Periodic Reporting for period 1 - HOPE (HOw Predictable are Earthquakes)
Reporting period: 2023-01-01 to 2025-06-30
Earthquakes are spectacular natural disasters, as exemplified by the 2004 Sumatra and 2011 Tohoku-Oki earthquakes. Predicting earthquakes remains one of the biggest societal challenges in natural science. This research project will attempt answering the following question: How predictable are earthquakes? We propose a multidisciplinary approach articulated around three main axes: (i) the deterministic predictability of earthquakes in simple, homogeneous faults, studied by reproducing and understanding earthquake phenomena in the laboratory, (ii) the deterministic predictability of earthquakes in complex, heterogeneous faults, studied by laboratory experiments producing multiple earthquake cycles on faults with controlled heterogeneities and (iii) the statistical predictability of earthquakes, studied by forecasting the spatial distribution of experimental seismicity using machine learning. At the core of this project lies the development of a new dedicated experimental setup to generate multiple earthquake cycles along a fault with prescribed complex geometry and rheology. With this new capability, we will conduct a threefold experimental program to: (i) compute the complete energy budget of laboratory earthquakes, (ii) study the sensitivity of rupture nucleation, propagation and arrest to heterogeneities, and (iii) study the effect of heterogeneities on the relation between fault seismic coupling and seismicity.
During the first two years of the ERC project, we successfully installed both the biaxial apparatuses Crackdyn and Cyclequakes in the host laboratory Geoazur. We also set up complete data acquisition systems for both devices, including high-frequency acquisition for strain gauge measurements, accelerometers, and displacement sensors. Additionally, we synchronized all acquisition systems with the ultra-high-speed camera TMX6410. We developed an advanced optical system leveraging the photoelastic properties of PMMA, using a highly stable polarized light source. This setup, combined with the high-speed camera, enabled us to capture the rupture front during dynamic rupture propagation. The precise synchronization between the camera and high-frequency sensors allowed us to image and accurately quantify the breakdown work during instabilities, demonstrating that frictional ruptures observed in our experiments mirror those in natural earthquakes.
Coupled with the development of the experimental apparatuses, we also designed two distinct inversion procedures to reconstruct the evolution of fault slip during both the nucleation and propagation of seismic ruptures.
First, we developed a custom approach to compute the Green’s function of our experimental medium using finite element methods. This method allows us to model the strain response expected at strain gauge locations during biaxial experiments with the CrackDyn apparatus, as well as triaxial experiments on saw-cut granite samples. Building on this Green’s function solution, and in collaboration with Pierre Dublanchet, we created the KISLAB algorithm — an innovative tool enabling static and quasi-static inversion of fault slip during earthquake nucleation. KISLAB combines deterministic and Bayesian approaches to determine the optimal fault slip evolution and is now openly available on GitHub.
Secondly, we developed an inversion procedure to reconstruct both kinematic and quasi-static fault slip during frictional rupture propagation in Crackdyn experiments. This algorithm integrates two key forward modeling solutions: the first uses the widely adopted Okada solution for Green’s functions, commonly applied to natural earthquakes, while the second relies on numerical modeling via COMSOL, capturing the complete experimental setup, including local loading and sample holder effects. By leveraging both approaches, our algorithm reconstructs the slip front evolution during rupture propagation, aligning well with direct measurements from the high-speed camera and displacement sensors. In addition, this solution allowed us for the first time to compute accurate solution of the moment rate function during laboratory earthquakes, that could help us to explore the possible self-similarity of earthquakes in the laboratory.
These two complementary inversion methods establish an essential bridge between laboratory-controlled earthquake data analysis and inversions applied to natural earthquake datasets — a key milestone planned in WP2 of our research proposal.
Coupled with the development of the experimental apparatuses, we also designed two distinct inversion procedures to reconstruct the evolution of fault slip during both the nucleation and propagation of seismic ruptures.
First, we developed a custom approach to compute the Green’s function of our experimental medium using finite element methods. This method allows us to model the strain response expected at strain gauge locations during biaxial experiments with the CrackDyn apparatus, as well as triaxial experiments on saw-cut granite samples. Building on this Green’s function solution, and in collaboration with Pierre Dublanchet, we created the KISLAB algorithm — an innovative tool enabling static and quasi-static inversion of fault slip during earthquake nucleation. KISLAB combines deterministic and Bayesian approaches to determine the optimal fault slip evolution and is now openly available on GitHub.
Secondly, we developed an inversion procedure to reconstruct both kinematic and quasi-static fault slip during frictional rupture propagation in Crackdyn experiments. This algorithm integrates two key forward modeling solutions: the first uses the widely adopted Okada solution for Green’s functions, commonly applied to natural earthquakes, while the second relies on numerical modeling via COMSOL, capturing the complete experimental setup, including local loading and sample holder effects. By leveraging both approaches, our algorithm reconstructs the slip front evolution during rupture propagation, aligning well with direct measurements from the high-speed camera and displacement sensors. In addition, this solution allowed us for the first time to compute accurate solution of the moment rate function during laboratory earthquakes, that could help us to explore the possible self-similarity of earthquakes in the laboratory.
These two complementary inversion methods establish an essential bridge between laboratory-controlled earthquake data analysis and inversions applied to natural earthquake datasets — a key milestone planned in WP2 of our research proposal.
The most significant achievements during the first two years of the project include the successful installation and calibration of the two experimental apparatuses funded by the project. These systems are now fully operational, delivering high-precision measurements on a daily basis.
We have also published five scientific papers and submitted three more, expected to be published in the coming months.
Firstly, we uncovered the origin of long-tailed weakening during frictional ruptures and its role in shaping the stress singular fields recorded at the rupture tip, influencing rupture velocities. These findings were published in Earth and Planetary Science Letters (EPSL) in 2023.
Secondly, we investigated the impact of this unconventional weakening on the potential extent of seismic ruptures. Both experimental and theoretical analyses demonstrated that this additional weakening significantly contributes to the rupture size. These results were published in 2024 in Journal of the Mechanics and Physics of Solids (JMPS). An outstanding question remains: how this long-tailed weakening influences the equations of motion — an area we continue to explore and aim to resolve within the next two years.
Thirdly, we examined how frictional heterogeneities influence seismic cycle dynamics. Initially, we hypothesized that brittle and ductile asperities would primarily affect rupture propagation. Instead, we observed that they predominantly control the emergence of fault afterslip. To our knowledge, this represents the first clear laboratory evidence of fault afterslip during seismic cycles. These results were published in January 2025 in EPSL.
Finally, we published a paper on the KISLAB algorithm (detailed in section 1.2) in Journal of Geophysical Research: Solid Earth (JGR Solid Earth). This work marks the first successful quasi-static inversion of fault slip evolution during laboratory earthquake nucleation, paving the way for new research directions in experimental rock mechanics.
We have also published five scientific papers and submitted three more, expected to be published in the coming months.
Firstly, we uncovered the origin of long-tailed weakening during frictional ruptures and its role in shaping the stress singular fields recorded at the rupture tip, influencing rupture velocities. These findings were published in Earth and Planetary Science Letters (EPSL) in 2023.
Secondly, we investigated the impact of this unconventional weakening on the potential extent of seismic ruptures. Both experimental and theoretical analyses demonstrated that this additional weakening significantly contributes to the rupture size. These results were published in 2024 in Journal of the Mechanics and Physics of Solids (JMPS). An outstanding question remains: how this long-tailed weakening influences the equations of motion — an area we continue to explore and aim to resolve within the next two years.
Thirdly, we examined how frictional heterogeneities influence seismic cycle dynamics. Initially, we hypothesized that brittle and ductile asperities would primarily affect rupture propagation. Instead, we observed that they predominantly control the emergence of fault afterslip. To our knowledge, this represents the first clear laboratory evidence of fault afterslip during seismic cycles. These results were published in January 2025 in EPSL.
Finally, we published a paper on the KISLAB algorithm (detailed in section 1.2) in Journal of Geophysical Research: Solid Earth (JGR Solid Earth). This work marks the first successful quasi-static inversion of fault slip evolution during laboratory earthquake nucleation, paving the way for new research directions in experimental rock mechanics.