Sparse Bayesian linearized amplitude-versus-angle inversion

The technique of seismic amplitude-versus-angle inversion has been widely used to estimate lithology and fluid properties in seismic exploration. The amplitude-versus-angle inversion problem is intrinsically ill-posed and generally stabilized by the use of L2-norm regularization methods but with drawback of smoothing important boundaries between adjacent layers. In this study, we propose a sparse Bayesian linearized solution for amplitude-versus-angle inversion problem to preserve the sharp geological interfaces. In this regard, a priori constraint term with two regularization functions is presented: the sparse constraint regularization and the low-frequency model information. In addition, to obtain high-resolution reflectivity estimation, the model parameters decorrelation technique combined with dipole decomposition method is employed. We validate the applicability of the presented method by both synthetic and real seismic data from the Gulf of Mexico. The accuracy improvement of the presented method is also confirmed by comparing the results with the commonly used Bayesian linearized amplitude-versus-angle inversion.     

实数编码遗传算法在大地电磁测深二维反演中的应用

作为全局非线性优化的新方法之一的遗传算法,近年来已从生物工程流行到大地电磁测深资料解释中.然而,大地电磁反演问题具有不适定性,解的非唯一性.通过结合求解不适定问题的Tikhonov正则化方法,本文采用实数编码遗传算法求解大地电磁二维反演问题.此算法在构建目标函数时引入正则化的思想,利用遗传算法求解最优化问题.常规的基于局部线性化的最优化反演方法易使解陷入局部极小值,而且严重的依赖初始模型的选择.与传统线性化的迭代反演方法相比,实数编码遗传算法能够克服传统方法的不足且能获得更好的反演结果.通过对大地电磁测深理论模型进行计算,结果表明:该算法具有收敛速度快、解的精度高和避免出现早熟等优点,可用于大地电磁资料解释.     

Inversion of DC resistivity data using neural networks

The inversion of geoelectrical resistivity data is a difficult task due to its non-linear nature. In this work, the neural network (NN) approach is studied to solve both 1D and 2D resistivity inverse problems. The efficiency of a widespread, supervised training network, the back-propagation technique and its applicability to the resistivity problem, is investigated. Several NN paradigms have been tried on a basis of trial-and-error for two types of data set. In the 1D problem, the batch back-propagation paradigm was efficient while another paradigm, called resilient propagation, was used in the 2D problem. The network was trained with synthetic examples and tested on another set of synthetic data as well as on the field data. The neural network gave a result highly correlated with that of conventional serial algorithms. It proved to be a fast, accurate and objective method for depth and resistivity estimation of both 1D and 2D DC resistivity data. The main advantage of using NN for resistivity inversion is that once the network has been trained it can perform the inversion of any vertical electrical sounding data set very rapidly.     

2-D crustal Poisson’s ratio from seismic travel time inversion in Changbaishan Tianchi volcanic region

Based on the inversion method of 2D velocity structure and interface, the crustal velocity structures of P-wave and S-wave along the profile L ₁ are determined simultaneously with deep seismic sounding data in Changbaishan Tianchi volcanic region, and then its Poisson’s ratio is obtained. Calculated results show that this technique overcomes some defects of traditional forward calculation method, and it is also very effective to determine Poisson’s ratio distribution of deep seismic sounding profile, especially useful for study on volcanic magma and crustal fault zone. Study result indicates that there is an abnormally high Poisson’s ratio body that is about 30 km wide and 12 km high in the low velocity region under Tianchi crater. Its value of Poisson’s ratio is 8% higher than that of surrounding medium and it should be the magma chamber formed from melted rock with high temperature. There is a high Poisson’s ratio zone ranging from magma chamber to the top of crust, which may be the uprise passage of hot substance. The lower part with high Poisson’s ratio, which stretches downward to Moho, is possibly the extrusion way of hot substance from the uppermost mantle. The conclusions above are consistent with the study results of both tomographic determination of 3D crustal structure and magnetotelluric survey in this region. Foundation item: Key Project from China Earthquake Administration and the Project (95-11-02-01) from Ministry of Science and Technology (2001DIA10003). Contribution No. RCEG200401, Geophysical Exploration Center, China Earthquake Administration.     

A numerical inversion scheme for geoelectrical soundings

A simple numerical inversion scheme for estimatingn-layer model parameters from observed geoelectrical resistivity data can be used in either the space or wavenumber domain. The technique utilizes Madden's Transmission Line Analogy to compute the resistivity transforms and linear filter theory to accomplish the excursions between the space and wavenumber domains. The inversion is effected by an iterative refinement scheme employing the stochastic inverse which is approximate to the generalized inverse. No singular decomposition analysis is required and the scheme is stable under ill conditions. The inversion scheme not only gives the desired estimates; it exposes redundant parameters and irrelevant data and is easily programmed on a desk-top mini computer. Examples of inverse modeling with hypothetical and field data are discussed.     

LOW-PASS FILTERING OF NOISY FIELD SCHLUMBERGER SOUNDING CURVES PART II: APPLICATION*

The basic principles of the application of the linear system theory for smoothing noise-degraded d.c. geoelectrical sounding curves were recently established by Patella. A field Schlumberger sounding is presented to demonstrate first their application and validity. To achieve this purpose, firstly it is pointed out that the required smoothing or low-pass filtering can be considered as an intrinsic property of the transformation of original Schlumberger sounding curves into pole-pole (two-electrode) curves. Then we sketch a numerical algorithm to perform the transformation, opportunely modified from a known procedure for transforming dipole diagrams into Schlumberger ones. Finally we show a field example with the double aim of demonstrating (i) the high quality of the low-pass filtering, and (ii) the reliability of the transformed pole-pole curve as far as quantitative interpretation is concerned.     

The neural network approximation method for solving multidimensional nonlinear inverse problems of geophysics

The iterative approximation neural network method for solving conditionally well-posed nonlinear inverse problems of geophysics is presented. The method is based on the neural network approximation of the inverse operator. The inverse problem is solved in the class of grid (block) models of the medium on a regularized parameterization grid. The construction principle of this grid relies on using the calculated values of the continuity modulus of the inverse operator and its modifications determining the degree of ambiguity of the solutions. The method provides approximate solutions of inverse problems with the maximal degree of detail given the specified degree of ambiguity with the total number of the sought parameters ~n × 10³ of the medium. The a priori and a posteriori estimates of the degree of ambiguity of the approximated solutions are calculated. The work of the method is illustrated by the example of the three-dimensional (3D) inversion of the synthesized 2D areal geoelectrical (audio magnetotelluric sounding, AMTS) data corresponding to the schematic model of a kimberlite pipe.     

Far-zone contributions to the gravity field quantities by means of Molodensky’s truncation coefficients

To reduce the numerical complexity of inverse solutions to large systems of discretised integral equations in gravimetric geoid/quasigeoid modelling, the surface domain of Green’s integrals is subdivided into the near-zone and far-zone integration sub-domains. The inversion is performed for the near zone using regional detailed gravity data. The farzone contributions to the gravity field quantities are estimated from an available global geopotential model using techniques for a spherical harmonic analysis of the gravity field. For computing the far-zone contributions by means of Green’s integrals, truncation coefficients are applied. Different forms of truncation coefficients have been derived depending on a type of integrals in solving various geodetic boundary-value problems. In this study, we utilise Molodensky’s truncation coefficients to Green’s integrals for computing the far-zone contributions to the disturbing potential, the gravity disturbance, and the gravity anomaly. We also demonstrate that Molodensky’s truncation coefficients can be uniformly applied to all types of Green’s integrals used in solving the boundaryvalue problems. The numerical example of the far-zone contributions to the gravity field quantities is given over the area of study which comprises the Canadian Rocky Mountains. The coefficients of a global geopotential model and a detailed digital terrain model are used as input data.     

Conditionality of inverse solutions to discretised integral equations in geoid modelling from local gravity data

The eigenvalue decomposition technique is used for analysis of conditionality of two alternative solutions for a determination of the geoid from local gravity data. The first solution is based on the standard two-step approach utilising the inverse of the Abel-Poisson integral equation (downward continuation) and consequently the Stokes/Hotine integration (gravity inversion). The second solution is based on a single integral that combines the downward continuation and the gravity inversion in one integral equation. Extreme eigenvalues and corresponding condition numbers of matrix operators are investigated to compare the stability of inverse problems of the above-mentioned computational models. To preserve a dominantly diagonal structure of the matrices for inverse solutions, the horizontal positions of the parameterised solution on the geoid and of data points are identical. The numerical experiments using real data reveal that the direct gravity inversion is numerically more stable than the downward continuation procedure in the two-step approach.     

Relative source time functions of seismic events at the Rudna copper mine, Poland: estimation of inversion uncertainties

The relative source time function (RSTF) inversion uncertainty assessment was performed for two small, mining-induced seismic events (M _W =2.9 and 3.0) that occurred at Rudna copper mine in Poland. The seismograms of selected events were recorded by the seismic net work composed of over 60, short-period, vertical seismometers, recording ground velocity, located in the distance ranging from 400 m up to 8 km from their hypocenters. The RSTFs were calculated for each seismic station independently, using the empirical Green’s function technique. The pseudospectral approximation of the sought RSTF by a finite sum of Gaussian kernel functions was used and the inverse problem was solved with the adaptive simulated annealing algorithm. Both methods improved the stability of the deconvolution procedure and physical correctness of the final solution in comparison to the classical deconvolution methods. To estimate the inversion uncertainties, classical Markov-chain Monte-Carlo techniques were used. The uncertainty analysis allows for improved selection of a priori data to the following inversion for kinematic rupture process.     

Inversions for MT data in 2D symmetrical anisotropic media

Introduction Ready and Renkin (1971) were the first to make the research on anisotropy problems in magnetotellurics (MT). The progress in the research is not evident because it is more complex and difficult than isotropic problems. Now, the one-dimensional (1D) anisotropy problems in MT have been well solved, while for the two-dimensional (2D) cases, the numerical solutions have only been obtained for some particular conditions (Ready and Renkin, 1975). As to the three-dimensional (3D) ani…     

GEOELECTRICAL MODELS INVOLVING LAYERS WITH A LINEAR CHANGE IN RESISTIVITY AND THEIR USE IN THE INVESTIGATION OF CLAY DEPOSITS*

A model has been set up for the interpretation of geoelectrical sounding data for certain kinds of clay deposits containing gradually varying amounts of sand. The model assumes that in the so-called gradient layer, resistivity varies linearly with depth. Model calculations show how such a layer can be replaced by two homogeneous layers. An inversion procedure using the Marquardt algorithm has been developed for the interpretation of sounding data obtained with the Schlumberger array; it assumes a gradient layer beneath several covering layers. The procedure is demonstrated on two clay deposits. A comparison is made between the newly developed interpretation, the traditional approach using model curves, and computerized inversion for a homogeneous layer model.     

Inversion of electromagnetic data: An overview of new techniques

This paper explores some of the newer techniques for acquiring and inverting electromagnetic data. Attention is confined primarily to the 2d magnetotelluric (MT) problem but the inverse methods are applicable to all areas of EM induction. The basis of the EMAP technique of Bostick is presented along with examples to illustrate the efficacy of that method in structural imaging and in overcoming the deleterious effects of near-surface distortions of the electric field. Reflectivity imaging methods and the application of seismic migration techniques to EM problems are also explored as imaging tools. Two new approaches to the solution of the inverse problem are presented. The AIM (Approximate Inverse Mapping) inversion of Oldenburg and Ellis uses a new way to estimate a perturbation in an iterative solution which does not involve linearization of the equations. The RRI (Rapid Relaxation Inverse) of Smith and Booker shows how approximate Fréchet derivatives and sequences of 1d inversions can be used to develop a practical inversion algorithm. The overview is structured to provide insight about the latest inversion techniques and also to touch upon most areas of the inverse problem that must be considered to carry out a practical inversion. These include model parameterization, methods of calculating first order sensitivities, and methods for setting up a linearized inversion.     

Two dimensional combined inversion of short- and long-normal dc resistivity well log data

A method is presented for the two-dimensional combined inversion of short- and long-normal tool direct current resistivity data with symmetry. The forward problem is solved using the finite element method in the cylindrical coordinates system. The inverse problem is solved using a conjugate gradient technique with the partial derivatives obtained using reciprocity. The parameters were obtained by means of both conjugate gradient relaxation and conventional conjugate gradient method. The solution of this highly underdetermined inverse problem is stabilized using Tikhonov regularization and the scheme yields a blurred image of the subsurface. The scheme is tested using synthetic data and field data. Tests using synthetic data suggest that traces of the horizontal boundaries are delineated in the range of the exploration distance while the resolution of vertical boundaries depends upon the solution regularization. Application to field data shows that additional information is necessary for resolving the resistivity structure when there are low resistivity contrasts between formation units.     

New advances in regularized inversion of gravity and electromagnetic data

The interpretation of potential and electromagnetic fields observed over 3D geological structures remains one of the most challenging problems of exploration geophysics. In this paper I present an overview of novel methods of inversion and imaging of gravity and electromagnetic data, which are based on new advances in the regularization theory related to the application of special stabilizing functionals, which allow the reconstruction of both smooth images of the underground geological structures and models with sharp geological boundaries. I demonstrate that sharp-boundary geophysical inversion can improve the efficiency and resolution of the inverse problem solution. The methods are illustrated with synthetic and practical examples of the 3D inversion of potential and electromagnetic field data.     

Recent developments in the direct-current geoelectrical imaging method

There have been major improvements in instrumentation, field survey design and data inversion techniques for the geoelectrical method over the past 25 years. Multi-electrode and multi-channel systems have made it possible to conduct large 2-D, 3-D and even 4-D surveys efficiently to resolve complex geological structures that were not possible with traditional 1-D surveys. Continued developments in computer technology, as well as fast data inversion techniques and software, have made it possible to carry out the interpretation on commonly available microcomputers. Multi-dimensional geoelectrical surveys are now widely used in environmental, engineering, hydrological and mining applications. 3-D surveys play an increasingly important role in very complex areas where 2-D models suffer from artifacts due to off-line structures. Large areas on land and water can be surveyed rapidly with computerized dynamic towed resistivity acquisition systems. The use of existing metallic wells as long electrodes has improved the detection of targets in areas where they are masked by subsurface infrastructure. A number of PC controlled monitoring systems are also available to measure and detect temporal changes in the subsurface. There have been significant advancements in techniques to automatically generate optimized electrodes array configurations that have better resolution and depth of investigation than traditional arrays. Other areas of active development include the translation of electrical values into geological parameters such as clay and moisture content, new types of sensors, estimation of fluid or ground movement from time-lapse images and joint inversion techniques. In this paper, we investigate the recent developments in geoelectrical imaging and provide a brief look into the future of where the science may be heading.     

Image-guided inversion in steady-state hydraulic tomography

In steady-state hydraulic tomography, the head data recorded during a series of pumping or/and injection tests can be inverted to determine the transmissivity distributions of an aquifer. This inverse problem is usually under-determined and ill-posed. We propose to use structural information inferred from a guiding image to constrain the inversion process. The guiding image can be drawn from soft data sets such as seismic and ground penetrating radar sections or from geological cross-sections inferred from the wells and some geological expertise. The structural information is extracted from the guiding image through some digital image analysis techniques. Then, it is introduced into the inversion process of the head data as a weighted four direction smoothing matrix used in the regularizer. Such smoothing matrix allows applying the smoothing along the structural features. This helps preserving eventual drops in the hydraulic properties. In addition, we apply a procedure called image-guided interpolation. This technique starts with the tomogram obtained from the image-guided inversion and focus this tomogram. These new approaches are applied on four synthetic toy problems. The hydraulic distributions estimated from the image-guided inversion are closer to the true transmissivity model and have higher resolution than those computed from a classical Gauss–Newton method with uniform isotropic smoothing.     

Estimation of Elastic Moduli in a Compressible Gibson Half-space by Inverting Rayleigh-wave Phase Velocity

A Gibson half-space model (a non-layered Earth model) has the shear modulus varying linearly with depth in an inhomogeneous elastic half-space. In a half-space of sedimentary granular soil under a geostatic state of initial stress, the density and the Poisson’s ratio do not vary considerably with depth. In such an Earth body, the dynamic shear modulus is the parameter that mainly affects the dispersion of propagating waves. We have estimated shear-wave velocities in the compressible Gibson half-space by inverting Rayleigh-wave phase velocities. An analytical dispersion law of Rayleigh-type waves in a compressible Gibson half-space is given in an algebraic form, which makes our inversion process extremely simple and fast. The convergence of the weighted damping solution is guaranteed through selection of the damping factor using the Levenberg-Marquardt method. Calculation efficiency is achieved by reconstructing a weighted damping solution using singular value decomposition techniques. The main advantage of this algorithm is that only three parameters define the compressible Gibson half-space model. Theoretically, to determine the model by the inversion, only three Rayleigh-wave phase velocities at different frequencies are required. This is useful in practice where Rayleigh-wave energy is only developed in a limited frequency range or at certain frequencies as data acquired at manmade structures such as dams and levees. Two real examples are presented and verified by borehole S-wave velocity measurements. The results of these real examples are also compared with the results of the layered-Earth model.     

2D data modelling by electrical resistivity tomography for complex subsurface geology

A new tool for two‐dimensional apparent‐resistivity data modelling and inversion is presented. The study is developed according to the idea that the best way to deal with ill‐posedness of geoelectrical inverse problems lies in constructing algorithms which allow a flexible control of the physical and mathematical elements involved in the resolution. The forward problem is solved through a finite‐difference algorithm, whose main features are a versatile user‐defined discretization of the domain and a new approach to the solution of the inverse Fourier transform. The inversion procedure is based on an iterative smoothness‐constrained least‐squares algorithm. As mentioned, the code is constructed to ensure flexibility in resolution. This is first achieved by starting the inversion from an arbitrarily defined model. In our approach, a Jacobian matrix is calculated at each iteration, using a generalization of Cohn's network sensitivity theorem. Another versatile feature is the issue of introducing a priori information about the solution. Regions of the domain can be constrained to vary between two limits (the lower and upper bounds) by using inequality constraints. A second possibility is to include the starting model in the objective function used to determine an improved estimate of the unknown parameters and to constrain the solution to the above model. Furthermore, the possibility either of defining a discretization of the domain that exactly fits the underground structures or of refining the mesh of the grid certainly leads to more accurate solutions. Control on the mathematical elements in the inversion algorithm is also allowed. The smoothness matrix can be modified in order to penalize roughness in any one direction. An empirical way of assigning the regularization parameter (damping) is defined, but the user can also decide to assign it manually at each iteration. An appropriate tool was constructed with the purpose of handling the inversion results, for example to correct reconstructed models and to check the effects of such changes on the calculated apparent resistivity. Tests on synthetic and real data, in particular in handling indeterminate cases, show that the flexible approach is a good way to build a detailed picture of the prospected area.     

An efficient non-linear least-squares 1D inversion scheme for resistivity and IP sounding data

Non-linear least-squares inversion operates iteratively by updating the model parameters in each step by a correction vector which is the solution of a set of normal equations. Inversion of geoelectrical data is an ill-posed problem. This and the ensuing suboptimality restrict the initial model to being in the near vicinity of the true model. The problem may be reduced by introducing damping into the system of equations. It is shown that an appropriate choice of the damping parameter obtained adaptively and the use of a conjugate-gradient algorithm to solve the normal equations make the 1D inversion scheme efficient and robust. The scheme uses an optimal damping parameter that is dependent on the noise in the data, in each iterative step. The changes in the damping and relative residual error with iteration number are illustrated. A comparison of its efficacy over the conventional Marquardt and simulated annealing methods, tested on Inman's model, is made. Inversion of induced polarization (IP) sounding is obtained by inverting twice (true and modified) DC apparent resistivity data. The inversion of IP data presented here is generic and can be applied to any of the IP observables, such as chargeability, frequency effect, phase, etc., as long as these observables are explicitly related to the DC apparent resistivity. The scheme is used successfully in inverting noise-free and noisy synthetic data and field data taken from the published literature.