MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 214: 237–251, 2001 Published April 26

INTRODUCTION

Coral reef habitats are highly heterogeneous andmay be subject to strong environmental gradients thatreflect the influence of exposure and reef position rel-ative to coastlines. Consequently reef fishes are sel-dom uniformly distributed and occur in conspicuouspatches over this multi-scalar environment. As the spa-

tial scales on which organisms are studied have majorimplications for our understanding of biological pro-cesses on coral reefs (Sale 1998), sampling designs thatincorporate scale are particularly valuable (Andrew &Mapstone 1987). Despite the acknowledged impor-tance of such designs in determining patterns of nat-ural variation in the distribution and abundance of reefbiota (e.g. Caley et al. 1996), very few studies in the lit-erature have covered more than 3 spatial scales, par-ticularly for reef fishes. The problem of natural spatial

© Inter-Research 2001

*E-mail: nick.gust.@hotmail.com

Spatial variability in reef fish distribution,abundance, size and biomass: a multi-scale analysis

Nick Gust*, J. Howard Choat, Mark I. McCormick

Department of Marine Biology, James Cook University, Townsville, Queensland 4811, Australia

ABSTRACT: This study used underwater visual census techniques to quantify the distribution, abun-dance, fork lengths and biomass of scarid fishes on multiple reefs across the continental shelf of thenorthern Great Barrier Reef. Spatial patterns in fish distribution were examined over a cross shelfenvironmental gradient using a hierarchical sampling design that covered a spectrum of scales rangingfrom metres to tens of kilometres. The design included replicate reefs within mid and outer conti-nental shelf positions and replicate sheltered and exposed sites within each reef. Most of the 21 spe-cies surveyed were found to be widely distributed across mid and outer shelf reefs and the number ofspecies did not change significantly between reefs, exposures or across the shelf. Despite a similarcomplement of species, scarid assemblages differed markedly in the relative and absolute abundanceof taxa between exposed and sheltered habitats on mid and outer shelf reefs. Principal componentsand cluster analyses indicated that changes in the numerical abundance of taxa resulted in consis-tently different assemblages from 9 sheltered mid shelf sites and 9 exposed outer shelf sites at oppo-site ends of the surveyed environmental gradient. Variance components calculated for scarid abun-dance and biomass across 4 spatial scales revealed differences in the scales at which significantvariability occurred between sheltered and exposed reef habitats. On exposed reef crests significantvariation occurred at both scales of tens of kilometres (shelf position) and hundreds of metres (sites),while on sheltered back reefs significant variation was attributable only to the site scale. Mean scaridabundance (±SE) for exposed outer shelf sites (3060 ± 460 ha–1) was on average 4 times higher than inall other habitat zones surveyed, while mean biomass (±SE) was only 3 times higher (920 ± 190 kg ha–1).The discrepancy between abundance and biomass estimates is explained by a 30% average declinein parrotfish fork lengths on exposed outer shelf reef crests relative to other habitats. Log-linearanalysis indicated that at least 8 species of scarids displayed significant changes in body size accord-ing to their location across the continental shelf and/or the prevailing exposure regime. The observedpatterns of variation in density and length frequency suggest density-dependent processes and thatchanges in mortality or growth rates may exist between habitats across the continental shelf.

KEY WORDS: Spatial scales · Exposure · Biomass · Environmental gradient · Body size · Tropical reef fish

Resale or republication not permitted without written consent of the publisher

Mar Ecol Prog Ser 214: 237–251, 2001

variability is especially acute for studies which seek tocompare undisturbed areas with those subject toanthropogenic disturbances.

The continuous expanse of the Great Barrier Reef(GBR) with strong cross shelf gradients provides anexcellent opportunity for exploring the effects of scaledependence in coral reef systems. The GBR marinepark is the largest underwater reserve in the world,stretching over 2000 km and comprising some 2500individual reefs separated by distances ranging from afew hundred metres to tens of kilometres. Understand-ing patterns of distribution and abundance of reeforganisms among reefs is crucial for managing thissystem (Hughes et al. 1999). Moreover for many groupsof reef species on the GBR anthropogenic disturbanceis minimal. These unique circumstances enable inves-tigation of patterns of spatial variation in many specieswithout the complication of direct human disturbance.

Strong latitudinal cross shelf gradients have previ-ously been described across the GBR and were re-viewed by Wilkinson & Cheshire (1988). These authorsdescribed changes in salinity, nutrient input, waterclarity and exposure to prevailing wind and waveswith increasing distance from the coast. They charac-terised inshore reefs as typically receiving variablefreshwater inputs from terrestrial run off which re-duced salinity and water clarity in comparison withmid and outer shelf reefs that experience more oceanicconditions, reduced inorganic and organic nutrientinputs, improved light transmittance and higher expo-sure to wind and waves.

Over the last decade a number of studies have inves-tigated the broad scale distribution and abundance ofcoral reef biota across the continental shelf of the cen-tral GBR that covary with these shifts in abiotic condi-tions. These studies have described the distributionand abundance of a wide variety of taxa including hardcorals (Done 1982), soft corals (Dinesen 1983), holo-thuroids (Hammond et al. 1985), crustaceans (Preston& Doherty 1990, 1994), sponges (Wilkinson & Trott1985), algae (Scott & Russ 1987, Klumpp & McKinnon1992, McCook et al. 1997) and reef fishes (Williams1982, 1983, Williams & Hatcher 1983, Russ 1984a,b,Newman & Williams 1996, Newman et al. 1997). Ineach of these studies major cross shelf differences incommunity structure were demonstrated and, in manycases, differences in the distribution and abundance ofindividual species were pronounced between shelflocations. In the case of reef fishes, none of these stud-ies have revealed changes in the body sizes of taxaacross the continental shelf.

In this study hierarchical or nested sampling wasused to measure spatial variation in the numericalabundance and biomass of reef fishes over spatialscales spanning 4 orders of magnitude. The target

group were fishes of the Family Scaridae, which are anabundant and functionally important element of coralreef fish communities (Williams & Hatcher 1983, Choat1991). These fishes are important targets of reef fish-eries in Asia and the Pacific (e.g. Jennings et al. 1995,Jennings & Polunin 1996, Russ & Alcala 1998), andincreasing human populations suggest increasingexploitation rates in the future. The impetus for thisstudy developed from the observation that a number offish taxa on reef crests of outer barrier reefs appearedmore abundant but smaller than conspecifics on theequivalent mid shelf habitats some 20 km away (Choat& Randall 1986). Scarid fishes were important ele-ments of this pattern, with numerous anecdotal ac-counts of extremely small terminal phase males onouter shelf reef crests exposed to full oceanic influ-ences. As most scarids are protogynous (Choat &Robertson 1975), this situation provided an opportunityto investigate the influence of spatial scale on the plas-ticity of growth, sex change and other life history fea-tures. The initial step in such an investigation requiresthe quantification of patterns of size and abundance inthe target species. This study provides the frameworkfor the analysis of population dynamics and demogra-phy in the species concerned.

The aims of this study were to describe patterns inthe distribution, abundance and species richness ofscarid fishes over a variety of spatial scales across thecontinental shelf of the northern GBR. Furthermore,we aimed to determine the length frequency distribu-tions of scarids on replicate reefs in mid and outer con-tinental shelf positions to examine potential shifts inbody sizes and biomass associated with this naturalenvironmental gradient.

MATERIALS AND METHODS

Study species. Scarid fishes are an excellent familyto target in a study that examines the influence of envi-ronmental gradients on abundance patterns and sizestructure since they are one of the most abundant andwidespread families of reef fishes (Williams & Hatcher1983, Choat 1991) and display particularly high localdensities (Meekan & Choat 1997). Adults of the 27 spe-cies on the GBR can be identified to species using thecriteria of Choat & Randall (1986) and Randall et al.(1997). In addition parrotfishes show little response tothe presence of SCUBA divers, so that underwatervisual censusing (UVC) techniques can provide reli-able measures of distribution, abundance, populationdensity and biomass (Samoilys & Carlos 1992). Existingevidence (reviewed by Parrish 1989) also suggests thatdemersal reef fish such as scarids rarely move betweenreefs, so that differences between reefs are likely to

238

Gust et al.: Spatial variability in reef fish demographics

reflect the dynamics of local populations and will notbe biased by individuals undertaking cross-shelf onto-genetic migrations. Furthermore, since scarids are notcurrently harvested on the GBR, differences cross thecontinental shelf can reasonably be expected to reflectresponses to environmental conditions without thepotentially confounding direct effects of fishing.

Most species of scarids occur within the size range of150 to 450 mm standard length (SL) as adults (Randallet al. 1997), although the very large but rare speciesBolbometopon muricatum exceeds 1000 mm SL. Thisspecies tends to be mobile, with schooling producinghighly clumped, shifting distribution patterns. In termsof body size, feeding activities (Bellwood & Choat 1990)and distribution patterns, B. muricatum represents adistinctive component of the scarid fauna which re-quires different methods for censusing and observation.Pilot studies investigating patterns of B. muricatumabundance and distribution suggest that most indi-viduals occur on outer reef fronts at lowdensities compared to smaller scarids(~0.87 ind. per 270 m2). The focus ofthis study is the abundant fauna ofsmaller scarids, comprising 21 specieswhich occur widely across mid andouter shelf reefs (Russ 1984a,b).

Sampling design. A large-scale visualsurvey was conducted in December1996 and February 1997 using 3 reefsin both mid and outer shelf locations asreplicate study units (Fig. 1). The 3 midshelf reefs surveyed (Lizard Island,North and South Direction Islands)were approximately 25 km from thecoast, while the 3 surveyed outer shelflocations (Hicks, Day and Yonge reefs)were approximately 45 km from themainland. Hierarchical sampling wasadopted since it enables partitioningof variability across a variety of spatialscales and can identify the scales atwhich important patterns and proces-ses occur (Sale 1998). The samplingdesign focused on variability at 4spatial scales: between shelf positions(20 km apart), between reefs withinthe same shelf position (5 to 10 kmapart), between sites (separated by100 to 1000 m), and variability withinsites at scales of tens of metres. Reefsand sites were randomly selected fromavailable localities subject to the logis-tical constraints of distance from theLizard Island Research Station andsuitable anchorages.

Surveyed habitats consisted of either windward reefcrests or the crests and shallow upper slopes of con-tiguous reef habitats on the leeward side of each reef.Surveyed reef areas were typically less than 5 m deepand habitats were selected on the basis that Russ(1984a) found that on the GBR they were characterisedby the highest scarid abundance and species richness.By surveying preferred scarid habitats, this study con-centrated on 4 key zones (mid shelf sheltered, midshelf exposed, outer shelf sheltered and outer shelfexposed) in order to maximise the statistical power ofcomparisons among the various spatial scales of inter-est, and allowed direct comparisons to the findings ofRuss (1984a,b) on the central GBR. Within each reef 3replicate sites were surveyed on both exposed (wind-ward) and sheltered (leeward) sides to address vari-ability within these important habitat zones. Easternsides of reefs on the northern GBR are exposed to pre-vailing southeasterly trade winds, while western sides

239

Fig. 1. Position of the 6 reefs surveyed in December 1996 and February 1997 across the continental shelf of the northern Great Barrier Reef (GBR)

Mar Ecol Prog Ser 214: 237–251, 2001

of reefs were generally sheltered from winds andwaves. In this region trade winds blow steadily ataround 15 knots from the southeast from April toNovember.

The senior author conducted all UVC surveys onSCUBA in 2 to 7 m of water at each of the 36 sites. Spe-cies abundance data at each site was recorded onwater proof data sheets from parrotfish encounteredon 6 replicated 45 × 6 m transect swims. This transectsize was chosen since a study on similar-sized mobileherbivorous fish on the GBR (Hart 1996) compared 4sizes of sampling units and concluded that 45 × 6 mtransects gave optimal precision and accuracy forabundance estimates. Individuals were identified tospecies and their fork lengths estimated to the nearest4 cm size class. Individuals were only included in thecensus if greater than 8 cm in fork length since theinclusion of small fish during the census of a large tran-sect width may result in errors because of the difficultyin detecting small individuals when surveying largeareas (Bellwood & Alcala 1988). The choice of 8 cmminimum size for survey inclusion reflected a conserv-ative approach that also aimed to avoid difficulties inidentifying small, typically cryptic juvenile scarids thatoften school together and share common colour pat-terns (Bellwood & Choat 1989). Furthermore it was notlogistically feasible to simultaneously survey small,cryptic individuals and large, conspicuous adults ofthese species.

To ensure consistency in length estimates, the forklengths of 20 wooden parrotfish models were esti-mated daily underwater prior to commencing surveys.This procedure involved attaching a line in an arc tothe substrate with model fish randomly attached. Thesenior author estimated fork lengths from a distance ofapproximately 5 m from the line, and then checkedestimates by swimming to the opposite side wherelengths were inscribed on each wooden fish. Tests ofthis procedure indicated that length estimations weregenerally accurate to within ±1.5 cm, and were charac-terised by small standard errors (see Fig. 2). Sincelength estimates were not consistently biased, no cor-rections were applied to the visual survey data. Bol-bometopon muricatum has been excluded from theseanalyses because reliable estimates of abundance andlength could not be obtained via these techniques.

Statistical analyses. To investigate cross shelf changesin scarid assemblages, a principal components analysis(PCA) was calculated for log-transformed abundancedata summed at the site level using the 15 most abun-dant species and a covariance matrix. By removing the6 least common species (which represented a com-bined total of less than 1% of individuals surveyed),problems associated with grouping the 36 sites byvirtue of their common absence of rare species were

greatly reduced. The first 2 principal components wereused to examine the patterns of community composi-tion among the 36 sites. A Euclidian bi-plot of speciesvectors was established to determine the contributionof individual taxa to the multivariate assemblage pat-terns. Following the PCA, hierarchical cluster analyseson logged abundance data were undertaken to iden-tify major groupings in the assemblage structure sug-gested by the PCA. A variety of distance measuresincluding Bray-Curtis and Manhattan distance andclustering methods including average linkage, wardsand complete linkage clustering were trialed to inves-tigate the consistency of assemblage clustering pat-terns. Results were broadly consistent among distancemeasures and the average linkage results for Bray-

240

Fig. 2. A comparison of actual and estimated fork lengths ofmodel fishes used to calibrate underwater visual census sizeestimates. Error bars represent ±1 standard error and the fit-ted line represents an ideal relationship between estimated

and actual model fish lengths

Table 1. Length-weight conversions used in biomass calcula-tions. Note that y = weight in g, where x = fork length in cm,R2 is the correlation coefficient for the power curve and n isthe number of individuals used to derive the relationships

Taxon Length-weight R2 nequation

All surveyed scarids y = 0.0181x3.0494 0.994 1078Chlorurus sordidus y = 0.0135x3.1638 0.994 240Scarus frenatus y = 0.0168x3.0725 0.994 150S. ghobban y = 0.0172x3.0627 0.987 20C. gibbus y = 0.0161x3.0922 0.987 150S. niger y = 0.0155x3.1076 0.986 166S. psittacus y = 0.0152x3.1039 0.990 105S. rivulatus y = 0.0256x2.947 0.957 35S. Schlegeli y = 0.0192x3.008 0.987 62

Gust et al.: Spatial variability in reef fish demographics

Curtis distance is presented. Statisticalanalyses were carried out using SPSS(version 8) and Statistica for Windows(version 4.5).

Abundance and biomass. Biomassestimates were made for each fish byconverting UVC length frequency datainto equivalent weights (Table 1). Thelength-weight conversions were de-rived from scarids collected on thenorthern GBR. Length-weight conver-sion equations were species specificfor 9 common parrotfish taxa, while forother species a general power curvederived for all scarids was used to con-vert fork lengths to weights. Weightcalculations assumed fish to be themean length of each 4 cm size classsurveyed.

Initially the number of species, totalabundance and biomass estimates werecompared among habitat categories across the shelfusing a 3-factor mixed-model ANOVA. The modelcontained terms representing the effects of shelf posi-tion (a fixed effect), exposure (also fixed) and reefs as arandom nested factor. Transects from each site arepooled in this broad spatial scale analysis. The appro-priate tests of each term required a variety of MS errorterms and were determined from Quinn & Keough (inpress). Prior to all ANOVA calculations the assumptionof homogeneity of variances was inspected via boxplots, and log(x+1) transformations were used to re-move heterogeneity of variances as required.

Inclusion of the orthogonal factors shelf and expo-sure in the full ANOVA model prevents the calculationof variance components. Therefore to calculate vari-ance components associated with the various spatialscales investigated, exposed windward reef crests andsheltered back reef habitats were examined sepa-rately. In both exposure regimes 3-factor fully nestedANOVA designs were tested with shelf position, reefs(shelf), and sites (reefs) as the main factors. Variancecomponents were derived according to Sokal & Rohlf(1995). Negative variance components for a factor arisewhen the MS between groups is less than the MSwithin groups and this suggests that the spatial scaleconcerned accounts for very little of the overall vari-ance. Negative variance components were convertedto zeros as recommended by Brown & Mosteller (1991).

Size structures. Log-linear modelling was used totest for differences in the length frequency of scarids inresponse to shelf position, exposure and their interac-tion effect. Underwater visual censusing data enabledthe generation of a multi-way contingency table ofscarid length frequencies. In simple terms, log-linear

models use logarithmic transformations to analysemultiway frequency tables and provide a systematicmethod of testing different hypotheses regardinginteractions between variables. Specifically, one maythink of the multi-way frequency table to reflect vari-ous main effects and interaction effects that addtogether in a linear fashion to bring about the observedtable of size frequencies. Shelf position and exposurewere both treated as explanatory variables, while forklength was treated as a response variable. A succes-sion of increasingly complex models (Table 2) were fit-ted to the observed data until there were no significantimprovements in the goodness-of-fit statistic (devianceor likelihood ratio chi square) from one model to thenext. Wrigley (1985) provides a detailed explanation ofconstructing and testing log-linear models using thismethod.

RESULTS

Cross shelf assemblages

PCA revealed that scarid assemblages differedmarkedly only between sheltered mid shelf sites andexposed outer shelf sites. The first 2 axes in the PCA(Fig. 3) explained a total of 53 % of the variability inscarid assemblages across the continental shelf, whilethe third axis represented only an additional 7% and isnot presented. The greatest differences in assemblageswere between sites at opposite ends of the surveyed en-vironmental gradient, from sheltered sides of mid shelfreefs and the exposed sides of outer shelf reefs. Much ofthis difference can be interpreted via the Euclidian

241

Table 2. Log-linear models used to test the effects of shelf position and exposureon the length frequencies of entire scarid assemblages and individual species.Models were tested sequentially until there was not further improvement in thefit of the model to the observed data. In each case ‘Length’ refers to fork length

estimates

Model Factors included Hypothesis tested

1 Shelf × Exposure + Length Length is independent of shelf position or exposure regime

2 Shelf × Exposure + Exposure × Length is influenced by expo-Length sure

3 Shelf × Exposure + Shelf × Length Length is influenced by shelf position

4 Shelf × Exposure + Shelf × Length Length is independently influ-+ Exposure × Length enced by both exposure and

shelf position

5 Shelf × Exposure × Length Length is influenced by an in-teraction between shelf positiontion and exposure (the saturatedmodel)

Mar Ecol Prog Ser 214: 237–251, 2001

bi-plot of the first 2 principal components (Fig. 3). Highabundances of Scarus niger and S. rivulatus were char-acteristic of the sheltered mid shelf reef sites whilethese species were either absent or rare on exposedouter reef crests. In contrast, the 9 sites on exposedouter shelf reef crests were characterised by particu-larly high abundances of S. globiceps and S. psittacus.

Mid shelf exposed sites and outer shelf shelteredsites displayed similar scarid assemblages and these 18sites were not clearly divided by either the PCA (Fig. 3)or cluster analysis (Fig. 4). Average and complete link-age clustering methods on Bray-Curtis, Wards and Man-hattan distance matrices resulted in a wide variety ofgroupings of mid shelf exposed and outer shelf shel-tered sites. This indicates no pronounced differencesin parrotfish assemblages between these 2 habitatzones. In contrast cluster analyses consistently identi-fied 2 zones as having distinct scarid assemblages. The9 mid shelf sheltered sites were identified as being themost dissimilar cluster, and the 9 outer shelf exposedsites were also clearly a distinct grouping (Fig. 4). Itshould also be noted that a single outer shelf shelteredsite from Day reef often clustered with outer shelfexposed sites, as indicated in Fig. 3. This result reflectsthe unusually high numbers of Scarus psittacus and S.globiceps which occurred in this back reef site.

Distribution and abundance of species

Scarid species were typically widely distributedacross the continental shelf, although their abun-dances varied greatly with both shelf position andexposure among the 4 major zones surveyed (Figs.5 & 6). Within each zone, assemblages were char-acterised by a few abundant and many rarer spe-cies, such that abundances were approximatelylog normally distributed between species. Assem-blages in each zone were species rich, typicallyincluding around 20 of the 27 scarid speciesrecorded on the GBR by Choat & Randall (1986).Furthermore there was no evidence of significantchanges in the number of species between reefswithin the same shelf position, or with changingexposure or distance from the coast (Table 3).

Widely distributed species varied considerably inabundance between habitat zones. For instanceChlorurus sordidus, Scarus frenatus and S. globi-ceps displayed consistently high abundances atsites across the northern continental shelf, while S.altipinnis and S. chameleon displayed medium den-sities and S. dimidiatus, S. oviceps and S. spinus dis-played consistently low densities in each of the 4major habitat zones surveyed (Figs. 5 & 6). Instancesof restricted distributions also existed, where Scarusrivulatus was largely restricted to mid shelf reefs,and S. flavipectoralis was only recorded from the

sheltered sides of mid and outer shelf reefs. Individualspecies varied in mean abundance across the continentalshelf by at most 2 orders of magnitude, and considerablechanges are evident in both the relative and absolutedensities of most scarids across this environmental gra-dient (Figs. 5 & 6). C. sordidus was consistently the most

242

Fig. 3. Results of a principal components analysis on logged abun-dance of the 15 most common scarid species across the continentalshelf. Each of the 36 study sites are plotted in relation to the first2 principal components axes (Prin 1 and Prin 2) and a bi-plot ofspecies vectors is included to aid interpretation of site groupings.The 2 major clusters of sites identified by cluster analysis are indi-cated by ellipses. For clarity only the 8 species with important con-tributions to variability between sites in axes 1 and 2 are indicated

Fig. 4. Dendrogram produced from a hierarchical clusteranalysis using log-transformed abundance of the 15 mostcommon scarid species, Bray-Curtis distance and the averagelinkage method. The most dissimilar clusters diverge initially

at the top of this vertical icicle plot

Gust et al.: Spatial variability in reef fish demographics

abundant scarid in 3 of the 4 surveyedhabitat zones in this region, althoughthe highest recorded mean densitiesfor any species existed for S. psittacus(136 ± 35 se per 270 m2) on exposedouter shelf reef crests (Fig. 6).

Cross shelf abundance and biomass

Significant changes in the totalabundance and biomass of scarids oc-

curred with changing shelf position and exposure,although no significant differences existed be-tween reefs nested within shelf position (Tables 4& 5, Fig. 7). A significant reef (shelf) × exposure in-teraction term existed for both scarid abundanceand biomass, and reflects the particularly highnumbers of individuals censused from the 3 ex-posed sites on Yonge reef. Total scarid abundanceestimates were similar amongst 3 of the 4 majorzones (sheltered and exposed mid shelf reef sitesand sheltered outer shelf sites) with mean abun-dances of between 15 and 30 ind. per 270 m2 tran-sect (Fig. 7A). In comparison exposed reef crestson outer shelf reefs displayed approximately 4times higher abundances with means of 70 to 100ind. transect–1 (Fig. 7A). Mean biomass estimatesbetween the 4 surveyed zones (Fig. 7B) broadly re-flected numerical abundance estimates andranged from 3 to 40 kg transect–1 or 11 to 148 g m–2.In comparison with other zones, exposed outershelf reefs displayed approximately a 4-fold in-crease in numerical abundance, but only a 3-foldincrease in biomass. This difference can be ex-plained by the reduced body sizes of many specieson exposed outer shelf reef crests (see ‘Results:Body sizes cross shelf’).

Components of variation calculated at 4 spatialscales are summarised in Fig. 8 along with signifi-cant nested ANOVA results. This figure indicatesthat scales of significant spatial variability in theabundance and biomass of scarids differed be-tween exposed and sheltered reef habitats. Ex-posed reef habitats (Fig. 8A,B) displayed signifi-cant spatial variability associated with both theshelf scale (tens of kilometres) and the site scale(hundreds of metres). In contrast sheltered reefhabitats (Fig. 8C,D) only displayed significant spa-tial variability at the site scale. There was also evi-dence of large but non-significant within-site vari-ation on sheltered reefs, where around 50 to 60%of total variation in abundance and biomass ex-

243

Table 3. A 3-factor ANOVA showing the influence of shelf position, reef withinshelf position and exposure on the number of scarid species across the northern

continental shelf. *p < 0.05; **p < 0.01;***p < 0.001

Source of variation MS df MS df F peffect effect error error

Shelf 0.0185 1 6.6018 4 0.003 0.960Reef (Shelf) 6.6018 4 2.8421 204 2.32 0.058Exposure 2.2407 1 7.5185 4 0.298 0.614Shelf × Exposure 3.6296 1 7.5185 4 0.482 0.525Reef (Shelf) × Exposure 7.5185 4 2.8420 204 2.64 0.034*Residual 2.8420

Fig. 5. Mean scarid abundances per transect (±SE) on (A) mid shelfsheltered sites and (B) outer shelf sheltered sites. Histograms rep-resent mean abundance per 270 m2 averaged across 3 reefs, 9 sites

and 54 transects

Mar Ecol Prog Ser 214: 237–251, 2001

isted between transects at the scale oftens of metres. Replicate reefs withineach shelf position contributed a negli-gible amount to variation in abundanceand biomass irrespective of exposureregime.

Body sizes cross shelf

A statistically significant differenceexisted in the size frequency (forklengths) of scarids according to shelf

position and exposure across the continental shelfof the northern GBR (Table 6, and see Fig. 9). Themost marked difference existed for exposed outershelf reefs, where over 95% of surveyed individu-als were distributed in the 4 smallest length cate-gories. By contrast the 3 other major zones dis-played a more even distribution of length frequen-cies, with the vast majority of individuals spreadthroughout seven 4 cm fork length categories(Fig. 9). This cross shelf pattern is also apparent ifwe simply consider the mean fork lengths of allscarids surveyed in the 4 habitat zones. Meanscarid fork lengths for mid shelf sheltered, midshelf exposed and outer shelf sheltered sites were22.1, 21.6 and 20.8 cm respectively. In contrast, theaverage fork length of all parrotfish on outer shelfexposed sites was only 15.0 cm. This represents anaverage 30% decline in fork length for scarids inthis habitat. The percentage contribution of eachsize class is presented for each reef and exposurecombination (Fig. 10), and indicates that differ-ences in the length frequency of assemblagesbetween replicate reefs within shelf position andexposure categories was small compared withcross shelf trends.

Log-linear analysis revealed a variety of species-specific patterns in body size across the continentalshelf in response to shelf position and/or exposure(Table 7). Sufficiently large sample sizes existed totest length frequencies in 9 species, 8 of whichshowed evidence of reduced fork length on outershelf reefs. This response was most marked for ex-posed outer shelf habitats, where length frequen-cies were markedly smaller in 5 species (Chlorurussordidus, Scarus frenatus, S. globiceps, S. nigerand S. psittacus). Three additional species (S. cha-meleon, S. schlegeli and S. spinus) displayed morecomplex patterns of length frequency distributionacross the northern continental shelf. Log-linearanalysis revealed that fork length in S. chameleonand S. schlegeli was best explained by shelf posi-tion alone (Model 3), and in both cases length was

244

Table 4. A 3-factor ANOVA showing the influence of shelf position, reef withinshelf position and exposure on the abundance of scarids (log transformed)across the northern continental shelf. (*p < 0.05; **p < 0.01; ***p < 0.001)

Source of variation MS df MS df F peffect effect error error

Shelf 1 4.766 4 0.059 80.635 0.0009***Reef (Shelf) 4 0.059 204 0.110 0.536 0.7093Exposure 1 3.842 4 0.391 9.834 0.0349*Shelf × Exposure 1 0.664 4 0.391 1.699 0.2623Reef (Shelf) × Exposure 4 0.391 204 0.110 3.543 0.0081**Residual 204 0.110

Fig. 6. Mean scarid abundances per transect (±SE) on (A) mid shelfexposed reef crest sites and (B) outer shelf exposed reef crest sites.Histograms represent mean abundance per 270 m2 averagedacross 3 reefs, 9 sites and 54 transects. To accommodate high val-ues, the abundance scale for (B) is 5 times larger than that shown

for the other habitats in Figs. 5 & 6

Gust et al.: Spatial variability in reef fish demographics

reduced on the outer shelf regardless ofexposure to prevailing trade winds. Incontrast Table 7 indicates that S. spinuslength frequency was best explainedby Model 4, which suggests that al-though there is no significant shelf byexposure interaction, both shelf posi-tion and exposure independently influ-ence fork length in this species. Only 1species (Scarus altipinnis) showed noevidence of changing length frequencyacross the surveyed habitats, althoughthis result may be compromised by thesmall sample size available.

DISCUSSION

This multi-scale study found that most scarid taxawere widely distributed among replicate reefs in midand outer continental shelf positions, and a similar spe-cies pool was shared between the 4 major habitats sur-veyed (mid shelf sheltered, mid shelf exposed, outershelf sheltered and outer shelf exposed). The only evi-dence of restricted distributions existed for Scarusrivulatus, which was largely confined to mid shelfreefs, and S. flavipectoralis, which was only recordedfrom sheltered reef sites regardless of shelf position.Although a similar complement of species occurred atsites on the exposed and sheltered sides of mid andouter shelf reefs, there were marked differences in theabsolute and relative abundance of key taxa betweenthe most sheltered and most exposed sites. Individualtaxa varied in numerical abundance between habitatsby up to 2 orders of magnitude.

Scarid assemblages from shallow reef habitats atopposite ends of the surveyed environmental gradient(approximately 20 km apart) were characterised bymajor changes in the abundance of 4 taxa. Shelteredmid shelf sites displayed high abundances of bothScarus niger and S. rivulatus while outer shelf exposedreef crests were characterised by high numbers of bothS. psittacus and S. globiceps. Multivariate analysesconfirmed that only assemblages on exposed outershelf reef crests and sheltered mid shelf reef habitatswere distinctively different. The cross shelf patterns ofdistribution and abundance described in this study areunlikely to result from cross shelf movement associatedwith ontogeny, which was suggested for some speciesof lutjanids (Williams 1991), although detailed studiesof scarid movement patterns have yet to be under-taken on the GBR.

On the northern GBR significant spatial variabilityexisted in the numerical abundance and biomass ofscarids on exposed reef crests at both the cross shelfscale of tens of kilometres and the between site scale ofhundreds to thousands of metres. In comparison, sig-

245

Table 5. A 3-factor ANOVA showing the influence of shelf position, reef withinshelf position and exposure on scarid biomass (log transformed) across the

northern continental shelf. (*p < 0.05, **p < 0.01, ***p < 0.001)

Source of variation df MS df MS F peffect effect error error

Shelf 1 3.119 4 0.207 15.006 0.0179*Reef (Shelf) 4 0.208 204 0.116 1.783 0.1334Exposure 1 6.296 4 0.567 11.103 0.0290*Shelf × Exposure 1 0.381 4 0.567 0.672 0.4583Reef (Shelf) × Exposure 4 0.567 204 0.116 4.866 0.0009***Residual 0.116

Fig. 7. (A) Mean abundance of all scarids per 270 m2 transect(±SE). (B) Mean biomass of all scarids per 270 m2 transect(±SE). Open bars: exposed reef crests; shaded bars: sheltered

back reef habitats

Mar Ecol Prog Ser 214: 237–251, 2001

nificant variability in abundance and biomass on backreef habitats was only apparent between sites. Thus itappears that important processes influencing the abun-dance of scarids varies between exposure regimescross shelf in this region. On exposed reef crests scaridnumbers and biomass varied considerably at all spatialscales investigated except among replicate reefswithin the same shelf position. Similarly, on shelteredback reef habitats only 3% of total variability innumerical abundance and 7% of total variability in

biomass could be attributed to the reef scale. Interest-ingly, Hughes et al. (1999) also found this scale failedto account for significant variability in scleractiniancoral abundance on the GBR. Together these studiessuggest that future investigations of patterns of distrib-ution in reef organisms on the GBR would benefit fromreduced replication of reefs within shelf positions, butincreased replication at the site level.

The distribution and abundance of species docu-mented in this study across the continental shelf are

broadly comparable with the findings ofRuss (1984a,b) on the central GBR. Bothstudies described a similar species list ofthe most abundant scarids, found no ev-idence of changing numbers of scaridspecies across the continental shelf andfound that most scarid taxa were widelydistributed between mid and outer shelfreefs. Nevertheless one major differ-ence existed between these studies.The present study described significantchanges in scarid numerical abundancebetween habitats across the continentalshelf. On the northern GBR the meanabundances of scarids on outer shelf

246

Fig. 8. Components of variation at4 spatial scales. Variation in (A) abun-dance and (B) biomass on exposed reefcrests, and (C) abundance and (D) bio-mass on sheltered back reef habitats.Overall variation is partitioned amongscales and expressed as a percentageof the total. Negative variance compo-nents were converted to zeros. Barswith asterisks indicate the spatial scaleswhere there is significant variation, i.e.where substantial differences occurredamong shelf positions, reefs, sites orwithin sites. (*p < 0.05; **p < 0.01;

***p < 0.001)

Table 6. Log-linear analysis of relationships between shelf position, exposureand length frequencies of scarid assemblages across the continental shelf. Themost appropriate model is in bold. A total of 7833 fish length estimates wereused in these analyses. (*p < 0.05, **p < 0.01, ***p < 0.001). A = Shelf

position, B = Exposure, C = Fork length

Model Factors included Deviance df Difference Changebetween models in df

1 A × B + C 2679 36 (1 & 2) 1372*** 12

2 A × B + B × C 1307 24 (1 & 3) 1205*** 12

3 A × B + A × C 1474 24 (2 & 4) 1105*** 12

4 A × B + A × C + B × C 369 12 (4 & 5) 0369*** 12

5 A ×× B ×× C 0.0 0

Gust et al.: Spatial variability in reef fish demographics

reef crests were 4-fold higher than from equivalent reefcrest habitats on mid shelf reefs. In contrast Russ(1984a,b) found no marked changes in the numericalabundance of scarid assemblages between habitatsacross the shelf of the central GBR.

There are 2 possible explanations for this apparentdiscrepancy. Firstly it may be that patterns of scaridabundance differ considerably between mid and outercontinental shelf positions on the central and northern

GBR. The second possibility is that differences inmethodology adopted by the 2 studies explain the dif-ferences in observed patterns. More specifically thediscrepancy between studies may simply reflect thedifferent surveying techniques adopted and varyingstatistical power to detect changes in numerical abun-dance. The present study benefited from censusingindividuals along replicate transects, while the use oflog abundance categories for visual surveys (Russ1984a,b) may have reduced the power to detect spatialchanges in abundance, a point recently acknowledgedby the author (Russ & Alcala 1998).

Mean scarid biomass estimates in this study weresimilar from 3 of the 4 major habitats surveyed (valuesranged from 11 to 40 g m–2), but were significantlyhigher for exposed outer shelf reef crests, where meanbiomass varied from 70 to 150 g m–2. These values areup to an order of magnitude higher than those previ-ously reported for the family from reefs in the Philip-pines, Seychelles and Fiji. In the Philippines, estimatesof the standing stock of scarids varied between loca-tions and researchers. On Apo island reported valuesvaried from 4 to 5.5 g m–2, while on Pamilacan islandthey varied from 3.4 and 4.6 g m–2 (Bellwood & Alcala1988), and on Sumilon island scarid biomass variedfrom 7.9 and 9.2 g m–2 (Russ 1985). Published biomassestimates from marine parks in the Seychelles (Jen-nings et al. 1996) varied between 10 and 22 g m–2,while mean scarid biomass from 6 localities in Fiji werefound to be between 4 and 6 g m–2 (Jennings & Polunin1996). Thus parrotfish biomass on the northern GBRappears exceptionally high, especially on outer shelfexposed reef crests.

It is clear that reef fish taxa vary in abundance anddistribution over a variety of spatial scales. However, todate, few studies on coral reefs have also identified themagnitude of natural variability in reef fish body sizes.This may compromise our ability to detect anthropo-genic impacts such as fishing if the simplistic assump-tion of uniform body sizes in reef fish taxa with widedistributions is accepted. This study provides the firstdetailed evidence of differences in the length frequencyand body sizes of a family of reef fishes across the con-tinental shelf of the GBR. Differences in body size werepronounced. On exposed outer shelf reef crests theaverage fork length of scarids declined by 30% in com-parison with other habitat zones 1 to 20 km away. Fur-thermore these changes in fork length correspond tomuch larger changes in body weight, since weight istypically a cubic relationship of length in teleosts. Suchnatural variability in the body size of reef fishes hasrarely been demonstrated in habitats of the samedepth, although past studies have identified changesin fish sizes associated with water depth (e.g. Clarke1977), or reef profile (Chabanet & Letourneur 1995).

247

Fig. 9. Size frequency distributions for all scarid species sur-veyed from (A) mid shelf sheltered sites, (B) mid shelfexposed sites, (C) outer shelf sheltered and (D) outer shelfexposed sites. Means of each 4 cm size class are representedon the x-axis, and the y-axis represents the total abundance ofscarids derived from 9 sites on either the sheltered or exposedsides of each reef. Note that the scale for (D) is 4 times larger

than in the other panels

Mar Ecol Prog Ser 214: 237–251, 2001

Cross shelf differences in fork lengths cannot be ex-plained by simple sampling artefacts since daily valida-tion with wooden fish models enabled accurate size es-timates to be made. This procedure minimised thelikelihood of sampling biases between days or siteswhich has been cited as a significant source of error inprevious studies using underwater visual surveys (Bellet al. 1985, St. John et al. 1990). Log-linear modellingindicated that the size of at least 8 species changed withshelf position and exposure regime. In 5 of these spe-cies a significant reduction in fork lengths occurred onouter shelf exposed reef crests. These size reductionscould have arisen from a variety of variable processesacross the continental shelf. For instance reduced sizeon outer shelf reef crests may reflect density-dependent

growth and the influence of exceptionally high scariddensities, or differences in growth rates resulting fromdifferent habitat associations across the continentalshelf. It is also possible that reduced fork lengths inthese habitats may reflect higher mortality regimes andhigher recruitment rates such that local populations onouter shelf reef crests are dominated by larger propor-tions of smaller, younger individuals. These potentialsare currently under investigation.

Described changes in the body size of scarids acrossthe continental shelf are likely to have major implicationsfor the way individuals interact with their environment(as reviewed by Werner & Gilliam 1984). Specifically,body size is known to impose important energetic con-straints on individual fishes and affects resource ex-

248

Fig. 10. Size frequency distrib-utions for all scarid species ateach study reef. Histogramsrepresent samples pooled from3 sites on each reef. Histo-grams on the left are from thesheltered back reef sites of thereefs indicated, while those onthe right are from exposed reefcrests sites. Mean lengths ofeach 4 cm size class are pre-sented on the x-axis, while they-axis shows the percentagecontribution of each size class

in the scarid assemblage

Gust et al.: Spatial variability in reef fish demographics

ploitation rates (e.g. Olson 1996).Changes in body size will also com-monly influence the strength, type andsymmetry of interactions with other spe-cies (Werner & Gilliam 1984), where re-duced body sizes on outer shelf reefcrests may for instance increase sus-ceptibility to predators (e.g. Caley1998). Furthermore since female gonadsize is strongly influenced by body sizein teleosts, we predict that individual fe-male scarids on outer shelf reefs mayexperience reduced reproductive out-put in comparison with larger con-specifics on nearby mid shelf reefs.

Depending on the causal mecha-nisms responsible for reduced bodysize, cross shelf changes in scarid bodysize may also have major implicationsfor local population dynamics and lo-cal population’s responses to distur-bance. For instance if density-depen-dent growth is operating on outer shelfreef crests then we may predict thatanthropogenic disturbances, such asfishing, which reduce population den-sities may facilitate compensatory in-creases in growth rates in these habi-tats. However if both higher mortalityand recruitment rates prevail for thesetaxa on outer shelf reefs, we suggestthat significant changes in the popula-tion dynamics and turnover of localpopulations could exist across the con-tinental shelf. In addition the changesin body size documented in this studymay influence the life history charac-teristics of individuals across the conti-nental shelf. General life history the-ory (e.g. Stearns 1992) predicts majorchanges in the timing of key life historyevents such as maturation if changesin mortality and growth regimes existbetween shelf positions. These poten-tial life history consequences are alsocurrently under investigation.

Acknowledgements. This study was sup-ported by a Lizard Island doctoral fellow-ship from the Australian Museum awardedto N.G. and a James Cook University pres-tige grant awarded to M.I.M. Thanks areextended to the staff of the Lizard Islandresearch station for assistance during thecourse of this study, and in particular forkeeping the Condor seaworthy. M. Thomas,

249

Table 7. Log-linear analysis of relationships between shelf position, exposureand estimated fork lengths of scarid species across the continental shelf. Themost appropriate model in each case is in bold. N is the number of length esti-mates obtained for each species. (*p < 0.05, **p < 0.01, ***p < 0.001; NS: non-significant). For explanation of the models tested see Table 2. A = Shelf position,

B = exposure and C = fork length

Model Factors included Deviance df Difference Change Nbetween models in df

Chlorurus sordidus1 A × B + C 394.4 15 (1 & 2) 260.5*** 5 21702 A × B + B × C 133.9 10 (1 & 3) 62.7*** 53 A × B + A × C 331.7 10 (2 & 4) 63.4*** 54 A × B + A × C + B × C 70.5 5 (4 & 5) 70.5*** 55 A ×× B ×× C 0.0 0

Scarus altipinnis1 A ×× B + C 25.3 21 (1 & 2) 7.0NS 7 1132 A × B + B × C 18.3 14 (1 & 3) 13.1NS 73 A × B + A × C 12.2 14 (2 & 4) 5.9NS 74 A × B + A × C + B × C 6.3 7 (4 & 5) 6.3NS 75 A × B × C 0.0 0

S. chameleon1 A × B + C 14.2 9 (1 & 2) 2.0NS 3 1002 A × B + B × C 12.2 6 (1 & 3) 9.9* 33 A ×× B + A ×× C 4.3 6 (2 & 4) 1.7NS 34 A × B + A × C + B × C 2.6 3 (4 & 5) 2.6NS 35 A × B × C 0.0 0

S. frenatus1 A × B + C 146.7 24 (1 & 2) 37.1*** 8 6542 A × B + B × C 109.6 16 (1 & 3) 61.1*** 83 A × B + A × C 85.5 16 (2 & 4) 31.2*** 84 A × B + A × C + B × C 54.3 8 (4 & 5) 54.3*** 85 A ×× B ×× C 0.0 0

S. globiceps1 A × B + C 246.8 12 (1 & 2) 91.7*** 4 8832 A × B + B × C 155.0 8 (1 & 3) 131.3*** 43 A × B + A × C 115.5 8 (2 & 4) 104.4*** 44 A × B + A × C + B × C 11.0 4 (4 & 5) 11.1* 45 A ×× B ×× C 0.0 0

S. niger1 A × B + C 207.0 21 (1 & 2) 96.0*** 7 2162 A × B + B × C 111.3 14 (1 & 3) 95.6*** 73 A × B + A × C 111.7 14 (2 & 4) 60.5*** 74 A × B + A × C + B × C 50.8 7 (4 & 5) 50.8*** 75 A ×× B ×× C 0.0 0

S. psittacus1 A × B + C 285.0 12 (1 & 2) 139.3*** 4 26262 A × B + B × C 145.7 8 (1 & 3) 129.9*** 43 A × B + A × C 155.1 8 (2 & 4) 100.3*** 44 A × B + A × C + B × C 54.8 9 (4 & 5) 54.8*** 45 A ×× B ×× C 0.0 0

S. schlegeli1 A × B + C 20.8 18 (1 & 2) 4.6NS 6 1902 A × B + B × C 16.2 12 (1 & 3) 14.4* 63 A ×× B + A ×× C 6.4 12 (2 & 4) 4.5NS 64 A × B + A × C + B × C 1.9 6 (4 & 5) 1.9NS 65 A × B × C 0.0 0

S. spinus1 A × B + C 32.8 18 (1 & 2) 15.2* 6 1042 A × B + B × C 17.6 12 (1 & 3) 15.7* 63 A × B + A × C 17.1 12 (2 & 4) 15.2* 64 A ×× B + A ×× C + B ×× C 1.8 6 (4 & 5) 1.8NS 65 A × B × C 0.0 0

Mar Ecol Prog Ser 214: 237–251, 2001

S. Watson and D. Brown provided invaluable field assistanceand we gratefully acknowledge P. Munday for constructivecriticism of this manuscript and assistance with log-linearmodelling techniques. Thanks also go to D. Williams and 2anonymous reviewers for constructive criticism of an earliermanuscript.

LITERATURE CITED

Andrew NL, Mapstone BD (1987) Sampling and the descrip-tion of spatial pattern in marine biology. Oceanogr MarBiol Annu Rev 25:39–90

Bell JD, Craik GJS, Pollard DA, Russell BC (1985) Estimatinglength frequency distributions of large reef fish under-waater. Coral Reefs 4:41–44

Bellwood DR, Alcala AC (1988) The effect of a minimumlength specification on visual estimates of density and bio-mass of coral reef fishes. Coral Reefs 7:23–27

Bellwood DR, Choat JH (1989) A description of the juvenilecolour patterns of 24 parrotfish species (family Scaridae)from the Great Barrier Reef, Australia. Rec Aust Mus 41:1–41

Bellwood DR, Choat JH (1990) A functional analysis of graz-ing in parrotfishes (family Scaridae): the ecological impli-cations. Environ Biol Fish 28:189–214

Brown C, Mosteller F (1991) Components of variance. In:Hoaglin DC, Mosteller F, Tukey JW (eds) Fundamentals ofexploratory analysis. Wiley, New York, p 59–84

Caley MJ (1998) Age specific mortality rates in reef fishes:evidence and implications. Aust J Ecol 23:241–245

Caley MJ, Carr MH, Hixon MA, Hughes TP, Jones GP,Menge BA (1996) Recruitment and the local dynamics ofopen marine populations. Annu Rev Ecol Syst 27:477–500

Chabanet P, Letourneur Y (1995) Spatial pattern in of size dis-tribution of four species on Reunion coral reef flats. Hydro-biologia 300–301:299–308

Choat J (1991) The biology of herbivorous fishes on coralreefs. In: Sale PF (ed) The ecology of fishes on coral reefs.Academic Press, San Diego, p 120–155

Choat JH, Randall JE (1986) A review of the parrotfishes(family Scaridae) of the Great Barrier Reef of Australiawith description of a new species. Rec Aust Mus 38:175–228

Choat JH, Robertson DR (1975) Protogynous hermaphro-ditism in fishes of the family Scaridae. In: Reinboth R (ed)Intersexuality in the animal kingdom. Springer Verlag,Heidelberg, p 263–283

Clarke RD (1977) Habitat distribution and species diversity ofchaetodontid and pomacentrid fishes near Bimini, Baha-mas. Mar Biol 40:277–289

Dinesen ZD (1983) Patterns in the distribution of soft coralsacross the central Great Barrier Reef. Coral Reefs 1:229–236

Done TJ (1982) Patterns in the distribution of coral communi-ties across the central Great Barrier Reef. Coral Reefs 1:95–107

Hammond LS, Birtles RA, Reichelt RE (1985) Holothuroidassemblages on coral reefs across the central section of theGreat Barrier Reef. Proc 5th Int Coral Reef Congr 5:285–290

Hart AM (1996) Response of herbivorous reef fish to out-breaks of crown of thorns starfish, with particular empha-sis on the brown surgeonfish (Acanthurus nigrofuscus).PhD thesis, James Cook University, Townsville

Hughes TP, Baird AH, Dinsdale EA, Moltschaniwskyj NA,Pratchett MS, Tanner JE, Willis BL (1999) Patterns of

recruitment and abundance of corals along the Great Bar-rier Reef. Nature 397:59–63

Jennings S, Polunin NVC (1996) Effects of fishing effort andcatch rate upon the structure and biomass of Fijian reeffish communities. J Appl Ecol 33:400–412

Jennings S, Grandcourt EM, Polunin NVC (1995) The effectsof fishing on the diversity, biomass and trophic structure ofSeychelles fish communities. Coral Reefs 14:225–235

Jennings S, Marshall SS, Polunin NVC (1996) Seychelles’marine protected areas: comparative structure and statusof reef fish communities. Biol Conserv 75:201–209

Klumpp DW, McKinnon AD (1992) Community structure, bio-mass and productivity of epilithic algal communities onthe Great Barrier Reef; dynamics at different spatialscales. Mar Ecol Prog Ser 86:77–89

McCook LJ, Price IR, Klumpp DW (1997) Macro algae on theGreat Barrier Reef: causes or consequences, indicators ormodels of reef degradation? Proc 8th Int Coral Reef Symp2:1851–1856

Meekan MG, Choat JH (1997) Latitudinal variation in abun-dance of herbivorous fishes: a comparison of temperateand tropical reefs. Mar Biol 128:373–383

Newman SJ, Williams DM (1996) Variation in reef associatedassemblages of the Lutjanidae and Lethrinidae at differentdistances offshore in the central Great Barrier Reef. Envi-ron Biol Fish 46:123–138

Newman SJ, Williams DM, Russ GR (1997) Patterns of zona-tion of assemblages of the Lutjanidae, Lethrinidae andSerranidae (Epinephelinae) within and among mid shelfand outer shelf reefs in the central Great Barrier Reef. MarFreshw Res 48:119–128

Olson MH (1996) Ontogenetic niche shifts in large mouthbass — variability and consequences for first year growth.Ecology 77:179–190

Parrish JD (1989) Fish communities of interacting shallowwater habitats in tropical oceanic regions. Mar Ecol ProgSer 58:143–160

Preston NP, Doherty PJ (1990) Cross-shelf patterns in thecommunity structure of coral dwelling Crustacea in thecentral region of the Great Barrier Reef. 1. Agile shrimps.Mar Ecol Prog Ser 66:47–61

Preston NP, Doherty PJ (1994) Cross-shelf patterns in thecommunity structure of coral dwelling Crustacea in thecentral region of the Great Barrier Reef. 2. Cryptofauna.Mar Ecol Prog Ser 104:27–38

Quinn G, Keough M (in press) Experimental design andanalysis for biologists. Cambridge University Press, Cam-bridge

Randall JE, Allen GR, Steene RC (1997) Fishes of the GreatBarrier Reef and Coral Sea. Crawford House Press,Bathurst

Russ GR (1984a) Distribution and abundance of herbivorousgrazing fishes in the central Great Barrier Reef. I. Levels ofvariability across the entire continental shelf. Mar EcolProg Ser 20:23–34

Russ GR (1984b) Distribution and abundance of herbivorousgrazing fishes in the central Great Barrier Reef. II. Patternsof zonation of mid-shelf and outer-shelf reefs. Mar EcolProg Ser 20:35–44

Russ GR (1985) Effects of protective management on coralreef fishes in the central Philippines. Proc 5th Int CoralReef Congr 4:219–224

Russ GR, Alcala AC (1998) Natural fishing experiments inmarine reserves 1983–1993: roles of life history and fish-ing intensity in family responses. Coral Reefs 17:399–416

Sale PF (1998) Appropriate spatial scales for studies of reeffish ecology. Aust J Ecol 23:202–208

250

Gust et al.: Spatial variability in reef fish demographics

Samoilys MA, Carlos G (1992) Development of an underwatervisual census method for assessing shallow water reef fishstocks in the south west Pacific. Queensland Departmentof Primary Industries, Cairns

Scott FJ, Russ GR (1987) Effects of grazing on species compo-sition of the epilithic algal community on coral reefs of thecentral Great Barrier Reef. Mar Ecol Prog Ser 39:293–304

Sokal RR, Rohlf FJ (1995) Biometry. WH Freeman and Com-pany, New York

Stearns S (1992) Age and size at maturity. In: Stearns S (ed)The evolution of life histories. Oxford University Press,Oxford

St. John J, Russ GR, Gladstone W (1990) Accuracy and bias ofvisual estimates of numbers, size structure and biomass ofa coral reef fish. Mar Ecol Prog Ser 64:253–262

Werner EE, Gilliam JF (1984) The ontogenetic niche and spe-cies interactions in size structured populations. Annu RevEcol Syst 15:393–425

Wilkinson CR, Cheshire AC (1988) Cross shelf variations incoral reef structure and function — influences of land and

ocean. Proc 6th Int Coral Reef Symp 1:227–233Wilkinson CR, Trott LA (1985) Light as a factor determining

the distribution of sponges across the central Great BarrierReef. Proc 5th Int Coral Reef Congr 5:125–130

Williams DM (1982) Patterns in the distribution of fish com-munities across the central Great Barrier Reef. Coral Reefs1:35–43

Williams DM (1983) Longitudinal and latitudinal variation inthe structure of reef fish communities. In: Baker JT, CarterRM, Sammarco PW, Stark KP (eds) The inaugural GreatBarrier Reef Conference. James Cook University Press,Townsville, p 265–270

Williams DM (1991) Patterns and processes in the distributionof coral reef fishes. In: Sale PF (ed) The ecology of fisheson coral reefs. Academic Press, San Diego, p 437–474

Williams D, Hatcher AI (1983) Structure of fish communitieson outer slopes of inshore, mid-shelf and outer shelf reefsof the Great Barrier Reef. Mar Ecol Prog Ser 10:239–250

Wrigley N (1985) Categorical data analysis for geographersand environmental scientists. Longman, London

251

Editorial responsibility: George Humphrey (ContributingEditor), Sydney, New South Wales, Australia

Submitted: October 3, 1999; Accepted: March 3, 2000Proofs received from author(s): March 28, 2001