Processing strategy for water-gun seismic data from the Gulf of Mexico

In order to study the regional distribution of gas hydrates
and their potential relationship to a large-scale sea-fl oor failures,
more than 1,300 km of near-vertical-incidence seismic
profi les were acquired using a 15-in3 water gun across the
upper- and middle-continental slope in the Garden Banks and
Green Canyon regions of the Gulf of Mexico. Because of the
highly mixed phase water-gun signature, caused mainly by a
precursor of the source arriving about 18 ms ahead of the
main pulse, a conventional processing scheme based on the
minimum phase assumption is not suitable for this data set. A
conventional processing scheme suppresses the reverberations
and compresses the main pulse, but the failure to suppress
precursors results in complex interference between the precursors
and primary refl ections, thus obscuring true refl ections.
To clearly image the subsurface without interference from
the precursors, a wavelet deconvolution based on the mixedphase
assumption using variable norm is attempted. This nonminimum-
phase wavelet deconvolution compresses a longwave-
train water-gun signature into a simple zero-phase wavelet.
A second-zero-crossing predictive deconvolution followed
by a wavelet deconvolution suppressed variable ghost arrivals
attributed to the variable depths of receivers. The processing
strategy of using wavelet deconvolution followed by a secondzero-
crossing deconvolution resulted in a sharp and simple
wavelet and a better defi nition of the polarity of refl ections.
Also, the application of dip moveout correction enhanced
lateral resolution of refl ections and substantially suppressed
coherent noise.