Detection of predator dive patterns from stationary echosounder data

Standard protocols for processing of echosounder data are geared towards the estimation of biomass indices of pelagic organisms. More transient visitors to the watercolumn however also give echoes, and the data provided by the echosounders may be used to assess the behaviour of these organisms, as well as their interactions with the pelagic residents. Using freely available software and neural networks developed for general image recognition, we detected diving events by air-breathing predators in a 1 year dataset from an acoustic lander located close to the South Orkneys. The analysis documented clear patterns in the dive data with regards to time of year as well as time of day. At the temporal scales analysed the data did not suggest a positive correlation between backscatter of pelagic aggregations and the detected diving activity at this location.