Tracking Local Sea Ice Extent in the Beaufort Sea Using Distributed Acoustic Sensing and Machine Learning

Abstract

Abstract Monitoring sea ice extent is critical to understand long-term trends in climate change. Here, we show that ambient noise recorded by fiber-optic sensing technology deployed in an Arctic shallow marine seafloor environment can track sea ice extent. We use a 37.4 km long section of fiber-optic cable deployed offshore of Oliktok Point, Alaska. Data are analyzed for two weeks: one in July 2021 and another in November 2021, when there is incomplete and evolving sea ice coverage. We apply different Machine Learning algorithms to identify types of ambient seismic noise in frequency–time scalogram images. We find evidence for two dominant noise types related to excitation of oceanic gravity waves in open water and the presence of sea ice with sufficient strength to suppress wave action. Comparison of the Distributed Acoustic Sensing (DAS) noise clustering results with satellite-based observations indicates that seafloor DAS can complement sea ice constraints from satellite imagery by locally increasing spatial and temporal resolution and tracking for which ice coverage is sufficient to diminish ocean waves.