Machine Learning Enhanced Earthquakes Reveal Stress Changes in Icelandic Geothermal Fields

Colin N. Pennington, Egill A. Guðnason, Þorbjörg Ágústsdóttir, Anette K. Mortensen, & Rögnvaldur L. Magnússon

Submitted August 30, 2026, SCEC Contribution #15100, 2026 SCEC Annual Meeting Poster #TBD

Krafla and Þeistareykir geothermal fields in northeast Iceland are respectively one of the oldest and one of the newest geothermal operations in Iceland. Since 2006, continuous monitoring by Iceland GeoSurvey (ÍSOR) for Landsvirkjun has produced a unique earthquake record of about 70,000 carefully reviewed events. This long-term dataset offers a rare chance to better understand how underground stresses, faults, geothermal operations, and volcanic activity interact over time. Using a machine learning-based workflow, we greatly increased the number of reliable earthquake focal mechanisms identified in the Krafla and Þeistareykir geothermal areas for 2013 to 2026. The new catalogue contains 3,457 focal mechanisms, including 3,089 from Krafla and 366 from Þeistareykir, a major increase over previously published, manually developed catalogues. The expanded dataset makes it possible to track changes in faulting behavior through time within key seismic clusters. In Krafla, we found a noticeable change in earthquake faulting style beginning around 2020, at the same time as a major inflation-deflation cycle. In Þeistareykir, the results show growing variety in earthquake behavior in recent years.

Key Words
Focal Mechanisms, Machine Learning, Volcano, Geothermal

Citation
Pennington, C. N., Guðnason, E. A., Ágústsdóttir, Þ., Mortensen , A. K., & Magnússon, R. L. (2026, 08). Machine Learning Enhanced Earthquakes Reveal Stress Changes in Icelandic Geothermal Fields. Poster Presentation at 2026 SCEC Annual Meeting.

Related Projects & Working Groups
Fault and Rupture Mechanics (FARM)