Cluster Analysis of Lithology Grouping Trends using Principal Component Spectral Analysis and Complex Seismic Attributes

Abstract

Cluster analysis is used to determine possible lithology groupings on the basis of information from seismic data. Specifically, k-means is used in the cluster analysis of different lithologies. The data center is determined randomly and updated through an iterative process (unsupervised). The cluster analysis process involves combinations of complex seismic attributes and spectral decomposition as inputs. The complex seismic attributes are reflection strength and cosine phase. Reflection strength clearly describes the lithology boundary while the cosine phase describes the lithologies. Spectral decomposition is used to detect the presence of channels. The resolution of seismic data generally reaches 90 Hz. Spectral decomposition can produce outputs with up to 1 Hz intervals. The spectral components are correlated and repeated. To reduce the repetition of spectral data and increase the trend within the data, we use principal component spectral analysis. We apply and validate the workflow using the seismic data volume acquired over Boonsville, Texas, USA. The results of the cluster analysis method show good consistency with existing lithological maps interpreted from well data correlations.