HR: 1340h
AN: H23A-1127    [Abstracts]
TI: Full waveform elastic inversion in a space frequency domain formulation: a powerful geotechnical tool for superficial reconstruction
AU: * Gelis, C
EM: gelis@geoazur.unice.fr
AF: Geosciences Azur, 250 rue A. Einstein Sophia Antipolis, Valbonne, 06560 France
AU: Virieux, J
EM: viri@geoazur.unice.fr
AF: Geosciences Azur, 250 rue A. Einstein Sophia Antipolis, Valbonne, 06560 France
AU: Grandjean, G
EM: g.grandjean@brgm.fr
AF: BRGM ARN, 3 avenue Claude Guillemin BP 6009, orleans cedex 2, 45060 France
AU: Leparoux, D
EM: leparoux@ipgp.jussieu.fr
AF: IPGP, 4 place Jussieu, paris, 75005 France
AU: Operto, S
EM: operto@obs-vlfr.fr
AF: Geosciences Azur, 2 quai de la Darse BP 48, Villefranche sur mer, 06235 France
AB: The superficial weathered zone, few hundreds meters thick, presents high variable and complex near-surface structures. This leads to an energetic seismic ground roll and therefore hide information coming from deeper areas. Moreover near-surface anomalies may characterize potentially dangerous structures as cavities or their surrounding altered media. Therefore detecting heterogeneities in near-surface areas and quantifying their physical properties will be of great help for seismic imaging and for natural hazard assessment. Since heterogeneities are located in near-surface areas, both surface and body waves induce complex footprints in seismic data. The propagation of 2D P-SV is performed thanks to a frequency domain modeling. This frequency formulation takes into account attenuating behavior and efficiently takes benefit of multisource and multireceiver configurations. A new finite-difference stencil of second order using rotated derivatives axes (Saenger et al., 2000) simulates surface waves very precisely and remains stable nearby the free surface and/or rapidly-varying zones. It will be the forward problem kernel of our approach. We follow the matrix formalism of Pratt et al. (1998) and perform a linearized inversion in the least-square sense, since heterogeneities of reasonable amplitudes towards the surrounding medium are considered, leading us to resort to the Born approximation. We use the gradient method to perform the full waveform inversion for elastic waves. In this formulation we take only the Hessian diagonal part and use a parabolic approximation to find the stepping in the gradient direction. Our selected inversion takes into account kinematic and amplitude information for waves coming from various reflection angles at different offsets. This allows to recover local parameters as P wave and S wave velocities from dense seismic experiments. Applications to realistic synthetic configurations illustrate the potentiality of the method when both backward and forward scatterings are encountered. Influences of data sampling, data geometry and data redundancy are, of course, critical but the initial model is a very sensitive key input for successful convergence to the minimum of our misfit function, taking into account the complexity of waves interaction and propagation. From these illustrations, we highlight the importance of data introduction into the inversion tool in order to avoid non-global minimum. Saenger E. H., Gold N. & Shapiro S. A., 2000. Modeling the propagation of elastic waves using a modified finite-difference grid. Wave Motion, 31, 77-92. Pratt G., Shin C. & Hicks G.J., 1998. Gauss-Newton and full Newton methods in frequence space seismic waveform inversion. Geophys. J. Int., 133, 341-362.
DE: 7255 Surface waves and free oscillations
DE: 6982 Tomography and imaging
DE: 7203 Body wave propagation
DE: 3025 Marine seismics (0935)
DE: 0902 Computational methods, seismic
SC: Hydrology [H]
MN: 2004 AGU Fall Meeting