Consumption of tabular acroporid corals by reef fishes: a comparison with plant–herbivore interactions

Authors

  • Andrew J. Cole,

    Corresponding author
    1. ARC Centre of Excellence for Coral Reef Studies, James Cook University, Townsville 4811, Queensland, Australia
    2. School of Marine and Tropical Biology, James Cook University, Townsville 4811, Queensland, Australia
    Search for more papers by this author
  • Rebecca J. Lawton,

    1. ARC Centre of Excellence for Coral Reef Studies, James Cook University, Townsville 4811, Queensland, Australia
    Search for more papers by this author
  • Shaun K. Wilson,

    1. Marine Science Program, Department of Environment and Conservation, Kensington, Western Australia, Australia
    2. Oceans Institute, University of Western Australia, Crawley, Western Australia 6009, Australia
    Search for more papers by this author
  • Morgan S. Pratchett

    1. ARC Centre of Excellence for Coral Reef Studies, James Cook University, Townsville 4811, Queensland, Australia
    Search for more papers by this author

Correspondence author. E-mail: andrew.cole3@my.jcu.edu.au

Summary

1. Interactions between primary producers and consumers (i.e. grazers) are of fundamental importance to the successful functioning of ecological communities. Plant–herbivore interactions have been extensively studied, and herbivory has been accepted as an important process contributing to the structure of terrestrial and aquatic ecosystems. In contrast, the functional importance of the ecologically equivalent interaction between scleractinian reef corals and polyp-feeding fishes is largely untested, but has generally been dismissed as unimportant.

2. This study quantified the amount of tabular acroporid coral tissue biomass consumed at the population level by corallivorous butterflyfishes and determined the proportion of both the standing biomass and productivity that is consumed annually at three exposed reef crest sites at Lizard Island, Great Barrier Reef and Australia.

3. Total daily coral consumption ranged from 18·6 (±1·6) to 27·4 (±1·5) g 200 m−2 day−1 with 61–68% of this consumption directed towards tabular acroporid corals. This selective feeding resulted in an annual consumption of between 8·9–13·5% of the total available tissue biomass and between 52–79% of the annual productivity of these tabular acroporid corals.

4. The proportion of standing coral tissue biomass removed by corallivorous butterflyfishes is similar to that removed from terrestrial plants by herbivores. However, the proportion of primary productivity consumed is considerably greater on coral reefs for both corallivorous and herbivorous fishes compared with terrestrial systems.

5. In terrestrial systems, even relatively low levels of defoliation can have significant effects on plant growth rates, seed production and overall fitness. Considering the high proportion of productivity that is consumed by polyp-feeding fishes, it would seem incongruous that these grazing fishes do not have similar effects on coral community structure and population dynamics. Our findings highlight the need to revisit previously held assumptions regarding the functional importance of corallivorous fishes to coral reef ecosystems.

Introduction

Predation is a common process structuring primary producers in all ecosystems. Trophic interactions between primary producers and consumers typically involve modular organisms in which predation results in only partial mortality, enabling the prey organism to regenerate lost tissue and maintain its place in the community (Henry & Hart 2005). One of the most extensively studied interactions between primary producers and consumers is that of herbivory (Crawley 1989, 1997; Huntly 1991; Bigger & Marvier 1998; Maron & Crone 2006). Herbivores exploit a food source that is fundamentally different from that of most other trophic levels. Plants represent a nutritionally poor but widely available, renewable food resource that is rarely consumed in its entirety by herbivores but rather only small proportions are consumed at any one time (Crawley 1983; Huntly 1991; Coley & Barone 1996). Understanding plant–herbivore interactions has been a major research focus of terrestrial ecologists, and there is a general acceptance that grazing herbivores consume a significant proportion of the edible plant biomass within an ecosystem (Crawley 1997; Maron & Crone 2006; Engelkes et al. 2008; Ancheta & Heard 2011). For example, annual defoliation rates typically range between 3–30% of the total leaf biomass which equates to an average consumption of between 10–20% of the annual primary productivity of plant communities (Crawley 1989, 1997; Huntly 1991; Bigger & Marvier 1998).

On coral reefs, primary productivity by microscopic algae living within coral tissue provides much of the energy for corals and reef growth (Muscatine & Porter 1977). Although corals and their symbiotic algae are consumed by numerous reef fishes, the relationship between corallivorous fishes and scleractinian reef corals has received considerably less attention compared with interactions between plants and herbivores in either aquatic or terrestrial systems despite the broad similarities between these interactions (e.g. Carpenter 1986; Hay 1991, 1997; Bigger & Marvier 1998; Maron & Crone 2006). Furthermore, the majority of research investigating the role of corallivorous fishes on reefs has focused on the effects of large bodied scarids and tetraodontids (e.g. Bulbometapon muricatum, Arothron meleagris) which can have strong direct effects on the coral community and limit the distribution and abundance of prey corals (e.g. Neudecker 1979; Wellington 1982; Littler, Taylor & Littler 1989; Miller & Hay 1998; Bellwood, Hoey & Choat 2003; Rotjan & Lewis 2008). These skeletal-feeding corallivores physically damage the coral skeleton when feeding and can remove a large biomass of coral material (Glynn, Stewart & McCosker 1972; Reyes-Bonilla & Calderon-Aguilera 1999, Bonaldo & Bellwood 2011).

In contrast, small-bodied corallivores (e.g. butterflyfishes) consume live coral polyps and tissues but do not harm the underlying coral skeleton. As this predation leaves no visible evidence of damage, the potential effects that polyp-feeding fishes have on prey corals has typically been overlooked by reef ecologists whose focus has been on the relatively large and obvious effects of skeletal feeders (Cole, Pratchett & Jones 2008; Rotjan & Lewis 2008). Furthermore, the amount of coral tissue consumed by these polyp feeders is likely to have been underestimated (Harmelin-Vivien & Bouchon-Navaro 1982; Harmelin-Vivien & Bouchon-Navaro 1983; Cole et al. 2011), which has lead to the assumption that these fishes are of no significance to energy pathways, and thus, polyp-feeding corallivores have generally been assigned a minimal role in the trophodynamics of reef systems (e.g. Hatcher 1988; Hixon 1997; but see Glynn 2004; Cole et al. 2011). Despite this, small-bodied corallivores feed continuously and at very high rates (400–700 bites per hour), are widely abundant on reefs throughout the Indo-Pacific and are ecologically similar to many small herbivores (Tricas 1985; Findley & Findley 2001; Gochfeld 2004; Cole, Pratchett & Jones 2010). Both consume a nutritionally poor food source, are small relative to the size of their prey, have high dietary selectivity and consume only a small fraction of (and rarely kill) potential prey organisms (Cole, Pratchett & Jones 2008). These fishes also forage over relatively small home ranges (circa 20–100 m2), meaning the same coral colonies are repeatedly consumed (Righton, Miller & Ormond 1998; Gochfeld 2004). As such, polyp-feeding corallivores are likely to play a significant role in coral reef trophodynamics and we expect that the effects of continual tissue loss on scleractinian corals from polyp-feeding corallivores to be comparable with that of small herbivores on plants in terrestrial systems.

Predation by polyp-feeding corallivorous fishes rarely results in direct coral mortality, but rather represents a long-term chronic stress on coral populations (Gochfeld 2004; Pratchett 2007). For example, a recent study found that each adult butterflyfish consumes up to 3 g of coral tissue each day, suggesting that these fishes represent an important trophic link between corals and higher consumers (Cole et al. 2011). Moreover, most species show highly convergent patterns of prey use, feeding disproportionately upon species of tabular acroporid corals (e.g. Acropora cytherea and Acropora hyacinthus) (Fig. 1b) (Irons 1989; Pratchett 2005, 2007; Berumen, Pratchett & McCormick 2005; Cole & Pratchett 2011). For example, the chevron butterflyfish, Chaetodon trifascialis, feeds almost exclusively upon these table corals regardless of their abundance (Irons 1989; Pratchett 2005, 2007). Even when tabular acroporids comprised only 0·32% of the coral community, this species still took 82·7% of its bites from these colonies (Irons 1989). The ecological significance of this highly selective tissue loss is currently unknown. However, we expect the constant loss of coral tissue to act as a drain on energy reserves, where investment in regeneration of lost tissue occurs at the expense of growth and reproduction. But before the ecosystem effects of chronic tissue consumption can be assessed, the magnitude of tissue loss relative to the rate of production first needs to be estimated (e.g. Harmelin-Vivien & Bouchon-Navaro 1983; Hay 1997).

Figure 1.

 (a) An adult corallivorous butterflyfish Chaetodon trifascialis (foreground) and Chaetodon baronessa (background). (b) Colony of a tabular acroporid (Acropora hyacinthus).

The aim of this study is to expand on the findings from Cole et al. (2011) which assessed the amount of coral tissue consumed by individual butterflyfishes and used estimates of corallivore abundances to estimate the amount of coral tissue consumed from the reef at the population level. In the current study, we extend these findings by determining the proportion of this consumption directed towards tabular acroporid corals and estimate the proportion of both the standing biomass and potential productivity that this consumption represents. Tabular acroporids are a functionally important group of corals on reef crest habitats, providing much of the topographic complexity which plays a key role in enhancing the diversity of reef fish assemblages (reviewed by Pratchett et al. 2008). Likewise these corals are an important food resource and support the high diversity and abundance of Chaetodon butterflyfishes that occupy these reef habitats (Pratchett 2005). The potential impacts of chronic tissue consumption on coral colonies and reef trophodynamics will be explored using existing knowledge of plant–herbivore interactions from terrestrial ecosystems.

Materials and methods

Study sites

This study was conducted at Lizard Island (14°409 S; 145°279 E), a continental island on the northern Great Barrier Reef (GBR), 35 km off the eastern coast of Australia. Sampling was conducted at three locations on the Northern, Eastern and Southern sides of Lizard Island: North Reef, South Island and South West Palfrey’s. These three locations are separated by distances of 2–8 km but are all characterized by extensive contiguous fringing reef. In order to estimate the amount of live tissue removed from tabular acroporid corals, data were collected on the size and density of all coral-feeding fishes, their feeding rates, and the amount of tissue removed per bite by fishes of different size. Estimates of the total annual biomass of live coral tissue removed by coral-feeding fishes were then compared with estimates of net accumulation (proxy for productivity) for tabular Acroporid corals, to determine the proportion of live coral growth removed by coral-feeding fishes.

Corallivore abundance and coral community structure

Absolute densities and total lengths of all obligate coral-feeding butterflyfishes were quantified using underwater visual census, with 10 replicate 50 × 4 m belt transects surveyed at each site. Replicate transects were laid parallel to and approximately 3 m from the edge of the reef crest. For all fishes recorded on transects, total length was visually estimated. Size estimations were validated by catching a subset of fish and comparing estimated vs. actual length; all cases were found to be accurate to within 5 mm.

The composition of the coral community on each of the ten transects used for fish census was assessed using the line point transect methodology. Line point transects have recently been assessed as the preferred method in sampling coral communities, giving the highest accuracy and precision when comparing between sites, they also have the advantage of sampling the coral community at the same scale as the fish community (Nadon & Stirling 2006). On these transects, the substrate directly underneath the tape was recorded to species level for scleractinian corals and to broad categories (e.g. sand, rubble, pavement and macroalgae) for all other substrate types every 25 cm, providing 200 sample points per transect. These data were used to determine the proportional cover of tabular acroporid corals at each site (Table 1, step 1).

Table 1.   Calculation of the proportion of tabular acroporid coral tissue biomass and productivity consumed annually by obligate coral-feeding butterflyfishes
 North ReefSouth IslandSouth-West Palfrey’s
  1. Values are the site means with standard error in parenthesis.

  2. A, converted using a 2D planar area to 3D skeletal surface area ratio of 1 cm2:4·94 cm2; B, converted using a tissue mass to surface area ratio of 1 g:36·68 cm2.

Coral availability
1. Proportional cover of tabular acroporids (%)18·6 (2·3)17·6 (2·5)24·8 (2·5)
2. Planar surface area of tabular acroporids (m2 200 m−2)37·2 (4·5)35·2 (5·1)49·6 (5·0)
3. Equivalent 3D surface area (m2 200 m−2) (A)183·8 (22·3)173·9 (25·0)245·0 (24·7)
Proportion of tissue biomass consumed
4. Mean coral tissue consumed (g 200 m−2 day−1)27·4 (1·5)18·6 (1·6)26·9 (1·3)
5. Proportion (%) of bites taken from tabular acroporids67·661·666·9
6. Total tissue removed from tabular acroporids (g 200 m−2 day−1)18·5 (1·2)11·5 (0·9)18·0 (1·2)
7. Total 3D area of tissue removed (cm2 200 m−2 day−1) (B)678·6 (43·0)421·8 (32·9)660·2 (42·3)
8. Total area of tissue removed annually (m2 200 m−2 year−1) (step 7 × 365 ÷ 10000)24·8 (1·6)15·4 (1·2)24·1 (1·5)
9. Proportion (%) of tabular acroporid tissue consumed annually (step 8 ÷ step 3)13·5 (0·7)8·9 (0·5)9·8 (0·4)
Proportion of productivity consumed
10. Increase in the proportional cover of tabular acroporids (%) (step 1 multiplied by 117%)21·8 (2·6)20·6 (3·0)29·0 (2·9)
11. Planar surface area of tabular acroporids (m2 per 200 m2)43·5 (5·3)41·2 (5·9)58 (5·8)
12. Change in planar surface area of tabular acroporids (m2 per 200 m2) (step 11 – step 2)6·3 (0·8)6 (0·9)8·4 (0·8)
13. Equivalent 3D surface area (m2 per 200 m2) (A)31·1 (4·5)29·6 (5·1)41·5 (5·0)
14. Proportion of tabular acroporid productivity consumed annually (%) (step 8 ÷ by step 13)79·7 (3·3)52·0 (2·4)58·1 (2·7)

Diet composition

To determine the proportional consumption of different coral species, 3 min feeding observations were performed on 40 individuals of each of the three most abundant obligate corallivorous butterflyfishes at each site: Chaetodon baronessa, Chaetodon lunulatus and C. trifascialis, following Pratchett (2005). These three species accounted for 90% (313/348 individuals) of the obligate coral-feeding butterflyfishes recorded on the reef crest at these three sites. During feeding observations, each individual was observed from a distance of 3 m and the substrate upon which all bites were taken was recorded (following Pratchett 2005). A total of at least 40 observations were conducted for each of the three species at each site. These data were used to determine the proportion of bites taken by each of the three species at each site on tabular acroporids (Table 1, step 5).

Proportion of tabular acroporid tissue biomass consumed

To assess the proportion of tabular acroporid biomass that is consumed annually by corallivorous butterflyfishes, we combined estimates of the amount of coral tissue eaten by fishes with the total available coral tissue biomass at each of the three sites. First, we converted the proportional cover of tabular acroporids into an equivalent planar surface area on a 200-m2 transect (Table 1, step 2). This is an appropriate conversion as tabular acroporids have a relatively flat two-dimensional table shape. As such the top down proportional cover will have only minor deviations from the equivalent planar surface area of substrate that is covered. Second, to assess the amount of coral tissue biomass available on a coral colony, we calculated coral tissue mass-to-surface area ratios and a conversion factor which enabled the actual three-dimensional skeletal surface area to be converted to an equivalent two-dimensional planar surface area. These ratios enabled us to accurately estimate the amount of coral tissue available as it incorporates coral tissue from all surfaces of the coral colony. To calculate these ratios, one large branch section (10–15 cm in length) was removed from 15 randomly chosen colonies of tabular acroporids. Each of these branches was given five abrupt shakes and then drip dried upside down on absorbent paper for 5 min to remove surface water. Each branch was then wet weighed with and without live tissue; coral tissue was removed by immersion in chlorine solution. The difference in weight between the two measurements corresponded to the amount of coral tissue on each branch. Top down digital photographs were taken of each branch which allowed the planar surface area to be calculated using the CPCe software (Kohler & Gill 2006). The three-dimensional skeletal surface area of each branch was determined using the parafin wax technique following Vytopil & Willis (2001). To relate the increase in mass (g) between wax coatings to surface area, a calibration relationship was determined using 15 PVC cylinders of known surface area ranging from 2·21 to 589·59 cm2. The regression relationship between increases in mass (x) and surface area (y) of the cylinders (y = 45·212x − 0·6756, r2 = 0·97) was then used to calculate the three-dimensional skeletal surface area of each coral branch. To calculate the tissue mass-to-surface area ratio of each branch, we divided the three-dimensional skeletal surface area by the tissue weight for each branch. This gave a mean tissue mass-to-surface area ratio of 1 g coral tissue to 36·68 cm2 (±1·68 SE). Likewise, the three-dimensional skeletal surface area of each branch was divided by its planar surface area to provide a mean conversion factor of 1 cm2 of two-dimensional planar area equates to 4·94 cm2 (±0·46 SE) of three-dimensional surface area. This conversion factor was multiplied by the planar surface area of tabular acroporid corals (calculated in step 2) for each of the three sites to provide the equivalent three-dimensional surface area of tabular acroporid coral tissue available at each site (Table 1, step 3).

To assess the amount of coral tissue consumed by corallivorous butterflyfishes, we used size-specific regression equations determined in a previous study (Cole et al. 2011) which link the total length of individual butterflyfishes of each species with their daily coral consumption (Table 2). These regression equations were used to calculate the amount of coral tissue removed per day by each coral-feeding butterflyfish observed on each transect to provide an overall estimate of the total amount of coral tissue removed by all coral-feeding butterflyfishes per day per 200 m2 at each of the three study sites (Table 1, step 4). Two butterflyfish species (Chaetodon rainfordi and C. trifascialis) were detected on these transects for which size-specific regression equations are not available. In these cases, regression equations of the closest related species were used (C. aureofasciatus and C. baronessa respectively) (Bellwood et al. 2010). The total amount of coral tissue removed per day by coral-feeding butterflyfishes at each site (Table 1, step 4) was then multiplied by the proportion of bites taken by each species on tabular acroporids (A. hyacinthus and A. cytherea) at each of the three sites based on our feeding observation data (Table 1, step 5) to determine the total amount of tissue removed from tabular acroporids per 200 m2 per day (Table 1, step 6). Chaetodon aureofasciatus, C. rainfordi and Chaetodon plebeius are generally rare on the reef crest habitat and accounted for <10% (31/344) of the total abundance of obligate corallivores in our study. As such we did not perform feeding observations on these species but rather used previously published estimates on the proportion of their diet that is focused on tabular acroporids, with these values ranging between 15% and 20% (Pratchett 2005).

Table 2.   Size-specific regression equations determining the amount of coral tissue (y) removed for a fish of a given size (x) for four species of corallivorous chaetodontids, based on data from Cole et al. (2011)
SpeciesConsumption equation
Chaetodon baronessay = 0·0348x − 1·2645
Chaetodon lunulatusy = 0·0308x − 1·1813
Chaetodon aureofasciatusy = 0·0313x − 1·1401
Chaetodon plebeiusy = 0·03x − 1·3947

The tissue mass-to-surface area ratio was multiplied by the daily amount (g) of coral tissue removed from tabular Acropora colonies (Table 1, step 6) to convert the total amount of coral tissue consumed per day to an equivalent three-dimensional surface area (cm2) of coral (Table 1, step 7). To calculate the proportion of tabular acroporid tissue consumed annually by corallivorous butterflyfishes, we multiplied the areal equivalents of daily coral consumption estimates by 365 (Table 1, step 8) and divided by the actual three-dimensional surface area of tabular acroporids tissue that was available at each site (Table 1, step 9). To calculate the error around our estimate of annual consumption, we repeated these calculations three times; once using the mean value for each variable in all calculations, once using the mean plus one standard error in all calculations and once using the mean minus one standard error, with our final error values representing the deviation from the mean in our three calculations. We do not expect the extrapolation from daily to yearly consumption estimates to be distorted by seasonal differences in feeding rates. Unlike herbivorous fishes, corallivorous butterflyfishes are surprisingly consistent, and no study has documented a significant effect of season on the feeding rates of these fishes (e.g. Cox 1986; Irons 1989; Pratchett 2005 etc.).

Proportion of coral productivity consumed

To estimate the proportion of primary productivity of tabular acroporids that is consumed by corallivorous fishes, we used the annual increase in proportional cover as a proxy for productivity. To calculate the average annual increase in proportional cover of the family Acroporidae, Thompson & Dolman (2010) used data collected annually (1985–2007) from 36 near shore reefs of the GBR (between 16°S and 23°S) by the Australian Institute of Marine Sciences long-term monitoring programme. Thompson & Dolman (2010) used data only during periods that had no major disturbance events (e.g. cyclones, crown-of-thorns starfish and coral bleaching events) to calculate the proportional increase of the coral community. The Acroporidae increased in proportional cover at an average annual rate of 16·9% (Thompson & Dolman 2010). While this estimate of annual growth does not include estimates from mid-shelf reefs, it does represent a wide spatial and temporal data set of growth rates and there are no other estimates of annual areal increase available for acroporid corals on the GBR. As such it is the best available estimate for determining the potential productivity of tabular acroporids at Lizard Island. To calculate the net accumulation and potential productivity of tabular acroporid corals at Lizard Island, the average proportional increase (16·9% per annum) was multiplied by the current proportional cover of tabular acroporids at each of the three sites (Table 1, step 10). This amount was then converted to a 2D planar surface area per 200 m2 (Table 1, step 11). The annual change in planar surface area of tabular acroporids was calculated based on the growth rates reported in Thompson & Dolman (2010) (Table 1, step 12), and then converted to an equivalent three-dimensional skeletal surface area (Table 1, step 13). To calculate the annual amount of potential productivity that is consumed by coral-feeding butterflyfishes (Table 1, step 14), the total area of tabular acroporid tissue consumed annually by corallivorous butterflyfishes (Table 1, step 8) was divided by the total annual increase in acroporid tissue due to growth (Table 1, step 13).

While the approach of using net accumulation as a proxy for productivity is not ideal and will result in an overestimation relative to the consumption of actual net primary productivity (which is measured as gross primary productivity minus respiration), it does represent a comparable method to previous studies which have typically used some aspect of biomass increase between caged and uncaged plots as a proxy for productivity (Cargill & Jefferies 1984; McNaughton 1985; McNaughton, Milchunas & Frank 1996; Russ & McCook 1999; Russ 2003; Vanderklift, Lavery & Waddington 2008; etc.). In our study, we have used the term ‘potential productivity’ and acknowledge that true net primary productivity will be higher than measured here.

Results

Corallivore abundance and coral community structure

A total of 348 obligate coral-feeding butterflyfishes were detected across the three sites with 90% of these belonging to just three species, C. baronessa, C. trifascialis and C. lunulatus. The mean total density of obligate coral-feeding butterflyfishes differed significantly (anovaF2,27 = 6·96, < 0·01) among the three sites with North Reef (13·2 ± 0·9 per 200 m2) and South West Palfrey’s (12·4 ± 0·7 per 200 m2) having a significantly (Tukey’s HSD P < 0·05) higher density compared with South Island (9·2 ± 0·8 per 200 m2). The community composition differed slightly between the three sites, with North Reef having a significantly (Tukey’s HSD P < 0·05) higher density of C. baronessa than either of the two other sites, with a mean density of just over double South Island (Fig. 2).

Figure 2.

 Mean density of corallivorous butterflyfish at the three study sites. Other obligate coral feeders included Chaetodon plebeius, Chaetodon aureofasciatus and Chaetodon rainfordi. Values are the means and standard errors of the number of fish counted on ten 50 × 4 m transects at each site.

Total hard coral cover was significantly higher at South West Palfrey’s 52% (±2·2), than South Island 38% (±2·9) or North Reef 38·2% (±3·4) (anovaF2,27 = 7·96, < 0·01). Coral community structure varied little between the three sites with corals from the Acropora genus accounting for 70–80% of the live coral cover. Tabular acroporids were the single most abundant group within this family accounting for 18–25% of the proportional cover among sites. The remaining coral community was made up of 25 species from seven families with these species having a low relative cover with no individual species ever accounting for more than 3% of the total cover (Fig. 3).

Figure 3.

 Diet composition of the three most abundant corallivorous butterflyfish; Chaetodon trifascialis, Chaetodon baronessa and Chaetodon lunulatus and coral availability at the three exposed reef crest habitats: (a) North Reef, (b) South West Palfrey’s and (c) South Island. Values are the proportion of the total bites taken on each coral resource category during 3 min observations of 40 individuals of each species at each site. Coral availability is presented as the mean and standard error of the proportional cover of each coral resource averaged across the 10, 50 m point-intercept transects at each site.

Diet composition

Among the three study species, a total of 14,113 bites were observed with 63% (8914/14,113) of these taken on colonies of tabular acroporids. Diet composition varied little between the three study sites for each species. Chaetodon trifascialis was the most selective taking between 83·9% and 91·6% of bites from tabular acroporids with the remaining bites focused primarily on other Acropora species. Chaetodon baronessa was also highly selective in its feeding and took between 68·3% and 80·3% of bites from tabular acroporids with its remaining diet evenly dispersed between other Acropora spp. and Pocillopora spp. Chaetodon lunulatus had a generalist diet and fed evenly between all available corals. Tabular acroporids accounted for 28·4% and 34·5% of C. lunulatus bites at South Island and North Reef respectively, but at South West Palfrey’s, the main dietary item was other Acropora spp. (50·7% of observed bites) and tabular acroporids only accounted for 25·7% of bites (Fig. 3).

Total biomass and annual consumption of productivity of tabular acroporids

Total daily coral consumption by all obligate corallivores varied significantly (anova, F2,27 = 13·42, < 0·001) between the three reefs and ranged from a low of 18·63 (±1·6) g 200 m−2 day−1 at South Island to highs of 26·9 (±1·3) g 200 m−2 day−1 and 27·4 (±1·5) g 200 m−2 day−1 at South West Palfrey’s and North Reef respectively. Approximately 60% of total coral consumption by obligate coral-feeding butterflyfishes is on tabular corals (Table 1) which relates to an annual consumption of 9–13·5% of the tabular acroporid tissue biomass available at the three sites (Table 1). If tabular acroporids are increasing in proportional cover at a rate of 17% per annum, this would increase the areal cover of tabular acroporids by 6–8·5 m2 which means corallivorous butterflyfish consume approximately 52–79% of the annual potential productivity of tabular acroporid corals (Table 1).

Discussion

This study demonstrates that coral-feeding butterflyfishes consume a significant proportion of both the standing biomass and annual potential productivity of tabular acroporid corals. This challenges the idea that coral-feeding butterflyfishes consume only minor amounts of coral tissue (Harmelin-Vivien & Bouchon-Navaro 1982, 1983). Further, these estimates are likely to be conservative as there are many other corallivorous fishes that feed on corals whose consumption rates were not considered in our calculations. For example, at our study sites, the facultative corallivore Chaetodon citrinellus was also relatively abundant with a mean density of between 3–5 individuals per transect and tabular acroporids account for approximately 20% of the diet of this species (Pratchett 2005). Likewise, many reefs throughout the Indo-Pacific also have obligate polyp feeders from other families (e.g. Labridae) (Cole, Pratchett & Jones 2010). In the present study, we could not assess the contribution of these corallivores, as we do not have any data on bite sizes and daily coral consumption rates. With these limitations in mind, if the corallivore guild is considered as a whole, the total biomass consumed will be considerably larger than the amount estimated here, calling into question the assumption that chronic tissue loss will have only minor impacts on coral populations.

In terrestrial systems, herbivores are often highly selective in the range of plant species that they consume and, generally, the smaller the herbivore the higher the level of dietary specialization (Crawley 1989, 1997; Bigger & Marvier 1998; Ancheta & Heard 2011). Similarly, coral-feeding fishes are highly selective in their feeding behaviour and consume preferred coral prey disproportionately to their abundance (Irons 1989; Berumen, Pratchett & McCormick 2005; Pratchett 2005, 2007; Lawton et al. in press). At Lizard Island, the proportional cover of tabular acroporids ranged from 17% to 25% of the total benthos, but 63% (8914/14,113) of the observed bites by the three most abundant corallivores (C. baronessa, C. lunulatus and C. trifascialis) were taken on these corals. This selective predation translates to an annual consumption of between 9–13·5% of the standing biomass of tabular acroporid colonies in reef crest habitats. These total consumption estimates are within the range observed in terrestrial systems where herbivores typically consume between 3–30% of the total leaf area in an ecosystem annually (Crawley 1983, 1997; Lowman 1992; Coley & Barone 1996; Ancheta & Heard 2011). In terrestrial systems, even relatively low levels of defoliation can have significant effects on the fitness of certain plant species. For example, persistent low levels of insect herbivory on oak trees removed 8–12% of the total leaf area annually; this tissue loss resulted in significantly reduced seed production relative to control trees (Crawley 1985). As such, it is likely that a corresponding level of tissue loss through corallivory could have comparable impacts upon corals.

While the total biomass consumed by corallivorous fishes is similar to terrestrial herbivores, our estimate that between 52–79% of the potential productivity of tabular acroporids is consumed by corallivorous butterflyfishes annually is much higher compared with similar studies in terrestrial systems. Herbivores, on average, consume only 10–20% of the annual productivity (Crawley 1983, 1997; Cyr & Pace 1993; Bigger & Marvier 1998). Within these estimates, though there is considerable variation between habitats, and the amount of productivity consumed can range as high as 60% for highly productive grazing lawns in the Serengeti (McNaughton 1985; Frank, McNaughton & Tracy 1998). Marine herbivores consume a much higher proportion of productivity than their equivalents in terrestrial ecosystems (Carpenter 1986; Cyr & Pace 1993; Hay 1991, 1997). Herbivorous fishes (F: Scaridae, Acanthuridae), for example, consume between 40–100% of the daily productivity of turf algal communities on coral reefs, with this consumption highest on the reef crest (Hatcher 1981; Carpenter 1986; Russ 1987; Klumpp & Polunin 1990; Van Rooij, Videler & Bruggemann 1998). Intense grazing by these fishes maintains turf algal communities in a highly productive but low standing biomass state, with these fishes representing the largest energy flux on coral reefs (Carpenter 1986; Russ 2003). In both marine and terrestrial systems fast growing, highly productive algal/plant species are frequently eaten while slower-growing species are generally avoided or consumed in low proportions as a consequence of chemical or physical defences which make them unpalatable (Gochfeld 2004; Endara & Coley 2011). In a meta-analysis that related terrestrial plant growth rates to the proportion of photosynthetic biomass consumed daily by herbivores, Cebrian & Duarte (1994) demonstrated that herbivory increased with plant turnover rate and fast-growing species supported a disproportionately larger herbivore pressure compared with slower-growing counterparts. They concluded that herbivory is likely to be an important mechanism depressing plant biomass in fast-growing plant communities (Cebrian & Duarte 1994). A similar result is likely occurring on coral reefs in which the biomass of herbivorous fishes correlates more strongly with algal productivity than algal biomass (e.g. Russ 2003). Likewise, a similar relationship with productivity may explain why corallivorous fishes reach their highest densities on the reef crest where the fast-growing Acropora corals also reach their highest densities (Pratchett & Berumen 2008).

The combined proportion of productivity that enters the foodweb through corallivorous and herbivorous fishes on coral reefs is much higher than terrestrial ecosystems (with the exception of highly productive grasslands) and raises the question of why such a high proportion of the primary productivity is consumed on coral reefs. The answer remains unclear but is likely related to the oligotrophic state of coral reefs. Under these conditions, efficient nutrient cycling and turnover of primary productivity is necessary to maintain the high consumer biomass characteristic of coral reefs. In terrestrial ecosystems, the majority (80–90%) of the biomass produced by plants is not consumed by herbivores. Rather, fallen leaves enter the foodweb through decomposers (Crawley 1983; Lowman 1992). Although primary consumers on coral reefs remove a much higher proportion of the productivity of coral and turf algal communities, it is unlikely that much of this consumption enters higher trophic levels directly. Predation rates upon adults of both polyp-feeding corallivores and larger herbivores (e.g. scarids, acanthuroids) are extremely low and these fish are rarely found in the stomachs of predators (e.g. Hiatt & Strasburg 1960; St John 1999; Mumby et al. 2006; but see Kingsford 1992). Furthermore, butterflyfishes are relatively long lived (10–14 years) (Berumen 2005; Zekeria et al. 2006), and much of the primary productivity consumed by these fishes will be used to meet the fishes’ growth and metabolic demands and will be effectively held static in this trophic level until they die through natural causes (e.g. disease, senescence, etc.). After this, the productivity consumed will be released through the detrital foodweb. Between 50–80% of the primary productivity on coral reefs is processed by detritivores (Hatcher 1983; Arias-Gonzalez et al. 1997). The importance of both the detrital pathway and small detritivores (e.g. gobies, blennies) in recycling and transferring primary productivity to higher consumers in the coral reef food chain has only recently been recognized (Wilson et al. 2003; Depczynski & Bellwood 2003; Wilson 2004; Depczynski et al. 2007). This similar dependence upon the detrital foodweb in both marine and terrestrial ecosystems indicates that these systems broadly function in similar ways. The main difference is the relatively large proportion of primary productivity that is consumed and incorporated into the standing biomass of corallivorous and herbivorous fishes before cycling through the detrital foodweb, whereas in terrestrial systems, the majority of productivity is consumed directly by decomposers (Crawley 1983).

It is likely that our study overestimates the proportion of primary productivity lost through corallivory; there are two main areas in our study that could contribute to this overestimation. First, increases in proportional cover represent growth in both coral tissue and skeleton, whereas the proportion of productivity consumed is only based on the tissue component. This assumption will result in productivity being underestimated as energy used to produce the carbonate skeleton of corals is not incorporated in our calculations. However, we do not expect this to be a major error in our study, as skeletal growth is energetically cheap and the growth of most branching corals is tissue dominated (Barnes & Chalker 1990; Anthony, Connolly & Willis 2002). For example, a model developed by Anthony, Connolly & Willis (2002) indicated that for every 100 J of energy used in growth only 2–4% of this is spent in calcification. Second, a more important source of error is that the annual growth rate estimates used in our study were calculated from healthy reefs which will have resident populations of corallivorous fishes, meaning that the estimated proportional increase of 17% is a net increase after some level of predation (Thompson & Dolman 2010). Consequently, our estimate of coral productivity is not a direct measure of net primary productivity (which is measured as gross primary productivity minus respiration); rather, we have used growth as a proxy for productivity. There are currently no estimates of coral growth rates on reefs without corallivorous fishes; likewise, aquarium-based estimates are lacking and their relevance to field-based estimates are questionable. To account for some of these limitations, we can recalculate the estimate of productivity consumed using a higher proportional increase per annum. If we double the increase in proportional cover of the tabular acroporids to 34% per annum, corallivorous butterflyfishes consume 31–44% of the productivity of these corals, which is still high when compared with similar-sized consumers in terrestrial ecosystems. When more accurate data on coral primary productivity become available, our estimates of productivity consumed by corallivorous butterflyfishes can be recalculated. However, as a first step, our study has demonstrated that corallivorous butterflyfish have the potential to consume a significant proportion of both the standing biomass and productivity of tabular acroporid corals.

The response of corals to the chronic grazing pressure exerted by small-bodied corallivores is currently unknown. Unlike the acute effects of the periodically abundant invertebrate corallivores such as Acanthaster planci or Druppella spp., which can dramatically reduce live coral cover over a relatively short period of time (Carpenter 1997), polyp-feeding fishes are a long term, chronic stressor on coral populations and predation only results in partial mortality of coral colonies. On healthy reef systems, this predation will most likely impact corals through sublethal effects, such as reduced growth and energy reserves, lowered fecundity and overall condition, rather than direct overgrazing and eventual death of a coral colony. The magnitude of these sublethal effects will depend upon the expense incurred by a coral from regenerating grazed tissue. Tissue regeneration is an energetically expensive process. For example, a 32% reduction in the growth rate of the coral Montastrea annularis occurred over a 2-month period following the creation of 1 cm2 lesions, and growth remained suppressed for a further 30 days after tissue regeneration had stopped (Meesters, Noordeloos & Bak 1994). Our study has demonstrated that the consumption side of the coral-corallivore relationship is considerably higher than previously thought. As such, we expect that this consumption will be a significant drain on energy reserves of tabular acroporid corals, although manipulative experiments are needed to determine whether the exclusion of coral-predators results in corresponding increases in growth, condition and reproductive output of these heavily consumed coral species.

Partial predation is a common feature of virtually all ecosystems, but it is in terrestrial systems that plant–herbivore interactions have been most extensively studied. These studies have created a general acceptance that herbivory can have major effects on plant communities, reducing growth rates, fitness and limiting the distribution and abundance of frequently consumed prey organisms (reviewed by Crawley 1989, 1997; Maron & Crone 2006). These studies have forced a rephrasing of the research question from ‘do herbivores have an effect?’ to ‘under what conditions do consumers have meaningful effects on plant dynamics?’ (Olff & Ritchie 1998; Maron & Crone 2006). In contrast, the equivalent relationship on coral reefs has been dismissed as largely unimportant to reef processes on the assumption that the amount of coral tissue consumed by corallivorous fishes is too low to have any meaningful effects on prey corals (e.g. Harmelin-Vivien & Bouchon-Navaro 1983; Hixon 1997). However, our study demonstrates that corallivorous fishes consume a major proportion of the standing biomass (9–13·5%) and potential productivity (52–71%) of tabular acroporid corals. Considering the large effects that can occur in plant communities from even relatively small annual defoliation rates (e.g. Marquis 1984, 1992; Crawley 1985), it is extremely likely that polyp-feeding corallivores will have similar effects to those seen in plants following herbivory. These grazing fishes are likely to limit the energy available for growth, reproduction and maintenance in potential prey organisms, such that grazing will have significant long-term consequences and may even reduce resilience of corals to other significant disturbances like climate-induced coral bleaching (e.g. Bellwood et al. 2006; Cole, Pratchett & Jones 2009; Gochfeld 2010).

Acknowledgements

This work was funded by the ARC Centre of Excellence for Coral Reef Studies. We thank the staff at Lizard Island Research Station for logistical support. This project was covered by the JCU Animal Ethics Review Committee no. A1306. We would also like to thank the three anonymous reviewers, whose comments greatly helped this work.

Ancillary