BULLETIN OF MARINE SCIENCE, 69(2): 745–758, 2001 MULTIVARIATE ANALYSIS OF REEF COMMUNITY STRUCTURE IN THE BELIZE BARRIER REEF COMPLEX M. D. McField, P. Hallock and W. C. Jaap ABSTRACT The Belize barrier reef complex includes approximately 250 km of barrier reef and three off-shelf atolls up to 40 km east of the barrier reef. Multivariate analysis techniques are being tested for their ability to discriminate patterns of geographical zonation in reef community structure. A stratified, haphazard video-based monitoring scheme has been established at 17 windward fore-reef sites. This preliminary community classification and ordination provides a baseline description of reef communities and indicates a sig- nificant difference (1) between atoll versus barrier reef sites, and (2) between southern, northern and central barrier reef sites, despite a relatively high degree of similarity. Two pairs of sites indicate there are significant differences in community structure between ‘impacted’ and ‘non-impacted’ sites. In addition, the key taxa responsible for these differ- ences (based on the Bray-Curtis similarity matrix) were identified and include Agaricia tenuifolia and macro algae. Most previous studies have examined stony coral species in such analyses. Because non-coral biota are normally the major community components on Caribbean reefs, they should be included in multivariate analyses of reef communi- ties. Reanalysis of these data using either major benthic substrate categories or only stony coral species produces less discriminating results and illustrates the importance of using a community approach with as much taxonomic detail as possible. Understanding the underlying similarities and differences in biological communities throughout this large reef area is the first step towards the goal of linking these patterns to environmental and management-linked influences. The Belize reef system includes the largest barrier reef in the Western Hemisphere (approximately 250 km), three well developed off-shelf atolls, numerous inner patch reefs and unusual rhomboid shaped reefs (faroes). While there have been major studies of the geology and morphology of Belize’s reefs and sediments (Stoddart, 1962; Stoddart 1963; and others), they have not provided a quantitative measure of reef community structure. Burke (1979) described seven transects from Gallows Point Reef to Queen Cayes (near Pompion Caye, see Fig. 1), recording bottom profiles, zonation of substrates and organ- isms, but giving few qualitative and no quantitative visual estimates of benthic commu- nity cover. Detailed taxonomic descriptions of reef biota and qualitative descriptions of reefs have been made near the Smithsonian Institution’s research station at Carrie Bow Cay (Rutzler and Macintyre, 1982). Aronson et al. (1994) and Aronson and Precht (1997) have also recorded quantitative community descriptions in the vicinity of Carrie Bow Caye. There have also been quantitative community descriptions in the vicinity of Wild- life Conservation Society’s research station on Glovers Reef Atoll (McClanahan and Muthiga, 1998) but little is published about the remainder (~90%) of the reef complex. An integrated coastal zone management program was initiated in 1991, which includes numerous measures aimed at coral reef conservation (Gibson et. al., 1998). Establish- ment of a long-term coral monitoring program is critical to the evaluation of these man- agement efforts and the health of the Belize reef system (McField et al., 1996). Our study represents the first reef-wide, quantitative description of reef community structure through- out the Belize Barrier Reef Complex (Fig. 1). The database created through this project, 745 746 BULLETIN OF MARINE SCIENCE, VOL. 69, NO. 2, 2001 including archived CD-ROM images, will serve as a baseline for future monitoring ef- forts. These data can be requested through the first author or through the Coastal Zone Management Institute, P.O. Box 1884, Belize City, Belize, Central America. Although numerous scientific investigations have been conducted in Belize, most have been iso- lated studies, with little continuity or comparison between projects, and with little col- laboration with national organizations involved in reef management (Wells, 1997). Thus there is an urgent need for scientific data to guide current management efforts in Belize and to evaluate the effectiveness of those efforts. This work applies a multivariate community approach to discriminating among geo- graphic regions, identifying key discriminant species, and, in the future, will be used to determine the relative importance of environmental and management-linked influences. It is hypothesized that reef communities along the 250 km barrier reef will be differenti- ated among northern, central, and southern sites (Fig. 1), based on a general zonation proposed by Burke (1979). Furthermore, the atoll sites are expected to have different communities than the barrier reef sites, due to the isolated, more oceanic locations of the atolls. Two sets of geographically paired sites are examined in this study: (1) Hol Chan (HolNbr) and Tackle box (TacNbr), off Ambergris Caye are separated by about 8 km, and (2) Gallows reef (GalCbr) and Goffs Caye (GofCbr), near Belize City are separated by approximately 15 km (Fig. 1). It is hypothesized that these geographically paired sites are naturally more similar to each other than to the other sites. An alternate hypothesis is that the two sites located in closest proximity to the major developed areas in Belize (TacNbr near the tourist resort of San Pedro and GalCbr near the major city, Belize City) would be more similar to each other than to their geographically paired sites if their biotic commu- nities were in fact ‘impacted’ by these developments. Lastly, this work compares different hierarchical (taxonomic) levels of analysis. It hypothesized that analyses based on the lowest taxonomic classification available, including different hierarchical levels for dif- ferent substrate categories, have the greatest ability to distinguish between geographic regions. METHODS The sampling scheme utilized a stratified (windward fore-reef spurs), haphazard video transect- based design, similar to that of several other current monitoring programs (Aronson et al., 1994; Wheaton et al., 1996). During the summer of 1997, 17 reef sites (Fig. 1) were assessed with 10 replicate 25 m transects conducted at each site, using the procedure described in Aronson, et al (1994). Transects were oriented along individual reef spurs near the reef slope in a depth range of 12 to 19 m. This relatively deep fore reef zone was selected because it is a zone of major reef development, it provides comparable communities throughout the reef system, and less is known about this reef zone. This zone is also critical for reef accretion and is also of primary management concern since tourist activities are focused in this area. Sites were selected to give full geographic coverage consistent with barrier reef regions as defined by Burke (1979) and Wantland and Pusey (1971). Hi-8 video was taken beside each transect line with the camera held 25 cm above the substrate with a scaled reference bar giving a swath width of approximately 25 cm. Two 30 w ultra- bright lights were used for illumination. This analysis includes the 12 sites for which data have been fully analyzed (Fig. 1). Additional data on species richness, rugosity and fish communities, were collected at each site but are not included in this analysis which is based on the multivariate abun- dance data (proportional cover). MCFIELD ET AL.: ANALYSIS OF REEF COMMUNITY STRUCTURE IN BELIZE 747 Figure 1. Location of Study Sites in Belize. 748 BULLETIN OF MARINE SCIENCE, VOL. 69, NO. 2, 2001 Image analysis followed protocols developed by the EPA Florida Keys National Marine Sanctu- ary monitoring project (Wheaton et al., 1996). Fifty video images per transect (500 per site) were frame-grabbed, processed (de-interlaced, sharpened, enhanced) and saved onto CD-ROM. These digital images are thus available for other analyses (sponges, algae, disease, etc.) and serve as a permanent archive. PointCount for Coral Reefs software developed by the Dustan Lab, University of Charleston (Wheaton et al., 1996) was used with Image Pro„ software to generate 10 random points on each image and to offer a menu selection of coral species and other community categories for each point. Organisms found under each point were then identified. When possible stony corals (Scleractinia and Milleporina) were identified to species. Unidentifiable specimens were called ‘Scleractinia’. The ‘macro algae’ category corresponds to algal specimens with blades large enough to be distinguished on the video. The ‘substrate’ category contains smaller turf algae, encrusting coralline algae and bare rock. Additional categories are ‘Porifera’, ‘Octocorallia’, and ‘sand’. Points that fell on the scale bar or transect line were called ‘equipment’ and those that could not be classi- fied within one of these groups were put in a ‘no count’ category. The percentage cover of each organism or category was calculated from the total 500 points per transect. Approximately 25 h of image analysis time was required to analyze each site (10 transects). The community data set, based on proportional cover of each species or category, was compiled into a matrix and imported into PRIMER ecological statistics software package (Clarke and Warwick, 1994; Clarke, 1993) for multivariate analysis. Raw biotic data values