High-throughput video and acoustic imaging from seafloor cabled observatories for benthic ecosystem monitoring in coastal and deep-sea settings

Ocean Networks Canada (ONC) operates large seafloor cabled observatories in the Arctic, Atlantic and Pacific, with some of its long-term observations surpassing 16 years. We present snapshot results from long-term video time-series observations and in-situ experiments studying the benthic boundary layer in two coastal and one continental margin setting of Canada¿s Atlantic and Pacific Oceans. Using video imagery spanning for 7 years (2013-2020) we studied the deep-sea pink urchin Strongylocentrotus fragilis with respect to the expanding oxygen minimum zone in Barkley Canyon (420 m), NE Pacific. In a second case study, we analyzed 6 months of hourly videos from the newly installed Holyrood observatory in Conception Bay, Newfoundland, Atlantic, to investigate benthic-pelagic coupling following the onset of the 2021 spring bloom. From a series of short-term experiments, we combined video and acoustic imagery (dual-frequency identification sonars) and passive acoustics data to better understand poorly understood fish vocalizations, overall temporal changes in benthic abundance and diversity, and behavioural responses to artificial lighting. In a first experiment, in turbid waters of the Fraser River Delta (150 m), Strait of Georgia, the acoustic camera proved to be the most efficient device for measuring faunal densities, while the video was more efficient in detecting a moderately diverse assemblage of fish and invertebrates. Light avoidance behaviour was detected in a large number of species while light attraction was verified for the spotted ratfish Hydrolagus colliei. In the second and third experiments, deployed at 640 m depth adjacent to Barkley Canyon, sequential bait-introduction was employed for the study of benthic successional patterns of deep-sea scavenger communities under limiting dissolved oxygen conditions. Lastly, we present an example of machine learning using a deep learning neural network applied to the automatic detection of commercially harvested s