Improving fault delineation using maximum entropy multispectral coherence

Abstract

Knowledge of fault geometry plays an important role in reservoir modeling and characterization. Seismic attributes, such as volumetric dip, coherence, and curvature, provide an efficient and objective tool to extract fault geometry attributes. Traditionally, we use noise-attenuated full-bandwidth seismic data to compute coherence followed by smoothing, sharpening, and skeletonization. However, different stratigraphic reflectors with relatively similar waveforms and amplitudes juxtaposing across a fault will algorithmically appear to be continuous, with the resulting fault image being broken. This leads to pseudo fault breakpoints and challenges the accurate extraction of other fault geometric attributes. Because the phase of the similar reflections across the faults varies with different spectral components, such nonstratigraphic alignments occur for only a few spectral components such that a multispectral coherence algorithm produces more continuous fault images. We have evaluated the influence of spectral voice selection and spectral decomposition algorithm on the quality of fault imaging in multispectral coherence images using a 3D seismic survey acquired in the Taranaki Basin, New Zealand. Of the algorithms evaluated, we find that the high-resolution maximum-entropy-based multispectral coherence method provides better results than those based on other spectral decomposition algorithms, which especially improves the fault continuity. However, the lateral resolution of fault imaging in multispectral coherence is decreased compared to the full-bandwidth coherence, because the fault image is smeared when we combine the coherence volumes computed using different spectral voices. We perform a fault enhancement workflow on the maximum-entropy-based multispectral coherence volume to improve the lateral resolution of fault imaging, which helps delineate the minor faults.