Search

3D CSEM modeling is usually performed by finite-difference (FD) discretization, but new players have entered the scientific community with finite-element (FE) modeling codes. FDs are suitable for simple cases, but are unsuitable for complex geometries, while FEs are very effective in the presence of high topography and/or bathymetry variations. The open-source 3D FE modeling code custEM has demonstrated its efficiency and reliability and is freely available to the scientific community. The recent work done by its development team has made it possible to integrate the inversion part with the Guauss Newton minimization algorithm, but there is not yet enough user feedback and tools available on this new module to be operational for the general public. On the other hand, POLYEM3D is a 3D FD modeling and inversion code that has a wide range of inversion algorithms and reparametrization and regularization tools but faces numerical errors in presence of high topography or bathymetry variations such as in coastal areas or volcanic environments due to the sea/land interface and large variations in near-shore bathymetry and topography. We tested a mixed FE and FD approach that aims to reduce the modeling errors inherent in the FD scheme in such complex problems. A secondary field formulation is available in POLYEM3D and instead of classically using a semi-analytical 1D code to compute the primary field, we use 3D fields computed in a simple isotropic homogeneous model with custEM incorporating topographic and bathymetric variations. The accuracy of the mixed approach was evaluated by modeling with a simple synthetic model incorporating topographic and bathymetric variations, and comparing the responses of the FE and mixed FE/FD codes. In addition, an inversion on a complex CSEM case demonstrated the good behavior of the mixed approach in inversion thanks to the panel of reparametrization and regularization tools offered by POLYEM3D.

Exploring for geothermal resources using land-based electromagnetic methods can be very challenging due to the presence of strong human-generated and "geological" noise. In such context, some passive EM techniques like the Magneto-Telluric method may not be applicable at all and a dedicated toolbox of EM techniques capable of dealing with these issues is required. In this paper, we will focus on the challenges encountered while exploring for geothermal resources in volcanic islands, namely a highly heterogeneous near-surface creating strong static-shifts, the presence of a sea/land transition zone and a high degree of urbanization over the area of interest. We will show that the combination of airborne with land-based controlled-source EM (CSEM) and MT surveying provide a robust approach and illustrate these aspects with two EM datasets acquired over the Martinique Island, in the French Lesser Antilles. The first study extends over the Lamentin bay area located near Fort de France, the main city of the Martinique Island. Here, exploratory drilling since the 1980's highlighted a polyphased hydrothermal system with fossilized hydrothermal features and a recent hydrothermal activity. The geothermal system is poorly understood and has prevented any exploitation of the geothermal resource to date. In an attempt to reduce the geological risk for future geothermal development, a joint 3D airborne and CSEM survey was performed. Here, the presence of anthropic EM noise made MT data hardly usable. Processing and inversion of the EM dataset allowed to map the geometry of the hydrothermal alteration "caprock" as well as underlying volcanic substratum and potential geothermal reservoir. The second study extends over the Petite-Anse area located in the SouthWest of the Martinique island. Here, evidences for high-temperature geothermal system (>200°C) are present and the design of exploratory boreholes requires to map accurately the geometry of the "caprock" and potential geothermal reservoir. To do so, a high-resolution 3D CSEM and MT survey was conducted in Spring 2019. Similarly to the Lamentin area, MT data suffered from strong cultural noise. CSEM data quality is sufficiently good to map the extent of the area of geothermal interest and areas to avoid drilling into (e.g. dykes).

International audience; The transition zone between the basement and the sedimentary cover is becoming an increasingly attractive target for the development of geothermal energy in deep sedimentary basin as encountered in the Upper Rhine Graben. Several geothermal power plants already exploit this target but the transition zone is however still poorly known with the presence of large heterogeneities. Studies need to be carried out in order to develop conceptual models on how it is formed and how it can be exploited to produce heat. In this study, we evaluate the ability of resistivity imaging by Controlled-Source Electromagnetic (CSEM) method, to identify favorable areas for the development of Enhanced Geothermal System (EGS) heat exchangers. We performed a land-CSEM survey on an analogue of the transition zone in a well-known catchment basin at Ringelbach, Vosges to assess the relevance of such data. Gathered data consist in a 3D-grid of 48 reception sites uniformly distributed over the whole basin, using a single transmitter. We performed 2.5D inversions of this dataset with the parallel adaptive finite-element code MARE2DEM to image the resistivity structure through a profile of interest and compared the result to a former Electrical Re-sistivity Tomography (ERT) inversion. CSEM inversion extends the shallow ERT image in depth and allowed to obtain a resistivity image of the transition zone. The integration of these results with existing geological and geophysical knowledge allowed identifying and mapping a fault zone as well as the fractured zone at the top of the unaltered granite basement. Results of this study demonstrate the importance of acquiring resistivity data at the target depth before drilling to maximize the success rate of a deep EGS project and point out the interest of pursuing the study with the 3D inversion of the whole set of data.

International audience; The transition zone between the basement and the sedimentary cover is becoming an increasingly attractive target for the development of geothermal energy in deep sedimentary basin as encountered in the Upper Rhine Graben. Several geothermal power plants already exploit this target but the transition zone is however still poorly known, with presence of large permeability heterogeneities. Studies in the framework of the CANTARE-Alsace project are currently on going in order to develop conceptual models on how it is formed and how heat can be exploited. In this study, we evaluate the ability of resistivity imaging by Controlled-Source Electromagnetic (CSEM) method in frequency domain, to identify favorable areas for the development of Enhanced Geothermal System (EGS). We performed a land-CSEM survey on a shallow analogue of the transition zone in a well-known catchment basin at Ringelbach (Vosges), to assess the relevance of such data. The catchment basin of Ringelbach is a well-known hydrogeological site, analogous in surface to our target in depth. According to previous geophysical studies (Two well logs, ERT…), the site is composed of resistive sandstone (>1600 Ω.m) superposed over two granitic facies with a NE-SW fault separating the site in two blocks. The first one is identified as a weathered granite being conductive (250-1600 Ω.m), whereas the second one appears as a block of fresh granite with a strong resistive signature (>1600 Ω.m). We know from former laboratory measurements (Belghoul, 2007) that conductivity of Ringelbach granites is related to its alteration degree and porosity. Furthermore, weathered granites show signs of former hydrothermal alteration succeeding the granite burial (Wyns, 2012). Fresh granitic basement is not reached by wells (150 meters depth) or former geophysical studies (Limited around 50 meters depth) The 3D Land-CSEM survey consists in a 3D-grid of 48 reception sites uniformly distributed over the whole basin and using a single transmitter located at 1.5 km north from the grid. We performed 2.5D inversions of a data subset with the 2.5D inversion code MARE2DEM (Key, 2016) to image resistivity structures through a profile of interest (Figure 1) and compared the result to a former Electrical Resistivity Tomography (ERT) inversion. 3D inversion of the entire dataset are currently ongoing using the BRGM proprietary software POLYEM3D for 3D CSEM and MT data inversion. Preliminary 3D results are consistent with the 2.5D resistivity image. CSEM inversion results extended shallow ERT images in depth and allowed to obtain a resistivity image of the transition zone analogue. Integration of these results with existing geological and geophysical knowledge allowed identifying and mapping the different resistivity blocks: the resistive sandstone; the altered granite area more conductive, altered and potentially more permeable; the resistive fresh granites with the "Bunker" block (Right on Figure 1) as well as the fresh granitic basement, less porous thus less favorable for a geothermal reservoir. Results of this analogue study demonstrate the importance of obtaining resistivity data at the target depth before drilling to maximize the success rate of a deep EGS project. Upscaling work will be part of the last work package of the CANTARE-Alsace project.

International audience; Anthropogenic noise, cost and logistical constrains generally limit to the use of land CSEM to a singletransmitter position for the deep imaging of the electrical conductivity. As the inversion of CSEM data inthe near field using a single transmitter position suffers from critical sensitivity singularities, we proposeda robust inversion framework adapted to this ill-conditioned inversion problem. The framework reliesspecifically on a robust Gauss-Newton solver, several model parameter transformations to compensate forthe heterogeneous sensitivities, and on the reformulation of the CSEM data under the form of a pseudo-MT tensor. We describe here the approach used for modelling and inversion implemented in our codePOLYEM3D and the new pseudo-MT formulation. We illustrate its application on a pathological syntheticcase inspired from Grayver et al. (2013) and then show the application of the process to a real CSEMdataset acquired in the context of thermal water prospection.

International audience; Anthropogenic noise, cost and logistical constrains generaly limit to the use of land CSEM to a few transmiter positions for the deep imaging of the electrical conductivity. The 3D inversion of CSEM data in the near field using a single transmiter position suffers from critical sensitivity singularities. We proposed a robust inversion framework adapted to this ill-conditioned inversion problem. The framework relies specificaly on a robust Gauss-Newton solver, model parameter transformations to compensate the heterogeneous sensitivies, and on the reformulation of the near field CSEM data under the form of a pseudo-MT tensor. We describe the approach used for modeling and inversion implemented in our code POLYEM3D and show the advantages of pseudo-MT tensor formulation. The strategy have been tested on a pathologic synthetic case inspired from grayver et al (2013), and then was successfully applied to a real CSEM dataset acquired in the context of thermal water prospection in a noisy environnement.

International audience; Sedimentary basins in Northwest Europe have significant potential for low to medium enthalpy, deep geothermal energy resources. These resources are generally assessed using standard seismic exploration techniques to resolve geological structures. The ElectroMagnetic campaign carried-out in Mol area (Belgium) has shown that despite the presence of high level of industrialization, the resistivity of deep formations (>3km) can be recovered from MT and CSEM methods and hence provide very valuable information for the assessment of geothermal resources. 1. INTRODUCTION Sedimentary basins in Northwest Europe have significant potential for low to medium enthalpy, deep geothermal energy resources. These resources are generally assessed using standard seismic exploration techniques to resolve geological structures. However, the electrical resistivity parameter, which can be directly impacted by the presence of a geothermal reservoir is rarely investigated in such context. Therefore, the development of alternative and complementary exploration techniques such as Electromagnetic (EM) techniques may have an important role in reducing the cost and uncertainty associated with geothermal resource assessment.

International audience; Land EM methods provide useful information in domains such as geothermal energy, CO2 storage or water ressources managements. However, the targets are usually close from urban areas where anthropogenic noise such as high power lines, industries and railways prevents from the use of Magnetotelluric (MT). Controlled-Source EM (CSEM) is thus generaly prefered to MT. However, unlike for marine CSEM surveys used in oil prospection, logistical constrains on land surveys limit to the use of a few transmitters, usually 2 polarizations at a single position. Again due to noise constrains, it is usually prefered to place the transmiter close from the stations, which make impossible the use of CSAMT processing, as far field assumption is not respected. Thus the 3D inverse problem associated to this kind of survey suffers from very high sensitivity heterogeneities and singularities, in particular in the vicinity of the transmiter, that prevents from good convergence of the inversion process. In order to achieve reliable local inversion with such pathologic sensitivities, several crucial aspect of the algorithm must be considered. Firstly, Gauss-Newton inversion must be prefered to a gradient-based method, and a very robust solver such as LSQR is necessary. Then a good parameterization must be chosen. Preconditionning of the system with model reparameterization must also be used to compensate for the sensitivity loss with depth. But we show that even with a Gauss-Newton resolution and efficient preconditionning of both data and model parameter, the inversion fails to converge to an accurate solution with single source position in cases where MT data inversion or CSEM with many transmiters provides sucessful results. In order to mitigate the sensitivity singularities, we propose to reformulate the CSEM inverse problem by recasting the data under the form of a pseudo-MT tensor. The associated sensitivities are a linear combinations of the common CSEM sensitivies for electric and magnetic fields. We show on a synthetic model that this formulation highly reduce the footprint of the tranmitter in the sensitivities even in the near field. It also provides a natural balance between the data and between the whole frequency range. Finaly the approach allows the inversion to converge to a reliable result. This approach has been successefully applied on a 3D land survey in France with very complex geology, where the common CSEM inversion had provided biased results.

International audience; Within the frame of the French Research Project founded by the group of scientific interest GEODENERGIES for a greater understanding of geothermal reservoirs in fault zones (REFLET), our research efforts have been focusing on improving electromagnetic methods for resistivity mapping in environments with a high level of electromagnetic noise. Indeed, the Magneto-Telluric (MT) method has been traditionally used for exploration in geothermal areas. The challenge is that such a passive method is severely affected by electromagnetic noise induced by anthropogenic activities such as in cities and industries. In particular, MT exploration is almost not feasible in countries where trains are powered by DC-current (such as in many countries in Europe). As an alternative, we propose to deploy active source EM techniques (Controlled Source EM) using surface and borehole high power emissions to improve the EM signal to noise ratios and allow mapping resistivity variations at the depth of the geothermal targets.Over the last few years, we have performed several combined MT and CSEM surveys to demonstrate the benefits and limitations of such methods for geothermal exploration in different geological settings, e.g. volcanic areas in the Caribbean, in sedimentary basins in France, Czech Republic, Belgium, and Switzerland. When possible, we have also deployed Long Electrode Mise-à-la-Masse (LEMAM) setups utilizing the casings of the geothermal wells as source electrodes to improve even more the CSEM signal to noise ratio. To process and interpret the large diversity of data acquired (near-field, far-field, plane wave, borehole etc.), we have developed our own processing and 3D modelling and inversion routines. In this paper, we will share the results and learning from these different surveys and show the complementarity of the different techniques for retrieving realistic resistivity measurements at the depth of the geothermal reservoirs. We will also show that in very noisy environments, despite the use of remote references and robust processing, some frequency bands of MT data are still corrupted by permanent anthropic noise sources and CSEM data must be preferred. Finally, we will illustrate with the different examples the need and benefits of developing more advanced acquisition, processing, modelling and inversion techniques to ensure EM techniques are considered as exploration techniques of prime choice to reduce the cost and uncertainty associated with geothermal resource assessment.

National audience; Within the frame of geothermal exploration, a CSEM survey was performed at the Le Lamentin area (Martinique, French Indies) using 400m long energized metallic casings of two deep exploration boreholes as long electrodes for current injection (the so-called double Longue Electrode Mise-à-la-Masse setup, hereafter 2xLEMAM). Apparent resistivity maps were derived from the in-phase electric fields. Frequency dependent apparent resistivity maps and profiles reveal a very conductive area north of the Fort de France Bay connected to a known poly-phased geothermal system and shallow salt water intrusion. The most conductive body is proposed to be the geo-electrical signature of an active hydrothermal system, superimposed on the signature of a conductive fossilized geothermal system. It is spatially well correlated with high temperature borehole logs.

National audience

International audience; Resistivity imaging is a key input in many geothermal exploration programs, and particularly in volcanic environment. It is assumed that resistivity variations allow imaging the caprock, but also hydrothermal weathering and preferential geothermal fluid flow. The latter is generally a brine which drastically decreases the electrical resistivity and the temperature increase also lowers resistivity. Several authors have described resistivity signatures for some geothermal reservoirs context. Magnetotelluric (MT) is generally used to image resistivity variations down to a relevant investigation depth of several kilometers. Within the framework of a global French program oriented towards the development of geothermal energy, Martinique Island (Lesser Antilles, France) has been extensively investigated (from 2012 to 2013) through an integrated multi-method approach, with the aim to define precisely the potential geothermal structures previously highlighted. Amongst the common investigation methods deployed, we carried out three magnetotelluric (MT) surveys located above three out of the most promising geothermal areas of Martinique, namely the Anses d'Arlet (focus of this presentation), the Montagne Pelée and the Pitons du Carbet prospects. Based on natural source signal, the quality of broad band MT data depends on the source activity (lightning strikes and sun) during the survey but also on the local noise conditions. In many cases, because industrial activities have been developed close to the geothermal resource, the noise conditions make it difficult to obtain reliable MT tensors and therefore a good resistivity distribution. A common way to improve the data quality is to record longer time series and to apply robust processing with remote reference including noise filtering.

International audience; The use of CSEM onshore monitoring for CO2 plume migration characterization completes the information brought by seismic methods (localisation of saturation fronts): it is sensitive to the positive resistivity variations in the aquifer induced by supercritical C02 injection. Once a baseline 3D seismic survey is done and the structural seismic horizons are known, CSEM monitoring could compete with seismic monitoring because of its really low cost and its ease of implementation. CSEM monitoring also brings complementary information to the electrical methods of monitoring: the depth of investigation is significantly higher than electrical methods, and CSEM permit a to inject multi-frequency currents in the ground (which allows a multi-scale scanning of the subsurface). Electrical methods have a very narrow resolution in the superficial subsurface while CSEM map greater depths and larger volumes. Moreover, CSEM surveys are also easy repeatable due to developments in current injections devices and the use of differential GPS monitoring of surface electrodes positions. Accuracy of CSEM double LEMAM (“Longue-Electrode-Mise-à-la Masse”, using boreholes metallic casings as long electrodes) and double MAM method (“Mise-à la Masse”, using downhole electrodes) for CO2 injection monitoring has been tested several times. While the repeatability noise is of a 1% order (shown by Bourgeois et al., 2012a), the electromagnetic response of a CO2 plume in a 1-km deep saline aquifer can reach several percents respect to the original baseline fields (ie pre-injection measurements), depending on the source and reservoir setup. Both double LEMAM and surface-to-LEMAM (or single-LEMAM) setups are particularly suited for deep CO2 injection monitoring because they are based on current injection directly in the deep aquifer layers using the borehole metallic casings as electrodes. The BRGM knowledge in this domain has been constructed on several consecutive and coherent projects (EMSAPCO2, CO2ReMoVe, CO2CARE and EM HONTOMIN) integrating field operations, data processing, uncertainties analysis and modeling (Bourgeois et al., 2012b).Double MAM CSEM monitoring proved its ability to map the CO2 plume and to characterize its migration during the Ketzin experiment (Girard et al., 2011). CSEM fields (electric and magnetic fields) measurements showed variations above the repeatability noise and spatially associated with the CO2 plume migration at short/medium term migration process (the CSEM method was at least able to detect a 17200 t CO2 bubble propagating at 700 m depth). At Hontomín, a CSEM baseline dataset has been acquired in December 2013 using a new type of CSEM source, involving surface electrodes and one borehole casing, the so-called surface LEMAM setup, which is supposed to be more sensitive to a CO2 plume than the classical double LEMAM setup (results to be published). The Hontomín experiment is challenging for the CSEM approach because a lower volume of CO2 than Ketzin will be injected at a depth twice the injection depth at Ketzin. Our strategy to overcome these conditions was to use a dense network of CSEM receiver station, with multi frequency acquisition, two different geometries of source polarization and current injection directly in the reservoir by the surface-LEMAM method. First results will be presented.In addition, in order to assess the feasibility of CSEM imaging for reservoir characterization, we will also present one application in the frame of geothermal exploration: a CSEM survey was performed at the Le Lamentin area (Martinique, French Indies) using 400m long energized metallic casings of two deep exploration boreholes as long electrodes for current injection (the so-called double Longue Electrode Mise-à-la-Masse setup, hereafter 2xLEMAM). Imaging results (frequency dependent apparent resistivity maps and profiles) reveal a very conductive area north of the Fort de France Bay connected to a known poly-phased geothermal system and shallow salt water intrusion. The most conductive body is proposed to be the geo-electrical signature of an active hydrothermal system, superimposed on the signature of a conductive fossilized geothermal system. It is spatially well correlated with high temperature borehole logs. Meanwhile, low temperature areas associated with the fossil hydro-thermal system plugged by clay minerals display less conductive values.Bourgeois, B., Coppo, N., Girard, .J.F. (2012). Tests de reproductibilité de la méthode LEMAM sur un site pétrolier près de Chuelles (Loiret) - Influence du bruit ambiant et des variations d'humidité saisonnières. BRGM report. BRGM/RP-60982-FR Bourgeois B., Rohmer J. et Girard J.F., 2012, Étude par modélisation numérique des performances de l’électromagnétisme appliqué au monitoring des stockages de CO2 - Influence du contexte géologique et effet des variations extérieures au réservoir. Rapport BRGM/RP-60981-FR. Girard, J. F., Coppo, N., Rohmer, J., Bourgeois, B., Naudet, V., & Schmidt-Hattenberger, C. (2011). Time-lapse CSEM monitoring of the Ketzin (Germany) CO 2 injection using 2× MAM configuration. Energy Procedia, 4, 3322-3329.

Submitted version.; International audience; Full waveform inversion is a powerful tool for quantitative seismic imaging from wide‐azimuth seismic data. The method is based on the minimization of the misfit between observed and simulated data. This amounts to the solution of a large‐scale nonlinear minimization problem. The inverse Hessian operator plays a crucial role in this reconstruction process. Accounting accurately for the effect of this operator within the minimization scheme should correct for illumination deficits, restore the amplitude of the subsurface parameters, and help to remove artefacts generated by energetic multiple reflections. Conventional minimization methods (nonlinear conjugate gradient, quasi‐Newton methods) only roughly approximate the effect of this operator. In this study, we are interested in the truncated Newton minimization method. These methods are based on the computation of the model update through a matrix‐free conjugate gradient solution of the Newton linear system. We present a feasible implementation of this method for the full waveform inversion problem, based on a second‐order adjoint state formulation for the computation of Hessian‐vector products. We compare this method with conventional methods within the context of 2D acoustic frequency full waveform inversion for the reconstruction of P‐wave velocity models. Two test cases are investigated. The first is the synthetic BP 2004 model, representative of the Gulf of Mexico geology with high velocity contrasts associated with the presence of salt structures. The second is a 2D real data‐set from the Valhall oil field in North sea. Although, from a computational cost point of view, the truncated Newton method appears to be more expensive than conventional optimization algorithms, the results emphasize its increased robustness. A better reconstruction of the P‐wave velocity model is provided when energetic multiple reflections make it difficult to interpret the seismic data. A better trade‐off between regularization and resolution is obtained when noise contamination of the data requires one to regularize the solution of the inverse problem.

International audience

Submitted version.; Full Waveform Inversion (FWI) is a powerful tool for quantitative seismic imaging from wide-azimuth seismic data. The method is based on the minimization of the misfit between observed and simulated data. This amounts to the resolution of a large-scale nonlinear minimization problem. The inverse Hessian operator plays a crucial role in this reconstruction process. Accounting accurately for the effect of this operator within the minimization scheme should correct for illumination deficits, restore the amplitude of the subsurface parameters, and help to remove artifacts generated by energetic multiple reflections. Conventional preconditioned gradient-based minimization methods only roughly approximate the effect of this operator. We are interested in this study to another class of minimization methods, named as truncated Newton methods. These methods are based on the computation of the model update through a matrix-free conjugate gradient resolution of the Newton linear system. The aim of this study is to present a feasible implementation of this method for the FWI problem, based on a second-order adjoint state formulation for the computation of Hessian-vector products. We compare this method with the nonlinear conjugate gradient and the l-BFGS method within the context of 2D acoustic frequency FWI for the reconstruction of P-wave velocity models. Two test cases are investigated. The first is the synthetic BP 2004 model, representative of the Gulf Of Mexico geology with high velocity contrasts associated with the presence of salt structures. The second is a 2D real data-set from the Valhall oil field in North sea. These tests emphasize the interesting properties of the truncated Newton method regarding conventional optimization methods within the context of FWI.

Imaging tectonic faults with small vertical offsets in argilittes (clay-rock) using geophysical methods is challenging. In the context of deep radioactive waste disposals, the presence of such faults has to be assessed since they can modify the rock confining properties. In the Tournemire Experimental Platform (TEP, France), fault zones with small vertical offsets and complex shape have been identified from underground works. However, 3D high-resolution surface seismic methods have shown limitations in this context that led us to consider the detection and characterization of the faults directly from underground works. We investigate here the potential of seismic full waveform inversion (FWI) applied in a transmission configuration to image the clay-rock medium in a horizontal plane between galleries, and compare it with first-arrival traveltime tomography (FATT). Our objective is to characterize seismic velocities of a block of argilittes crossed by a subvertical fault zone with a small vertical offset. The specific measurement configuration allows us to neglect the influence of the galleries on the wave propagation and to simplify the problem by considering a 2D isotropic horizontal imaging domain. Our FWI scheme relies on a robust adaptation of early-arrival waveform tomography. The results obtained with FATT and FWI are in accordance and both correlate with the geological observations from the gallery walls and boreholes. We show that even though various simplifications are done in the inversion scheme and only a part of the data is used, FWI allows to get higher resolution images than FATT, and is especially less sensitive to the incomplete illumination as it uses also diffracted energy. The results provided in this study highlight the complexity of the fault zone, showing a complex interaction of the main fault system with a secondary system composed of decimetric fractures associated with the presence of water.

International audience

Standard seismic methods are generally not well adapted to provide sharp quantitative images of the first few metres of underground. A two-dimensional full-waveform inversion of land seismic data, based on frequency-domain viscoelastic modelling, offers a promising approach to take advantage of the full complexity of seismograms and to simultaneously build 2D images of P and VS parameters. In order to understand the behaviour of this method in a near-surface context and anticipate the corresponding field applications, we perform this investigation by applying waveform inversion on a simple layered medium. We first use synthetic data obtained from numerical modelling and then we employ laboratory data obtained by small-scale physical modelling. We demonstrate that such a near-surface 2D model can be quantitatively determined even in a realistic situation where the data are dominated by high-amplitude surface waves. A comparison of results derived for the same medium from ideal synthetic data and noisy experimental data allows detecting anomalies in the reconstruction of velocity models due to the experimental nature of the data used.

International audience; In near-surface quantitative seismic imaging, the mechanical properties of an heterogeneous medium are usually inferred from the measure of the normal velocity component at different locations. In this study, it is proposed to investigate the benefits of measuring also the tangential velocity component. For that purpose, a realistic synthetic model is defined and the benefits of each component are analyzed in the framework of seismic imaging by Full Waveform inversion. The model is a two-layer medium including lateral variations close to the surface and the synthetic data are generated using a visco-elastic finite elements code. An analysis of the information contained for the different components is carried out and the behavior of the inversion algorithm is studied for each component. To conclude this part, inversion results from inversion strategies integrating the two components are presented. The last part concerns the experimental modeling facility developed in order to experimentally validate the imaging methods. This measurement bench reproduces seismic measurement configurations at a reduced scale using an ultrasonic source and a laser interferometer. This facility has already been validated for the case of the measurement of the vertical component, and first experimental results of the horizontal component are presented.

Recent advances in seismic inversion methods allow us to get high-resolution quantitative representations of mechanical parameters of the Earth. One of the most promising inversion method is the Full-waveform inversion (FWI) method which basically consists of an iterative data fitting process based on a local optimisation (Tarantola, 1984) and unlike conventional methods (e.g. first arrival times tomography) all the information available in the data can be handled. This method has been first applied to depth exploration (Virieux and Operto, 2009), but some authors already applied it to quantitatively image the near-surface. For example, G'elis (2005) succeeded to image cavities using synthetic data including surface waves arrivals. However one of the most important remaining challenges is the validation of the method with real field data that presents many difficulties: field data are often measured on not well-known media composed of high heterogeneity degree, important intrinsic attenuation and with prominent three dimensional effects. Considering all these difficulties, the necessity to have an experimental intermediate validation step becomes obvious. The aim is to reproduce the field experiments in a well-controlled environment at a smaller scale using a well-defined spacio-temporal scalling. The physical reduced scale modeling has several advantages compared to field-measurements: first the model physical parameters are well controlled, thus numerically repeatable; secondly, the great flexibility of the setup enables to experiment easily novel measurement configurations. This validation approach has already been first applied for geophysical imaging purposes by French (1974). Recently Bretaudeau (2010) designed an ultrasonic measurement bench called MUSC with several features that enable an excellent reproducibility of field measurements from the surface. For instance, a special attention has been paid to select a source that can be assimilated to a point source and the sensors displacements are controlled with an accuracy of 10 m. Using this experimental setup, Bretaudeau (2010) showed promising results on the inversion using the vertical component. Due to recent technological improvements (Blum et al., 2010), it is now possible to record the vertical and the horizontal components simultaneously using a new laser interferometer device which has been implemented in the MUSC measurement bench. In this context, the aim of the study presented here is to investigate numerically the benefits of the each component in the frame of the FWI method and to present the recent advances on measurements of both components on reduced physical scale models. First, the physical reduced scale model and its numerical representation will be described, then a numerical study will present the benefits provided by each component for velocities reconstructions and finally measurements of the vertical and horizontal components on the reduced physical scale model will be presented and compared to numerical simulations.

Full Wave Inversion (FWI) is a multiparameter quantitative method of seismic imaging that is currently booming. Taking into account all propagation phenomena recorded, especially the surface waves, makes it a promising technique for near surface applications on geotechnical and environmental issues. However, adapting a seismic method validated on synthetic data to the context of near surface or on deep imaging applications remains a difficult task. What we are proposing here is a reduced scale physical modelling approach using a laboratory of non contact laser-ultrasonic measurement, which will have been set up as part of this thesis. The various elements of the laboratory were validated for reduced scale seismic modelling, and particular attention has been paid to the simulation of the source. The acquired data has been validated by comparison with data obtained by numerical modelling using algorithmes provided by Géoazur within the context of the ANR SEISCOPE project. In a second step, we evaluated the potential of the use of FWI for near surface applications by using a joint approach of numerical modelling and experimental modelling. By this dual approach, we have thus shown that FWI allows to obtain information on the morphology of underground cavities by taking into account surface waves. Furthermore, on the various multilayer models we studied, the FWI was able to exploit the information on the most superficial areas supplied by dispersive surface waves.; L'inversion des formes d'ondes (FWI) est une méthode d'imagerie sismique quantitative multiparamètre actuellement en plein essor. La prise en compte de l'ensemble des phénomènes de propagation enregistrés, en particulier des ondes de surface, en font une technique prometteuse pour des applications de la subsurface sur des problématiques géotechniques ou environnementales. Cependant, l'adaptation au contexte de la subsurface d'une méthode sismique validée numériquement ou sur des applications d'imagerie profonde est une tâche difficile. Nous proposons ici une approche par modélisation physique à échelle réduite à l'aide d'un laboratoire de Mesures laser-Ultrasonore Sans Contact (MUSC) mis en place dans le cadre de cette thèse. Les différents éléments du laboratoire ont été caractérisés pour la modélisation sismique à échelle réduite et une attention particulière a été portée à la simulation de la source. Les données acquises ont été validées par confrontation avec des données obtenues par modélisation numérique à l'aide des codes fournis par Géoazur dans le cadre du projet ANR SEISCOPE. Dans un deuxième temps, le potentiel de la FWI pour des applications de la subsurface a été évalué par l'approche conjointe de la modélisation numérique et de la modélisation expérimentale. Nous avons ainsi pu montrer par cette double approche que la FWI permet d'obtenir une information sur la morphologie de cavités souterraines grâce à la prise en compte des ondes de surface. Par ailleurs, sur les différents modèles multicouches étudiés, l'information sur les zones les plus superficielles fournie par les ondes de surface dispersives a pu être exploitée par la FWI.

Interpretation, 4 (4), ISSN:2324-8858