Accounting for spatial scale and temporal variation in fish-habitat analyses using baited remote underwater video stations (BRUVS)

Abstract

With the implementation of marine spatial planning in many coastal regions of the world, there is a need to understand how marine species and communities respond to environmental heterogeneity. Predictive modelling approaches are one efficient method for associating marine communities with variations across the seascape. These approaches, along with increasing access to spatially explicit environmental data, provide improved opportunities for modelling fish assemblages. Baited remote underwater video stations (BRUVS) are a popular means of gathering fish assemblage data in the coastal zone, but have biases in bait attraction, trophic groups sampled, and behavioral conditions. To account for these biases, spatial and temporal scales of analyses must be considered. In this study, we combined time-series BRUVS observations with seafloor and oceanographic variables in generalized additive models to model patterns of relative species richness and abundance in temperate coastal fish assemblages across multiple habitat types, functional trophic groups, and spatial scales from 5-500 m. We show that the spatial and temporal scale of analyses and behavioral characteristics of target species (such as mobility) are important considerations when predicting the spatial distribution of a particular assemblage or functional subset. The resulting models performed well, with prediction accuracies up to 79% while explaining between 24 and 83% of variance. These models were then used to extrapolate assemblage characteristics over broader areas of the seafloor to expand our understanding of fish distributions, providing valuable insights for marine spatial planning, including marine protected area assessment.