, Figure G 1). I used a similar model, but with the added random effect of Pair, to assess differences in the abundance of Stegastes nigricans BEFORE treatment implementation, while accounting for variation associated with reef size (July 2006, April 2007, July 2007; Appendix G: Table G 4 and G 5, Figure G 2). I used the same model to assess the efficacy of removals after the EXPERIMENTAL period (i.e., whether reefs differed by Treatment, rather than Pair after removals) (April 2008; Appendix G: Table G 6 and G 7, Figure G 3). Mortality Prior to assessing mortality, I restricted the data set to exclude fragmented colonies and their progeny (i.e., sing le colonies that split into two or more smaller colonies), as this would skew estimation of the size specific mortality rate (e.g., Hughes 1984). I assessed the frequency of mortality as a function of Treatment, Period, Previous size, and all of thei r two and threeway interactions, as well as, an additive
86 effect of Reef size. The full model was: Mortality ~ Treatment + Period + Previous Size +Reef size + Treatment*Period + Previous Size*Period + Period*Previous Size + Treatment*Period*Previous Siz PAIR REEF Pair (N=8) and Reef (N=16) ( Acropora, N=192; Pocillopora N=381, Appendix H: Table H 3 and H 4, Figures 4 1 and 42). In addition, I performed follow up analyses for relevant subsets of the data. F or each genus (i.e., Acropora and Pocillopora, independently), I examined the BEFORE data in isolation to test whether mortality differed among treatments or reefs prior to actual treatment implementation. For this analysis, Period was not relevant and the model was constrained accordingly. Thus, the model for this data subset included Treatment, Previous Size, their interaction, Reef Size, and random effects of Pair (N=6) and Reef (N=12) ( Acropora, N=66; Pocillopora, N=134, Table H 5 and H6, Figures 41 and 42, respectively). In addition, I examined the PRESENT only reefs (across both periods), to test whether the mortality rate varied between years within S. nigricans territories. For this test, Treatment was not relevant and the model i ncluded only Period, Previous Size, their interaction, Reef Size, and the random effect of Reef (N=16) ( Acropora, N=94; Pocillopora, N=206; Tables H 5 and H6, respectively). The random effect of Pair was also omitted from this model, as it is red undant without the reefs associated with the other treatment. Both of these tests using reduced data sets provide verification of the experimental design and evaluate the baseline mortality rates in the PRESENCE of S. nigricans; however, neither evaluates the main inference of interest (the Treatment x Period interaction). For all mortality models, I used a Generalized Linear Mixed Model (GLMM) (binomial distribution, logit link) fit with the Laplace approximation (glmer, lme4 package, R v
87 2.9.2). I per formed Likelihood Ratio Tests (LRT) of the models defined a priori to isolate which experimental variables best predicted variation in the frequency of mortality for the entire data set, as well as for relevant subsets of the data. Change in size of corals I followed the change in size of all corals, but excluded from analyses those colonies that experienced complete mortality. Live areas (only) were measured for colonies that underwent partial mortality. I also removed fragmented colonies from the data set as these skew the rate of negative growth in an artificial manner. Instead, I report the sum and mean size of resulting fragmented colonies, as well as the average number and size of fragmented daughter colonies, by Treatment and Period in Table 4 2. I used a Linear Mixed Model (LMM) defined a priori to isolate which experimental variables significantly affect the chang Previous Size) among and within reefs. For both Acropora (N=123) and Pocillopora (N=272), the full model tested was: Change in Size ~ Treatment + Period + Previous Size + Reef size + Treatment*Period + Previous Size*P eriod + Period*Previous Size + PAIR REEF Reef (N=16) and Pair (N=8) (Appendix H: Table H 7; Figure 43 and 44, respectively). I also performed similar additional analyses, where I c onstrained the dataset to the before period and present treatment only, using the same models reported for mortality frequency above, but using LMMs to examine change in size (lme, nlme package, R v 2.9.2; Appendix H: Table H 8 and H 9; Figure 43 and 44) Because variability was not constant between periods, I added a term to specify constant variance within (but not
88 between) periods to all model s that included Period as a term (weights=varident (form=~1|Period) lme, nlme package, R v 2.9.2). Matrix Models I compared sizespecific demographic rates (e.g., survival, growth, and transition rates including looping (or stasis), shrinkage, fragmentation, and mortality, s ensu Hughes 1990) on reefs where S. nigricans is naturally present and experimentally removed. Due to differences in sample sizes, the size classes estimated varied by taxon, with one more size class for Pocillopora than for Acropora (Appendix I: Figure I 1). As is standard for matrix population models, I computed the dominant eigenvalue of usual way as an intrinsic population growth rate; instead, it is essentially a measure of population decay and is expected to be less than 1.0 (i.e., recruits were not included in fragmentation alone can maintain a growing population, with zero recruitment of new genets from larvae (Hughes and Tanner 2000). I report the mean population de and 95% confidence intervals as an alternative summary of the previously described continuous processes (see Appendix I: Tables I 2 through I 7 for transition matrices). This approach integrates the effects of the (possibly sizedependent) changes in growth and mortality between treatments and years into a summary measure of the effects on overall population demography (Figure 45). Recruitment and Fragmentation Given the low rates of both fragmentation (Table 42) and recruitment (Table 43) to experimental reefs, I refrained from formal testing and instead reported the raw
89 counts, and mean size with 95 % Confidence Intervals (1.96 standard error), and made basic comparisons as outlined briefly below. Results Life Table Response Experiment Acropor a mortality The effect of Stegastes nigricans Treatment varied by experimental Period for Acropora (Figure 4 1), with mortality rates considerably higher on reefs where S. nigricans were REMOVED (i.e., there was a significant Treatment by Period inte raction; 2=11.51, df=1, P =0.0007; Appendix H: Tables H 1 & H 2). In the BEFORE period, total mortality was low (7 15% across all size classes and both Treatments, Table 42) and the model estimated that very small individuals (1.0 cm) incurred mortality at relatively low levels (~ 1030% on average) across experimental reefs (from both treatments) (Figure 41). The estimated average mortality probability for Acropora was higher on reefs in which Stegastes nigricans was REMOVED (~75 90%), relat ive to those with S. nigricans PRESENT (~0 45%) (Figure 41). In particular, the vulnerability of larger Acropora varied with the presence of S. nigricans 2=4.74, df=1, P =0.03; Appendix H: Table H 1). During the BEFORE period, large colonies (> ~12 cm previous maximum length) all escaped mortality, although there were very few such colonies and the estimat es of mortality were uncertain (Figure 41). In contrast, mortality of large colonies was 100% on reefs in which S. nigricans was REMOVED during the EXPERIMENTAL period, when the sample size was also larger (Figure 4 1 ). Thus, in the EXPERIMENTAL period total mortality of Acropora (across size classes) was nearly 5 times
90 higher in removal (75%) relative to control/present (17%) reefs (Table 42; Appendix H: Figure H 1). The Acropora data from the BEFORE period showed no significant effect of Treatment (LR 2=1.23, df=1, P =0.26) or any other factor tested (Appendix H: Table H 3), confirming that mortality rates did not differ among reefs assigned to alternate treatments prior to S. nigricans manipulations (Figure 41, Table 42). When the data were const rained to examine the mortality probability of Acropora on reefs with S. nigricans PRESENT 2=24.36, df=1, P <0.0001), but the effect of 'Period' was not significant (nor were any other factors, Appendix H: Table H 3). This suggests mortality pressure varied little across years for Acropora (15% and 17%, respectively; Table 42). Despite an apparent size refuge from predation in the presence of S. nigricans (Figure 4 1), corals may suffer overgro wth by turf, causing individuals to shrink (see Change in Coral Size) or fragment (see Fragmentation). Pocillopora mortality The results for Pocillopora were similar to those for Acropora in that the effects of Stegastes nigricans territorial defense var ied between the BEFORE and EXPERIMENTAL 2=17.54, df=1, P <0.0001; Appendix H: Tables H1 and H 2). Across size classes, total mortality was approximately 3 times higher on reefs with S. nigricans REMOVED (59%) versus PRESENT (17%) (Table 43). The predicted mortality probability was higher for small individuals of Pocillopora (1 cm) on reefs where S. nigricans was REMOVED relative to PRESENT during the EXPERIMENTAL period (~75% and 30%, respectively; Figur e 42). These differences were similar in magnitude to the differences observed for Acropora (Figure 4 1). In contrast to the
91 patterns observed for Acropora, removal of S. nigricans did not significantly affect the mortality rates for larger Pocillopora an d there was no significant interaction between Previous size and Treatment (Appendix H: Table H 1). There is, however, some evidence that the mortality rates for Pocillopora were generally higher during the EXPERIMENTAL period, including slightly higher mortality in the presence of S. nigricans (i.e., total mortality increased from 6% to17%, Table 43). Thus, although there appear to be higher overall mortality rates for Pocillopora in the EXPERIMENTAL period, regardless of S. nigricans defense (Table 43), there were no differences between periods when the data were constrained to the PRESENT treatment (Figure 42; Appendix H: Table H4). However, as for Acropora larger corals have a lower probability of mortality for Pocillopora when S. nigricans is P RESENT (Figure 4 2; Previous size, LRT, 2=58.22, df=1, P<0.001; no effect of other factors tested, Appendix H: Table H 4). The overall mortality rates of Pocillopora were low across reefs in the BEFORE period (6% 12%, Table 43). When the data were constrained to examine the effects of S nigricans in the BEFORE 2=2.05, df=1, P=0.15) or any other factor, reinforcing that the experimental design was also valid for Pocillopora (Appendix H: Table H 4). Fragmentation I did not observe any fragmentation of Acropora on reefs with Stegastes nigricans experimentally REMOVED. The five Acropora that fragmented were all located on reefs with S. nigricans PRESENT and were of similar, relatively large size, with resulting daughter colonies also of simi lar size (Table 44). The initial cause was presumably predation, or some other stress, and fragmentation was maintained because turf overgrowth segregated the fragments (pers. obs).
92 No Pocillopora colonies fragmented during the BEFORE period across experimental reefs (Table 44), despite the considerably larger number of colonies monitored (Table 4 3). Following the S. nigricans removals, six times more individuals fragmented on removal (n=6) than control (n=1) reefs (Table 4 4). Change in coral size As with mortality, the experimental design was validated in that there was not a significant effect of Treatment on change in coral size when the data were constrained to examine only the BEFORE period for both corals (Appendix H: Table H 5). Change in size did vary as a function of Previous size for both Acropora (LLM, F1,42=6.81, P=0.01) and Pocillopora for this data subset (LLM, F1,106=9.33, P=0.003; Appendix H: Tables H 6 and H 7, respectively). There was a decline in change in size with an increase in previous size for both corals in the BEFORE period (Figures 43 and 44) and this may be related to the strong storms during that time. Larger individuals can also be more susceptible to fragmentation (Table 41) or mortality (Figures 4 1 and 4 2) and bot h fragmented and dead individuals were excluded from these analyses (for details please see Statistical Analyses ). Experimental change in Acropora size In Moorea today, Acropora generally have very few large representatives in the lagoon due to their slow recovery from earlier natural disturbances (Berumen and Pratchett 2006). For this relatively delicate branching coral, the change in sizedependent growth in the PRESENCE and REMOVAL of Stegastes nigricans diverged, as expected, between the BEFORE versus E XPERIMENTAL periods. That is, the Treatment by Period by Previous size interaction was significant for the full data set (LLM, F1,100=18.61, P<0.0001; Appendix H: Table H 5). In the EXPERIMENTAL period, the
93 trajectory of change in size as a function of previous size changed dramatically for both treatments for Acropora. In the REMOVED treatment, only a few large individuals were included in the analysis due to the high mortality rates (Figure 41). Those individuals that survived (and did not fragment Table 44) suffered considerable shrinkage (Figure 4 3). When I examined the change in size of Acropora using data from the PRESENT treatment in isolation, there was a significant Period by Previous Size interaction (LLM, F1,67=13.48, P=0.0005; Appendix H: Table H 6). This suggests growth rates were higher for Acropora within S. nigricans territories during the EXPERIMENTAL period (Figure 4 3). Although growth rates were lower in the before period, potentially due to a higher storm frequency, the impact of S. nigricans removal was substantially higher than the temporal variation in their presence (Figure 43). Experimental change in Pocillopora size Sample sizes of Pocillopora were more than twice that of Acropora thereby increasing confidence in growth estimates. Sizedependent growth varied according to experimental period (Period effect, LLM, F1,250=4.85, P=0.03) and previous size (LMM, F1,250=8.12, P=0.004). In this case, there was a slight decrease in change of size as a function of previous si ze across both treatments, and considerably higher variation in the EXPERIMENTAL period, versus the BEFORE period (Figure 44, Appendix H: Table H 5), despite an almost doubling of the sample size. When I examined growth rates for Pocillopora in the presence of S. nigricans (across both periods), there was a significant effect of both Previous Size (LMM, F1,162=11.00, P=0.001) and Period (LMM, F1,162=5.31, P=0.02). The small negative effect of previous size was consistent for both constrained data sets (BEFORE period and S. nigricans PRESENT, Appendix H: Table H -
94 7). This offers additional evidence that the design was valid. The significant effect of Period when S. nigricans was present (Appendix H: Table H 7), as well as the higher variation (despite higher sample sizes), does raise concerns about processes constraining growth during the EXPERIMENTAL period. Matrix Models Matrix models corroborated that among reefs, on average, the rat e of population Stegastes nigricans, and increased, for both Acropora and Pocillopora, when S. nigricans were removed (Figure 45). Further, there 95% C.I., see Table 46) be tween the two periods for reefs with Stegastes nigricans present, despite changes in abiotic (storm) and biotic ( Acanthaster planci presence) factors (For transition matrices see Appendix I: Figures I 2 to I 7). Recruitment Patterns Recruitment to experimental reefs was very low for Acropora regardless of Period or Treatment (Table 4 3). For Pocillopora recruitment was similarly low for both Treatments in the BEFORE period. In the EXPERIMENTAL period recruitment was approximately ten times higher than in the BEFORE period for both reef treatments. Discussion I compared growth and survival of two recovering branching coral genera ( Acropora and Pocillopora ) within S. nigricans territories versus reefs with S. nigricans removed, and compared differences i n key demographic rates to highlight factors contributing to their differential recovery. Survival and growth of Acropora were substantially higher inside S. nigricans territories across sizes. Mortality was low for small sizes of Acropora and zero for lar ger individuals (> 15 cm) in the presence of
95 Stegastes nigricans (Figure 4 1, Appendix H: Figure H 1). During the EXPERIMENTAL period, juvenile Acropora suffered ~3 times higher mortality and nearly all (12/13 or 92%) of the large corals died on reefs with S. nigricans REMOVED (Figure 4 1). Pocillopora also incurred less mortality in the presence of S. nigricans (Figure 4 2). For both corals, there were no differences between treatments in the BEFORE period, which confirms reefs did not differ prior to mani pulations (see Results, Appendix H). For Pocillopora, there were no differences between periods when I restricted the analyses to only reefs with S. nigricans present. However, there was evidence the growth trajectory varied between these periods for Acrop ora (Figure 4 3), potentially a consequence of storm damage during the observational (BEFORE) period. Overall, Acropora, and large individuals in particular, enjoyed reduced mortality, and higher growth rates, in the presence of S. nigricans, despite the preponderance of turf. both Acropora and Pocillopora S. nigricans (Figure 4 5). Observed Community Response In Moorea, Steg astes nigricans territories can occupy up to 45% of the back reef environment (Done et al. 1991; Gleason 1994). Following a series of disturbances in the 1980s on, disturbance tolerant massive Porites were less affected and became the dominant hermatypic c oral in this system (Bouchon 1985; Berumen and Pratchett 2006), as has been observed at other disturbed sites in the West Pacific (Guam: Colgan 1987) and Indian Ocean (McClanahan et al. 2007). In addition, Pocillopora, which was relatively rare in the syst em prior to these disturbances (Bouchon 1985), recovered better, and is now more abundant than the once dominant Acropora (Berumen and
96 Pratchett 2006). While Pocillopora is also more abundant within S. nigricans territories than outside (Glynn and Colgan 1 988; Done et al. 1991; Shima et al. 2008), the pattern is not nearly as striking as for Acropora, which is rarely found today outside S. nigricans territories in the backreef, except in areas of particularly high water flow (Galzin et al. In preparation), or in the few existing aborescent Acropora thickets, which are colonized by both S. lividus and S. nigricans (Mapstone et al. 2007). This differential recovery is likely due to either (1) variation in post settlement mortality, or (2) differential recruitm ent processes. The first is plausible because Pocillopora is protected by tightly branching morphology and a relatively dense skeleton, and was observed to be less affected by fish predation (Chapter 2) than Acropora, which has a more fragile, spongy skeleton (Veron 2000). Similarly, recruitment processes could contribute because the life history strategies of these two genera differ substantially. Acropora is a broadcast spawner, releasing relatively small ova in hermaphroditic bundles, development takes a couple of weeks and cannot be extended as larvae do not feed and lack zooxanthellae (e.g., Hughes et al. 2002). In contrast, Pocillopora broods to the planula stage, releasing large larvae that are competent to settle, but that also can delay settlement for months because they harbor zooxanthellae, and are often asexually produced (e.g., Stoddart 1983; Hughes et al. 2002). While I did observe higher recruitment across experimental reef types for this taxon in the EXPERIMENTAL period (Table 45), recruitm ent processes are notoriously variable and longer term measurements are required to make any inference. Differential Susceptibility to Predation It is clear that, of the common genera in Moorea, Acropora is the most susceptible to natural disturbances (Table 4 1). It appears they are also more susceptible to
97 predators, given they experience a drastic loss when S. nigricans and their associated territorial defense, is removed (Figure 41). Within this system, the main coral predators for Acropora include polyp feeding butterflyfishes (Chaetodontidae), as well as skeletal excavators, including triggerfishes (Balistidae) and pufferfishes (Tetradontidae) (pers. obs.). S. nigricans reduces foraging rates of herbivores, including surgeonfish (Acanthuridae) and parrotfish (Scaridae), as well as butterflyfishes (Chapter 2). However, the corallivorous fishes that can inflict the damage observed (triggerfishes and pufferfishes) are less common and considerably more shy than butterflyfishes and thus their impact on cor als in damselfish territories is difficult to quantify. Nevertheless, I have observed S. nigricans successfully defend against these coral predators (e.g., Arothron meleagris and A. hispidus (Tetradontidae); Balistipus undulatus and Melicthys vidua (Balist idae)), though several bites may be taken before the intrusion is successfully thwarted. Thus, fish predation on coral can occur within territories, but at a considerably reduced rate (See Appendix H for representative photos of predation observed during each period: Figure H 3). Pocillopora, on the other hand, is mostly fed upon by butterflyfishes, which do not inflict skeletal damage (reviewed in Cole et al. 2008; Rotjan and Lewis 2008). Gleason (1994) reported skeletal loss due to fish predation for juvenile Pocillopora. However, when corallivores were presented with Acropora, Montipora, Pocillopora, and Porites simultaneously on reefs with S. nigricans removed or naturally absent, only the delicate acroporids incurred observable (i.e. skeletal) damage ( Chapter 2). This lends further credence that Pocillopora is less susceptible than Acropora to predation. For both taxa, coral spat and small juveniles can also incur predation via nondiscriminate foraging by
98 herbivores (e.g., parrotfishes and surgeonfishes: Randall 1974, Bak and Engel 1979). As these fish are also excluded by S. nigricans (Chapter 2), this type of predation is also expected to be lower within territories. Acropora and Pocillopora are also vulnerable to predation by the corallivorous crown of thorns seastar, Acanthaster planci although P. eydouxi, the most robust Pocillopora species, is effectively protected by large crustacean exosymbionts (C.S. McKeon, pers. comm.), while other Pocillopora are protected less effectively (Glynn 1985). H owever, P. eydouxi was not included in analyses due to their rarity in the lagoon. During 2008, an Acanthaster outbreak was underway in Moorea and we observed several individuals within the lagoon near experimental reefs. Previous research suggests that S. nigricans can defend against Acanthaster (Glynn and Colgan 1988) ; however, I did observe feeding by Acanthaster on some reefs with S. nigricans Further, the type of partial mortality incurred by Pocillopora on experimental reefs was consistent with Acant haster feeding mode (i.e., no skeleton loss, oval death zone of appropriate size; See Appendix H for representative photos of predation observed during each period: Figure H 4). Both Acropora and Pocillopora are preferred food for Acanthaster (e.g, Saipan : Goreau et al. 1972; Panama: Glynn et al. 1972, Porter 1972; and Hawaii: Chess et al. 1972), and preference appears to increase with prey availability (e.g., see Porter 1972 and Goreau et al. 1972). Pocillopora is more than twice as abundant as Pocillopor a (Tables 42 and 23) and attains larger sizes within territories (Figures 43 and 44). The type of predation (Appendix H: Figure H 4), higher mortality, and variation in negative growth experienced by Pocillopora during the experimental period is consis tent with
99 predation by Acanthaster. Whereas, both small and especially larger size classes of Acropora were protected by S. nigricans defense consistently across both periods (Figure 4 1). The differential predation pressure by Acanthaster may have been dr iven by differences in prey availability (i.e., Pocillopora were larger and more abundant, thus were targeted more frequently). If Acanthaster had not been present in the system, we may not have seen the same degree of mortality on large Pocillopora. Recru itment Failure Acropora experienced very low levels of both asexual (Fragmentation, Table 42) and sexual (Recruitment, Table 43) recruitment to experimental reefs. In combination with the observed distribution of only small and denuded Acropora outside o f territories (unpubl. data), this suggests Acropora are highly susceptible to predation, and associated population declines (Figure 45), in the absence of S. nigricans defense. For Pocillopora, both fragmentation and recruitment were very low in the bef ore period (Table 42 and 4 3). During the experimental period, fragmentation (via partial predation) increased substantially on reefs in which S. nigricans were removed, but not where present. Interestingly, rates of ambient recruitment to both reefs incr eased dramatically for Pocillopora in the experimental period, suggesting larval supply was likely higher (given the poor survival of larger size classes on removal reefs, a reduction of post settlement mortality is not as likely of an explanation). Given Pocillopora brood, reduced planktonic dispersal may also increase the probability those few individuals that do survive may better contribute to the next generation (Knowlton 2001). However, long term recruitment studies are necessary before any inference can be made. It seems Pocillopora has the advantage over Acropora of a stronger skeleton that minimizes susceptibility to skeleton damaging predation, though their susceptibility to
100 predation by Acanthaster can be high without S. nigricans defense (Figure 4 2). In Moorea, the intense predation pressure is likely the most important selective pressure, as persistence of large colonies is expected to have the biggest impact on population dynamics for long lived, sessile species (Tanner and Hughes 2000) and future research is warranted to tease apart the possible drivers of this high predation pressure. Conclusions A central goal in ecology is to understand the mechanisms underlying the distribution and abundance of organisms well enough to predict populationlevel effects of perturbations. The use of manipulative removals of a key engineer can identify the mechanisms underlying documented community changes and provide a solid foundation for understanding impacts within a community context (e.g., Paine 1969). W e can move toward a more predictive framework for nontrophic interactions (e.g., ecosystem engineers, mutualisms, etc.) by merging the interaction strength framework commonly used in consumer resource interactions (e.g., Paine 1992; Wootton 1997; Berlow et al 2005) and populationlevel measurements obtained using a BeforeAfter Control Impact design (Osenberg et al. 2006) within a Life table Response Experiment framework (Caswell 2001). By comparing corals that vary in both disturbancesensitivity and rec overy rates, it is possible to isolate factors that constrain community resilience. This mechanistic understanding facilitates explorations of what processes are driving the observed changes in community structure (Chapter 5).
101 Table 41. General suscepti bility of Acropora, Pocillopora, and Porites to natural disturbances. Natural Disturbance Documented (Relative) Susceptibility References Acropora Pocillopora Porites Storm damage High Intermediate Low 1 Bleaching events High Intermediate Low 2 Aca nthaster planci predation High Intermediate High* Low 3, 4 Destructive predation (fishes) High Intermediate Low 5 Overgrowth by algal turf Intermediate Low High 5 Notes: Natural disturbance is extended to include predation and competition with S tegast es nigricans associated algal turf. References are not exhaustive but exemplary: 1Hughes 1989; 2Gleason 1993, 3Porter 1972, 4Goreau et al. 1972, 5White and ODonnell In press. *This generalization excludes Pocillopora eyudouxi and depends upon whether colo nies have large crustacean guards present (e.g., Glynn 1985, Pratchett 2001); In addition, preference appears to increase with prey availability (e.g., see Porter 1972 and Goreau et al. 1972).
102 Table 42. Overall mortality patterns for all sizes of Acrop ora. Before Experimental Stegastes nigricans NMortality NTotal Rate NMortality NTotal Rate PRESENT 5 33 0.15 11 64 0.17 REMOVED 2 31 0.07 52 69 0.75 Notes: Measurements were repeated Before manipulations (20062007) and during the Experimental period (20072008) on reefs with Stegastes nigricans Present and those from which S. nigricans were Removed. Please note that no S. nigricans were removed in the BEFORE period and, in this context, Removed refers to reefs on which S. nigricans were sc heduled to be removed during the EXPERIMENTAL period.
103 Table 43. Overall mortality patterns for all sizes of Pocillopora. Before Experimental Stegastes nigricans NMortality NTotal Rate NMortality NTotal Rate PRESENT 4 71 0.06 23 137 0.17 REMOVED 8 68 0.12 68 116 0.59 Notes: Measurements were repeated Before manipulations (20062007) and during the Experimental period (20072008) on reefs with Stegastes nigricans Present and those from which S. nigricans were Removed (as in Table 42).
104 Tabl e 44. Record of fragmentation of corals by period and treatment Coral Period Stegastes nigricans Num. of Ind. Fragmented Mean Size of Ind. (cm) Mean Num of Fragments (Total) Mean Frag. Size Acropora Before Present 0 NA (0) NA Removed 2 14.05 4.1 8 3.0 0.0 (6) 2.63 2.45 Experimental Present 3 13.40 7.34 4.3 7.9 (13) 2.65 3.12 Removed 0 NA (0) NA Pocillopora Before Present 0 NA (0) NA Removed 0 NA (0) NA Experimental Present 1 12.30 0.00 2.0 0.0 (2) 5.30 4.16 Remove d 6 14.05 26.93 2.2 0.8 (13) 3.66 4.30 Notes: The number and mean size of Acropora or Pocillopora individuals that fragmented by Period and Stegastes nigricans treatment. In addition, I report the mean number of fragments (as well as the total number of fragments) and the mean size of fragments. For all mean values, 95% confidence intervals are reported as 1.96 standard error. Please note that no S. nigricans were removed in the BEFORE period and, in this context, Removed refers to reefs on which S. nigricans were scheduled to be removed during the EXPERIMENTAL period.
105 Table 45. Record of coral recruitment to reefs by period and treatment Coral Period Stegastes nigricans Num. of Recruits Mean Size of Recruits (cm) Acropora Before Presen t 3 2.90 0.36 Removed 0 NA Experimental Present 4 3.48 1.0 Removed 2 4.30 0.24 Pocillopora Before Present 2 1.90 0.03 Removed 0 NA Experimental Present 30 2.24 0.36 Removed 25 2.09 0.36 Notes: The total number, and mean s ize, of Acropora or Pocillopora individuals that recruited to experimental reefs by Period and Stegastes nigricans treatment. For all mean values, 95% confidence intervals are reported as 1.96 standard error. Please note that no S. nigricans were r emoved in the BEFORE period and, in this context, Removed refers to reefs on which S. nigricans were scheduled to be removed during the EXPERIMENTAL period.
106 Table 4Pocillopora and Acropora. Pocillopor a Acropora Stegastes nigri cans Mean min max Mean min max Before (Present) 0.96 0.89 0.99 1.03 0.86 1.20 Present 0.98 0.81 1.07 0.84 0.71 0.99 Removed 0.59 0.43 0.75 0.35 0.07 0.51 Notes: Matrice analyses were performed on all colonies for all reefs Before manipulations (2 0062007, i.e., with Stegastes nigricans Present ) and during the Experimental period (20072008) on reefs with Stegastes nigricans Present and those from which S. nigricans were Removed (as in Table 42).
107 Figure 41. Mortality rates of Acropora. M ean mortality probability as a function of previous maximum linear length (cm) during the Before (20062007) and Experimental (20072008) periods, between which clove oil was applied to reefs and the Dusky farmerfish, Stegastes nigricans was allowed t o recover ( Present, purple) or Removed (gray) from half of each pair of experimental reefs. Points are jittered to reduce overlap and represent mortality, (=1) or survival (=0). The effect of S. nigricans Treatment varied by experimental Period for Acropora with mortality rates considerably higher on reefs where S. nigricans were REMOVED (i.e., there was a significant Treatment by Period 2=11.51, df=1, P =0.0007). In particular, the vulnerability of larger Acropora varied with the presence of S. nigricans (Treatment by 2=4.74, df =1, P =0.03; Appendix H: Tables H 1 & H 2).
108 Figure 42. Mortality rates of Pocillopora Mean mortality probability as a function of previous maximum linear length (cm) during the Before (20062007) and Experimental (20072008) periods, between which clove oil was applied to reefs and the Dusky Farmerfish, Stegastes nigricans, was allowed to recover (Present, blue) or Removed (gray) from half of each pair of experimental reefs. Points are jittered to reduce overlap and represent mortality, (=1) or survival (=0). The effects of S. nigricans on mortality varied between the BEFORE and EXPERIMENTAL periods, with higher mortality on reefs where S. nigricans were REMOVED 2=17.54, df=1, P <0.0001; Appendix H: T ables H 1 and H 2).
109 Figure 43. Change in size (Delta) of Acropora as a function of previous maximum linear length (cm) during the Before ( 2007Size2006) and Experimental ( 2008Size2007) periods, between which clove oil was applied to ree fs and the Dusky Farmerfish Stegastes nigricans was allowed to recover (Present, purple) or Removed (gray) from half of each pair of experimental reefs. For Acropora, the change in sizedependent growth (of survivors) in the PRESENCE and REMOVAL of S nigricans diverged, as expected, between the BEFORE versus EXPERIMENTAL periods, with considerably lower growth (higher shrinkage) when S. nigricans was REMOVED (i.e., the Treatment by Period by Previous size interaction was significant for the ful l data set; LLM, F1,100=18.61, P<0.0001; Appendix H: Table H 5)
110 Figure 44. Change in size (Delta) of Pocillopora as a function of previous maximum linear length (cm) during the Before ( 2007Size2006) and Experimental ( 2008Size2007) peri ods, between which clove oil was applied to reefs and the Dusky Farmerfish, Stegastes nigricans was allowed to recover (Present, blue) or Removed (gray) from half of each pair of experimental reefs. For Pocillopora, size dependent growth (of survivors ) varied according to experimental period (Period effect, LLM, F1,250=4.85, P=0.03) and previous size (LMM, F1,250=8.12, P=0.004; Appendix H: Table H 5).
111 Figure 45. Mean population decay rate of Acropora and Pocillopora before manipulations (all reefs combined), as well as in the presence (control) and removal of the farmerfish, Stegastes nigricans.
CHAPTER 5 CONCLUSIONS Coral reefs are declining worldwide due to natural and anthropogenic disturbances (B ellwood et al. 2004, HoeghGuldberg et al. 2007). Stressors include terrigenous runoff, nutrient pollution, overextraction by humans, and higher abiotic stress induced by global change (Knowlton and Jackson 2008). Resilience and recovery rates vary, and ec ologists seek to understand the mechanisms influencing these processes. High cover of macroalgae is an especially important proximate cause of coral decline, as it can inhibit coral recruitment and survival (reviewed in McCook et al. 2001). Thus, herbivorous fishes and urchins (which can suppress algal biomass via grazing) have long carried a reputation for indirectly maintaining the abundance and diversity of reef building corals (Hixon 1997, Hughes et al. 2007). Although herbivory is well accepted as a do minant process structuring coral reefs, it is important to differentiate two broad behavioral groups of her bivorous coral reef fishes: mobile foragers versus siteattached farmerfish (Ceccarelli et al. 2005) (Table 51). Foragers are generally larger, herbivorous fishes that forage typically in loose singleor mixedspecies schools over relatively large home ranges (e.g., parrotfishes (Scaridae), surgeonfishes (Acanthuridae), and rabbitfishes (Siganidae), Ceccarelli et al. 2005). Farmerfish are smaller, highly territorial damselfish that cultivate preferred turf algae via active farming and defense behaviors in restricted areas (Ceccarelli et al. 2001, 2005). Farmerfish and foragers overlap in distribution but impact benthic communities in fundamentally different ways (Ceccarelli et al. 2005). While all territorial damselfish occupy defined areas which they actively defend from competitors and egg predators (e.g., Ebersole 1977), only territorial algal farmers engineer the benthic structure of
coral reefs (reviewed in Ceccarelli et al. 2001). These farmerfish develop and maintain conspicuous filamentous algal mats (i.e., turf) through a variety of behavioral adaptations, including substratum preparation (killing coral and removing sediment), weeding (selectively removing unpalatable algae), fertilizing (through defecation), and reducing herbivory through active defense (Ceccarelli et al. 2001). The benthic algal biomass (Hixon and Brostoff 1981, Hata and Nishihiri 2002, Hata et al. 2002), diversity (Hixon and Brostoff 1983, 1996; Sammarco 1983), and productivity (Klumpp et al. 1987, Lison de Loma and Ballesteros 2002) of filamentous algal turfs are higher within farmerfish territories than in nondefended areas. With farmerfish removals, foraging rat es of mobile foragers increase in these gardens and turf cover declines rapidly (Ceccarelli et al. 2001, Chapter 2), suggesting that territorial behavior is a key determinant of the efficacy of benthic engineering. In contrast, experimental removals of mobile foragers effectively enhances the abundance of fleshy and filamentous algae (e.g., Lewis 1986, Hughes et al. 2007). Further, different species of foragers appear to target specific taxa of algae (Mantyka and Bellwood 2007), and mixed assemblages are mo re effective at maintaining low algal cover and promoting coral (Burkepile and Hay 2008). Variable Outcomes of Farming Behaviors In some Pacific reefs, there is a paradoxical pattern of higher recovery of branching corals within algal farms of territorial damselfish (Panama: Wellington 1982; Japan: Suefuji and van Woesik; Great Barrier Reef (GBR): Sammarco and Carleton 1981, Sammarco and Williams 1982; Moorea: French Polynesia: Done et al. 1991, Gleason 1994; Tahiti, French Polynesia: Glynn and Colgan 1988) Generally, when farmerfish inhabit arborescent branching corals of the genus Acropora, the corals suffer
from farming behaviors (e.g., biting coral tips to increase territory size) and associated increased algal abundance (e.g., Kaufman 1977, Potts 1977, Knowlton et al.1988). After a series of disturbances in Moorea, French Polynesia, the mortality risk for branching Acropora and Pocillopora corals (via direct or incidental predation) was higher outside territories than the mortality risk associated with algal overgrowth within territories (Chapters 2 and 4). This pattern was also observed of Acropora in Okinawa, Japan, following a bleaching event (Suefuji and van Woesik 2001). In Panama, coral zonation, was reportedly controlled by another farmerfish ( Eup omacentrus acapulcoensis ) via protection of Pocillopora ( Wellington 1982). This indirect positive facilitation is opposite to the typical (negative) interaction, in which live coral is reduced (and algae are cultivated) (Ceccarelli et al. 2001). Indeed, in these same systems, there was a detrimental impact on the other dominant hermatypic coral ( Pavona in Panama: Wellington 1982; Porites in Tahiti: Glynn and Colgan 1988, and Moorea: Done et al. 1991, Chapter 3; however there was no mention of other taxa in the short Reef Site from Okinawa; Suefuji and van Woesik 2001). These patterns suggest that changes in community structure, due to natural or anthropogenic disturbance, may change the strength and direction of farmerfish engineering impacts. Further, species may respond differently, due to variation in population dynamics and life history traits. Farmerfish Effect Varies with Community Structure Roughly half of all herbivorous damselfish farm algal turf (tallied from Ceccarelli et al. 2001). These farmerf ish function as ecosystem engineers (Jones et al. 1994), although their roles apparently shift depending on community structure. In areas with high coral cover and diversity (i.e., low disturbance sites), farmerfish are a nuisance community member for the fast growing arborescent Acropora because they reduce live
coral (e.g., Kaufman 1977, Potts 1977, Knowlton et al. 1988, Lison de Loma and Ballesteros 2002). This could benefit other corals, in that their competitors are adversely affected. In some disturbed communities, reefs have shifted from acroporiddominated (Acropora and Montipora) to a community structure dominated by disturbancetolerant, or fast recovering weedy, coral species (e.g., Indian Ocean: McClanahan et al. 2007; Moorea: Berumen and Pratc hett 2006), due to differential susceptibility (Huston 1985), thereby shifting the size and direction of the farmerfish effect. Colonization by farmerfish negatively affects large corals that are amenable to turf farming, e.g., Pavona and Porites, by reduc ing cover of live coral (Wellington 1982, Done et al. 1991, Chapter 3). However, some branching coral taxa may be necessary because they offer important habitat structure for farmerfish, though this has only been tested explicitly in Panama (Wellington 198 2). The increased abundance of branching corals in these systems is presumably due to a higher grazing rate outside territories (Panama: Pocillopora is protected from pufferfishes, Wellington 1982; Okinawa: Acropora is protected from urchin grazing, Suefuj i and van Woesik 2001; Moorea, Acropora (Chapter 2 and 4) is protected from corallivorous fishes, while Pocillopora is also protected from Acanthaster planci (Chapter 4). Expanding the Grazing Continuum Recent studies across archipelagos indicate that changes in benthic reef community structure can be clearly associated with changes in the composition and trophic structure of coral reef fishes in both the Caribbean (Mora 2008) and the Pacific (Friedlander and DeMartini 2002, Dulvy et al. 2004, DeMartini et al 2008, Sandin et al. 2008). Coral health is lower with lower biomass of top predators and corresponding higher biomass of herbivorous fishes (Sandin et al. 2008). Similarly, the biomass of
farmerfish is higher in fished islands (DeMartini et al. 2008) and is negatively correlated with predator biomass across U.S. reefs (Zgliczynski and Sandin, unpublished data). In addition to cultivating algae, farmerfish effectively deter territory use by urchins (Sammarco and Williams 1982, Mapstone et al. 2007) and large herbivorous and corallivorous fishes (Ebersole 1977, Wellington 1982, Chapter 2). Thus, farmerfish indirectly alter the magnitude and frequency of interactions for other community members. For example, corals must contend with higher abundances of algal turf, but enjoy reduced interactions with foraging herbivores and corallivores. Thus, largescale changes in fish community structure may predictably alter the magnitude and direction of farmerfish engineering (i.e., the outcome of their engineering beha viors) on benthic composition. That is, in the absence of high predation pressures, the high algal abundance is expected to have a net negative effect; whereas, in the presence of high corallivory or incidental predation on coral spat, turf tolerant taxa m ay benefit from the indirect protection of territorial defense (i.e., a net positive effect). The observed differences in fish community structure highlights that different processes can structure benthic communities. A recent metaanalysis of 54 field ex periments suggests grazers are the key determinant of algal abundance in nutrient poor tropical systems and that the effects of nutrient enrichment are enhanced with losses of herbivores (Burkepile and Hay 2006). Given variation in time since beginning of fishing, grazing as a driver of community structure likely exists along a longer continuum than previously considered. For example, in the Caribbean, marine protected areas enhance the biomass of parrotfishes and increase grazing rates relative to unprotec ted areas (Mumby et al. 2006) thereby indirectly facilitating higher coral
recruitment (Mumby et al. 2007). This is a classic example of an essentially grazer limited system, where ecologists expect herbivores to maintain coral dominance by reducing algal biomass (e.g., Hay 1981; Hughes 1994; McCook 1996, 1999; Hughes et al. 2007). Grazing by reef fishes is largely unaccepted as a process limiting coral recovery, despite early reports (Darwin 1842, p. 14) and subsequent documented changes in reef fish trophic structure (i.e., removal of piscivorous predators) that would favor a prey release of coral predators and herbivores (e.g., Sandin et al. 2008). Alternatively, high disturbance frequency may augment algal turf food resources (e.g., bleaching events, Gl eason 1993), thereby supplementing herbivorous populations (Hixon 1997) or facilitating differences in diet for facultative corallivores (e.g., Guzman and Robertson 1989). However, there is evidence mounting that coral predation can play an important regul atory role for coral community structure within Caribbean reefs, including Belize (Littler et al. 1989) and the Florida Keys (Miller and Hay 1998); as well as central Pacific coral reef communities: Hawaii (Frydl 1979, Jawardene et al. 2009), Panama (Welli ngton 1982), Moorea, French Polynesia (White and ODonnell, in press), and outer zones of the Great Barrier Reef (Hoey and Bellwood 2008). Given farmerfish can effectively exclude most other grazing fishes (Brawley and Adey 1977, Wellington 1982, Letourneur 2000, Chapter 2), increases in their abundance will further concentrate foraging outside of territories (Steneck and Sala 2005). Evidence Based Upon Research in Moorea Darwin (1842) reports second hand (citing Mr Couthouy, p. 128131) that at Tahiti and Moorea (Eimeo) the space between reef and the shore has been nearly filled up by the extension of those coral reefs (madrepores with tops of branches exposed = Acropora), which within most barrier reefs merely fringe the land. Indeed, later reports
by Cros sland (1928) corroborates that, historically, the most abundant reef builders in Tahiti and Moorea were corymbose Acropora, branching species of Pocillopora, and Porites (though he notes Porites were less important than conspicuous, P. 722) As late as 1979, Bouchon (1985) reports both Acropora and Porites dominated the reef flat within Moorea lagoons, whereas, the fringing reef was already dominated by only Porites. Since that time, coral reefs in French Polynesia suffered a series of natural disturbance s since the 1980s, including cyclones, bleaching events, and crownof thorn seastar, Acanthaster planci outbreaks (Berumen and Pratchett 2006). Within these lagoons, the recovery of corals varies with the presence of a common territorial farmerfish, Stega stes nigricans. There is lower cover of disturbance tolerant massive corals (Chapter 3), but higher abundances of recovering branching corals within territories (Gleason 1994,1996; Done et al. 1991). Recent tests confirmed that corals respond differently t o S. nigricans presence: massive Porites experienced higher mortality due to competition with turf, while branching Acropora and Montipora survived and grew better in the presence of farmerfish due to corallivore and grazer exclusion (Chapter 2). Impacts will likely vary among life stages depending on vulnerability to competition or predation / grazing impact, e.g., larger corals generally grow more slowly (depending upon how you measure growth) and incur less mortality than smaller corals (Hughes 1990). I ndeed, within given taxa different size classes can vary in their response. For example, previous research suggests recruitment of branching Pocillopora is higher within territories (Gleason 1994, 1996). However, a recent test found no evidence of a farmer fish effect for the juvenile size class (Chapter 2). In addition, important
demographic rates may vary spatiotemporally, due to spatial or temporal differences in physical (e.g., temperature, flow) or biological factors (e.g., corallivory or algal growth r ates). For example, in a test of the rate of turf overgrowth for massive Porites the rate varied dramatically within a given colony (~ 0 vs. 5 cm overgrowth in a six week period) (Chapter 2, Figure 51). Overgrowth was lowest with manual sediment removal, and highest under normal turf associated sediment loads, i.e., in the absence of vermetid snails, ( Dendropoma spp.), which cast a mucus web and inadvertently remove sediment from turf. This suggests that sediment is an important driver and that these inve rtebrates may contribute to spatiotemporal variation in this rate of overgrowth by indirectly reducing mortality associated with algal competition (Chapter 2). Clearly, the snapshot approach of measuring interactions in one time and place gives us little understanding of the true per capita effects on corals. To scaleup these previously isolated mechanisms to the population level, I combined demographic field surveys, experimental removals, and mathematical modeling. This isolated the processes underlying higher growth and survival of two recovering branching coral genera ( Acropora and Pocillopora) within S. nigricans territories relative to outside. As Pocillopora has recovered better than once dominant Acropora (Berumen and Pratchett 2006), comparing dif ferences in key demographic rates may highlight factors limiting recovery. Indeed, both Pocillopora and Acropora suffered significantly higher mortality after the removal of Stegastes nigricans (Chapter 4). However, the effect was stronger for Acropora, and only this taxa also suffered lower growth after removals (Chapter 4). Whereas, Pocillopora suffered high mortality after removal of farmerfish, but also suffered lower growth during the experimental period in the presence of farmerfish, and
it is likely both of these effects were driven by the presence of the crownof thorn seastar (COTS) in the system, rather than corallivorous fishes. For example, when nubbins of Pocillopora verrucosa and the more delicate acroporids (Acropora striata and Montipora flow erii ) were presented to corallivorous fishes on the same experimental reefs, only the delicate acroporids suffered a reduction in skeletal mass due to predation (Chapter 2, in 2005, prior to the COTS outbreak). This higher mortality and shrinkage suggests the delicate skeletal structure is constraining recovery of Acropora in this high grazing environment (Figure 51), while Pocillopora, with a dense lobed skeletal structure, can tolerate the high grazing pressure (or is not preferred by corallivores) and h as a higher current distribution relative to historic records (Berumen and Pratchett 2006). By comparing coral species that vary in disturbancesensitivity and recovery rates in the presence and removal of the territorial farmerfish S. nigricans I identif ied that high grazing pressure on coral colonies may be constraining community resilience. Thus, taxa with a higher resistance to competition with turf can utilize dead substrata made available by farmerfish gardening. This increase in substrate availability, when coupled with lower mortality rates, has led to enhanced recovery of branching corals within farmerfish territories. Within the lagoon, in areas outside of farmerfish territories, branching corals are relegated to high flow, or cracks and crevices, suggesting grazing pressure is constraining recovery in the absence of this ecosystem engineer. This system supports the idea that disturbance can alter the engineering role. S. nigricans adversely affects branching corals in relatively undisturbed habitats, but can offer a less stressful environment when there is a shift in the community structure of grazing populations. Of course, this pattern is not expected
for all farmerfish species, as they may vary in their engineering impact. Good predictors of engineering capacity (or magnitude of effect sizes) are: 1) large size and aggressive territorial defense, 2) communal or shared defense, and 3) resulting large territory sizes. Ideally, future research will take the ecological history into account and advoca te a precautionary, adaptive approach to management (e.g., Doak et al. 2008).
Table 51. Summary of differences between foraging and farming coral reef fishes. Character Foragers Farmers Home range Varies; 10 to 100s m square 1,2 Small; 0 3 m square 1,2 Diet Variable; Functional Redundancy 1,3,4 Filamentous reds 2 Size Larger (Post Juvenile = 15 70 cm) 1,4 Smaller ( Post Juvenile: 5 30 cm) 1,4 Reproduction Broadcast Spawners 1,4 Demersal Egg Layers 1,4 Lifespan Long? (20 40 years) 1,4 Long (~20 years) 1,4 Dominant Families Acanthuridae (Surgeonfishes), Scaridae (Parrotfishes), & Siganidae (Rabbitfishes) Pomacentridae (Damselfishes) References: 1Randall 2005, 2Ceccarelli et al. 2001, 3Burkepile and Hay 2008, 4www.fishbase.org
Figure 51. Schematic of Stegastes nigricans direct and indirect interactions. Direct (solid arrows) and indirect (dashed arrows) effects of the dusky farmerfish ( Stegastes nigricans ) documented on the north shore of Moorea, French Polynesia (Chapters 2, 3, and 4). The size of the arrow indicates the relative magnitude of the response: Net effects of farmerfish range from negative (grey) to positive (black) for massive Porites and branching Acropora coral, respectively, which is opposite their relative susceptibility to natural di sturbances, such as predation, bleaching events, storm damage, or Crownof Thorn Seastar (COTS) break outs.
APPENDIX A SAMPLING METHODS AND RESULTS CONCERNING E XPERIMENTAL REEF CHARACTERISTICS PRIO R TO AND AFTER MANIP ULATIONS Methods I identified and marked 32 reefs colonized by Stegastes nigricans ( Figure A 2 ) and 8 interspersed reefs naturally devoid of S. nigricans and associated turf (i.e., healthy Porites reefs, Figure A 2). Prior to any manipulations, I measured the density of S. nigricans and phys ical characteristics of each reef, including: footprint (maximum length and perpendicular width, cm), average and maximum heights (cm), water depth (cm), distance to the three nearest neighbors (cm), and percent composition (using fixed point contact, at 1 0 cm intervals, of 3 transects perpendicular to the longest reef axis). For reef composition, corals living upon the reefs belonging to Acropora Montipora, or Pocillopora were combined into one category (Branching corals) Similarly, I combined dead coral and crustose coralline algae into one category. Macroalgae were in very low abundance on any reef type and I combined the following genera: Dictyota, Halimeda, Turbinaria, and unidentified cyanobacteria. Results Table A 1. Physical f actors of a ll reefs be fore m anipulations (GLM) Physical Reef Characteristic Variable F value df Pr > F Reef size (length x width) Treatment 0.49 4, 39 0.74 Depth to sand Treatment 1.94 4, 39 0.13 Average Height Treatment 1.19 4, 39 0.33 Maximum Height Treatment 1.24 4, 39 0 .31 Reef isolation (avg. dist. to 3 nearest reefs) Treatment 2.38 4, 39 0.07 Notes: Differences in average physical characteristics among experimental reefs prior to manipulations (PROC GLM, SAS 9.13). Mean values ( 95%CI) are reported in the Experiment al Reefs section.
Table A 2. Composition of Stegastes Reefs before removals (GLM) Benthic Category Variable F value df Pr > F Algal Turf Treatment (before) 1.03 3, 28 0.39 Porites spp. Treatment (before) 1.60 3, 28 0.21 Branching Corals Treatment (bef ore) 0.17 3, 28 0.92 DeadCoral/ CrustoseCorallineAlgae Treatment (before) 0.89 3, 28 0.46 Macroalgae Treatment (before) 0.77 3, 28 0.52 Notes: Differences in average (arc sine square root) benthic composition among Stegastes nigricans reefs prior to manipulations (PROC GLM, SAS 9.13). Benthic categories were arcsine squareroot transformed prior to analyses. The positive control reefs (i.e., those naturally devoid of Stegastes and Turf = Porites reefs) were not included as I expected differences due to the higher coral cover and absence of turf (Appendix A: Table A 3 and Figure A 2 ).
Table A 3. Comparisons of Stegastes and Porites r eefs before removals (t tests) Benthic Category Variable t df Pr > t Algal Turf Reef Type 9.76 38 <0.0001 Porites spp. Re ef Type 15.63 38 <0.0001 Branching Corals Reef Type 3.21 38 <0.003 DeadCoral/ CrustoseCorallineAlgae Reef Type 1.27 38 0.21 Macroalgae Reef Type 0.51 38 0.61 Notes: Differences in average (arc sine square root) benthic composition bet I en experimental a nd control reefs prior to manipulations (PROC TTEST, SAS 9.13) (Figure A 4) Reef type refers to reefs colonized by Stegastes nigricans (prior to any manipulations) versus control reefs (uncolonized by S. nigricans = Porites reefs). Benthic categories were arcsine square root transformed. Sample sizes are not equal, but assumptions of normality and equal variance were met.
Table A 4. Composition of Stegastes Reefs after removals (GLM) Benthic Category Variable F value df Pr > F Algal Turf Treatment (afte r) 10.25 3, 28 0.0001 Porites spp. Treatment (after) 1.61 3, 28 0.21 Branching Corals Treatment (after) 0.47 3, 28 0.71 DeadCoral/ CrustoseCorallineAlgae Treatment (after) 40.25 3, 28 <0.0001 Macroalgae Treatment (after) 3.17 3, 28 0.04 Notes: Differe nces in average (arc sine square root) benthic composition among Stegastes nigricans reefs after experimental manipulations (PROC GLM, SAS 9.13). Benthic categories were arcsine squareroot transformed prior to analyses. The positive control reefs (i.e., those naturally devoid of Stegastes and Turf = Porites reefs) were not compared in this analysis as I was testing the effects of the removal treatments and initial differences were already established (Table A 3 and Figures A 4 and A 5 ).
Figure A 1. Des ign schematic. A ) Factorial manipulation of Stegastes nigricans and algal turf presence and removal on 32 colonized reefs. Naturally absent Porites reefs served as a positive control (N=8 for all treatments). B) For each of the 40 experimental reefs, we outplanted one of each subtreatment (cage, cage control, and open) for each species ( Acropora striata, Montipora floweri, Pocillopora verrucosa, and Porites australiensis ) yielding 480 experimental nubbins total.
Figure A 2 Photographic example of exp erimental reefs. A) Stegastes reef (N=32) B) Porites reef (N=8). Nubbins were outplanted randomly by species and subtreatment (orange flags).
Figure A 3 Photographic example of experimental coral nubbins. Each square represents the species x su b treatments combinations that were outplanted to EACH experimental reef type (N=12 nubbins / reef).
Before Manipulations 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 + Stegastes/ + Turf + Stegastes/ Turf Stegastes/ + Turf Stegastes/ Turf 0 Stegastes/ 0 Turf Treatment Mean % Reef Composition Turf Porites Branching Corals Dead Coral and CCA Macroalgae Figure A 4 Mean reef composition before removals. Mean percent benthic cover of reefs ( 95% CI) by treatment prior to experimental manipulations (see Appendix A: Tables A2 & A3).
After Manipulations 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 + Stegastes/ + Turf + Stegastes/ Turf Stegastes/ + Turf Stegastes/ Turf 0 Stegastes/ 0 Turf Treatment Mean % Reef Composition Turf Porites Branching Corals Dead Coral and CCA Macroalgae Figure A 5 Mean reef composition after removals. Mean percent benthic cover of experimental and control reefs ( 95% CI) one month after experimental manipulations (see Appendix A: Table A 4 )
APPENDIX B TABLES AND FIGU RES SUPPORTING THE EFFICACY OF REMOVAL TREATMENTS Table B 1. Density of adult Stegastes nigricans (GLMM) Effect F value df Pr > F Treatment 45.91 4,314 <0.0001 Period 74.26 1,324 <0.0001 Notes: Differences in the average density of Stegastes nigricans among reefs throughout the experiment by fixed effects of Treatment and Period (before or after removals), and Day as a random effect as well as relevant contrasts within each main effect (PROC GLIMMIX, Poisson error distribution, SAS 9.13)
Table B 2 Contrasts for density of adult Stegastes nigricans (GLMM) Contrasts Estimate Range t value df Pr > t Present vs Removed 7.3 5.3 9.9 12.56 314 <0.0001 Before vs After 0.5 0.5 0.6 8.62 314 <0.001 Notes: Differences in the average density of Stegastes nigricans among reefs throughout the experiment by fixed effects of Treatment and Period (before or after removals), and Day as a random effect as well as relevant contrasts within each main effect (PROC GLIMMIX, Poisson error distribution, SAS 9.13)
Table B 3 Density of adult Stegastes nigricans before removals ( 2) Variable N X 2 df P > X 2 Stegastes densities BEFORE removals 667 0.84 3 0.84 Treatment Count Percent Cumulative Count Cumulative Percent 1 ( Stegastes +/ Turf + ) 172 25.8 172 25 .8 2 ( Stegastes + Turf ) 159 23.8 331 49.6 3 (Stegastes Turf + ) 163 24.4 494 74.1 4 (Stegastes Turf ) 173 25.9 667 100.0 5 (Stegastes Turf )* 0 NA NA NA Notes : Chi square contingency analyses compare densities of adult Stegastes nigricans on treatment reefs Before and After removals (PROC FREQ, SAS 9.13).* Treatment 5 was not tested because cumulative counts across reefs and sampling observations were extremely low (0 and 1 in the before and after periods, respectively).
Table B 4. De nsity of adult Stegastes nigricans after removals ( 2). Variable N X 2 df P > X 2 Stegastes densities AFTER removals 1075 625.74 3 <0.0001 Treatment Count Percent Cumulative Count Cumulative Percent 1 ( Stegastes +/ Turf + ) 522 48.6 522 48.6 2 ( Stegastes + Turf ) 419 39.0 941 87.5 3 (Stegastes Turf + ) 72 6.7 1013 94.2 4 (Stegastes Turf ) 62 5.8 1075 100.0 5 (Stegastes Turf )* 1 NA NA NA Notes : Chi square contingency analyses compare densities of adult Stegastes nigricans on treatment reefs Before and After removals (PROC FREQ, SA S 9.13).* Treatment 5 was not tested because cumulative counts across reefs and sampling observations were extremely low (0 and 1 in the before and after periods, respectively).
Table B 5 Mean mass of dried algal turf (GLM) Effect F value df Pr > F Cora l 0.79 3, 418 0.50 Treatment 128.45 4, 418 <0.0001 Sub Treatment 29.31 2, 418 <0.0001 Coral x Treatment 0.64 12, 418 0.81 Coral x Sub Treatment 0.79 6, 418 0.58 Treatment x Sub Treatment 7.21 8, 418 <0.0001 Coral x Treatment x Sub Treatment 0.90 24, 418 0.60 Notes: Differences in mean mass of dried algal turf scraped from nubbins across coral species, treatments, subtreatments, and all interactions (PROC GLM, SAS 9.13).
Figure B 1. Mean number of adult Stegastes nigricans ( 95 % C.I.). R emovals occurred at Day 0. Squares represent reefs with S. nigricans present while circles represent reefs with S. nigricans removed (or naturally absent). Symbol color designates the level of turf in each treatment: ambient (black), partially removed (grey), and naturally absent (light grey).
Figure B 2. Mean mass of dried algal turf ( 95 % C.I.) Algae was scraped from the bases of nubbins at the termination of the experiment. Treatments included Stegastes nigricans present or removed, crossed with farmed a lgal turf present or removed and natural absent Porites reefs as a positive contol. On each reef, nubbins were placed in a full cage (black), partial cage (control=grey) or open (without cage=beige). Given the lack of differences among coral taxa tested (A ppendix B: Table B3) we pooled these data to ease graphical presentation.
APPENDIX C EXPERIMENTAL METHODS DETAILS, AND RESUL TS FOR DAILY CHANGE IN CORAL MASS Methods We used the buoyant mass technique (Davies 1989) to quantify the change in skeletal mas s of nubbins. At the termination of the experiment, we quantified surface area of nubbins using the dyedipping method (HoeghGuldburg 1988). The high predation frequency in exposed and cage control subtreatments (see Results) restricted our use of this procedure to caged nubbins. The procedure was repeated twice for each nubbin and the average of these values was then converted to surface area based on a linear calibration curve (r2 = 0.99) derived from cylinders of known surface area. We tested for an ef fect of surface on the calculated change in dry mass for each species of caged nubbins and found insignificant or extremely weak relationships. As a result, we have omitted surface area from our growth estimates and instead report daily growth as the chang e in mass ov er the course of the experiment (MassFinal MassInitial / 32 days) by treatment. Results Table C 1. Daily coral growth (GLM) Variable F value df Pr > F Coral 14.87 3, 418 <0.0001 Treatment 2.92 4, 418 0.02 Sub Treatment 12.97 2, 418 <0.0001 Coral Treatment 1.76 12, 418 0.05 Coral Sub Treatment 4.78 6, 418 <0.0001 Treatment Sub Treatment 1.79 8, 418 0.08 Coral Treatment Sub Treatment 0.96 24, 418 0.52 Notes: Variation in skeletal growth/ loss among corals, treatments, subtreat ments, and all interactions (PROC GLM, SAS 9.13).
Table C 2. Tests of growth effect slices for each coral species (GLM). Effect Species F value df Pr > F Coral TTT* SUBTTT Acropora striata 3.26 14, 417 <0.0001 Coral TTT* SUBTTT Montipora flowerii 3. 63 14, 417 <0.001 Coral TTT* SUBTTT Pocillopora verrucosa 0.88 14, 417 0.58 Coral TTT* SUBTTT Porites australiensis 1.18 14, 417 0.28 Notes: Where TTT = Treat ment and SUBTTT = SubTreatment (PROC GLM, SAS 9.13).
Table C 3. Daily growth of caged i ndividuals (GLM) Effect Species F value df Pr > F Treatment Acropora striata 0.48 4, 35 0.75 Treatment Montipora flowerii 0.58 4, 35 0.68 Treatment Pocillopora verrucosa 0.89 4, 35 0.48 Treatment Porites australiensis 0.90 4, 35 0.48 Notes: Variation in skeletal growth or loss in CAGED individuals (only) among experimental treatments by species (PROC MIXED, SAS 9.13).
Figure C 1. Mean daily coral growth rates ( 95% CI). A) Pocillopora verrucosa B) Porites australiensis The average daily change in mass for each species is illustrated as a function of experimental treatment for Caged (black) versus Exposed (grey = Control + Open) nubbins.
APPENDIX D RESULTS OF FULL MODE LS TESTING THE FREQU ENCY OF PREDATION AND OVERGROWTH Table D 1. Predation freq uency (Logisitic Regression). Taxon Variable Estimate S.E. Wald 2 df Pr > 2 Acropora striata Stegastes 4.68 1.15 16.44 1 <0.0001 Stegastes 0.037 1.07 0.0012 1 0.97 Turf 1.20 0.82 2.12 1 0.15 Caging 4.58 0.99 21.37 1 <0.0001 Exposure 1.25 0.86 2.11 1 0.15 Montipora flowerii Stegastes 3.30 0.99 11.09 1 0.0009 Stegastes 0.18 0.99 0.03 1 0.86 Turf 1.78 0.70 6.44 1 0.01 Caging 4.33 0.96 20.37 1 <0.0001 Exposure 0.97 0.64 2.26 1 0.13 Notes: Results of logistic regression isolating which experimental variables best predicted variation in predati on frequency among and within reefs (PROC LOGISTIC, SAS 9.13). Pocillopora verrucosa and Porites australiensis each incurred zero skeletal predation across all treatments and subtreatments and no analyses were performed.We used binomial (1 or 0) coding an d dummy variables to isolate the comparisons of interest. Stegastes refers to reefs with Stegastes present vs. removed or absent. Stegastes isolates differences between Stegastes removed and naturally absent reefs. Turf isolates differences between reefs in which turf was present versus removed or naturally absent. Caging refers to nubbins protected by cages vs. cage control and exposed nubbins. Exposure isolates differences between cage control and exposed nubbins. Pocillopora verrucosa and Porites australiensis did not experience any predation.
Table D 2. Frequency of algal overgrowth (Logistic Regression) Taxon Variable Estimate S.E. Wald 2 df Pr > 2 Acropora striata Stegastes 0.58 1.27 0.21 1 0.64 Stegastes 1.36 1.44 0.88 1 0.35 Turf 1.94 0.84 5.47 1 0.02 Caging 0.91 0.81 1.28 1 0.25 Exposure 0.36 0.86 0.17 1 0.67 Montipora flowerii Stegastes 2.41 1.12 4.65 1 0.03 Stegaste s 0.61 1.18 0.27 1 0.60 Turf 0.70 0.53 1.69 1 0.19 Caging 1.22 0.61 3.90 1 0.04 Exposure 0.24 0.69 0.11 1 0.73 Pocillopora verrucosa Stegastes 10.39 216.7 0.002 1 0.96 Stegastes 11.13 216.7 0.003 1 0.96 Turf 0.74 0.90 0.67 1 0.41 Caging 0 .43 0.95 0.20 1 0.65 Exposure 0.75 1.25 0.35 1 0.55 Porites australiensis Stegastes 1.42 0.84 2.86 1 0.09 Stegastes 0.92 0.85 1.17 1 0.27 Turf 1.23 0.45 7.40 1 0.007 Caging 0.28 0.52 0.28 1 0.60 Exposure 0.28 0.52 0.28 1 0.60 Notes: Results of logistic regression isolating which experimental variables best predicted variation in the frequency of algal overgrowth of corals among and within reefs (PROC LOGISTIC, SAS 9.13). We used binomial (1 or 0) coding and dummy variables to isolate the comp arisons of interest. Stegastes refers to reefs with Stegastes present vs. removed or absent. Stegastes isolates differences between Stegastes removed and naturally absent reefs. Turf isolates differences between reefs in which turf was present versus removed or naturally absent. Caging refers to nubbins protected by cages vs. cage control and expose d nubbins. Exposure isolates differences between cage control and exposed nubbins.
Table D 3. Turf abundance and coral overgrowth (Linear Regression) Notes: Results of linear regressions testing the degree to which turf abundance explains variation in (ar c sine square root transformed) algal overgrowth of corals (PROC REG, SAS 9.13). Taxon R2 Adj. R2. F value df Pr > F Acropora striata 0.51 0.46 13.26 1, 13 0.0 03 Montipora flowerii 0.19 0.14 3.21 1, 13 0.10 Pocillopora verrucosa 0.04 0.03 0.55 1, 13 0.47 Porites australiensis 0.56 0.53 16.56 1, 13 0.001
Figure D 1. Frequency of algal overgrowth for Pocillopora verrucosa. Mean frequency of algal overgrowth ( 95% CI) for Pocillopora verrucosa as a function of Treatment and Sub Treatment (Caged vs. Exposed=Cage control and Open)
APPENDIX E RESULTS OF FULL MODE LS TESTING THE TERRI TORIAL EFFORT EXPEND ED BY STEGASTES NIGRICANS AND RESPONSES BY IMPORTANT MOBILE FISHES (DENSITY, FORAGING, REEF USE) Table E 1. Chase f requency (G LMM) Notes: Differences in frequency of chases by Stegastes nigricans using a fixed effect of Reef Type (S. nigricans present, removed, or absent) and Day as a random effect. Chase frequency was calcu lated as the sum of the number of reefs on which chases occurred divided by the number of reefs in that category and averaged through time (PROC GLIMMIX, SAS 9.13, binomial error distribution). Effect F value Df Pr>F Reef Type 5.15 2,597 0.006
Table E 2 C ontrasts of c hase f requency (GLMM) Contrasts E stimate Range t value df Pr > t Absent vs Present 0.07 0.009 0.5 2.62 597 0.009 Absent vs Removed 0.15 0.02 1.1 1.84 597 0.07 Present vs Removed 0.47 0.2 0.9 2.15 597 0.03 Notes : Differences in frequency of chases by Stegastes nigricans using a fixed effect of Reef Type ( S tegastes nigricans present, removed, or absent) and Day as a random effect. Chase frequency was calculated as the sum of the number of reefs on which chases occurred divided by the number of reefs in that category and averaged t hrough time (PROC GLIMMIX, SAS 9.13, binomial error distribution). Estimate refers to odds ratios with range indicating the 95% lower and upper confidence intervals.
Table E 3 List of species documented for each target f ish f amily Acanthuridae: Acanthur us nigrofuscus,Ctenochaetus binotatus, Naso lituratus Chae odontidae: Chaetodon citrinellus, C. lunulatus, C. ornatissimus, C. reticulates, C. vagabundis Scaridae : Scarus altipinnis, S. oviceps, S. psittacus, S. sordidus
Table E 4. Densities of target fam ilies of coral reef fishes (GLMM) Family Variable F value df Pr > F Acanthuridae (surgeonfishes) Reef Type 3.81 2,193 0.02 Chaetodontidae (butterflyfishes) Reef Type 2.94 2,193 0.06 Scaridae (parrotfishes) Reef Type 2.86 2,193 0.06 Notes: Densities of fishes were calculated as the mean number of the total individuals observed in a given fish family per reef per day. We tested fish density using a fixed effect of Reef Type ( Stegastes nigricans present, removed, or absent) and Day as a random effect ( after manipulations only: Day 0 30), for each taxonomic fish Family (Acanthuridae, Chaetodontidae, and Scaridae) (PROC GLIMMIX, SAS 9.13, Poisson error distribution).
Table E 5. Contrasts of d ensities of target families of coral reef fishes (GLMM) Fa mily Contrasts Est. Range t value df Pr > t Acanthuridae Absent vs Present 1.22 1.0 1.5 2.13 193 0.03 (surgeonfishes) Absent vs Removed 1.00 0.8 1.2 0.03 193 0.98 Present vs Removed 1.21 1.0 1.4 2.53 193 0.01 Chaetodontidae Absent vs Present 0.55 0. 3 1.1 1.73 193 0.09 (butterflyfishes) Absent vs Removed 0.44 0.2 0.9 2.40 193 0.01 Present vs Removed 1.24 0.8 1.9 1.00 193 0.31 Scaridae Absent vs Present 1.11 0.7 1.7 0.05 193 0.62 (parrotfishes) Absent vs Removed 0.76 0.5 1.1 1.37 193 0.17 Pr esent vs Removed 1.46 1.0 2.0 2.31 193 0.02 Notes: Differences in densities of fishes using a fixed effect of Reef Type ( Stegastes nigricans present, removed, or absent) and Day as a random effect, for each taxonomic fish Family (PROC GLIMMIX, SAS 9 .13, Poisson error distribution). Densities of fishes were calculated as the mean number of the total individuals observed in a given fish family per reef per day. Estimate refers to the odds ratio. Range indicates the 95% lower and upper confidence interv als.
Table E 6 Reef u se by target families of coral reef fishes (GLMM) Family Variable F value df Pr > F Acanthuridae (surgeonfishes) Reef Type 30.13 2,197 <0.0001 Chaetodontidae (butterflyfishes) Reef Type 5.61 2,197 0.004 Scaridae (parrotfishes) R eef Type 29.03 2,197 <0.0001 Notes: Differences in frequency of reef use using a fixed effect of Reef Type ( Stegastes nigricans present, removed, or absent) and Day as a random effect, for each taxonomic fish Family (PROC GLIMMIX, SAS 9.13, binomial error distribution) with relevant contrasts reported below. During counts fish were assigned a microhabitat use code to designate the location of fish when sighted (Table 1). Reef use summarizes the frequency at which individuals were observed using dead portions of the reef, living coral, and turf (versus using the water column or sand / rubble / pavement substrates within 1 m of the reef base).
Table E 7 Contrasts of reef u se by target families of coral reef fishes (GLMM) Family Contrasts Est. Range t value df Pr > t Acanthuridae Absent vs Present 5.41 2.2 13.1 3.78 197 0.0002 (surgeonfishes) Absent vs Removed 0.17 0.06 0.4 3.93 197 0.0001 Present vs Removed 31.82 13.2 77.0 7.72 197 <0.0001 Chaetodontidae Absent vs Present 0.64 0.2 2.5 0.65 197 0.52 (butterflyfishes) Absent vs Removed 0.20 0.05 0.7 2.50 197 0.01 Present vs Removed 3.25 1.4 7.7 2.71 197 0.007 Scaridae Absent vs Present 0.83 0.3 2.4 0.35 197 0.73 (parrotfishes) Absent vs Removed 0.055 0.02 0.2 5.56 197 <0.0001 Presen t vs Removed 14.96 6.8 32.9 6.77 197 <0.0001 Notes: Differences in frequency of reef use using a fixed effect of Reef Type ( Stegastes nigricans present, removed, or absent) and Day as a random effect, for each taxonomic fish Family (PROC GLIMMIX, SA S 9.13, binomial error distribution) with relevant contrasts reported below. During counts fish were assigned a microhabitat use code to designate the location of fish when sighted (Table 1). Reef use summarizes the frequency at which individuals were observed using dead portions of the reef, living coral, and turf (versus using the water column or sand / rubble / pavement substrates within 1 m of the reef base). Estimate refers to the odds ratio. Range indicates the 95% lower and upper confidence interva ls.
Table E 8 Foraging f requency of target families of coral reef fishes (GLMM) Family Variable F value df Pr > F Acanthuridae (surgeonfishes) Reef Type 9.48 2,197 <0.0001 Chaetodontida (butterflyfishes) Reef Type 6.01 2,197 0.002 Scaridae (parrotf ishes) Reef Type 20.15 2,197 <0.0001 Notes: Differences in frequency of foraging by fishes using a fixed effect of Reef Type (Stegastes nigricans present, removed, or absent) and Day as a random effect, for each taxonomic fish Family (PROC GLIMMIX, SAS 9.13, binomial error distribution). Each sampling period, observed individuals were noted as foraging (1) or not foraging (0).Foraging represents the frequency we observed individuals of a given family actively foraging according to treatment (n=8 re efs / treatment) throughout the experiment (n=5 sampling periods). Estimate refers to the odds ratio. Range indicates the 95% lower and upper confidence intervals.
Table E 9. Contrasts of foraging f requency of target families of reef fishes (GLMM) Family Contrasts Est. Range t value df Pr > t Acanthuridae Absent vs Present 1.24 0.5 3.1 0.47 197 0.64 (surgeonfishes) Absent vs Removed 0.24 0.09 0.6 3.01 197 0.002 Present vs Removed 5.16 2.4 11.3 4.12 197 <0.0001 Chaetodontidae Absent vs Present 0.76 0.2 2.7 0.42 197 0.68 (butterflyfishes) Absent vs Removed 0.21 0.06 0.7 2.56 197 0.01 Present vs Removed 3.61 1.5 8.6 2.91 197 0.004 Scaridae Absent vs Present 2.15 0.8 5.5 1.61 197 0.10 (parrotfishes) Absent vs Removed 0.17 0.07 0.4 3.88 197 0.0 001 Present vs Removed 12.2 5.4 27.3 6.10 197 <0.0001 Notes: Differences in frequency of foraging by fishes using a fixed effect of Reef Type (Stegastes nigricans present, removed, or absent) and Day as a random effect, for each taxonomic fish Fami ly (PROC GLIMMIX, SAS 9.13, binomial error distribution). Each sampling period, observed individuals were noted as foraging (1) or not foraging (0).Foraging represents the frequency we observed individuals of a given family actively foraging according t o treatment (n=8 reefs / treatment) throughout the experiment (n=5 sampling periods). Estimate refers to the odds ratio. Range indicates the 95% lower and upper confidence intervals.
157 APPENDIX F DETAILED METHODOLOGY FOR LAB AND FIELD OBSERVATIONS OF COMPE TITION BETWEEN PORITES AND ALGAL TURF Table F 1. Test of average change in live Porites (LMM) Data Variable num DF den DF F value Pr > F Full dataset Treatment 3 21 3.33 0.04 Normal data only Treatment 3 15 3.28 0.05 Notes: Results of LMM to evaluate the average linear change in live Porites by Treatment (lme (package: nlme), R v. 2.9.2). Treatments included a CONTROL (+Turf, +Sediment, Vermetid snails); REMOVAL ( Turf, Sediment, Vermetid snails); SEDIMENT (+Turf, Sediment, Removals); and VERMETID (+T urf, +Sediment,+Vermetid) (See Table 31). This model included a random effect of Quadrat nested within Reef. Note, the data were not normal and the analysis was repeated for the subset of the data that satisfied normality, with no change in the outcom e, thus full model results are reported with confidence.
158 Table F 2. Contrasts for change in live Porites (LMM) Comparison Effect DF t value Pr > t Control vs. Removal Turf Effect 21 1.04 0.30 Control vs. Sediment Sediment Effect 21 2.27 0.03 Control vs. Vermetid Vermetid Effect 21 0.41 0.68 Notes: Results of LMM contrasts to isolate effects of interest on the average linear change in live Porites (see Table 31, lme (package: nlme), R v. 2.9.2). Treatments included a CONTROL (+Turf, +Sediment, Verme tid snails); REMOVAL ( Turf, Sediment, Vermetid snails); SEDIMENT (+Turf, Sediment, Removals); and VERMETID (+Turf, +Sediment,+Vermetid).
159 Table F 3. Test of average change in bleaching front (LMM) Data Variable num DF den DF F value Pr > F Full data s et Treatment 3 21 0.26 0.86 Notes: Results of LMM to evaluate the average linear change in live Porites by Treatment (lme (package: nlme), R v. 2.9.2). Treatments included a CONTROL (+Turf, +Sediment, Vermetid snails); REMOVAL ( Turf, Sediment, Vermet id snails); SEDIMENT (+Turf, Sediment, Removals); and VERMETID (+Turf, +Sediment,+Vermetid) (See Table 31). This model included a random effect of Quadrat nested within Reef
160 Figure F 1. Paired reef surveys: Reef Size. Mean reef volume (maximum length x perpendicular width x average height 95% CI) of paired Porites reefs colonized by Stegastes nigricans (Colonized) and uncolonized Porites reefs (Uncolonized) did not differ significantly (t=1.87, df=22, P >0.05).
161 A) B) C) D) Figure F 2. Example photo quadrats from a sample experimental reef at Time=0. (A) Treatment 1: Sediment; N=8). (B) Treatment 2: removed, N=9). C) Treatment 3: vermetid sn ails/+Turf/ removal, N=9). D) Treatment 4: +vermetid snails/+turf/+sediment (Vermetid snails present, N=7). Note in the removal treatment bands of thick and thin dead skeleton were evident when turf was removed, suggesting rates of overg rowth are not constant and there may be seasonal, in addition to spatial, variation in the rate of overgrowth.
162 Figure F 3. Schematic of line measurements for transition zone experiment. Yellow lines illustrate representative measurements of combined turf and substrate (turf or dead coral). Red lines indicate length of substrate (turf or dead coral) alone. Small transparent red lines adjacent to the meeting point of these two lines estimated the size of the dead tissue zone. Linear measurements were taken from initial (Febuary 14th, 2007) and final (March 30th, 2007) photographs of each Treatment quadrat (see Table 31 and Figure F 2) randomly assigned to each experimental reef (N=9).
163 APPENDIX G EXPERIMENTAL REEFS BEFORE AND AFTER FARMERFISH REMOVALS Table G 1. Experimental reef characters BEFORE manipulations. Pair Treatment Num Reef Size (m 2 ) Mean STNI ( 95%C.I.) Acropora Pocillopora A Present 1 13.87 18.3 2.8 7 20 A Removed 17 12.59 15.0 3.9 18 18 B Present 3 5.58 8.0 1.1 5 21 B Removed 1 3 5.37 6.7 1.3 6 5 C Present 4 4.68 6.0 0.0 1 5 C Removed 5 4.10 6.7 1.3 3 11 D Present 8 13.40 19.3 0.7 6 25 D Removed 16 14.08 23.0 0.0 9 22 E Present 15 3.13 6.7 1.7 6 8 E Removed 9 3.08 6.7 0.7 3 10 F Present 14 7.88 11.3 1.3 5 14 F Removed 7 7.43 12.7 0.7 8 15 G Present 22 9.78 21.5 1.0 17 18 G Removed 24 13.54 22.0 0.0 13 20 H Present 26 3.76 10.0 0.0 17 26 H Removed 25 3.21 11.0 0.0 9 15 Notes: Pairs (A H) of experimental reefs were randomly assigned to a Stegastes nigricans ( STNI ) Treatment. Num is the arbitrary field designation for reefs. The main factors of interest were Reef Size (length x width), mean number of adult STNI, and in situ abundances of Acropora (N=133) and Pocillopora (N=253) in Jul y 2007, prior to farmerfish removals. The last two pairs (G & H) were added in July 2007, prior to manipulations, after storm damage reduced sample sizes of larger corals, in particular.
164 Table G 2. Differences in the average size of experimental reefs (LM) Variable DF Sum Sq Mean Sq F value Pr > F Treatment 1 0.37 0.37 0.3 1 0.5 9 Pair 7 274.74 39.25 33.52 0<0.0001 Residuals 7 8.20 1.17 Notes: Differences in the average size of experimental (length x width, m2) as a function Treatment ( Stegastes nigricans PRESENT versus REMOVED) and Pair (A H; N=8; see Table G 1, G 5) (Linear Model ( LM), lm, R v. 2.9.2).
165 Table G 3. Reef size model estimates (LM). Variable Estimate Std.Error t value Pr > t Pair A 13.38 0.81 16.49 <0.0001 Pair B 7.76 1.08 7.17 0.0001 Pair C 8.84 1.08 8.17 <0.0001 Pair D 0.51 1.08 0.47 0.65 Pair E 10.12 1.08 9.36 <0.0001 Pair F 5.58 1.08 5.15 0.001 Pair G 1.57 1.08 1.45 0.19 Pair H 9.75 1.08 9.01 <0.0001 Treatment:Present 0.30 0.54 0.56 0.59 Notes: E stimates o f differences in reef size (length x width, m2) as a function Treatment ( Stegastes nigricans PRESENT versus REMOVED) and Pair (A H; N=8; see Tables G 1 and G 2 ) (Linear Model (LM), lm, R v. 2.9.2).
166 Table G 4. Average density of Stegastes nigricans b efore removals (LMM). Variable numDF denDF F value Pr > F Treatment 1 7 0.0 9 0. 77 Pair 7 7 68.21 <0.0001 Notes: Differences in the density of adult Stegastes nigricans on experimental reefs Before removals as a function Treatment ( Stegastes nigrica ns PRESENT versus REMOVED) and Pair (A H; N=8; see Tables G 1, G 2, and G 3), with Reef as a random effect ( Linear Mixed Model ( L M M), lm e R v. 2.9.2).
167 Table G 5. Stegastes nigricans density estimates before removals (LMM) Variable Estimate Std.Erro r df t value Pr > t Pair A 0.93 0.16 7 5.91 0.006 Pair B 1.33 0.21 7 6.37 0.004 Pair C 1.91 0.21 7 9.11 <0.0001 Pair D 0.04 0.21 7 0.18 0.87 Pair E 3.07 0.21 7 14.67 <0.0001 Pair F 0.68 0.21 7 3.27 0.01 Pair G 0.16 0.21 7 0.73 0.49 Pair H 2.66 0. 21 7 12.68 <0.0001 Treatment: Present 0.03 0.10 7 0.30 0.77 Notes: Estimates of differences in density of adult Stegastes nigricans during the observation period ( i.e., Before removals) as a function Treatment ( Stegastes nigricans PRESENT versus REMO VED) and Pair (A H; N=8; see Tables G 1 through G 4 ) with Reef as a random effect (Linear Mixed Model (LMM), lm, R v. 2.9.2).
168 Table G 6. Average density of Stegastes nigricans after removals (LMM) Variable numDF denDF F value Pr > F Treatment 1 7 2 6.82 0. 001 Pair 7 7 055 0.78 Notes: Differences in the density of adult Stegastes nigricans on reefs at the termination of the Experimental period (i.e., after farmerfish were removed) as a function Treatment Stegastes nigricans PRESENT versus REMOVED) and Pair (A H; N=8; see Tables G 1 through G 5 ) (Linear Model ( LM), lm, R v. 2.9.2).
169 Table G 7. Stegastes nigricans density estimates after removals (LMM) Variable Estimate Std.Error t value Pr > t Pair A 0.06 0.48 0.13 0.90 Pair B 0.08 0.64 0.1 2 0.91 Pair C 0.25 0.64 0.39 0.71 Pair D 0.14 0.64 0.21 0.83 Pair E 0.23 0.64 0.36 0.72 Pair F 0.13 0.64 0.20 0.85 Pair G 0.26 0.64 0.40 0.69 Pair H 0.82 0.64 1.30 0.23 Treatment: Present 1.62 0.32 5.18 0.001 Notes: Estimates of differences in abundance of adult Stegastes nigricans at the termination of the Experimental period ( i.e., after farmerfish were removed) as a function Treatment ( Stegastes nigricans PRESENT versus REMOVED) and Pair (A H, see Table G 1, G 2) with Reef as a random effect ( Linear Mixed Model ( L M M), lm, R v. 2.9.2).
170 Figure G 1. Mean size of experimental reefs. Mean reef size (length x width, m2; 95% C.I.) did not differ between Treatment (F1,7=0.31, P =0.59) but size did vary by Reef Pair (Appendix G: Tables G 2 and G 3). Measurements were taken at the end of the BEFORE observational period (i.e., prior to experimental removals July 2007, N=8 reefs per Treatment).
171 Figure G 2. Mean density of Stegastes nigricans before removals. Mean density of adult far merfish ( 95% C.I.) on experimental reefs did not differ throughout the observational period (i.e., prior to experimental removals, July 2006 July 2007, N=6 reefs per Treatment; Appendix G: Tables G 4 and G 5).
172 Figure G 3. Mean density of Stegastes nigricans at the termination of the experiment. Mean density of adult farmerfish ( 95% C.I.) was significantly lower on reefs with S. nigricans removed (i.e., at the termination of the Experimental period, April 2008, N=8 reefs per Treatment; Appendix G: Tables G 6 and G 7).
173 APPENDIX H POPULATION MEASURES FOR ACROPORA AND POCILLOPORA IN THE PRESENCE AND REMOVAL OF FARMERFISH Table H 1 Mortality frequency for Acropora & Pocillopora (GLMM, LRT). Taxon Variable 2 df Pr > 2 Acropora Treatment 6. 3 1 12.61 1 0.0003 Period 2.29 4.58 1 0.03 Previous Size 59.92 119.84 1 <0.00 0 1 Reef Size 0.05 0.00 1 1.00 Treatment:Period 5.75 11.51 1 0.0007 Treatment:Previous Size 2.37 4.74 1 0.03 Period:Previou s Size 0.001 5e 04 1 0.98 Treatment:Period:Previous Size 1.02 2.03 1 0.15 Pocillopora Treatment 7.60 15.18 1 <0.0001 Period 3.00 6.00 1 0.01 Previous Size 74.79 149.59 1 <0.0001 Reef Size 0.43 0.86 1 0.36 Treatment:Period 8.77 17.54 1 <0.0001 Treatment:Previous Size 0.58 1.16 1 0.28 Period:Previous Size 1.43 2.68 1 0.10 Treatment:Period:Previous Size 0.22 0.44 1 0.51 Notes: Results of Likelihood Ratio Tests (LRT) using backward reduction of the full Generalized Linear Mixed Model (GLMM) defined a priori to isolate which experimental variab les best predicted variation in the frequency of mortality among and within reefs. For both Acropora (N=192) and Pocillopora (N=381), the full model tested was: Mortality ~ Treatment + Period + Previous Si ze +Reef size + Treatment*Period PAIR REEF, where ( glmer, lme4 package, binomial distribution, logit link, fit with Laplace approximation, R v 2.9. 2). Treatment refers to the presence or removal of Stegastes nigricans from experimental reefs (during experimental period only). Period distinguishes between the 200607 observational period before removals occurred (versus the 200708 experimental period). Previous Size is a continuous covariate testing whether the size of an individual at the previous time step influences the likelihood of mortality. Reef size isolates differences due to increases in the size of experimental reefs (Appendix G: Table G 1). In addition, we tested all interactions of the factors of interest (Treatment, Period, and Previous Size). Significant interactions and all main effects were retained in the final model (Table H 4).
174 Table H 2 Mortality frequency model estim ates (GLMM). Taxon Variable Estimate S.E. z value df Pr > z Acropora Before (Present Size=1.0) 0.89 0.93 0.95 1 0.34 Treatment (Removed) 2.86 1.04 2.72 1 0.006 Period (Experimental) 0.61 0.66 0.92 1 0.36 Previous size 0.14 0.08 1.88 1 0.06 Reef size 0.01 0.06 0.21 1 0.83 TTTRemoved:PERIODExptal 3.11 1.06 2.91 1 0.003 TTTRemoved :Previous size 0.17 0.09 1.85 1 0.06 Pocillopora Before (Present Size=0) 1.79 0.70 2.56 1 0.01 Treatment (Removed) 0.90 0.59 1.51 1 0.13 PERIOD (Experime ntal) 1.34 0.68 1.97 1 0.05 Previous size 0.06 0.04 1.42 1 0.16 Reef size 0.06 0.04 1.60 1 0.11 TTTRemoved:PERIODExptal 1.25 0.74 1.70 1 0.09 PeriodExptal:Previous Size 0.003 0.05 0.08 1 0.94 Notes: Parameter estimates for final GLMM logistic regressions used to isolat e which experimental variab les best predicted variation in the frequency of mortality among and within reefs (Wald z, 2 distribution) For Acropora (N=65), the final model was: Mortality ~ Treatment + Period + Previous Size +Trea PAIR REEF, of Pair (N=8) and Reef (N=16) ( glmer, lme4 package, binomial distribution, logit link, R v 2.9.2, Table H 3 ) For Pocillopora (N=134), the model was the same, but did not contain the Period*Previous Size interaction (please see Table H 3). Before ( Stegastes ) establishes reefs with Stegastes present in the before period as a baseline (Size=0, Intercept ). Treatment (Removed) isolates differences between Steg astes removed and naturally present reefs. Period (Experimental) isolates differences between the before and experimental periods. Previous size refers to differences due to increases in the previous size of individuals. Reef size isolates differences due to increases in the size of experimental reefs (Appendix G: Table G 1). Abbreviations: TTT=Treatment and Exptal=Experimental.
175 Table H 3 Follow up mortality analyses for Acropora (GLMM, LRT) Subset Variable 2 df Pr > 2 Acropora BEFORE ONLY Treatment 0.61 1.23 1 0.26 Previous Size 1.04 2.08 1 0.15 Reef Size 0.01 0.01 1 0.92 Treatment : Previous Size 0.04 0.09 1 0.77 Acropora PRESENT ONLY Period 0.39 0.77 1 0.38 Previous Size 12.18 24.36 1 <0.001 Reef Size 0.51 2.42 1 0.12 Per iod : Previous Size 0.64 1.29 1 0.26 Notes: Results of Likelihood Ratio Tests (LRT) using backward reduction of the GLMM models defined a priori to isolate which experimental variables best predicted variation in the frequency of mortality for select subset s of the data. For Acropora in the before period only (both treatments) (N=65), the full model PAIR REEF, 2). For Acropora within reefs with Stegastes nigricans present only (N=94), the full model tested was: Mortality ~ Period + Previous Size PAIR REEF, (Note: Pair was omit ted, as it is redundant without the other Treatment). I used logistic regression for both models (glmer, lme4 package, binomial distribution, logit link, fit with Laplace approximation, R v 2.9.2).Treatment refers to the presence or removal of Stegastes nigricans from experimental reefs. Period distinguishes between the 200607 observational period before removals occurred (versus the 200708 experimental period). Previous Size is a continuous covariate testing whether the size of an individual at the previous time step influences the likelihood of mortality. Reef size isolates differences due to increases in the size of experimental reefs (Appendix G: Table G 1).In addition, I tested all interactions of the relevant of interest for each subset of the data.
176 Table H 4 Follow up mortality analyses for Pocillopora (GLMM, LRT) Subset Variable 2 df Pr > 2 Pocill opora BEFORE ONLY Treatment 1.03 2.05 1 0.15 Previous Size 1.60 3.20 1 0.07 Reef Size 0.17 0.34 1 0.56 Treatment : Previous Size 0.13 0.26 1 0.61 Pocillopora PRESENT ONLY Period 0.39 0.78 1 0.38 Previous Size 29.11 58 .22 1 <0.0001 Reef Size 0.00 1e 04 1 0.99 Period : Previous Size 0.05 0.10 1 0.75 Notes: Results of Likelihood Ratio Tests (LRT) using backward reduction of the GLMM models defined a priori to isolate which experimental variables best predicted variatio n in the frequency of mortality for select subsets of the data. For Pocillopora in the before period only (both treatments) (N=134), the full model tested was: Mortality ~ Treatment + Period + Previous Size +Reef size + Previous Size*Period + PAIR REEF, Pocillopora within reefs with Stegastes nigricans present only (N=206), the full model tested was: Mortality ~ Treatment + PAIR REEF, effects of Reef (N=8) (Note: Pair was omitted, as it is redundant without the other Treatment). I used logistic re gression for both models (glmer, lme4 package, binomial distribution, logit link, fit with Laplace approximation, R v 2.9.2).Treatment refers to the presence or removal of Stegastes nigricans from experimental. Period distinguishes between the 200607 observational period before removals occurred (versus the 200708 experimental period). Previous Size is a continuous covariate testing whether the size of an individual at the previous time step influences the likelihood of mortality. Reef size isola tes differences due to increases in the size of experimental reefs (Appendix G: Table G 1).In addition, I tested the interactions of interest for each subset of the data.
177 Table H 5 Differences in final size of Acropora and Pocillopora (LMM). Taxon Vari able Est. S.E. num DF den DF F value Pr > F Acropora Before (Present Size=1.0) 2.06 0.78 1 101 14.30 0.0003 Treatment (Removed) 0.51 0.93 2 6 14.89 0.008 Period (Experimental) 1.98 0.94 1 10 1 10.81 0.001 Previous Size 0.11 0.05 1 10 1 70.99 <0.00 0 1 Reef Size 0.04 0.05 1 6 0.21 0.66 Treatment : Period 6.92 1.51 1 10 1 38.07 < 0.00 0 1 Treatment : Previous Size 0.03 0.08 1 10 1 131.1 3 <0.000 1 Period:Previous Size 0.24 0.07 1 101 13.30 0.0004 Treatment:Period:Previous Size 1.20 0.11 1 101 116.6 1 <0 .000 1 Pocillopora Before (Present Size=1.0) 1.20 0.60 1 250 3.33 0.07 Treatment (Removed) 0.27 0.65 1 6 0.06 0.81 Period (Experimental) 0.80 0.83 1 250 4.85 0.03 Previous Size 0.06 0.03 1 250 8.12 0.004 Reef Size 0.02 0.04 1 6 0. 24 0. 64 Treat ment : Period 1.27 1.38 1 250 0.50 0.48 Treatment : Previous Size 0.01 0.04 1 250 0.63 0.43 Period:Previous Size 0.02 0.05 1 250 1.31 0.25 Treatment:Period:Previous Size 0.15 0.08 1 250 3.42 0.07 Notes: Results of the Linear Mixed Model (LLM) defined a priori to isolate which experimental variables best predicted variation in the final size of individuals (max length, cm) among and within reefs. For both Acropora (N=122) and Pocillopora (N= 272), the full model tested was: Size ~ Treatment + Period + P revious Size + Reef size + Treatment*Period + Previous Size*Period + Period*Previous Size + PAIR REEF), within Pair (N=8) ( lme, nlme package, R v 2.9.2). The Interc ept refers to differences for very small (Size=0) individuals. Treatment refers to the presence or removal of Stegastes nigricans from experimental reefs (which occurred during the experimental period only). Period isolates differences in size between the 200708 experimental period, and the 200607 observational period (before removals occurred). Previous Size is a continuous covariate isolating differences in the final size due to the size of an individual at the previous time step. Reef size i solates differences due to increases in the size of experimental reefs (Table G 1). In addition, we tested all interactions of the factors of interest (Treatment, Period, and Previous Size). Please note, because variability was not constant between periods I added a term to specify constant variance within (but not between) periods (weights=varident(form=~1|Period)).
178 Table H 6 Follow up change in size analyses for Acropora (LMM) Subset Variable Est. S.E. num DF den DF F value Pr > F Acropora BEFORE ONLY Before (Present Size=0) 2.44 0.86 1 42 19.43 0.0001 Treatment 0.43 0.94 1 3 1*10 6 0.99 Previous Size 0.11 0.04 1 42 6.81 0.01 Reef Size 0.006 0.06 1 3 0.02 0.90 Treatment:Previous Size 0.03 0.08 1 42 0.12 0.73 Acropora PRESENT ONLY Befo re (Present Size=0) 2.43 0.81 1 67 30.01 <0.0001 Period 2.06 0.93 1 67 1.07 0.30 Previous Size 0.11 0.04 1 67 0.005 0.92 Reef Size 0.005 0.06 1 6 0.12 0.94 Period:Previous Size 0.24 0.06 1 67 13.48 0.0005 Notes: Results of the Linear Mixed Model (LLM) defined a priori Previous Size) for select subsets of the data. For Acropora in the BEFORE period only (both treatments) + Previous Size*Period PAIR REEF, Acropora within reefs with Stegastes nigricans PRESENT Previous Size +Reef size + Previous REEF, (N=8) (Note: Pair was omitted, as it is redundant without the other Treatment). Intercept refers to differences for very small (Size=0) individuals. Treatment refers to the presence or removal of Stegastes nigricans on experimental reefs. Period distinguishes between the 200607 observational period before removals occurred (versus the 200708 experimental period). Previous Size is a continuous covariate testing whether the size of an individual at the previous time step influences the likelihood of mortality. Reef size isolates differences due to increases in the size of experimental reefs (Appendix G: Table G 1). In addition, I tested relevant interactions for each subset of the data as reported above. Please note, because variability was not constant between periods, I added a term to specify constant variance within (but not between) periods to the latter model (weights=varident(form=~1|Period)).
179 Table H 7 Follow up change in s ize analyses for Pocillopora (LMM) Subset Variable Est. S.E. num DF den DF F value Pr > F Pocillopora BEFORE ONLY Before (Present Size=1.0) 1.59 0.76 1 106 4.59 0.03 Treatment (Removed) 0.34 0.70 1 4 0.56 0.49 Previous Size 0.06 0.03 1 106 9.33 0.0 03 Reef Size 0.02 0.06 1 4 0.06 0.82 Treatment:Previous Size 0.004 0.04 1 106 0.008 0.92 Pocillopora PRESENT ONLY Before (Present Size=1.0) 0.77 0.56 1 162 6.78 0.01 Period (Experimental) 0.76 0.81 1 162 5.31 0.02 Previous Size 0.06 0.02 1 162 11.00 0.001 Reef Size 0.07 0.05 1 6 2.05 0.20 Period:Previous Size 0.01 0.05 1 162 0.09 0.76 Notes: Results of the Linear Mixed Model (LLM) defined a priori Previous Size) for select subsets of the data. For Pocillopora in the BEFORE period only (both Previous Size*PerioPAIR REEF, For Pocillopora within reefs with Stegastes nigricans PRESENT R EEF, effects of Reef (N=8) (Note: Pair was omitted, as it is redundant without the other Treatment). Intercept refers to differences for very small (Size=0) individuals. Treatment refers to the presence or removal of Stegastes nigricans on experimental reefs. Period distinguishes between the 200607 observational period before removals occurred (versus the 200708 experimental period). Previous Size is a continuous covariate testing whether the size of an individual at the previous time step influences the likelihood of mortality. Reef size isolates differences due to increases in the size of experimental reefs (Appendix G: Table G 1). In addition, I tested relevant interactions for each subset of the data as reported abov e. Please note, because variability was not constant between periods, I added a term to specify constant variance within (but not between) periods to the latter model (weights=varident(form=~1|Period)).
180 Figure H 1. Raw mortality values for Acropora Mo rtality (=1) versus survival (=0) as a function of Previous Size (cm) by Treatment ( Stegastes nigricans PRESENT or REMOVED) and Period (BEFORE or EXPERIMENTAL).
181 Figure H 2 Raw mortality values for Pocillopora Mortality (=1) versus survival (=0) a s a function of Previous Size (cm) by Treatment ( Stegastes nigricans PRESENT or REMOVED) and Period (BEFORE or EXPERIMENTAL).
182 Figure H 3 Representative photos of Acropora predation before and after removals of Stegastes nigricans No te: There are a few branch tips missing before, but substantial skeletal loss and development of crustose coralline algae after removals, which is consistent high grazing pressure by excavating corallivores and herbivores.
183 Figure H 4 Representative photos of Pocillo pora predation before and after removals of Stegastes nigricans Note: There is some predation by butterflyfishes before and drastic shrinkage, consistent with crown of thorns seastar Acanthaster planci feeding scars, after removals.
184 APPENDIX I TRANSITIO N MATRICES FOR ACROPORA AND POCILLOPORA IN THE PRESENCE AND REMOVAL OF FARMERFISH Figure I 1. Schematic of size classes and types of transitions for Acropora and Pocillopora.
185 Figure I 2. Size structure transition matrix for Acropora on all reefs prio r to experimental manipulations. This matrix combines colonies on both control and (scheduled to be) removal reefs (i.e., with Stegastes nigricans present during the BEFORE Period, July 2006April 2007). Individuals that fragmented are represented by re d dots, while all others are in black. Size classes, marked by grey lines on the X (Previous size, cm) and Y (Current size, cm) axes, correspond to the size class schematic in Figure I 1. The black line marks the 1:1 line and runs through those cells that represent looping, or staying in the same size class. Cells below this diagonal vector represent shrinkage from a larger size class to a smaller size class, while cells above represent growth in the opposite direction. For each cell, raw transition v alues (bold) and probabilities (small) are in red. Raw mortality (bold) and probabilities (small) are in blue, and also represented by small black lines, along the x axis. The pink line represents the Size~Previous size (lm), with no mortalities or fragments included. The teal line represents the same relationship with fragments (but no mortalities).
186 Figure I 3. Size structure transition matrix for Acropora following application of clove oil to experimental CONTROL reefs (i.e., with Stegastes nigricans pr esent during the EXPERIMENTAL Period, July 2007Apr 2008). Individuals that fragmented are represented by red dots, while all others are in black. Size classes, marked by grey lines on the X (Previous size, cm) and Y (Current size, cm) axes, correspond t o the size class schematic in Figure I 1. The black line marks the 1:1 line and runs through those cells that represent looping, or staying in the same size class. Cells below this diagonal vector represent shrinkage from a larger size class to a small er size class, while cells above represent growth in the opposite direction. For each cell, raw transition values (bold) and probabilities (small) are in red. Raw mortality (bold) and probabilities (small) are in blue, and also represented by small black lines, along the x axis. The pink line represents the Size~Previous size (lm), with no mortalities or fragments included. The teal line represents the same relationship with fragments (but no mortalities).
187 Figure I 4. Size structure transition matrix for Acropora following application of clove oil and removal of farmerfish (i.e., experimental reefs in which Stegastes nigricans was REMOVED during the EXPERIMENTAL Period, July 2007Apr 2008). Individuals that fragmented are represented by red dots, while all others are in black. Size classes, marked by grey lines on the X (Previous size, cm) and Y (Current size, cm) axes, correspond to the size class schematic in Figure I 1. The black line marks the 1:1 line and runs through those cells that represent lo oping, or staying in the same size class. Cells below this diagonal vector represent shrinkage from a larger size class to a smaller size class, while cells above represent growth in the opposite direction. For each cell, raw transition values (bold) and probabilities (small) are in red. Raw mortality (bold) and probabilities (small) are in blue, and also represented by small black lines, along the x axis. The pink line represents the Size~Previous size (lm), with no mortalities or fragments included. The teal line represents the same relationship with fragments (but no mortalities).
188 Figure I 5. Size structure transition matrix for Pocillopora on all reefs prior to experimental manipulations. This includes control and (scheduled to be) removal reefs ( i.e., with Stegastes nigricans present during the BEFORE Period, July 2006April 2007). Individuals that fragmented are usually represented by red dots (none present), while all others are in black. Size classes, marked by grey lines on the X (Previous size, cm) and Y (Current size, cm) axes, correspond to the size class schematic in Figure I 1. The black line marks the 1:1 line and runs through those cells that represent looping, or staying in the same size class. Cells below this diagonal vector repr esent shrinkage from a larger size class to a smaller size class, while cells above represent growth in the opposite direction. For each cell, raw transition values (bold) and probabilities (small) are in red. Raw mortality (bold) and probabilities (sm all) are in blue, and also represented by small black lines, along the x axis. The pink line represents the Size~Previous size (lm), with no mortalities or fragments included.
189 Figure I 6. Size structure transition matrix for Pocillopora following appli cation of clove oil to experimental CONTROL reefs (i.e., with Stegastes nigricans present during the EXPERIMENTAL Period, July 2007Apr 2008). Individuals that fragmented are usually represented by red dots (none present), while all others are in black. Size classes, marked by grey lines on the X (Previous size, cm) and Y (Current size, cm) axes, correspond to the size class schematic in Figure I 1. The black line marks the 1:1 line and runs through those cells that represent looping, or staying in the same size class. Cells below this diagonal vector represent shrinkage from a larger size class to a smaller size class, while cells above represent growth in the opposite direction. For each cell, raw transition values (bold) and probabilities (small) are in red. Raw mortality (bold) and probabilities (small) are in blue, and also represented by small black lines, along the x axis. The pink line represents the Size~Previous size (lm), with no mortalities or fragments included.
190 Figure I 7 Size struct ure transition matrix for Pocillopora following application of clove oil and removal of farmerfish (i.e., experimental reefs in which Stegastes nigricans was REMOVED during the EXPERIMENTAL Period, July 2007Apr 2008). Individuals that fragmented are represented by red dots, while all others are in black. Size classes, marked by grey lines on the X (Previous size, cm) and Y (Current size, cm) axes, correspond to the size class schematic in Figure I 1. The black line marks the 1:1 line and runs through tho se cells that represent looping, or staying in the same size class. Cells below this diagonal vector represent shrinkage from a larger size class to a smaller size class, while cells above represent growth in the opposite direction. For each cell, raw transition values (bold) and probabilities (small) are in red. Raw mortality (bold) and probabilities (small) are in blue, and also represented by small black lines, along the x axis. The pink line represents the Size~Previous size (lm), with no mortalit ies or fragments included. The teal line represents the same relationship with fragments (but no mortalities).
191 LIST OF REFERENCES Adjeroud, M., Y. Chancerelle, M. Schrimm, T. Perez, D. Lecc hini, R. Galzin, and B. Salvat. 2005. Detecting the effects of natural disturbances on coral assemblages in French Polynesia: a decade survey at multiple scales. Aquatic Living Resources 18:111123. Altieri, A. H., B. R. Silliman, and M. D. Bertness. 2007. Hierarchical organization via a facilitation cascade in intertidal cordgrass bed communities. American Naturalist 169: 195206. Bak, R. P. M. and M. S. Engel. 1979. Distribution, abundance, and survival of juvenile hermatypic corals (scleractinia) and the importance of life history strategies in the parent coral community Marine Biology 54:341352. Bellwood D. R., T. P. Hughes C. Folke, and M. Nystrm. 2004. Confronting the coral reef crisis. Nature 429:827833. Berlow, E L S A Naverrete, C J. Briggs, M E Power, B A Menge.1999. Quantifying variation in the str engths of species interactions. Ecology 80: 22062224. B ertness, M. D. and G. H. Leonard. 1997. The role of positive interactions in communities : Lessons from intertidal habitats. Ecology 78: 19761989. Bertness, M. D., G. C. Trussell, P. J. Ewanchuk, and B. R. Silliman. 2002. Do alternate stable community states exist in the Gulf of Maine rocky intertidal zone? Ecology 83:34343448. Bertness, M. D., G. C. Trussell, P. J. Ewanchuk, B. R. Silliman, and C. M. Crain. 2004. Consumer controlled community states on Gulf of Maine rocky shores. Ecology 85:13211331. Berumen, M. L. and M. S. Pratchett. 2006. Recovery without resilience : persistent dist urbance and long term shifts in the structure of fish and coral communities at Tiahura Reef, Moorea. Coral Reefs 25:647653. Birrell, C. L L J. McCook, and B L Willis 2005. Effects of algal turfs and sediment on coral settlement. Marine Pollution Bulletin 51:408414. Bolker, B. M, M. E. Brooks, C. J.Clark, S. W.Geange, J. R. Poulsen, M. H. H. Stevens, and J. S. S. W hite 2009 Generalized linear mixed models : a practical guide for ecology and evolution. Trends in Ecology and Evolution 24: 127 135. Boogert, N. J., D. M. Paterson, and K. N. Laland. 2006. The implications of niche construction and ecosystem engineering f or conservat ion biology. Bioscience 56:570578.
192 Bouchon, C. 1985 Quantative study of scleractinian coral communities of Tiahura Reef (Moorea, French Polynesia). Pages 279290 in Proceedings of the 5 th International Coral Reef Symposium Volume 6. Papeete, Tahiti. Boyer, S. E., J. S. S. White, A. C. Stier, and C. W. Osenberg 2009. Effects of the fish anesthetic clove oil (eugenol) on coral health and growth. Journal of Experimental M arine Biology & Ecology 369:5357. Bruno, J. F., J. J. Stachowicz, and M. D. Bertness. 2003. Inclusion of facilitation into ecological theory. Trends in Ecology and Evolution 18:119125. Bruno, J. F. and E. R. Selig. 2007. Regional decline of coral cover in the IndoPacific: timing, extent, and subregional comparisons. P ublic L i brary o f S cience One 2: e711. doi:10.1371/journal.pone.0000711. Burkepile, D. E. and M. E. Hay. 2006. Herbivore vs. nutrient control of marine primary producers: context dependent effects. Ecology 87:31283139. Burkepile, D. E. and M. E. Hay. 2008. Herbivore species richness and feeding complementarity affect community structure and function on a coral reef. Proceedings of the National Academy of Sciences 105:1620116206. Caswell, H. 2001. Li fe table response experiments, Pages 258278 in Matrix Population Models 2nd Ed., Sinauer, Sunderland, Massachusetts, USA Chase, J. M., M. A. Liebold, A. L. Downing, and J. B. Shurin. 2000. The effects of productivity, herbivory, and plant species turnover in grassland food webs. Ecology 81:24852497. Ceccarelli, D. M. G. P. Jones, and L. J. McCook. 2001 Territorial damselfishes as determinants of the structure of benthic communities on coral reefs. Oceanography and Marine Biology 39:355389. Ceccarelli, D. M., G. P. Jones, and L. J. McCook. 2005. Foragers versus far mers: contrasting effects of two behavioural groups of herbivores on coral reefs. Oecologia 145:445453. Ceccarelli, D. M. 2007. Modification of benthic communities by territorial damselfish: a multispecies comparison. Coral reefs 26:853866. Cole, A. J., M. S. Pratchett, and G. P. Jones 2008. Diversity and functional importance of coral feeding fishes on tropical coral reefs. Fish and Fisheries Bulletin 9 : 286 307. Coles, S. L., and R. Strathmann. 1973. Observations on coral mucus flocs and their potent ial trophic significance. Limnology and Oceanography 18:673678.
1 93 Colgan, M. W. 1987. Coral reef recovery on Guam (Micronesia) after catastrophic predation by Acanthaster planci Ecology 68:15921605. Colgan, M. W. 1985. Growth rate reduction and modificati on of a coral colony by a vermetid mollusc, Dendropoma maxima. Pages 205210 in Proceedings of the 5th International Coral Reef Symposium, Volume 6. Papeete, Tahiti. Connell, J. H. 197 3. Population ecology of reef building corals. Pages 205245 in O. A. Jones and R. Endean, eds. Biology and geology of coral reefs, Volume 2. Academic Press, New York. Connell, J. H. 1978 Diversity in tropical rain forests and coral reefs. Science 199: 13021310. Connell, J. H., T. P. Hughes, and C. C. Wallace. 1997. A 30yea r study of coral abundance, recruitment, and disturbance at several scales in space and time. Ecological Monographs 67:461488. Crain, C. M. and M. D. Bertness 2005 Community impacts of a tussock sedge: Is ecosystem engineering important in benign habitat s? Ecology 86: 26952704. Crain, C. M. and M. D. Bertness 2006 Ecosystem engineering across environmental gradients : i mplications for conservation and management. Bioscience 56:211218. Crossland, C. 1928. Notes on the ecology of the reef builders of Tahi ti. Proceedings of the Zoological Society 3:717735. Crouse, D T L B Crowder, H Caswell. 1987. A stage based population model for loggerhead sea turtles and implications for conservation. Ecology 68:14121423. Darwin, C. 1842. The structure and distr ibution of coral reefs. Smith, Elder & Co., London, England. Dayton P. K. 1972. Toward an understanding of community resilience and the potential effects of enrichments to the benthos at McMurdo Sound, Antarctica Pages 81 96 in Proc eedings of the Colloq uium on Cons ervation Prob lems in Antartica, Allen Press Davies, P. S. 1989. Short term growth measurements of corals using an accurate buoyant weighing technique. Marine Biology 101:389395. DeMartini, E. E., A. M. Friedlander, S. A. Sandin, and E. Sala. 2008. Differences in fish assemblage structure between fished and unfished atolls in the northern Line Islands, central Pacific. Marine Ecology Progress Series 365:199215.
194 Doak, D. F., J. A. Estes, B. S. Halpern, U. Jacob, D. R. Lindberg, J. Lovvorn, D. H Monson, M. T. Tinker, T. M. Williams, J. T. Wootton, I. Carroll, M. Emmerson, F. Micheli, and M. Novak. Understanding and predicting ecological dynamics: are major surprises inevitable? Ecology 89: 952961. Doherty, P. J. 1981. Coral reef fishes: recruit ment limited assemblages? Pages 465 470 in Proceedings of the 4th International Coral Reef Symposium, Volume 2. Manila, Phillipines. Done, T. J., P. K. Dayton, A. E. Dayton, and R. Stegner 1991. Regional and local variability in recovery of shallow coral reef communities. Coral Reefs 9:183192. Dulvy, N. K., R. P. Freckleton, and N. V. C. Polunin. 2004. Coral reef cascades and the indirect effects of predator removal by exploitation. Ecology Letters 7:410416. Ebersole, J. P. 1977. The adaptive signficance of interspecific territoriality in the reef fish Eupomacentrus leucostictus Ecology 58:914920. Edmunds, P. J. 2005. The effect of sublethal increases in temperature on the growth and population trajectories of three scleractinian corals on the southern Great Barrier Reef. Oecologia 146:350364. Elahi, R. and P. J. Edmunds 200 7a. Consequences of fission in the coral Siderastrea sidereal : growth rates of small colonies and clonal input to population structure. Coral Reefs 26:271276. Elahi, R. and P. J. Edmunds 200 7b. Tissue age affects calcification in the scleractinian coral, Madracis mirabilis Biological Bulletin 212:2028. Ellison, A. M., M. S. Bank, B. D. Clinton, E. A. Colburn, K. Elliot, C. R. Ford, D. R. Foster, B. D. Kloeppel, J. D. Knoepp, G. M Lovett, J. Mohan, D. A. Orwig, N. L. Rodenhouse, W. V. Sobczak, K. A. Stinson, J. K. Stone, C. M. Swan, J. Thompson, B. Von Holle, and J. R. Webster. 2005. Loss of foundation species: consequences for the structure and dynamics of forested ecosystems. Fr ontiers in Ecology and the Environment 3: 479486. Eriksson, B. K., A. Rubach, and H. Hillebrand. 2006. Biotic habitat complexity controls species diversity and nutrient effects on net biomass production. Ecology 87:246254. Friedlander, A. M. and E. E. DeM artini. 2002. Contrasts in density, size, and biomass of reef fishes between the northwestern and the main Hawaiian islands: the effects of fishing down apex predators. Marine Ecology Progress Series 230: 253264. Frydl, P. 1978. The effect of parrotfish ( Scaridae) on coral in Barbados, W.I. International Review of Hydrobiology 64: 737748.
195 Galzin, R. and J. P. Pointier 1985. Moorea Island, Society Archipelago. Pages 157 164 in Proceedings of the 5 th International Coral Reef Symposium Volume 2. Papeete, Ta hiti. Galzin, R. 1987. Structure of fish communities of French Polynesian coral reefs. II. Temporal scales. Marine Ecology Progress Series 41: 137145. Galzin R., J. S. S. White, and T. Lison de Loma. In Preparation. Can grazers limit coral recovery? Spat iotemporal trends in Moorea, French Polynesia (For submission to Science ). Gleason, M. G. 1993. Effects of disturbance on coral communities : bleaching in Moorea, French Polynesia. Coral Reefs 12: 193201. Gleason, M. G. 1994. Factors influencing the recover y of corals after natural disturbance on reefs in Moorea, French Polynesia. Ph.D. dissertation, University of California, Berkeley, 358 pp. Gleason, M. G. 1996. Coral recruitment in Moorea, French Polynesia: the importance of patch type and temporal variat ion. Journal of Experimental Marine Biology & Ecology 207:79101. Glynn, P. W. 1985. El Nio associated disturbance to coral reefs and post disturbance mortality by Acanthaster planci. Marine Ecology Progress Series 26:295300. Glynn, P. W. and M. W. Colga n 1988 Defense of corals and enhancement of coral diversity by territorial damselfishes. Pages 157 163 in Proceedings of the 6 th International Coral Reef Symposium, Volume 2 Townsville, Australia Goreau, T. F., J. C. Lang, E. A. Graham, and P. D. Gorea u. 1972. Structure and ecology of the Saipan reefs in relation to predation by Acanthaster planci (Linneus) Bulletin of Marine Science 22:113152. Green, A. L. 1992. Damselfish territories: focal sites for studies of the early life history of labroid fish es. Pages 601605 in Proceedings of the 7th International Coral Reef Symposium, Volume 1. Guam, Micronesia. Green, A. L. 1998. Spatiotemporal patterns of recruitment of labroid fishes (Pisces: Labridae and Scaridae) to damselfish territories. Environmental Biology of Fishes 51:235244. Haley, M. P. and C. R. Mller. 2002. Territorial behaviour of beaugregory damselfish ( Stegastes leucostictus ) in response to egg predators. Journal of Experimental Marine Biology and Ecology. 273:151159. Hall, VR and TP Hug hes. 1996. Reproductive strategies of modular animals: comparative studies of reef building corals. Ecology 77:950963.
196 Halpern, B. S., K. Cottenie, and B. R. Broitman. 2006. Strong topdown control in southern California kelp forest ecosystems. Science 312:12301232. Harrington, L, K Fabricius, G Death, and A Negri. 2004. Recognition and selection of settlement substrata determine post settlement survival in corals. Ecology 85:34283437. Hartman C. D., J. S. S. White, and C. W. Osenberg. In Preparation Effects of the farmerfish Stegastes nigricans on size and abudance of vermetid snails. (For submission to Marine Ecology Progress Series ). Hata H. and M. Kato. 2002. Weeding by the herbivorous damselfish Stegastes nigricans in nearly monocultural algae far ms. Marine Ecology Progress Series 237:227231. Hata, H., and M. Nishihira. 2002. Territorial damselfish enhances multi species co existence of foraminifera mediated by biotic habitat structuring. Journal of Experimental Marine Biology and Ecology 270:215 240. Hata, H., M. Nishihira, and S. Kamura. 2002. Effects of habitat conditioning by the damselfish Stegastes nigricans (Lacepede) on the community structure of benthic algae. Journal of Experimental Marine Biology and Ecology 280:95 116. Hata, H and M Kato. 2003. Demise of monocultural algal farms by exclusion of territorial damselfish. Marine Ecology Progress Series 263:159167. Hata, H. and M. Kato. 2004. Monoculture and mixedspecies algal farms on a coral reef are maintained through intensive and extensive management by damselfishes. Journal of Experimental Marine Biology and Ecology 313:285296. Hata, H. and M. Kato 2006. A novel obligate cultivation mutualism between damselfish and Polysiphonia algae. Biology Letters 2: 593 596. Hay, M. E. 1981. H erbivory, algal distribution, and the maintenance of betweenhabitat diversity on a tropical fringing reef. American Naturalist 118: 520540. Highsmith, R. C. 1982. Reproduction by fragmentation in corals. Marine Ecology Progress Series 7: 207 227. Hixon, M. A. 1997. Effects of reef fish on corals and algae. Pages 230248 in Birkeland C., ed. Life and death of coral reefs, Chapman and Hall, New York, U.S.A Hixon, M. A. and W. N. Brostoff. 1981. Fish grazing and community structure of Hawaiian reef algae. Pages 507514 in Proceedings of the 4th International Coral Reef Symposium, Volume 2. Manila, Phillipines. Hixon, M. A. and W. N. Brostoff. 1983 Damselfish as keystone species in reverse : i ntermediate disturbance and diversity of reef algae. Science 220: 511 513.
197 Hixon, M. A. and W. N. Brostoff. 1996. Succession and herbivory: effects of differential fish grazing on Hawaiian coral reef algae. Ecological Monographs 66:6790. HoeghGuldburg, O. 1988. A method for determining the surface area of corals. Coral Reefs 7:113 116. HoeghGuldburg, O., P. J. Mumby, A. J. Hooten, R. S. Steneck, P. Greenfield, E. Gomez, C. D. Harvell, P. F. Sale, A. J. Edwards, K. Caldiera, N. Knowlton, C. M. Eakin, R. Iglesias Prieto, N. Muthiga, R. H. Bradbury, A. Dubi, and M. E. Ha tziolos. Science 318:17371742. Hoey, A.S. and D.R. Bellwood. 2008. Cross shelf variation in the role of parrotfishes on the Great Barrier Reef. Coral Reefs 27:3747. Hughes R. N. and A. H. Lewis 1974. On the spatial distribution, feeding and reproduction of the vermetid gastropod dendropoma maximum. Journal of Zoology, London 172:531547. Hughes, T. P. 1984. Population dynamics based on individual size rather than age: a general model with a coral reef example. American Naturalist 123:778795. Hughes T. P 1989 Community Structure and diversity of coral reefs : the r ole of history. Ecology 70:275279. Hughes, T. P. 1990. Recruitment limitation, mortality, and population regulation in open systems: a case study. Ecology 71:1220. Hughes, T. P. 1994. Catastr ophes, phase shifts, and largescale degradation of a Caribbean coral reef Science 265:15471551. Hughes, T. P. and J. H. Connell. 1999. Multiple stressors on coral reefs: a long term perspective. Limnology and Oceanography 44:932940. Hughes, T. P. and J. B. Jackson. 1980. Do corals lie about their age? Some demographic consequences of partial morality, fission, and fusion. Science 209: 713715. Hughes, T. P. and J. E. Tanner. 2000. Recruitment failure, life histories, and long term decline of Caribbean corals. Ecology 81:2250 2263. Hughes, T. P., A. H. Baird, E. A. Dinsdale, V. J. Harriot, N. A. Moltschaniwskyj, M. S. Pratchett, J. E. Tanner, and B. L. Willis. 2002. Detecting regional variation using meta analysis and largescale sampling: latitudinal patterns in recruitment. Ecology 83: 436451.
198 Hughes, T. P., M. Rodrigues, D. R. Bellwood, D. Ceccarelli, O. HoeghGuldberg, L. McCook, N. Moltschaniwskyj, M. S. Pratchett, R. S. Steneck, and B. Willis. 2007. Phase shifts, herbivory, and the resilience of co ral reefs to climate change. Current Biology 17:360365. Huston, M. A. 1985. Patterns of species diversity on coral reefs. Annual Review of Ecology and Systematics 16: 149177. Idjadi, J. A., and P. J. Edmunds. 2003. Freeliving colonies of Porites in Moorea, French Polynesia. Bulletin of Marine Science 72:10251031. Jan, R. Q., C. T. Ho, and F. K. Shiah. 2003. Determinants of territory size of the dusky gregory. Journal of Fish Biology 63:15891597. Jayewardene, D., M. J. Donahue, and C. Birkeland. 2009. Effects of frquent fish prdation on corals in Hawaii. Coral Reefs 28:499506. Jones, C. G. J. H. Lawton, M. Shachak 1994. Organisms as ecosystem engineers. Oikos 69: 373 386. Jones C. G., J. H. Lawton, M. Shachak 1997 Positive and negative effects of organisms as physical ecosys tem engineers. Ecology 77:19461957. Jones, G. P., L. Santana, L. J. McCook, and M. I. McCormick 2006 Resource use and impact of three herbivorous damselfishes on coral reef communities. Marine E cology Progress Series 328:215224. Kappner, I., S. Al Moghrabi, C. Richter. 2000. Mucus net feeding by the vermetid gastropod Dendropoma maxima in coral reefs. Mari ne Ecology Progress Series 204: 309313. Kaufman, L. 1977 The three spot damselfish: Effects on benthic biota of Caribbean coral reefs. Pages 559 564 in Proceedings of the 3rd International Coral Reef Symposium, Volume 1. Miami, Florida, USA. Klumpp, D. W., D. McKinnon, and P. Daniel 1 987. Damselfish territories: zones of high productivity on coral reefs. Marine Ecology Progress Series 40: 41 51. Klumpp, D. W., A. D. McKinnon, and C. N. Mundy. 1987. Damselfish territories: zones of high productivity on coral reefs. Marine Ecology Progress Series 40:4151. Klumpp, D. W., A. D. McKinnon, and C. N. Mundy. 1988. Motile cryptof auna of a coral reef: abundance and trophic potential. Marine Ecology Progress Series 45:95108. Klumpp, D., W., and N. V. C. Polunin. 1989. Partitioning among grazers of food resources within damselfish territories on a coral reef. Journal of Experimental Marine Biology and Ecology 125:145169.
199 Knowlton, N. 2001. The future of coral reefs. Proceedings of the National Academy of Sciences 98:54195425. Knowlton, N., J. C. Lang, and B. D. Keller. 1988. Fates of staghorn coral isolates on hurricanedamaged ree fs in Jamaica: the role of predators. Pages 8388 in Proceedings of the 6th International Coral Reef Symposium, Volume 2. Townsville, Australia. Knowlton, N. and J. B. C. Jackson. 2008. Shifting baselines, local impacts, and global change on coral reefs. P ublic L ibrary o f S cience Biology 6: e54. doi:10.1371/journal.pbio.0060054. Kohler, K. E. and S. M. Gill. 2006. Coral Point Count with Excel extensions: A visual basic program for the determination of coral and substrate coverage using random point count methodology. Computers and Geosciences 32:12591269. Lessios, H.A., D. R. Robertson, and J. D. Cubit. 1984. Spread of Diadema mass mortality through the Caribbean. Science 226:335337. Letourneur, Y R. Galzin, and M. Harmelin Vivien. 1997. Temporal variati ons in the diet of the damselfish Stegastes nigricans (Lacepde) on a Runion fringing reef. Journal of Experimental Marine Biology and Ecology 217:118. Letourneur, Y 2000. Spatial and temporal variability in territoriality of a tropical benthic damselfi sh on a coral reef (Reunion island). Environmental Biology of Fishes 57:377391. Lewis, S. M. 1986. The role of herbivorous fishes in the organization of a Caribbean reef community. Ecological Monographs 56:184200. Lison de Loma, T., M. L. HarmelinVivien O. Naim, and M. F. Fontaine. 2000. Algal food processing by Stegastes nigricans a herbivorous damselfish: differences between an undisturbed and a disturbed coral reef site (La Runion, SW Indian Ocean). Oceanologica Acta 23:793804. Lison de Loma, T. and E. Ballesteros 2002. Microspatial variability inside epilithic algal communities within territories of the damselfish Stegastes nigricans at La Runion. Indian Ocean). Botanica Marina 45:316323. Littler, M. M., P. R. Taylor, and D. S. Littler. 1989. C omplex interactions in the control of coral zonation on a Caribbean reef flat. Oecologia 80: 1432 1939. Lobel, P. S. 1980. Herbivory by damselfishes and their role in coral reef community ecology. Bulletin of Marine Science 30: 273289. Mantyka, C. S. and D. R. Bellwood. 2007. Direct evaluation of macroalgal removal by herbivorous coral reef fishes. Coral Reefs 26:435442.
200 Mapstone, B. D., N. L. Andrew, Y. Chancerelle, B. Salvat 2007 Mediating effects of sea urchins on interactions among corals, algae and herbivorous fish in the Moorea lagoon, French Polynesia. Marine Ecology Progress Series 332:143153. McClanahan, T. R., M. Ateweberhan, N. A. J. Graham, S. K. Wilson, C. Ruiz Sebastin, M. M. M. Guillaume, and J. H. Bruggemann 2007. Western Indian Ocean coral communities : bleaching responses and susceptibility to extinction. Marine Ecology Progress Series 337:113. McCook, L. J. 1996. Effects of herbivores and water quality on Sargassum distribution on the central Great Barrier Reef: cross shelf transplants. Marine Ecology Progress Series 139: 179 192. McCook, L. J. 1999. Macroalgae, nutrients and phase shifts on coral reefs: scientific issues and management consequences for the Great Barrier Reef. Coral Reefs 18: 357 367. McCook, L. J., J. Jompa, and G. D iaz Pulido 2001. Competition between corals and algae on coral reefs : a review of evidence and mechanisms. Coral Reefs 19:400417. Miller, M. W. and M. E. Hay. 1998. Effects of fish predation and seaweed competition on the survival and growth of corals. O ecologia 113:231238. Mills, L S M E Soul, D F Doak .1993. The keystonespecies concept in ecology and conservation. Bioscienc e 43:219 224. Mora, C. 2008. A clear human footprint in the coral reefs of the Caribbean. Proceedings of the Royal Society B 275:767773. Morton J. E. 1965. Form and function in the evolution of the vermetidae. Bulletin of the British Museum of Natural History, Zoology 11:583630. Mumby, P. J., C. P. Dahlgren, A. R. Harborne, C. V. Kappel, F. Micheli, D. R. Brumbaugh, K. E. Ho lmes, J. M. Mendes, K. Broad, J. N. Sanchirico, K. Buch, S. Box, R. W. Stoffle, and A. B. Gill. 2006. Fishing, trophic cascades, and the process of grazing on coral reefs. Science 311: 98101. Mumby, P.J., A. R. Harborne, J. Williams, C. V. Kappel, D. R. B rumbaugh, F. Micheli, K. E. Holmes, C. P. Dahlgren, C. B. Paris, and P. G. Blackwell. Trophic cascades facilitates coral recruitment in a marine reserve. Proceedings of the National Academy of Sciences 104: 83628367. Nelder, J. A. 1998. The great mixedm odel muddle is alive and flourishing, alas! Food Quality and Preference 9:157159. Nugues, M. M. and C. M. Roberts 2003. Coral mortality and interaction with algae in relation to sedimentation. Coral Reefs 22:507516.
201 ODonnell J. L. and J. S. S White. I n P reparation. Habitat selectivity and size allometry of the Christmas tree worm, Spirobranchus giganteous (Polychaeta: Serpulidae) (For submission to the Bulletin of Marine Science ). Osenberg, C. W., B. M. Bolker, J. S. S. White, C. M. St. Mary, and J. S. Shima. 2006. Statistical issues and study design in ecological restorations: lessons learned from marine reserves. Pages 280302 in Falk, D. A., H., M. A. Palmer, J. B. Zedler, eds., Foundations of Restoration Ecology, Island Press, Washington D.C., U.S.A Paine, R. T. 1966. Food web complexity and species diversi ty. American Naturalist 100:6575. Paine, R. T. 1969. A note on trophic complexity and community stability. American Naturalist 103: 9193. Paine, R. T. 1992. Foodweb analysis through field measur ement of per capita interaction strength. Nature 355: 7375. Penin, L., M. Adjeroud, M. Schrimm, and H. S. Lenihan. 2007. High spatial variability in coral bleaching around Moorea (French Polynesia): patterns across locations and water depths. Comptes Rendus Biologies 330:171181. Peyrot Clausade, M., P. Hutchings, and G. Richard. 1992. Temporal variations of macroborers in massive Porites lobata on Moorea, French Polynesia. Coral Reefs 11:161166. Porter, J. W. 1972. Predation by Acanthaster and its effect on coral species diversity. American Naturalist 106:487492. Potts, D. C. 1977. Suppression of coral populations by filamentous algae within damselfish territories. Journal of Experimental Marine Biology and Ecology 28:207216. Power, M., D. Tilman, J. A. Estes, B. A. Menge, W. J. Bond, L. S. Mills, G. Daily, J. C. Castilla, J. Lubchenco, and R. T. Paine. 1996. Challenges in the quest for keystones. Bioscience 46:609620. Power, M. 1997. Estimating impacts of a dominant detritivore in a neotropical stream, Trends in Ecology and Evolution 12:47 48. Pratchett, M. S. Influence of coral symbionts on feeding preferences of crownof thorn starfish Acanthaster planci in the western Pacific. Marine Ecology Progess Series 214:111119. Purcell, S. W. 2000. Association of epilithic algae with sediment distribution on a windward reef in the northern Great Barrier Reef, Australia. Bulletin of Marine Science 66:199214.
202 Randall, J. E. 1977. The effect of fishes on coral reefs. Pages 159166 in Proceedings of the 2nd Inter national Coral Reef Symposium, Volume 1. Brisbane, Australia. Randall, J. E. 2005. Reef and shore fishes of the South Pacific: New Caledonia to Tahiti and the Pitcairn Islands. University of Hawaii Press, Honolulu, HI, 707 pp. Ribak, G., J. Heller, A. Gen in. 2005. Mucus Net Feeding on Organic Particles by the Vermetid Gastropod Dendropoma Maximum In and Below the Surf Zone. Mari ne Ecology Progress Series 293: 77 87. Risk, M. J. and P. W. Sammarco 1982 Bioerosion of corals and the influence of damselfish territoriality : a preliminary study. Oecologia 52 : 376 380. Ritchie, M. E., D. Tilman, and J. M. H. Knops. 1998. Herbivore effects on plant and nitrogen dynamics in oak savanna. Ecology 79:165 177. Robertson, D. R., S. G. Hoffman, and J. M. Sheldon. 1981. Availability of space for the territorial Caribbean damselfish Eupomacentrus planifrons Ecology 62:11621169. Robertson, D. R. 1996. Interspecific competition controls abundance and habitat use of territorial Caribbean damselfishes. Ecology 77:885899. Ro tjan, R. D. and S. M. Lewis 2008. Impact of coral predators on tropical reefs. Marine Ecology Progress Series 367: 73 91. Sale, P. F. 1977. Maintenance of high diversity in coral reef fish communities. American Naturalist 111: 337359. Sale, P. F. 1982. St ock recruit relationships and regional coexistence in a lottery competitive system: a simulation study. American Naturalist 120:139 159. Sammarco, P. W. 1983. Effects of fish grazing and damselfish territoriality on coral reef algae. I. Algal community str ucture. Marine Ecology Progress Series 13:114. Sammarco, P. W. and J. H. Carleton 1981. Damselfish territoriality and coral community structure: reduced grazing, coral recruitment, and effect on coral spat. Pages 525 535 in Proceedings of the 4th International Coral Reef Symposium Volume 2. Manila, Philippines. Sammarco, P.W., J. H. Carleton, and M. J. Risk. 1986. Effects of grazing and damselfish territoriality on bioerosion of dead corals: direct effects. Journal of Experimental Marine Biology and Eco logy 98:119. Sammarco, P. W., and M. J. Risk. 1990. Largescale patterns in internal bioerosion of Porites : cross continental shelf trends on the Great Barrier Reef. Marine Ecology Progress Series 59:145156.
203 Sammarco, P. W. and A. H. Williams 1982. Dams elfish territoriality : Influence on Diadema distribution and implications for coral community structure. Marine Ecology Progress Series 8: 5359. Sandin, S.A., J.E. Smith, E.E. DeMartini, E.A. Dinsdale, S.D. Donner, A.M. Friedlander, T. Konotchick, M. Malay J.E. Maragos, D. Obura, O. Pantos, G. Paulay, M. Richie, F. Rohwer, R.E. Schroeder, S. Walsh, J.B.C. Jackson, N. Knowlton, E. Sala. 2008. Baselines and degradation of coral reefs in the northern Line Islands. P ublic L ibrary o f S cience O ne 3: e1548. Shima J. S., C. W. Osenberg, C. M. St. Mary and L. Rogers. 200 6. Implication of changing coral communities: do larval traits or habitat features drive variation in density dependent mortality and recruitment of a juvenile reef fish? Pages 226231 in Proceedings of the 10th International Coral Reef Symposium, Volume 1. Okinawa, Japan. Shima, J. S., C. W. Osenberg, and C. M. St. Mary 2008. Quantifying site quality in a heterogeneous landscape: recruitment of a reef fish. Ecology 89:86 94. Sluka, R. D., and M. W. Miller. 2001. Herbivorous fish assemblages and herbivory pressure on Laamu Atoll, Republic of Maldives. Coral Reefs 20: 255262. Smith, J E M. Shaw, R. A. Edwards, D. Obura, O. Pantos, E. Sala, S. A. Sandin, S. Smriga, M. Hatay, and F. L. Rohwer 2006 Indirect effects of algae on coral : algaemediated, microbeinduced coral mortality. Ecology Letters 9:835845. Soul, M. E., J.A. Estes, J. Berger, and C. Martinez del Rios 2003. Ecological effectiveness : Conservation goals for interactive species Co nservation Biology 17:12381250. Sousa, W. P. 1979a. Experimental investigations of disturbance and ecological succession in a rocky intertidal algal community. Ecological Monographs 49:228254. Sousa, W. P. 1979b. Disturbance in marine intertidal boulder fields: the nonequilibrium maintenance of species diversity. Ecology 60:12251239. Spooner, D. E., and C. C. Vaughn. 2006. Context dependent effects of freshwater mussels on stream benthic communities. Freshwater Biology 51:10161024. Suefu ji, M. and R. va n Woesik 2001. Coral recovery from the 1998 bleaching event is facilitated by Stegastes territories, Okinawa, Japan. Coral Reefs 20: 385386. Steneck, R. S. and Sala, E. A. 2005. Large marine carnivores: trophic cascades and top down controls in coastal ec osystems past and present. Pages 110137 in Ray, J., K. Redford, R. Steneck and J. Berger, eds., Large Carnivores and the conservation of biodiversity, Island Press.
204 Stewart, H. L., S. J. Holbrook, R. J. Schmitt, and A. J. Brooks. 2006. Symbiotic crabs ma intain coral health by clearing sediment. Coral Reefs 25:609615. Stoddart, J. 1983. Asexual production of planulae in the coral Pocillopora damicornis Marine Biology 76: 279 284. Veron, J. E. N. 2000. Corals of the world, Volumes 13, 1381 pp. Australian Institute of Marine Science, Queensland, Australia. Vine, P. J. Effects of algal grazing and aggressive behavior of the fishes Pomacentrus lividus and Acanthurus sohal on coral reef ecology. Marine Biology 24:131136. Vuxton, E. A., J. S. S. White and B .M. Bolker In Preparation. Effects of the dusky farmerfish, Stegastes nigricans on algal community structure (For submission to Bulletin of Marine Science ). Warner, R. R., and P. L. Chesson. 1985. Coexistence mediated by recruitment fluctuations: a field guide to the storage effect. American Naturalist 125:769787. Webster, M. S. and G. R. Almany. 2002. Positive indirect effects in a coral reef fish community. Ecology Letters 5:549557. Werner, E. E. and J. F. Gilliam 1984 The ontogenetic niche and spec ies interactions in size structured populations. Annual Review of Ecology, Evolution, & Systematics 15: 393 425. Wellington, G. M. 1982 Depth zonation of corals in the Gulf of Panama: control and facilitation by r esident reef fishes. Ecological Monogr aphs 52:223241. White, J.S.S. and B.M. Bolker. In Preparation. Key engineer indirectly alters demographic rates and recovery of corals (For submission to American Naturalist) White, J.S.S., S.E. Boyer, and A. Delval. In Preparationa Indirect effects of Stegastes nigricans on massive Porites coral and implications for other community members (For submission to Coral Reefs). White, J. S. S., A. R. Thompson, and S. Ferse. In P reparationb Hierarchical sampling design tests spatial v ariation in distribution and abundance of sessile epifauna on dominant massive Porites spp. (For submission to the Marine Ecology Progress Series). White, J. S. S. and J. L. ODonnell. 2010. Indirect effects of a key ecosystem engineer alter survival and growth of foundation coral species. Ecology In press. Williams, A. H. 1980. The threespot damselfish: a noncarnivorous keystone species. American Naturalist 116:138142.
205 Williams, A. H. 1981. An analysis of competitive interactions in a patchy back reef environment. Ecology 62:1107112 0. Wilson, S. and D. R. Bellwood. 1997. Cryptic dietary components of territorial damselfishes (Pomacentridae, Labroidei). Marine Ecology Progress Series 153:299310. Woodley, J. D., E. A. Chornesky, P. A. Clifford, J. B. C. Jackson, S. Kaufman, N. Knowlton, J. C. Lang, M. P. Pearson, J. W. Porter, M. C. Rooney, K. W. Rylaarsdam, V. J. Tunnicliffe, C. M. Wahle, J. L Wulff, A. S. G. Curtis, M. D. Dallmeyer, B. P. Jupp, M. A. R. Koehl, J. Neigel, and E. M. Sides. 1981. Hurricane Allens impact on Jamaican cor al reefs. Science 214:749755. Wootton, J. T. 1997. Estimates and test of per capita interaction strength: diet, abundance, and impact of intertidally foraging birds. Ecological Monographs 67:4564. Wright, J.P., C. G. Jones, and A. S. Flecker. 2002. An ec osystem engineer, the beaver, increases species richness at the landscape scale. Oecologia 132:96101. Zuur, A., F., E. N. Leno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2009. Mixed effects models and extensions in ecology with R. Springer, New York, U.S.A. 574 pp.
206 BIOGRAPHICAL SKETCH JadaSimone Shanti White was born in Phoenix, Arizona, and raised in northern California. She was first introduced to Polynesia through several school related trips to Hawaii. These experiences had a profound impact and Jada attended Hawaii Pacific University after graduating from Mercy High School in 1995. At the time, she was PreProfessional, on track to Medical School, but she realized through her experiences at H.P.U. that there might be other options to study and contribute to the biological sciences. She returned to northern California and pursued basic study in biology at Butte Community College. At Butte, she was given hands on opportunities to learn about field ecology by her mentor, Albin Bills, and she never looked back. She started working at the Department of Fish and Game Habitat Conservation Division collecting basic population data for four runs of Chinook salmon in the Sacramento River Valley, undertook additional projects studying the California Newt on Table Mountain, and rounded out her experience by studying Steelhead and Rainbow Trout on the Feather River and its tributaries with the Department of Water Resources. When she had exhausted all coursework at Butte College, and gained considerable field experience, she transferred to University of California Santa Barbara. Within the first semester she joined the Schmitt Holbrook lab, which had a large research program conducting fieldwork in Moorea, French Polynesia. Volunteer work became a technician pos ition, which blossomed into a senior honors thesis opportunity, and ultimately resulted in additional professional experiences in both French Polynesia and Santa Barbara. In 2003, she moved to Gainesville, Florida to develop strong quantitative and natural history skills at the University of Florida and continue her work contributing to our understanding of community change in Polynesia.