AI uncovers hidden fault movements along the San Andreas

A deep-learning algorithm trained on continuous borehole strainmeter data has uncovered dozens of previously undetected slow slip events beneath the Parkfield section of California's San Andreas Fault, according to a study published in Nature Communications. The research demonstrates that these silent fault movements are systematically followed by increased low-frequency earthquake activity, offering new insight into how stress is released along one of the world's most closely watched fault systems.nature+1

Detecting the Invisible

The study, led by Zahra Zali of the GFZ German Research Centre for Geosciences along with co-authors Patricia Martínez-Garzón, David Mencin, and Gregory C. Beroza, applied deep learning directly to high-sensitivity strainmeter observations to identify short-duration slow slip events that conventional detection methods had missed. Unlike regular earthquakes, slow slip events release energy gradually over hours rather than seconds, making them invisible to standard seismic monitoring.linkedin+2

The researchers built what amounts to an automated catalog of these events at Parkfield — a stretch of the San Andreas Fault that has long served as a natural laboratory for earthquake science. Their algorithm learned patterns from the raw strain data without relying on predefined templates, allowing it to pick up subtle deformation signals buried in noise.gfzpublic.gfz+1

A Link to Low-Frequency Earthquakes

The most striking finding is the temporal relationship between slow slip and low-frequency seismicity. The team found that detected slow slip events are consistently followed by upticks in low-frequency earthquakes — small, deep tremor-like signals that occur in the fault's transition zone. This connection suggests that slow slip may load adjacent patches of the fault, triggering the low-frequency activity and pointing to a more dynamic interplay between different modes of fault movement than previously understood.nature+1

Implications for Fault Monitoring

The work arrives at a moment of heightened attention to California's seismic hazards. A separate study published in June found that stress on the San Andreas and San Jacinto faults has reached its highest level in a thousand years. While slow slip events relieve rather than build stress, understanding their patterns could improve models of how strain accumulates and dissipates along the fault.cnn+1

The study underscores AI's expanding role in geophysics, where machine-learning tools are increasingly able to extract information from datasets too large and complex for human analysts to process manually. As EarthScope, which operates the borehole strainmeter network used in the research, noted, such instruments provide the sensitivity needed to detect these "silent" movements when paired with modern computational techniques.linkedin+1