Experience with four-component iterative Gabor deconvolution

David C. Henley, Gary F. Margrave

The seismic processing operation of deconvolution is aimed at removing as many of the earth's filtering effects on a set of seismic data as possible. The Gabor algorithm was first incorporated into deconvolution to accommodate the known non-stationarity of seismic data. Subsequent improvements enabled it to accommodate localized surface- related effects, as well as the more slowly varying Q effects. An algorithm recently developed for ProMAX derives, in the Gabor magnitude spectral domain, individual deconvolution operators which consist of four components; two attributable to source and receiver surface locations, one to midpoint location, and one to source-receiver offset. Because the initial factorized estimates of these components are somewhat ad hoc, the latest version of the algorithm allows iterative improvements of the components using a scheme resembling the ART algorithm used in transmission tomography and other applications. We show here results of using the new algorithm, compared to the older Gabor2 algorithm, which was capable of only single trace deconvolution or ensemble average deconvolution. All results are real data. In general, we find the four-component algorithm provides broader bandwidth than the ensemble average Gabor operation and better event phase stability than the single-trace Gabor algorithm. Also, no more than two iterations appear to be necessary to reach a stable result in the Gabor_sc algorithm.