Nathan Angelakis ^1,2* Gretchen L. Grammer ^1,3 Sean D. Connell ¹ Frédéric Bailleul ³ Leonardo M. Durante ⁴ Roger Kirkwood ² Dirk Holman ⁵ Simon D. Goldsworthy ^1,2

• ¹ Ecology and Evolutionary Biology, School of Biological Sciences, University of Adelaide, Adelaide, South Australia, Australia
• ² Marine Ecosystems, South Australian Research and Development Institute, Adelaide, Australia
• ³ Pelagic Fisheries, South Australian Research and Development Institute, Adelaide, Australia
• ⁴ Division of Aquatic Resources, Department of Land and Natural Resources, Honolulu, United States
• ⁵ Department for Environment and Water, South Australia, Australia

Across the world’s oceans, our knowledge of the habitats on the seabed is limited. Increasingly, video/imagery data from remotely operated underwater vehicles (ROVs) and towed and drop cameras, deployed from vessels, are providing critical new information to map unexplored benthic (seabed) habitats. However, these vessel-based surveys involve considerable time and personnel, are costly, require favorable weather conditions, and are difficult to conduct in remote, offshore, and deep marine habitats, which makes mapping and surveying large areas of the benthos challenging. In this study, we present a novel and efficient method for mapping diverse benthic habitats on the continental shelf, using animal-borne video and movement data from a benthic predator, the Australian sea lion (Neophoca cinerea). Six benthic habitats (between 5-110m depth) were identified from data collected by eight Australian sea lions from two colonies in South Australia. These habitats were macroalgae reef, macroalgae meadow, bare sand, sponge/sand, invertebrate reef and invertebrate boulder habitats. Percent cover of benthic habitats differed on the foraging paths of sea lions from both colonies. The distributions of these benthic habitats were combined with oceanographic data to build Random Forest models for predicting benthic habitats on the continental shelf. Random forest models performed well (validated models had a >98% accuracy), predicting large areas of macroalgae reef, bare sand, sponge/sand and invertebrate reef habitats on the continental shelf in southern Australia. Modelling of benthic habitats from animal-borne video data provides an effective approach for mapping extensive areas of the continental shelf. These data provide valuable new information on the seabed and complement traditional methods of mapping and surveying benthic habitats. Better understanding and preserving these habitats is crucial, amid increasing human impacts on benthic environments around the world.