In the initial stages of an exploration project, a seismic data processor is often presented with a set of seismic data acquired over a geological prospect whose subsurface features and properties are either partially or completely unknown. It is then the processor's job to extract as much detail as possible through appropriate data processing. To simulate and illustrate the process, two data sets from the CREWES physical modeling facility were acquired by Wong (henceforth known as the Modeller) and presented to Henley (termed the Processor) for imaging, but with no information conveyed about the physical model structure or materials, except that the two surveys simulated 'baseline' and 'time-lapse' realizations of the same model. Each set of seismic data was processed independently, and algorithm parameters were deduced only from the data characteristics. Features of the processed data were then used to evaluate three different proposed 'models' supplied by the Modeller. Only one of these was consistent with images and other information obtained from the processed seismic data. This experiment was a very instructive illustration of the seismic exploration process.
The 'time-lapse' version of the model created a family of strong multiple reflections in the corresponding seismic data. As part of our analytic processing, we used this opportunity to test a multiple-elimination technique applied in the common-ray-parameter domain. Two variations of the technique were found effective on surface-related multiples in this geometrically simple model. We expect the same technique to be useful on seismic field data as well, but it probably loses its effectiveness with increasing geological complexity, decreasing regularity of acquisition geometry, and increasing levels of noise.
View full article as PDF (14.77 Mb)