Online Journal for E&P Geoscientists

Aldawood, Ali A.^*1; Alshuhail, Abdulrahman ¹; Hanafy, Sherif ¹
(1) Physical Science and Engineering, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia.

Building an accurate near-surface velocity model is an essential step for many applications in seismic data processing such as wave-equation redatuming, statics correction, and tomographic inversion. Achieving an accurate model requires precise first-break picking of refracted energy. However, refracted energy can be degraded due to geometrical spreading and anelastic attenuation. Furthermore, refraction first breaks at far offsets are relatively weak in amplitude (longer traveled distance) and usually tend to be obscured by noise. This limits the maximum offset that can be incorporated into first-break picking since distinction between signal and noise then becomes subjective.

It follows from basic raypath principles that far-offset traces bear the signature of deeper parts of the near-surface geologic section. To maximize the usable offset aimed at accurate deeper velocity estimate of the near surface, we propose the method of super-virtual refraction interferometry to enhance the weak energy at far offsets.

We use interferometric Green's functions here to redatum sources by cross-correlating two traces recorded at receiver stations, say A & B, due to a source at location W. The result is a redatumed trace with a source at A and receiver at B. Similarly, the same redatumed trace can be obtained by cross-correlating two traces recorded at A & B but due to different shots. Intuitively, stacking a number of them would enhance the signal-to-noise ratio of this new redatumed "virtual" trace.

We next augment redatuming with convolution and stacking. The trace recorded at B due to a virtual source at A is convolved with the original trace recorded at A due to a source at W. The result is a "super-virtual" redatumed trace at B in the far-offset due to a source at W. Stacking N traces, again, enhances the signal by a factor of √N.

We applied our method to both synthetic and field data recorded over a complex near-surface. The refracted energy at far offsets was degraded heavily by noise. Unraveling the refracted energy initially masked by noise, we were predictably able to pick more traces in the far offset. Lastly, and using real data first breaks, we generated two tomograms, a conventional and its counterpart via our method. It was possible with our method to easily and reliably accommodate more far-offset picks and, hence, sample deeper parts of the near surface, resulting in a better lateral and vertical coverage than the conventional method.