**Robust Methods of Multistep Velocity Analysis**

**KATZ, SIMEON, University of Southern California, Los Angeles, CA**

The stability of velocity analysis results obtained in the presence of strong noise may be improved if several adjacent CMP gathers are processed, thus increasing the number of the input records for velocity analysis, and if a better velocity analysis operator is utilized. The two classes of the velocity analysis techniques discussed here are based on the joint processing of a group of adjacent CMP gathers and are characterized by high stability of the output results.

The algorithms of the first class are less sensitive to perturbations in dips of the reflectors, so that approximate values of the dips may be used. The algorithms of the second group are dip sensitive so that depth dependent dips should be taken into account.

In each of these classes different velocity operators may be used to generate an output velocity analysis section. I consider four types of the velocity analysis operators: (1) modified semblance, (2) penalty function defined as a monotonic function of the signal-to-noise ratio estimate, (3) linear stack velocity function, and (4) linear stack modulated by the penalty function.

The results of the experiments with these velocity analysis techniques indicate that in the case of semihorizontal interfaces the best results are obtained via use of the operator (4). If the dip of the reflector is unknown and may be large, the velocity operators (2) and (1) may give better results.

AAPG Search and Discovery Article #91009©1991 AAPG-SEPM-SEG-SPWLA Pacific Section Annual Meeting, Bakersfield, California, March 6-8, 1991 (2009)