Introduction

Reef fish ecologists have long debated the roles of stochastic versus deterministic processes in structuring reef fish communities (Sale 1977, 1978; Anderson et al. 1981; Warner and Chesson 1985; MacNeil et al. 2009). Whilst stochasticity in larval supply is clearly important for some species and at some scales (Sale 1977, 1978; Doherty and Williams 1988), some coral reef fishes exhibit striking patterns in their distribution and abundance corresponding with marked gradients in environmental and habitat characteristics, including aspect, depth and coral cover (Williams 1982, 1991; Williams and Hatcher 1983; Russ 1984a, b; MacNeil et al. 2009), pointing to a high degree of determinism in population and community structure. Understanding key attributes of habitats and environments that structure reef fish assemblages is therefore very important for projecting changes in the distribution and abundance of fishes associated with widespread habitat degradation (Pratchett et al. 2008a, b) and highlights the need for effective habitat management in conserving coral reef fish assemblages (Graham et al. 2008).

One of the most striking patterns in the structure of reef fish communities is the systematic replacement of species within dominant families (Pomacentridae, Chaetodontidae, Acanthuridae and Scaridae) along a cross-shelf gradient on the Great Barrier Reef (Anderson et al. 1981; Williams 1982; Russ 1984a, b; Fulton et al. 2001; Hoey and Bellwood 2008). Such cross-shelf variation in community structure of fishes closely corresponds with changes in coral community types and coral cover (Done 1982; Wismer et al. 2009), algal abundance (Russ 2003) and wave energy (Williams 1982), and it appears likely that the distributions and abundances are structured according to specific environmental tolerances and resource requirements of individual species (Anderson et al. 1981; Fulton et al. 2001).

The degree to which populations and assemblages of coral reef fishes are influenced by physical and biological habitat characteristics is likely to vary spatially, temporally and taxonomically. Importantly, there are certain groups of fishes that are likely to be more constrained by habitat characteristics, due to highly specific resource requirements (Pratchett and Berumen 2008). Fishes from the family Chaetodontidae (butterflyfishes) are among the most specialised of all coral reef fishes. Many are obligate corallivores, feeding almost exclusively on corals (Cole et al. 2008), but even among coral feeders, feeding is concentrated on relatively few coral species (Pratchett 2005). As a consequence, butterflyfish abundances are often positively associated with the percentage of reef substrate occupied by scleractinian (“hard”) corals (e.g., Harmelin-Vivien and Bouchon-Navaro 1983; Bouchon-Navaro et al. 1985; Pratchett et al. 2006; Pratchett and Berumen 2008). However, given marked differences in the feeding ecology of butterflyfishes (Pratchett 2005, 2007), changes in coral composition independent of differences in total coral cover should also influence the species composition and community structure of butterflyfish assemblages (Pratchett and Berumen 2008).

The Great Barrier Reef (GBR) is one of the world’s largest and most complex reef systems, extending >2,000 km along the north-east coast of northern Australia. The GBR occupies the width of the continental shelf, from turbid, coastal waters to clear oceanic waters at the edge of the shelf, providing a multitude of different habitat types. Thus, it provides an ideal opportunity to test associations of reef fishes with different habitat types over large spatial scales. The objectives of this study were:

  1. 1.

    to examine cross-shelf and latitudinal patterns in community structure of benthic reef habitat and butterflyfish assemblages and

  2. 2.

    to assess the extent to which differences in habitat structure account for spatial variation in the distribution and abundance of coral reef butterflyfishes.

It is predicted that community structure of butterflyfish assemblages will differ between inner-shelf, mid-shelf and outer-shelf reefs, as shown in previous studies (e.g., Anderson et al. 1981; Wismer et al. 2009). However, previous studies on cross-shelf variation in fish and benthic communities have been conducted within restricted latitudinal bands (e.g., Anderson et al. 1981; Williams and Hatcher 1983; Russ 1984a, b; Hoey and Bellwood 2008), and so it remains to be seen whether these cross-shelf patterns are consistent along the entire length of the GBR. If so, this would demonstrate the over-arching importance of environmental and habitat conditions in structuring reef fish assemblages.

Methods

Butterflyfishes and coral assemblages have been surveyed annually at 47 distinct reefs along the length and breadth of the Great Barrier Reef (GBR) (Fig. 1) since 1993 as part of the Long Term Monitoring Program (LTMP) by the Australian Institute of Marine Science (AIMS). The 47 study reefs are distributed across six latitudinal sectors, Cooktown/Lizard Island, Cairns, Townsville, Whitsunday, Swain and Capricorn-Bunker (Fig. 1). Within each sector, reefs were surveyed within three cross-shelf positions (inner-shelf, mid-shelf and outer-shelf), where inner-shelf reefs were 5–30 km from land, mid-shelf reefs were 20–200 km from land and outer-shelf reefs were 40–210 km from land, depending on the latitudinal sector (Fig. 1). Typically, three reefs were surveyed in each combination of sector and shelf position. One inner-shelf reef from each of the Cooktown/Lizard Island and Townsville sectors was discounted from the analysis due to reduced visibility precluding surveys being conducted in many years. Also, there are a few suitable inner-shelf reefs in the southern Great Barrier Reef, so sampling was restricted to mid-shelf and outer-shelf reefs; in the Swain sector, surveys were conducted at five mid-shelf and two outer-shelf reefs, and in the Capricorn-Bunker sector, only four outer-shelf reefs were surveyed.

Fig. 1
figure 1

Latitudinal sectors and cross-shelf positions of 47 reefs surveyed annually between 1993 and 2007, to quantify spatial and temporal variation in the abundance and structure of butterflyfishes and coral communities

Full size image

At each reef, annual surveys were conducted at three replicate sites along the northeast flank, wherein five 50-m permanent transects were marked using metal stakes (n = 15 transects per reef). Transects were located within standard reef slope habitat at depths ranging between 6 and 10 m and orientated parallel to the reef crest. All adult butterflyfishes observed within 2.5 m either side of the transect line were recorded by a single diver moving at approximately 10 m per minute. As different sectors were surveyed at different stages of the annual recruitment cycle, all juvenile fishes (<1 year old) that are easily recognisable based on size (Pratchett et al. 2008a) were excluded from counts. Benthic habitats were sampled concurrently on the same transects as the butterflyfishes using digital video recordings of a 30-cm-wide belt centred along each transect line. Video transects were processed using uniform point sampling, whereby 40 evenly spaced frames from each of these video transects were projected and the benthic organisms beneath five equidistant points on each frame were identified to the highest possible taxonomic resolution yielding 200 points per transect. Benthic taxa were categorised as hard corals (Acroporidae, Agariciidae, Dendrophylliidae, Euphyllidae, Faviidae, Fungidae, Merulinidae, Mussidae, Oculinidae, Pectiniidae, Pocilloporidae, Poritidae and Siderastreidae), soft corals (Alcyoniidae, Asterospiculariidae, Briareidae, Clavulariidae, Gorgoniidae, Nephtheidae, Tubiporidae, Xeniidae and other soft corals), fire corals (Milleporidae), algae (coralline algae, green macroalgae, other macroalgae, turf algae), sponges, other (rare benthic organisms of very low abundance e.g., ascidians, anemones) or unknown (organisms unable to be identified due to poor image resolution).

Spatial variability in univariate summary statistics of butterflyfish total abundance and species richness and percentage hard coral cover were assessed using a linear mixed model ANOVA in SPSS v.17. Because we were interested in only the statistical significance of spatial factors and fixed transects were visited annually, we averaged counts from each transect through time for each of butterflyfish abundance, species richness and hard coral cover. This enabled the most robust examination of the spatial factors. Initial analyses indicated the spatial factors accounted for a very large proportion of the variation, whilst temporal factors accounted for only a very small proportion of the overall explained variation (shelf + sector + sector × shelf = 73% of variation explained compared to 1.7% explained by time), a finding collaborated for LTMP data (Burgess et al. 2010). Therefore, we focused solely on the time-averaged spatial factors. The factors sector and shelf were treated as fixed, whilst reef was treated as a random factor and the error term in the analysis. Sites were treated as a random factor nested within reef within shelf within sector. Any test of the interaction terms should be treated with caution due to missing data cells in the sampling design, i.e., not all shelf positions were sampled in every sector (Milliken and Johnson 1992). Butterflyfish abundance was row-centred and fourth root–transformed prior to analysis to homogenise variance and meet the assumption of normality. Hard coral cover was row-centred and square root–transformed to meet the assumptions of normality of data and homogeneity of variances.

Spatial variability in community structure of both butterflyfishes and benthic communities was analysed using non-parametric multivariate analysis of variance (MANOVA). Statistical significance was evaluated using restricted permutation of observations (Anderson and ter Braak 2002). Because sampling was conducted annually along fixed transects (i.e., repeated measures), among-reef variability was used as the error term for the spatial effects of shelf, sector and their interaction. Models were fitted separately for butterflyfishes and benthic communities using the DISTLM program (McArdle and Anderson 2001; Anderson 2001). Data for butterflyfishes were row-centred and fourth root–transformed in order to reduce the over-riding influence of highly abundant species and to adjust for the large numbers of zeros commonly encountered in this type of ecological data.

In order to examine the effects of benthic habitat type on butterflyfish communities, we first examined how the reefs related to each other based on the dominant benthic habitat type rather than the constraints of the sampling design. A hierarchical cluster analysis with complete linkage grouped reefs based on the average annual percentage cover of the benthic groups defined above. Data were row-centred and then square root–transformed before analysis and based on a Euclidian dissimilarity matrix using S-PLUS 2000. The cluster allocations were then imposed on PCA ordination plot based on a Euclidean dissimilarity matrix from the butterflyfish abundance data, which were row-centred and transformed using a Hellinger metric. This allowed a visualisation of the separation of butterflyfish communities based on the dominant benthic habitat type.

To assess the similarity of butterflyfish and benthic communities among the dominant benthic habitat types, an Analysis of Similarity (ANOSIM) was conducted using Primer 6 (Warwick and Clarke 2001). ANOSIM allows tests of the null hypothesis of no assemblage differences between groups of samples specified by levels of factors in the design. Butterflyfish abundance data were row-centred then fourth root–transformed, and the ANOSIM conducted on a Euclidean-based similarity matrix.

Results

Benthic habitats

Average GBR-wide hard coral cover was 29.7 ± 1.2 95% CI per reef per year from the 9,450 transects surveyed since 1993. The average annual percentage cover of scleractinian corals was highest (33.6 ± 0.6 95% CI per reef per year) on outer-shelf reefs and relatively lower on mid-shelf (27.2 ± 0.5 95% CI per reef per year) and inner-shelf reefs (27.0 ± 0.8 95% CI per reef per year), although there was no statistical difference in hard coral cover across the shelf (Mixed model ANOVA: df = 2, F = 0.4, P = 0.7). Nonetheless, a significant sector–shelf interaction term highlighted considerable variation in hard coral cover in different sector–shelf combinations (Mixed model ANOVA df = 7, F = 2.988, P = 0.016). For example, in the Cooktown/Lizard Island and Cairns sectors, the highest average hard coral cover occurred on outer-shelf reefs, whilst in the Swain and Whitsunday sectors, highest average hard coral cover occurred on mid-shelf reefs (Fig. 2). Additionally, there was substantial spatial variation in hard coral cover among sectors (Table 1—mixed model ANOVA: df = 5, F = 534.4, P < 0.001). Average annual coral cover was highest on the southernmost reefs in the Capricorn-Bunker sector and lowest in the Cairns sector, but there were no clear latitudinal gradient (Table 1).

Fig. 2
figure 2

Average annual percentage hard coral cover of reefs in each sector–shelf position. Error bars are 95% CI

Full size image
Table 1 GBR Butterflyfish species abundance by sector, shelf and benthic community type
Full size table

Community structure of reef benthos varied among sector–shelf combinations (non-parametric MANOVA: df = 10, pseudo-F = 6.31, P = 0.001), meaning any cross-shelf patterns varied with latitude (sector). However, this term accounted for only 6% of the total variation explained in this analysis. Additionally, benthic community structure varied among shelf positions (non-parametric MANOVA: df = 2, pseudo-F = 107.4, Perm P = 0.001) and sectors (non-parametric MANOVA: df = 5, pseudo-F = 19.8, Perm P = 0.001). Shelf position had the strongest influence on community structuring accounting for 24% of the variation in benthic communities, compared to sector that accounted for 15%.

Overall, the survey reefs clustered into four main habitat types (Fig. 3). Whilst there was spatial overlap in these habitat types, there was a general cross-shelf progression from coastal to outer-shelf environments: 1. Poritidae/ Alcyoniidae habitats, 2. mixed taxa habitats, 3. soft coral habitats and 4. Acropora-dominated habitats. There were distinct differences in community structure among the four habitat types (ANOSIM Global R = 0.7, Sig. = 0.01%).

Fig. 3
figure 3

Principal Components Analysis (PCA) ordination of benthic communities to illustrate the relationship among reefs with similar benthic habitats. Filled circles depict the average annual mean community type of each survey reef and are coloured by community type. Orange Acropora habitats; green soft coral habitats; maroon mixed taxa habitats; blue Poritidae/Alcyoniidae habitats. Vectors are the benthic families driving the ordinations

Full size image

Poritidae-Alcyoniidae habitats characterised three inner-shelf reefs in the Townsville sector and two in the Whitsunday sector, where the benthic community was dominated by turf algae and hard coral, with high abundance of Alcyoniid soft corals. Among the hard coral community, Goniopora and massive Porites spp. were most abundant. These habitats were also distinguished from others by the presence of Agariciid and Dendrophyllid corals, which were much less abundant in all other habitat types (Fig. 3). Macroalgae occurred in much higher abundance in this habitat (Fig. 3) due mostly to a single reef (Havannah Island) that has experienced a regime shift towards an algal-dominated community (AJ Cheal pers. comm.).

Mixed taxa habitats were dominated by turf algae as well as low abundance of a number of hard coral families, including Acroporidae, Faviidae, Mussidae, Pocilloporidae, Poritidae and Pectiniidae (Fig. 3). These habitats were found predominantly on the mid-shelf reefs of the Cooktown/Lizard Island (3 reefs), Townsville (3 reefs), Whitsunday (3 reefs) and Swain (3 reefs) sectors and reefs on the inner-shelf of the Cairns (3 reefs) and two inner-shelf reefs of the Cooktown/Lizard Island sectors.

Soft coral habitats were composed of a relatively high proportion of Alcyoniidae, Xeniidae and sponges. The proportion of hard coral, soft coral and algae was approximately equal, and coralline algae were more abundant than mixed taxa or Poritidae/Alcyoniidae reefs (Fig. 3). A large proportion of these reefs were positioned on the outer-shelf of the Cairns (3 reefs), Townsville (3 reefs) and Whitsunday (3 reefs) sectors and the mid-shelf of the Cairns (3 reefs) and Swain sectors (2 reefs).

Acropora-dominated habitats had the highest hard coral cover and were comprised of a high proportion of Acroporidae (Fig. 3), mainly tabulate and branching Acropora colonies, (including A. hyacinthus and A. cytherea) and coralline algae, with smaller amounts of Pocilloporidae (Fig. 3). These communities occurred on outer-shelf reefs of the Cooktown/Lizard Island (3 reefs) and Capricorn-Bunker sectors (4 reefs).

Butterflyfish communities

A total of 58,817 butterflyfishes from 30 different species were counted on 8,915 transects over the course of this study, corresponding with a mean density of 26.4 ± 1.2 butterflyfishes per 1,000 m2. Abundance varied among species, and of the nine most abundant species, seven were obligate corallivores (Table 1). Overall abundance of butterflyfishes recorded at each reef in each year was strongly and positively correlated with percentage live coral cover (N = 595, Pearson’s r 2 = 0.67, P < 0.001—Fig 4).

Fig. 4
figure 4

Correlation between annual butterflyfish abundance (mean no. per 1,000 m2) and annual average percentage cover of hard corals across all 47 reefs from 1993 to 2007

Full size image

There was strong spatial structure evident in GBR butterflyfish communities (Fig 5). Total abundance varied among sectors (Mixed model ANOVA sector: df = 5, F = 17.67, P < 0.001) and shelf position (Mixed model ANOVA shelf: df = 2, F = 5.79, P = 0.007), and there was no significant interaction between these two factors (Mixed model ANOVA sector–shelf: df = 7, F = 1.997, P = 0.087). Similarly, species richness varied among sectors (Mixed model ANOVA: df = 5, F = 10.497, P < 0.001) and shelf position (Mixed model ANOVA: df = 2, F = 24.398, P < 0.001); however, there was a significant interaction term (Mixed model ANOVA sector–shelf: df = 7, F = 2.463, P = 0.039), meaning any cross-shelf patterns varied with latitude (Fig. 5).

Fig. 5
figure 5

Average butterflyfish abundance per 1,000 m2 (black bars) and species richness (grey bars) of reefs at sector–shelf localities. Error bars are 95% CI

Full size image

There was a general increase in abundance and species richness from inner-shelf to outer-shelf reefs (Table 1). Whilst cross-shelf patterns of species richness were consistent in each latitudinal sector, there was no such consistency to patterns of abundance (Fig. 5). Mean abundance of butterflyfishes per 1,000 m2 was highest in the Capricorn-Bunker and Cooktown/Lizard Island sectors, whilst Townsville and Whitsunday always had lowest abundance (Table 1). Average annual species richness per reef also varied among sectors; however, there was no consistent pattern with latitude (Fig. 5). The patterns of abundance and species richness were driven in part by the truncated latitudinal distributions of a number of species. Most species were unevenly distributed among sectors (Table 1). Species with northern distributions included Chaetodon auriga, C. ephippium, C. lunulatus and Hemitaurichthys polylepis, whilst southern distributions were displayed by only C. flavirostris (Table 1).

Community structure of butterflyfishes varied among sector–shelf combinations (non-parametric MANOVA sector–shelf: df = 10, pseudo-F = 34.31, Perm P = 0.001), sector (non-parametric MANOVA sector: df = 5, pseudo-F = 33.26, Perm P = 0.001) and shelf position (non-parametric MANOVA shelf: df = 2, pseudo-F = 202.04, Perm P = 0.001). However, shelf position was the strongest structuring factor, accounting for 33% of the variation in butterflyfish, compared to sector and the interaction sector–shelf, which accounted for 23 and 17% of variation in butterflyfish communities, respectively.

Differences in butterflyfish community structure across the shelf and with latitude result in part from the truncated distributions and spatial replacement of some species (Table 1). Butterflyfish communities on inner-shelf reefs were less speciose than those on the mid-shelf and outer-shelf (Table 1). These communities were characterised by two species, Chaetodon aureofasciatus and Chelmon rostratus, which were most abundant and had 63 and 49% of their distribution recorded on inner-shelf reefs. These two species had distributions truncated to the inner- and mid-shelf reefs, rarely being recorded on outer-shelf reefs (Table 1). Of the 30 species surveyed, eight were never recorded on inner-shelf reefs. Outer-shelf reefs had communities characterised by species with distributions truncated to the outer reefs. There were 11 species that were most abundant with greater than 80% of their distributions recorded on outer-shelf reefs (Table 1), and six of these were obligate corallivores. One species of planktivorous butterflyfish, Hemitaurichthys polylepis, was only ever recorded on outer-shelf reefs (Table 1). All 30 surveyed species were recorded on outer-shelf reefs at some time. Butterflyfish communities on mid-shelf reefs appeared to be a crossover of inshore and outer-shelf communities, with no species being exclusively found on these reefs. There were, however, four species with greater than 50% of their distributions found on these reefs including C. rainfordi, C. baronessa, C. plebeius and C. rafflesii, although all of these species were also found on reefs inner-shelf and outer-shelf (Table 1).

The effect of habitat type on community structure was apparent among all shelf positions. Outer-shelf reefs divided into butterflyfish communities associated with Acropora and soft coral habitats; mid-shelf reefs divided into mixed taxa and soft coral communities; and inner-shelf reefs divided into Poritidae/Alcyoniidae, mixed taxa and soft coral communities. Despite hard coral feeders being numerically dominant in all habitats, the dominant species and community composition differed markedly among the four broad habitat types (Fig. 6; Table 1; ANOSIM Global R = 0.6, Sig = 0.1%). The lowest abundance and species richness of butterflyfishes were recorded in Poritidae/Alcyoniidae habitats (Table 1), which were the predominant habitat of inshore reefs. Chaetodon aureofasciatus, a facultative hard coral feeder (Pratchett 2005), was the dominant species in this habitat type and accounted for 62% of butterflyfishes recorded (Table 1). In contrast, Acropora-dominated habitats supported the highest total abundance and species richness of butterflyfishes and were dominated by C. trifascialis (Table 1), a specialist obligate corallivore (Pratchett 2005). C. lunulatus, an obligate corallivore, was the most conspicuous species in soft coral habitats, whilst C. rainfordi and C. aureofasciatus were both the most abundant species in mixed taxa habitats.

Fig. 6
figure 6

Principal Components Analysis (PCA) ordination of butterflyfish communities to illustrate the relationship among reefs with similar benthic habitats. All symbols as for Fig. 4. Vectors are the top 25% butterflyfish species contributing to the overall patterns. The capitalised letters in brackets behind butterflyfish species names indicate trophic affiliations of each species. HC hard coral feeder, GF generalist feeder, NC Non-coral feeder and SC soft coral feeder

Full size image

Discussion

The distribution and abundance of coral reef fishes is often associated with spatial variation in habitat structure (Bell and Galzin 1984; Bouchon-Navaro et al. 1985; Jennings et al 1996; Munday et al. 1997; Pratchett and Berumen 2008; Wilson et al. 2008), and this is particularly true for butterflyfishes. Butterflyfishes, especially obligate corallivores, have a strong association with live coral cover (Halford et al. 2004; Pratchett et al. 2006; Graham et al. 2007; Pratchett and Berumen 2008). Similarly, this study has revealed a strong positive correlation between the overall abundance of butterflyfishes and live coral cover on 47 reefs of the GBR surveyed over 15 years. This suggests that coral cover, and the habitat structure and food resources it provides, is a major factor that regulates abundance of coral reef butterflyfishes. Whilst butterflyfishes generally have low rates of recruitment and may therefore be expected to be recruitment-limited, they are quite long-lived (Berumen 2005), and their populations are very stable independent of major disturbances (Halford et al. 2004). Thus, relatively low rates of replenishment may be sufficient to maintain these populations. Moreover, the exposed reef flanks surveyed in this study are among the habitats that receive highest rates of recruitment by coral-dependent butterflyfishes (Pratchett et al. 2008a).

Despite predictable patterns in overall abundance of butterflyfishes associated with variation in live coral cover, the community structure of butterflyfishes varied among reefs, mostly in response to changes in shelf position and habitat type. Shelf position accounted for the greatest amount of variation in butterflyfish communities, and different community structure was evident in each of the three cross-shelf positions. Other work on the GBR has highlighted cross-shelf differences in community structure of not only reef fishes but also hard corals, soft corals and algae (Anderson et al. 1981; Done 1982; Williams 1982; Williams and Hatcher 1983; Russ 1984a, b; Fulton et al. 2001; Hoey and Bellwood 2008) and identified cross-shelf differences in physical factors such as exposure, water clarity and depth as causal factors. However, differences in butterflyfish community structure across the shelf are largely a result of differential distributions of individual species, whereby some species (e.g., Chaetodon rainfordi) are found predominantly on inner-shelf reefs, whilst others (e.g., Hemitaurichthys polylepis) are only seen on outer-shelf reefs. Such differences are likely to be caused by broad scale larval habitat selection and survival (dependent on specialised adult habitat and food requirements), which are further altered by post-settlement competitive interactions.

Cross-shelf variation in the structure of GBR reef fish assemblages is largely attributable to differences in habitat structure, which is very apparent given fishes associate with reefs of a particular habitat type within each shelf position. Whilst habitat type is strongly confounded with shelf position (e.g., Poritidae/Alcyoniidae habitats were only ever found on inner-shelf reefs, and Acropora-dominated reefs were only ever found on outer-shelf reefs), there were apparent differences among reefs within shelf positions depending on habitat type. Therefore, variation in community structure of butterflyfishes among inner-, mid- and outer-shelf reefs appears to result mostly (if not exclusively) from differences in habitat structure, rather than differences in physicochemical conditions and individual tolerances of butterflyfishes. Among butterflyfishes, which tend to be very similar in overall morphology and physical capabilities (Anderson et al. 1981), it seems the most likely determinant of their distributions are spatial patterns in the abundance of preferred prey (Pratchett and Berumen 2008), though these patterns may be further reinforced by larval settlement preferences over evolutionary time. For many reef species, larvae settle preferentially into adult habitats (Sweatman 1983; Jones 1987; Booth 1992; Lecchini et al. 2005; Pratchett et al. 2008a), and reef fish larvae are quite capable of distinguishing between and navigating to distinct reefs (Jones et al. 2005; Almany et al. 2007; Dixson et al. 2008).

Despite variation in community structure with habitat type, the dominant butterflyfishes species in every habitat type were hard coral feeders. This finding may point to functional replacement of corallivorous species on different reefs (e.g., across the shelf), as was originally suggested by Anderson et al (1981). In Acropora-dominated habitats, Chaetodon trifascialis was always the most abundant butterflyfish and was up to four times more abundant than the next most abundant species. This is perhaps not surprising given that C. trifascialis is one of the most highly specialised butterflyfishes, feeding almost exclusively on Acropora spp. (Pratchett 2007; Berumen and Pratchett 2008). C. trifascialis is also one of the most aggressive species of butterflyfish on the GBR and tends to win competitive interactions with other corallivores, including C. aureofasciatus, C. lunulatus, and C. rainfordi (Berumen and Pratchett 2006b). On reefs where Acropora was less abundant, more generalist corallivorous butterflyfishes dominated; in contrast to C. trifascialis, other dominant corallivores (e.g., C. aureofasciatus, C. lunulatus and C. rainfordi) tend to feed on a wide range of corals (Pratchett 2005, 2007), and their dietary composition often reflects the local availability of different coral prey (Pratchett et al. 2004). Whilst the abundance of dominant competitors (mainly C. trifascialis) is probably regulated by the availability of key resources, these dominant species, in turn, limit the abundance of subordinate species (Berumen and Pratchett 2008), leading to distinct differences in dominance among different habitat types.

Reefs with soft coral habitat types are typically found in locations exposed to moderate wave or current, physical settings that are common on the mid-shelf and outer-shelf (Fabricius and De’ath 1997). Soft coral habitats supported the greatest diversity of different butterflyfishes. The butterflyfish assemblages on these reefs were not dominated by soft coral feeders but by corallivores, a fact that was mirrored in each of the habitat types and shelf position. This suggests that the moderate level of hard coral and Acropora in particular on these reefs is enough to support the habitat requirements of all GBR butterflyfishes surveyed. Conversely, Poritidae/Alcyoniidae communities and inner-shelf reefs generally were less speciose than reefs of other community type and shelf position, suggesting that these reefs failed to provide necessary resources for many species of butterflyfishes. Accordingly, relatively few butterflyfishes will typically feed on Porites (Pratchett 2005), and Berumen and Pratchett (2008) showed that butterflyfishes that specialise on Acropora cannot survive on a diet of Porites. The dominant species, C. aureofasciatus, in Poritidae/Alcyoniidae habitats is one of the few species that actively targets Porites corals (Pratchett 2007), though this probably reflects its poor competitive ability and lack of access to Acropora corals, rather than an innate preference for Porites.

Although marine ecologists continue to seek general explanations for striking and consistent patterns in the abundance of coral reef fishes, it appears likely that the processes that determine the distribution of species vary spatially, temporally and especially taxonomically. In contrast to many smaller species of coral reef fishes (e.g., Pomacentridae and Apogonidae) that exhibit highly dynamic patterns of abundance driven by stochasticity in larval supply, the community structure of butterflyfish assemblages appears to be strongly structured by habitat structure and prey availability. These assemblages with high levels of determinism will also be disproportionately affected by habitat degradation, which is increasingly being caused by climate change and other anthropogenic disturbances (Jackson et al. 2001; Hughes et al. 2003). Accordingly, butterflyfishes are among the most vulnerable of coral reef fishes to climate-induced coral depletion, with many species going locally extinct following severe episodes of coral bleaching (Pratchett et al. 2008b). Marked differences in the structure of butterflyfish assemblages within contrasting habitat types of the GBR may also illustrate likely changes in the structure of fish assemblages, as coral communities adapt to altered environmental regimes (Berumen and Pratchett 2006a). Notably, inshore reefs with lowest abundance and diversity of butterflyfishes were dominated by generalist coral feeding species, whereas offshore reefs with high abundance of corals and butterflyfishes are dominated by high specialised species.