Search

International audience; From numerous studies, geophysical methods based on seismic surveying appear to be well-adapted to investigate the morpho-structure of landslides and to progress in understanding the related mechanisms. Indeed, these methods allow direct and non-intrusive measurements of acoustic (Vp) or shear (Vs) wave velocity, two important physical parameters for estimating mechanical properties of reworked moving materials. Different processing techniques and inversion strategies were applied on the La Valette and Super-Sauze mudslides (French South Alps) as well as on the Ballandaz landslide (Savoie, France) to retrieve these parameters. On each of these sites, measurements were recorded along 2D profiles of several hundred meters length, with sensor spacing from 2 to 5m. (of about few meters). A first approach, based on first breaks acoustic inversion for estimating Vp distribution on the Super-Sauze and La Valette sites was carried out; then, SASW (spectral analysis of surface waves) was performed to image Vs distribution on the same site. In order to produce a more geotechnical diagnosis of these sections, a fuzzy logic fusion was used to assimilate both of these parameters into a highest level of interpretation. This approach has (also) the advantage to take into account the resolution and accuracy of each individual method. Finally, a 2D elastic full-waveform inversion test was carried out on a synthetic seismic data set computed from a conceptual Super-Sauze velocity model. This test highlights the difficulty to manage highly contrasted media in terms of velocity but also of topography. Nevertheless, the integration in the inversion process of the whole seismic signal produce a more coherent model in terms of Vp and Vs distribution compared to above-cited conventional techniques.

Swelling soils contain clay minerals that change volume with water content and cause extensive and expensive damage on infrastructures. Based on spatial distribution of infrastructure damages and existing geological maps, the Bureau de Recherches Géologiques et Minières (BRGM, the French Geological Survey) published in 2010 a 1:50 000 swelling hazard map of France. This map indexes the territory to low, intermediate, or high swell susceptibility, but does not display smallest and isolated clays lithologies. At local scale, identification of clay minerals and characterization of swell potential of soils using conventional soil analysis (DRX, chemical, and geotechnical analysis) are slow, expensive, and does not permit integrated measurements. Shortwave infrared (SWIR:1100-2500 nm) spectral domains are characterized by significant spectral absorption bands that provide an underused tool for estimate the swell potential of soils. Reflectance spectroscopy, using an ASD Fieldspec Pro spectrometer, permits a rapid and less expensive measurement of soil reflectance spectra in the field and laboratory. In order to produce high precision map of expansive soils, the BRGM aims to optimize laboratory reflectance spectroscopy for mapping swelling soils. Geotechnical use of laboratory reflectance spectroscopy for local characterization of swell potential of soils could be assessable from an economical point of view. A new high resolution airborne hyperspectral survey (covering ca. 280 km², 380 channels ranging from 400 to 2500 nm) located at the W of Orléans (Loiret, France)will also be combined with field and laboratory measurements to detect and map swelling soils.

The detection of underground cavities is still a challenging problem despite the good potential of geophysical techniques. The seismic surface waves analysis method can be acceptable for the cavity localisation if observables in data are affected by the void. We propose here a combined approach based on numerical 3D simulations (Spectral Finite Element Method) and reduced-scale experimental data to assess the possible observables affected by the cavity presence. The measurement bench MUSC is designed to effectuate seismic data measurements on reduced scale models on the two component of a P-SV configuration. We highlight different observables all visible in numerical or experimental cases: 1) the diffraction event is higher on the horizontal component; 2) there is a perturbation of the direct wavefield amplitude above the cavity and after the cavity position; 3) the analyse of the elliptic particular movement is perturbed in terms of amplitude and direction above and few meters after the cavity. Test results demonstrate the capability of using different observables of the Rayleigh wavefield in presence of a cavity to detect the object. These qualitative characteristics confirm the interest to study and compare observables on the horizontal and vertical components in a medium containing a cavity.

The approach presented here deals with the inversion of Rayleigh wave dispersion (SASW: Spectral Analysis of Surface Waves) in laterally heterogeneous media. The traditional method consists of computing the frequency versus phase-velocity curve, known as the dispersion curve, from seismic records, then inverting to obtain a 1D shear-wave velocity model. When media are laterally heterogeneous, S-wave velocity changes lead to phase variations in Rayleigh waves in the offset dimension. This phenomenon can drastically alter the dispersion image, since it affects the local slopes of Rayleigh waves in the shot gather, and thus their phase-velocity dispersion properties. In the case of multifold acquisitions (2M-SASW: Multifold and Multichannel SASW), we redefine the manner in which the dispersion-image calculation is formulated in order to compute the local dispersion stack (LDS). In principle, this consists of gathering receivers within a restricted window for a series of shots and computing the local dispersion images that are then stacked in order to improve the signal-to-noise ratio. Even though local dispersion diagrams are noisy because they are computed from a limited number of traces, the summation over multiple dispersion images enables the quality of the final LDS to be improved. Because it is related to the dispersion properties of the windowed wavefield, the LDS is an efficient input for the 1D inversion process. The method is tested on synthetic data to demonstrate its contribution compared to that of the traditional SASW technique. An application of the 2M-SASW method to the Super-Sauze earthflow confirms that is well-suited to inverting the shear-wave velocity from Rayleigh-wave dispersion when high-contrast media are considered.

International audience; Guided waves are the major source of coherent noise, either on terrestrial or on marine seismic data, because the signals they produce are much stronger in amplitude than the reflected ones. In marine contexts, these waves exhibit characteristics that depend on the water depth, on the geometry and on the material properties of the substrata. Guided waves - or ground roll - are dispersive, which constitutes their main property. This means that each frequency component of the wave travels at a different velocity, in the sense that smaller and larger wavelengths are respectively influenced by the seismic properties of the shallower and deeper parts of the media. Due to their linear moveout-versus-offset characteristics, it may be possible to suppress the guided waves by dip filtering techniques. Unfortunately, this kind of filtering causes serious distortion of the signal when the amplitude of guided waves is much stronger than that of reflection (Liu 1999). The method developed here consists of extracting the guided waves from common-shot gathers without disturbing the reflection signals.

International audience; Frequency analysis of surface waves propagation is an efficient tool to retrieve the vertical shear-wave velocity profile. Surface waves result from the interaction of elastic body waves with a free surface. They are commonly generated by impulsive active sources but are also present in ambient noise. Here we compare the results of interferometric estimation of surface-wave dispersion from controlled-noise sources by using both cross-correlation and deconvolution methods. The crucial step in both methods lies in the dispersion image computation that conditions the reliability of the dispersion curve, and therefore the inversion results. We have found that the dispersion characteristics of surface waves are better defined when the deconvolution method is used. This better definition is probably due to the relative independence of the deconvolved output from the source function.

Quantifies penetration capabilities of radar systems in arid areas, for sub-surface imaging purposes. Data acquisition systems considered range from ground penetrating radars to airborne and orbital synthetic aperture radars. First fieldwork results are presented for two study sites: one is located in France (Pilat dune near Bordeaux) and the other one is located in East Africa (Republic of Djibouti). Data obtained are used to constrain theoretical radar backscattering models, including both surface and volume scattering terms.

International audience; Various studies operated seismic methods for imaging landslide structures. Among them, different approaches can be used to process the data, analyze and exploit the different kind of waves associated to particular propagation phenomena. Ferucci et al. (2000) used the late arrivals of P-waves to produce a reflective seismic section of the landslide. Glade et al. (2005) used the refracted waves to delimit main landslide's layers. Grandjean et al. (2011) studied the first arrival traveltimes to recover P-wave velocity distribution across landslides. Finally, based on the work of Virieux and Operto (2009), Romdhane et al. (2011) showed the possibility to perform a full elastic waveform inversion (FWI) on a real dataset acquired over an earthflow. All those methods are more or less based on strong approximations and require complex data preprocessing. The issue of recovering the structural image of a landslide from the seismic velocity field estimated with an accurate, but not too unstable, method is thus posed. To solve it, we choose to revisit first arrival tomography approach which is a good compromise between the strong assumptions featuring simple refraction methods and the complexity of FWI approach when used in very heterogeneous soils. The proposed method is based on a gradient / Hessian formulation (Tarantola, 1987) to ensure an optimum convergence of the velocity model. We only use here the first arrivals of the seismic signal due to direct or refracted waves. Nevertheless, we show that some regularization strategies allow detecting sharp velocity variations tending to reach FWI inversion performances.

Hydrogeophysics typically consists in the combination of hydrological and geophysical methods for the understanding of a hydrogeological system. Landslides are typical water catchments which benefit from such a combined approach. Among key parameters that can provide time-lapse sections of the underground and help the modeling, we consider the electric conductivity for its sensitivity to soil water content. In the present work, we consider a new electrical monitoring experiment on the Super-Sauze landslide, running over a longer period for the observation of natural rainfall effect. We present experimental set up: the GEOMON4D device developed by the Austrian Geological Survey, and several sensors that measure continuously the parameters affecting apparent electrical resistivity. Then we show the first results in terms of electric resistivity but also streaming-potential (SP) and discuss about perspectives. The first results indicate that apparent and inverted resistivities show low variations correlated with rainfall and large variations correlated with the groundwater table level. The SP data could be used to provide information on the infiltration processes and top soil saturation. We plan to use the SP data to understand the complex relation between rainfall and soil water content in the top soil.

We tackle the problem of characterizing the subsurface, more specifically detecting shallow buried objects, using seismic techniques. This problem is commonly encountered in civil engineering when cavities or pipes have to be identified from the surface in urban areas. Our strategy consists of processing not only first arrivals, but also later ones, and using them both in tomography and migration processes, sequentially. These two steps, which form the basis of seismic imaging, can be carried out separately provided that the incident and diffracted wavefields are separated in the data space. Tomography is implemented here as an iterative technique for reconstructing the background velocity field from the first-arrival traveltimes. The later signals are then migrated by a Kirchhoff method implemented in the space domain. To study the reliability of this methodology, it is first applied to synthetic cases in the acoustic and elastic approximation. Both the background velocity field and the local impedance contrasts are reconstructed as defined in the predicted model. An experimental case, specifically designed for the purpose, is then considered in order to test the algorithms under real conditions. The resulting image coincides well with the predicted model when only P-waves are generated. In the elastic mode, surface waves make P-wave extraction difficult, so that the reconstruction remains incomplete. This is confirmed by the real data example. Finally, we demonstrate the appropriateness of the proposed method under such circumstances, provided that suitable preprocessing of data is carried out, in particular, the removal of surface waves.

Received 15 February 2002; revised 20 July 2002; accepted 28 August 2002; published 6 March 2003. [1] Low-frequency sounding radars should be able to probe the Martian subsurface layers down to varying depths, depending on the geoelectrical properties of the sounded sites. We present in this work four frequency-dependent geoelectrical models of the Martian subsurface in the 1–20 MHz frequency band, based on laboratory electromagnetic characterization of Martian soil analogues. Those models correspond to local Martian sites that we considered to be of particular interest in the search for water using mainly the Ground-Penetrating Radar (GPR) instrument of the Netlander mission. Results and discussion are also valid for both sounding experiments MARSIS and SHARAD. The four models of the Martian subsurface are designed to represent terrains where recent fluviallike features suggest the presence of near-subsurface ground ice and probably liquid water. We performed measurements on volcanic and sedimentary materials that may be present on these sites under the appropriate geophysical conditions that may exist in those terrains. We then simulated the backscattered radar echo arising from each site in the 2 MHz frequency band, using the Finite Difference Time Domain (FDTD) algorithm, in order to evaluate the instrument performances to probe the subsurface stratigraphy of each site. Our results confirm that the near-subsurface rich iron oxide mineralogy controls the instrument performances in terms of penetration depth and signal-to-noise ratio in the 2 MHz frequency band. We finally discuss the geophysical and geoelectrical sounding conditions that could lead to an ambiguous detection of shallow subsurface water on Mars for the Netlander GPR. INDEX TERMS: 3210 Mathematical Geophysics: Modeling; 1794 History of Geophysics: Instruments and techniques; 5144 Physical Properties of Rocks: Wave attenuation; 5109 Physical Properties of Rocks: Magnetic and electrical properties; KEYWORDS: Mars, hydrology, GPR, sounding, simulation, FDTD

Mineralogical clays are intimately mixed in soils. Until now, there is not a clear view on which unmixing methods are the most effective in case of intimate mixtures with both clays and with other mineral. A laboratory approach is proposed in order to compare existing linear and nonlinear unmixing approaches in terms of performances. Laboratory images of pure clays of Montmorillonite, Illite and Kaolinite were preprocessed using 6 transformations commonly used in literature (SNV, CR, CWT, Hapke, 1st SGD, Log(l/R)). Prediction of abundances has been performed on non-preprocessed and preprocessed data with 2 linear unmixing methods (FCLS and MESMA) and 2 nonlinear unmixing methods (GBM and MLM). Results show a decrease of unmixing performance due to sample variability, which can be reduced using the appropriate preprocessing. However, no particular advantage has been found on using nonlinear over linear unmixing approaches for clay minerals.

L'étude des sols argileux fait l'objet de nombreux travaux motivés par leur rôle dans les processus d'érosion, les catastrophes naturelles et l'agriculture de précision. La caractérisation du contenu en argiles gonflantes du sol est aussi nécessaire pour évaluer la traficabilité d’une région ou le risque de retrait-gonflement des sols, responsable d’engendrer des dégâts sur le bâti. En effet, les argiles gonflantes sont des smectites qu'il faut différencier des autres types d'argiles telles que l'illite ou la kaolinite, en milieu tempéré. Les techniques traditionnelles pour réaliser la cartographie des minéraux argileux des sols sont en général couteuses, financièrement et en temps et sont basées sur des campagnes terrain intensives simultanément à des acquisitions photographiques afin de spatialiser l'information qui reste qualitative. La télédétection hyperspectrale est une technique potentiellement intéressante pour obtenir des cartes d'argile plus précises et à moindre coût. Néanmoins, elle est limitée par le fait que (i) les minéraux sont mélangés de manière intime dans les sols avec d’autres composants; mais aussi que (ii) à l’échelle aéroportée, le signal réfléchi au sein d’un pixel (résolution spatiale de l’ordre du mètre) comprend de la végétation en plus du sol nu. Ces phénomènes de mélange, aux échelles microscopique et macroscopique, rendent difficile l’estimation du contenu en argiles minéralogiques. Le développement des drones ainsi que leur possibilité d'embarquer de nouvelles caméras hyperspectrales couvrant l'ensemble du spectre [0,4 - 2,5 µm] avec une haute résolution spatiale (environ 10 cm) et un signal à bruit élevé ouvrent la voie à un inventaire plus précis des argiles.L'objectif de cette thèse est de montrer l'intérêt de l'utilisation de méthodes de démélange sur des données hyperspectrales à très haute résolution spatiale pour estimer le contenu en minéraux argileux du sol, et plus précisément des argiles responsables du retrait-gonflement, les smectites. Dans un premier temps, les méthodes existantes de détection, de caractérisation des différents types d'argile et d'estimation de leur abondance sont présentées. Les potentialités des méthodes de démélange existantes dans la littérature pour l’estimation du contenu en minéraux argileux des sols sont mises en avant. Dans un second temps, les méthodes de démélange sont utilisées sur une base de données d’images hyperspectrale acquises en laboratoire de mélanges contrôlés minéraux contenant des argiles (montmorillonite, illite, kaolinite) et d’autres minéraux présents dans les sols (quartz, calcite). Comme les minéraux sont mélangés de manière intime, des méthodes de démélange, linéaires et non-linéaires sont décrites et comparées. Néanmoins, les algorithmes non-linéaires ont des performances similaires aux algorithmes linéaires. De plus, l’effet de la variabilité des données sur la précision de l’estimation des abondances a pu être réduit en utilisant des prétraitements spectraux. Dans une dernière étape, la comparaison des méthodes de démélange sont étendues à des mesures en environnement extérieur. Cette analyse repose sur une campagne de mesure en extérieur réalisée pour la mesure d'images hyperspectrales acquises depuis une nacelle (12 m de hauteur environ, 1,5 cm de résolution spatiale), et l’acquisition d’échantillons prélevés et analysés par DRX (données quantitatives des abondances des minéraux) pour validation. Cette dernière phase permet d'analyser l'impact d'un sol naturel (composé d'un mélange minéralogique, des matières carbonées telles que la cellulose, et ayant une rugosité de surface…) sur les méthodes de démélange. Les performances obtenues (moins de 15% RMSE sur l’estimation de la montmorillonite) permettent d’ouvrir des perspectives quant à l’application de ces méthodes sur des capteurs embarqués par drone, pour la cartographie de la traficabilité et de l’aléa de retrait-gonflement des sols. Clayey soils are studied because of the importance of soils in erosion processes, natural disasters and precision agriculture. Mapping of clay mineralogy is essential for surveying and predicting trafficability and ground instability hazards, such as shrink-swelling, in order to cope with damages caused by expansive soils on infrastructures. Clay minerals in temperate zone soils are mainly divided in smectites, which highly contribute to soil swelling, illite and kaolinite. Geotechnical engineering practice for clayey soils mapping are expensive and time-consuming. Indeed, it is based on field and extensive laboratory studies. In addition, spatial distribution of clay is assessed using aerial photographs and low-scale geological maps. Thereby, small heterogeneities in geological features are rarely detected, and spatial information remains qualitative. Hyperspectral remote sensing could be an alternative to conventional methods for clay mapping. However, this method is limited by two facts: (i) soils are an intimate mixture of minerals, and (ii) vegetation is mixed with bare soil within airborne sensors pixels (meter range of spatial resolution). Those mixtures (at microscopic and macroscopic scales) mask clays specific spectral signatures and limit clay mineral quantification. Recent development in UAV offers new possibilities for carrying hyperspectral cameras in the reflective domain [0.4 – 2.5 µm], and obtaining data with higher SNR and resolution (10 cm). These advances open new perspectives for accurate and less expensive clay maps. This PhD thesis aims to present the potentiality of clay mapping in soils using very-high spatial resolution hyperspectral data, and more specifically, to estimate swelling clay minerals (smectite) abundances. First, existing methods of detection, and abundance estimation of clay minerals are presented. Second, unmixing methods are used on a database of hyperspectral images of controlled mixtures with different abundances of clay minerals, (illite, montmorillonite and kaolinite) and other minerals existing in soils (quartz, calcite). Due to the intimate nature of mixtures, linear and non-linear unmixing methods are described and compared. However, linear and nonlinear algorithms exhibit similar performances. Moreover, the accuracy of estimation of abundances of mineral clay increased using spectral preprocessings. Regarding the third step, field measurements are used to assess clay unmixing methods. This study is based on an outdoor experiment which acquired hyperspectral images from a bucket truck (12 m elevation, 1.5 cm ground sampling distance), and a sampling collection analyzed by XRD (quantitative analysis of mineral abundances) for validation. This step analyzed effects of a natural soil, (with organic matter, a larger diversity of mineralogical components and with surface roughness) on unmixing methods tested with laboratory data. Obtained performances (less than 15% RMSE for montmorillonite estimation) allow perspectives to apply these methods on data obtained with UAV sensors, for trafficability or expansive soils mapping purposes.

Clayey soils are studied because of the importance of soils in erosion processes, natural disasters and precision agriculture. Mapping of clay mineralogy is essential for surveying and predicting trafficability and ground instability hazards, such as shrink-swelling, in order to cope with damages caused by expansive soils on infrastructures. Clay minerals in temperate zone soils are mainly divided in smectites, which highly contribute to soil swelling, illite and kaolinite. Geotechnical engineering practice for clayey soils mapping are expensive and time-consuming. Indeed, it is based on field and extensive laboratory studies. In addition, spatial distribution of clay is assessed using aerial photographs and low-scale geological maps. Thereby, small heterogeneities in geological features are rarely detected, and spatial information remains qualitative. Hyperspectral remote sensing could be an alternative to conventional methods for clay mapping. However, this method is limited by two facts : (i) soils are an intimate mixture of minerals, and (ii) vegetation is mixed with bare soil within airborne sensors pixels (meter range of spatial resolution). Those mixtures (at microscopic and macroscopic scales) mask clays specific spectral signatures and limit clay mineral quantification. Recent development in UAV offers new possibilities for carrying hyperspectral cameras in the reflective domain [0.4 – 2.5 mm], and obtaining data with higher SNR and resolution (10 cm). These advances open new perspectives for accurate and less expensive clay maps. This PhD thesis aims to present the potentiality of clay mapping in soils using very-high spatial resolution hyperspectral data, and more specifically, to estimate swelling clay minerals (smectite) abundances. First, existing methods of detection, and abundance estimation of clay minerals are presented. Second, unmixing methods are used on a database of hyperspectral images of controlled mixtures with different abundances of clay minerals, (illite, montmorillonite and kaolinite) and other minerals existing in soils (quartz, calcite). Due to the intimate nature of mixtures, linear and non-linear unmixing methods are described and compared. However, linear and nonlinear algorithms exhibit similar performances. Moreover, the accuracy of estimation of abundances of mineral clay increased using spectral preprocessings. Regarding the third step, field measurements are used to assess clay unmixing methods. This study is based on an outdoor experiment which acquired hyperspectral images from a bucket truck (12 m elevation, 1.5 cm ground sampling distance), and a sampling collection analyzed by XRD (quantitative analysis of mineral abundances) for validation. This step analyzed effects of a natural soil, (with organic matter, a larger diversity of mineralogical components and with surface roughness) on unmixing methods tested with laboratory data. Obtained performances (less than 15 % RMSE for montmorillonite estimation) allow perspectives to apply these methods on data obtained with UAV sensors, for trafficability or expansive soils mapping purposes., L’étude des sols argileux fait l’objet de nombreux travaux motivés par leur rôle dans les processus d’érosion, les catastrophes naturelles et l’agriculture de précision. La quantification des argiles gonflantes est aussi nécessaire pour évaluer la traficabilité d’une région ou le risque de retrait-gonflement des sols, responsable d’engendrer des dégâts sur le bâti. En effet, les argiles gonflantes sont des smectites qu’il faut différencier des autres types d’argiles telles que l’illite ou la kaolinite, en milieu tempéré. Les techniques traditionnelles pour réaliser la cartographie des minéraux argileux des sols sont en général couteuses, financièrement et en temps, et sont basées sur des campagnes terrain. La télédétection hyperspectrale est une technique intéressante pour obtenir des cartes d’argile plus précises et à moindre coût. Néanmoins, elle est limitée par le fait que (i) les minéraux sont mélangés de manière intime dans les sols avec d’autres composants ; mais aussi que (ii) à l’échelle aéroportée, le signal réfléchi du sol nu est mélangé à celui de la végétation au sein d’un pixel (résolution spatiale de l’ordre du mètre). Ces phénomènes de mélange, aux échelles microscopique et macroscopique, rendent difficile l’estimation du contenu en argiles minéralogiques. Le développement des drones ainsi que leur possibilité d’embarquer de nouvelles caméras hyperspectrales couvrant l’ensemble du spectre [0,4 - 2,5 mm] avec une haute résolution spatiale (environ 10 cm) et un signal à bruit élevé ouvrent de nouvelles perspectives. L’objectif de cette thèse est de montrer l’intérêt de l’utilisation de méthodes de démélange sur des données hyperspectrales à très haute résolution spatiale pour estimer le contenu en minéraux argileux du sol, et plus précisément des argiles responsables du retrait-gonflement, les smectites. Dans un premier temps, les potentialités des méthodes de démélange existantes dans la littérature pour l’estimation du contenu en minéraux argileux des sols sont présentées. Dans un second temps, les méthodes de démélange sont utilisées sur une base de données d’images hyperspectrale acquises en laboratoire de mélanges contrôlés minéraux contenant des argiles (montmorillonite, illite, kaolinite) et d’autres minéraux présents dans les sols (quartz, calcite). Comme les minéraux sont mélangés de manière intime, des méthodes de démélange, linéaires et non-linéaires sont décrites et comparées. Néanmoins, les algorithmes non-linéaires ont des performances similaires aux algorithmes linéaires. De plus, l’effet de la variabilité des données sur la précision de l’estimation des abondances a pu être réduit en utilisant des prétraitements spectraux. Dans une dernière étape, la comparaison des méthodes de démélange sont étendues à des mesures en environnement extérieur. Cette analyse repose sur une campagne de mesure en extérieur réalisée pour la mesure d’images hyperspectrales acquises depuis une nacelle (12 m de hauteur environ, 1,5 cm de résolution spatiale), et l’acquisition d’échantillons prélevés et analysés par DRX (données quantitatives des abondances des minéraux) pour validation. Cette dernière phase permet d’analyser l’impact d’un sol naturel (composé d’un mélange minéralogique, des matières carbonées telles que la cellulose, et ayant une rugosité de surface. . . ) sur les méthodes de démélange. Les performances obtenues (moins de 15 % RMSE sur l’estimation de la montmorillonite) permettent d’ouvrir des perspectives quant à l’application de ces méthodes sur des capteurs embarqués par drone, pour la cartographie de la traficabilité et de l’aléa de retrait-gonflement des sols.

Full waveform inversion has shown a great suitability in the acoustic case for imaging large and deep kilometric structures. Contrary to first traveltimes tomography, this imaging technique allows to process the whole seismic signal in order to produce a high resolution image of the medium. In the particular context of subsurface prospecting, the elastic approximation needs to be considered to take into account the surface waves which constitute the most part of the seismic signal, with the offset range used for such studies. In our approach, the forward problem is numerically solved in the frequency domain in order to take into account a multi-source and multi-receiver seismic device efficiently. In front of the difficulties inherent to the finite difference method with regard to complex topographies, a formulation based on finite elements discontinuous Galerkin technique is finally used. The inverse problem requires an initial guess of the inverted parameters to minimize the objective function in an iterative way. The method is first applied to simple numerical models to assess the role of the different parameters which control the inversion process. A synthetic model of the Super-Sauze earthflow, characterized by strong contrasts and a complex topography is afterwards considered to evaluate the performances of the code. This study allows us to highlight the impact of the optimisation technique used, the acquisition device, the quality of the initial model and the contribution of the preconditioning strategy used in the inversion process. The approach is then applied to examples using real data acquired over the Jargeau site (45) in the presence of a cavity, and over the Super-Sauze earthflow (04). The difficulties inherent to real conditions require an important data pre-processing. The source estimation is also a critical issue that directly impacts the convergence of the inversion. In the first case, the proposed methodology succeeds to reconstruct an image of the seismic parameters distributions (Vp and Vs) with a relatively good accuracy, the final quality of these images being dependant on the efforts produced to constrain the inversion. The second case reveals the difficulties related to complex media ; seismic data show a low signal to noise ratio, the topography is very sharp and generates a strong level of diffracted surface waves, the seismic source varies from one shot to another. Despite the complexity of the context, the inversion results are in good agreement with the geotechnical knowledge of the site., L'inversion des formes d'onde sismique a démontré toute sa pertinence dans le cas acoustique pour imager des structures kilométriques. Contrairement aux méthodes de tomographie des premiers temps d'arrivée, cette technique d'imagerie à l'avantage d'exploiter l'ensemble du signal sismique pour produire une image haute résolution du milieu. Dans le contexte de la proche surface, l'approximation élastique devient nécessaire afin de prendre en compte les ondes de surface, qui prédominent le signal sismique avec les gammes de déports utilisées. Dans notre approche, le problème direct est résolu numériquement dans le domaine fréquentiel afin de prendre en charge efficacement des acquisitions multisources et multi-récepteurs. Devant les complications générées par une approche en différences finies pour traiter les topographies complexes, une formulation en éléments finis Galerkin discontinus est finalement utilisée. Le problème inverse nécessite un modèle initial pour les paramètres inversés et permet de minimiser la fonction objectif de façon itérative. Cette méthode est appliquée dans un premier lieu à des modèles numériques simples afin de contrôler le rôle des différents paramètres de l'inversion. Un modèle synthétique du glissement de Super-Sauze, caractérisé par de forts contrastes et une topographie complexe, est par la suite utilisé pour évaluer les performances du code. Cette étude met en évidence l'importance de la méthode d'optimisation utilisée, du dispositif d'acquisition, de la qualité du modèle initial et de l'apport des stratégies de pré-conditionnement utilisées dans le processus d'inversion. La méthode est ensuite appliquée à deux jeux de données réelles, acquis respectivement au-dessus d'une cavité maçonnée à Jarjeau (45), et sur le glissement de terrain de Super-Sauze (04). Les difficultés relatives aux conditions réelles nécessitent un prétraitement important des données. L'estimation de la source constitue aussi un problème majeur qui impacte directement la convergence de l'inversion. Dans le premier cas, la méthode proposée permet de produire une image plus ou moins fidèle des distributions des vitesses du milieu, la qualité finale de ces images variant suivant les efforts mis en oeuvre pour contraindre l'inversion. Le deuxième cas illustre quant à lui toutes les difficultés relatives aux milieux complexes : les données montrent un faible rapport signal sur bruit, la topographie très découpée génère un fort taux de diffraction des ondes de surface, la signature de la source admet de fortes variations d'un tir à l'autre. Malgré la complexité du contexte, les résultats de l'inversion des formes d'onde sont en bon accord avec la connaissance géotechnique du site.

Imaging the soil physical parameters with surface seismic recordings is a non linear inverse problem. The elastic full waveform inversion is a quantitative multiparameters imaging method of diffractors that requires beforehand the knowledge of an initial velocity model. The corresponding forward problem, the elastic waves propagation, is solved in the frequency domain, allowing to efficiently take into account multisources and multireceivers acquisitions, with a finite difference method that modelizes the propagation of all wave types (body waves, surface waves, diffrated waves, refracted waves ...). The chosen finite difference stencil precisely simulates the free surface and surface waves propagation. This linearized inversion is based on a gradient method that minimizes a misfit function containing the difference between observed and calculated data. Two seismic parameters are reconstructed from vertical and horizontal seismograms. The inversion is performed from low to high frequencies, allowing to progressively introduce higher wavelengths in the parameters images. The latter are sensitive to the choice of the physical approximations chosen to calculate the misfit function gradient (Born or Rytov approximation), to the acquisition geometry, to the data preconditionning and to the inverted parameters choice. This method is then applicated to media containing a free surface. The free surface is a very contrasted interface that generates very energetic surface waves in seismograms. Near-surface media are complex, propagating waves are reflected or diffrated several times. When the medium contains two anomalies whose velocity contrast is equal to $20 \%$, the inversion correctly localize them in a known backgound model. Images are improved and anomalies amplitudes are very well recovered when the number of inverted frequencies increases and the data are selected from low to high offset. With a strongly contrasted anomaly like an empty cavity, the inversion correctly recovers the object position, shape and size but its amplitude is underestimated.The application to subsurface real vertical data recorded in a complex medium containing a cavity shows that the inversion constructs a heterogeneous medium allowing to better fit data and allows to reproduce inverted waves. Nevertheless the cavity is not imaged., L'imagerie des paramètres physiques du sous-sol à partir d'enregistrements sismiques de surface constitue un problème inverse non linéaire. L'inversion des formes d'onde élastique est une méthode d'imagerie quantitative multiparamètres de diffractants, nécessitant au préalable la connaissance précise d'un macromodèle de vitesse. Le problème direct associé, la propagation des ondes élastiques, est résolu dans le domaine fréquentiel, permettant la prise en compte efficace d'acquisitions multisources et multirécepteurs, par une méthode numérique de différences finies modélisant la propagation de tous les types d'onde (ondes de volume, de surface, diffractées, réfractées ...). Le stencil de différences finies choisi permet simule précisément la surface libre et la propagation des ondes de surface. Cette inversion linéarisée s'appuie sur une méthode de gradient, qui minimise une fonction coût contenant les différences entre données observées et calculées. Deux paramètres sismiques sont reconstruits à partir de sismogrammes verticaux et horizontaux. L'inversion est effectuée des basses fréquences vers les hautes fréquences, introduisant des longueurs d'onde de plus en plus courtes dans les images des paramètres. Ces dernières sont sensibles au choix des approximations physiques effectuées pour calculer le gradient de la fonction coût (approximation de Born ou de Rytov), au dispositif d'acquisition, au préconditionnement des données et au choix des paramètres inversés. Cette méthode est ensuite appliquée à des milieux contenant une surface libre. La surface libre est une interface très contrastée qui donne lieu à des ondes de surface très énergétiques dans les sismogrammes. Les milieux de subsurface sont complexes, les ondes qui s'y propagent subissent des réflexions ou diffractions multiples. Lorsque le milieu contient deux anomalies dont le contraste en vitesse vaut $20 \%$, l'inversion les localise correctement dans un macromodèle connu. Les images sont améliorées et l'amplitude des anomalies est très bien reconstruite lorsque le nombre de fréquences inversées augmente et les données sont sélectionnées des faibles déports vers les grands déports. Avec une anomalie fortement contrastée comme une cavité vide, l'inversion retrouve correctement la position, la forme et la taille de l'objet mais son amplitude est sous-estimée.L'application à des données réelles verticales de subsurface acquises dans un milieu complexe contenant une cavité maçonnée montre que le milieu hétérogène issu de l'inversion ajuste mieux les données et permet de bien reproduire les ondes inversées. Néanmoins, la cavité n'est pas imagée.