A novel habitat-based approach for combining indices of abundance from multiple fishery-independent video surveys

Abstract

Three surveys in the eastern Gulf of Mexico use baited remote underwater video systems to assess changes in abundance of reef fishes: the National Marine Fisheries Service (NMFS) Southeast Area and Monitoring and Assessment Program Reef Fish Video Survey, the NMFS Panama City (Florida) survey, and the Fish and Wildlife Research Institute of the Florida Fish and Wildlife Conservation Commission survey. These surveys use similar sampling gear and video-processing protocols, but they vary in spatial extent and habitats sampled. Each survey has been used individually to produce indices of relative abundance to assess various reef fish, but species trends may vary across surveys, possibly making subsequent assessment models more complex. A combined index could yield a more representative and statistically powerful characterization of the relative abundances of commercially important species. We developed a method for combining video count data from these surveys for managed reef-fish species into a combined index for the eastern Gulf using habitat data in classification and regression trees (CART) and general linear models (GLMs). CART results indicated that several site-specific and landscape-level habitat variables could be used to predict site occupancy of target species. We then used the CART-derived habitat groups as a variable shared among surveys in fitting a GLM to catch data to derive estimated annual abundances. We evaluated models’ potential and utility for a single estimated relative-abundance index for key managed reef species in the region compared to a suite of alternative GLMs of less complexity. Models that incorporated habitat covariates across the surveys showed better fits than models that did not incorporate habitat information. We also developed model-weighting methods that allowed us to account for the variation in spatial footprint in the surveys when combining data, allowing for what is likely a more representative index of regional relative abundance trends. Our results indicated that the data can be reliably combined into a single index. These methods should be evaluated for similar instances of combining survey data in other species, ecosystems, and management frameworks.