Using Principal Component Analysis to Decouple Seismic Diffractions from Specular Reflections
This presentation was offered as part of the Technical Program at the SEG 2020 Annual Meeting.
Methods developed for diffraction imaging are commonly based on a separation technique between continuous and discontinuous structural elements, based on their seismic response. Seismic imaging in the dip-angle domain decomposes direction-dependent images where the response of different structural elements is clearly distinguished. The subsurface structural geometry dictates a local dip-angle seismic signature of preferable scattering directions, accordingly. Assuming these signatures are uncorrelated, we propose a practical workflow for dip-angle domain principal component analysis as a comprehensive structural feature separator. Once the dip-angle images are transformed into their principal components, a back projection yields structural-specific images of the subsurface main building blocks. Our proposal emphasizes the strength of principal component analysis as a producer of probable geologic features from pre-stack dip-angle data. These should be incorporated as structural attributes to enhance seismic interpretation, reduce uncertainty and enable automation.
Dr. Raanan Dafni is R&D Project Manager, Seismic Imaging in the Emerson E&P Software group. Among other interests, his research expertise includes seismic processing and inversion, high-performance computing, automatic interpretation, and machine learning. He holds a PhD in Applied Geophysics from Tel-Aviv University, and served as a postdoctoral research associate in Rice University, Houston. He is the author or co-author of numerous peer-viewed articles.