Does habitat availability determine geographical-scale abundances of coral-dwelling fishes?

Transcription

1 Coral Reefs (2002) 21: DOI /s y REPORT P.L. Munday Does habitat availability determine geographical-scale abundances of coral-dwelling fishes? Received: 22 March 2001 / Accepted: 30 July 2001 / Published online: 2 February 2002 Ó Springer-Verlag 2002 Abstract The role of local-scale processes in determining large-scale patterns of abundance is a key issue in ecology. To test whether habitat use determines local and large-scale patterns of abundance of obligate coraldwelling fishes (genus Gobiodon), the author compared habitat availability with the abundance of four species, G. axillaris, G. brochus, G. histrio, and G. quinquestrigatus, among four locations, from the southern Great Barrier Reef to northern Papua New Guinea. Habitat availability, measured at tens of meters, explained 47 65% of the variation in abundance of these species among geographic locations spanning over 2,000 km. Therefore, local-scale patterns of habitat use appear to determine much larger-scale patterns of abundance in these habitat-specialist fish. The abundances of all species, except G. brochus, were also closely associated with particular exposure regimes, independently of the abundance of corals. Broad-scale habitat selection for reef types within locations can most easily explain this pattern. The abundances of all species, except G. brochus, also varied among geographic locations, independently of coral abundances. Therefore, the abundances of these species are influenced by either geographic variation in local-scale processes that was not measured, or additional processes acting at very large spatial scales. Keywords Habitat use Æ Specialisation Æ Local versus regional processes Æ Spatial scale Æ Coral-reef fish Æ Gobiidae P.L. Munday Centre for Coral Reef Biodiversity and School of Marine Biology and Aquaculture, James Cook University, Townsville, Queensland 4811, Australia philip.munday@jcu.edu.au Tel.: Fax: Introduction Both local- and large-scale processes have been emphasised in attempts to explain similarities and differences among widely separated assemblages of animals (Ricklefs 1987; Menge and Olson 1990; Ricklefs and Schluter 1993; Caley 1995a; Hubbell 1997). For example, efforts to understand regional variation in the structure of animal assemblages have sought to integrate ecological processes that operate on a local scale (e.g., habitat selection, competition, predation) with processes that can act on much larger scales (e.g., larval supply) (Caley 1995b; Alexander and Roughgarden 1996; Connolly and Roughgarden 1998). However, difficulties in conducting properly replicated experiments at large spatial scales have limited progress in determining the processes underlying patterns of abundance among geographic locations. Consequently, the relative importance of local-scale processes in determining patterns of abundance at geographic spatial scales (thousands of kilometers) remains largely unanswered. Availability of suitable habitat is a widely recognised factor underlying the distribution and abundance of many species (Pulliam and Danielson 1991; Rosenzweig 1991; Verner and Fahrig 1996). However, the effect of habitat use on patterns of abundance is likely to vary with spatial scale (Morris 1992; MacNally 1995). As spatial scale increases, factors such as dispersal characteristics are expected to have increasing importance (Palmer et al. 1996). For species with a dispersive larval phase, such as many marine organisms, processes causing variation in larval supply and recruitment can have major effects on patterns of abundance at large spatial scales (Butman 1987; Cowen and Castro 1994; Doherty and Fowler 1994; Caselle and Warner 1996; Connolly and Roughgarden 1998; Connolly et al. 2001). The degree of habitat specialisation exhibited by a species is also predicted to have scale-related implications for patterns of abundance (Fox and Morrow 1981; Brown 1984; Gaston and Lawton 1990). The abundances of

2 106 species with very specialised patterns of habitat use are more likely to be closely linked to the availability of preferred habitats, at both small and large spatial scales, than are the abundances of species with broad patterns of habitat use. Comparisons of population abundance and habitat correlates, sampled in the same manner at replicate sites within different locations, are useful for interpreting the importance of habitat availability to population dynamics at local and larger spatial scales (Caselle and Warner 1996; Tolimieri 1998). Multiscale analyses of variation in abundance can also provide important insights into the spatial scales where other important ecological processes operate (Underwood and Petraitis 1993; Underwood and Chapman 1996; Hughes et al. 1999; Maurer 1999). In this study, I use a multiscale approach to examine the role of habitat use and other processes in structuring patterns of abundance of four species of coral-reef fish among geographic locations spanning several thousand kilometers. I define the maximum spatial extent of this comparison as a geographical spatial scale. Habitat associations in coral-reef fish Microhabitat associations have been widely reported among coral-reef fishes (Jones 1991; Williams 1991), and the influence of these associations on the distribution of reef fishes at local spatial scales (i.e., tens to hundreds of meters) has been demonstrated by empirical observations (e.g., Williams 1980; Sale et al. 1984; Shulman 1984; Booth 1992; Munday et al. 1997) and experimental manipulations (e.g., Sweatman 1983; Fautin 1992; Wellington 1992; Tolimieri 1995; Gutie rrez 1998). Local-scale abundances of coral-reef fishes have been correlated with coral cover (e.g., Bell and Galzin 1984; Bouchon-Navaro and Bouchon 1989; Jennings et al. 1996; Munday et al. 1997; Munday 2000), availability of shelter holes (Roberts and Ormond 1987; Buchheim and Hixon 1992; Hixon and Beets 1993; Clarke 1996), structural complexity (Luckhurst and Luckhurst 1978; Friedlander and Parish 1998), and microhabitat heterogeneity (Kaufman and Ebersole 1984). However, the effects of habitat availability on the distribution and abundance of coral-reef fishes at larger spatial scales are very poorly understood (Fowler et al. 1992). There is some indication that habitat associations become less important in determining patterns of abundance with increasing spatial scale (Tolimieri 1995; Caselle and Warner 1996), but at least one study (Holbrook et al. 2000) found a strong correlation between the abundance of a coral-reef fish and the availability of suitable habitats at large spatial scales. Obligate coral-dwelling gobies from the genus Gobiodon are some of the most habitat-specialised fishes on coral reefs and their local-scale abundances are influenced by habitat availability (Munday et al. 1997, 2001; Munday 2000). The abundances of these species also appear to be influenced by physical characteristics of the reef location, such as exposure to prevailing winds (Munday et al. 1997; Munday 2000). Therefore, the distribution and abundance of coral-dwelling gobies among widespread locations might be influenced by a combination of factors including habitat availability, exposure regime, and processes that act on very large spatial scales. To examine the relationship between localscale habitat availability and geographical-scale patterns of abundance of coral-dwelling gobies, I compared habitat availability with the abundances of these fishes at four locations separated by approximately 2,000 km, from the southern Great Barrier Reef to northern Papua New Guinea. To test whether processes other than habitat use affect patterns of abundance, I removed the variation in abundance that could be explained by habitat availability and then examined residual variation in abundance within and among locations. I used a hierarchical sampling design to identify the spatial scale where processes other than habitat availability were acting and to help identify what these processes were likely to be. Because changes in habitat associations might influence the relationship between habitat availability and patterns of abundance among locations, I also determined habitat associations for each species of fish at each location and then tested whether changes in habitat associations might have influenced the sensitivity of analyses at the largest spatial scale. Methods Study species and locations Gobiodon are small (<60 mm total length), obligate, coraldwelling fishes (family Gobiidae) that live among the branches of corals mostly of the genus Acropora (Munday et al. 1997, 1999). These fish are highly sedentary and rely on their host coral colony for protection from predation and the provision of breeding sites. Gobiodon are almost entirely confined to coral reefs as a result of their obligate association with living corals. They can also be found on isolated patches of coral habitat that is not associated with the principal reef matrix, such as isolated coral colonies within reef lagoons. Lassig (1981) suggested that coral gobies feed on coral tissue; however, Harold and Winterbottom (1999) found only copepods, foraminifera, and unidentified material in the guts of G. brochus, and coral nematocysts were found in the guts of only one of five species examined at Lizard Island on the Great Barrier Reef (Munday, unpublished data). Therefore, corallivory is either facultative or restricted to just a few species of Gobiodon. Monitoring of tagged fish at Lizard Island has shown that at least one species, G. histrio, can live for over 4 years (Munday, unpublished data). Thirteen species of Gobiodon are currently recognised from the Great Barrier Reef and Papua New Guinea (Munday et al. 1999) and most species occur throughout this range. Four species of Gobiodon that are abundant in Papua New Guinea and the Great Barrier Reef are considered in this study: G. axillaris, G. brochus, G. histrio, and G. quinquestrigatus (Munday et al. 1999). Abundances of these four Gobiodon species and the corals they inhabit were estimated at four widely separated geographic locations: (1) One Tree Island (23 30 S; E) on the southern Great Barrier Reef, (2) Lizard Island (14 40 S; E) on the northern Great Barrier Reef, (3) Bootless Bay (9 31 S; E) in southern Papua New Guinea, and (4) Kimbe Bay (5 15 S;

3 E) in northern Papua New Guinea (Fig. 1). Due to their geographic separation it was not practical to sample locations concurrently. Patterns of habitat use and abundance were estimated during March 1995 (Lizard Island), September 1996 and April 1997 (Kimbe Bay), April 1997 (Bootless Bay), and October 1997 (One Tree Island). The temporal variation in the sampling regime is not expected to substantially affect the spatial patterns of Gobiodon abundance observed within and among locations because: (1) these fish are relatively long-lived (see above) compared to the temporal variation in sampling among locations; (2) habitat associations of these four species of Gobiodon are very conservative within locations (Munday et al. 1997; Munday 2000) and among seasons (personal observations); (3) juveniles have similar habitat associations to adults at Lizard Island (Munday et al. 1997) and elsewhere (personal observations), therefore, sampling period cannot be confounded with ontogenetic habitat shifts; (4) most coral colonies of the types inhabited by these species of Gobiodon are occupied, i.e., there are relatively few vacant corals that may be inhabited in some years but not others (Munday et al. 1997, 2001; Munday 2000); and (5) sampling at Lizard Island was conducted prior to an Acanthaster planci outbreak that occurred in this area during One Tree and Lizard Islands are located in the outer half of the Great Barrier Reef lagoon, which extends from the coast to the 200-m-depth isobath along the north coast of Queensland. One Tree Island has a well-developed, contiguous reef enclosing a large ponding lagoon. Lizard Island is a continental island with welldeveloped fringing reef, extending to form a large lagoon on the southern side. Sampling was not conducted within the One Tree or Lizard Island lagoons. Bootless Bay is located in the Papuan Barrier Reef lagoon, which extends from the coast to the 200-mdepth isobath along the south-east coastline of Papua New Guinea. Three small continental islands (Loloata, Motupore and Lion Islands) occur in the Bootless Bay area and each is surrounded by well-developed fringing reef. Numerous patch reefs (tens to hundreds of meters in diameter) occur between these continental islands and the Papuan Barrier Reef. Kimbe Bay is not enclosed by a barrier reef, but is protected from prevailing winds by the mainland of New Britain and the Willaumez Peninsula. A dense network of small patch reefs (tens to hundreds of meters in diameter) and several small continental islands surrounded by well-developed fringing reef occur between the coast and the 200-m isobath. One small continental island (Kimbe Island) and a number of pinnacle and larger patch reefs (hundreds of meters across) occur beyond the 200-m isobath. The reefs at all locations can be clearly split into a zone comprising the reef flat and a zone comprising the reef crest and slope. Sampling was conducted on the reef flat because this was the most readily distinguished and uniform reef zone at all locations. Detailed descriptions of the morphology and structure of reefs at each location are given in Weber (1973), Heatwole (1981), Pichon and Morrissey (1981), Meekan et al. (1995), and Holthus and Maragos (1996). Coral cover is typically between 20 and 40% in the Lizard Island and One Tree Island sections of the Great Barrier Reef and above 40% in Papua New Guinea (Maniwavie et al. 2000). The number of species of Acropora, which are the primary habitat of Gobiodon species, increases from 36 in the vicinity of One Tree Island, 49 at Lizard Island, 52 in the vicinity of Motupore Island, to 72 species in Kimbe Bay (Veron 1993; Wallace and Wolstenholme, personal communication), providing a gradient of habitat diversity which might influence patterns of habitat use and abundance of coral-dwelling gobies. Furthermore, there are changes in the relative abundances of species of corals across this geographic gradient, which might influence the abundances of coral-dwelling gobies. Exposure to prevailing winds appears to influence abundances of Gobiodon among reefs within a location (Munday et al. 1997; Munday 2000), either by direct effects on patterns of water movement, or indirectly through differences in the coral assemblage or physical structure of corals at sites with different exposures to prevailing winds. Therefore, the reefs at each geographic location were separated into exposed, moderate, and sheltered exposures. Exposed reefs were classified as those reefs exposed to the prevailing winds and where waves break on the reef crest and flat for most of the year. Sheltered reefs were classified as reefs mostly protected from prevailing winds and where waves rarely break on the reef crest or flat. Moderate reefs were intermediate in exposure. At each location, the south-easterly trade wind is the prevailing wind during the austral winter but tends more northerly during the austral summer. Three representative sites were selected within each of these exposure regimes at each location (i.e., 36 sites in total) (Fig. 1). Although there is undoubtedly considerable variation among locations in features, such as the gross structure of reefs and the duration of prevailing winds, reefs within the three exposure regimes appeared to be broadly similar among locations. Estimates of abundance The species of coral to be censused for Gobiodon were determined during a preliminary investigation of habitat use by Gobiodon at each location, where all species of Acropora were carefully searched for the presence of Gobiodon. In this preliminary investigation, ten species of coral, Acropora cerealis, A. digitifera, A. humilis, A. loripes, A. millepora, A. nasuta, A. secale, A. selago, A. tenuis, and A. valida were found to be inhabited by the four species of Gobiodon considered here. To examine relationships between the abundances of Gobiodon and the abundances of the coral species they inhabit, I censused Gobiodon in all colonies of these ten species of Acropora, within random transects on the reef flat at each site. At each site, five replicate 10 1-m belt transects were used on the outer reef flat, running roughly parallel to the reef crest. A 10-m tape was placed along the center of the transect and a 1-m plastic bar was used to measure the transect width. Each coral colony of Acropora species used by Gobiodon, located at least half within the transect and with a diameter greater than 5 cm, was carefully searched for Gobiodon with the aid of a small underwater light. Gobies always remained within coral colonies during the census. For each transect I recorded the total number of coral colonies of each species of Acropora and the number of gobies within each coral colony. The species of Gobiodon considered here usually occur singly or in pairs and coral colony size does not appear to control the upper boundary of social group size (Munday et al. 1998). Although the total number of gobies (all species) per coral may be influenced by coral colony size to some degree (Munday et al. 1998), the number of coral colonies per transect has been found to be a reasonable measure of habitat availability for Gobiodon in previous studies (Munday et al. 1997; Munday 2000) and was used here for simplicity. Analysis of variation in abundance I used multiple regression, followed by ANOVA of regression residuals, to investigate the relationships between habitat availability and the abundances of Gobiodon among and within geographic locations. If variation in abundance of each species of Gobiodon within and among geographic locations is mostly determined by habitat availability, then patterns of abundance should be closely correlated with the abundance of the coral colonies inhabited. I used best-subsets multiple regression to determine the percent of variation in abundance of each species of Gobiodon that could be attributed to variation in the abundance of corals. Best subsets of the ten species of Acropora were determined by maximum adjusted R 2. Because I was only interested in positive associations between the abundance of each species of goby and the species of coral they inhabit, I excluded any variables with negative regression coefficients when calculating the adjusted R 2. If processes acting on large spatial scales also influence, or are correlated with, the abundance of each species of Gobiodon, then there should be significant additional variation in abundance among exposure regimes or among geographic locations that is not explained by the multiple regression. ANOVA on the

4 108 residuals from the multiple regression analysis was used to determine whether the abundances of each species of Gobiodon were associated with particular sites, exposure regimes, or geographic locations independently of the availability of corals. The ANOVA model included geographic location and exposures as fixed factors and sites as a random factor nested within

5 b Fig. 1. Study sites: at Kimbe Bay; exposed sites: KI, OR, MR; moderate sites: SR, VR, DR; sheltered sites: RR, NN, CR; at Bootless Bay; exposed sites: EB, HR, SP; moderate sites: BR, LS, MS; sheltered sites: LN, MN, LIN; at Lizard Island; exposed sites: LH, BI, SR; moderate sites: WM, NR, MC; sheltered sites: OI, VR, HR; and at One Tree Island; exposed sites: ES, EM, EN; moderate sites: ME, MM, MW; sheltered sites: SN, SM, SS geographic locations and exposures. Residuals were from the regression model that contained only positive regression coefficients. Regression and ANOVA analyses were conducted using Statistica 99 for Windows. Habitat associations Because changes in habitat associations among locations might weaken relationships between habitat availability and abundances of coral-dwelling gobies among locations, I examined habitat associations of each species of fish at each location. Resource selection ratios (Manly et al. 1993) were used to determine whether species of Gobiodon associated more frequently than expected with any of the ten species of Acropora censused at each location, and whether these associations changed among geographic locations. Selection ratios (w i ) were estimated using the formula w i ¼ o i =a i, where o i is the proportion of the total coral colonies occupied by a species of Gobiodon for coral species i, and a i is the proportion of total available coral colonies for coral species i. To ensure independence of habitat use observations, only the presence or absence of each Gobiodon species per colony was used in this analysis. Data at the transect level were pooled for each location to estimate patterns of habitat use at each location. To determine whether corals were used in proportion to their availability, used more frequently than expected from their availability, or used less frequently than expected, a Bonferroni-corrected 95% confidence interval was estimated for each selection ratio using the formula: Z a=2k qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi fo i ð1 o i Þ= ðu þ a 2 i Þg where z a=2k is the critical value of the standard normal distribution corresponding to an upper tail area of a/2k, a=0.05, k is the number of coral species, o i is the proportion of the total coral colonies occupied by a species of Gobiodon for coral species i, u + is the total number of coral colonies used, and a i is the proportion of total available coral colonies for coral species i. In order to conform with the spatial scale at which habitat associations were calculated, confidence intervals were calculated using transect data pooled at the level of locations. A coral species was considered to be used in proportion to its availability when the 95% confidence interval of its selection ratio encompassed 1 and more frequently than expected when the 95% confidence interval of its selection ratio exceeded 1 (Manly et al. 1993). To determine whether any changes in habitat associations among locations tended to weaken the relationship between habitat availability and patterns of abundance detected in the among location regression analysis (above), I conducted separate regression analyses for each location and then compared the mean adjusted R 2 for these locations to the adjusted R 2 in the among location regression analysis. If changes in habitat use had significant effects on the ability of the among location analysis to detect associations between habitat availability and patterns of abundance then I expected that the mean R 2 from the separate analyses would be higher than the R 2 in the among location analysis. Best-subsets regression was used as described above. Results Variation in abundance 109 The abundance of each species of Gobiodon varied among geographic locations and exposures (Fig. 2). Between 47 and 65% of this variation in abundance was ð1þ Fig. 2. Mean number of individuals per transect (±SE), for each species of Gobiodon, at each geographic location and exposure level. N Total number of individuals observed; south One Tree Island; north Lizard Island; south Bootless Bay; north Kimbe Bay

6 110 explained by the abundance of the ten species of corals (Table 1). For G. axillaris and G. quinquestrigatus, 34 and 47%, respectively, of the variation in abundance was explained by the abundance of four species of Acropora. F or G. brochus, 59% of the variation in abundance was explained by the abundance of only three species of Acropora. F or G. histrio, 62% of the variation in abundance was explained by the abundance of five species of Acropora (Table 1). The abundances of G. axillaris, G. histrio, and G. quinquestrigatus varied significantly among exposure regimes independently of the availability of corals (Table 2). G. axillaris was most abundant on exposed reefs and least abundant on sheltered reefs at each location (Fig. 3). In contrast, G. histrio and G. quinquestrigatus were more abundant on sheltered reefs compared to exposed and moderate reefs at each location (Fig. 3). The abundance of G. brochus did not differ among exposure regimes independently of habitat availability. The abundances of G. axillaris, G. histrio, and G. quinquestrigatus also varied significantly among geographic locations independently of the availability of corals (Table 2). G. axillaris was generally more abundant on the Great Barrier Reef () compared to Papua New Guinea (). In contrast, G. histrio was more abundant in compared to the (Fig. 4). G. quinquestrigatus was least abundant on the northern and most abundant in northern (Fig. 4). The abundance of G. brochus did not differ among geographic locations independently of habitat availability. G. axillaris was the only species for which abundance varied among sites independently of the availability of corals (Table 2). Habitat associations Gobiodon brochus and G. histrio exhibited very conservative patterns of habitat use among locations (Table 3). G. brochus used A. loripes and A. tenuis more frequently than expected at all locations. G. histrio used A. nasuta more frequently than expected at all locations, but also used a number of other species in accordance with their availability. In contrast, patterns of habitat use by G. axillaris and G. quinquestrigatus varied among locations. G. axillaris inhabited A. nasuta more frequently than expected at most locations but exhibited an increasing use of A. digitifera and a decreasing use of A. valida from the southern to northern (Table 3). G. quinquestrigatus was the most generalist species, using between five and eight species of corals in Table 1. Multiple regression analyses showing proportion of variation in abundance of four species of Gobiodon explained by ten species of Acropora (Full model) and the best subset of these ten species (Best subset). The best subset model includes only variables with positive regression coefficients Species Adjusted R 2 Variables and regression coefficients Full model Best subset G. axillaris A. digitifera 0.22, A. humilis 0.12, A. nasuta 0.44, A. valida 0.18 G. brochus A. loripes 0.61, A. millepora 0.11, A. tenuis 0.34 G. histrio A. cerealis 0.26, A. loripes 0.15, A. millepora 0.40, A. nasuta 0.29, A. tenuis 0.14 G. quinquestrigatus A. cerealis 0.23, A. millepora 0.33, A. nasuta 0.36, A. tenuis 0.12 Table 2. ANOVA on residual abundance from best-subsets multiple regression for each species of Gobiodon Species Source df MS F P G. axillaris Location Exposure Site (location, exposure) Location exposure Residual G. brochus Location Exposure Site (location, exposure) Location exposure Residual G. histrio Location Exposure Site (location, exposure) Location exposure Residual G. quinquestrigatus Location Exposure Site (location, exposure) Location exposure Residual

7 111 Fig. 3. Mean residual abundance per transect (±SE) across exposure regimes for each species of Gobiodon after removing the effects of habitat availability by multiple regression. Sample size as per Fig. 2 Fig. 4. Mean residual abundance per transect (±SE) among geographic locations for each species of Gobiodon after removing the effects of habitat availability by multiple regression. Sample size as per Fig. 2 proportion to their availability or more frequently than expected at each location (Table 3). Changes in habitat use among locations appeared to have little effect on the ability of the among location regression to detect associations between habitat availability and the abundances of each species of fish. For all four species of Gobiodon the mean R 2 calculated using separate analyses for each location (Table 4) was similar to the best subset R 2 in the among locations analysis (Table 1). For G. axillaris and G. quinquestrigatus, which exhibited changes in habitat use among locations, the mean R 2 from separate analyses was slightly higher than the among location R 2 for one species (G. axillaris: 0.39 vs 0.34) and slightly lower than the among location R 2 for the other species (G. quinquestrigatus: 0.40 vs 0.47). The similarity of the R 2 from the two methods of analysis indicates that any changes in patterns of habitat use among locations did not seriously bias the results of the among location regression. The abundances of Acropora species explained relatively little variation in abundance of G. axillaris in south (Table 4: R 2 =0.14) and G. quinquestrigatus in north (Table 4: R 2 =0.14). The weak relationship between habitat availability and patterns of abundance for these species correlates with low abundance at these locations (Fig. 1).

8 112 Species Location Coral species Table 3. Significance of habitat (coral species) use by Gobiodon species at four geographic locations. NS Habitat used in proportion to availability; + habitat used significantly more than expected; - habitat used significantly less than expected. Note: A. loripes and G. brochus were not found in north and A. selago was not found in south and north A. cerealis A. digitifera A. humilis A. millepora A. nasuta A. secale A. tenuis A. valida A. loripes A. selago G. axillaris South South G. brochus South South G. histrio South South G. quinquestrigatus South South NS NS + + NS NS NS + + NS NS + NS NS + NS + NS NS NS + NS NS NS NS + NS NS NS + NS NS NS NS + + NS NS NS NS + NS NS NS NS + NS + NS NS NS + NS NS NS + + Table 4. Best-subsets multiple regression analyses showing proportion of variation in abundance of four Gobiodon species that is explained by the abundance of Acropora species (variables) at four different locations. The best-subset models include only variables with positive regression coefficients. Mean adjusted R 2 is the average from all locations for each species of fish. Note: G. brochus was not found in north Species Mean R 2 South South Adj. R 2 Variables Adj. R 2 Variables Adj. R 2 Variables Adj. R 2 Variables G. axillaris A. digitifera, A. humilis, A. nasuta, A. valida G. brochus A. loripes, A. tenuis G. histrio A. loripes, A. nasuta, A. tenuis G. quinquestrigatus A. loripes, A. nasuta 0.34 A. millepora, A. nasuta 0.71 A. loripes, A. nasuta, A. tenuis 0.37 A. cerealis, A. loripes, A. millepora, A. nasuta, A. tenuis 0.14 A. nasuta 0.39 A. digitifera 0.59 A. humilis, A. loripes, A. millepora, A. tenuis, A. valida 0.69 A. cerealis, A. humilis, A. loripes, A. millepora, A. nasuta 0.04 A. nasuta 0.66 A. humilis, A. millepora, A. nasuta, A. tenuis, A. valida 0.67 A. digitifera, A. humilis, A. millepora, A. nasuta 0.58 A. cerealis, A. humilis, A. nasuta, A. tenuis, A. valida

9 113 Discussion Comparisons of data collected in the same way at replicate sites and locations can provide insights into the processes that influence patterns of distribution and abundance among widespread locations, and can focus attention on the spatial scales where these processes operate (Ricklefs and Schluter 1993; Caselle and Warner 1996; Underwood and Chapman 1996; Hughes et al. 1999). The spatial scale over which habitat availability determines the distribution and abundance of coral-reef fishes has remained a widely debated question (Jones 1991; Williams 1991). The local-scale abundances of some coral-reef fishes have been closely linked to the availability of suitable habitat (e.g., Kuwamura et al. 1994; Clarke 1996; Munday et al. 1997; Schmitt and Holbrook 1999; Munday 2000); however, some previous studies have suggested that the effect of habitat availability on the abundances of reef fishes declines with increasing spatial scale (Tolimieri 1995; Caselle and Warner 1996). In this study, habitat availability explained approximately half the variation in the abundances of coral-dwelling fishes among locations separated by thousands of kilometers. In other words, patterns of habitat use at tens of meters appear to play a major role in determining regional-scale patterns of abundance of these habitat-specialist fishes. The abundance of another coral-dwelling fish, Dascyllus aruanus, has also been found to be closely correlated with the availability of live corals at large spatial scales (Holbrook et al. 2000). In conjunction, these two studies indicate that habitat availability can be a major factor determining the distribution and abundance of coraldwelling fishes at geographical spatial scales. Each species of Gobiodon considered here inhabited one or more species of Acropora more frequently than expected from the availability of these corals; however, the consistency of habitat use varied among the species studied. Two species, G. histrio and G. brochus, exhibited very conservative patterns of habitat use among geographic locations. Despite an increasing diversity of suitable acroporid species from the southern to northern, and changes in the relative abundances of corals among locations, these species of fish consistently inhabited the same species of coral. Such conservative patterns of habitat use among widespread locations might be expected if (1) the corals inhabited offer considerable fitness benefits and/or (2) interspecific interactions operate in a similar manner at each location, thereby producing similar patterns of habitat use. At Lizard Island, G. histrio and G. brochus compete for colonies of Acropora nasuta (Munday et al. 2001), a species of coral that favors growth and survival of these species (Munday 2001). Furthermore, G. histrio was found to be a superior competitor to G. brochus (Munday et al. 2001). The consistent use of A. nasuta by G. histrio appears to be the result of a superior competitor using the best habitat at all geographic locations. In contrast, the consistent pattern of habitat use for G. brochus is likely to be the result of interspecific interactions with G. histrio producing similar patterns of habitat use at each location. Habitat use by two species, G. axillaris and G. quinquestrigatus, changed among locations; however, these changes did not appear to have substantial effects on the regression analysis among locations. Plasticity in habitat use would enable these species to make use of new or abundant coral species at any location. This strategy may be favored where habitats are a limited resource, as appears to be the case for these species. G. quinquestrigatus appears to be more generalist in habitat use than the other species considered, which could be a mechanism to reduce competition for space. Three of the four species considered here, G. axillaris, G. histrio, and G. quinquestrigatus, were associated with particular exposure regimes independently of coral abundances. Although a variety of processes could produce this pattern, it is most easily explained by broad-scale habitat selection. Habitat selection at the scale of individual coral heads is common among coral reef fishes at the time of settlement (Sweatman 1983; Sale et al. 1984; Shulman 1984; Booth 1992; O hman et al. 1998). Habitat selection is also known to occur at the scale of reef zones (Wellington 1992; Gutie rrez 1998) and some species appear to select habitats for settlement at whole-reef scales (Doherty et al. 1996). Indeed, larvae of reef fishes competent to settle are capable of directed movement at scales that can influence their distributions over whole reefs (Leis et al. 1996; Stobutzki 1998). Alternatively, they might make use of hydrodynamic characteristics that result in differential reef-scale distributions (Cowen and Castro 1994). Gobiodon settling to the reef might first select the broad environment preferred, either by directed movement to that environment, or by selection of hydrodynamic characteristics that favor transport to those sites. Within these sites they may then select habitat at the scale of individual coral colonies. For all species except G. axillaris, the lack of significant variation in abundance among sites, once the effect of habitat availability was removed, indicates that habitat availability is a major determinant of abundance among sites separated by hundreds to thousands of meters. The abundances of G. axillaris, G. histrio, and G. quinquestrigatus also varied among geographic locations, independently of the availability of coral habitats. These differences in abundance could be the result of geographic variation in local-scale processes (e.g., competition or predation), or due to factors acting on much larger spatial scales (e.g., large-scale patterns of larval supply or abiotic conditions). Except for the absence of G. brochus from Kimbe Bay, the species composition of Gobiodon changes very little over the geographic range of this study (Munday et al. 1999). Therefore, there are no species replacements that might dramatically alter the nature of interactions among G. axillaris, G. histrio, and G. quinquestrigatus at each location. G. axillaris and

10 114 G. histrio are known to compete for habitat space (Munday et al. 2001) and variation in the relative abundances of preferred corals among locations could influence the intensity of these competitive interactions. However, these two species also partition their use of reef zones (Munday et al. 1997; Munday 2000) and tend to inhabit different exposure regimes (above), which might mediate competitive effects. There is little evidence that the intensity of competitive interactions varies among locations; however, manipulative experiments at each location would be required to test this idea. Processes acting on large spatial scales could also account for the observed differences in abundance among locations that are not explained by habitat availability. Processes influencing the abundance of coral-reef fishes on these spatial scales are not well understood (Thresher 1991); however, patterns of larval supply help determine geographical patterns of abundance in other marine animals (Alexander and Roughgarden 1996; Palmer et al. 1996; Connolly and Roughgarden 1998) and could account for the patterns observed here. Larvae may be preferentially advected to particular locations, or conditions in the plankton might favor survival in some locations but not others. For example, food for larvae may be generally abundant in some locations but sparse in others. Changes in the physical environment among widespread locations make it likely that physiological tolerances will also help determine patterns of abundance among regions (Menge and Olson 1990). Water temperatures, salinity, and a variety of other physical variables differ among the locations in this study and may influence the proportion of larvae or post-settlement fishes surviving at each location. The clear distinction between abundances of G. histrio at locations in compared to the supports the idea that regional-scale processes are interacting with habitat availability to determine broadscale patterns of abundance for this species. However, the very different patterns of abundance among locations when a comparison is made among G. histrio, G. axillaris, and G. quinquestrigatus indicate that the processes influencing regional-scale variation in abundances, above and beyond the effects of habitat availability, are likely to vary among species. The close association between habitat availability and patterns of abundance of coral-dwelling fishes at large spatial scales (Holbrook et al. 2000, this study) has implications for understanding how habitat specialist species may respond to increasing levels of disturbance. Firstly, large-scale disturbances that reduce the number of living coral colonies are likely to have widespread flowon effects to the abundance of fishes that usually inhabit these corals. Human activities are widely attributed to have caused direct declines in the extent of coral cover in many regions, and large-scale declines in coral cover as a result of coral bleaching have also been widely reported in recent years (Wilkinson 2000). Such declines in coral cover are likely to have direct impacts on the abundances of coral-dwelling fishes at the scale of whole populations. Second, many other reef taxa have specialist patterns of habitat use and changes in habitat availability could have similar flow-on effects for a wide range of species. Acknowledgments Thanks to staff at Mahonia Na Dari Research Station, Motupore Island Research Station, Lizard Island Research Station, One Tree Island Research Station, Walindi Diving, and Loloata Island Resort for support throughout this project. Thanks also to F. Kroon, K. Munday, A. Persson, and T. Sin who assisted in the field and to C. Wallace and J. Wolstenholme who identified corals and tutored me on field identifications. Field work was made possible by a Lizard Island Research Station Doctoral Scholarship from the Australian Museum, a One Tree Island Doctoral Fellowship from the Mazda Foundation, and a Merit Research Grant from James Cook University. Thanks to Mahonia Na Dari Research Station and Max Benjamin of Walindi Plantation for continued support of my research in. D. Booth, J. Caley, S. Connolly, G. Jones, M. Hixon, T. Hughes, M. McCormick, D. Morris, P. Sale, J. Tanner, R. Warner, and two anonymous reviewers gave valuable comments on various versions of this paper. This is contribution number 1 from the Centre for Coral Reef Biodiversity. References Alexander SE, Roughgarden J (1996) Larval transport and population dynamics of intertidal barnacles: a coupled benthic/oceanic model. Ecol Monogr 66: Bell JD, Galzin R (1984) Influence of live coral cover on coral-reef fish communities. Mar Ecol Prog Ser 15: Booth DJ (1992) Larval settlement patterns and preferences by domino damselfish Dascyllus albisella Gill. J Exp Mar Biol Ecol 155: Bouchon-Navaro Y, Bouchon C (1989) Correlations between chaetodontid fishes and coral communities of the Gulf of Aqaba (Red Sea). Environ Biol Fishes 25:47 60 Brown JH (1984) On the relationship between abundance and distributions of species. Am Nat 124: Buchheim JR, Hixon MA (1992) Competition for shelter holes in the coral-reef fish Acanthemblemaria spinosa Metzelar. J Exp Mar Biol Ecol 164:45 54 Butman CA (1987) Larval settlement of soft-sediment invertebrates: the spatial scales of pattern explained by active habitat selection and the emerging role of hydrodynamic processes. Oceanogr Mar Biol Annu Rev 25: Caley MJ (1995a) Reef-fish community structure and dynamics: an interaction between local and larger-scale processes? Mar Ecol Prog Ser 129:19 24 Caley MJ (1995b) Community dynamics of tropical reef fishes: local patterns between latitudes. Mar Ecol Prog Ser 129:7 18 Caselle JE, Warner RR (1996) Variability in recruitment of coral reef fishes: the importance of habitat at two spatial scales. Ecology 77: Clarke RD (1996) Population shifts in two competing fish species on a degrading coral reef. Mar Ecol Prog Ser 137:51 58 Connolly SR, Roughgarden J (1998) A latitudinal gradient in northeast Pacific intertidal community structure: evidence for an oceanographic based synthesis of marine community theory. Am Nat 151: Connolly SR, Menge BA, Roughgarden J (2001) A latitudinal gradient in recruitment of intertidal invertebrates in the northeast Pacific Ocean. Ecology 82: Cowen RK, Castro LR (1994) Relation of coral reef fish larval distributions to island scale circulation around Barbados, West Indies. Bull Mar Sci 54: Doherty PJ, Fowler T (1994) An empirical test of recruitment limitation in a coral reef fish. Science 263: Doherty PJ, Kingsford M, Booth D, Carleton J (1996) Habitat selection before settlement by Pomacentrus coelestis. Mar Freshwater Res 47:

11 115 Fautin DG (1992) Anemonefish recruitment: the roles of order and chance. Symbiosis 14: Fowler AJ, Doherty PJ, Williams DMcB (1992) Multi-scale analysis of recruitment of a coral reef fish on the Great Barrier Reef. Mar Ecol Progr Ser 82: Fox LR, Morrow PA (1981) Specialization: species property or local phenomenon. Science 211: Friedlander AM, Parish JD (1998) Habitat characteristics affecting fish assemblages on a Hawaiian coral reef. J Exp Mar Biol Ecol 224:1 30 Gaston KJ, Lawton JH (1990) Effects of scale and habitat on the relationship between regional distribution and local abundance. Oikos 58: Gutie rrez L (1998) Habitat selection by recruits establishes local patterns of adult distribution in two species of damselfishes: Stegastes dorsopunicans and S. planifrons. Oecologia 115: Harold AS, Winterbottom R (1999) Gobiodon brochus, a new species of gobiid (Teleostei: Gobioidei) from the western South Pacific, with a description of its unique jaw morphology. Copeia:49 57 Heatwole H (1981) A coral island: the story of One Tree Island and its reef. Collins, Sydney, Australia Hixon MA, Beets JP (1993) Predation, prey refuges, and the structure of coral-reef fish assemblages. Ecol Monogr 63: Holbrook SJ, Forrester GE, Schmitt RJ (2000) Spatial patterns in abundance of a damselfish reflect availability of suitable habitat. Oecologia 122: Holthus P, Maragos J (1996) Kimbe Bay rapid ecological assessment: coral reef habitat, volume 2. Nature Conservancy, Auckland, New Zealand Hubbell SP (1997) A unified theory of biogeography and relative species abundance and its application to tropical rain forests and coral reefs. Proc 8th Int Coral Reef Symp 1:33 42 Hughes TP, Baird AH, Dinsdale EA, Moltschaniwskyj N, Pratchett MS, Tanner JE, Willis B (1999) Patterns of recruitment and abundance of corals along the Great Barrier Reef. Nature 397:59 63 Jennings S, Boullé DP, Polunin NVC (1996) Habitat correlates of the distribution and biomass of Seychelles reef fishes. Environ Biol Fishes 46:15 25 Jones GP (1991) Postrecruitment processes in the ecology of coral reef populations: a multifactorial perspective. In: Sale PF(ed) The ecology of fishes on coral reefs. Academic Press, San Diego, pp Kaufman LS, Ebersole JP (1984) Microtopography and the organisation of two assemblages of coral reef fishes in the West Indies. J Exp Mar Biol Ecol 78: Kuwamura T, Yogo Y, Nakashima Y (1994) Population dynamics of goby Paragobiodon echinocephalus and host coral Stylophora pistillata. Mar Ecol Progr Ser 103:17 23 Lassig B (1981) Significance of the epidermal ichthyotoxic secretion of coral-dwelling gobies. Toxicon 19: Leis JM, Sweatman HPA, Reader SE (1996) What the pelagic stages of coral reef fishes are doing out in blue water: daytime field observations of larval behavioural capabilities. Mar Freshwater Res 47: Luckhurst BE, Luckhurst K (1978) Analysis of the influence of substrate variables on coral reef fish communities. Mar Biol 49: MacNally RC (1995) Ecological versatility and community ecology. Cambridge University Press, Cambridge Maniwavie T, Sweatman H, Marshall P, Munday P, Rei V (2000) Status of coral reefs of Australasia: Australia and Papua New Guinea. In: Wilkinson CR (ed) Status of coral reefs of the world: Australian Institute of Marine Science, Townsville, pp Manly B, McDonald L, Thomas D (1993) Resource selection by animals, statistical design and analysis for field studies. Chapman Hall, London Maurer BA (1999) Untangling ecological complexity. University of Chicago Press, Chicago Meekan MG, Steven ADL, Fortin MJ (1995) Spatial patterns in the distribution of damselfishes on a fringing coral reef. Coral Reefs 14: Menge BA, Olson AM (1990) Role of scale and environmental factors in regulation of community structure. Trends Ecol Evol 5:52 57 Morris DW (1992) Scales and costs of habitat selection in heterogeneous landscapes. Evol Ecol 6: Munday PL (2000) Interactions between habitat use and patterns of abundance in coral-dwelling fishes. Environ Biol Fishes 58: Munday PL (2001) Fitness consequences of habitat use and competition among coral-dwelling fishes. Oecologia 128: Munday PL, Jones GP, Caley MJ (1997) Habitat specialisation and the distribution and abundance of coral-dwelling gobies. Mar Ecol Progr Ser 152: Munday PL, Caley MJ, Jones GP (1998) Bi-directional sex change in a coral dwelling goby. Behav Ecol Sociobiol 43: Munday PL, Harold AS, Winterbottom R (1999) Guide to coraldwelling gobies, genus Gobiodon (Gobiidae) from Papua New Guinea and the Great Barrier Reef. Rev Fr Aquariol 26:49 53 Munday PL, Jones GP, Caley MJ (2001) Interspecific competition and coexistence in a guild of coral-dwelling fishes. Ecology 82: O hman MC, Munday PL, Jones GP, Caley MJ (1998) Settlement strategies and distribution patterns of coral-reef fishes. J Exp Mar Biol Ecol 225: Palmer MA, Allan JD, Butman CA (1996) Dispersal as a regional process affecting the local dynamics of marine and stream invertebrates. Trends Ecol Evol 11: Pichon M, Morrissey J (1981) Benthic zonation and community structure of South Island Reef, Lizard Island (Great Barrier Reef). Bull Mar Sci 3: Pulliam HR, Danielson BJ (1991) Sources, sinks, and habitat selection: a landscape perspective on population dynamics. Am Nat 137:S50 S66 Ricklefs RE (1987) Community diversity: relative roles of local and regional processes. Science 235: Ricklefs RE, Schluter D (1993) Species diversity and ecological communities. University of Chicago Press, Chicago Roberts CM, Ormond RFG (1987) Habitat complexity and coral reef fish diversity and abundance on Red Sea fringing reefs. Mar Ecol Prog Ser 41:1 8 Rosenzweig ML (1991) Habitat selection and population interactions: the search for mechanism. Am Nat 137:S5 S28 Sale PF, Doherty PJ, Eckert GJ, Douglas WA, Ferrell DJ (1984) Large scale spatial and temporal variation in recruitment to fish populations on coral reefs. Oecologia 64: Schmitt RJ, Holbrook SJ (1999) Settlement and recruitment of three damselfish species: larval delivery and competition for shelter space. Oecologia 118:76 86 Shulman MJ (1984) Resource limitation and recruitment patterns in a coral reef fish assemblage. J Exp Mar Biol Ecol 74: Stobutzki IC (1998) Interspecific variation in sustained swimming ability of late pelagic stage reef fish from two families (Pomacentridae and Chaetodontidae). Coral Reefs 17: Sweatman HPA (1983) Influence of conspecifics on choice of settlement sites by larvae of two pomacentrid fishes (Dascyllus aruanus and D. reticulatus). Mar Biol 75: Thresher RE (1991) Geographic variability in the ecology of coral reef fishes: evidence, evolution and possible implications. In: Sale PF(ed) The ecology of fishes on coral reefs. Academic Press, San Diego, pp Tolimieri N (1995) Effects of microhabitat characteristics on the settlement and recruitment of a coral reef fish at two spatial scales. Oecologia 102:52 63 Tolimieri N (1998) Contrasting effects of microhabitat use on largescale adult abundance in two families of Caribbean reef fishes. Mar Ecol Prog Ser 167: