Recycling data: An annotated marine acoustic data set that is publicly available for use in classifier development and marine mammal research

Abstract

Barkley Canyon is a productive submarine canyon approximately 60 km southwest of Vancouver Island, Canada. The canyon's nutrient flow is affected by multiple regional currents, and draws aggregations of euphausiids, hake, herring, and various marine mammal species. A subset of the acoustic data collected from the 2013–2015 hydrophone deployment on the Barkley Canyon Upper Slope platform of Ocean Networks Canada's North-East Pacific Time-series Undersea Networked Experiments (NEPTUNE) observatory was manually annotated for marine mammal presence to investigate marine mammal habitat use and in support of development of a random forest classifier. This dataset, which is being made publicly available for further use, includes strong-label annotations of phonations from blue whales, fin whales, humpback whales, sperm whales, orcas, Pacific white-sided dolphins, Risso's dolphins, and other delphinids that could not be identified to species. All regional orca communities are represented within the dataset, and phonations are labelled to ecotype and pod level when possible. This dataset could be further used in a number of ways, including classifier development, investigations into habitat use and seasonality, or combining the acoustic data with data collected from co-located oceanographic instruments to investigate links between marine mammal presence and oceanographic conditions.