Search

International audience; Recently, we have developed a localized adaptive waveform inversion method (LAWI) to tackle the cycle-skipping issue in velocity reconstruction by waveform inversion. In LAWI, the Gabor deconvolution is applied to compute a local matching filter, whose centroid time is used for measuring the instantaneous time shift between observed and calculated data. Different from AWI that is based on a stationary convolutional model, LAWI can take the non-stationarity nature of seismic data into account, therefore performs better in handling realistic cycle skipping problem. Numerical tests show that, compared with AWI, the application of LAWI seems to require a higher signal-to-noise ratio (SNR) of observed data. To make LAWI work for low-SNR data, a delta-type regularization is developed to deal with the noise problems inherent in the Gabor deconvolution. Despite a slight resolution loss and a "layer-stripping principle break" induced by this new regularization illustrated numerically, we present how this method can be useful to invert for low-SNR data on the Chevron benchmark dataset.

International audience; Full waveform inversion (FWI), as an efficient seismic imaging tool, is widely used in the investigation of the structure of the Earth. For the oil & gas industry, in addition to the subsurface image, a quantitative estimation of elastic properties from seismic data is crucial for the geophysical characterization and monitoring of the subsurface lithology and reservoirs. In marine surveys, with the emergence of the fourcomponent (4C) ocean-bottom acquisition systems (deploying 1C hydrophone plus 3C geophone on the seabed), more elastic wave propagation effects can be recorded in the seismic data for an elastic property estimation of the subsurface, i.e. extracting medium shear modulus from the S-wave velocity model. Multi-parameter elastic FWI offers the possibility to reconstruct the P-wave (V p) and S-wave velocity (V s) models jointly. However, compared with the mono-parameter acoustic FWI, the multi-parameter elastic FWI can be more vulnerable due to the parameter coupling, wave modes conversion and interference. For the purpose of building a robust multi-parameter elastic inversion of the 4C seismic data, we consider a workflow design from the aspect of model parameterization analysis. Three different elastic model parameterizations, m 1 = (V p ,V s), m 2 = (V p ,V p /V s) and m 3 = (V p , σ) (σ is the Poisson's ratio), are analysed in terms of data sensitivity and model gradient feature. Together with synthetic case studies on a series of overburden models with increasing elastic effects and a realistic geological model, we conclude that a workflow of first inverting hydrophone data with (V p , V p /V s) parameterization and then 3C geophone data with (V p , V s) parameterization contributes to a robust and reliable V p and V s model reconstruction.

International audience; Subduction zones around the world remain one of the most important targets of geophysical investigations. This is because the megathrusts at convergent margins are the sites of the largest earthquakes ever recorded. The factors that control these earthquakes, as well as the evolution of accretionary wedges are still subject of geological and geophysical interpretation. Given that subduction zones are generally located offshore, approaches from the field of marine geophysical imaging-in particular seismic acquisition and processing techniques-are the methods of choice to investigate these sites. In this study we combine the results of waveform modeling, inversion and migration methods applied to the ocean-bottom seismometer (OBS) data and multi-channel seismic (MCS) data acquired in the Tokai segment of the eastern Nankai Trough. We use high-resolution velocity model derived with full-waveform inversion (FWI) and its attributes together with the reflectivity profile derived with Kirchhoff pre-stack depth migration (K-PSDM) to build a data-driven geological interpretation of the region. We focus on the nature of the subducting oceanic crust, geometry of the megathrust and splay faults, subducting front, complex accretionary wedge and the paleo-prism acting as a backstop. In particular we are able to image in great detail the complex deformation associated to the subduction of seamounts. We perform waveform and ray modeling in the FWI velocity model to analyse wave propagation within the accretionary wedge, making it possible to associate different structures directly to the arrivals in the OBS data. We support our seismic-derived results with bathymetry, magnetic and gravity data and show that they are consistent with our geologic interpretation of the structural elements that shape the Tokai segment.

International audience; Main objectives Recasting polarization-based multi-component inversion in a time-domain compatible framework. Showing, through a numerical example, the superior robustness of the proposed approach with respect to phase and amplitude ambiguities compared to more conventional approaches. New aspects covered Introducing a strategy for evaluating the distance between polarization states in the time-frequency domain. Presenting the associated misfit function and adjoint sources. Application to a realistic case study with a direct comparison against the classical multi-component full waveform inversion.

International audience; In Manceau et al. (2021), 1 we measured the mass-dependent fractionation (MDF) of mercury (Hg) stable isotopes, noted as δ 202 Hg, in 22 biological tissues from eight giant petrels (Macronectes spp.

We derive a three-dimensional shear-wave velocity model of the Ligurian-Provence back-arc basin (Northwestern Mediterranean Sea) using ocean-bottom seismometers (AlpArray OBSs) and land stations from permanent and temporary seismic networks. The quality of OBS continuous records is enhanced by a specific processing that reduces instrumental and seabed-induced noises (transients, tilt, compliance). To further improve the resolution of ambient-noise tomography in the offshore area, we compute the Rayleigh-wave part of the Green functions for OBS-OBS pairs by using onshore stations as virtual sources. 2-D group-velocity maps and their uncertainties are computed in the 4 - 150 s period range by a transdimensional inversion of Rayleigh-wave travel times. The dispersion data and their uncertainties are inverted for a probabilistic 3-D shear-wave velocity model that includes probability densities for Vs and for the depth of layer interfaces. The probabilistic model is refined by a linearized inversion that accounts for the water layer in the Ligurian Sea. Our S-wave velocity and layer boundary probability models correspond well to a recent, high-resolution P-wave velocity cross-section derived from controlled-source seismic profiling along the Ligurian-Provence basin axis. A joint interpretation of the P- and S-wave velocity sections along this profile reveals a thin, anomalous oceanic crust of low P-wave velocities but high S-wave velocities, intruded by a gabbroic body. The illuminated part of the upper mantle appears to be devoid of serpentinization. Key Points Efficient processing scheme to remove transients and reduce tilt and compliance from continuous ambient noise recorded by OBSs Computation of iterative correlations between OBSs based on a virtual reconstruction of the Rayleigh waves Thin, anomalous oceanic crust with gabbroic intrusions evidenced in the basin axis from a joint interpretation of Vs and Vp models Plain Language Summary The Ligurian-Provence basin (Northwestern Mediterranean Sea) is one of the Miocene-Pliocene back-arc basins that resulted from the retreat of the Adria subduction in the plate reorganization due to Africa-Europe convergence. The crustal structure of the basin is still debated, even though it has been probed by active seismic profiling. We compute a high-resolution shear-wave velocity model of the Ligurian-Provence basin and its margins by making optimal use of ambient-noise recordings of seafloor broadband seismometers. In particular, we improve the usually low quality of surface-wave signals in noise correlations between seafloor stations by involving correlations with land stations. The joint interpretation of our S-wave velocity model with a P-wave velocity section obtained in the basin axis by controlled-source seismic profiling provides compelling evidence for the presence of a thick sediment pile above a thin, ∼4.5 km-thick oceanic crust, intruded by gabbroic bodies emplaced at the crust-mantle transition. These results show the potential of a joint interpretation of P- and S-wave velocity models since they provide reliable answers to a number of debated questions on the petrological nature of the crust and uppermost mantle, in particular in the Ligurian-Provence basin.

International audience; Main objectives This study analyses the influence of different model parameterizations in the elastic full-waveform inversion of multi-component ocean-bottom seismic data, so as to design an effective FWI workflow to reconstruct P-and S-wave velocity models of the subsurface jointly and robustly, applicable in scenarios of both weak and strong elastic effects. New aspects covered We analyse the pros and cons of three different model parameterizations in the elastic full-waveform inversion and build hierarchical FWI workflows over different parameter classes and data components for multi-component ocean-bottom seismic data. Poisson's ratio is used as the indicator to design the overburden models that produce elastic effects varying from weak to strong in the robustness test of the proposed inversion workflows. An effective hierarchical elastic FWI workflow is found for reconstructing P-and S-wave velocity models jointly and robustly in scenarios of both weak and strong elastic effects.

International audience; Main objectives This abstract presents how a nonlinear anisotropic diffusion filtering technique, originally designed in the field of image processing, can be used for high resolution seismic imaging by full waveform inversion. We show how the application of this technique to full waveform inversion can be done at two (potentially complementary) levels: as a structure preserving smoothing technique, in the model space, and as a low frequency enhancing filtering technique, in the data space. New aspects covered 1. Introducing nonlinear anisotropic diffusion filtering to the community of seismic imaging. 2. Showing how this can be applied as a structure preserving/enhancing smoothing technique in the model space, i.e. to the gradient in the FWI workflow, or to an estimated subsurface model, or to a seismic image obtained by migration. 3. Sowing how this can be applied as a low frequency enhancing filtering technique in the data space for FWI applications.

International audience; Main objectives A 3D modeling and inversion code named SEM46 has been developed since few years in the SEISCOPE project. Currently this code implements a coupled visco-elastic/acoustic wave propagation, with generic types of anisotropy in the solid part. For practical usage, such 3D modeling and inversion code needs to be able to exploit efficiently the performance of modern large scale computing facilities. This study aims at presenting the recent HPC implementations and performance optimizations done on this code to achieve this goal. New aspects covered (1) Presentation of several optimizations and quantification of the performance gains. (2) New porting on the TGCC center (AMD Rome partition) and new parallel efficiency assessed to compare acoustic, elastic and acoustic-elastic coupling modeling.

Notre objectif de long terme est de contribuer à la construction d'un géomodèle des Alpes. Ici nous présentons un modèle 3-D de la vitesse des ondes S de la lithosphère alpine obtenu par tomographie de bruit sismique ambiant. Pour ce faire, nous utilisons l'un des meilleurs jeux de données sismologiques disponibles aujourd'hui constitué de 4 ans d'enregistrements de plus de 900 stations permanentes européennes, des ~350 stations temporaires terrestres et fond de mer du projet européen AlpArray ainsi des ~110 stations des projets CIFALPS et CIFALPS2 (profils denses aux travers des Alpes occidentales). Nous tirons profit de la qualité de ce jeu de données pour mettre en place une méthode innovante permettant de construire des cartes de vitesse de groupe probabiliste en 4s et 150s de période à partir d'une inversion 2-D Transdimensionelle hiérarchique (Bodin et Sambridge, 2009). Pour tenir compte de la non-linéarité du problème inverse, nous avons combiné cette méthode avec la Fast Marching Methode (Rawnilson et Sambridge, 2005). Nous construisons le modèle Vs 3D probabiliste à partir des cartes de vitesse de groupe en inversant à chaque pixel la courbe de dispersion locale par une inversion Bayésienne. Un traitement spécifique a été appliqué sur les données brutes de fond de mer afin de supprimer les bruits océaniques et instrumentaux (Crawford et Webb, 2000). Ce modèle sera utilisé pour modéliser la propagation des ondes de surface issues de corrélations de bruit et des ondes de volumes issues de séismes régionaux à partir des outils développés dans le cadre du consortium SEISCOPE. Ces modèles sismologiques serviront d’appuis pour produire un géomodèle 3D de référence à l’échelle des Alpes occidentales dans le cadre du chantier RGF, Alpes et bassins périphériques.