Pattern analysis of seismic records

Abstract

The three steps in automated analysis of seismic records are trace matching, event detection, and seismic zoning. Our algorithm matches traces by comparing sequences of peaks on neighboring traces and edits pairs of peaks identified as similar for consistency in trend across and down the record. By combining connected pairs, we obtain laterally coherent events of varying quality which divide a record into zones that may have lithologic significance. These zones can be obtained automatically by applying cluster analysis to seismic attributes and other discriminating properties of these events. On common‐shot gathers from a four‐fold survey in Saskatchewan, Canada, the coherent events which we detected this way are consistent with those recognized visually and are associated with clear formation boundaries. Strong reflections correspond to events detected with high confidence, and the seismic zones show good correlation among shot records. High detection confidence and good correlation indicate that the application of automatic analysis procedures to common‐midpoint gathers from surveys of higher multiplicity is possible and may lead to a new and quick way of obtaining velocity profiles. Zoning, if done properly, provides a compact description of seismic records that may simplify telemetry of large volumes of data.