Postmigration multiple prediction and removal in the depth domain

Abstract

We have developed a new methodology for predicting and removing multiples in the postmigration depth domain based on wavefield extrapolation and attribute-based subtraction. The inputs for the multiple prediction are a 3D prestack depth-migrated stack volume and the corresponding migration velocity volume. The output is the predicted multiple model in the migration depth domain. In some cases, the strong residual top of salt multiple may be erroneously picked as the base of salt reflection. With the predicted multiple model available for comparison during the salt model building stage, there is a better chance of building an accurate salt model and avoid picking multiple events. In an effort to further improve the final migrated images, the predicted multiple model is used to remove residual multiples in the migration depth domain. A poststack wavefield extrapolation-based multiple prediction is used to identify and confirm the multiple events in the migration depth domain. Once multiple events are identified, an effective and efficient demultiple technique is applied to remove the residual multiples from the final migration. The key ingredient of this new demultiple methodology is the attribute-based subtraction. We describe the main steps of this methodology and demonstrate its effectiveness by showing some field data applications.