AVO-sensitive semblance analysis for wide-azimuth data

Main Content:

Conventional semblance-based moveout analysis models prestack reflection data
with events that have hyperbolic moveout and no amplitude variation with offset
(AVO). It has been shown that substantial amplitude variation and even phase
change with offset do not significantly compromise the semblance operator.
However, polarity reversal associated with a change in the sign of the reflection
coefficient may cause conventional semblance to fail. An existing modification
of the semblance operator that takes amplitude variations into account (so-
called “AK semblance”) is limited to narrow-azimuth data and cannot handle
nonhyperbolic moveout.
Here, we extend the AK semblance algorithm to long-spread (nonhyperbolic)
moveout and 3D multiazimuth data. To preserve velocity resolution in the pres-
ence of substantial AVO signature, we keep the ratio K of the AVO gradient and
intercept constant within each semblance window. In the presence of azimuthal
anisotropy, however, the parameter K has to be azimuthally dependent. In our
implementation, the modified semblance operator is combined with a nonhy-
perbolic moveout inversion algorithm devised for wide-azimuth data.
Synthetic tests confirm that the distortions in moveout analysis caused by po-
larity reversals are more common for long-spread data. Conventional semblance
produces substantial errors in both the NMO ellipse and azimuthally varying
parameter  not just for type 2 AVO response, but also for some models with
type 1 AVO. In contrast, the AK semblance algorithm gives accurate estimates
of the moveout parameters even when the position of the polarity reversal varies
with azimuth. The AK method not only helps to flatten wide-azimuth reflection
events prior to stacking and azimuthal AVO analysis, but also provides input
parameters for the anisotropic geometrical-spreading correction.