MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 493: 165–177, 2013doi: 10.3354/meps10510

Published November 20

INTRODUCTION

In marine environments, grazing by herbivores,omnivores, and micro-predators can strongly influ-ence both settlement and post-settlement success ofbenthic organisms (Dayton 1971, Connell 1985, Hunt& Scheibling 1997, O’Connor et al. 2011). The avail-

ability of open space for settlement is often an indi-rect effect of grazing, which removes competitorsand reduces the accumulation of sediments. Forexample, in temperate intertidal or subtidal systems,adult abalone in California as well as some AtlanticSouth African sea urchins promote abalone settle-ment by creating settlement space and removing

© Inter-Research 2013 · www.int-res.com*Email: jkoleary@stanford.edu

Fish and sea urchin grazing opens settlement spaceequally but urchins reduce survival of

coral recruits

Jennifer K. O’Leary1,3,*, Donald Potts1, Kathryn M. Schoenrock1, Timothy R. McClahanan2

1Department of Ecology and Evolutionary Biology, University of California, Santa Cruz, California 95064, USA2Marine Programs, Wildlife Conservation Society, Bronx, New York 10460, USA

3Present address: Hopkins Marine Station, Stanford University, Pacific Grove, California 93950, USA

ABSTRACT: Grazing fishes and invertebrates are influential in marine ecosystems because theyopen space for benthic recruits, alter post-settlement recruit survival, and can often determinebenthic community composition. On tropical reefs, grazing fishes and sea urchins can play keyroles in limiting growth of fleshy macro-algae, thereby facilitating coral recruitment and maintain-ing coral-dominated communities. However, as grazer abundance increases, grazer influence oncorals (or other settlers) may shift from being positive and indirect by reducing space competitors,to being negative and direct by damage or removal of coral recruits. Fishing can alter both theabundances and types of dominant grazers, with potential cascading effects on coral recruitmentand subsequent community organization. In Kenya, sea urchins dominate grazing on heavilyfished reefs, while herbivorous and omnivorous fishes dominate grazing within fisheries closures(marine protected areas). We used reefs under these 2 fishery management systems to investigatethe effects of fish versus sea urchin grazing on the availability of settlement substrate for corals,subsequent coral settlement, and mortality of coral recruits. Fish and sea urchin grazers wereequally effective at clearing benthic space for coral settlement. However, grazing associated withhigh densities of sea urchins on fished reefs removed many coral recruits after settlement. In con-trast, fish grazing within fisheries closures enhanced coral survival compared to non-grazed treat-ments. We conclude that the effects of reduced abundance of grazing fishes (on available spacefor coral settlement) may be initially offset by increased sea urchin grazing, but that higher urchinabundances may ultimately reduce coral cover by their negative influence on post-settlement survival.

KEY WORDS: Coralline algae · Diadema · Herbivory · Predation · Phase shifts · Trophic cascades ·Marine protected areas · Western Indian Ocean · Kenya

Resale or republication not permitted without written consent of the publisher

FREEREE ACCESSCCESS

Mar Ecol Prog Ser 493: 165–177, 2013166

sediment (Day & Branch 2002, Miner et al. 2006).Grazers can also have direct effects on settled organ-isms through feeding on or dislodging new recruits.Intertidal herbivorous limpets, for example, havebeen shown to incidentally dislodge new barnaclerecruits (Connell 1961, Stimson 1970). Grazers mayalso increase post-settlement survival indirectly byremoving competing benthic organisms that canovergrow new settlers. Consequently, changes in thetype or abundance of the dominant grazer can causelarge community shifts through multiple pathways(Babcock et al. 1999, Mumby 2006, Hughes et al.2007, Salomon et al. 2007, Duffy et al. 2003).

On tropical coral reefs, grazing fishes and seaurchins often have key roles in limiting fleshy macro-algal growth, thereby indirectly facilitating coral re -cruitment and maintaining coral dominance (Edmunds& Carpenter 2001, Wright et al. 2005, Mumby et al.2007, Coma et al. 2011). When grazers are scarce,space for coral recruitment is reduced and recruitstend to be outcompeted and overgrown by faster-growing organisms. However, new coral recruits arehighly vulnerable to removal by grazing when omniv-orous fishes (especially parrotfish) or sea urchins arehighly abundant (Rotjan & Lewis 2008, McCauley etal. 2010, Cover 2011, Penin et al. 2011). Intense seaurchin grazing can also act to reduce coral recruit-ment indirectly by reducing the quantity and alteringthe species composition of crustose coralline algae(O’Leary & McClanahan 2010, O’Leary et al. 2012), astrongly preferred settlement substrate for manycorals (Morse et al. 1994, Heyward & Negri 1999,Harrington et al. 2004).

Various biological or anthropogenic factors, suchas disease or fishing, can alter both the type and theabundance of the dominant grazers. For example, following the massive die-off of Diadema antillarumin the 1980s, fish grazing has dominated most Carib -bean reefs (Lessios 1988, Hughes 1994). Fishing andmanagement can alter grazer dominance where bothfish and sea urchin grazers remain abundant, as inmany Indo-Pacific reefs (McClanahan et al. 1994,Harborne et al. 2009, Vermeij et al. 2010, Dee et al.2012). On Kenyan reefs in the Western Indian Ocean,for example, grazing fishes are abundant only in fish-eries closures (i.e. fully protected national marineprotected areas, MPAs), while sea urchins are muchmore abundant on fished reefs where predatoryfishes have been removed (Mc Clanahan et al. 2007).Altering the abundance and types of grazers canhave subsequent effects on recruitment processes,including those of important foundational taxa, suchas corals. Reefs where both fish and sea urchin graz-

ers are abundant offer opportunities to investigatethe relative effects and the pathways of influence ofthese grazers (McClanahan & Muthiga 1989). Suchinformation is likely to become increasingly impor-tant for management of reefs, especially in the Carib -bean, as D. antillarum recovers and potentially be -comes the dominant grazer (Idjadi et al. 2010). Whileseveral studies have looked at the combined and dif-ferential effects of grazers on (1) removal of algae(McClanahan 1997), (2) benthic composition of reefs(Smith et al. 2010, Sjöö et al. 2011), and (3) erosion ofreefs (Peyrot-Clausade et al. 2000, Carreiro-Silva &McClanahan 2001), few empirical investigationshave distinguished between the effects of fish andsea urchin grazing on early coral recruitment (Map-stone et al. 2007).

Understanding the mechanisms through whichgrazing either promotes or inhibits coral recruitmentrequires an experimental design that isolates thegrazers’ influences at pre- and post-settlement stages.We took advantage of reefs in Kenya where fisheriesmanagement has created areas strongly dominatedby either fish or sea urchin grazers. We used anorthogonal (fully crossed) experimental design toquantify how fish and sea urchin grazers influencecoral recruitment indirectly (by changing availabilityof settlement habitat and competitive dynamics), anddirectly (through subsequent re moval of recruits).Based on previous findings, we hypothesized thatfish grazing would allow higher coral recruitmentthan sea urchin grazing, and we made 4 specificstage-dependent and mechanistic predictions:

(1) Fish and sea urchin grazing both increase spaceavailable for coral recruitment.

(2) Space cleared by fish versus sea urchin grazingdiffers in quality: sea urchins create more bare oralgal turf substrate, while fish create more crustosecoralline algal (CCA) substrate.

(3) Increased CCA cover increases coral recruit-ment.

(4) Post-settlement mortality of recruits is higherunder sea urchin than fish grazing, due to greaterdirect removal of newly settled corals by urchins.

MATERIALS AND METHODS

Study sites

The experimental sites were in the Kenyan backreef lagoon protected by a fringing reef spanning250 km of the coast from Malindi in the north to theTanzanian border in the south (Fig. 1) (McClanahan

O’Leary et al.: Comparing grazer effects on coral recruitment

& Arthur 2001). Most of the lagoon is

Mar Ecol Prog Ser 493: 165–177, 2013

using two 1-way ANOVAs with biomass as thedependent and fisheries management (fished orclosed) as the independent variable. The size distri-bution of Acanthuridae and Scarinae were evaluatedto determine the proportion of the population thatcould conceivably graze within experimental coralrecruitment cavities.

We used data from an annual monitoring programto estimate sea urchin density and biomass in 9 non-overlapping, circular 10 m2 areas at 2 locations perreef site (McClanahan & Shafir 1990) between Feb-ruary and March of 2007, and 2008 at each siteexcept Kisite, which was surveyed in February 2006and March 2008. The center of each area was deter-mined haphazardly by tossing a weight. Sea urchinswere counted and identified to species, and densitieswere averaged for each location at a site. Biomass(g m−2) of each species was estimated by multiplyingthe average density by the average wet mass of 20 to

200 haphazardly selected individuals per species(depending on abundance). Biomass samples werecollected only at fished reefs, as we have notobserved large differences in urchin size amongreefs (T. McClanahan unpubl. data). We tested fordifferences in sea urchin biomass using a 1-wayANOVA with biomass as the dependent and fisheriesmanagement as the independent variable.

Experimental design

We used a 2-stage sequential design to decouplethe effects of fish and sea urchin grazing on develop-ment of potential settlement substrates (Stage 1) fromthe effects of post-settlement mortality (Stage 2;Fig. 2). During the first 6 mo (Stage 1; April to Octo-ber 2007) benthic communities became establishedon experimental surfaces (PVC tiles) in 2 grazingtreatments (ungrazed and grazed) created by cagingin both fisheries management treatments (fished andunfished reefs). Fished reefs represent reefs wheresea urchins were the dominant grazers, and unfishedreefs represent reefs where fishes were the dominantgrazers. At the end of Stage 1, just before the maincoral recruitment season began, we recorded thecommunity present on tiles under each regime. Priorto Stage 2, half of the treatments at each site wereswapped: half of the cages around ungrazed (caged)treatments were opened to allow grazing and half ofthe grazed treatments were caged to prohibit graz-ing. The other half of treatments were left in theiroriginal state. During the 6 mo of Stage 2 (October2007 to April 2008), corals settled on the experimen-tal substrates and coral recruitment and substrate

168

Management Total Herbivorous Sea Site fish fish urchin

Fished reefsVipingo 204.2 ± 31.7 9.3 ± 1.1 2817.9 ± 307.5Kanamai 87.2 ± 9.0 9.5 ± 1.6 4551.5 ± 328.5Diani 43.3 ± 13.1 7.0 ± 5.6 3104.9 ± 1103.1Average 111.6 ± 23.1 8.6 ± 1.9 3464.3 ± 331.5

Fisheries closuresMalindi 1127.2 ± 186.8 221.4 ± 27.9 66.9 ± 20.6Mombasa 1411.5 ± 300.4 185.1 ± 26.4 1059.7 ± 168.3Kisite 916.4 ± 148.9 164.0 ± 45.6 636.1 ± 255.3Average 1151.7 ± 130.9 190.1 ± 19.3 587.6 ± 153.6

Table 1. Total biomass (kg ha−1 ± SE) of all fish (2007), herbi -vorous fish (2007), and all sea urchin species (2006 to 2008)at 6 study sites. Herbivorous fish are the 2 dominant herbi-

vorous families (Acanthuridae and Scarinae)

Fig. 2. Experimental design showing replication numbers

O’Leary et al.: Comparing grazer effects on coral recruitment

community composition were evaluated for eachtreatment at the end of Stage 2. The sequentialdesign made it possible to distinguish the effect ofavailable settlement substrate (established duringStage 1 treatments) on coral recruitment from theeffect of grazing on settled coral recruits (duringStage 2) under the 2 management systems domi-nated by different grazers (fishes and sea urchins).

The experimental surfaces were PVC tiles (10.3 ×15.7 cm and 2 mm thick) attached by a large plasticcable tie (zip tie) to the insides of cavities in concreteblocks (Fig. 2). Tiles were affixed to blocks with a ca-ble tie inserted through the hole in the tile and theconcrete block, with the head of the cable tie holdingthe tile in place (such that no cable tie lay across thetile), and with another cable tie attached outside theblock. The area of each tile available for settlementwas 159.7 cm2 (after accounting for a 5 mm diameterhole through which tiles were secured to blocks).Tiles were sanded before use to provide a rougher sur-face. Each block (20 × 25 × 50 cm) was made of con-crete poured into molds to create 2 cavities (13 × 13 cmopenings and 20 cm long) forming horizontal tunnelsthrough the block (Fig. 2). Holes for attaching tilesand cages were made during the molding process.Each block was attached using a 50/50 mixture of ce-ment and sand to a horizontal, dead, hard substratewithin coral-dominated habitat. Blocks were ~2 mapart and placed in a line parallel to the reef margin.

A preliminary experiment was conducted betweenJanuary and April 2006 using 10 cement blocks(16 cavities) at 6 sites (Fig. 1) to determine coral set-tlement orientation within block cavities. In the pre-liminary experiment, tiles were attached to the innertop, 1 inner side, and inner bottom of each cavity asdescribed above. Virtually all coral recruits werefound on tiles in the inner top (upside-down) andinner vertical orientations, a pattern consistent with ageneral tendency of corals to settle on the undersidesof overhangs (Wallace 1985, Raimondi & Morse 2000,Mangubhai et al. 2007, Price 2010). Thus, in the finalexperiment, 2 PVC tiles were attached to the innertop and one inner side of each cavity. During both thepreliminary and final experiments, both fish and seaurchins were observed grazing on the tiles withinuncaged cavities. On reefs closed to fishing (MPAs),parrotfish of multiple species and the red-lined trig-gerfish were the dominant fish observed entering thecavities, and these had an approximate maximumlength of 40 cm. On fished reefs, open cavities typi-cally had between 1 and 3 sea urchins within them.The most common sea urchin species observedwithin cavities was Diadema savignyi, with Echino -

metra ma thaei and D. setosum seen in cavities onlyoccasionally.

For the experiment (2007 and 2008), we deployed16 cement blocks (32 cavities) at each of the 6 sites inApril 2007 (Fig. 1), 6 mo before the main coral settle-ment season. During Stage 1, each block had 2 treat-ments: grazed and ungrazed. Grazed cavities wereopen at both ends and exposed to fishes and seaurchins. Ungrazed cavities were caged at each end toexclude large grazers, using plastic garden meshwith 2.5 × 2.5 cm2 openings. The mesh was fastenedto the block with zip ties through holes made in theblocks as they were cast. The mesh was cleanedinside and out monthly throughout Stages 1 and 2 tominimize fouling that could reduce light and watermovement within the caged treatments. At the endof Stage 1 (October 2007), we removed and photo -graphed each tile, then returned it to its cavity. Atthe end of Stage 2 (April 2008), we removed, photo -graphed, and air-dried all tiles and then stored themin individual plastic zip lock bags for subsequent lab-oratory analysis. For each Stage 1 plus Stage 2 graz-ing combination, the design provided 8 replicate cav-ities per treatment per reef.

Data collection

We used a point sampling method to estimate thearea of each tile with suitable coral settlement sub-strate (bare or CCA) from photographs taken at theend of Stages 1 and 2. We overlaid each tile photo-graph with a grid of 50 uniformly distributed points(using the program CPCe; Kohler & Gill 2006) andrecorded the organism type under each point. Weused the photographs of tiles taken after Stage 1 toestimate the initial area available for settlement (bareor CCA habitat) at the beginning of Stage 2. We usedthe photographs taken at the end of Stage 2 to esti-mate the final area of open space (bare and CCA)remaining at the end of the experiment on whichcorals could have survived without being overgrown.

Two independent observers searched each tile forcoral recruits under a dissecting microscope. Eachrecruit was identified to family following Babcocket al. (2003). We then lifted all organisms (exceptcoral and CCA) from the tiles, and searched forcoral recruits covered by encrusting organisms. Werecorded the type of organism overgrowing any cov-ered coral recruit and the substrate beneath eachcoral recruit. For each coral recruit, we assumed therecruit was alive at the time of collection if it wasuncovered, and that it was dead if it was overgrown.

169

Mar Ecol Prog Ser 493: 165–177, 2013170

Data analyses

In the subsequent analyses, weaveraged the number of recruitsand the substrate percent coverfor the 2 tiles within a singleopening, so that there were 16replicate treatments (caging ×stage) per site per stage (Fig. 2).Before analyses, total re cruits,live recruits, and recruit density(live recruits per open area afterStage 2) data were fourth-roottransformed to meet assumptionsof normality. The area of opensubstrate (bare and CCA) at theend of Stages 1 and 2 were eachsquare root transformed. We useda series of linear mixed models todetermine how fish and sea urchingrazing affected settlement sub-strate availability and post-settle-ment mortality. Al though cementblocks (with cavities) were evenlyspaced along the reef, to accountfor potential spatial variabilityalong the line of cement blocks,we established 4 statistical blocksper site that we used in the linear mixed models. Eachstatistical block, which we termed ‘group,’ was madeup of 4 adjacent cement blocks (8 cavities) that com-prised all possible Stage 1 and 2 caging combinations.

Coral settlement rates

We evaluated whether coral settlement was differ-ent between fisheries closures and fished reefsregardless of the presence of any grazers. We did thiswith a linear mixed model using the total number ofrecruits (live and dead) at the end of Stage 2 as thedependent variable, site nested within managementand group nested within site nested within manage-ment as a random factors, and management as afixed factor (Model 1; Table 2).

Effects of grazers on the amount and type of settlement substrate

We next tested whether grazers in the 2 manage-ment systems had different effects on the quantity ofsuitable settlement substrate (bare and CCA) as

measured at the end of Stage 1. We did this using alinear mixed model (Model 2; Table 2) with fisheriesmanagement system, Stage 1 treatment (grazed ver-sus ungrazed), and the management × treatmentinteraction as fixed factors, and other organismcover, site nested within management, and groupnested within site nested within management as ran-dom factors.

We then performed a linear regression analysisbetween the percent cover of CCA (dependent vari-able) and the percent cover of open space (indepen-dent variable) at the end of Stage 1. A linear relation-ship would indicate that the grazing treatmentsinfluenced the CCA percent cover proportionally tothe cover of open substrate.

Coral settlement as a function of habitat availability

We analyzed whether the number of recruits variedas a function of the habitat created during Stage 1.Because open space and CCA cover were stronglycorrelated, we used open space as the dependentvariable. We used a linear mixed model with the

Model 1: Coral recruitment in the 2 fisheries management systemsDependent variable: Number of recruits (all)Fixed factor: Fisheries management systemRandom factors: Site nested within management

Group nested within site nested within management

Model 2: Effects of grazers on available settlement space, end of Stage 1Dependent variable: Open space (CCA and bare combined) at end of Stage 1Fixed factors: Fisheries management system

Stage 1 treatment (caged/open)Management × treatment interaction

Random factors: Site nested within managementGroup nested within site nested within management

Model 3: Recruits as a function of open space after Stage 1Dependent variable: (a) Number of all recruits or (b) number of live recruitsFixed factor: Open space at the end of Stage 1Random factor: Site

Model 4: Factors associated with live recruit density, end of Stage 2Dependent variable: Live recruit density (no. m−2 open space at end of Stage 2)Fixed factors: Fisheries management system

Stage 1 treatmentStage 2 treatmentStage 1 treatment × managementStage 2 treatment × managementStage 1 treatment × Stage 2 treatmentStage 1 treatment × Stage 2 treatment × management

Random factor: Site nested within managementGroup nested within site nested within management

Table 2. Components of linear mixed models used in analyses

O’Leary et al.: Comparing grazer effects on coral recruitment

number of all coral recruits (live and dead) as thedependent variable, open space at the end of Stage 1as the fixed factor, and site as the random factor(Model 3; Table 2). We also repeated this analysisusing the number of live (uncovered) recruits as thedependent variable.

To determine whether coral recruits settled prefer-entially on CCA or bare substrate, we conducted a χ2

analysis. The expected distribution was determinedby multiplying the total number of recruits (found atthe end the experiment) by the percent cover of CCAand by the cover of bare habitats (measured at theend of Stage 1). For the observed distribution, wecounted the number of recruits on each substratetype—bare and CCA.

Importance of pre- and post-settlement factors incoral recruitment

To determine if differential post-settlement mortal-ity occurred during Stage 2, we needed to calculaterecruit density for the amount of available open sub-strate on which recruits could still be alive at the endof the experiment. The density was thus calculated asthe number of live recruits at the end of Stage 2divided by the final area of open space on tiles at theend of Stage 2. By using this density measure, weaccounted for differences in open space created inthe treatments, allowing us to evaluate whether anyof the experimental treatments (caging: grazer pres-ence, and management: type of grazers) in any stagehad effects on post-settlement mortality.

To determine what factors contributed to recruit-ment success (live recruit density), we used a linearmixed model (Model 4; Table 2) that evaluated all thefactors and combinations of factors within the exper-iment. The dependent variable was density of livecoral recruits; the fixed factors were managementsystem, Stage 1 treatment, Stage 2 treatment, Stage 1treatment × management, Stage 2 treatment × man-agement, and Stage 1 treatment × Stage 2 treatment;the random factors were site nested within manage-ment, and group nested within site nested withinmanagement.

RESULTS

Photographs of 39 tiles (9% of tiles) at the end ofStage 1 were unusable due to camera malfunction orfogging. When the tiles were collected at the end ofStage 2, 41 (11%) of the 384 tiles were missing or had

broken off the block. Missing tiles and photos weredistributed broadly across management regimes andtreatments (Fig. 2). Because we averaged data for 2tiles within each cement block cavity, where therewas 1 tile missing within a single cavity, we simplyused data for the remaining tile.

Fish and sea urchin populations

The biomass of all fish families was significantlyhigher on closed reefs than on fished reefs (ANOVAdf = 1,22, F = 62.3, p < 0.0001; Table 3). Similarly, thebiomass of the 2 dominant grazing fish families(Acanthuridae and Scarinae) was also significantlyhigher on closed than on fished reefs (ANOVAdf = 1,22, F = 87.2, p < 0.0001; Table 3). The size fre-quency distributions of the dominant fish grazer families in the system (Acanthuridae and Scarinae)showed that 95% of acanthurid grazers and 51% ofscarinid grazers within fisheries closures were

Mar Ecol Prog Ser 493: 165–177, 2013

The vast majority of all recruits were either Pocillo-poridae (58%) or Porites (40%). There were only 2Acroporidae, and 11 recruits could not be identified.Of the live recruits, 74% were Pocilloporidae and25% were Porites. Most coral recruits (89%) wereattached to CCA, and 10% of recruits were attachedto bare substrate. Four recruits were on Peyssonneliaspp., 2 on barnacles, and 2 on bryozoans. The samegeneral pattern held when we looked at only the liverecruits with 91% on CCA substrate. There was nostatistically significant difference in the total numberof recruits (live and covered) that settled on tiles bythe end of Stage 2 on fished reefs and in closed reefs(Model 1: df = 1,4, F = 4.22, p = 0.11). Variance incoral settlement was 9 times greater between sitesnested within management treatments than betweenadjacent groups of cavities (the 4 statistical blocksper site) nested within sites.

Effects of grazers on the amount and type of settlement substrate

There was no difference in the amount of opensubstrate created by the grazer types in the 2 man-agement systems during Stage 1 (Model 2; Table 4).However, grazing had a strong effect on open spacefor both grazer types (fisheries management systems,Model 2; Table 4), with more open substrate ingrazed treatments. Variance in open space was 2times greater between sites nested within manage-

172

(A) Overall recruit densityManagement Total recruits Live recruits Live recruits on

open substrate

Fished reefs 149 ± 23 31 ± 5 611 ± 143Fisheries closures 32 ± 6 11 ± 3 111 ± 28

(B and C) Recruit density and open substrateTreatment Total Live Open substrate (%) Live recruits after

Stage 1 Stage 2 recruits recruits After Stage 1 After Stage 2 Stage 2 (m−2 open)

(B) Fished reefsUngrazed Ungrazed 234 ± 61 44 ± 17 28 ± 15 15 ± 10 543 ± 156Ungrazed Grazed 67 ± 16 25 ± 12 27 ± 21 26 ± 18 481 ± 274Grazed Ungrazed 240 ± 61 32 ± 5 49 ± 25 19 ± 13 1054 ± 454Grazed Grazed 63 ± 17 23 ± 6 43 ± 23 30 ± 22 440 ± 214

(C) Fisheries closuresUngrazed Ungrazed 10 ± 4 3 ± 2 27 ± 13 17 ± 11 80 ± 63Ungrazed Grazed 57 ± 21 21 ± 8 34 ± 18 22 ± 11 191 ± 75Grazed Ungrazed 24 ± 8 7 ± 3 38 ± 18 22 ± 17 65 ± 34Grazed Grazed 35 ± 8 12 ± 4 42 ± 21 31 ± 14 106 ± 34

Table 3. Summary of data: (A) overall densities (m−2 ± SE) of coral recruits; (B and C) coral recruit densities and open substrate (% cover ± SE) in each Stage 1 × Stage 2 treatment on (B) fished reefs and in (C) fisheries closures

Fig. 3 Size frequency distribution of the 2 major groups ofgrazing fishes on Kenyan reefs (no. fish per 500 m2): (a)Acanthuridae (surgeonfishes) and (b) Scarinae (parrotfishes)

O’Leary et al.: Comparing grazer effects on coral recruitment

ment treatments than between groups of adjacentcavities nested within sites. There was a strong posi-tive correlation between the amount of open sub-strate and the amount of CCA substrate (linear re -gression analysis r = 0.93, p < 0.0001) indicating thatover the 6 mo of Stage 1, grazers created CCA andbare substrates in similar proportions.

Settlement as a function of habitat availability

There was no correlation between the total numberof coral recruits and the amount of open settlementsubstrate at the end of Stage 1 (Model 3a: df = 1,170,F = 0.10, p = 0.76), or between the number of liverecruits and open substrate at the end of Stage 1(Model 3b: df = 1,170, F = 2.53, p = 0.11). Coralrecruits were found more frequently than expectedby chance on CCA substrate than bare substrate (χ2 =363.1, df = 1, p < 0.001), with 89% of all recruits and91% of live recruits found on CCA substrate.

Importance of pre- and post-settlement factors in recruitment

When evaluating which experimental factors con-tributed to differences in the density of live coralrecruits at the end of the experiment (Model 4), theinteraction between the Stage 2 grazer treatmentand the fisheries management system was the onlysignificant factor (p = 0.002; Table 5). There was ahigher live recruit density in grazed treatments thanin ungrazed (caged) treatments in fisheries closureswith fish grazers, but there was a higher recruit den-sity in ungrazed treatments than in grazed treat-ments on fished reefs with sea urchin grazers (Fig. 4).Variation in the density of live recruits was approxi-mately 1000 times greater between sites nestedwithin management treatments than between groupsof adjacent cavities nested within sites.

DISCUSSION

Fish and sea urchin grazers can alter the amountand quality of substrate available for coral settle-ment, as well as the survival of newly settled recruits.As grazer abundance increases or grazer dominancechanges, grazer influences may shift from positiveand indirect (via promotion of settlement substrateand removal of space competitors) to direct and neg-ative (via removal of coral recruits). We used theexisting differences in the identity and abundancesof dominant grazers on fished and unfished reefs toexplore the differential influence of fish and seaurchin grazers on substratum and early coral settle-ment as well as subsequent survival. Fish and seaurchin grazers were equally effective in creatingcoral settlement habitat during the pre-settlementstage (Stage 1) but differentially influenced recruitdensities during the post-settlement stage (Stage 2).Coral recruit density was higher in the presence of

173

Effect df F p

Fisheries management 1,4 2.61 0.63Stage 1 treatment 1,152 44.67

Mar Ecol Prog Ser 493: 165–177, 2013

grazers when fish were the dominant grazers, buthigher in the absence of grazers when sea urchinswere the dominant grazers (Fig. 4). It appears that, atleast in Kenya, sea urchin grazing is detrimental tocoral recruitment due to the direct removal ofrecruits, but that fish grazing is beneficial overall viaan indirect pathway of removing other benthicorganisms (algal turf and invertebrates) that eitherprevent settlement or overgrow recruits.

Settlement space was not the driving factor of coralrecruitment within the time frame of this study. Graz-ing by fishes and sea urchins equally increased spaceavailable for coral settlement by the end of Stage 1,and there was no correlation between the amount ofopen space and number of coral recruits settled. Fur-thermore, there was no indication that grazer typealtered the ratio of CCA to bare substrate withincleared areas. Our previous work, investigating 18 yrof altered grazing regimes in fished and closed reefs,contrasts with the results of this short-term study.Over longer time periods, sea urchin grazers greatlyreduced the amount of CCA on fished Kenyan reefs(O’Leary & McClanahan 2010). One possible expla-nation for this discrepancy is that this experimentwas too short to adequately test the pre-settlementeffects of grazers in establishing habitat becauseCCA grow slowly and therefore take much longer tobecome well established (McClanahan 1997). Ourexperiment may best represent recruit responses tonewly cleared substrate rather than to climax sub-strate conditions that develop over long time periods.In a Caribbean coral settlement study with minimalgrazing, there was a recruitment window of 9 to14 mo after open space was created (longer than ourpre-settlement Stage 1) where habitat for coral set-tlement was optimal (Arnold & Steneck 2011).

Though CCA abundance did not have an influenceon coral recruitment, CCA was an important settle-ment habitat: the majority of coral recruits (~90%)settled on CCA, regardless of grazer presence oridentity. Similar results have been reported in a num-ber of other studies (Raimondi & Morse 2000, Har-rington et al. 2004, Ritson-Williams et al. 2010,O’Leary et al. 2012). Our previous work on naturalsubstratum in Kenya also found that corals werestrongly associated with CCA in general, and specif-ically with 2 species of CCA in the genus Hydrolithonthat are known to induce settlement of invertebratelarvae (O’Leary et al. 2012). In our previous work,coral recruitment was correlated with the cover ofCCA. In this experiment, while most corals recruitedto CCA, coral recruitment was not correlated withCCA cover. It may be that presence of inductive CCA

taxa rather than absolute abundance facilitates coralrecruitment. We suggest that over reef scales, induc-tive CCA are sparse and infrequently encounteredby settling recruits, so that higher CCA cover leads togreater recruitment. Over small spatial scales, wherethe amount of CCA is not limiting, increasing coverof CCA may not induce further recruitment. Thiscould be an important finding for management ofgrazing communities on reefs to optimize coralrecruitment, and suggests the need for further con-trolled experiments testing whether CCA cover orthe presence of inductive CCA is more important.

Fish and sea urchin grazing had significantly dif-ferent influences on the post-settlement mortality ofcorals. The density of coral recruits was higher in thepresence of grazers (fishes) within fisheries closures,but was higher in the absence of grazers (seaurchins) on fished reefs. Furthermore, the spatial dis-tribution of grazing pressure within sites was largelyuniform regardless of grazer type. There was muchgreater variability in grazing between sites withinmanagement systems than on smaller scales (of~16 m between groups of cavities within sites). Forfish grazers, this finding is consistent with our previ-ous work that showed spatial variation in fish grazingwas greatest at scales of at least 1000 m (O’Leary &Potts 2011). These results suggest that the mecha-nisms by which grazers affected coral survival weredifferent. On Kenyan reefs, sea urchin grazingappears to be detrimental to coral recruitment bydirectly causing mortality of juvenile corals, whilefish grazing is apparently beneficial — possibly indi-rectly by removing other benthic organisms (algalturf and invertebrates) that overgrow coral recruits.Sea urchins are less selective and more intense graz-ers than fish: they feed in a swath and ingest non-algal materials while feeding (Lewis 1964, McClana-han 1992, Bak 1994), likely causing higher recruitremoval. Coral recruit removal probably continues tobe high on urchin dominated-fished reefs for at leastthe first 1 to 2 yr after settlement. In field studieslooking at coral recruits up to 30 mm size, recruitdensity on fished reefs was half that in closed reefs,and very low overall (

O’Leary et al.: Comparing grazer effects on coral recruitment

tality, either by settling later or by growing fasterthan poritids. Arnold & Steneck (2011) found thatPorites was at a competitive disadvantage in resistingovergrowth compared to other corals, and attributedthis to slow growth rates. This may mean that somejuvenile corals benefit disproportionally from re -duced competition under higher grazing levels. Theresponse of coral recruitment to changes in grazinglikely depends on the type and biomass of grazers,the abundance of benthic competitors, and the domi-nant species of coral recruits.

Literature from various regions with different graz-ing abundances may help identify grazer density orbiomass switch points where grazer effects on coralrecruitment change from indirect and positive todirect and negative. Over the last few decades, fishbiomass on Kenyan reefs and possibly throughout theWestern Indian Ocean has been higher than in theCaribbean (McClanahan et al. 2009). With a typicalherbivore biomass of 500 to 700 kg ha−1 in Kenyanfisheries closures, grazing fish have maintained acropped substrate with 12 cm total length) ranged from20 kg ha−1 in Jamaica to a high of 170 kg ha−1 in Bar-bados (Williams & Polunin 2001). Even where fish bio-mass was highest, fish maintained only ~40 to 60% ofreef substratum in a cropped state (Williams &Polunin 2001), and many Caribbean reefs shifted fromcoral dominance to fleshy algal dominance followingthe loss of the sea urchin D. antillarum (Lessios 1988,Hughes 1994). Furthermore, the importance of largefish grazers, particularly parrotfishes (Scarinae), hasbeen emphasized in the literature (Lokrantz et al.2008, Mumby et al. 2012). Our data suggest that incryptic habitat where corals prefer to settle, smallergrazing fishes (16 m−2

(Sammarco 1980, 1982). This effect may be less markedfor smaller urchins, such as Echinometra viridis, whichhad no effect on post-settlement mortality at densitiesup to 50 m−2 (Sammarco 1982). In our Kenyan experi-ment, D. savignyi was the most common urchin withinexperimental treatments on fished reefs, and it seemsto have strong effects on post-settlement mortality atrelatively low densities of ~5 m−2.

Understanding density-dependent thresholds ofurchin predation on corals may be critical for manag-ing fisheries on coral reefs. Declines in herbivorousfish populations may be offset to some extent by in-creased echinoid grazing that benefits corals in theshort-term by limiting overgrowth by fleshy algae orbenthic invertebrates. However, at least one otherstudy has shown that herbivorous fish are more effi-cient at removing fleshy algae than sea urchins, evenwhen fish biomass was much lower than that ofurchins (Jessen & Wild 2013). Furthermore, caution isneeded when making long-term projections because,in both the Caribbean (Brown-Saracino et al. 2007)and East Africa (McClanahan 1997, 1999, McClana-han et al. 2007), removal of predatory fishes that eatsea urchins has resulted in high abundances of seaurchins that are eroding the reef structure (Carreiro-Silva & McClanahan 2001, 2012), reducing CCA settlement habitat (O’Leary & McClanahan 2010,O’Leary et al. 2012), and reducing coral recruit sur-vival. In Kenya, both fish and sea urchin grazer densi-ties seem to be above thresholds needed to maintainspace free of fleshy algae, particularly during thecooler and less productive southeast monsoon season(McClanahan 1997). Our study suggests that main-taining higher grazing fish densities is likely to havethe most beneficial effects on coral ecology by pro-moting coral recruitment. However, managing fish-eries to increase the abundance of sea urchin preda-tors while also restricting the capture of herbivorousfishes may reduce fisheries yields (McClanahan 1995).Evaluating the tradeoffs posed by this fisheries–ecology conflict remains an important area for futureinvestigation.

175

Mar Ecol Prog Ser 493: 165–177, 2013176

Acknowledgements. We thank the University of California atSanta Cruz, Robert and Patricia Switzer Foundation, ARCSFoundation, Project AWARE, and the Wildlife ConservationSociety for financial support. Research clearance was pro-vided by Kenya’s Ministry of Science and Technology andpermission to work in the parks by the Kenya Wildlife Serv-ice. We thank J. Mariara, B. Wong, M. Goodman, C. Bednarand numerous associates who assisted with the field experi-ment and with laboratory sample processing. We thank P.Raimondi, M. Carr, J. Pearse, and F. Micheli for advice onexperimental design and P. Raimondi for statistical advice.

LITERATURE CITED

Arnold SN, Steneck RS (2011) Settling into an increasinglyhostile world: the rapidly closing “recruitment window”for corals. PLoS ONE 6: e28681

Babcock RC, Kelly S, Shears NT, Walker JW, Willis TJ (1999)Changes in community structure in temperate marinereserves. Mar Ecol Prog Ser 189: 125−134

Babcock RC, Baird AH, Piromvaragorn S, Thomson DP,Willis BL (2003) Identification of scleractinian coralrecruits from Indo-Pacific reefs. Zool Stud 42: 211−226

Bak RPM (1994) Sea-urchin bioerosion on coral reefs: placein the carbonate budget and relevant variables. CoralReefs 13: 99−103

Brown-Saracino J, Peckol P, Curran HL, Robbart ML (2007)Spatial variation in sea urchins, fish predators, and bio-erosion rates on coral reefs of Belize. Coral Reefs 26: 71−78

Carpenter RC, Edmunds PJ (2006) Local and regional scalerecovery of Diadema promotes recruitment of scleractin-ian corals. Ecol Lett 9: 271−280

Carreiro-Silva M, McClanahan TR (2001) Echinoid bioero-sion and herbivory on Kenyan coral reefs: the role of pro-tection from fishing. J Exp Mar Biol Ecol 262: 133−153

Carreiro-Silva M, McClanahan TR (2012) Macrobioerosionof dead branching Porites, 4 and 6 years after coral massmortality. Mar Ecol Prog Ser 458: 103−122

Coma R, Serrano E, Linares C, Ribes M, Diaz D, BallasterosE (2011) Sea urchins predation facilitates coral invasionin a marine reserve. PLoS ONE 6: e22017

Connell JH (1961) The influence of interspecific competitionand other factors on the distribution of the barnacleChthamalus stellatus. Ecology 42:710–723

Connell JH (1985) The consequences of variation in initialsettlement vs. post-settlement mortality in rocky inter-tidal communities. J Exp Mar Biol Ecol 93: 11−45

Cover WA (2011) Species interactions affecting corals andrecruitment on a protected, high-latitude reef: herbivory,predation, and competition by fishes, urchins, macro-algae and cyanobacteria. PhD dissertation, University ofCalifornia, Santa Cruz, CA

Day E, Branch GM (2002) Effects of sea urchins (Parechinusangulosus) on recruits and juveniles of abalone (Haliotismidae). Ecol Monogr 72: 133−149

Dayton PK (1971) Competition, disturbance, and communityorganization: provision and subsequent utilization ofspace in a rocky intertidal community. Ecol Monogr 41: 351−389

Dee LE, Witman JD, Brandt M (2012) Refugia and top-downcontrol of the pencil urchin Eucidaris galapagensis inthe Galapagos Marine Reserve. J Exp Mar Biol Ecol416−417: 135−143

Duffy JE, Richardson JP, Canuel EA (2003) Grazer diversity

effects on ecosystem functioning in seagrass beds. EcolLett 6: 637−645

Edmunds PJ, Carpenter RC (2001) Recovery of Diademaantillarum reduces macroalgal cover and increasesabundance of juvenile corals on a Caribbean reef. ProcNatl Acad Sci USA 98: 5067−5071

Harborne AR, Renaud PG, Tyler EHM, Mumby PJ (2009)Reduced density of the herbivorous urchin Diademaantillarum inside a Caribbean marine reserve linked toincreased predation pressure by fishes. Coral Reefs 28: 783−791

Harrington K, Fabricius J, De’ath G, Negri AP (2004) Recog-nition and selection of settlement substrata determinepost-settlement survival in corals. Ecology 85: 3428−3437

Heyward AJ, Negri AP (1999) Natural inducers for coral lar-val metamorphosis. Coral Reefs 18: 273−279

Hughes TP (1994) Catastrophes, phase-shifts, and large-scale degradation of a Caribbean coral-reef. Science 265: 1547−1551

Hughes TP, Rodrigues MJ, Ceccarelli D, Hoegh-Guldberg Oand others (2007) Phase shifts, herbivory, and the resili-ence of coral reefs to climate change. Curr Biol 17: 360−365

Hunt HL, Scheibling RE (1997) Role of early post-settlementmortality in recruitment of benthic marine invertebrates.Mar Ecol Prog Ser 155: 269−301

Idjadi JA, Haring RN, Precht WF (2010) Recovery of the seaurchin Diadema antillarum promotes scleractinian coralgrowth and survivorship on shallow Jamaican reefs. MarEcol Prog Ser 403: 91−100

Jessen C, Wild C (2013) Herbivory effects on benthic algalcomposition and growth on a coral reef flat in the Egypt-ian Red Sea. Mar Ecol Prog Ser 476: 9−21

Kohler KE, Gill SM (2006) Coral Point Count with Excelextensions (CPCe): a Visual Basic program for the deter-mination of coral and substrate coverage using randompoint count methodology. Comput Geosci 32: 1259−1269

Knowlton N (2001) Sea urchin recovery from mass mortality: New hope for Caribbean coral reefs? Proc Natl Acad SciUSA 98: 4822−4824

Lessios HA (1988) Mass mortality of Diadema antillarum inthe Caribbean: What have we learned? Annu Rev EcolSyst 19: 371−393

Lewis JB (1964) Feeding and digestion in tropical sea urchinDiadema antillarum Philippi. Can J Zool 42: 549−557

Lokrantz J, Nystrom M, Thyresson M, Johnansson C (2008)The non-linear relationship between body size and func-tion in parrotfishes. Coral Reefs 27: 967−974

Mangubhai S (2009) Reproductive ecology of the scleractin-ian corals Echinopora gemmacea and Leptoria phrygia(Faviidae) on equatorial reefs in Kenya. Invertebr ReprodDev 53: 67−79

Mangubhai S, Harrison PL (2008) Gametogenesis, spawningand fecundity of Platygyra daedalea (Scleractinia) onequatorial reefs in Kenya. Coral Reefs 27: 117−122

Mangubhai S, Harrison PL (2009) Extended breeding sea-sons and asynchronous spawning among equatorial reefcorals in Kenya. Mar Ecol Prog Ser 374: 305−310

Mangubhai S, Harrison PL, Obura DO (2007) Patterns ofcoral larval settlement on lagoon reefs in the MombasaMarine National Park and Reserve, Kenya. Mar EcolProg Ser 348: 149−159

Mapstone BD, Andrew NL, Chancerelle Y, Salvat B (2007)Mediating effects of sea urchins on interactions amongcorals, algae and herbivorous fish in the Moorea lagoon,French Polynesia. Mar Ecol Prog Ser 332: 143−153

http://dx.doi.org/10.3354/meps332143http://dx.doi.org/10.3354/meps07090http://dx.doi.org/10.3354/meps07910http://dx.doi.org/10.1007/s00338-007-0297-8http://dx.doi.org/10.1080/07924259.2009.9652292http://dx.doi.org/10.1007/s00338-008-0394-3http://dx.doi.org/10.1139/z64-048http://dx.doi.org/10.1073/pnas.091107198http://dx.doi.org/10.1016/j.cageo.2005.11.009http://dx.doi.org/10.3354/meps10157http://dx.doi.org/10.3354/meps08463http://dx.doi.org/10.3354/meps155269http://dx.doi.org/10.1016/j.cub.2006.12.049http://dx.doi.org/10.1126/science.265.5178.1547http://dx.doi.org/10.1007/s003380050193http://dx.doi.org/10.1890/04-0298http://dx.doi.org/10.1007/s00338-009-0516-6http://dx.doi.org/10.1073/pnas.071524598http://dx.doi.org/10.1046/j.1461-0248.2003.00474.xhttp://dx.doi.org/10.1016/j.jembe.2012.02.016http://dx.doi.org/10.2307/1948498http://dx.doi.org/10.1016/0022-0981(85)90146-7http://dx.doi.org/10.1371/journal.pone.0022017http://dx.doi.org/10.3354/meps09726http://dx.doi.org/10.1016/S0022-0981(01)00288-Xhttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=16958892&dopt=Abstracthttp://dx.doi.org/10.1007/s00338-006-0159-9http://dx.doi.org/10.1007/BF00395440http://dx.doi.org/10.1007/BF00300768http://dx.doi.org/10.3354/meps189125http://dx.doi.org/10.2307/2937107http://dx.doi.org/10.1371/journal.pone.0028681http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=22194883&dopt=Abstract

O’Leary et al.: Comparing grazer effects on coral recruitment

McCauley DJ, Micheli F, Young HS, Tittensor DP and others(2010) Acute effects of removing large fish from a near-pristine coral reef. Mar Biol 157: 2739−2750

McClanahan TR (1988) Seasonality in East Africa’s coastalwaters. Mar Ecol Prog Ser 44: 191−199

McClanahan TR (1992) Resource utilization, competitionand predation: a model and example from coral reefgrazers. Ecol Model 61: 195−215

McClanahan TR (1995) A coral reef ecosystem fisheriesmodel: impacts of fishing intensity and catch selection onreef structure and processes. Ecol Model 80: 1−19

McClanahan TR (1997) Primary succession of coral-reefalgae differing patterns on fished versus unfished reefs.J Exp Mar Biol Ecol 218: 77−102

McClanahan TR (1999) Predation and the control of the seaurchin Echinometra viridis and fleshy algae in the patchreefs of Glovers Reef, Belize. Ecosystems 2: 511−523

McClanahan TR (2000) Recovery of a coral reef keystonepredator, Balistapus undulatus, in East African marineparks. Biol Conserv 94: 191−198

McClanahan TR (2008) Response of the coral reef benthosand herbivory to fishery closure management and the1998 ENSO disturbance. Oecologia 155: 169−177

McClanahan TR, Arthur R (2001) The effect of marinereserves and habitat on populations of east African coralreef fishes. Ecol Appl 11: 559−569

McClanahan TR, Kaunda-Arara B (1996) Fishery recovery ina coral-reef marine park and its effect on the adjacentfishery. Conserv Biol 10: 1187−1199

McClanahan TR, Muthiga NA (1989) Patterns of predationon sea urchin, Echinometra mathaei (Deblainville), onKenyan coral reefs. J Exp Mar Biol Ecol 126:77–94

McClanahan TR, Shafir SH (1990) Causes and consequencesof sea urchin abundance and diversity in Kenyan coralreef lagoons. Oecologia 83: 362−370

McClanahan TR, Nugues M, Mwachireya S (1994) Fish andsea urchin herbivory and competition in Kenyan coralreef lagoons: the role of reef management. J Exp MarBiol Ecol 184: 237−254

McClanahan TR, Graham NAJ, Calnan JM, MacNeil MA(2007) Toward pristine biomass: reef fish recovery incoral reef marine protected areas in Kenya. Ecol Appl 17: 1055−1067

McClanahan TR, Graham NAJ, Wilson SK, Letourneur Y,Fisher R (2009) Effects of fisheries closure size, age, andhistory of compliance on coral reef fish communities inthe western Indian Ocean. Mar Ecol Prog Ser 396: 99−109

Miner CM, Altstatt JM, Raimondi PT, Michinton TE (2006)Recruitment failure and shifts in community structurefollowing mass mortality limit recovery prospects ofblack abalone. Mar Ecol Prog Ser 327: 107−117

Morse DE, Morse ANC, Raimondi PT, Hooker N (1994) Mor-phogen-based chemical flypaper for Agaricia humiliscoral larvae. Biol Bull (Woods Hole) 186: 172−181

Mumby PJ (2006) The impact of exploiting grazers (scari-dae) on the dynamics of Caribbean coral reefs. Ecol Appl16: 747−769

Mumby PJ (2009) Phase shifts and the stability of macroalgalcommunities on Caribbean coral reefs. Coral Reefs 28: 761−773

Mumby PJ, Harborne AR, Williams J, Kappel CV and others(2007) Trophic cascade facilitates coral recruitment in amarine reserve. Proc Natl Acad Sci USA 104: 8362−8367

Mumby PJ, Steneck RS, Edwards AJ, Ferrari R, Coleman R,Harborne AR, Gibson JP (2012) Fishing down a Carib -

bean food web relaxes trophic cascades. Mar Ecol ProgSer 445: 13−24

O’Connor NE, Donohue I, Crowe TP, Emmerson MC (2011)Importance of consumers on exposed and shelteredrocky shores. Mar Ecol Prog Ser 443: 65−67

O’Leary JK, McClanahan TR (2010) Trophic cascades resultin large-scale coralline algae loss through differentialgrazer effects. Ecology 91: 3584−3597

O’Leary JK, Potts DC (2011) Using hierarchical sampling tounderstand scales of spatial variation in early coralrecruitment. Coral Reefs 30: 1013−1023

O’Leary JK, Potts DC, Braga JC, McClanahan TR (2012)Indirect consequences of fishing: reduction of corallinealgae suppresses juvenile coral abundance. Coral Reefs31: 547−559

Penin L, Michonneau F, Carrol A, Adjeroud M (2011) Effectsof predator and grazer exclusion on early post-settlementmortality. Hydrobiologia 663: 259−264

Peyrot-Clausade M, Chabanet P, Conand C, Fontaine MF,Letourneur Y, Harmelin-Viven M (2000) Sea urchin andfish bioerosion on La Reunion and Moorea reefs. BullMar Sci 66: 477−485

Price N (2010) Habitat selection, facilitation, and biotic set-tlement cues affect distribution and performance of coralrecruits in French Polynesia. Oecologia 163: 747−758

Raimondi PT, Morse ANC (2000) The consequences of com-plex larval behavior in a coral. Ecology 81: 3193−3211

Ritson-Williams R, Paul VJ, Arnold SN, Steneck RS (2010)Larval settlement preferences and post-settlement sur-vival of the threatened Caribbean corals Acroporapalmata and A. cervicornis. Coral Reefs 29: 71−81

Rotjan RD, Lewis SM (2008) Impact of coral predators ontropical reefs. Mar Ecol Prog Ser 367: 73−91

Salomon AK, Tanape NM, Huntington HP (2007) Serialdepletion of marine invertebrates leads to the decline ofa strongly interacting grazer. Ecol Appl 17: 1752−1770

Sammarco PW (1980) Diadema and its relationship to coralspat mortality: grazing, competition, and biological dis-turbance. J Exp Mar Biol Ecol 45: 245−272

Sammarco PW (1982) Echinoid grazing as a structuring forcein coral communities: whole reef manipulations. J ExpMar Biol Ecol 61: 31−55

Sjöö GL, Mork E, Andersson S, Melander I (2011) Differ-ences in top-down and bottom-up regulation of macro-algal communities between a reef crest and back reefhabitat in Zanzibar. Estuar Coast Shelf Sci 91: 511−518

Smith JE, Hunter CL, Smith C (2010) The effects of top-down versus bottom-up control on benthic coral reefcommunity structure. Oecologia 163: 497−507

Stimson J (1970) Territorial behavior of owl limpet, Lottiagigantea. Ecology 51: 113−118

Vermeij MJA, Dailer ML, Walsh SM, Donaan MK, Smith CM(2010) The effects of trophic interactions and spatial com-petition on algal community composition on Hawaiiancoral reefs. Mar Ecol 31: 291−299

Wallace CC (1985) Reproduction, recruitment and fragmen-tation in nine sympatric species of the coral genus Acro-pora. Mar Biol 88: 217−233

Williams ID, Polunin NVC (2001) Large-scale associationsbetween macroalgal cover and grazer biomass on mid-depth reefs in the Caribbean. Coral Reefs 19: 358−366

Wright JT, Dworjanyn SA, Rogers CN, Steinberg PD,Williamson JE, Poore AGB (2005) Density-dependent seaurchin grazing: differential removal of species, changesin community composition and alternative communitystates. Mar Ecol Prog Ser 298: 143−156

177

Editorial responsibility: Christine Paetzold,Oldendorf/Luhe, Germany

Submitted: April 15, 2013; Accepted: August 8, 2013Proofs received from author(s): November 12, 2013

http://dx.doi.org/10.3354/meps298143http://dx.doi.org/10.1007/BF00392585http://dx.doi.org/10.1111/j.1439-0485.2009.00343.xhttp://dx.doi.org/10.2307/1933604http://dx.doi.org/10.1007/s00442-009-1546-zhttp://dx.doi.org/10.1016/j.ecss.2010.12.003http://dx.doi.org/10.1016/0022-0981(82)90020-Xhttp://dx.doi.org/10.1016/0022-0981(80)90061-1http://dx.doi.org/10.1890/06-1369.1http://dx.doi.org/10.3354/meps07531http://dx.doi.org/10.3354/meps306103http://dx.doi.org/10.1007/s00338-009-0555-zhttp://dx.doi.org/10.1890/0012-9658(2000)081[3193%3ATCOCLB]2.0.CO%3B2http://dx.doi.org/10.1007/s00442-010-1578-4http://dx.doi.org/10.1007/s10750-010-0569-0http://dx.doi.org/10.1007/s00338-012-0872-5http://dx.doi.org/10.1007/s00338-011-0789-4http://dx.doi.org/10.1890/09-2059.1http://dx.doi.org/10.3354/meps09412http://dx.doi.org/10.3354/meps09450http://dx.doi.org/10.1073/pnas.0702602104http://dx.doi.org/10.1007/s00338-009-0506-8http://dx.doi.org/10.1890/1051-0761(2006)016[0747%3ATIOEGS]2.0.CO%3B2http://dx.doi.org/10.2307/1542051http://dx.doi.org/10.2307/1941634http://dx.doi.org/10.3354/meps327107http://dx.doi.org/10.1016/j.jembe.2007.05.014http://dx.doi.org/10.1007/s004420050373http://dx.doi.org/10.1007/s003380000129http://dx.doi.org/10.1016/0022-0981(94)90007-8http://dx.doi.org/10.1007/s00338-004-0471-1http://dx.doi.org/10.1046/j.1523-1739.1996.10041187.xhttp://dx.doi.org/10.3354/meps08279http://dx.doi.org/10.1890/06-1450http://dx.doi.org/10.1890/1051-0761(2001)011[0559%3ATEOMRA]2.0.CO%3B2http://dx.doi.org/10.1007/s00442-007-0890-0http://dx.doi.org/10.1016/S0006-3207(99)00176-7http://dx.doi.org/10.1007/s100219900099http://dx.doi.org/10.1016/S0022-0981(97)00069-5http://dx.doi.org/10.1016/0304-3800(94)00042-Ghttp://dx.doi.org/10.1016/0304-3800(92)90018-Ahttp://dx.doi.org/10.3354/meps044191http://dx.doi.org/10.1007/s00227-010-1533-2

cite10: cite14: cite21: cite16: cite30: cite32: cite27: cite41: cite4: cite43: cite36: cite8: cite38: cite52: cite34: cite61: cite45: cite29: cite70: cite49: cite63: cite72: cite67: cite81: cite76: cite83: cite90: cite87: cite1: cite89: cite5: cite11: cite13: cite20: cite22: cite15: cite24: cite17: cite2: cite26: cite40: cite19: cite28: cite42: cite35: cite37: cite51: cite33: cite39: cite53: cite46: cite55: cite48: cite44: cite64: cite71: cite66: cite59: cite73: cite68: cite75: cite82: cite77: cite84: cite79: cite86: cite88: cite7: