Dept. of Geosciences Colloquium: Stochastic Impedance Inversion Taking into Account Diffractive Component of the Total Wavefields

Zoom: https://zoom.us/j/99062683146?pwd=YWlkcUNYSndSWkNuSzY4dUx0NXJjQT09

Abstract:

Seismic diffraction encodes information regarding medium- and small-scale subsurface objects in the subsurface. The last decade of diffraction imaging has been used for fault, pinch-out, and fracture detection and localization. Very little research, however, has been undertaken with regard to taking into account diffraction in the impedance inversion. Usually, in the standard inversion scheme, the input to the inversion is the migrated data, and it is assumed that the energy of diffraction is focused during the migration process. However, this is true only when a perfect velocity model is provided and the true amplitude migration algorithm is implemented, which is rare in practice. Therefore, a novel approach is proposed to implement poststack impedance inversion considering the unmigrated input data and accounting for the diffractive component of the wavefield. The input data for the inversion is a zero-offset section. The accurate approximation of the zero-offset section can be obtained using a special case of a multifocusing stack with controlled reflection point smearing. Inversion, as usual, consists of minimizing the difference between observed and modeled data. Forward modeling for the impedance inversion includes the specular reflection and diffraction components of the total wavefield. The inversion result is composed of impedance perturbation and a low-frequency model. The impedance perturbation is estimated in the unmigrated domain and mapped to the depth domain by the ray migration procedure, while the low-frequency model is built using geological interpretation and well-log information. Additionally, under a Bayesian framework, stochastic inversion is utilized to estimate model perturbation with given prior information. During the optimization process, to generate and efficiently update the obtained models, the improved Markov chain Monte Carlo simulation is implemented by adaptive particle swarm optimization. Numerical modeling tests were performed to verify the proposed method and to find the best inversion methodology and optimal parameters. A special dataset obtained by a physical modeling experiment was used in this study. Data obtained from physical modeling can be considered as a bridge connecting the field seismograms and geological features. The physical model was designed to simulate carbonate karst reservoirs, and model construction included a 3D printer. Finally, the proposed inversion method was applied to several field datasets. The results demonstrated that this method can improve the resolution and accuracy of the impedance model.

Event Organizers: Dr. Roy Barkan and Dr. Asaf Inbal