Translation invariant attributes for effective lithofacies discrimination

Abstract

This study introduces translation-invariant attributes using the Hadamard and Rapid transforms of seismic data to discriminate lithofacies. Unlike the Fourier transform, which projects the data onto a set of orthogonal sinusoidal waveforms, the Hadamard transform projects data onto a set of square waves called Walsh functions. The Hadamard transform is particularly good at finding repeating, stacked vertical sequences. Crossplot analysis indicates translation-invariant attributes for horizon-based time windows are less sensitive to horizon interpretation errors in the reservoir interval. Therefore, translation-invariant attributes can be used to find a particular geologic pattern of interest in a reservoir interval. These attributes have been applied successfully to discriminate the lithofacies in a reservoir interval of interest for a 3D seismic survey in the upper Assam basin of India. The sand-shale thickness ratio at the drilled locations within the reservoir interval is used to define the lithofacies categories. Using the discriminant function analysis of each of the conventional-seismic, principal-component, and translation-invariant sets of attributes, we create a corresponding discriminant score map. The probability density function (PDF) of the discriminant score at the drilled locations transforms the PDF to lithofacies categories. The comparative analysis presented in this study indicates that translation-invariant attributes are superior to conventional-seismic and principal-component attributes in discriminating the lithofacies.