Automatic identification of biophonics and sea ice processes in large datasets from the High Arctic Ocean

• Published in: International Conference on Underwater Acoustics 2024
• Main Authors: Cleverly, Jonathan, Blondel, Philippe, Sagen, Hanne, Dzieciuch, Matt, Storheim, Espen
• Format: Article in Journal/Newspaper
• Language: English
• Published: Institute of Acoustics 2024
• Subjects: underwater acoustics; climate change; Arctic; sea ice; biodiversity; /dk/atira/pure/subjectarea/asjc/3100/3102; name=Acoustics and Ultrasonics; /dk/atira/pure/subjectarea/asjc/1700/1702; name=Artificial Intelligence; /dk/atira/pure/sustainabledevelopmentgoals/life_below_water; name=SDG 14 - Life Below Water; /dk/atira/pure/sustainabledevelopmentgoals/climate_action; name=SDG 13 - Climate Action; albedo; Arctic Ocean; Climate change; Fram Strait; Sea ice
• Online Access: https://researchportal.bath.ac.uk/en/publications/b2c42cee-a0ba-4d3d-a8f8-9f63f283117b; https://doi.org/10.25144/22225; https://purehost.bath.ac.uk/ws/files/331373855/Cleverly_etal_Proc_IOA_2024_ArcticSoundscapes.pdf

The extent of sea ice in the Arctic Ocean has severely declined due to the amplification of climate change. Sea ice plays a pivotal role in the ice-albedo feedback process, limiting the planet’s absorption of solar radiation. Its reduction has been linked to rising sea levels and extreme weather events at more temperate latitudes. Many species are strongly associated with seasonal variation of ice cover, primarily for feeding grounds. With passive acoustic monitoring, acoustic signatures of marine mammals and sea ice processes can be identified to further understand the state of the Arctic. This study uses recordings from the long-term research projects “Acoustic Ocean Under Melting Ice” (UNDER-ICE, 2014-2016) and “Coordinated Arctic Acoustic Thermometry Experiment” (CAATEX, 2019-2020). UNDER-ICE, in the Fram Strait, recorded for 130 seconds every 3 hours, at 1,953 Hertz from September 2014 to March 2016, whereas CAATEX, along the Eurasian Basin, recorded for 45 minutes every 12 hours at 976 Hertz from August 2019 to August 2020. The selection of recording stations presented here corresponds to 50+ GB of data, encompassing a wide range of soundscapes and acoustic sources (including ships and geophonics). Due to the high duty cycle, these large datasets require automatic processing to highlight processes or signals of interest. This study will consider conventional detection methods, such as energy sum detectors, alongside the potential use of acoustic indices (such as Acoustic Complexity Index or Entropy) for identifying significant audio files for further analysis. Biophonic and cryophonic soundscape contributions will be linked with their seasonal variations, environmental changes in the region and future uses to train AI approaches for large-scale measurements. The usability of acoustic indices in analysing low-sample rate data (relative to the terrestrial ecological studies from which they were derived) will also be discussed.