Magnetotelluric study of seismic swarm areas (Maurienne and Ubaye valleys). Inversion studies.

premier semestre 2021, deux-trois mois

This project is aiming at the 3D resistivity imaging of two areas of seismic swarms in the French Alps : Maurienne and Ubaye valleys. The Ubaye swarm has been studied for nearly 20 years, while Maurienne swarm lasted only from 2016 till 2018 and represents a new target. The seismicity behaviour in Ubaye is peculiar, the swarms of micro-seismicity alternate with main shock - aftershock sequences, suggesting as possible source mechanisms transfer of coseismic stress as well as fluid-pressure diffusion. The Maurienne swarm is not yet fully understood due to the complex tectonic situation there. To better understand the role of fluids in these two areas, we propose to carry out broad band magneto-telluric (MT) measurements. MT will be used to characterize the electrical resistivity distribution of the Earth, extremely sensitive to the presence of fluids.
The student will take part in the field work to complete the existing set of magneto-telluric data and process the new data using available code (FFMT).
(S)he will then perform 1- and 3D inversion of the completed data set using the stochastic Markov chain Monte Carlo inversion code RJMCMC_MT and modEM software respectively. Once a model is produced, it is important to characterize its uncertainty and depth of investigation. Uncertainty quantification is possible using stochastic approaches, e.g., Markov chain Monte Carlo techniques (e.g. Mandolesi et al, 2018). These techniques allow to fully explore the parameter space for 1D applications with no need of keeping close to some arbitrary chosen smooth prior model (on the contrary to 2-and 3D deterministic inversions). For 3D inversion however, the computational cost of sampling the full parameter space makes direct use of stochastic techniques impossible. The simplest method to evaluate the sensitivity (e.g., the depth of investigation) achieved by the model is to produce an ensemble of models explaining the observations using various prior models of resistivity (same idea as in Oldenburg &Li, 1999). Beyond the depth of investigation, the data do not constrain the inversion and resulting model converges to the prior model. The variance of the ensemble therefore can be used as visualization of the sensitivity loss, while the weighted average of the ensemble can be considered as final model (e.g., with weights inversely proportional to the rms).
Finally, the obtained model of resistivity will be discussed in terms of physical parameters (e.g., presence of water at depth).
We are interested into continuing the subject during a M2 project.

References :
Brodie & Jiang, 2018, Trans-Dimensional Monte Carlo Inversion of Short Period Magnetotelluric Data for Cover Thickness Estimation, https://github.com/GeoscienceAustralia/rjmcmcmt
Kelbert et al (2014) Modem : a modular system for inversion of electro-magnetic geophysical data, Comput Geosci 66:40–53
Thouvenot et al., (1998). The Ml 5.3 Epagny (French Alps) earthquake of 1996 July 15 : a long awaited event on the Vuache fault. Geophysical Journal International, 135, 876-892.