--> Abstract: Iterative Interactive Interpretation of Principal Component Clusters from Seismic Amplitude Data: A Rapid Geologic Feature Extraction Method, by Sebastien Bombarde, Jay R. Scheevel, and Karen Payrazyan; #90914(2000)

Datapages, Inc.Print this page

Sebastien Bombarde1, Jay R. Scheevel2, Karen Payrazyan1
(1) Chevron Petroleum Technology Company, San Ramon, CA
(2) Chevron Petroleum Technology Company, Rangely, CO

Abstract: Iterative Interactive Interpretation of Principal Component Clusters from Seismic Amplitude Data: A Rapid Geologic Feature Extraction Method

We apply Principal Component Analysis (PCA) to the problem of structural and stratigraphic interpretation from 3D seismic amplitude data. We use PCA in a novel way with the purpose of identifying detailed reservoir structures in a channellized sand and shale reservoir that is further complicated by faulting. We track these features using an interactive seed-picking method we call Interactive Iterative Interpretation (III)

In order to apply PCA methods to a single-property, 3D-seismic amplitude volume, we sample the volume with a small sliding sampling window that captures multiple, vertically-adjacent, amplitude samples as a vector. The sampled vectors from all such windows in the seismic dataset forms the set of input vectors for the PCA algorithm. Final output from the PCA algorithm is a cube of assigned classes of clustered Principal Components (PC’s). Applying such a method will allow for identifying unique vertical sequences rather than just local amplitude variations.

Interactive Iterative Interpretation (III) of geological features is performed on the PCA clusters. Using a combination of automatic seed-picking and manual interpretation (mainly for faults), allows us to rapidly interpret unique geological events (shales, channels, etc.). Fault-block specific picking is ensured using III, because it automatically uses picked faults as barriers to limit seed-pick propagation.

We find this method useful because high-resolution regions of unique seismic character are clustered categorically by PCA, in contrast to the continuously varying character of the amplitude. This method of feature mapping is especially useful in areas where rapid lateral geological variation and faulting exists.

AAPG Search and Discovery Article #90914©2000 AAPG Annual Convention, New Orleans, Louisiana