"Fish abundance and diversity" is one of ten Essential Biodiversity Variables (EBVs) used for global ocean assessments, recognizing their importance in trophic pathways in all marine systems. Selecting EBVs for coral reef ecosystems is challenging because of their high inherent variation, particularly for fishes. Here we measured the species diversity, density and size of fish species at reefs in Comoros, Madagascar, Mozambique, Tanzania and Chagos, together with 16 bio-physical variables for each site, to develop a model that tests the sensitivity of reef fish indicators and hence their potential as coral reef EBVs. The analytical approach used space rather than time to represent the wide range of reef conditions across the western Indian Ocean. We first modelled diversity, density and biomass of fish species to determine which bio-physical variables were significant in determining fish assemblages. Geographic location and reef geomorphology were the most significant drivers, explaining 32% of the observed variation in densities and biomass. Species richness also varied geographically with Comoros being the least diverse. Fishing effects only became significant, explaining 6% of variation in biomass when data were analysed at smaller spatial scales. Only in Chagos, with zero fishing pressure and relatively homogeneous geomorphology, did live coral cover correlate with fish assemblage structure. The second analysis used PCA to try to condense key benthic variables and found benthic attributes delineated clearly across two primary PCA axes (48% of total variation): i) from high turf algal cover to exposed sites with high CCA cover; ii) from high coral cover and rugosity to high fleshy algal cover and rubble. This provides a matrix for preliminary modelling which is now testing how responsive are different aggregate fish indicators to variation in environmental conditions and anthropogenic pressures. This paper shows that different geographic / morphological reef systems support different fish assemblages. The broad spatial approach enabled a wide range of benthic variables to be simplifed into two values per site and provides the basis for testing commonly used aggregate fish indicators. These results can contribute to the development of global fish EBVs and to biodiversity targets set by the CBD.