In anisotropic media, when the reflector is dipping with respect to the normal to the isotropic axis of symmetry (horizontal direction for VTI) the incident and reflected aperture angles differ. This difference is caused by the fact that, although the phase slowness is function of the propagation angle, Snell law requires that the components parallel to the reflector of the incident and reflected slowness vectors must match at the interface. However, we can still define an ``average'' aperture angle and ``average'' dip angle using the following relationships:
(8) |
Figure shows the geometric interpretation of these angles. Notice that the average dip angle is different from the true geological dip angle ,and that the average aperture angle is obviously different from the true aperture angles and .However, these five angles are related and, if needed, the true angles can be derived from the average angles, as shown in Appendix A.
Prestack images defined in the subsurface-offset domain are transformed into the angle domain by applying slant stacks. The transformation axis is thus the physical dip of the image along the subsurface offset; that is, .The dip angles can be similarly related to the midpoint dips in the image; that is, .Following the derivation of acoustic isotropic ADCIGs by Sava and Fomel (2003) and of converted-waves ADCIGs by Rosales and Rickett (2001), we can write the following relationships between the propagation angles and the derivative measured from the wavefield:
(9) | ||
(10) | ||
(11) |
(12) | ||
(13) |
Solving for and we obtain the following:
(14) | ||
(15) |
Substituting equation 15 in equation 14, and equation 14 into equation 15, we get the following two quadratic expressions that can be solved to estimate the angles as a function of the dips measured from the image:
(16) | ||
(17) |
For anisotropic velocities, the slownesses depend on the propagation angles, and thus the normalized difference depends on the unknown and .In practice, these equations can be solved by a simple iterative process that starts by assuming the ``normalized difference'' to be equal to zero. In all numerical test I conducted this iterative process converges to the correct solution in only a few iterations, and thus is not computationally demanding.
If the anisotropic slowness function were spatially homogeneous, equations 16 and 17 could be solved iteratively in the Fourier domain, and the transformation to the average angles and could be computed exactly without the need of estimating the apparent reflector dip in the space domain. When the anisotropic parameters are a function of the spatial variables; that is, in the majority of the real situations, the solution of equations 16 and 17 requires the estimation of the local reflector dip in the space domain. If necessary, the reflectors' dip can be either extracted from the interpretation of the horizons of interest, or can be automatically estimated from the image by applying one of the several methods that have been presented in the literature (see for example Fomel (2002)). In practice, the estimation of the reflector dip is seldom necessary. The numerical and real-data examples shown below indicate that for practical values of the anisotropy parameters the dependency of the estimate from the dip angles can be safely ignored for small dips, and it is unlikely to constitute a problem for steep dips.