← Back to Sessions

Unsupervised Detection and Characterization of Glaciogenic Noise Sources in Greenland During Winter

Description: 

Glaciers and ice sheets are seismologically rich environments, where ice fracturing, stick-slip basal motion, and subglacial fluid flow all generate distinct signals which can be used to characterize different glaciogenic processes. Recently it was found that meltwater produced during summer and stored in partially frozen surface lakes can be released in winter, resulting large scale ice-flow accelerations that potentially impact long-term ice dynamics and sea-level rise. To better understand these unique events and other poorly observed wintertime processes, we develop an unsupervised signal detection and classification workflow using UMAP and ST-DBSCAN algorithms to identify and characterize glaciogenic noise sources in Greenland. Here, we focus on multiple years of data from a single seismic station near the terminus of Jakobshavn Isbræ, the fastest-flowing outlet glacier in Greenland. We show the wintertime noisescape rapidly transitions from a multi-month quiescent period to that with frequent impulsive events and background noise changes. We use UMAP and ST-DBSCAN to sequence these changing events into different classes of signals which we interpret using a combination of polarization analysis, cross correlation, modeling, and supplementary datasets of ice dynamic and temperature changes. The results highlight both the benefits of using an unsupervised tool to parse glaciogenic signals which would likely go undetected elsewise, as well as the difficulty of using a single station to fully attribute them to specific glacier processes.

Session: Applications and Discoveries in Cryoseismology Across Spatial and Temporal Scales [Poster Session]

Type: Poster

Date: 5/2/2024

Presentation Time: 08:00 AM (local time)

Presenting Author: Nathan

Student Presenter: No

Invited Presentation: 

Authors