Project #3

The seismic signatures of aseismic processes with deep learning powered monitoring

Main Supervisor: Men-Andrin Meier (ETH Zurich)
Co-Supervisor: David Marsan (UGA)

Location: Swiss Federal Institute of Technology (Switzerland)

Duration of the PhD: 3 years

The doctoral candidate will be enrolled in a PhD program at the Swiss Federal Institute of Technology

Objectives: Aseismic processes can play a first-order role in the build-up to large earthquakes, but they are hard to detect and monitor. The strong recent advances in seismic monitoring with deep learning (DL) techniques is an opportunity to improve the detection and characterisation of the subtle seismic signatures that aseismic processes leave behind. The doctoral candidate will i) develop new DL methods that are tailored to characterise the seismic signatures of aseismic slip during earthquake sequences. This may include stress migration and rotation, strain acceleration as captured by repeating earthquakes, and fluid pressure build-up as evidenced by seismic swarms; and ii) study the predictive value of aseismic observations for anticipating large earthquakes. We will use some of the recent, exceptionally well recorded earthquake sequences (natural or stimulated) to constrain transient aseismic deformation, including deformation caused by underground fluid flow. From this observational basis we will develop mixture seismicity models that account for the observed triggering of earthquakes by aseismic transients and previous shocks. This will allow us to study how the total deformation is partitioned into seismic and aseismic contributions, in space and in time. The goal is to understand the physics of the hard-to-observe aseismic deformation, and to design a seismicity model that provides substantially improved probability gain, compared to state-of-the-art models.

Expected Results:

  1. Development and implementation of DL monitoring method to generate next-level, deep seismicity catalogues;
  2. Monitoring and inference of aseismic deformation, and their underlying driving mechanisms such as fluid flow;
  3. Operational seismicity model for predicting the evolution of earthquake sequences;
  4. Improved understanding of the interactions between seismic and aseismic deformation mechanisms;
  5. Data-driven, objective inference of fault structures and geometries at small, decametre scale.

Planned secondments: UGA (12 months, D. Marsan, M19-30, Developing the mixture seismicity model, and its physical interpretation in terms of seismic and aseismic deformation processes).