EM Imaging: Inversion and Interpretation Approaches Chairpersons: P.P. Lugao and P.Weidelt 5.1 USE OF MPI AND GALERKIN PROJECTION METHODS IN THREE- DIMENSIONAL MODELING AND INVERSION OF ELECTROMAGNETIC DATA Patricia Pastana de Lugao Exploration and Reservoir Services Baker Hughes, 11111 Katy Freeway, Suite 200, Houston, TX 77079, USA lugao@geosignal.com This work presents an algorithm for three-dimensional modeling and inversion of electromagnetic field data that incorporates a conjugate gradient solution to solve multiple linear systems and is parallelized using MPI, Message Passing Interface, libraries to run in a cluster of Linux boxes. The forward code is based on a finite difference approximation solving for the electric field components. The symmetry of the matrix allows for forward calculation of multiple frequencies and calculation of sensitivities using Galerkin projection methods. Galerkin projection methods have been applied for calculation of sensitivities when the coefficient matrix is the same for different right-hand-sides. When the matrices of coefficients differ only by a parameterized identity matrix or a low rank matrix, Galerkin projection methods can be used to solve multiple systems with different coefficient matrices and right-hand- sides. This is the case for forward modeling of multiple frequencies. In Galerkin projection methods a seed system is solved and the solution is projected to solve other systems. Further speed can be obtained if the next systems are solved in parallel by a cluster of PC's with Linux environments. MPI libraries are used to pass information from a master system onto slave systems. This type of environment is proven to be cost and time efficient, making inversion of large data sets affordable. 5.2 APPROXIMATE EQUALITY CONSTRAINTS IN THE INVERSION OF ANISOTROPIC MT DATA Cicero R.T. Regis and Luiz Rijo Universidade Federal do Para, Belem, Brazil cicero@ufpa.br We have studied the inversion of synthetic anisotropic magnetotelluric data in models composed of horizontal layers. In each layer we have three values for the resistivities, corresponding to the main directions of the resistivity tensor, and two angles that define those directions. There is a high degree of ambiguity between the resistivities and the angles, this contributes to make the solution of this problem more unstable than that of the isotropic case. We have found that, for anisotropic data, it is better to use the components of the complex MT impedance tensor as observations in the inversion, rather than their representation as apparent resistivities and phases, because this transformation increases the level of ambiguity. To achieve stable solutions we have to add a priori information to the inversion. A priori information can be acquired from well logging or from other kinds of geophysical data or geological knowledge. If we have this information and if it is really representative, we use it to constrain the inversion parameters in a least squares sense. This results in solutions that are not only stable, but also geologically meaningful. We determine which parameters should be constrained by studying the singular value decomposition of the sensitivity matrix obtained in the last iteration of the unconstrained inversion, which yields unstable solutions. The parameters or combinations of parameters that generate the smallest singular values are the ones that need to be constrained. 5.3 STUDY OF THE DIURNAL VARIATION AT A GLOBAL SCALE Naphsica Grammatica(1), Michel Menvielle(2), and Pascal Tarits(1) (1) IUEM - UBO, UMR6538 "Domaines Oceaniques", Place Nicolas Copernic, F29280, Plouzane , France (2) CETP, 4 Avenue de Neptune, F-94107 Saint Maur des Fosses, France naphsica@sdt.univ-brest.fr The diurnal variation recorded at satellite altitude is the sum of a primary field, of ionospheric origin, and a secondary field induced in the conducting Earth. It therefore carries information about the conductivity structure of the Earth. However, during daytime, the diurnal variation also acts as a contaminating factor of the magnetic data for main field and crustal anomaly field calculations. We therefore address the question of the description of the diurnal variation. We use 3-component recordings from magnetic oservatories. We conduct a spherical harmonic analysis, up to degree and order 4. We develop two different methods for the determination of the spherical harmonic coefficients. The first one is based upon a singular value decomposition approach. The second one is a MCMC or Monte Carlo Markov Chain method, which provides a posteriori probability density functions of the harmonic coefficients. Synthetic data corresponding to realistic situations have been analysed using the two methods. The results are presented and discussed in terms of a possibility of a routine modelling of the diurnal variation at the global scale using geomagnetic observatory data. 5.4 QUASI-ANALYTICAL APPROXIMATION IN 3-D ELECTROMAGNETIC MODELING AND INVERSION Michael Zhdanov and Gabor Hursan University of Utah, USA mzhdanov@mines.utah.edu In this paper we address one of the most challenging problems of electromagnetic geophysical methods - 3-D inversion of EM data over inhomogeneous geological formations. The difficulties in the solution of this problem are two-fold. On the one hand, 3-D EM forward modeling is an extremely complicated and time consuming mathematical problem itself. On the other hand, the inversion is an unstable and ambiguous problem. To overcome these difficulties we suggest using for forward modeling the new quasi-analytical approximation developed recently by Zhdanov et. al., 1999. It is based on ideas similar to those developed by Habashy et al., 1993, for localized nonlinear approximation, and by Zhdanov and Fang, 1996, for quasi-linear approximation. We assume that the anomalous electrical field within an inhomogeneous domain is linearly proportional to the background field through an electrical reflectivity coefficient, which is a function of the background geoelectrical cross-section and the background EM field only. This approach leads to a construction of the quasi-analytical expressions for an anomalous EM field and for the Frechet derivative operator of a forward problem, which simplifies dramatically the forward modeling and inversion. To obtain a stable solution of a 3-D inverse problem we apply the regularization method based on using a focusing stabilizing functional introduced by Portniaguine and Zhdanov, 1999. This stabilizer helps generate a sharp and focused image of anomalous conductivity distribution. The inversion is based on the re-weighted regularized conjugate gradient method. 5.5 MATHEMATICAL MODELING OF THE CHARACTERISTIC SPACE-TIME DISTRIBUTION IN TRANSIENT PROCESSES OF SURFACE AND BOREHOLE SOUNDING V. Hallbauer-Zadorozhnaya(1) and W.S. Mogilatov(2) (1) Terra Sounding and Analytical Ltd., P.O. Box 2142, Somerset West, South Africa (2) Institute of Geophysics, Russian Academy of Science, Novosibirsk, Russia dkh@mweb.co.za When working with non-coaxial loops, particularly with a fixed- current source, difficulties arise in the interpretation of field data with wrong interpretations caused by anomalous signals. However, such anomalies could be caused by a difficult distribution of the e.m. field, even in a D1 medium. Programs, which we created, allow to calculate the intensity of an electrical field, E, and the current density, J, for each point of a real polarized medium for any moment of time. These programs also allowed to visualize the results as a three-dimensional realization, i.e. depth, distance and transient time, and present that realization in a time sequence. Compared to the concept of displaced smoke rings, our model shows that the distribution of eddy currents in each conductive layer consists of few ring currents separated by an intermediate zone. Their positions change with time. Our modeling further demonstrates that small IP effects can only be detected in conditions of loop separation. In such a case new ring currents form within the layer, which have opposite sign. The program compiled allowed an interpretation of surface and bore hole TDEM data. It has also been tested using actual borehole data and, it produced qualitative results for the lithological cross section. 5.6 3D MT INVERSION USING NONLINEAR CONJUGATE GRADIENTS Randall Mackie(1) and William Rodi(2) (1) GSY-USA, Inc., PMB #643, 2261 Market St., San Francisco, CA 94114-1600, USA (2) Earth Resources Laboratory, Massachusetts Institute of Technology, Cambridge, MA 02139, USA randy@gsy-usa.com We investigate an algorithm for computing smooth solutions to the 3-D magnetotelluric (MT) inverse problem. The algorithm employs the method of nonlinear conjugate gradients (NLCG) to minimize an objective function that sums the squared norm of the data residuals and the squared norm of spatial derivatives of the model function (log resistivity vs. position). The primary difference between NLCG and traditional Gauss-Newton methods is that each step of the iterative algorithm entails solving a one-dimensional nonlinear inverse problem (i.e., line minimization) rather than a multi-dimensional linearized problem. As a result, the NLCG approach does not require explicit computation of the sensitivity matrix (which entails solving a reciprocal forward problem for each frequency and station site) nor the solution of a large linear system at each iteration step. We have implemented the NLCG approach in 2-D MT inversion and demonstrated its advantages in problems involving real and synthetic data. Our efforts in implementating NLCG in the 3-D problem have focussed mainly on improving the line minimization and preconditioning algorithms we developed for 2-D, and on optimizing the forward modeling computations for 3-D. 5.7p JOINT INVERSION USING GLOBAL OPTIMIZATION OF DC RESISTIVITY AND MT DATA FOR RESOLVING THIN RESISTIVE AND CONDUCTING LAYERS S.P. Sharma Dept. of Geology and Geophysics, Indian Institute of Technology, Kharagpur 721302, India spsharma@gg.iitkgp.ernet.in Joint inversion is an effective approach for enhancing the resolution of a thin resistive and conductive layer. However, a solution obtained by linearized inversion can not be reliable even after joint inversion. A number of solutions can be derived using newly emerged global inversion methods and they are widely used now a days to interpret various geophysical data. Therefore in the present study the efficacy of joint inversion has been studied in enhancing the resolution of thin layers and reliability of the results. Global optimization using very fast simulated annealing, VFSA, is performed for 1-D Earth structures. The resolution problems in MT and direct current resistivity methods are studied. The synthetic noise-free and noisy apparent resistivity data from DC resistivity, Radial dipole array, and MT soundings are inverted individually and jointly over different types of layered Earth structures. The study reveals that global optimization of individual data sets cannot resolve the layers properly. Joint inversion of MT and DC measurements can overcome the problem. The study reveals that the resolution problem associated with a thin resistive layer can effectively solved by joint inversion. However, for a thin conducting layer it can only be reduced. It is rather surprising that even though MT method or in general EM method is insensitive for a resistive layer it can improve its resolution through joint inversion with DC resistivity data better than that for a conducting layer. 5.8p JOINT INVERSION OF ELECTROMAGNETIC AND SEISMIC DATA Dhananjai Pandey(1), Martin Sinha(2), Satish Singh(3), and Lucy Macgregor(1) (1) Bullard Labs., Madingley Road, Department of Earth Sciences, University of Cambridge, Cambridge, CB3 0EZ, UK (2) Southampton Oceanographic Centre, University of Southampton, Southampton, UK (3) IPG, Paris, France pandey@esc.cam.ac.uk Since the inception of geophysics, several techniques have been discovered to determine the substructure of the earth but overall, none of these techniques individually can give the complete and unambiguous subsurface picture. At this stage the idea of combining two or more techniques together becomes essential to get a reliable information. An approach to combine the seismic and electromagnetic method for joint interpretation is one further step towards this which will result in as a characterising information for the subsurface investigation. The regions with complex geology like flood basalts and a high velocity layer overlying a low velocity layer, the beauty of resolution of the subsurface parameters through the seismic method, which is one of the most useful geophysucal tool, is lost and the method fails to reveal completely the structures below the surface. Such geological examples can be well constrained by the magnetotelluric method because of the good resistivity contrast between these layers. Thus both of these techniques can be combined and both types of data sets can be inverted jointly to constrain the individual informations i.e. seismic velocities and the electrical resistivities from each other. This joint inversion and their interpretation will give the lithological information of the area. This kind of work will be very usefull for further resesrch works and hydrocarbon exploration in such areas. 5.9p NEW METHOD FOR 3-D ELECTROMAGNETIC INVERSION BASED ON ANALYTICAL APPROXIMATION P.S. Martyshko and A.L. Rublev Institute of Geophysics, Ural Branch of Russian Academy of Science, 620016, Amundsen str. 100, Ekaterinburg, Russia pmart@igeoph.mplik.ru Inversion of electromagnetic (EM) data in geophysical prospecting involves solution of a nonlinear-operator equation of the first kind (with an implicit, ill-conditioned, operator). Numerical solution of such equations can require considerable computer time. We do EM data inversion by two-stage interpretation method: (1) analytical approximation of the observed data with the fields of singular sources; (2) construction of equivalent objects with different physical parameters values. We have obtained the new explicit inverse problem equations of electromagnetic fields (relative to the boundary of anomalous object). There are the first generation equations with explicit operators. The material properties of the anomalous region are assumed to be parameters, i.e. the solution of the Inverse Problem holds for various values of conductivity and permability. As a result of interpretation we obtain the bodies with different values of conductivity (permeability) which generate the same electrical (or magnetic) field. We have obtained some numerical examples (for various boundaries between air and earth). We have studied the possibilities for solution uniqueness seeking on the based jont interpretation of different EMD. 5.10p INVERSION OF TIME DOMAIN INDUCED POLARIZATION DATA M. Hoenig, S. Recher, B. Tezkan and F.M. Neubauer University of Cologne, Institute of Geophysics and Meteorology, Germany hoenig@geo.uni-koeln.de Recording of the whole time series for induced polarization surveys allows the computation of the Cole-Cole relaxation parameters from the transient decay curves. A modified time domain electromagnetics algorithm is used to calculate the response of layered polarizable ground. The transient signals are then inverted using the Marquart-method to give the Cole-Cole parameters of each layer. Sensitivity studies with synthetic datasets for 1D polarizable media show a significant dependence of the parameter resolution on the magnitude of the IP-effect. Application of the algorithm to field data is used to give pseudosections of the Cole- Cole parameters for near surface structures. 5.11p INVERSION OF EMAP DATA Cicero R.T. Regis, Luiz Rijo Universidade Federal do Para, Belem, Brazil cicero@ufpa.br We present a method for inverting EMAP data which consists of first building an interpretative model from the filtered sections, then applying approximate equality constrains to achieve a stable solution to a parametric inversion. One characteristic of our interpretative models is the presence of a layer formed by small outcropping bodies, one for each dipole in the line, which are intended to simulate the effects of the static distortions in the inversion. We term that layer the "static shift layer". The resistivities in this layer tend to converge to the values of the features that generate the static distortions, where they occur, and to the value of the upper layer in those positions that are free of static shifts. The use of approximate equality constrains is necessary to stabilize the solutions of our least squares procedure. To determine which parameters need to be constrained we perform the singular value decomposition of the sensitivity matrix obtained in the last iteration of the unconstrained inversion, which generates unstable solutions. The parameters or combinations of parameters that need to be constrained in the inversion are the ones that generate the smaller singular values. We applied this method to synthetic data and to real field data from the Parana Basin. 5.12p INVERSION OF 2-D MT DATA CONSTRAINED BY DERIVATIVE OPERATORS OF ORDER GREATER THAN ONE Jose Gouvea Luiz and Luiz Rijo Departamento de Geofisica, Universidade Federal do Para, Belem, Brazil gouvea@ufpa.br Inversion of geophysical data requires a priori information about the parameters as a mean to stabilize the inversion process. In MT inversion problems a priori information is frequently introduced by constraining the parameters with an operator of first derivative, which imposes smoothness on the parameters yielding a poor delineation of the anomaly sources. In order to circumvent the delineation problem and still keep the inversion process stable, we carried out a study on the application of derivative operators of order greater than one as constraints to the parameters. Such operators can improve delineation because they incorporate information about the source distribution like symmetry and convexity, features usually observed on MT data. The study was done using the finite element algorithm and bi-dimensional models to represent a subsurface with one and two prismatic heterogeneity. The results show that the delineation of the anomaly sources increases with the use of derivatives of higher order when compared to that obtained with the conventional first derivative operator. They also show that stabilization can be achieved almost everywhere in the finite element grid even with operators of order 4. The effectiveness of the inversion technique has been evaluated by applying it on both noisy and noise-free data, and on COPROD2 profile real data. 5.13p IMPROVED TECHNIQUE OF 2D INVERSION OF MAGNETOTELLURIC DATA Ludmila Porokhova(1) and Irina Mardanova(2) (1) St. Petersburg State University, Dept. of Physics of Earth, Russia (2) University of Texas at Austin, Dept. of Computational and Applied Mathematics, USA Ludmila.Perokhova@pobox.spbu.ru At the present time, there are a lot of different methods for a 2D inversion of magnetotelluric data. Different codes give different results (for example, an inversion of COPROD2), which obviously depend on an algorithm used as well as on a technique of inversion. Our study for a number of models has showed that one needs a certain technique to use a code for the interpretation of MT data according to a type of a 2D cut. We propose the following technique. First of all, one needs to carry out a preliminary data analysis to split the maximal and minimal impedance curves into the curves, which correspond to the TE and TM modes. Then, one needs to consider amplitude and phase pseudosections to choose strategy of interpretation depending on the location of the anomalies. For example, if anomalies are in a deep part of a cut and sediments are horizontally uniform, then a better result is obtained after the alternative inversion of the modes: TE, TM, etc. If sediments have horizontal heterogeneities as well as a depth, then the independent inversion of TE and TM data is more effective. From the data on TE mode, we find the conductivity of a depth, while from the data of TM mode, we find the conductivity of sediments. Then two solutions are to be combined into the one and then, if necessary, the alternative inversion can be performed. We found, that the average inversion (when the whole dataset on TE and TM mode is used and the common misfit is calculated) did not give satisfactory results. The proposed technique was applied to the data observed on the South slope of Baltic shield. 5.14p AN APPROXIMATE THIN PLATE MODEL IN EM PROSPECTING Markku Pirttijarvi and Sven-Erik Hjelt Department of Geophysics, Institute of Geosciences, University of Oulu, POB 3000, FIN-90014 University of Oulu, Finland markku.pirttijarvi@oulu.fi An approximate method is presented for modeling and interpreting near-surface EM prospecting data using a conductive plate model embedded in two-layer host medium. In this method the thin plate is represented by a lattice structure composed of 2-D surface elements. The integral equation governing the EM problem is solved for the surface current densities in the lattice elements using point collocation. Analytic solutions for the primary surface Green's functions are solved using the approximate equations of the electric field of an dipole in conductive whole-space. To speed up the computation the secondary Green's functions are excluded and a simple depth compensation method is used to take the layered host medium into account. The forward problem is solved in frequency domain but TEM responses can be computed via cosine-transformation. Practical interpretation of field data is made possible using a parameter optimization method, which is based on linearized inversion. Examples of forward and inverse solutions are presented for the Slingram (HCPL) system. The SVD information obtained from the inversion method is used to analyze the resolution and the correlation of the plate model parameters. A comparison between the resolving capabilities of the FEM and TEM systems is also made. 5.15p EFFECTS OF UNDETECTED ANISOTROPY IN 2D INVERSION MODELLING Wiebke Heise and Jaume Pous Departament de Geodinamica i Geofisica, Universitat de Barcelona, Martí i Franqués s/n, 08028 Barcelona, Spain wiebke@geo.ub.es We present a simulation of two-dimensional models with lateral anisotropy and the 2D inversion of their synthetic data responses assuming isotropic structures. The data responses are obtained using the finite-difference algorithm of Pek y Verner 1997. Once obtained the anisotropic data we carried out the following steps, (a) dimension analysis and determination of the optimal 2D strike after the Swift criterion, Groom-Bailey analysis and induction arrows; (b) rotation of the impedance tensor in direction of the obtained strike; (c) 2D inversion using the REBOCC algorithm. For anisotropic models with an anisotropy strike different from the 2D strike direction we find that there are significant differences between the analysed models concerning the behaviour of the induction arrows. The induction arrows are severely affected by anisotropic structures. Their direction depends on the anisotropy strike, the resistivity contrast and the period. Although the model can not be considered 2D, assuming field data with noise we could determine a preferred strike direction. In this case the Swift criterion indicates a direction about the anisotropy strike. When the data are rotated to the determined optimal strike angle, the inversion reproduces approximately the initial model with the typical sequence of conductive and resistive dikes -macroscopic anisotropy-, although the data fit is not satisfying. 5.16p BOUNDARY ELEMENT METHOD (BEM) AND 2D OR 3D GEOELECTRICAL STRUCTURE ANALYSIS Qian Jiadong(1), Ma Qingzhong(2), and Mao Xianjin(3) (1) Center for Analysis and Prediction, China Seismological Bureau, P.O. Box 166, 63 Fuxing Ave., Beijing, 100036, China (2) Department of Geology and Geography, Lanzhou University, Lanzhou, 730000, China (3) Seismological Bureau of Yunnan Province, Kunming, Yunnan Province, China jdqian@sdb.cdsi.ac.cn This paper deals with the applications of Boundary Element Method in the survey of apparent resistivity on the site with the 2D or 3D geoelectrical structure with better accuracy, more convenience and less computations than the previous ones such as Finite Element Method, Finite Difference Method which demand huge amount calculations with inconveniences. Furthermore, much more complicated cases in horizontally layered models with 2D or 3D heterogeneous bodies contained inside some of layers have been also dealt with in the paper with the treatment of the infinitely extended interfaces between adjacent layer. The special basic solutions in BEM have been taken to let the infinitely extended interfaces disappeared, so that the integral equations only contain the term of integral along the boundaries of heterogeneous bodies without the demands for taking the integral along the infinitely extended interfaces into account. The method has provided good tools for solving the problems of heterogeneous structures and a good base either for forward problems or inversion. 5.17p SENSITIVITY OF 2D GEOELECTRICAL MODELING TECHNIQUES TO IMAGE SHALLOW STRUCTURES HAVING LOW RESISTIVITY CONTRASTS Ana Osella, Nestor Bonomo, Patricia Martinelli and Eugenia Lascano Dpto. de Fisica, Universidad de Buenos Aires, Conicet, Ciudad Universitaria, Pab. 1, Buenos Aires, Argentina osella@df.uba.ar A geoelectrical survey has been carried out to image shallow buried structures which results will be applied to characterize an archaeological site in South Argentina. This first study was focused to improve the sensitivity of the methodology. According to it, we selected a sub-area of approximately 20 m long and 6 m width. Two resistivity profiles distributed along and across the structure, respectively, were carried out. To have a better resolution when inverting the data, we deployed both parallel and perpendicular configurations at each site. A two-dimensional method previously developed was modified in order to allow calculating the responses of shallow structures for both configurations. For each profile, we first inverted the data obtained along perpendicular deployments at each site and with the results we constructed a tentative cross- section. Then, we used this cross-section as input for the inversion process, where we considered both configurations to make the fitting to the data. To compare the results, we applied this methodology both for Schlumberger as well as Wenner geometries. In this way, we could minimize uncertainties in the model, obtaining a better definition of the structure. 5.18p DETECTION OF 3D CONDUCTIVE STRUCTURES BY 2D INVERSION OF MT DATA Ana Osella, Patricia Martinelli, and Denise Tortorella Dpto. de Fisica, Universidad de Buenos Aires, Conicet, Ciudad Universitaria, Pab. 1, Buenos Aires, Argentina osella@df.uba.ar Our study was focused on the analysis of the sensitivity of 2D inversion methods to image a magmatic chamber. This structure was modeled as a deep 3D conductive body embedded in a medium which parameters were selected taking into account characteristic values for this kind of environment. A numerical simulation of the electromagnetic response was performed using a 3D Rayleigh-Fourier forward method. These synthetic data were calculated along two different profiles and then they were inverted using 2D codes. We analyzed the results at each profile when inverting the parallel and perpendicular components of the apparent resistivity and phase curves in order to determine the best methodology to properly image this deep conductive structure. 5.19p LONG PERIOD EFFECTS OVER 2D STRUCTURES DUE TO SPATIALLY NON-UNIFORM 2D AND 3D INDUCING FIELDS Patricia Martinelli and Ana Osella Dpto. de Fisica, Universidad de Buenos Aires, Ciudad Universitaria, Pab. 1, Buenos Aires, Argentina osella@df.uba.ar In previous works, we presented a 2D magnetotelluric modeling method based on the application of Rayleigh scattering theory; later, we generalized this method for the calculation of the electromagnetic response of 2D structures to spatially non-uniform 2D and 3D inducing fields. These sources are characteristic of low and high latitude regions. Here we calculate the response to different 2D and 3D sources, of various 2D structures representative of conductivity distributions which could be found in subduction zones, including deep conductive anomalies in the lower crust and upper mantle. We investigate the effect that the spatial dependence of the inducing fields exerts on the long period range of the responses, as a function of the characteristic lengths of the external fields along and across the symmetry direction of the structure. We also analyze the coast effect present in almost all these regions. 5.20p NEW ASPECTS OF SOUNDINGS AT LOW INDUCTION NUMBERS Francisco J. Esparza and Enrique Gomez-Trevino CICESE, Km 107 Carr. Tij.-Ensenada, Ensenada, B.C., 22860 Mexico fesparz@cicese.mx Low-frequency electromagnetic data can be described in terms of a null real component of the electric field, an imaginary component of the magnetic field proportional to frequency, and the real part of the magnetic field independent of the subsurface electrical conductivity. By imposing this properties of the electromagnetic fields to Maxwell's equations, a general relation can be found between the imaginary part of the magnetic field and a vertical profile. By using standard integral transforms, the integral equation for the electrical conductivity can be formally inverted in terms of the apparent conductivity. In this paper we present a formal inversion for the horizontal magnetic dipole source. This solution and the corresponding formal inversion for the vertical magnetic dipole source found earlier, completes the formal inversion at low induction numbers. Besides the analytical inversion, we present a practical inversion method for finding the subsurface conductivity based in linear programming techniques. The numerical method is applied both to synthetic data and to field data. 5.21p SOME MT INTERPRETATION PROBLEMS I.S. Feldman, B.A. Okulessky Electromagnetic Research Center, 3, Jubileinaya str, 142092, Troitsk, Moscow region, Russia emrcingf@mtu-net.ru Applicability of 2-D and 3-D inversions is essentially limited by the influence of inhomogeneity of the upper part of a section (so called geologic noise). In most cases, taking into consideration of such influence by calculation is impossible since the inhomogeneities are characterized with higher frequency spatial spectrum and 3-D shape. It is proposed to consider such influence as a summarized action of multiplicative (shift effect) and additive random noise. Various algorithms of interaction and filter are considered; the effectiveness of those are shown on theoretical models and field data. A method of correction of 1-D interpretation for quasi-E- polarized MTS curves is considered. The action of induction effects in the low-frequency domain is corrected for by the use of 2-D modeling on a big grid (200x600) with a detailed frequency step (NT"100) in an interactive selection regime. The problem of stabilization of the 1-D inverse problem solution at a profile by means of interactive selection of smooth functions of 2-D model parameters is discussed. 5.22p IS THE SINGULAR VALUE DECOMPOSITION SUITABLE FOR RESOLUTION STUDIES IN THE 2D-CASE? Katrin Schwalenberg(1) and Volker Rath(2) (1) GFZ Potsdam, Telegrafenberg, 14473 Potsdam, Germany (2) FU Berlin, Fachrichtung Geophysik, Malteserstr. 74-100, Haus D, 12249 Berlin, Germany Katrin@gfz-potsdam.de Being the most powerful tool for linear numerical resolution studies we used the Singular Value Decomposition to calculate eigenvectors as well as resolution- and covariance matrices from the sensitivity matrices. The latter is the connecting term between model and data space and can be calculated for any given model. First we apply these quantities in the one-dimensional case, were only a few parameters describe the problem. A 1-D inversion was carried out with different regularization schemes incorporated to study the relationship between the singular value spectrum and the regularization parameters. In a final step we present the sensitivity and resolution matrices for a 2-D model from the Central Andes, where the large parameter space renders the visualization more difficult.