Clustering of attributes by projection pursuit for reservoir characterization

From the values of each trace occupying a reservoir interval (or other time segment) on a (flattened) seismic section a set of p attributes can be computed and assigned to, say, the CDP location of the trace. Attributes could, for example, include energy and bandwidth. For a set of N traces, N points in a p-dimensional space are obtained. The ability to determine the clustering of such points in the high-dimensional space then corresponds to recognizing the locations of traces that have similar properties (governed by the attribute values) over the reservoir interval. This ability provides a useful first step in characterizing the reservoir interval.