Abstract
Knowledge of the vertical migration pattern of sea lice (Lepeophtheirus salmonis) copepodites is necessary for designing efficient measures to prevent lice infestations on farmed Atlantic salmon (Salmo Salar) in sea-cages. However, data can be challenging to acquire at a large scale under realistic circumstances without interfering with the natural behavior of the specimen. A mesocosm platform was built to help acquire this data consisting of a sensor package in an underwater housing being pulled up and down along a 11-meter-long transparent tube containing planktonic organisms while collecting image-, temperature- and spectrometer data. The platform was placed at a salmon farm and the acrylic tube was filled with L. salmonis copepodites and was pre-programmed to run a profile scan twice per hour for four consecutive days. Using a fully convolutional neural network, the copepodites were automatically counted – creating a depth profile of detected lice and measured light specter. The final results showed a diurnal migration pattern throughout the test period. • Capable of acquiring vertical density profiles of any aquatic species between 0,5 – 10 mm down to 11 m below the surface. • Fully automated and can be left unintended for weeks while collecting data.