Proceedings of the 11 th International Coral Reef Symposium, Ft. Lauderdale, Florida, 7-11 July 2008 Session number 17 Remote sensing for studies of the spatial distribution of coral reef fishes A. Knudby 1 , C.M. Newman 1 , E. LeDrew 1 1) University of Waterloo, 200 University Avenue West, Waterloo, Ontario, N2L3G1, Canada Abstract. Reef fish biodiversity is influenced by habitat, including such variables as coral cover, depth, rugosity, and the distance to the reef edge. Commercially available satellite imagery can be used to map these habitat variables, and fish biodiversity can be estimated using the resulting habitat maps. We present a case study from two reefs in Zanzibar, Tanzania, based on IKONOS imagery acquired concurrently with habitat and fish surveys. The influence of some habitat variables, difficult or impossible to measure in-situ but mappable using satellite imagery, such as habitat diversity and depth variability at a range of spatial scales, are also explored. We illustrate how well each habitat variable can be estimated using remote sensing, and how accurately a variety of models can predict the spatial distribution of reef fish biodiversity. High-resolution satellite imagery can map species richness with a Residual Standard Error of <5 species at the study site. Future refinement of habitat maps and of predictive models is expected to reduce prediction error. Key words: Remote sensing, Habitat mapping, Reef fish biodiversity, Predictive models. Introduction The influence of habitat on reef fish diversity has been demonstrated in experimental and observational studies (Roberts and Ormond 1987; Friedlander and Parrish 1998; Gratwicke and Speight 2005). Habitat variables found to influence fish diversity typically include live coral cover (henceforth: coral cover), depth, rugosity, and habitat heterogeneity (Huston 1994; Chabanet et al. 1997; Jones et al. 2004). The predictive strength of these variables, individually or combined, has varied between studies (Jones and Syms 1998; Knudby et al. 2007). It is therefore prudent that any relationships be establish locally before they feed into management decisions, e.g. designation of ‘high-diversity habitat’ as protected. We established habitat-fish relationships using in- situ data on a range of habitat and fish diversity variables, from two reefs in Zanzibar, one protected (Chumbe) and one un-protected (Bawe) (Fig. 1). We then assessed the feasibility of predicting spatial variation in fish diversity using remotely sensed estimations of the most important habitat variables. Material and Methods Three data sets were used for this study. Fish point counts were made at 93 sites on Chumbe and 51 sites on Bawe. Sites were located at a random number of fin kicks, in a random direction, from the previous site. Habitat data were also collected at these sites, as well as at 347 additional sites on Chumbe and 56 sites on Bawe, covering all major habitat types. IKONOS satellite data, from 2007 for Chumbe and 2005 for Bawe, covered both reefs. Figure 1: Study sites. The reef around Chumbe Island is effectively protected from direct human damage; the reef around Bawe Island is fished and used for dive and snorkel tourism. Fish point counts Fish point counts were carried out in 5m radius circles according to Bohnsack and Bannerot (1986). The location of each site was found by snorkeling in a random direction for a random number of fin-kicks from the previous site. Abundance and average length of all species were recorded. 631