View in EDS

The Influence of Ice Coverage, Calving, and Melt on Underwater Ambient Sound in a Glacierized Fjord.

Saved in:
Bibliographic Details
Title: The Influence of Ice Coverage, Calving, and Melt on Underwater Ambient Sound in a Glacierized Fjord.
Authors: Zeh, Matthew C.1,2,3 (AUTHOR) matthew.zeh@belmont.edu, Pettit, Erin C.4 (AUTHOR), Ballard, Megan S.3 (AUTHOR), Wilson, Preston S.2,3 (AUTHOR), Amundson, Jason M.5 (AUTHOR)
Source: Journal of Geophysical Research. Earth Surface. Jan2026, Vol. 131 Issue 1, p1-26. 26p.
Subject Terms: *Fjords, *Ice fields, Ice calving, Melting, Ecosystem dynamics, Machine learning, Underwater acoustics
Geographic Terms: Southeast Alaska
Abstract: Noise from calving icebergs, cracking ice, and melting ice dominates the underwater soundscape of glacierized fjords creating one of the loudest recorded ambient ocean environments. While progress has been made toward identifying and describing individual sound sources—including the automatic detection of calving and quantification of ice‐mass loss—the relative contributions of multiple, simultaneous processes, and how these contributions evolve over time, remain underexplored, limiting robust interpretation of ice‐ocean interactions. Here, we show that unsupervised machine learning separates a series of recordings captured over 8 months into five dominant sound profiles related to glacier activity. We deployed an array of hydrophones approximately 400 m from the terminus of Xeitl Sít' (LeConte Glacier) in Southeast Alaska and recorded sound regularly between October 2016 and May 2017. Using the k‐means clustering algorithm, we cluster spectral shapes of 10,440 background acoustic spectra, defined as the 25th‐percentile spectral level of each recording. We identify five distinct acoustic clusters and relate their temporal occurrence to environmental time series including ice movement, meteorology, and oceanographic data. We further link spectral shapes to known glacier sources such as calving and ice melt. Our analysis reveals that these clusters correspond more closely with glacier and ice‐mélange activity than with other environmental variables, confirming the dominance of glacier behavior on fjord soundscapes. This research demonstrates the effectiveness of clustering passive acoustic data and provides a framework for analyzing large, complex acoustic data sets of undersampled environments—such as glacierized fjords—to guide interpretation and track changes in dominant environmental processes. Plain Language Summary: The underwater environment near a tidewater glacier is a cacophony of sounds and forms one of the ocean's loudest ambient soundscapes. Icebergs frequently break off from the glacier face ("calving") and crash into the ocean. As these icebergs melt, they crack and "fizz" like seltzer water, as pressurized air bubbles in the ice escape into the water. These sounds contain information about glacier melt rates and the stability of the glacier‐ocean boundary. Researchers have made advances identifying specific glacier sounds—such as automatically detecting and estimating ice loss from calving—but it remains difficult to tell how much overlapping processes contribute to the overall soundscape, especially over longer timescales. To explore this, we deployed underwater acoustic recorders near Xeitl Sít' (LeConte Glacier) in Southeast Alaska, regularly recording over 8 months. Using an unsupervised machine–learning tool—one not given prior information about the data—we sorted nearly 430 hr of recordings into five groups based on frequency content. Comparing these groups with local environmental conditions reveals several relationships between recorded sounds and environmental processes. The presence and melt of unmoving icebergs and calving events show the strongest ties to the local sound field. Key Points: K‐means clustering, a machine‐learning tool, can categorize ambient sound in glacier fjords into five prototypes linked to glacier activityAn 8‐month acoustic data set shows glacier and ice‐mélange activity dominate the fjord soundscape over atmospheric and ocean processesThe clustering‐based approach presented here provides a framework for interpreting large acoustic data sets from undersampled environments [ABSTRACT FROM AUTHOR]
Copyright of Journal of Geophysical Research. Earth Surface is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: GreenFILE
Description
Abstract:Noise from calving icebergs, cracking ice, and melting ice dominates the underwater soundscape of glacierized fjords creating one of the loudest recorded ambient ocean environments. While progress has been made toward identifying and describing individual sound sources—including the automatic detection of calving and quantification of ice‐mass loss—the relative contributions of multiple, simultaneous processes, and how these contributions evolve over time, remain underexplored, limiting robust interpretation of ice‐ocean interactions. Here, we show that unsupervised machine learning separates a series of recordings captured over 8 months into five dominant sound profiles related to glacier activity. We deployed an array of hydrophones approximately 400 m from the terminus of Xeitl Sít' (LeConte Glacier) in Southeast Alaska and recorded sound regularly between October 2016 and May 2017. Using the k‐means clustering algorithm, we cluster spectral shapes of 10,440 background acoustic spectra, defined as the 25th‐percentile spectral level of each recording. We identify five distinct acoustic clusters and relate their temporal occurrence to environmental time series including ice movement, meteorology, and oceanographic data. We further link spectral shapes to known glacier sources such as calving and ice melt. Our analysis reveals that these clusters correspond more closely with glacier and ice‐mélange activity than with other environmental variables, confirming the dominance of glacier behavior on fjord soundscapes. This research demonstrates the effectiveness of clustering passive acoustic data and provides a framework for analyzing large, complex acoustic data sets of undersampled environments—such as glacierized fjords—to guide interpretation and track changes in dominant environmental processes. Plain Language Summary: The underwater environment near a tidewater glacier is a cacophony of sounds and forms one of the ocean's loudest ambient soundscapes. Icebergs frequently break off from the glacier face ("calving") and crash into the ocean. As these icebergs melt, they crack and "fizz" like seltzer water, as pressurized air bubbles in the ice escape into the water. These sounds contain information about glacier melt rates and the stability of the glacier‐ocean boundary. Researchers have made advances identifying specific glacier sounds—such as automatically detecting and estimating ice loss from calving—but it remains difficult to tell how much overlapping processes contribute to the overall soundscape, especially over longer timescales. To explore this, we deployed underwater acoustic recorders near Xeitl Sít' (LeConte Glacier) in Southeast Alaska, regularly recording over 8 months. Using an unsupervised machine–learning tool—one not given prior information about the data—we sorted nearly 430 hr of recordings into five groups based on frequency content. Comparing these groups with local environmental conditions reveals several relationships between recorded sounds and environmental processes. The presence and melt of unmoving icebergs and calving events show the strongest ties to the local sound field. Key Points: K‐means clustering, a machine‐learning tool, can categorize ambient sound in glacier fjords into five prototypes linked to glacier activityAn 8‐month acoustic data set shows glacier and ice‐mélange activity dominate the fjord soundscape over atmospheric and ocean processesThe clustering‐based approach presented here provides a framework for interpreting large acoustic data sets from undersampled environments [ABSTRACT FROM AUTHOR]
ISSN:21699003
DOI:10.1029/2025JF008435