Geophysical applications of AI for natural hazard and georesources


With our present day methodologies, the dynamics of the processes in Earth are still hidden in rapidly accumulating seismic and geodetic data. AI will bridge the gap between the amount of observations and the incomplete understanding of the processes.
New weak signals highlight the complex spatiotemporal evolution of the deformation. Our group is at the forefront for massive processing of seismological (passive imaging and monitoring) and geodetic data. Our goals are: the detection of known and new classes of signals, the disentanglement of the physical processes at work, the upscaling of our workflows.


Multiscale representation of signals
Continuous records clustering
Detection with vector auto-regression
Ambient noise modeling
The applications cover the micro-earthquake detection and classification in mining and fluid injection environments and the improvement of ambient noise based imaging and monitoring, a technique developed in our group.
Our cross disciplinary group implement the tools of AI to handle the complexity and the imbrication of scales of the different interacting non-linear processes acting on the solid and fluid components of the actual medium.


Joint work seminars

American Geophysical Union, Fall Meeting 2019, abstract #S52A-05, Towards Systematic Classification of Seismic Signals with Deep Neural Networks , Beauce, E.; Cougoulat, G.; Poli, P.; Seydoux, L.; van der Hilst, R. D.; Campillo, M.

American Geophysical Union, Fall Meeting 2019, abstract #S52A-04
Seismic signals and noises clustering with unsupervised deep representation learning, Seydoux, L.; Balestriero, R.; Poli, P.; De Hoop, M. V.; Baraniuk, R.; Campillo, M.

European Geosciences Union General Assembly 2019 Unsupervised learning for identification of seismic signals Maarten de Hoop, Richard Baraniuk, Joan Bruna, Michel Campillo, Hope Jasperson, Stephane Mallat, TanNguyen, and Leonard Seydoux


  • El Bouch S. & al. (2021) “A Normality test for multivariate dependant samples”, submitted to J. Of Computational stat. and data analysis, Elsevier.

  • El-Bouch S et al. 'Joint Normality Test via Two-dimensional Projection. R Steinmann, L Seydoux, M Campillo (2021) Hierarchical exploration of single station seismic data with unsupervised learning.

  • L. Seydoux, M Campillo, R Steinmann, R Balestriero, M de Hoop (2021) Observing seismic signatures of slow slip events with unsupervised learning.

  • R. Steinmann, L Seydoux, M Campillo (2021) Revealing the signature of ground frost in continuous seismic data with machine learning.

  • L. Seydoux, R Steinmann, M Campillo, M de Hoop, R Balestriero (2021) Learning the signature of slow-slip events and slow earthquakes from seismic and geodetic data.

Published on  January 25, 2024
Updated on January 25, 2024