Cooperative deep learning inversion of controlled-source electromagnetic data for salt delineation

Various geophysical data types have advantages for exploring the subsurface, and more reliable exploration can be realized through integration of such data. However, the imaging of physical properties based on deep learning (DL) techniques, which has received considerable attention because of its enormous potential, has generally been performed using only a single type of data. We have developed a cooperative inversion method based on supervised DL for salt delineation. Controlled-source electromagnetic (CSEM) data, which can effectively distinguish a salt body with high electrical resistivity from the surrounding medium, are used as data for cooperative inversion, with high-resolution information derived from seismic data used as the constraint. This approach can entrain seismic information into a fully convolutional network designed to invert CSEM data to reconstruct the resistivity distribution. The inversion network is trained using large synthetic data sets, including the seismic information derived from seismic data as well as CSEM data and resistivity models. A cooperative strategy for reasonable entrainment of seismic information into the inversion network is established based on analysis of the network and kernels of the convolutional layers. The performance of the proposed method is demonstrated through experiments on test data generated for resistivity models for complex salt structures. The trained cooperative inversion network shows improved salt delineation results compared to the independent inversion network, irrespective of noise levels added to the test data, due to restriction of the resistivity model to fit seismic information. Moreover, training with noise-added data decreased the effects of noise on prediction results, similar to the case of adversarial training. We develop the possibility of combining geophysical data with a constraint using DL-based techniques.

Sensitivity analysis of optimization parameters on cooperative inversion of seismic reflection and gravity data

<p>Reflections of seismic waves are strongly distorted by the presence of complex geological structures (e.g. salt bodies) and their vertical resolution is usually of the order of a few tens of meters, imposing limitations in the construction of subsurface models. One way to improve the reliability of such models is to integrate reflection seismic data with other types of geophysical data, such as gravimetric data, since the latter provide an additional link to map geological structures that exhibit density contrasts with respect to their surroundings. In a previous study, we developed a cooperative inversion method of 2D post-stack and migrated reflection seismic data, and gravimetric data. Using that inversion method, we minimize two problems: (1) the problem of the distortion of reflection seismic data due to the presence of complex geological bodies and (2) the problem of the greater ambiguity and the commonly lower resolution of the models obtained only from gravimetric anomalies. The method incorporates a technique to decrease the number of variables and is solved by optimization of the gravity inverse problem, thus reducing computing time. The objective function of cooperative inversion was minimized using three different methods of optimization: (1) simplex, (2) simulated annealing, and (3) genetic algorithm. However, these optimization methods have internal parameters which affect the convergence rate and objective function values. These parameters are usually chosen accordingly to previous references. Although the usage of these standard values is widely accepted, the best values to assure effectiveness and stability of convergence are case-dependent. In the present study, we propose a sensitivity analysis on the internal parameters of the optimization methods for the previously presented cooperative inversion. First, we developed the standard case, which is an inversion performed using all parameters at their standard values. Then, the sensitivity analysis is performed by running multiple inversions, each one with a set of parameters. Each set is obtained by modifying the value of a single parameter either for a lower or for a higher value, keeping all other values at their standard values. The results obtained by each setting are compared to the results of the standard case. The compared results are both the number of evaluations and the final value of the objective function. We then classify parameters accordingly to their relative influence on the optimization processes. The sensitivity analysis provides insight into the best practices to deal with object-based cooperative inversion schemes. The technique was tested using a synthetic model calculated from the Benchmark BP 2004, representing an offshore sedimentary basin containing salt bodies and small hydrocarbons reservoirs.</p>

Decoupling strategy for large-scale multiphysics joint inversion

<p>We present a generalized 3-D multiphysics joint inversion scheme with a focus on large-scale regional problems. One of the key features of this scheme is the formulation of the structure coupling as a sparsity-promoting joint regularization. This approach makes it possible to simplify the structure of the objective function and to keep the number of hyperparameters relatively low, so that the inversion framework complexity scales well with respect to the number of geophysical methods and possible reference models used. To further simplify adding geophysical solvers to the framework and to optimize the discretization, we propose an alternating minimization scheme that decouples the inversion and the joint regularization steps. Decoupling is achieved by introducing an auxiliary multi-parameter model. This allows the individual subproblems to make use of problem-tailored grids and specialized optimization algorithms. As we will see, this is in particular important for the regularization subproblem. In contrast to straightforward 'cooperative inversion' formulation, decoupled inversion steps appear to be regularized by a standard quadratic model-norm penalty, and as a result existing separate inversion codes can be used with minimal, if any, modifications. The developed scheme is applied to magnetotelluric, seismic and gravity data and tested on synthetic model examples.</p>

Cooperative inversion of gravity and seismic data with different spatial coverage

<p>Combining two or more geophysical datasets with different resolutions and characteristics is now a common practice to recover one or more physical properties. Building 3D geological models for mineral exploration targeting is often an expensive task even for inversion of a single dataset, because of extremely complicated structures with small scale targets. In this context, seismic methods, among all other traditional techniques in mineral exploration, are receiving increasing attention due to their higher resolution in depth. With more limited spatial coverage and higher resolution, they are usually used to refine the interpretation of potential field data.</p><p>As each seismic survey is designed for a particular intention with specific targets and may not be available in all regions of interests, we develop an iterative cooperative inversion algorithm for inverting gravity and seismic travel-time data. This enables the utilization of localized high-resolution seismic data in a larger full 3D volume which is covered by gravity data. Geological information in the form of probabilistic geological modelling is used to extend information away from the high-resolution data and constrain the inversion result. We use these data as the prior model and to derive constraints incorporated into the objective function of gravity inversion. This allows us to obtain information about the probability of the presence of lithologies associated with the formation of mineral systems. To ensure structural consistency between density and velocity we develop a geologically constrained structure-based coupling technique following the same principle as the cross-gradient technique but with a higher degree of freedom in spatial directions. We apply local structure-based constraints conditioned by a geological probability distribution, which is considering direction and magnitude and provide a higher degree of freedom for model variations. An investigation of the proposed methodology and a proof-of-concept using realistic synthetic data are presented. Our results reveal that the methodology has the potential to constrain the gravity inversion results using the limited seismic data.</p>

New perspectives on Solid Earth Geology from Seismic Texture to Cooperative Inversion

Seismic and electromagnetic methods are fundamental to Solid Earth research and subsurface exploration. Acquisition cost reduction is making dense 3D application of these methods accessible to a broad range of geo-scientists. However, the challenge of extracting geological meaning remains. We develop the concept of “textural domaining” for 3D seismic reflectivity data. Dip-steered seismic texture attributes are combined with unsupervised learning to generate sets of volume rendered images accompanied by a seismic texture reference diagram. These methods have the potential to reveal geological and geotechnical properties that would otherwise remain hidden. Analysis of seismic texture presents particular value in hard-rock settings where changes in velocity may be negligible across rock volumes exhibiting significant changes in rock mass texture. We demonstrate application and value of textural domaining with three industry-scale field examples. The first example links seismic texture to rock type along a 400 km long transect through central Australia. The second and third examples partition dense 3D seismic data based on texture for complex hard rock terrains in Nevada, USA and Kevitsa, Finland. Finally, we demonstrate application of domaining within texture guided cooperative inversion of 3D seismic reflectivity and magnetotelluric data to provide new perspectives on Solid Earth geology.