京都大学 白眉センター

Faults can accommodate stress in a variety of slip modes, from devastating earthquakes to slow aseismic events. Among these slip modes, slow events remain among the most elusive and poorly understood.

Unraveling the interactions between slip modes is key to progress towards a better understanding of earthquakes and how they start: while laboratory experiments point to slow nucleation generally preceding dynamic rupture, observations in the field are far from systematic, and more the exception than the rule.

However, the difficulty in detecting transient slow slip events points to a possible observational gap that may explain the rarity of slow deformation detected prior to dynamic earthquakes.

In this presentation, the use of machine learning to monitor the earthquake cycle and improve the detection of slow deformation will be discussed, from the laboratory to the field.