Interval velocities from moveout velocities over a seismic reflection survey area.

Moveout velocities sampled frequently along seismic horizons on a selection of seismic lines are used to derive interval velocities in an 'inversion' algorithm developed from work published by Hubral. This inversion is based on zero-offset raytrace modelling in a simplistic local ground model. The 'Hubral algorithm' is incorporated into a database which allows spatial smoothing of velocities. The spatial consistency of derived interval velocities can then be assessed by reference to mis-ties at line intersections, while interval velocities from well data can be used to check their validity. These principles have been used to derive interval velocities both from real data and from 'synthetic' data generated by common mid-point raytracing over schematic ground models. The latter study reveals that the procedure performs well if the local subsurface sampled by the CMP gather conforms approximately to the simplistic ground model assumed by the Hubral algorithm. The method is unsuitable in areas of faulting and interval velocity heterogeneity, and may yield spurious results over fold axes. Application of the procedure to real data indicates that it is generally desirable to smooth both moveout velocities before inversion and interval velocities after inversion. Comparison with well information shows that interval velocities derived by the Hubral algorithm are consistently higher than those measured from calibrated velocity logs. This observation is disturbing, since the derived interval velocities require a correction if they are to be used for depth conversions, but the discrepancy cannot be explained by ray theoretical considerations. No advantage appears to be gained by the 'layer-by-layer' mode of inversion over the 'direct' inversion, despite the greater potential for error propagation anticipated in the latter. Further work on different data sets is required to justify general use of the layer-by-layer mode of inversion