Spatial patterns in abundance of a damselfish reflect availability of ...

Oecologia (2000) 122:109–120

© Springer-Verlag 2000

Sally J. Holbrook · Graham E. Forrester Russell J. Schmitt

Spatial patterns in abundance of a damselfish reflect availability of suitable habitat

Received: 18 May 1999 / Accepted: 9 January 1999

Abstract For species with metapopulation structures, variation in abundance among patches can arise from variation in the input rate of colonists. For reef fishes, variability in larval supply frequently is invoked as a major determinant of spatial patterns. We examined the extent to which spatial variation in the amount of suitable habitat predicted variation in the abundance of the damselfish Dascyllus aruanus, an abundant planktivore that occupies live, branched coral throughout the Indo-Pacific. Reef surveys established that size, branching structure and location (proximity to sand) of the coral colonies together determined the “suitability” of microhabitats for different ontogenetic stages of D. aruanus. Once these criteria were known, patterns of habitat use were quantified within lagoons of five Pacific islands. Availability of suitable habitat generally was an excellent predictor of density, and patterns were qualitatively consistent at several spatial scales, including among different lagoons on the same island, among different islands and between the central (French Polynesia and Rarotonga) and western (Great Barrier Reef, Australia) South Pacific. A field experiment that varied the amount of suitable coral among local plots indicated that habitat for settlers accounted for almost all of the spatial variation in the number of D. aruanus that settled at that location, suggesting that spatial patterns of abundance can be established at settlement without spatial variation in larval supply. Surveys of four other species of reef-associated fish revealed that a substantial fraction of their spatial variation in density also was explained by availability of suitable reef habitat, suggesting that habitat may be a prevalent determinant of spatial patterns. The results underscore the critical need to identify accurately the resource requirements of different species and life stages when S.J. Holbrook (✉) · R.J. Schmitt Department of Ecology, Evolution and Marine Biology, and Coastal Research Center, Marine Science Institute, University of California, Santa Barbara, CA 93106, USA G.E. Forrester Department of Biological Sciences, University of Rhode Island, Kingston, RI 02881, USA

evaluating causes of spatial variation in abundance of reef fishes. Key words Habitat availability · Resource limitation · Recruitment limitation · Coral reefs · Damselfish

Introduction Many species have metapopulation structures where the abundance and dynamics of a sub-population are influenced by events that occur both within and external to the patch. Most marine reef organisms have spatially sub-divided populations that are linked via planktonic dispersal of early developmental stages, and the causes of spatial variation in local abundance of such species have been examined extensively (Connell 1985; Doherty and Williams 1988; Mapstone and Fowler 1988; Menge 1991; Grosberg and Levitan 1992; Olafsson et al. 1994; Booth and Brosnan 1995; Caley et al. 1996; Chesson 1998a,1998b). It is well appreciated that the abundance in a patch is set by the balance between the externally derived input and subsequent losses (Robertson 1988; Underwood and Fairweather 1989; Doherty 1991; Jones 1991; Grosberg and Levitan 1992; Doherty and Fowler 1994; Forrester 1995; Schmitt and Holbrook 1996, 1999a,1999b; Steele 1997). With respect to reef fishes, considerable debate continues concerning the extent to which post-settlement losses may be density-dependent (Caley et al. 1996). However, it is widely believed that much of the variation in the input rate among patches, which often can be large, results from spatial and temporal variability in “larval supply” (Doherty 1991; Doherty and Fowler 1994). A substantial body of work has revealed that several attributes of reef fish assemblages also can be influenced greatly by variation in habitat features. The distribution of species and the composition and diversity of local assemblages can reflect spatial patterns in certain reef attributes (Hixon and Beets 1989; Holbrook et al. 1990, 1994; Wellington 1992; Caley and St. John 1996;

110

Munday et al. 1997; Ault and Johnson 1998; Gutierrez 1998). Fewer studies have explored the degree to which patterns of local abundance reflect variation in habitat features (Carr 1994; Ault and Johnson 1998), although several mechanisms could result in a causal relationship. For example, a larva may need an environmental cue to settle (Sweatman 1988; Elliott et al. 1995), and spatial variation in the intensity of such a cue can alter input rates to a patch (Sweatman 1985; Booth and Wellington 1998). Early survivorship can be enhanced if appropriate nursery or shelter habitat is available (Sale and Dybdahl 1975; Sale et al. 1984; Eckert 1985; Tolimieri 1995; Holbrook and Schmitt 1997; Risk 1997; Ohman et al. 1998), and the amount of such habitat can vary substantially in space (Carr 1994; Nemeth 1998). Finally, older stages of many species also can exhibit strong habitat preferences (Sale 1972; Holbrook et al. 1990; Schmitt and Holbrook 1990; Ault and Johnson 1998), and movement after settlement can result in positive relationships between habitat availability and adult abundance. These findings are consistent with the idea that demographic rates can be affected by the availability of suitable habitat, which in turn can influence local patterns of abundance. When examined in survey studies, the amount of variation in the abundance of fish among patches explained by habitat availability has ranged from extremely high (e.g., Carr 1994; Ault and Johnson 1998) to relatively little (e.g., Ault and Johnson 1998). Drawing inferences from the latter outcome can be particularly problematic. There are numerous factors (e.g., larval supply, predator density) that can decouple a relationship between variation in structural aspects of the local environment and abundance, and clearly the factors that set abundance may not involve quantitative variation in a habitat attribute. However, it also could arise if patterns were estimated using incorrect criteria or an inappropriate spatial scale. For example, reef habitats frequently are classified either into location categories (e.g., fore-reef, reef crest) or into relatively broad microhabitat groupings (e.g., live coral, rubble, sand). Such classification schemes could obscure existing relationships for species with quite specific microhabitat requirements. Further, it is possible that the suitability of a given microhabitat is context-specific, such as when a species inhabits a particular microhabitat only when it occurs in a particular reef area (for examples, see Elliott et al. 1995; Doherty et al. 1996). Here we explore the extent to which spatial variation in densities of a damselfish, Dascyllus aruanus (humbug dascyllus), can be explained by availability of suitable habitat. D. aruanus is a reef-associated planktivore in one of the most speciose families of reef fish (Pomacentridae), which are small-bodied reef fish that have been the focus of numerous ecological studies (Sale 1991; recent reviews by Booth and Brosnan 1995; Booth and Wellington 1998; Caley et al. 1996). Like many other species of reef fish, D. aruanus inhabits specific types of microhabitats (live heads of branched coral) that serve as shelter from predators (Sale 1971, 1972). Further, D.

aruanus is widespread throughout the Indo-Pacific (Allen 1991), providing an opportunity to examine the consistency in habitat relationships among different portions of its geographical range and across several spatial scales. Our study involved several steps. First, reef surveys were conducted to establish precisely what constituted “suitable” microhabitat for different ontogenetic stages of D. aruanus. Subsequently, additional survey data were collected to determine the extent to which spatial patterns in abundance of various life stages could be predicted by the amount of microhabitat that was classified as suitable. Microhabitat-fish abundance relationships were explored across several spatial scales (e.g., among transects, lagoons, islands, geographic areas). Third, a field experiment was done to assess whether microhabitat-fish abundance relationships could arise from settlement patterns in the absence of substantial spatial variation in larval supply. Finally, we conducted complementary studies on several other species of reef fish to assess the generality of results obtained for D. aruanus. Relationships between the availability of preferred substrate and spatial patterns in abundance were established for several species with life histories similar to and different from D. aruanus. These included the damselfishes D. flavicaudus (yellow-tail dascyllus), D. trimaculatus (three-spot dascyllus), and Stegastes planifrons (threespot damselfish), and the arceye hawkfish Paracirrhites arcatus (Family Cirrhitidae). Repeated surveys of D. aruanus at one location and S. planifrons at another also allowed us to determine whether observed spatial relationships for these species were temporally consistent.

Methods Study areas Sampling of fish and habitats took place in lagoons of five islands in the South Pacific and on the fringing reef of one island in the Caribbean. Three of the islands – Heron Island (23°27’S, 151°55’E), One Tree Island (23°30’S, 152°06’E), and Lady Elliot Island (24°307’S, 152°43’E) – are on the southern Great Barrier Reef, Australia. The other two Pacific Islands were Rarotonga (21°14’S, 159°46’W) in the Cook Islands, and Moorea, French Polynesia (17°30’S, 149°50’W). The Caribbean island was Guana Island, British Virgin Islands (18°29’N, 64°35’W). Water depths were 10 m or less in all areas sampled. Spatial patterns of occurrence of D. aruanus: coral morphology, size and location Our own observations and those of other workers (Sale 1972; Shpigel 1982; Sweatman 1985; Forrester 1990) indicated that D. aruanus occurs in fairly stable social groups (with up to 80 individuals, but typically less than 10) in coral within shallow, sheltered lagoon areas. Occupied corals usually have a branched morphology or some other shape that provides a network of crevices within which the fish shelter when threatened. These general observations were not sufficient for the precise assessment of what constitutes suitable habitat for the species, so we conducted surveys at Heron Island to address this issue. A total of 38 band transects, each 100×2 m, were established in seven different locations

111 in the lagoon, from nearshore and mid-lagoon (where patch reefs were common) to pavement areas behind the reef crest. All corals encountered on each transect were categorized by divers according to morphology, colony size, presence of D. aruanus, and surrounding substrate. In the surveys each coral was examined to determine whether it was occupied by D. aruanus. It was then assigned to one of six morphological types, including (1) mound (=massive of Veron 1986, p. 60), (2) fine-branched (=branching of Veron, with small spaces between branches), (3) coarse-branched (=branching of Veron, with large spaces between branches), (4) lobed (similar to columnar of Veron), (5) coarse-branched-lobed (intermediate between types 3 and 4), and (6) plate (=laminar of Veron). Colonies were categorized according to size along their longest dimension (5 m) as well as surrounding substrate (on sand, continuous reef pavement, or small patch reefs). D. aruanus occupied four of the six morphological types of corals. To quantify differences in the structure of these four types, we measured attributes related to overall size of the colony and quantity of space available for fish to seek refuge between the branches or lobes. A total of 146 colonies representing the four types (fine-branched, n=38; coarse-branched, n=73; lobed, n=9; coarse-branched-lobed, n=26) was measured. The measurements included height, diameter and circumference of the colony, crevice depth (maximum depth of the crevices between branches or lobes), distances between adjacent branch or lobe tips (n=10 measurements per colony), and diameter of branches (n=5 per colony). All of the corals were occupied by D. aruanus at the time of measurement. Discriminant analysis explored whether the morphological categories we had defined reflected quantifiable differences in shape and branching pattern.

particular morphological type (fine-branched, coarse-branched, and so on), and the percent of the colony that was alive was estimated. The number (if any) and sizes of D. aruanus occupying each coral were determined. The few D. aruanus not clearly associated with a coral colony were noted. These individuals all were adults on small patch reefs. Size classes used for analysis were recruits (40 mm SL). Recruits were almost certainly young-of-the-year (Forrester 1990). In total, 132 transects were included in the analyses (One Tree 1987 n=33; One Tree 1993 n=18; Heron n=26; Lady Elliot n=11; Raratonga n=18; Moorea Teavaro n=13; Moorea Tiahura n=13). Patterns of occupancy of different corals by each of the size classes of D. aruanus were quantified. Relationships between habitat features and density of fish were explored with linear multiple regression models (transects as replicates). These models were constructed to explain the density of a size class of fish using the number of colonies and volume of each of the four morphological types of coral habitat as predictors (fine-branched, coarsebranched, lobed, coarse-branched-lobed). Analyses reported here considered all of the islands with transects as replicates, with location (island or lagoon) as a categorical variable. Since the full models for all of the Pacific data included some location effects, additional regression analyses were performed for each island (or lagoon) separately. Multiple regression models were constructed in a forward stepwise fashion to generate the simplest models that could reliably predict fish densities. The insight gained from our preliminary data on habitat use, and our previous experience with this species, were used in the selection of habitat attributes to add to the models, yielding models that were interpretable biologically. Experimental test of the effect of habitat availability on settlement

Patterns of occupancy of branched corals by D. aruanus The coral surveys revealed the sizes, morphologies and locations of coral used by D. aruanus and yielded an explicit, operational definition of suitable habitat. This information enabled us to explore patterns of occupancy of branched corals. First, the proportion of all (live) branched corals on each transect (each 200 m2 in area) occupied by D. aruanus was determined, and these values were regressed against the proportion of corals on each transect deemed suitable by size and location criteria. Second, the proportion of suitable habitat occupied on each transect was regressed against the number of suitable corals on the transect to determine whether the intensity of occupation varied with habitat availability. Proportions were angular transformed for statistical analyses. Relationships between habitat availability and abundance of D. aruanus by life stage Ontogenetic patterns of habitat use in six lagoons were quantified in 1993 at two lagoon areas in Moorea (Tiahura on the west coast and Teavaro on the east coast), and one lagoon on each of four other islands (Heron, One Tree, Lady Elliot, and Rarotonga). One Tree Island was sampled in 1987 and 1993. Up to 13 sites were selected within the lagoon(s) of each island and within each site counts were made on two to six transects. Transects were rectangular, but were variable in size due to restrictions imposed by lagoon topography. Most transects were 10 m wide (range 5–15 m) and 85 m long (range 37–85 m). After marking the perimeter of the transect, it was searched systematically. All corals that were suitable habitat for D. aruanus, based on the criteria established in the previous surveys, were counted and measured. Volumes of corals were determined by assigning the coral to a shape (sphere, cylinder, ellipsoid, rectangle) at the time of sampling, and making measurements that allowed later calculation of volume with standard formulas. Corals were identified to genus. Commonly-occupied genera included Pocillopora, Acropora, Porites, and Seriatopora. Each coral was assigned to a

A field experiment tested the effect of habitat abundance on settlement in the absence of substantial spatial variation in larval flux. Habitat availability was manipulated by transplanting colonies of branched coral (Pocillopora sp.) to six 5×5 m reef areas that lacked coral in Viapahu Lagoon, Moorea. Between 1 and 14 live coral heads were transplanted during July 1995 to each plot by attaching them with marine epoxy to cinder blocks. Corals were measured and their volumes calculated; the range in volume mirrored that obtained among the transects in the Tiahura lagoon at Moorea. Six additional sites served as controls; these lacked coral (natural or transplanted). The plots were arranged in a two by six grid with 5 m spacing between adjacent plots. The highly localized configuration of this experiment was intended to minimize the likelihood of gross spatial differences in larval supply. Groups of D. aruanus (as well as several other species of coral-dwelling fish including D. flavicaudus) became established on the outplanted corals via settlement from the plankton. One year after the plots were established, daily settlement of D. aruanus was assessed each morning during a settlement pulse. After the week-long settlement event, the total number of settlers to each coral and plot was calculated and plotwise data were regressed against total volume of live coral. Studies of variation in abundance of other species The relationship between habitat availability and abundance for four additional species of reef-associated fish was explored. Two were congeners of D. aruanus (D. trimaculatus and D. flavicaudus) and so are relatively similar in several facets of their lifehistory and ecology. Notably, both are planktivorous and live in stable social groups for at least part of their lifespan. By contrast, Stegastes planifrons (Family Pomacentridae) is an aggressive territorial herbivore, and Paracirrhites arcatus (Family Cirrhitidae) is a benthic sit-and-wait predator. S. planifrons was studied on Guana Island and the other three species were investigated in lagoons on Moorea. For each species,

112 band transects were established at a range of sites, and fish and their habitats were enumerated. D. trimaculatus settles to sea anemones and associates with them until adulthood, at which point it becomes more mobile. Anemones were counted and measured on each transect (n=19 locations, four 40 m×2 m transects per location), to estimate habitat availability as anemone cover per square meter. D. flavicaudus occupies fine- and coarse-branched coral, and P. arcatus lives in Pocillopora sp. and Acropora sp.. The volume and number of these corals on each 10 m×5 m transect (n=22 for D. flavicaudus and n=13 for P. arcatus) were quantified as described previously. S. planifrons recruits are positively associated with colonies of Montastrea annularis, a massive coral with crevices that provide shelter (Tolimieri 1995; Booth and Beretta 1994), and they may also favor other live corals (Booth and Beretta 1994). As adults their microhabitat associations are less distinct, but in some areas adult territories are common on coral rubble or colonies of Acropora cervicornis or A. palmata (Itzkowitz 1977; Williams 1978). S. planifrons and reef attributes were enumerated each July from 1992 to 1996 at eight sites around Guana Island. All sites consisted of fringing reef with continuous hard substrate at 8–10 m depth. Counts were made on three to six replicate 1.5 m×30 m transects per site. S. planifrons were recorded as recruits if 40 mm SL. Percent cover of taxa of live coral plus other types of substrates were estimated on transects by line intercept. The type of coral, substrate or organism (sand, coral pavement, dead coral, coral rubble, algae, sponge, soft coral, hard coral, anemone) under a tape placed in the middle of the band transect was recorded every 0.25 m (120 points per transect) to give an estimate of cover of different habitat categories. Linear regression analyses explored the relationship between habitat availability and abundance of each species. For D. trimaculatus, D. flavicaudus and P. arcatus, we generated models predicting spatial patterns in the density of adult fishes using transects (or locations in the case of D. trimaculatus) as replicates. For S. planifrons, we had temporal as well as spatial information on patterns of abundance of fish and habitat features. These data spanned at least one generation (K. Clifton, personal communication), allowing a test of whether time-averaged density of recruits at each site was related to time-averaged cover of M. annularis. We also tested the relationships between density of adult S. planifrons in 1996 and the history of recruitment to the site (1992 to 1996) or the availability of preferred habitat (M. annularis). For these, transects at each site were pooled, and site means were used as replicates.

Table 1 Classification of corals to determine usable habitat. Given are the percent (and number) of coral colonies occupied by Dascyllus aruanus on Heron Island when corals were categorized by morphology. Total number of corals in the survey was 3675 (FB fine-branched, CB coarse-branched, CB-L coarse-branched-lobed) Coral morphology

%Occupied n

Massive

Plate

FB

CB

CB-L

Lobed

0 482

0 266

5.6 936

38.6 422

8.8 1063

0.2 503

Table 2 Classification of corals to determine usable habitat. Given are the percent (and number) of coral colonies occupied by D. aruanus on Heron Island when corals were categorized by location. Total number of corals in the survey was 3675 (FB finebranched, CB coarse-branched, CB-L coarse-branched-lobed) Location of corals Patch reef

Pavement

>5 m 5 m) patch reefs. Rather, they occurred mainly on corals located on or immediately adjacent to sand (Table 2). Further, the size of the coral colony mattered. Small corals (diameter