Habitat Associations and Aggregation of Recruit Fishes on Hawaiian Coral Reefs

Associations with structural habitat complexity and the aggregation of individuals lessen the risk of predation, and both are commonly observed for recently settled fishes on coral reefs. On reefs fringing Hawaii Island, such recruits of many fishes, particularly two species of surgeonfishes, exhibited strong affinities for branching finger coral (Porites compressa Dana, 1846), the most structurally complex coral available in the relatively low-diversity assemblage of Hawaiian reef corals. Recruits of these species also aggregated with conspecific and heterospecific recruits as they associated with this coral. That recruits of reef fishes take refuge in the best available shelter, however, is not a novel observation. More noteworthy is the frequent co-occurrence of recruits with other recruits of the same and other species. Our observations suggest that small juvenile reef fishes aggregate near preferred coral habitat and perhaps also with one another independently of habitat per se, the latter a phenomenon that has been generally overlooked by researchers focusing on evidence for density-dependent negative interactions within juvenile reef fishes. Because interactions between habitat structure and aggregative behaviors have the potential to importantly modify the survivorship of recruits that may influence year-class strength, they merit further study. The distributions of adult reef fishes, like the adults of most other organisms, are primarily related to the distribution of food and mating opportunities. The distributions of juvenile reef fishes, however, are more proximally related to the distribution of structurally complex habitats and other factors that directly influence survival under intense predation (Shulman, 1985; Hixon, 1991; Caley and St. John, 1996; Nemeth, 1998; Almany and Webster, 2006). Safe havens from predators provided by habitat features have been documented by many studies (e.g., Carr and Hixon, 1995; Beukers and Jones, 1997; Carr et al., 2002; Webster, 2002), and the anti-predator benefits of schools and less structured aggregations of individuals are also clear (Hobson, 1978; Webster and Almany, 2002; Sandin and Pacala, 2005). Surprisingly few studies, however, have described the undoubtedly complex interplay between refuging and group behaviors for organisms in general (Krause and Ruxton, 2002) and for juvenile reef fishes in particular. Numerous studies have evaluated relations between the distribution, abundance, and diversity of fishes and live coral cover on shallow tropical reefs (reviewed by Wilson et al., 2006), often with an emphasis on relations of recently settled recruit stages (Tolimieri, 1998; Jones et al., 2004). It is generally accepted that structurally complex coral reefs provide biologically important physical structure in the form of shelter holes and habitat for prey including small fishes (Caley and St. John, 1996). In this study, we describe the interrelated sheltering and aggregating behavior of recruits (benthic juveniles < 2 mo old) of fishes commonly encountered on the fringing coral reefs of the Big Island of Hawaii (Hawaii Island) in the Main Hawaiian Islands (MHI), along with implications for reef fishes throughout the archipelago and elsewhere. BULLETIN OF MARINE SCIENCE, VOL. 81, NO. 1, 2007 140 Materials and Methods Study Area.—Surveys of fishes and habitats were conducted using SCUBA in 8–13 m depths at three locations spanning 25 km, each within extensive continuous fringing reefs and representative of wave exposures on the leeward coast of northwest Hawaii Island: Point at End-of-the-Road, Puako Beach Drive (Location 1: exposed), Holoholokai Beach Park (Location 2: partially exposed), Mahukona (Location 3: protected). Both recruit fishes and habitat were surveyed during a 19-d period from 23 May to 10 June 2005. Supplemental surveys of recruit-habitat associations were conducted biweekly to monthly at Location 3 through midOctober 2005. Survey Protocols.—The densities of recruits and other small-bodied fishes were estimated using 50-m2 belt transects (25-m long × 2-m wide) within 1.0–1.5 ha areas at each location. The starting points and bearings of transects were chosen haphazardly following minimal criteria (> 50% consolidated substratum and a constant depth ± 1 m). Eight to ten transects were surveyed per location. Fish were tallied by species or lowest taxon and size (1-cm total length [TL] for individuals ≤ 10 cm TL) encountered as the transect line was laid. Fishes were classified as either recruits (≤ 5 cm TL) or as older (≥ 6 cm TL) juveniles. Two acanthurids—the goldring surgeonfish, Ctenochaetus strigosus (Bennett, 1828), endemic to the Hawaiian Islands and Johnston Atoll (Randall and Clements, 2001), and the yellow tang, Zebrasoma flavescens (Bennett, 1828), a species broadly distributed throughout the tropical central Pacific—were targeted for special consideration because of their abundance at the time surveys were conducted. Each species settles from the plankton at ~3 cm TL beginning in late May–early June and continuing throughout the summer on leeward Hawaii Island (Walsh, 1987; E. DeMartini, unpubl. data). To assess coral habitat, the percentage cover of major coral taxa and other predominant substratum types were recorded using 1-m2 quadrats. The major corals included four taxa ranging in structural complexity from prostrate (lobe coral, Porites lobata Dana, 1846 and Montipora spp.), to digitate (finger coral, Porites compressa Dana, 1846) and discrete (cauliflower coral, Pocillopora meandrina Dana, 1846). The two additional major substratum types were coral limestone with short (< 1-cm high) turf algae, and sand. Percentage cover of each of the six substrata was estimated visually; accuracy was evaluated previously by comparing visual estimates with a random-point-contact method (Dethier et al., 1993). Within each quadrat, a rugosity index ratio of conforming-to-straight line distance (Risk, 1972; Andrews and Anderson, 2004) was determined using a 1.25 cm-link chain laid over the substratum. Three 1-m2 reference quadrats, positioned randomly along each transect line, and six 1-m2 target quadrats, centered on recruits of target species haphazardly encountered within 5 m of each transect line, were surveyed for coral cover and rugosity. Prior to the placement of target quadrats, individual recruits were sighted from a distance of 1.5–3 m. Once recruits were sighted, the species, number, and size (cm TL) of fish were recorded, along with the distance from the target individual to the closest substratum type (cm), and the median distance (cm) to other recruits in a “group” (defined as one or more individuals ≤ 10 cm TL of the same or other species present within 10 cm of the target individual). Any aggressive acts also were noted. Target quadrats were then placed over the substratum, centered on the exact point at which the recruit(s) were first sighted; 1-m2 quadrats were sufficient in area because recruit ambits were typically ≤ 50 cm. Additional data on proximity to substratum type and median distance among aggregated fishes were collected on supplemental dives dedicated to finding individuals of target species both shoreward (to 5-m depth) and seaward (to 16 m) of transected areas at Location 3. The latter dives were necessary to augment the number of recruit sightings. All estimates of distances to substratum and among recruits were made as previously described. Supplemental observations were limited to one site (Location 3) because of limited personnel and easier access. Statistical Analysis.—Standard parametric (Sokal and Rohlf, 1981) and nonparametric (Siegel and Castellan, 1988) tests were used, with choice based on the metric and its caseDEMARTINI AND ANDERSON: HABITAT RELATIONS AND AGGREGATIONS OF FISH RECRUITS 141 specific sampling distribution. Variance heterogeneity was evaluated using Cochran’s C test (Underwood, 1997). Except where noted, the variances in densities of fish became homogeneous after log-transformation. Recruit-habitat relations and recruit aggregations were evaluated using several categorical analyses. Contingency, goodness-of-fit, and heterogeneity Chi-square and G-tests (Sokal and Rohlf, 1981) were used, as appropriate, if contrasts lacked a response variable. Logistic regression was used if a response variable was clearly distinguishable among the explanatory variables (Allison, 1999). We assumed that the expected numbers of singleton and grouped recruits followed a Poisson distribution, and we compared their observed frequencies against this expectation. Poisson regression (Allison, 1999; proc GENMOD: SAS, 2000) was used to test whether the number of recruits in a group was predictably related to the proximity of the substratum. Sokal and Rohlf ’s (1981, p 779) statistic was used for combining probabilities resulting from analogous tests. The Bonferroni correction (pcrit = 100 α/m %), where m = number of comparisons; Manly, 1991) was used to adjust P-values wherever multiple testing was unavoidable. Interaction effects in factorial ANOVAs and Gtests were not listed if non-significant.