Je suis
Citoyen / Grand public
Étudiant / Futur étudiant
Partenaire public
Enseignant / Elève

Multicomponent observables : full waveform inversion and laboratory scale measurements.


IPGP - Îlot Cuvier


Séminaires de Sismologie

Salle 310

Raphaël Valensi


The first part of this talk concerns the physical reproduction of multicomponent seismic experiments at the laboratory scale (reduced scale modelling). An ultrasonic experimental facility (MUSC) dedicated to the reproduction of seismic experiments at the laboratory scale has been developed at IFSTTAR Nantes (formerly LCPC). To enable high fidelity multicomponent measurements, we take advantage of a recent ultrasonic-laser interferometry technique which enables to simultaneously measure the vertical and the horizontal components (P-SV configuration). The fidelity of the measurements is demonstrated with a quantitative analysis of reccorded waveforms and polarisation observables obtained on a reference model. I will conclude this section by showing some results of a massive multicomponent data set obtained on a physical model representing a near surface structure. The second part of my talk concerns the inversion of alternative multicomponent observables, with a special emphasis on polarization observables. I will first derive a new kind of "well-posed" polarization misfit function which enables to consider all the different polarization observables in an unified and consistent way. I integrated these alternative objective functions to a full waveform inversion code in order to quantitatively estimate the elastic parameters (collaboration with R. Brossier). First, with a canonical example, I illustrate some properties of the polarization observables. Then, I will introduce an example consisting in a near surface imaging configuration where the signals are energetically dominated by the contributions of the surface waves. Using this example, I will discuss some benefits of multicomponent observables to reconstruct the elastic parameters of the shallow subsurface (Vs, Vp) when the density/attenuation parameters and the source wavelet are misestimated.