Temporal and spatial correlations in earthquakes dynamics : physical modelling and data analysis

One of the most distinctive and poorly understood feature of earthquakes is the significant increase of theseismic rate observed after large events. Well established empirical laws of aftershocks occurrencedemand for a physical explanation. Foreshocks are also observed before a large event but their statisticalfingerprints, potentially important for human security, are much more elusive. In this project, using themethods developed in the Statistical Physics we will design a model of the fault able to reproduce complexspatio-temporal patterns with foreshocks, mainshocks and aftershocks. Using Machine Learning we willunderstand the statistical properties of the short sequence of foreshocks. First, using our syntheticsequences, we determine how much information is needed to predict the following events. Then we willuse actual data: on one side to calibrate the model on the real fault activity, on the other side to predicthow dangerous is a real sequence of foreshocks. The PhD student is based at LPTMS and work both at theLPTMS and at LISN.

Earthquakes, foreshocks, statistical physics, machine learning, disordered systems

Partenaires du projet

ROSSO Alberto
(UMR8626) Orsay France
Schoenauer Marc
LRI (UMR8623) France
Libre de droits
Crédit photo : Libre de droits