- Top of page
Environmental disturbances are an integral component of ecology that influence the dynamics and structure of both terrestrial and aquatic communities (Sousa 1984). In marine systems, disturbances often result in the loss of habitat-forming structures such as kelp (Dayton 1985; Steneck et al. 2002), seagrass (Duarte 2002), mangroves (Alongi 2002) and coral (Bruno & Selig 2007), which has an indirect effect on associated faunal communities. Understanding the impacts of disturbance-induced habitat changes on the distribution and abundance of species requires a comprehensive knowledge of species’ dependence on local habitat features. In general, those species that are habitat specialists are more likely to be governed by habitat availability than generalists that utilize a range of habitat types (Brown 1984). Dramatic changes in the availability of habitat should therefore have greater impact on the abundance of habitat specialists (Vazquez & Simberloff 2002), as observed in communities of birds (Julliard, Jiguet & Couvet 2004), mammals and amphibians (Swihart et al. 2003), butterflies (Cleary & Genner 2004) and coral reef fishes (Munday 2004). Consequently, habitat specialists are more susceptible to extinction (McKinney 1997), and increasing levels of habitat loss and fragmentation raise concerns about the future of many of these species (Tilman et al. 1994; Travis 2003).
For mobile animals, habitat preferences can also change with ontogeny, and thus the degree of habitat specialization may differ among life stages. While the adults of some species may be habitat generalists, juveniles of the same species may have quite specific habitat requirements, potentially creating a bottleneck for these populations (Halpern, Gaines & Warner 2005). However, the influence of juvenile specialization on adult stocks may vary depending on the relative availability of adult and juvenile habitats and on species longevity (Halpern, Gaines & Warner 2005). For example, specific habitat requirements of adults rather than juveniles regulate populations of stone crabs when adult sites are limited (Beck 1995). Thus specialization at different life-history stages can influence population size, and habitat preferences should be examined at a variety of life stages to assess the influence of specialization on species demographics.
Coral reefs provide a unique opportunity to explore habitat associations of fauna, as they are systems of extremely high biodiversity where species exhibit an array of habitat relationships. Reefs are also subject to high levels of both anthropogenic and natural disturbances, which can result in dramatic shifts in habitat type (Hughes 1994; Pandolfi et al. 2005). Of particular concern are predictions that climate change will increase the frequency of disturbances such as coral bleaching and storms (Hoegh-Guldberg 1999; Goldenberg et al. 2001; Webster et al. 2005) and that eutrophication (Birkeland 1982) and fishing (Dulvy, Freckleton & Polunin 2004) may encourage outbreaks of coral feeding seastar, Acanthaster planci L., also known as crown-of-thorns seastar (COTS). Major changes in reef habitats have already been documented, the most pronounced being loss of coral cover (Gardner et al. 2003; Bruno & Selig 2007), which affects reef-associated fish and invertebrate communities (Wilson et al. 2006; Pratchett et al., in press), and can lead to changes in community structure (Bellwood et al. 2006). The species most susceptible to coral loss are those that depend on live coral for food and shelter (Williams 1986; Halford et al. 2004). Previous studies have focused on specialization within one life-history stage of coral feeding butterflyfish or obligate coral-dwelling gobies (Munday 2004; Pratchett 2005). However, many fish undertake ontogenetic shifts in habitat use (Lecchini & Galzin 2005) and the extent of specialization may vary among life-history phases.
Here we examined habitat specialization and the effects of disturbance on one of the most abundant and ecologically diverse families of coral reef fishes, the pomacentrids, commonly known as the damselfish. Fish from this family display a wide array of habitat associations and diets (Allen 1991), and territorial behaviour by many species has a major influence on the structure of benthic reef communities (Ceccarelli, Jones & McCook 2001). This influence can extend over large spatial and temporal scales, as territories often cover extensive areas of reef (Ceccarelli, Jones & McCook 2005) and individuals may live for more than 20 years (Meekan, Ackerman & Wellington 2001). Most pomacentrids are also small-bodied and highly abundant, making them major prey for many reef predators (Hiatt & Strasburg 1960). Thus pomacentrids are an ecologically diverse and important family on coral reefs, and disturbance-induced changes to pomacentrid communities may subsequently affect the composition of benthic communities and reef trophodynamics.
We used a unique, long-term data set to test the hypothesis that habitat specialists are more susceptible to disturbance than generalists. Specialization was quantified based on the strength of pomacentrid associations with live coral and niche breadth of both adult and juveniles. The relationship between fish abundance and habitat availability was then assessed to gauge the dependence of the abundance of specialist species on specific microhabitat types. Finally, breadth of habitat use was used to examine the relationship between habitat specialization and the effects of COTS-instigated habitat degradation on fish populations. Disturbances caused by COTS have resulted in extensive loss of live coral on reefs throughout the Indo-Pacific, and are known to have a detrimental impact on elements of pomacentrid communities (Williams 1986; Sano, Shimizu & Nose 1987; Feary et al. 2007a). This study provides a comprehensive evaluation of reliance on coral within an ecologically important family of reef fish, and examines the implications of disturbance for persistence of reef fish.
- Top of page
Underwater visual surveys were used to assess microhabitat use of pomacentrids at 10 mid-shelf reefs within the northern section of the Great Barrier Reef (GBR) during December 2005 (Fig. 1). Within each reef a standard reef slope habitat was surveyed on the north-eastern flank, and within this habitat three sites were sampled, each containing five permanent 50-m transects lying approximately parallel to the reef crest. Transects were set along the slope at depths ranging between 6 and 9 m. All pomacentrids within transects were identified to species level, and habitat use for each fish was assessed by recording microhabitats directly beneath each fish. Microhabitats were categorized as: branching, plate, submassive, massive, encrusting, soft or dead coral, rubble or consolidated pavement. Microhabitat use by fish >1 m above the substratum was not recorded as it could not be assigned reliably for these individuals. The presence of these fish was noted and used to provide total counts of each species within transects, which were subsequently used to assess the relationship between abundance and microhabitat availability, although they were not included in selectivity calculations. These fish accounted for 30% of all pomacentrids observed.
Figure 1. Location of reefs surveyed. *, Reefs where habitat availability, pomacentrid abundance and habitat use were surveyed in December 2005 and used to assess habitat associations and niche breadth. #, Reefs where coral cover and pomacentrid abundance were collected between 1997 and 2003 and used to assess impact of Acanthaster planci-mediated coral decline on specialist species abundance.
Download figure to PowerPoint
Fish were identified as juveniles or adults based on body size and coloration. The number of individuals within fish groups sharing the same microhabitat was also recorded, as many pomacentrid species form large schools, and habitat use by juveniles can be influenced by the presence of conspecifics (Sweatman 1985; Booth 1992; Wellington 1992). Differences in adult and juvenile distribution patterns were examined using χ2 homogeneity tests.
Microhabitat cover on reefs was assessed using a video camera held ≈50 cm from the substratum to record substrata along each transect. Coverage of each microhabitat category was expressed as a percentage of total substrate area based on the substratum recorded under five points on each of 40 systematically selected frames from each transect.
Selectivity indices were used to determine if species used any of the microhabitats preferentially. For each species, selectivity indices (w) were calculated for each microhabitat using the equation:
where oiis the proportion of used microhabitat and πiis the proportion of microhabitat available for microhabitat type i (Manly, McDonald & Thomas 1993). Values of oiand πiwere calculated as the mean of the proportion of microhabitat used and available at the 10 reefs surveyed in 2005. Bonferroni-corrected confidence intervals encompassing selectivity indices were used to assess if fish used any microhabitat significantly more often than expected, based on its availability. Confidence intervals (CI) were calculated using the formula:
- CI = za/2[oi(1 – oi)/u+πi2]−2((eqn 2))
where u+ was the total number of microhabitats used by a pomacentrid species (Manly, McDonald & Thomas 1993). Fish were considered to be using a microhabitat significantly more often than expected if selectivity indices and associated confidence intervals were >1. Indices were calculated only for those species seen on >10 transects and more than three reefs, and were based on the number of fish groups rather than individuals occurring in each microhabitat, alleviating the influence of conspecifics on habitat use. Because niche breadth and selectivity indices for both adults and juveniles are based on the proportion (rather than total numbers) of fish of each type seen in each microhabitat type, comparisons between species, adults and juveniles were possible.
To determine if pomacentrid abundance is influenced by the availability of preferred microhabitats, fish abundance data collected from the 10 reefs during December 2005 were fitted to an exponential increase to maximum model:
- y = a[1 – exp(–b × x)] + c((eqn 3))
where x is the proportion of preferred habitat available on each transect and y is the number of fish aggregates present on that transect. In the model, a represents asymptotic population levels; b the rate of population change with increasing habitat availability; and c the abundance when habitat is absent. This model was chosen in preference to a linear relationship, as although habitat availability may directly affect coral-dependent species when coral cover is low, the effect of high coral cover is expected to be negligible (Holbrook, Brooks & Schmitt 2006). Parameters were constrained such that only a positive relationship between abundance and the cover of a preferred microhabitat was allowed.
The response of coral-reliant fish to declines in coral cover was assessed using data collected from seven reefs on the GBR where there was >50% decline in coral cover (Table 1; Fig. 1). Coral decline was primarily attributed to COTS outbreak (>1500 COTS km−2), although coral bleaching and storms are likely to have contributed to coral loss, particularly at Fitzroy Island and the Low Isles (Sweatman et al. 2003). For this analysis, coral cover was taken as the summed percentage coverage of those coral growth forms preferentially used by fish (branching, plate and submassive). Data were collected by the long-term monitoring team (Australian Institute of Marine Science), 2–8 years before habitat and fish abundance data were collected for assessment of microhabitat use and niche breadth of pomacentrids. The sampling protocol and design for collection of both data sets were, however, identical (for detailed description of sampling protocols and species lists see Halford & Thompson 1996; Abdo et al. 2003).
Table 1. Severe decline in mean coral cover at seven GBR reefs
|Year||Percentage coral||Year||Percentage coral|
|John Brewer||2001||16·1 (0·9)||2003||0·8 (0·4)||95|
|Low Isles||1997||14·3 (1·0)||2000||1·5 (0·5)||89|
|Thetford||1999||22·8 (1·6)||2002||2·4 (0·7)||89|
|Rib||1999||37·5 (1·7)||2001||5·5 (1·0)||85|
|Fitzroy Island||1999||14·6 (1·1)||2000||3·0 (0·9)||79|
|Gannet Cay||1997||28·3 (1·6)||1998||12·4 (1·8)||56|
|Horseshoe||1997||26·0 (1·4)||1999||11·9 (1·5)||54|
The response to coral decline by pomacentrid species was calculated as the percentage change in abundance of fish divided by the percentage change in coral cover at each of the seven reefs. This accounted for differences in the abundance of various fish species and percentage coral cover between sites. Large positive values indicated declining fish abundance, while negative values were indicative of increasing abundance.
The niche breadth was calculated for each pomacentrid species using the proportional similarity index (Feinsinger, Spears & Poole 1981), which considers both proportional use and availability of resources. This metric ranges between 0 and 1; lower values indicate smaller niche breadths and greater habitat specialization.
To determine if there was a relationship between habitat specialization and disturbance, niche breadth of the six coral-dwelling pomacentrid species was plotted against changes in fish abundance both before and after COTS-induced coral decline. Comparisons made before coral decline were used to assess natural variation in pomacentrid abundance. Calculations for before impact assessment were based on comparing data collected 3 years before disturbance with data 1 year before disturbance at each of the impacted reefs. Regression was used to assess the significance of the relationship between change in species abundance and niche breadth. Fish abundance data were log(x + 1)-transformed prior to analysis to meet the assumptions of homogeneity of variance and normality, examined using residual plots.
Microhabitat selectivity indices of all species selected for this analysis had strong affinities with live coral as adults. However, two species with strong affinity for live coral, Plectroglyphidodon dickii Liénard and Dascyllus reticulatus Richardson, were excluded from the analysis as they were of low abundance before COTS outbreaks and occurred on fewer than two of the affected reefs.
- Top of page
Resource specialization confers an advantage over generalists within a subset of resources; however, generalists are expected to outperform specialists over a wider range of resources (Caley & Munday 2003). In particular, generalist species should be more resilient to changing resource availability (Vazquez & Simberloff 2002). Consequently, dietary and habitat specialization are strong predictors of response to disturbance by animals in both terrestrial and aquatic systems (Swihart et al. 2003; Cleary & Genner 2004; Julliard, Jiguet & Couvet 2004; Munday 2004). However the strength of the specialization–disturbance relationship may vary between communities, even when within the same system and subjected to the same disturbance (Taki & Kevan 2007).
We found varying levels of habitat specialization and response to disturbances within a prominent and ecologically important family of reef fish, the Pomacentridae. Some species were strongly associated with coral as both adults and juveniles, while others used a range of microhabitats, many of which were non-coral. Overall, coral dependency within the pomacentrids is high compared with other coral reef fishes, and ≈20% of all pomacentrid species on the GBR rely on coral for food or shelter (Munday et al. 2007). This value is lower than our estimate of 40%; however, meta-analysis of studies documenting fish responses to coral loss indicate that the percentage of species declining in abundance exceeds that of known coral dependents (Wilson et al. 2006), indicating that other species are somehow reliant on coral.
Reduced abundance of adult fish not known to associate with coral on perturbed reefs may be partially attributed to juveniles’ preference for coral habitats (Jones et al. 2004). We found that a higher percentage of species and individuals relied on coral microhabitat as juveniles compared with adults, including some not known to associate with hard coral as adults. These findings are consistent with other studies showing strong juvenile associations with live coral in fish with adult conspecifics that do not favour coral habitats (Booth & Beretta 1994; Gutiérrez 1998; Feary et al. 2007b). Loss of live coral and increased algal cover can also result in an overall reduced abundance of fish recruits and a shift to juvenile communities dominated by algae-associated species (Feary et al. 2007a, 2007b). Extensive and protracted declines in coral cover may therefore result in changes to adult communities, whereby fish that associate with corals as juveniles decline in abundance, irrespective of their habitat associations as adults.
All coral-dwelling adults displayed a preference for live coral as juveniles. This is partially due to juveniles recruiting to sites occupied by conspecifics (Sweatman 1985; Booth 1992). However, juveniles had a narrower niche breadth than adults and a stronger association with plating corals, suggesting preference for a subset of corals used by adults. Juvenile fish may be better suited to coral plate refuges, as fish tend to choose shelters that match their body size (Hixon & Beets 1993; Friedlander & Parrish 1998), and spaces between branches of plating corals provide better shelter for small juveniles than do larger branching colonies. Importantly, narrow niche breadth of juveniles relative to adults suggests that they are more specialized and therefore more susceptible to disturbance.
Species with ontogenetic shifts in coral preferences may be particularly susceptible to coral loss, as declines in either habitat type will ultimately influence adult populations. Density dependence during early life history regulates demographics at later life stages of some amphibians (Altweg 2003), and availability of juvenile habitat can influence adult populations of reef fishes (Halpern 2004; Mumby et al. 2004). Models predict that adult abundance will be limited by juvenile habitat when adult habitat size is greater than that of juveniles (Halpern, Gaines & Warner 2005), which is likely for coral-dependent pomacentrid species that have a narrower niche breadth as juveniles.
For pomacentrids, availability of preferred coral microhabitat was, however, a poor predictor of abundance, particularly that of juveniles, suggesting that factors other than microhabitat are more important in determining juvenile abundance. Caselle & Warner (1996) found that microhabitat failed to predict recruitment patterns of coral reef fish at different sites, concluding that physical oceanic processes were more important in determining recruitment patterns at large spatial scales. Microhabitat could, however, predict recruit density at the smaller spatial scale of transects (Caselle & Warner 1996). Poor relationships between the abundance of specialists and their preferred habitat probably reflect the balancing of pre- and postrecruitment processes in driving the recruitment patterns of these fish. At small scales, postrecruitment processes such as habitat selection, predation and competition, and the interaction of these processes with habitat complexity, are dominant (Almany 2004), whereas the abundance of juveniles at larger scales are probably driven by supply rates from the plankton. Thus small-scale dependence of specialist species on live coral suggests that comprehensive coral loss is still expected to have serious consequences for coral-dependent juveniles and ultimately adult abundance of these species.
Although microhabitat availability was a poor predictor of fish abundance at larger scales, niche breadth of microhabitat use provided estimates of habitat specialization that are compatible with the specialization-disturbance hypothesis (Vazquez & Simberloff 2002). As predicted, versatility in resource use improved resilience of coral-dependent species to COTS-mediated coral declines, with more generalist species showing reduced declines in abundance relative to specialists. However, the response of generalist coral-dwelling species was variable, and the abundance of these species occasionally increased. Munday (2004) found that, although the extent of population declines in coral-dwelling Gobiodon was related to habitat specialization, abundance of all coral-dwelling species declined following coral loss. Similarly, Pratchett, Wilson & Baird (2006) found that when coral loss was severe, all obligate coral-feeding butterflyfish declined in abundance irrespective of diet breadth. Thus the extent of habitat disturbance may sometimes outweigh differential impacts on habitat or diet specialists.
Habitat specialists are likely to take longer than generalists to recover from disturbances, because the continued absence of these species allows invasion by habitat generalists (Marvier, Kareiva & Neubert 2004), as observed among butterfly communities following extensive forest fires (Charrette, Cleary & Mooers 2006). On coral reefs, this may translate to fish communities dominated by generalist species, which are not reliant on live coral at any stage in their life history, and a higher extinction risk for coral specialists. This prediction is supported by local extinctions of coral specialists (Munday 2004; Graham et al. 2006) and an increased proportion of habitat generalists (Bellwood et al. 2006) following disturbances.
Our study supports the hypothesis that habitat specialists are at greater risk due to disturbance than are generalists. Analysis of changes to adult fish abundance following coral decline found that highly specialized coral-associated pomacentrids consistently declined in abundance, while the response from generalist habitat-users was more variable and may relate to the severity of the disturbance. A high proportion of pomacentrids are closely associated with live coral, although the type of coral they associate with varies among species and often changes ontogenetically. Importantly, associations with live coral were especially high among juveniles, suggesting that this is a life-history phase more vulnerable to coral loss, which may have serious consequences for future adult stocks.