Corentin Gouache

Stochastic generator of earthquakes in low-to-moderate seismic activity context: from data to hazard. Study case in French territories.

Supervisor: Jean-Marc Montel (CRGP)

Co-supervisors: François Bonneau (RING) - Pierre Tinard (CCR)

Objectives:

French mainland seismicity is considered as low-to-moderate. Indeed, only a few damaging earthquakes occurred in this territory over the last century. This lack of data (earthquakes recorded but also strain rates, active faults known...) is the main issue for who deals with seismic hazard assessment. This observation leads us to query the applicability of the classical seismic hazard assessment scheme, mainly developed in high seismicity regions (Californie, Japan...), to low-to-moderate seismicity areas. This classical seismic hazard assessment scheme is fond of parameters hard to determinate from sparse data otherwise arbitrarily.

Our approach is to use the non parametric inter event time method in order to compute objective recurrence rates of earthquakes in function of their magnitude. We developed an implementation to stabilize the results when data are really spare: for large magnitudes. This method allows us to calculate recurrence rates without increasing the uncertainties and using arbitrary parameters. A Monte Carlo Markov Chain is applied on initial data to propagate the magnitude uncertainties. This stochastic process leads to a probability density function of recurrence rate for each magnitude step instead of only one deterministic result.

For these pdf of reccurrence rate we generate synthetic earthquakes on the French mainland territory in order to overcome the lack of data. Then for each earthquake the hazard (Peak Ground Acceleration, PGA) is calculated thanks to the Ground Motion Prediction Equations (GMPE). The large amount of GMPE and the simplicity of these equations regarding the complex physical processes involved in the seismic waves propagation force the seismic hazard community to use several weighted GMPE. These weights are set arbitrarily from a group of experts. We want to overcome this subjectivity by weight a set of pre-selected GMPE thanks to numerical simulations of seismic waves propagation through the RegSEM code (Spectral Element Method) (Paul Cupillard).