Estimating and interpreting velocity uncertainty

in migrated images and AVO attributes H. Grubb⁄, A. Turaz, and C. Hanitzsch⁄⁄ Estimating a suitable velocity field for use in prestack depth migration is inherently uncertain because of limi-tations on the available data and estimation techniques. This uncertainty affects both the migrated depth of struc-tures and their amplitudes in the inverted images. These effects can be estimated by performing multiple migra-tions with a set of velocity fields and colocating features in the migrated images. This lets us examine the imaging procedure’s sensitivity to changes in the velocity field so we can assess both structural and amplitude uncertain-ties in migrated images. These two types of uncertainties affect interpreta-tion in different ways. For instance, with structural un-certainty interpretation we consider the change in mi-grated location of structures when deciding on drilling locations, optimizing well trajectories, or computing un-certainty in volumetric calculations. With amplitude un-certainty or amplitude versus offset (AVO) uncertainty interpretation, we consider (1) uncertainty in crossplots of pairs of AVO attributes at a point of interest or (2) uncertainty of the attribute values along identified struc-tures. For any interpretation informing a decision, the uncertainty can help estimate risk. Our data processing approach is based on amplitude-preserving prestack depth migration followed by AVO inversion, or AVO migration=inversion. It is valid for estimating AVO attributes in simple to moderately com-plex structural settings. Our methods of assessing the effect of velocity uncertainty can also be applied when obtaining structural uncertainties for a complex overbur-den geology or amplitude uncertainties in conventional NMO-based AVO analysis. They may also be applied straightforwardly to any poststack attribute analysis. Key to the approach is the availability of multiple velocity fields to generate multiple migrated images. In our ap-plication, an automatic algorithm samples possible fields, but the set of fields to consider could be generated from another source, such as interpretation