INTRODUCTION
Fish species interact closely with their habitat for the majority of their lives (Jones & Syms, Reference Jones and Syms1998) and therefore, there is reason to hypothesize that the distribution and structure of reef fish communities should correlate with variables of the habitats present (McGehee, Reference McGehee1994; Öhman, Reference Öhman1998). Factors previously found to influence reef fish community structure include benthic diversity, habitat complexity, live coral cover, macroalgal cover, depth and exposure. Topographically complex reef habitats or those with high numbers of growth forms or high benthic diversity should provide more microhabitats, refuge sites, and food resources for a higher number of individuals and species. Research in the literature supports this in general, with positive relationships being found with fish diversity, species richness and abundance (Luckhurst & Luckhurst, Reference Luckhurst and Luckhurst1978; Carpenter et al., Reference Carpenter, Miclat, Albaladejo and Corpuz1981; Ault & Johnson, Reference Ault and Johnson1998; Lara & Gonzalez, Reference Lara and Gonzalez1998; Öhman & Rajasuriya, Reference Öhman and Rajasuriya1998; Lirman, Reference Lirman1999; Nagelkerken et al., Reference Nagelkerken, Dorenbosch, Verberk, Cocheret de la Morinière and van der Velde2000; García-Charton & Pérez-Ruzafa, Reference García-Charton and Pérez-Ruzafa2001; McClanahan & Arthur, Reference McClanahan and Arthur2001; Friedlander et al., Reference Friedlander, Brown, Jokiel, Smith and Rodgers2003; Dominici-Arosemena & Wolff, Reference Dominici-Arosemena and Wolff2005, Reference Dominici-Arosemena and Wolff2006; Espinoza & Salas, Reference Espinoza and Salas2005; Gratwicke & Speight, Reference Gratwicke and Speight2005a; Brokovich et al., Reference Brokovich, Baranes and Goren2006; Kuffner et al., Reference Kuffner, Brock, Grober-Dunsmore, Bonito, Hickey and Wright2007).
Live bottoms offer food sources and often higher structural variability and shelter (Huntsman & Waters, Reference Huntsman, Waters, Polovina and Ralston1987) and positive relationships have been found between live coral cover and reef fish diversity, richness and abundance (Bell & Galzin, Reference Bell and Galzin1984; Bouchon-Navaro & Bouchon, Reference Bouchon-Navaro and Bouchon1989; Chabanet et al., Reference Chabanet, Ralambindrainy, Amanieu, Faure and Galzin1997; Connell & Kingsford, Reference Connell and Kingsford1998; Jones et al., Reference Jones, McCormick, Srinivasan and Eagle2004; Bozec et al., Reference Bozec, Dolédec and Kulbick2005) with declines in fish communities found after coral cover loss (Wilson et al., Reference Wilson, Graham, Pratchett, Jones and Polunin2006, Reference Wilson, Burgess, Cheal, Emslie, Fisher, Miller, Polunin and Sweatman2008), although this was not found in other studies (Roberts & Ormond, Reference Roberts and Ormond1987; Chapman & Kramer, Reference Chapman and Kramer1999; Lecchini et al., Reference Lecchini, Adjeroud, Pratchett, Cadoret and Galzin2003). Williams & Polunin (Reference Williams and Polunin2001) reported positive relationships between algae cover and biomass of reef fish, but other researchers found no such relationship (Chabanet et al., Reference Chabanet, Ralambindrainy, Amanieu, Faure and Galzin1997; Friedlander & Parrish, Reference Friedlander and Parrish1998; Eagle et al., Reference Eagle, Jones and McCormick2001). Depth and exposure has also been found to positively correlate with species richness, diversity and abundance on coral and rocky reefs (Roberts & Ormond, Reference Roberts and Ormond1987; Adjeroud et al., Reference Adjeroud, Letourner, Procher and Salvat1998; Friedlander & Parrish, Reference Friedlander and Parrish1998; Öhman & Rajasuriya, Reference Öhman and Rajasuriya1998; Ferreira et al., Reference Ferreira, José, Gonçalves and Coutinho2001; Arreola-Robles & Elorduy-Garay, Reference Arreola-Robles and Elorduy-Garay2002; Lecchini et al., Reference Lecchini, Adjeroud, Pratchett, Cadoret and Galzin2003; Dominici-Arosemena & Wolff, Reference Dominici-Arosemena and Wolff2005, Reference Dominici-Arosemena and Wolff2006; Brokovich et al., Reference Brokovich, Baranes and Goren2006).
Studying fish–habitat associations can provide insight into the effects of habitat structure that can be accounted for when other variables are examined; it enables a fuller understanding of how disturbances affect fish communities; and an improved understanding of the spatial distribution of reef fish has implications for reef management. Despite an increasing number of studies on reef fish–habitat associations, a broader range of species, habitats, and biogeographical regions must be studied to better understand how reef fish communities are structured (Jones & Syms, Reference Jones and Syms1998; Öhman, Reference Öhman1998). In addition, studies on reefs of non-biotic origin in tropical areas also require study (Öhman, Reference Öhman1998; Ferriera et al., 2001).
The reef-associated fish communities of the Las Perlas Archipelago (LPA), Panama have received little attention. Previous coral reef studies were concentrated around the northern islands (Glynn & Stewart, Reference Glynn and Stewart1973; Wellington, Reference Wellington1982). Two dominant reef types are present in the archipelago: coral reefs (CRs) and coral communities (CCs). Although the CRs of the LPA are not as diverse or extensive as those found in the western Pacific or the Caribbean, their simplicity make them ideal for investigating reef fish–habitat relationships (Glynn & Maté, Reference Glynn and Maté1996). CCs are the term given in this study to rocky reefs composed of sandstone and basaltic rock provides a substratum that supports scleractinian corals and gorgonians. They differ from CRs because they lack live coral frameworks. CCs fringe much of the coastline of the islands and provide an extensive reef habitat for fish.
This study is the first investigation of reef fish assemblages throughout the whole of the LPA. The aim was to describe and compare the fish communities found on CRs and CCs around the LPA in relation to their habitat. Specifically, we addressed the following questions: (1) do reef fish assemblages of CRs and CCs differ in their composition, diversity, abundance, and fish length?; and (2) do the differences observed between and within these two habitats correlate to measured habitat variables?
MATERIALS AND METHODS
Study site
The LPA lies between 08°40′19″N 79°03′49″W and 08°11′46″N 78°46′31″W within the Gulf of Panama (Figure 1). It is composed of 250 mostly uninhabited basaltic rock islands and islets that lie within the 50 m isobath and fall within the Tropical Eastern Pacific (TEP) biogeographical zone. The LPA experiences an upwelling period during the dry season (January–April) that results in plankton blooms and high marine productivity. The mean tidal range for the archipelago is 3.8 m (Glynn & Maté, Reference Glynn and Maté1996).
Fig. 1. Coral reef and coral community sites surveyed in Las Perlas Archipelago.
Historically, no CRs were thought to exist within the TEP because of the cool currents and upwelling. However, studies conducted over the last 30 years have found extensive reefs in the Gulf of Chiriquí and in the Gulf of Panama (Glynn & Stewart, Reference Glynn and Stewart1973; Glynn, Reference Glynn1976; Guzman et al., Reference Guzman, Guevara and Breedy2004). The CRs of this region are characterized by their small size (a few hectares), discontinuous distribution, and low species diversity (Cortés, Reference Cortés1997). The largest aggregation of CRs in the Gulf of Panama occurs in the LPA (Glynn & Maté, Reference Glynn and Maté1996). The main reef-building corals are branching species of Pocillopora, but massive Pavona coral species also occur in deeper areas (Glynn & Stewart, Reference Glynn and Stewart1973; Glynn, Reference Glynn1976). CCs also contain Pocillopora, and Pavona species, as well as gorgonians (Guzman et al., Reference Guzman, Benfield, Breedy and Mair2008). Currently, the CRs and associated marine habitats of the LPA are affected by El Niño Southern Oscillation events, sedimentation, pollution, overfishing, and coastal tourism development. Detailed information was not available on the degree to which these factors affected our study sites, but they are believed to be minimally fished in an artisanal nature (personal observation).
Survey method
We selected 45 survey locations (31 CRs and 14 CCs) based on where reefs existed (Figure 1), as identified by manta tow surveys of the coastline and by satellite image classification (Benfield et al., Reference Benfield, Guzman, Mair and Young2007; Guzman et al., Reference Guzman, Benfield, Breedy and Mair2008). Surveys occurred in August 2003, May 2004 and August 2004. The depth range surveyed was narrow, with a maximum survey depth of 12 m, but most sites had a maximum depth of less than 6 m due to the shallow distribution of CRs.
We adapted our survey methods from the Atlantic and Gulf Rapid Reef Assessment methodology (AGRRA, Kramer et al., Reference Kramer, Lang, Marks, Garza-Pérez and Ginsburg2005). At each site, we surveyed four 2 m × 30 m transects, which was a revision to the AGRRA methodology that recommends ten. Ten surveys were not possible on most reefs as their small and narrow size would have led to overlapping transects. We used stratified random sampling to place transects, which ran parallel to the shore. When a slope was present, we placed two transects near the top and two closer to the bottom, although the depth variation between the two sections was never more than 5 m. Replicate transects were placed randomly within the stratified zones, separated by a minimum of 5 m (estimated by fin kicks) to prevent overlap between the fish counted.
In this study we examined all diurnally active post-recruitment individuals from 1 cm long upwards, excluding gobies and blennies because these small, cryptic species were too difficult to identify rapidly in the field. This approach was a revision to the AGRRA protocol, which was developed for a different geographical location and recommends counting only selected families. Individuals were identified to species level, with identifications based on Allen & Robertson (Reference Allen and Robertson1994); counted; and placed in a size category (0–5, 6–10, 11–20, 21–30, 31–40 and >40 cm) according to their total length. For fish over 40 cm, estimated lengths to the nearest centimetre were recorded. We estimated lengths using a graduated PVC t-piece with an arm width of 1 m (sensu Kramer et al., Reference Kramer, Lang, Marks, Garza-Pérez and Ginsburg2005). When individuals were present in a large school, a small group was counted and that number was multiplied up to the size of the school. The lengths of large groups were classified into one or more size-categories as necessary. Test dives took place before surveying commenced for divers to familiarize themselves with species and equipment and to assess the divers' surveying accuracy (sensu Kramer et al., Reference Kramer, Lang, Marks, Garza-Pérez and Ginsburg2005).
The bottom composition of each site was assessed simultaneously by another team of two divers who surveyed 90 1-m2 quadrats that were placed end to end along transects that ran parallel to the shore in the same areas where the fish surveys were conducted (sensu Guzman et al., Reference Guzman, Guevara and Breedy2004). Percentage cover of scleractinian and soft coral species were recorded to species level, and the cover of sponges, crustose coralline algae (CCA), macroalgae, turf algae and dead coral also were recorded. Wave exposure, tide state, sea state, topography and slope were graded (Table 1), and depth below chart datum was calculated.
Fig. 2. MDS ordinations of species abundance and habitat variables selected through BIO-ENV. Triangles, coral reef; circles, coral community. Ordinations for habitat variables that showed the highest correlation with the clustering of the species abundance ordination also are shown. Bubble size is proportional to the habitat value at each site.
Fig. 3. MDS ordinations of species length and habitat variables selected through BIO-ENV. Triangles, coral reef; circles, coral community. The ordinations for the habitat variables that showed the highest correlation with the pattern of clustering of the MDS ordination of species length also are given. Bubble size is proportional to the habitat value at each site.
Table 1. Environmental variables recorded for each survey site.
Data analysis
For each site we calculated species richness, Shannon diversity (log10), and mean abundance of each fish species. We calculated the mean length for each species per site by multiplying the number of fish in each size-category by the mid-point of the size-category (2, 7, 15, 25 and 35 cm), summing these values, and then dividing it by the total number of the species counted. We also calculated the mean length of all fish per site. Additionally, fish species were allocated into families and broad trophic groups based upon Allen & Robertson (Reference Allen and Robertson1994) and FishBase (Froese & Pauly, Reference Froese and Pauly2005).
We examined the relationships between habitat type, benthic composition, and fish community parameters using univariate and multivariate methods. Shannon diversity, species richness, mean abundance, and mean length met the requirements for ANOVA (i.e. homogeneity of variance and normality), although mean abundance and mean length per site had to be square-root transformed to allow the use of parametric statistics. ANOVAs were performed to confirm that no significant difference existed in Shannon diversity, species richness, mean abundance, and mean length among the survey seasons (P > 0.05 in all cases), which allowed data from habitat sites sampled at different times to be combined in the same analyses. ANOVAs on sea state and t-tests for tide state confirmed that these variables had no significant effect on the aforementioned parameters (P > 0.05). Thus, the data from all survey seasons, tide states, and sea states were pooled for use in the analyses. We used a two-sample t-test to test for significant differences in diversity, species richness, mean abundance, and mean length between the two habitats. Relationships between habitat variables (benthic diversity (Shannon diversity, log10) and the percentage cover of branching coral, massive coral, macroalgae, CCA, soft coral and sponge) and fish parameters (fish diversity, species richness, mean abundance, mean length, mean abundance of families and mean abundance of trophic groups) were investigated using Spearman's rank correlation within each habitat.
We implemented non-parametric multi-dimensional scaling (MDS) using the multivariate statistical software PRIMER v5 to further test for differences in fish community composition between and within coral reef and coral community sites and to examine the relationships with habitat variables (Clarke & Ainsworth, Reference Clarke and Ainsworth1993; Clarke & Gorley, Reference Clarke and Gorley2001; Clarke & Warwick, Reference Clarke and Warwick2001). The abundance data for species, families, and trophic groups and the length data for species were square-root transformed prior to the implementation of the Bray–Curtis similarity measure to create a similarity matrix. This technique was used because it down-weighted the importance of very abundant variables, which we believed was a valuable step as the samples often were dominated by a few species. The similarity matrices then were converted to MDS ordinations and an analysis of similarity (ANOSIM) test was used to test for significant differences in species, family, or trophic composition between the two habitats. We also produced similarity matrices using normalized Euclidean distance (as recommended for environmental data by Clarke & Warwick (Reference Clarke and Warwick2001)) for the habitat data for both reef types combined. The habitat variables examined were wave exposure, depth, topography, slope, benthic diversity, and the percentage cover of branching coral, massive coral, soft coral, macroalgae, CCA, and sponge. The percentage cover data were square-root transformed prior to analysis because they differed by an order of magnitude, and data were normalized because they were not in the same units. We used the BIO-ENV module in PRIMER to determine the combination of environmental variables whose similarity matrix had the highest Spearman rank-correlation with the fish similarity matrix. The significance of the correlation between the habitat and fish matrices was tested using the RELATE command in PRIMER and a weighted Spearman rank correlation, ρ (Clarke & Ainsworth, Reference Clarke and Ainsworth1993; Clark & Gorley, 2001; Clarke & Warwick, Reference Clarke and Warwick2001). This procedure was repeated within coral reef and coral community sites to examine the influence of habitat variables on any of the observed differences in species, family, or trophic composition. We used the SIMPER command of PRIMER to identify those species, families, and trophic groups that contributed to 90% of the fish composition of each habitat and those that played a part in any observed differences between CRs and CCs.
RESULTS
Descriptive results
During the surveys we counted 34,336 fish (26 families, 79 species), with an average of 715.3±304.5 SD per site (Table 2). The three most abundant fish were Thalassoma lucasanum, Chromis atrilobata and Stegastes acapulcoensis, which constituted up to 66.5% of the total number of fish found. Thalassoma lucasanum was the only species found at all 45 sites. A few rare species were found only at one site in low numbers (less than 30 individuals): Apogon pacifici, Caranx caballus, Scarus compressus, Lutjanus aratus, Caranx sexfasciatus, Gnathanodon speciosus, Lutjanus inermis, Mycteroperca xenarcha and Anisotremus taeniatus. Over 90% of the fish surveyed were members of the Labridae, Pomacentridae, or Haemulidae families. The family with the greatest abundance was the Labridae (45%), followed by Pomacentridae (36.3%). Species from Labridae, Serranidae, and Pomacentridae families were recorded at all sites, and members of Tetraodontidae were recorded at all but one site. Members of Haemulidae were found at only 28 of the 45 sites, but often occurred in large aggregations. The most dominant trophic groups found across all sites combined were plantivores (53.5%), invertivores (20.1%) and herbivores (18.3%), with the rest of the fish surveyed being omnivores.
Table 2. Species found during surveys of CRs and CCs in Las Perlas Archipelago, Pacific Panama.
The reef fish surveyed were dominated by small species and individuals, with 72% being ≤10 cm in length and the majority being 6–10 cm long. We found very few fish >40 cm long, and these were generally fish such as Fistularia commersonii and members of the Scaridae.
The greatest mean abundance of fish occurred at site 14 (CC), and the top three sites with the highest abundance were CCs. Site 19 had the highest mean fish length per site; although this site was a CR, three out of the five sites with the highest mean fish length were CCs. The majority of sites with low mean length were CRs in the northern part of the archipelago. The mean species richness per site was 20±4.9 SD and the most species rich sites were generally CCs. Site 40 was the most species rich site, and site 21 had the least species. The five sites with the lowest richness, each with <15 species, lie close together around Isla Mogo Mogo, Isla Chapera and Isla Bolaños in the northern area of the archipelago, and all were CRs. Isla Elefante and Isla Casaya had the highest fish diversity with the lowest being found at CR sites 22 and 26.
COMPARISON OF REEF FISH PARAMETERS IN CORAL REEF AND CORAL COMMUNITY HABITATS AND RELATIONSHIPS WITH HABITAT VARIABLES
A two-sample t-test revealed a significant difference (P < 0.05) in diversity (mean H = 0.819 for CCs, 0.669 for CRs, T = 3.64, P < 0.001, df = 28) and species richness (mean species richness = 22.5 for CCs, 18.2 for CRs, T = 2.16, P = 0.045, df = 15) between CRs and CCs. However, we found no significant difference between the habitats for mean abundance (T =−0.31, P = 0.759, df = 15) and mean length (T = 1.63, P = 0.120, df = 18).
The clustering of sites found using non-parametric MDS ordination based on the abundance of fish species surveyed was related significantly to the habitat type (ANOSIM, R = 0.296, P < 0.001; Figure 2). Exposure, topography and coverage of massive corals were generally greater in CCs than CRs, whereas branching coral cover was higher in the coral reef sites. BIO-ENV confirmed that the similarity matrix produced from exposure, depth, topography, percentage branching and massive coral cover was significantly correlated with the fish abundance matrix (Table 3; Figure 2). There was also a significant difference in family composition (R = 0.174, P = 0.026) between the habitats although this was not great as the difference found using species abundance. Similar habitat variables were found to correlate with this, with the addition of benthic diversity (Table 3). A significant difference in composition between CRs and CCs was also found based on trophic group abundance (R = 0.16, P = 0.026). The low R value indicates the two groups were not well separated but similar habitat variables were found to be influential in explaining this pattern, with the addition of sponge cover which was greater in CCs (Table 3). The length composition of fish species on CRs and CCs also differed significantly (R = 0.295, P < 0.001) and this was correlated with slope, topography, and the percentage cover of branching coral and CCA (Table 3; Figure 3).
Table 3. BIO-ENV and RELATE results of the similarity between the multivariate pattern of the fish community and habitat variables. Significance level of P < 0.05 is denoted by * and P < 0.001 **. ρ, Spearman rank-correlation; H, Shannon diversity of benthic composition; CCA, crustose coralline algae. Stress of the fish and habitat MDS relate to the 2-D MDS configuration.
The SIMPER analyses revealed that the average dissimilarity between the two habitats was 51.4%, 36.0%, and 43.8% for species, family, and trophic group abundance, respectively. Differences between the two habitats were due to variation in abundance of T. lucasanum, C. atrilobata, Halichoerus dispilus and S. acapulcoensis. Thalassoma lucasanum contributed most to the dissimilarity and was more abundant in CRs, whereas the other three species were more abundant in CCs. Members of the Labridae contributed the most to the dissimilarity in family abundance between the habitats and was the most common family in CRs, whereas members of the Pomacentridae were more abundant in CCs. The abundance of planktivores was important in accounting for the dissimilarity in trophic group composition; they were the most commonly found trophic group in CRs, whereas herbivores were the most abundant in CCs. Plantivores were the dominant trophic group in both habitats although this was higher in CRs (55.4%) compared to CCs (48.2%). Herbivores were the second most dominant group in CCs (22.7%), unlike CRs where invertivores were (20.1%). Herbivores composed 18.3% of the fish in CRs and invertivores 21.1% in CCs.
RELATIONSHIPS BETWEEN REEF FISH PARAMETERS AND HABITAT VARIABLES IN CRS
A positive, significant correlation was found between species richness and the percentage cover of massive corals and Labridae abundance and benthic diversity as well as other non-significant but notable correlations (Table 4). A significant positive correlation existed between Pomacentridae abundance and percentage of CCA. We hypothesized that CCA cover might be positively correlated with topography because the area available for CCA colonization would increase, but an ANOVA revealed no significant difference in mean CCA cover in response to topography category (F = 0.08, P = 0.782, df = 1). Several significant correlations for trophic groups were also found (Table 4). MDS ordinations were run using all variables except topography, slope, and exposure, as these did not change among coral reef sites. Combinations of depth, benthic diversity and the branching coral and macroalgae cover provided the highest significant correlation with the species, family and trophic group abundance similarity matrices (Table 3). Massive and laminar coral cover was found to significantly influence species length composition.
Table 4. Spearman rank-correlation results for coral reef fish parameters and habitat variables. H, Shannon diversity; CCA, crustose coralline algae; *, significant at P < 0.05; values in roman type, correlation value; values in italic type, significance value.
SIMPER analysis revealed that average similarity among coral reef sites was 56.1%, 67.6%, and 61.3% for species, family, and trophic group abundance, respectively. Thalassoma lucasanum, S. acapulcoensis and C. Atrilobata contributed to over 50% of the similarity in species composition among CR sites, with T. lucasanum being the most dominant. Labridae and Pomacentridae contributed to over 70% of the similarity in family composition and planktivores contributed most out of the trophic groups to the similarity among sites.
RELATIONSHIPS BETWEEN REEF FISH PARAMETERS AND HABITAT VARIABLES IN CCS
Species richness, mean length and abundance of Pomacentridae, Scaridae and piscivores were positively correlated with CCA cover (Table 5). Unlike CRs an ANOVA test identified a significant difference in the mean CCA cover in response to topography category (F = 4.75, P = 0.035, df = 2), with CCA cover being greater in more topographically complex sites. Seranidae and Chaetodontidae abundance were positively correlated with the coverage of coral cover (Table 5). Herbivore abundance showed a positive significant correlation with the cover of branching coral and a negative correlation with macroalgae cover (Table 5).
Table 5. Spearman rank-correlation results for coral community fish parameters and habitat variables. H, Shannon diversity; CCA, crustose coralline algae; *, significant at P < 0.05; values in roman type, correlation value; values in italic type, significance value.
The association between habitat variables and fish community composition was not as significant for CCs as it was for coral reef sites. Although combinations of exposure, slope, benthic diversity, and branching and massive coral cover provided the best correlation with the MDS ordinations of species and family abundance in coral community sites, the correlations were not significant (Table 3). However, exposure and the cover of massive coral and sponge were significant in explaining the observed pattern in the ordination produced from trophic group abundance and species length (Table 3).
SIMPER analysis revealed that average similarity among coral community sites was 49.6%, 64.2%, and 57.4% for species, family, and trophic group abundance, respectively. The top three species characterizing CCs were the same as for CRs, except that S. acapulcoensis was the main contributing species. CCs also had a greater number of typical species compared to CRs, and in particular Halichoerus species were important. The main families were similar to those from CRs, except that the Pomacentridae was the top contributory family. Additionally, the Chaetodontidae was as a typical family for CCs but not for CRs.
DISCUSSION
Composition
Generally, the CRs and communities of the LPA surveyed in this study were dominated by relatively few species, by three main families (Labridae, Pomacentridae and Haemulidae) similar to Dominici-Arosemena & Wolff (Reference Dominici-Arosemena and Wolff2006), and by small individuals and species. For the sites surveyed, coral community sites had higher mean abundance, length, species richness, and diversity compared to CRs. We found a significant difference in composition between the two habitats for species, family, and trophic group abundance and for species length, with the largest difference between the habitats being found for species abundance and length composition. Although the ANOSIM analyses detected a significant difference in species composition, the fairly low R value indicates that an overlap between the habitats exists. The habitats were dominated by similar species, although the abundances varied and CCs had more typifying species than CRs, meaning that the similarity of species composition among coral community sites was lower than that of CRs. The composition of family and trophic group composition was similar both between habitats and within habitats.
Reef fish–habitat relationships
The results of this research suggest that reef fish communities in the LPA are influenced to some degree by the structure of their habitat. A combination of topography, exposure, depth, and the percentage of branching and massive corals consistently correlated with the pattern of clustering observed in the fish ordinations for both habitats. The relative importance of these environmental variables could not be easily separable without experimental manipulation, but we concluded that they act together and can account for some of the observed differences in community structure. However, the literature (Luckhurst & Luckhurst, Reference Luckhurst and Luckhurst1978; Chabanet et al., Reference Chabanet, Ralambindrainy, Amanieu, Faure and Galzin1997; Connell & Kingsford, Reference Connell and Kingsford1998; Friendlander & Parrish, 1998; Lara & Gonzalez, Reference Lara and Gonzalez1998; Öhman & Rajasuriya, Reference Öhman and Rajasuriya1998; Nagelkerken et al., Reference Nagelkerken, Dorenbosch, Verberk, Cocheret de la Morinière and van der Velde2000; Dominici-Arosemena & Wolff, Reference Dominici-Arosemena and Wolff2005, 6; Gratwicke & Speight, Reference Gratwicke and Speight2005a; Brokovich et al., Reference Brokovich, Baranes and Goren2006; Garpe et al., Reference Garpe, Yahya, Lindahl and Öhman2006; Kuffner et al., 2007) provides reason to believe that topographical complexity is one of the most influential environmental factors on fish community composition and where it is reduced by disturbance it has negative effects (Wilson et al., Reference Wilson, Graham, Pratchett, Jones and Polunin2006).
At the outset of this research we noted the striking physical difference between the CRs and CCs of the LPA. The CCs have greater vertical height and topography with outcrops, crevices, and boulders, around which fish were seen to aggregate (sensu Grigg Reference Grigg1994; Parrish & Boland Reference Parrish and Boland2004; Gratwicke & Speight, Reference Gratwicke and Speight2005b), in contrast to the structurally simple CRs. More rugose sites offer a larger surface area for colonization by algae, gorgonians and sponges. Thus, these sites can have higher productivity and provide more microhabitats and resources to support fish populations, thereby resulting in higher diversity, greater species richness, larger fish, and higher abundances (Connell & Kingsford, Reference Connell and Kingsford1998; Dominici-Arosemena & Wolff, Reference Dominici-Arosemena and Wolff2005; Espinoza & Salas, Reference Espinoza and Salas2005; Gratwicke & Speight, Reference Gratwicke and Speight2005a, Reference Gratwicke and Speightb). The higher complexity of coral community sites is an explanation for the differences observed (Öhman & Rajasuriya, Reference Öhman and Rajasuriya1998) and the lower complexity of CRs may account for the greater similarity in species composition amongst coral reef sites.
Branching coral cover and exposure were correlated with habitat type. CRs were found in more sheltered sites compared to CCs due to their inability to develop in locations exposed to upwelling (Glynn & Stewart, Reference Glynn and Stewart1973). Exposure also contributed to differences in species length composition between CC sites. In studies of coral and rocky reefs, researchers found that exposure influenced species composition and fish length (Ferreira et al., Reference Ferreira, José, Gonçalves and Coutinho2001; Arreola-Robles & Elorduy-Garay, Reference Arreola-Robles and Elorduy-Garay2002; Dominici-Arosemena & Wolff, Reference Dominici-Arosemena and Wolff2005), which concurs with our findings. CRs had a much higher coverage of branching coral than CCs, which we expected because these corals form the basic structure of CRs in the LPA. Branching coral cover helps explain the pattern of clustering we observed in the CR fish ordinations and supports previous work showing its importance (Bouchon-Navaro & Bouchon, Reference Bouchon-Navaro and Bouchon1989; Chabanet et al., Reference Chabanet, Ralambindrainy, Amanieu, Faure and Galzin1997; Connell & Kingsford, Reference Connell and Kingsford1998; Jones et al., Reference Jones, McCormick, Srinivasan and Eagle2004; Bozec et al., Reference Bozec, Dolédec and Kulbick2005; Wilson et al., Reference Wilson, Graham, Pratchett, Jones and Polunin2006, Reference Wilson, Burgess, Cheal, Emslie, Fisher, Miller, Polunin and Sweatman2008). This parameter also correlated negatively with invertivores within coral reef sites, which contradicts findings by Connell & Kingsford (Reference Connell and Kingsford1998). We did find a positive relationship between live branching coral cover and the Serranidae, in agreement with Connell & Kingsford (Reference Connell and Kingsford1998). In the current research, members of the Serranidae associated closely with branching corals in both habitats and used these structures as refuge sites (e.g. Epinephelus analogus and Serranus psittacinus). Herbivores also were positively correlated with live branching coral cover within coral community sites, and a similar pattern was observed by Öhman & Rajasuriya (Reference Öhman and Rajasuriya1998). Filamentous algae grow on and between the colonies and branches of the coral provide a food source (Öhman & Rajasuriya, Reference Öhman and Rajasuriya1998) and many of the herbivorous members of the Pomacentridae, which were common in this habitat, utilize branching coral for shelter (Wellington, Reference Wellington1982; Öhman & Rajasuriya, Reference Öhman and Rajasuriya1998).
The percentage cover of live massive coral was important in explaining the difference in community structure between CRs and CCs and explained some of the observed differences in species length composition in CR sites and trophic group composition within CCs. Massive corals generally were more abundant in CCs as their colonization and growth are not as restricted compared to CRs where they are limited to deeper areas because of the dominance of branching corals (Glynn & Maté, Reference Glynn and Maté1996; Maté, Reference Maté and Cortés2003). During the surveys we found different species (e.g. Microspathodon dorsalis, Holacanthus passer, Abudefduf troschelii and members of Balistidae) aggregating around the massive coral species compared to those generally found around the Pocillopora coral framework. This is supportive of the work of Dominici-Arosemena & Wolff (Reference Dominici-Arosemena and Wolff2005) who found the presence of massive corals was related to higher levels of species richness and diversity. Thus, massive corals, especially when large (>1 m diameter) increase the variety of growth forms and provide areas of topographical complexity in a landscape of low variability, which attract fish. Massive corals provide microhabitats that offer refuge and food (e.g. crabs, tunicates, bivalves and brittlestars; Glynn & Maté, Reference Glynn and Maté1996) that boost fish species richness and alter the abundance of certain groups, e.g. Balistidae, invertivores and omnivores. The positive relationship between Tetradontidae and the abundance of massive corals can be explained by the preference of this family for this food resource (Guzman & Robertson, Reference Guzman and Robertson1989). In our study, the species found associated with the massive corals were generally larger which would account for the correlation between massive corals and the species length ordination for coral community sites.
We also found that benthic diversity, macroalgal cover, and CCA cover showed correlations in the MDS ordinations and univariate methods. Previous studies have also reported positive correlations between benthic diversity and fish diversity, species richness, and abundance on coral and rocky reefs (Roberts & Ormond, Reference Roberts and Ormond1987; Chabanet et al., Reference Chabanet, Ralambindrainy, Amanieu, Faure and Galzin1997; Öhman & Rajasuriya, Reference Öhman and Rajasuriya1998; García-Charton & Pérez-Ruzafa, Reference García-Charton and Pérez-Ruzafa2001). Sites with higher benthic diversity offer a wider variety of habitats and resources to meet the needs of more species. The negative relationship between algae cover and herbivore abundance in CCs found in this study is similar to that reported by Williams & Polunin (Reference Williams and Polunin2001), but it contradicts other results (Chabanet et al., Reference Chabanet, Ralambindrainy, Amanieu, Faure and Galzin1997; Friedlander & Parrish, Reference Friedlander and Parrish1998).
Crustose corralline algae was found to correlate positively with several fish parameters and families in CCs with Pomacentridae and herbivore abundance on CRs. CCA is a food resource for herbivores, such as the Scaridae, which scrape it from substrate surfaces (Williams & Polunin, Reference Williams and Polunin2001). Correlations with the Pomacentridae on CRs can be attributed to space: members of the Pomacentridae family guard algal territories, where there is potential for macroalgae and CCA to take hold where coral is unable to colonize. In the CCs, CCA cover was significantly higher in more topographically complex sites and therefore, some of the correlations detected for CCs may relate to topography.
CONCLUSION
Our results suggest that the structure reef fish communities in the LPA are determined more by the large scale structural features that characterize the two habitats (e.g. topographical complexity, exposure and branching coral cover) than by those that vary over small scales (i.e. the size of a transect). However, we found small-scale differences, e.g. in massive coral cover, had a localized impact on fish communities. Based on these findings and the dominance of CCs in the LPA (Benfield et al., Reference Benfield, Guzman, Mair and Young2007; Guzman et al., Reference Guzman, Benfield, Breedy and Mair2008), we conclude that the contribution of tropical rocky reefs to fish diversity should not be overlooked, especially in biogeographical regions where CRs are impoverished.
ACKNOWLEDGEMENTS
Funding for this research was provided by the Engineering and Physical Sciences Research Council, Department for Environment, Food and Rural Affairs' Darwin Initiative Fund, Smithsonian Tropical Research Institute, and the Heriot-Watt University Alumni Fund. The authors thank C. Guevara, A. Lam, and the crew of the R/V ‘Urraca’ who assisted the authors during field work. We thank the Government of Panama for giving permission to work in the archipelago and the anonymous referees of this paper who provided constructive comments.
