Tracking Iceberg Calving Events in Greenland from 2013 to 2024 Using Seismic Data and Machine Learning

The retreat of Greenland’s glaciers is accelerating due to climate change, driven not only by rising temperatures but also by processes such as iceberg calving. These events contribute significantly to the Greenland Ice Sheet mass loss, a critical factor in global sea level rise. Identifying as many iceberg calving events as possible is essential for reducing the uncertainty in mass loss estimates, ultimately helping to improve our understanding of their cumulative impact on sea level rise and climate change.

We use seismic data to detect signals generated by time-varying forces during iceberg calving on marine-terminating glacier termini, known as glacial earthquakes. By applying a detection algorithm based on the Short-Time Average over Long-Time Average (STA/LTA) method, combined with a supervised machine learning approach (Random Forest), we successfully differentiate glacial earthquakes from tectonic earthquakes. Despite limited recordings per event, we can locate them using a non-linear location methodology (NonLinLoc).

Applying this methodology to continuous seismic data from 2013 to 2024, we identify more than 4500 previously undocumented glacial earthquakes along Greenland's coastline. While the yearly and monthly event counts are strongly influenced by the availability of seismic stations, seasonal variations in iceberg calving activity are clearly observed. This trend is further supported by an observed increase in detected events over time when focusing on a continuously available subset of stations. In addition, we will present the spatio-temporal evolution of detected events, providing further insights into the dynamics of iceberg calving activity.

These findings lay the groundwork for future work, including characterizing iceberg volume and shape to enhance our understanding of Greenland’s ice mass loss dynamics.