MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 546: 61–74, 2016doi: 10.3354/meps11657

Published March 21

INTRODUCTION

The relative inaccessibility of oceanic coral reefecosystems has limited our understanding of whethercoral communities in these systems are structured inthe same way as continental reefs. If they are, then anumber of relatively predictable biotic and hydro -dynamic factors are likely to influence coral zonationpatterns. For example, physical disturbance fromwave energy and light availability are important se-lective pressures that lead to characteristic and direc-tional patterns of diversity and structure across depthand exposure gradients (Connell 1978, Huston 1985).

Generally, in shallow habitats (<5 m), coral speciesdiversity is low and increases with depths to 15−30 mbefore decreasing again (Done 1983, Karlson 1999).Diversity is lowest near the ocean surface due towave-induced stress. Corals are particularly vulnera-ble to physical breakage and thus diversity and mor-phological variation is limited in areas that receiveconstant swell (e.g. exposed reef flats) (Madin & Con-nolly 2006). Storms or cyclones can also cause varioustypes of damage to corals through mechanical destruc-tion, scouring and abrasion to changes in sedimenta-tion and increases in turbidity (Harmelin-Vivien 1994).However, the responses of coral communities to phys-

© Inter-Research 2016 · www.int-res.com*Corresponding author: zoe.richards@museum.wa.gov.au

Diversity on the edge: non-linear patterns of coralcommunity structure at an isolated oceanic island

Zoe T. Richards1,*, Nicole M. Ryan2, Euan S. Harvey3, Rodrigo Garcia4, Jean-Paul A. Hobbs3

1Department of Aquatic Zoology, Western Australian Museum, 49 Kew Street, Welshpool, WA 6106, Australia2The Oceans Institute and School of Plant Biology, University of Western Australia, Crawley, WA 6009, Australia

3Department of Environment and Agriculture, Curtin University, Bentley, WA 6102, Australia4School for the Environment, University of Massachusetts, Boston, MA 02125, USA

ABSTRACT: Coral communities are expected to display predictable patterns of diversity andstructure across depth and exposure gradients, yet these predictions have rarely been tested atoceanic locations. Here we tested 5 common ecological predictions about coral community struc-ture at the remote Chrismas Island in the eastern Indian Ocean. Our results suggest that not all ofthe predictions hold true at this oceanic location, primarily because the community is structured innon-linear ways. Each surveyed depth zone (5, 12, 20 m) supported a distinctive community, yetspecies diversity and coral abundance did not show the expected linear increase with depth.Habitat complexity was also shown to respond non-linearly with depth, with the highest habitatcomplexity occurring in the intermediate 12 m zone. We discuss the ‘cliff-edge effect’ as a possibleexplanation for the high diversity and abundance of corals at 20 m, while physical stress and com-petitive exclusion may explain the low diversity at 12 m. The cliff-edge habitat provides a narrowzone in between the wave-swept shallows and the low light/high shade environment of the steepouter reef walls, and this zone of heterogeneous environmental conditions supports a wide diver-sity of corals. If future storm events, bleaching, disease or predator outbreaks were to impact thecorals living in the cliff-edge habitat, this may have a disproportionate impact on the coral reefcommunity as a whole. Monitoring the status of corals on the cliff edge is important for under-standing and predicting how oceanic reef systems will be affected by climate change.

KEY WORDS: Abundance · Christmas Island · Cliff-edge effect · Community ecology · Diversity ·Habitat complexity · Indian Ocean

Resale or republication not permitted without written consent of the publisher

This authors' personal copy may not be publicly or systematically copied or distributed, or posted on the Open Web, except with written permission of the copyright holder(s). It may be distributed to interested individuals on request.

Mar Ecol Prog Ser 546: 61–74, 2016

ical disturbances can differ because a range of diver-sity−disturbance relationships are possible dependingon the frequency and intensity of disturbances andangle and profile of the reef (Woodley et al. 1981,Done 1982, Huston 1985, Hall et al. 2012).

High wave energy also influences colony morphol-ogy whereby coral communities on shallow, exposedreefs are dominated by robust wave-tolerant specieswhich exhibit simple encrusting, digitate and mas-sive morphologies (Adjeroud et al. 2009, Williams etal. 2013, Madin et al. 2014). In contrast, on deeper orsheltered reefs, a reduction in wave exposure allowscorals to grow into large and/or complex ‘canopy-forming’ morphologies (Goatley & Bellwood 2011). Inaddition, light becomes increasingly limited withdepth, which enables shade-tolerant species to dom-inate on deeper reefs and grow as large flat horizon-tal plates (Wallace 1999, Veron 2000). Spatial gradi-ents in wave energy and light also affect habitatcomplexity, which is defined by the morphology ofindividual corals and the complexity of the under -lying topographic reef structure (Wilson et al. 2007,Alvarez-Filip et al. 2009). Consequently, there is aclose association between some morphologies (e.g.branching coral) and habitat complexity (Bergman etal. 2000, Graham & Nash 2013).

Coral colony abundance (i.e. the total abundanceof all species) follows a similar pattern to diversitywhereby it increases with depth (peaking at ~30 m)and decreases with exposure. This pattern is largelycaused by wave action that decimates large or fragilecoral colonies in shallow and exposed habitats(Madin 2005, Madin et al. 2014). However, colonyabundance is also influenced by biological inter -actions that are not strictly depth-dependent, such aspredation, herbivory and disease (Connell et al. 1997,Rudolf & Antonovics 2005). For example, the feedingpreferences of corallivorous predators, such as thecrown-of-thorns starfish Acanthaster planci, are notgreatly influenced by depth (De’ath & Moran 1998).Grazing by herbivorous fishes influences competitiveinteractions between corals and macroalgae (Lirman2001), and the removal of herbivores can lead tophase shifts that are not depth restricted (Russ 2003,Fox & Bellwood 2007). Further, coral disease is influ-enced by demographic and environmental factors,but depth does not seem to be the most importantfactor (Bruno et al. 2007, Hobbs & Frisch 2010).

Due to climate change, the frequency and intensityof disturbance events is predicted to increase(Emanuel 2005). Consequently, Madin et al. (2008)have predicted that future reefs will be dominated bysmall, morphologically simple and robust coral spe-

cies with low abundance. Branching coral speciessuch as Acropora are particularly susceptible to dis-turbance because they often have limited attachmentto the substrate (Marshall & Baird 2000, Carpenter etal. 2008). As a result, Acropora-dominated communi-ties, particularly those growing in sheltered habitats,can be severely damaged when exposed to rarestorm events (Jackson 1979, Madin & Connolly 2006,Fabricius et al. 2008). Furthermore, on contemporaryreefs, the range and velocity of disturbances occur-ring at depth is increasing (e.g. coral bleaching,Hoegh-Guldberg & Salvat 1995, Penin et al. 2007;cyclones, Bongaerts et al. 2013; and corallivore pre-dation, Kayal et al. 2012); hence, those species thathave previously found refuge at depth may also bemore vulnerable in the future.

Detecting change in coral communities is crucial,but documenting patterns in coral community com-position has been limited by difficulties in identifyingcorals to species level. Valuable insights have beengained from species level studies conducted on con-tinental reefs (e.g. Great Barrier Reef, Connell et al.1997, De Vantier et al. 2006, Wakeford et al. 2008).While some studies have examined spatial patternsin the benthic composition of oceanic Pacific islandsat various levels of resolution (Williams et al. 2008,Gove et al. 2015), the species level responses of coralcommunities have rarely been examined in oceanicIndian Ocean communities (see the seminal studiesof Spencer-Davies et al. 1971, Sheppard 1980, Bou-chon 1981). Given the increasing scale, frequencyand intensity of disturbances on coral reefs over thelast 2 decades, it is timely to re-evaluate the spatialdistribution and abundance of coral communities atremote Indian Ocean locations to better predict howthese disturbances will impact the coral communities(Kenyon et al. 2006, Adjeroud et al. 2010) and associ-ated taxa.

Christmas Island is an isolated oceanic island situ-ated on the Indo-Pacific biogeographic border, andas a result, the coral reef ecosystem there includes aunique mix of Indian and Pacific Ocean species andnumerous endemics (Allen 2008, Brewer et al. 2009,Hobbs et al. 2012). Preliminary benthic surveys con-ducted at Christmas Island between 2005 and 2012documented relatively high cover of live coral (41 to58%, Gilligan et al. 2008, Speed et al. 2013). Taxo-nomic surveys conducted by the Western AustralianMuseum in 1987 identified 88 species of scleractiniancorals across 38 genera at Christmas Island (Done &Marsh 2000). More recently, Richards & Hobbs(2014) documented 145 species of scleractinian coralfrom 51 genera, bringing the total number of sclerac-

62A

utho

r cop

y

Richards et al.: Coral community structure at an oceanic island

tinian corals known from the island to 169. However,these benthic and taxonomic surveys did not assessthe spatial variation in coral community compositionor structure around the island.

While Christmas Island is relatively pristine, im -pacts from disease outbreaks and coral bleachinghave been documented (Hobbs & Frisch 2010, Hobbs2014, Hobbs et al. 2015) and increases in the fre-quency and intensity of storm activity and bleachingevents are predicted for the coming decades (Shep-pard et al. 2002, Hyder Consulting 2008, AECOMAustralia 2010, NOAA Coral Reef Watch 2015). Suchburgeoning threats highlight the need to study hownatural processes structure coral communities beforetheir influence becomes intermixed with anthro-pogenic disturbances. Examining the way coral com-munities are structured will help predict which taxaand habitats will be most affected by increasingimpacts. Moreover, understanding the response ofthe local community to disturbances is particularlyimportant at remote locations because isolated coralcommunities are expected to have limited resiliencedue to the reliance on self-recruitment, rather thanlong-distance dispersal, for recovery (Ayre & Hughes2000, Gilmour et al. 2013).

Here we tested the following 5 hypotheses aboutthe coral community structure at Christmas Island,where coral communities grow along steep depthand wave-exposure gradients: (1) coral diversity de -creases with exposure and increases with depth; (2)colony abundance decreases with exposure andincreases with depth; (3) morphological complexityde creases with exposure and increases with depth;(4) habitat complexity decreases with exposure andincreases with depth; and (5) coral community struc-ture is predictable across depth and exposure gradi-ents as a linear trend.

MATERIALS AND METHODS

Study site

Located in the eastern Indian Ocean, 1400 kmnorth-west of Western Australia (Berry & Wells2000), Christmas Island (10° 30’ S, 105° 39’ E) is thesummit of a submarine mountain that rises to a cen-tral plateau of approximately 361 m above sea level.Its jagged limestone coastline is fringed by a coralreef that extends 0 to 200 m seaward before a rapid

drop in depth from 12−20 m to severalthousand metres. With no lagoonsthere is a restricted subset of habitattypes. The typical tidal range is ap -proximately 1 m (Mean Lower LowWater [MLLW] to Mean Higher HighWater [MHHW]).

Survey methodology

Eight sites and 3 depths (5, 12 and20 m) were surveyed around Christ-mas Island between 25 April and 7May 2013. The survey sites spannedthe east, west and north coast of theisland and represented different lev-els of wave exposure (Fig. 1); thesouthern coast was inaccessible dueto constant large swell.

To test for depth-related changes incoral community composition, 3 depthzones were surveyed: the reef flat(5 m), reef crest/upper reef slope(12 m) and reef slope/wall (20 m).Three replicate 15 × 2 m wide belttransects (total 90 m2) were surveyedwithin each depth zone. Every colony

63

Fig. 1. Location of the 8 survey sites distributed around the west (Tom’s Point,TP; Jackson’s Point, JP), north (Thundercliff, TC; Million Dollar Bommie, MDB;Eidsvold, Eis; Flying Fish Cove, FFC; Chinese Literary Association, CLA) andeast (Ethel Beach, EB) coasts of Christmas Island in the eastern Indian Ocean

Aut

hor c

opy

Mar Ecol Prog Ser 546: 61–74, 2016

within the belt transect was counted and identified tospecies; however, juvenile corals (<5 cm diameter)were not included in the assessment due to the diffi-culty of identifying them accurately to species level(Richards 2013). In the case of large thickets ofbranching coral where the beginning and end ofcolonies could not be determined, 2 colonies wererecorded for every 1 m2 of thicket.

Habitat complexity was assessed every 5 m on eachtransect using a 5 point scale (Richards 2013) as fol-lows: 1 is extremely low habitat complexity (e.g. sandand rubble); 2 is scattered coral outcrops among rub-ble; 3 is numerous low profile coral outcrops; 4 is de -fined as well developed reef structure with over-hangs; and 5 is the highest possible complexitydefined as a large complex reef structure with welldeveloped crevices and caves.

Wave energy

Following Williams et al. (2013), wave power perunit length of wave crest was calculated for Christ-mas Island using the archived NOAA Wavewatch IIIglobal, 30 min arc dataset (ftp://polar.ncep.noaa.gov/pub/ history/ waves/). The hindcasts at 3 h intervalsfrom January 2010 to July 2015 of significant waveheight and peak wave period were used to calculatewave power at each timestamp. The average direc-tion of the wave peak period was used to bin theresultant dataset into 16 segments of 22.5° each. Theaverage wave power and proportional occurrence foreach segment were then calculated.

Statistical analysis

The mean and standard error of the number of spe-cies (species density), the number of individuals andthe habitat complexity scores were calculated foreach depth zone at each site. The experimentaldesign consisted of 2 factors: ‘exposure’ (2 levels,fixed and crossed) and ‘depth’ (3 depths, fixed andcrossed) with 3 replicates at each site by depth com-bination. In order to determine sampling adequacy,rarefaction was calculated from raw data (Gotelli &Colwell 2011) using sample-based abundance meth-ods (Colwell et al. 2012). The species density resultswere then plotted against sampling unit (transects).To determine if coral species diversity and colonyabundance varied between exposures and depths, a2-way permutational multivariate analysis of vari-ance (PERMANOVA) (Anderson & Millar 2004) was

undertaken with PRIMER-E version 6 with thePERMANOVA+ add-on (Anderson et al. 2008). Wechose this approach due to the predominance ofzeros in the datasets, the highly skewed relativeabundances of corals and the variability betweensamples (Anderson 2001).

The Euclidean distance matrix (Anderson & Millar2004) based on raw data units was used to analysethe total number of species, the total number of indi-viduals and average habitat complexity. The Bray-Curtis dissimilarity measure was used on the trans-formed coral species assemblage and morphologyassemblage data because it provides a robust meas-ure of ecological distance in terms of communitystructure and can manage joint absences (doublezeros) (Anderson & Robinson 2003). A distance-based test for homogeneity of multivariate disper-sions (PERMDISP) was performed on all datasetsprior to testing the data with PERMANOVA. Thecoral species assemblage data and habitat complex-ity data yielded a significant result, so the data weretransformed using a fourth-root and a square-roottransformation, respectively. The species densitydata also showed a significant result after the PERM-DISP test; however, the data were left untransformedas subsequent transformations retained a significantresult. Due to the relatively small sample size (n =72), the test was done using un restricted permuta-tions of the raw data (Anderson 2001) with 9999 per-mutations of each term in the analysis to obtain p-values (Anderson & Millar 2004). Significant termswere further investigated with pair-wise compar-isons. In analyses where possible permutations wereless than 100, Monte Carlo p-values were used(Anderson & Robinson 2003).

To visualise patterns in the coral species assem-blages and morphologies, a 2-dimensional, non-metric multidimensional scaling (MDS) plot was con-structed and key taxa and morphologies contributingto differences between depths and exposures wererecognised using a similarity of percentages analysis(SIMPER) in PRIMER-E version 6. To further visualisewhich components are contributing to the dissimilari-ties, we conducted a constrained canonical analysisof principal coordinates (CAP) (Anderson & Robinson2003, Anderson & Willis 2003, Anderson & Millar2004). These CAP analyses were conducted using theoriginal replicate observations (n = 72). The leave-one-out allocation success provides a statistical esti-mate of error (proportion of incorrect allocations) forthis classification method and gives a reasonablemeasure of how distinct groups are in multivariatespace (Anderson & Willis 2003, Anderson & Robinson

64A

utho

r cop

y

Richards et al.: Coral community structure at an oceanic island

2003). The appropriate number of axes (m) to use forthe canonical analysis was chosen based on minimis-ing misclassification error and keeping m small rela-tive to n (Anderson & Robinson 2003, Anderson &Willis 2003). The null hypothesis of no difference inthe coral assemblages from different exposures anddepths was tested by calculating the trace statisticand obtaining a p-value by permutation (Anderson &Robinson 2003, Anderson & Willis 2003). Individualspecies likely to be responsible for any observed dif-ferences were determined by examining Pearson cor-relations of untransformed species counts with canon-ical axes. Spearman rank correlation values greaterthan rS = 0.36 or less than rS = −0.36, and greater thanrS = 0.51 or less than rS = −0.51 were used as an arbi-trary cut-off for the colony morphology and commu-nity structure correlation vectors, respectively. ABEST test was undertaken to determine whetherdepth or exposure were most important in determin-ing the patterns in the species assemblage matrix,along with a DistLM test to examine how much of thevariation was explained by each variable.

RESULTS

Wave energy

Christmas Island experiences south-easterly tradewinds throughout the dry season (April to Novem-ber). The prevailing wave energy (63.55%) propa-gates from the south-west (202.5 to 225°) with anaverage power of 44 kW m−1 (Fig. 2). Overall, 94.67%of waves propagate from the south-west, south andsouth-east (157.5 to 248°) with an average power of40 kW m−1. These data indicate that the southerncoastline is constantly exposed to large wave swell.Waves propagating from the north-east are rare(0.03%) but can have high energy (61 kW m−1).Hence, rare north-east swell events can be extremelydamaging for the east coast of the island. Waves fromthe north-west are also relatively rare (3.2%) andnormally associated with wet season tropical cy -clones which occasionally pass close to the island(Department of Transport 2009). These waves canalso have high energy, ranging from 41 to 51 kW m−1

and impact the north-west of the island.

Diversity

Rarefaction analyses indicated that the local spe -cies diversity was adequately sampled in our surveys

(see Fig. S1 in the Supplement at www. int-res. com/articles/ suppl/ m546p061_ supp. pdf). Approximately75% of the recorded species were detected duringthe first 18 (25%) of the 72 transects, suggesting thatadditional sampling would not substantially increase

65

Fig. 2. Directional plots of wave power and normalised fre-quency for Christmas Island. The wave direction is given indegrees and uses the oceanographic convention, such thatwaves propagating north would be located at 0°. The meanwave power and normalised wave frequency binned into 16discrete sectors each spanning 22.5° were computed fromNOAA’s Wavewatch III from January 2010 to July 2015.(a) Wave power. The black numbers represent the wavepower in kW m−1 from 0 at the centre to 70 at the circumfer-ence. The numbers in white in each sector represent the per-centage occurrence of the wave events in that sector.(b) Normalised wave frequency, the number of wave events(3 h output values) in each 22.5° sector divided by the totalnumber events over the 5 yr time period. Here the values in

black represent the fractional number of waves

Aut

hor c

opy

Mar Ecol Prog Ser 546: 61–74, 2016

the number of species recorded. Coral diversity (spe-cies density) did not decrease linearly with depth aspredicted, but was greater at sheltered sites. Hence,the results only partially confirm the hypothesis thatdiversity increases with depth up to 20 m, becausespecies density was lowest at the 12 m depth zone,intermediate at the 5 m depth zone, and highest atthe 20 m depth zone (Table 2, Fig. 3a). Mean speciesdensity peaked at 20 m depth at the sheltered siteThundercliff (TC) (37.6 ± 0.6) and was lowest at 5 mdepth at the exposed site Ethel Beach (EB) (9.6 ± 4.1)(see Fig. S2 in the Supplement). The diversity ofcorals was significantly higher at sheltered sites(mean species density = 26.917 ± 3.538 per 30 m2)than at exposed sites (22.75 ± 3.668 per 30 m2) (p <0.005, Table 1, Fig. 3a). Overall, there was no signifi-cant interaction between depth and exposure forspecies density (see Table S1a in the Supplement).

Abundance

A total of 9265 coral colonies from 125 coral speciesand 43 genera were recorded across the 72 belt tran-sects. On average, 128 (±7.97 SE) coral colonies wererecorded per transect. As expected, colony abun-

dance was highest at the 20 m depth zone (178.917 ±8.533), but contrary to expectations, colony abun-dance was significantly higher at the 5 m depth(152.542 ± 11.739) than at 12 m (54.583 ± 2.190) depthzone. Overall, colony abundance was highest at shel-tered sites (138.667 ± 10.615 colonies per 30 m2); andthere was considerable variability in colony abun-dance among exposed sites (118.694 ± 33.984)(Table 1). The number of colonies recorded were sig-nificantly different between the 5 and 12 m and the12 and 20 m depth zones (all p < 0.001) (Table 2).There was no significant difference in the abundanceof colo nies between 5 and 20 m and there was no sig-nificant interaction between exposure and depth (seeTable S1a in the Supplement).

Colony morphology

The complexity of coral morphologies conformed tothe expectations of being highest at the 20 m depthzone and decreasing with exposure. The morpholog-ical types that were present differed significantlybetween depth and exposures (p < 0.001, Tables 1 &2), and a greater variety of morphological types was

66

Species Abundance Colony Habitat density morphology complexity

ShelteredExposed 0.004** 0.039* <0.001*** <0.001***

Table 1. Significance levels and p-values for pairwise com-parisons of the factor ‘exposure’ in PERMANOVA tests car-ried out on species density, abundance, colony morphologyand habitat complexity derived from data collected across2 levels of exposure (sheltered and exposed) at Christmas

Island.*p < 0.05, **p < 0.005, ***p < 0.0005

5 m 12 m

Species density 12 m 0.065ns 20 m <0.001*** <0.001***

Abundance 12 m <0.001*** 20 m 0.064ns <0.001***

Colony morphology 12 m <0.001*** 20 m <0.001*** <0.001***

Habitat complexity 12 m <0.001*** 20 m <0.001*** <0.001***

Table 2. Significance levels and p-values for pairwise compar-isons of the factor ‘depth’ in PERMANOVA tests carried outon species density, abundance, colony morphology and habi-tat complexity derived from data collected across 3 depths (5,12 and 20 m) at Christmas Island. ns: p > 0.05, ***p < 0.0005

Fig. 3. Spatial differences in (a) species density (mean ± SE)per 30 m2 and (b) Habitat complexity score (mean ± SE)from 3 depth zones (5, 12 and 20 m) and 2 levels of exposure

(exposed and sheltered)

Aut

hor c

opy

Richards et al.: Coral community structure at an oceanic island

recorded at the 20 m depth zone compared to theshallower zones (see Fig. S3 in the Supplement).However, there was no significant interaction be -tween depth and exposure (see Table S1b). Differ-ences in exposure and depth are illustrated in theCAP plot (Fig. 4) and in the MDS (see Fig. S4). Theseplots and the SIMPER analysis (see Tables S2 & S3),identify encrusting colonies as contributing to thedistinctness of the 20 m depth zone and columnarand massive colonies as characteristic of the 5 m

zone. The columnar species Isopora palifera was oneof the 5 most abundant species at all sheltered sites inthe 5 m depth zone and contributed to the character-izing of these sites (see Tables S3, S4 & S5), and theencrusting species Pavona varians was one of the 5most abundant species at 20 m depth across themajority of sites (see Tables S2, S3 & S4).

Habitat complexity

Spatial patterns in habitat complexity at ChristmasIsland conformed to the hypothesis of decreasingcom plexity with exposure (p < 0.001, Table 1,Fig. 3b). However, contrary to our hypothesis thathabitat complexity would increase with depth, thehighest level of complexity was recorded at 12 m, thelowest at 5 m and intermediate levels at 20 m(Fig. 3b). There were also significant differencesacross the 3 depth zones examined (Table 2), but nosignificant interaction between exposure and depth(see Table S1a in the Supplement).

Community structure

Contrary to our hypothesis, the structure of coralcommunities was not consistent or predictable acrossdepth gradients. There was a significant interactionbe tween exposure and depth (p < 0.05, see Table S1bin the Supplement), which is illustrated in the CAPplot (Fig. 5) and in the MDS (see Fig. S5 in the Supplement at www. int-res. com/ articles/ suppl/ m546p061_ supp. pdf). The BEST analysis also indicatedthat both depth and exposure were correlated withpatterns in the community composition (R2 = 0.294);however, depth explained most of the correlation(R2 = 0.275) because there was little improvement inthe correlation when exposure was added (R2 =0.107) (see Table S5). This result was corroborated bythe DistLM test, which showed significant correla-tions between community structure and each of thesevariables (all p < 0.001); however, the correlation wasnot very high (R2 = 0.205) (see Table S6). The sequen-tial test also showed that depth contributed the mostto the correlation.

The uniqueness of the coral community structure atthe 5 m depth zone is likely due to the high abun-dances of Goniastrea retiformis, Montipora informis(Fig. 5) and I. palifera (see Tables S7 & S8). The com-munity composition at 20 m depth was distinct due tothe high abundances of P. varians and Cyphastreaserailia (Fig. 5, see Table S8). There was also a high

67

Fig. 4. Canonical analysis of principal coordinates (CAP) ofcolony morphology for depth and exposure explained94.62% of the total variation in the first 8 principal coordi-nates (squared canonical correlation = 0.895) with the small-est cross-validation error achieved when m = 3. (a) Biplot ofspecies community structure correlations with the canonical

axes and (b) Pearson’s Overlay rS > 0.36

Aut

hor c

opy

Mar Ecol Prog Ser 546: 61–74, 201668

level of variation in the dominant species betweensites, which was most evident at Ethel Beach where asingle species (Galaxea as tre ata) was almost 7 timesmore abundant than any other species at the 5 m

depth zone (see Table S2). Furthermore, the mostdominant species in the 12 m depth zone tended tohave 2 to 3 times lower abundance than thatrecorded in the 5 and 20 m depth zones.

DISCUSSION

Maintaining healthy coral communities is impera-tive for securing the wider biodiversity and socio-economic benefits of coral reefs. Over the last fewdecades corals have proven extremely susceptible toclimatic change. Being able to forecast, detect andmitigate departures from ‘natural’ is crucial and rep-resents a great challenge for ecologists, conservationscientists and managers. Theory of coral reef ecologyis generally based on studies undertaken in rela-tively sheltered and shallow continental reefs thathave often been exposed to an undisclosed level ofanthropogenic disturbance. Here, we tested ecologi-cal theory in an oceanic location characterised bysteep gradients in physical factors (e.g. depth andexposure) and minimal anthropogenic impacts. Bytesting 5 traditional hypotheses of coral communitystructure at the relatively pristine Christmas Islandwe find that not all of the expectations of traditionalecological theory are supported. Understanding howand why coral communities are structured differentlyon oceanic versus continental reefs is important forpredicting how increasing anthropogenic distur-bances will impact different types of reefs.

Diversity and abundance

The results of this study only partially met expecta-tions of the traditional hypotheses of coral diversityand abundance. We found abundance and diversitywere highest at the 20 m depth zone across all sites,which concurs with earlier seminal studies (Loya1972, Glynn 1976, Sheppard 1980, Ross & Hodgson1982) that found these variables were highestbetween 15 and 30 m depth. Our finding of highestcoral diversity at the edge of the steepest habitatzone (almost vertical walls) is consistent with the‘cliff-edge’ effect. In a study of patterns of coral species diversity on the San Blas Islands in Panama,Porter (1972) proposed that the peak in species rich-ness that was consistently observed at 25 m was dueto abrupt changes in light and turbulence (i.e. watercirculation which delivers nutrients). Similarly, in astudy of reef community structure in north Jamaica,the highest diversity of corals was recorded at 22 m

Fig. 5. Canonical analysis of principal coordinates (CAP) ofcoral community structure for depth and exposure explained97.33% of the total variation in the first 8 principal coordi-nates (squared canonical correlation = 0.947) with the small-est cross-validation error achieved when m = 16. (a) Biplot ofspecies community structure correlations with the canonical

axes and (b) Pearson’s Overlay rS > 0.5

Aut

hor c

opy

Richards et al.: Coral community structure at an oceanic island 69

and this was attributed to an edge effect (Liddell &Ohlhorst 1987).

At Christmas Island, the upper section of the cliffoccurs between 15 and 20 m and this depth also rep-resents the zone where physical impacts are less pro-nounced. This zone is deep enough to escape theeffect of waves but has much more light availablethan on the steep walls beyond 20 m. The zone rep-resents a sharp change in the angle of the reef wherewater from on top of the reef mixes with oceanicwater washíng against the outer reef walls. Conse-quently, this zone experiences abrupt changes intemperature from thermoclines and other variablesassociated with different water bodies. It is thereforelikely that this narrow zone represents an ecotonewhere there is a very sharp transition in environmen-tal conditions resulting in a heterogeneous environ-ment that contains a range of ecological niches suitedto a wider diversity of corals.

Edge effects have been shown to increase diversityin seagrass systems (Tanner 2005, Macreadie et al.2010) and are also well known in terrestrial ecosys-tems (Harris 1988, Laurance & Yensen 1991). Giventhat oceanic islands are exposed to considerableswell and have steep drop-offs, it is likely that edgeeffects are a common feature of oceanic coral reefs,with diversity peaking in the narrow zone wherephysical impacts from waves are less pronouncedand where light remains at suitable levels for optimalgrowth. At Christmas Island, we detected the signalof edge effects at 20 m, beyond which depth thesteep wall becomes shaded and the conditions be -come suboptimal for photosynthetic organisms likecorals. However, the precise depth range of the edgezone will vary between oceanic islands and will bedependent on the amount of wave energy receivedand the depth of the outer reef wall. Further investi-gation is required to test whether the cliff edge effectdetected at Christmas Island and elsewhere (Porter1972, Liddell & Ohlhorst 1987) is a general phenom-enon of oceanic coral reefs.

Coral abundance and diversity did not increasewith depth in a linear trend, but rather they werelowest at the intermediate depth 12 m, suggesting aJ-curve distribution may best describe the relation-ship between these variables and depth at this iso-lated oceanic location (Fig. 6c). The reason for the J-curve distribution may relate to 2 features that aretypical of oceanic islands. Firstly, the heterogeneityof environmental conditions at the cliff edge mayprevent species from dominating or establishinglarge colonies, hence enabling high diversity. Sec-ondly, oceanic islands are exposed to persistent swell

and the role of wave action creates sufficient distur-bance to facilitate turnover on the reef flat (Madin etal. 2014). At 12 m depth near the reef crest, wave-tol-erant massive species (such as massive Porites) growinto large colonies, and fast growing and competitivebranching species (such as Acropora abrotanoides)tend to dominate and appear to competitively ex -clude other corals. While the size of corals was notspecifically measured here, we anecdotally observedcoral colonies to be larger in the 12 m depth zonethan the other 2 depth zones, particularly at shel-tered sites. Hence, competitive exclusion is likely tobe a factor limiting the diversity of corals at this depth(Connell 1978). Thus, the less physical but moreenvironmentally variable conditions associated withthe cliff-edge habitat (at 20 m) sustains a more het-erogeneous community, whilst the wave-exposedhabitats at 5 m experience a high turnover and a limited number of species dominate at 12 m.

Colony morphology

The relative dominance of branching, foliose,corymbose, columnar and massive morphologies at5 m was exposure-dependent. In contrast to expecta-tions, the more structurally complex branching andcorymbose morphologies were most common at ex-posed sites. Generally, branching morphologies aretypical in sheltered, shallow areas due to their struc-tural fragility and vulnerability to breakage fromwave energy (Madin & Connolly 2006). However, atChristmas Island, the main taxa contributing to thismorphology were Pocilloporids and A. abro ta no ides,which have robust and/or highly compact branchesthat are resistant to strong wave action (Dollar 1982,Veron 1993, Kaandorp 1999). However, the exposedsite Ethel Beach was dominated by corals with mas-sive morphology. This site experiences constant waveaction and rare high energy wave events; hence,massive forms that are more re sistant to dislodgementdominate (Connell 1978, Madin & Connolly 2006). At20 m depth, tabulate and free-living colonies, whichare vulnerable to dislodgement or being overturned(Madin & Connolly 2006), occurred in higher abun-dance, most likely be cause they are protected fromwave energy at this depth.

Habitat complexity

As predicted by established hypothesis, habitatcomplexity was generally lower at exposed sites and

Aut

hor c

opy

Mar Ecol Prog Ser 546: 61–74, 2016

in the shallow waters (5 m). However, contrary toexpectations (Fig. 6b), there was no linear pattern ofincrease in habitat complexity with depth, as habitatcomplexity was highest at the 12 m depth zone acrossall sites, suggesting a negative skew distribution maybest describe the relationship between habitat com-plexity and depth at this isolated oceanic location(Fig. 6d). This unexpected result may be explainedby the coral community at this depth. Althoughabundance and species diversity (density) was lowestat 12 m, the number of corals with complex colonymorphologies was higher at this depth and this willhave contributed to the higher habitat complexityscore. The most abundant morphological types at the12 m depth zone were the branching, columnar andmassive morphologies (i.e. A. abrotanoides, Isoporapalifera and P. lobata). These growth forms are con-sidered to create complex habitats as they can formlarge 3-dimensional habitats with numerous intricatespaces (Veron 1993).

Community structure

The distinct coral community across depth andexposure highlights the importance of wave energyin structuring community composition at this oceaniclocation. Species that are tolerant of varied and harshconditions, such as massive corals like P. lutea orencrusting species like Montipora informis, werecharacteristic of the 5 m depth zone at exposed sites.At exposed sites, the community at 12 m was similarin functionality to that at 5 m, albeit composed of dif-ferent species (P. lobata and M. grisea). Acroporaspecies often dominate shallow habitat zones at shel-tered sites (Huston 1985, Toda et al. 2007, Williams etal. 2013), so we expected to find this pattern; how-ever, Acropora were rarely the most dominant spe-cies at sheltered sites. Moreover, Acropora was morecommonly recorded in abundance at the exposedsites where A. nana, A. abrotanoides and A. papillarewere all recorded among the top 5 dominant taxa. A.

70

Fig. 6. (a) Schematic cross section of the Christmas Island fringing reef depicting the 3 reef zones examined at 5, 12 and20 m. (b) The expected linear relationship between coral diversity, abundance or habitat complexity and increasing depth.(c) Graphical representation of the observed relationship between diversity or abundance and depth, which resembles aJ-curve. (d) Graphical representation of the observed relationship between habitat complexity and depth, which resembles

a negatively skewed curve

Aut

hor c

opy

Richards et al.: Coral community structure at an oceanic island

papillare is a primarily encrusting species that occursin shallow marginal habitats (Wallace 1999), so itspresence is not surprising. However, the presence ofcorymbose and branching Acropora in shallow zonesat more exposed sites goes against predictions basedon mechanical vulnerability (Madin 2005, Madin etal. 2014). Nevertheless, in this case, the branchingcoral A. abrotanoides is an exceptionally robust spe-cies with thick branching units approximately 10 cmin diameter (Wallace 1999) that is well suited tostrong current and high wave energy conditions. Thepresence of A. nana in abundance at Jackson’s Pointmay be a result of a recent pulse recruitment event ofthis ‘weedy’ (opportunistic or early colonizing) spe-cies (Darling et al. 2013) and temporal data areneeded to examine the longer term persistence ofthis species at this site (Kenyon et al. 2006, Smith etal. 2008).

The coral community growing at 20 m depth wascharacterised by encrusting species that make effi-cient use of light at lower saturations and can growon vertical faces without risk of toppling, such asCyphastrea serailia, Leptastrea pruinosa, Pavonavari ans and Porites lobata (Bouchon 1981, Wellington1982, Titlyanov & Latypov 1991). Our results suggestthat the 20 m depth zone appears to be a refuge for awide range of species that are not found in shallowerzones. We infer, therefore, that any disturbances thataffect the reef at this depth (i.e. increasing frequencyor intensity of storms, coral bleaching, coral disease,predator outbreaks) could have major and lastingramifications for the diversity and abundance of thewider coral reef community (Graham et al. 2006).Corals occurring across all depth zones in shelteredhabitats along the north coast (i.e. Thundercliff, Mil-lion Dollar Bommie and Eidsvold), are also particu-larly vulnerable to changes in wave frequency orintensity. While the loss of corals from storm eventsmay open up space for coralline and turf algae (seeFig. S6 in the Supplement for more detail about thecurrent level of benthic cover of these taxa), it mayalso drive fishes into deeper water and changetrophic relationships (Friedlander et al. 2003). Oursurveys revealed that turf algae can reach approxi-mately 60% cover in places and indicates that it hasbecome well established and may be outcompetingcorals at some sites. The reason for such high turfcover at this remote and relatively pristine location isunknown but it is not necessarily an indicator ofdegradation given that both natural and anthro-pogenic disturbances can provide opportunities forturf algae to dominate coral reef communities (Barottet al. 2012). Our results show that if the predicted

increases in bleaching events at Christmas Islandeventuate (Sheppard et al. 2002, NOAA Coral ReefWatch 2015), it will be particularly important to mon-itor the status of corals and coral−algal interactions atthe 20 m depth zone.

CONCLUSIONS

This species-level study at an oceanic IndianOcean location found that natural environmental fac-tors such as exposure and depth strongly influencecoral community structure. Contrary to expectations,however, there was no clear, consistent or linearincrease in diversity, abundance or habitat complex-ity across the depth gradient examined. This findingof non-linear patterns in community structure at thisisolated oceanic location concurs with patternsreported for coral communities in the northern LineIslands (Gove et al. 2015) and the Hawaiian Archi-pelago (Jouffray et al. 2015). We identify a cliff-edgeeffect where coral diversity is greatest in a narrowzone between the wave-dominated shallows and thelight-limited outer reef wall. Further investigations ofcommunity structure at other oceanic locations areneeded to determine the generality of the edge ef -fect. The need to understand natural processes struc-turing oceanic coral communities is reinforced whenhuman impacts are considered, because these im -pacts have been found to decouple natural biophysi-cal relationships on coral reefs (Williams et al. 2015).Species currently finding refuge in the cliff-edgezone may be impacted by changing climatic condi-tions (e.g. increasing frequency and intensity ofstorm and bleaching events). In order to make accu-rate predictions about the future of isolated oceaniccoral reefs there is a need to re-evaluate and updateour understanding of how these coral communitiesare structured. In addition, species diversity andabundance in cliff-edge habitats (beyond the reef flatand crest) must be monitored in order to predict ordetect how increasing climate and local impacts mayaffect these unique coral reef ecosystems.

Acknowledgements. This research was undertaken as partof N.M.R.’s honour research programme at the University ofWestern Australia. It was conducted under the appropriatepermits from Parks Australia, the Western Australia Depart-ment of Fisheries. We acknowledge and thank Parks Aus-tralia, Christmas Island, for providing logistical support.Financial support was provided by an AIMS-CSIRO-UWAFellowship to J.-P.A.H. Funding for the post-fieldworkwrite-up of this data was provided to Z.T.R. by WoodsideEnergy. Z.T.R. thanks Diana Jones and Jane Fromont for

71A

utho

r cop

y

Mar Ecol Prog Ser 546: 61–74, 2016

their support of this project. Thanks also to Loisette Marshfor interesting discussions about the Christmas Island coralsand to the anonymous reviewers whose constructive com-ments helped to improve this manuscript.

LITERATURE CITED

Adjeroud M, Michonneau F, Edmunds PJ, Chancerelle Yand others (2009) Recurrent disturbances, recovery tra-jectories, and resilience of coral assemblages on a SouthCentral Pacific reef. Coral Reefs 28: 775−780

Adjeroud M, Fernandez JM, Carroll AG, Harrison PL, PeninL (2010) Spatial patterns and recruitment processes ofcoral assemblages among contrasting environmentalconditions in the southwestern lagoon of New Caledonia.Mar Pollut Bull 61: 375−386

AECOM Australia (2010) Indian Ocean territory climatechange risk assessment (2010 update version). Report tothe Attorney-General’s Department, June 2010, Can-berra

Allen GR (2008) Conservation hotspots of biodiversity andendemism for Indo-Pacific coral reef fishes. Aquat Con-serv Mar Freshw Ecosyst 18: 541−556

Alvarez-Filip L, Dulvy NK, Gill JA, Côté IM, Watkinson AR(2009) Flattening of Caribbean coral reefs: region-widedeclines in architectural complexity. Proc R Soc B 276: 3019−3025

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance. Austral Ecol 26: 32−46

Anderson MJ, Millar RB (2004) Spatial variation and effectsof habitat on temperate reef fish assemblages in north-eastern New Zealand. J Exp Mar Biol Ecol 305: 191−221

Anderson MJ, Robinson J (2003) Generalized discriminantanalysis based on distances. Aust NZ J Stat 45: 301−318

Anderson MJ, Willis TJ (2003) Canonical analysis of princi-pal coordinates: a useful method of constrained ordina-tion for ecology. Ecology 84: 511−525

Anderson M, Gorley RN, Clarke RK (2008) Permanova+ forPrimer: Guide to software and statistical methods.PRIMER-E, Plymouth

Ayre DJ, Hughes TP (2000) Genotypic diversity and geneflow in brooding and spawning corals along the GreatBarrier Reef, Australia. Evolution 54: 1590−1605

Barott KL, Williams GJ, Vermeij MJA, Harris J, Smith JE,Rohwer FL, Sandin SA (2012) Natural history of coral−algae competition across a gradient of human activity inthe Line Islands. Mar Ecol Prog Ser 460: 1−12

Bergman KC, Öhman MC, Svensson S (2000) Influence ofhabitat structure on Pomacentrus sulfureus, a western In-dian Ocean reef fish. Environ Biol Fishes 59(3): 243–252

Berry PF, Wells FE (2000) Survey of the marine fauna ofChristmas Island, Indian Ocean. Rec West Aust MusSuppl No. 59: 71−74

Bongaerts P, Muir P, Englebert N, Bridge TCL, Hoegh-Guld-berg O (2013) Cyclone damage at mesophotic depths onMyrmidon Reef (GBR). Coral Reefs 32: 935

Bouchon C (1981) Quantitative study of the scleractiniancoral communities of a fringing reef of Reunion Island(Indian Ocean). Mar Ecol Prog Ser 4: 273−288

Brewer DT, Potter A, Skewes TD, Lyne V and others (2009)Conservation values in Commonwealth waters of theChristmas and Cocos (Keeling) Islands remote AustralianTerritories. Report to Department of Environment andWater Resources. CSIRO, Cleveland

Bruno JF, Selig ER, Casey KS, Page CA and others (2007)Thermal stress and coral cover as drivers of coral diseaseoutbreaks. PLoS Biol 5: e124

Carpenter KE, Abrar M, Aeby G, Aronson RB and others(2008) One-third of reef-building corals face elevatedextinction risk from climate change and local impacts.Science 321: 560−563

Colwell RK, Chao A, Gotelli NJ, Lin SY, Mao CX, ChazdonRL, Longino JT (2012) Models and estimators linkingindividual-based and sample-based rarefaction, extrapo-lation and comparison of assemblages. J Plant Ecol 5: 3−21

Connell JH (1978) Diversity in tropical rain forests and coralreefs. Science 199: 1302−1310

Connell JH, Hughes TP, Wallace CC (1997) A 30-year studyof coral abundance, recruitment, and disturbance at sev-eral scales in space and time. Ecol Monogr 67: 461−488

Darling ES, McClanahan TR, Côté IM (2013) Life historiespredict coral community disassembly under multiplestressors. Glob Chang Biol 19: 1930−1940

De’ath G, Moran PJ (1998) Factors affecting the behaviour ofcrown-of-thorns starfish (Acanthaster planci) on theGreat Barrier Reef: 2: Feeding preferences. J Exp MarBiol Ecol 220: 107−126

Department of Transport (2009) Christmas Island coastalengineering investigation. Technical Report No. DOT R482. Government of Western Australia.

DeVantier LM, De’ath G, Turak E, Done TJ, Fabricius KE(2006) Species richness and community structure of reef-building corals on the nearshore Great Barrier Reef.Coral Reefs 25: 329−340

Dollar SJ (1982) Wave stress and coral community structurein Hawaii. Coral Reefs 1: 71−81

Done TJ (1982) Patterns in the distribution of coral commu-nities across the central Great Barrier Reef. Coral Reefs1: 95−107

Done TJ (1983) Coral zonation: its nature and significance.In: Barnes D (ed) Perspectives on Coral Reefs. BrianClouston, Manuka, p 107−147

Done TJ, Marsh L (2000) Reef-building corals of ChristmasIsland. Rec West Aust Mus Suppl No. 59: 79−81

Emanuel K (2005) Increasing destructiveness of tropicalcyclones over the past 30 years. Nature 436: 686−688

Fabricius KE, De’ath G, Puotinen ML, Done TJ, Cooper TF,Burgess SC (2008) Disturbance gradients on inshore andoffshore coral reefs caused by a severe tropical cyclone.Limnol Oceanogr 53: 690−704

Fox RJ, Bellwood DR (2007) Quantifying herbivory across acoral reef depth gradient. Mar Ecol Prog Ser 339: 49−59

Friedlander AM, Brown EK, Jokiel PL, Smith WR, RodgersKS (2003) Effects of habitat, wave exposure and marineprotected area status on coral reef fish assemblages inthe Hawaiian archipelago. Coral Reefs 22: 291−305

Gilligan JJ, Hender J, Hobbs J-PA, Neilson J, McDonald C(2008) Coral reef surveys and stock size estimates ofshallow water (0−20m) marine resources at ChristmasIsland, Indian Ocean. Report to Parks Australia North

Gilmour JP, Smith LD, Heyward AJ, Baird AH, Pratchett MS(2013) Recovery of an isolated coral reef system follow-ing severe disturbance. Science 340: 69−71

Glynn PW (1976) Some physical and biological determinantsof coral community structure in the eastern Pacific. EcolMonogr 46: 431−456

Goatley CHR, Bellwood DR (2011) The roles of dimensional-ity, canopies and complexity in ecosystem monitoring.PLoS One 6: e27307

72A

utho

r cop

y

Richards et al.: Coral community structure at an oceanic island

Gotelli NJ, Colwell RK (2011) Estimating species richness.In: Magurran AE, McGill BJ (eds) Biological diversity: frontiers in measurement and assessment. Oxford Uni-versity Press, Oxford, p 39−54

Gove JM, Williams GJ, McManus MA, Clark SJ, Ehses JS,Wedding LM (2015) Coral reef benthic regimes exhibitnon-linear threshold responses to natural physical driv-ers. Mar Ecol Prog Ser 522: 33−48

Graham NAJ, Nash KL (2013) The importance of structuralcomplexity in coral reef ecosystems. Coral Reefs 32: 315−326

Graham NA, Wilson SK, Jennings S, Polunin NV, Bijoux JP,Robinson J (2006) Dynamic fragility of oceanic coral reefecosystems. Proc Natl Acad Sci USA 103: 8425−8429

Hall AR, Miller AD, Leggett HC, Roxburgh SH, Buckling A,Shea K (2012) Diversity−disturbance relationships: fre-quency and intensity interact. Biol Lett 8: 768−771

Harmelin-Vivien ML (1994) The effects of storms andcyclones on coral reefs: a review. J Coast Res 12: 211−231

Harris LD (1988) Edge effects and conservation of bioticdiversity. Conserv Biol 2: 330−332

Hobbs JPA (2014) A glaring omission in Australia’s marineconservation planning. Mar Policy 44: 149−151

Hobbs JPA, Frisch AJ (2010) Coral disease in the IndianOcean: taxonomic susceptibility, spatial distribution andthe role of host density on the prevalence of white syn-drome. Dis Aquat Org 89: 1−8

Hobbs JPA, Jones GP, Munday PL, Connolly SR, SrinivasanM (2012) Biogeography and the structure of coral reef fishcommunities on isolated islands. J Biogeogr 39: 130−139

Hobbs JPA, Frisch AJ, Newman SJ, Wakefield CB (2015)Selective impact of disease on coral communities: out-break of white syndrome causes significant mortality ofAcropora plate corals. PLoS One 10: e0132528

Hoegh-Guldberg O, Salvat B (1995) Periodic mass-bleach-ing and elevated sea temperatures: bleaching of outerreef slope communities in Moorea, French Polynesia.Mar Ecol Prog Ser 121: 181−190

Huston MA (1985) Patterns of species diversity on coralreefs. Annu Rev Ecol Syst 16: 149−177

Hyder Consulting (2008) The impacts and managementimplications of climate change for the Australian govern-ment’s protected areas. Report to the Department of theEnvironment, Water, Heritage and the Arts and theDepartment of Climate Change, Canberra, Australia

Jackson JBC (1979) Morphological strategies of sessile ani-mals. In: Larwood G, Rosen BR (eds) Biology and system-atics of colonial organisms. Academic Press, New York,p 499−555

Jouffray JB, Nyström M, Norström AV, Williams ID, Wed-ding LM, Kittinger JN, Williams GJ (2015) Identifyingmultiple coral reef regimes and their drivers across theHawaiian archipelago. Phil Trans R Soc Lond B Biol Sci370: 20130268

Kaandorp JA (1999) Morphological analysis of growth formsof branching marine sessile organisms along environ-mental gradients. Mar Biol 134: 295−306

Karlson RH (1999) Dynamics of coral communities, Vol. 23.Population and community biology series. Kluwer Aca-demic Publishers, Dordrecht

Kayal M, Vercelloni J, Lison de Loma T, Bosserelle P andothers (2012) Predator crown-of-thorns starfish (Acan-thaster planci) outbreak, mass mortality of corals, andcascading effects on reef fish and benthic communities.PLoS One 7: e47363

Kenyon JC, Vroom PS, Page KN, Dunlap MJ, Wilkinson CB,Aeby GS (2006) Community structure of hermatypiccorals at French Frigate Shoals, northwestern HawaiianIslands: capacity for resistance and resilience to selectivestressors. Pac Sci 60: 153−175

Laurance WF, Yensen E (1991) Predicting the impacts ofedge effects in fragmented habitats. Biol Conserv 55: 77−92

Liddell WD, Ohlhorst SL (1987) Patterns of reef communitystructure, north Jamaica. Bull Mar Sci 40: 311−329

Lirman D (2001) Competition between macroalgae andcorals: effects of herbivore exclusion and increased algalbiomass on coral survivorship and growth. Coral Reefs19: 392−399

Loya Y (1972) Community structure and species diversityof hermatypic corals at Eilat, Red Sea. Mar Biol 13: 100−123

Macreadie PI, Hindell JS, Keough MJ, Jenkins GP, ConnollyRM (2010) Resource distribution influences positive edgeeffects in a seagrass fish. Ecology 91: 2013−2021

Madin JS (2005) Mechanical limitations of reef corals duringhydrodynamic disturbances. Coral Reefs 24: 630−635

Madin JS, Connolly SR (2006) Ecological consequences ofmajor hydrodynamic disturbances on coral reefs. Nature444: 477−480

Madin JS, O’Donnell MJ, Connolly SR (2008) Climate-mediated mechanical changes to post-disturbance coralassemblages. Biol Lett 4: 490−493

Madin JS, Baird AH, Dornelas M, Connolly SR (2014)Mechanical vulnerability explains size-dependent mor-tality of reef corals. Ecol Lett 17: 1008−1015

Marshall PA, Baird AH (2000) Bleaching of corals on theGreat Barrier Reef: differential susceptibilities amongtaxa. Coral Reefs 19: 155−163

NOAA Coral Reef Watch (2015) updated daily. NOAACoral Reef Watch four-month coral bleaching thermalstress outlook December−March 2016. College Park,Maryland: NOAA Coral Reef Watch. Data set accessed15 December 2015 at http: //coralreefwatch.noaa.gov/satellite/ bleachingoutlook _ cfs/outlook_cfs_east.php

Penin L, Adjeroud M, Schrimm M, Lenihan HS (2007) Highspatial variability in coral bleaching around Moorea(French Polynesia): patterns across locations and waterdepths. C R Biol 330: 171−181

Porter JW (1972) Patterns of species diversity in Caribbeanreef corals. Ecology 53: 745−748

Richards ZT (2013) A comparison of proxy performance incoral biodiversity monitoring. Coral Reefs 32: 287−292

Richards ZT, Hobbs JPA (2014) The status of hard coral bio-diversity at Christmas Island. Raffles Bull Zool S30: 376−398

Ross MA, Hodgson G (1982) A quantitative study of her-matypic coral diversity and zonation at Apo Reef, Min-doro, Philippines. Proc 4th Int Coral Reef Symp, Manila2: 281−291

Rudolf VHW, Antonovics J (2005) Species coexistence andpathogens with frequency-dependent transmission. AmNat 166: 112−118

Russ GR (2003) Grazer biomass correlates more stronglywith production than with biomass of algal turfs on acoral reef. Coral Reefs 22: 63−67

Sheppard CRC (1980) Coral cover, zonation and diversity onreef slopes of Chagos Atolls, and population structures ofthe major species. Mar Ecol Prog Ser 2: 193−205

Sheppard CR, Spalding M, Bradshaw C, Wilson S (2002)

73A

utho

r cop

y

Mar Ecol Prog Ser 546: 61–74, 2016

Erosion vs. recovery of coral reefs after 1998 El Niño: Chagos reefs, Indian Ocean. Ambio 31: 40−48

Smith LW, Wirshing HH, Baker AC, Birkeland C (2008)Environmental versus genetic influences on growth ratesof the corals Pocillopora eydouxi and Porites lobata(Anthozoa: Scleractinia). Pac Sci 62: 57−69

Speed CW, Babcock RC, Bancroft KP, Beckley LE and others(2013) Dynamic stability of coral reefs on the west Aus-tralian coast. PLoS One 8: e69863

Spencer-Davies P, Stoddart DR, Sigee DC (1971) Reef formsof Addu Atoll, Maldive Islands. In: Stoddart DR, YongeCM (eds) Regional variation in Indian Ocean reefs. SympZool Soc Lond 28: 217–259

Tanner JE (2005) Edge effects on fauna in fragmented sea-grass meadows. Austral Ecol 30: 210−218

Titlyanov EA, Latypov YY (1991) Light-dependence in scle-ractinian distribution in the sublittoral zone of SouthChina Sea Islands. Coral Reefs 10: 133−138

Toda T, Okashita T, Maekawa T, Kee Alfian BAA and others(2007) Community structures of coral reefs aroundPeninsular Malaysia. J Oceanogr 63: 113−123

Veron JEN (1993) Corals of Australia and the Indo-Pacific.University of Hawaii Press, Honolulu

Veron JEN (2000) Corals of the World. Australian Institute ofMarine Science, Townsville

Wakeford M, Done TJ, Johnson CR (2008) Decadal trends ina coral community and evidence of changed disturbanceregime. Coral Reefs 27: 1−13

Wallace CC (1999) Staghorn corals of the world: a revision ofthe genus Acropora. CSIRO Publishing, Melbourne

Wellington GM (1982) An experimental analysis of theeffects of light and zooplankton on coral zonation.Oecologia 52: 311−320

Williams GJ, Maragos JE, Davy SK (2008) Characterizationof the coral communities at Palmyra Atoll in the remoteCentral Pacific. Atoll Res Bull 557: 1−30

Williams GJ, Smith JE, Conklin EJ, Gove JM, Sala E, SandinSA (2013) Benthic communities at two remote Pacific coralreefs: effects of reef habitat, depth, and wave energy gra-dients on spatial patterns. PeerJ 1: e817717/ peerj.81

Williams GJ, Gove JM, Eynaud Y, Zgliczynski BJ, SandinSA (2015) Local human impacts decouple natural bio-physical relationships on Pacific coral reefs. Ecography38: 751−761

Wilson SK, Graham NAJ, Polunin NVC (2007) Appraisal ofvisual assessments of habitat complexity and benthiccomposition on coral reefs. Mar Biol 151: 1069−1076

Woodley JD, Chornesky EA, Clifford PA, Jackson JBC andothers (1981) Hurricane Allen’s impact on Jamaican coralreefs. Science 214: 749−755

74

Editorial responsibility: Charles Birkeland, Honolulu, Hawaii, USA

Submitted: July 22, 2015; Accepted: February 6, 2016Proofs received from author(s): March 11, 2016

➤

➤

➤

➤

➤

➤

➤

➤

➤

➤

➤

Aut

hor c

opy