Joint Bayesian AVO and RMO inversion

Abstract

Seismic amplitude variation with offset (AVO) has proved to be a useful technique in oil exploration. However, the sensitivity of AVO to residual move-out (RMO) is considered a severe limitation. Small misalignments between corresponding reflections have a severe impact on the gradient, which may cause the user to acquire the wrong conclusion. To make AVO analysis more robust, Semblance, Cross correlation and Swans residual velocity indicator (RVI) have earlier been proposed as methods to correct for RMO. These existing methods are reviewed. In addition, a new joint Bayesian AVO and RMO inversion is developed to find the RMO and elastic parameters. The methods are all tested on one seismic section where the intercept and gradient follow a colinear relationship, and another one containing an AVO class 2p anomaly deviating from a colinear relationship between the intercept and gradient. Semblance has traditionally been used during processing of the data and calculates the normalized stacking amplitude along different trajectories. However, the method lacks the sensitivity needed to properly condition the data for AVO and mishandles the anomaly. Cross correlation is a simple technique to quantify the similarity between the near and far traces relative to each other in time, but is distorted by heavy noise and the differences between the traces due to the anomaly. Swans RVI provides the necessary sensitivity to the small velocity errors Semblance lacks, but is heavily reliant on a co-linear relationship between the intercept and gradient. The newly developed joint Bayesian AVO and RMO inversion provides the necessary sensitivity, in addition to handling the AVO anomaly properly.