previous up next print clean
Next: RESULTS Up: Karrenbach, Nichols & Muir: Previous: FINITE DIFFERENCE METHOD

AMPLITUDE EXTRACTION

Reflection amplitudes at the interface are retrieved by operating in the $\tau-p$ domain. Looking at the recorded time domain CMP records ( Figures [*] and [*]) we see an amplitude estimation problem near 1.75 sec caused by triplicating branches of reflection hyperbolae. Transforming those records to the $\tau-p$ domain gets rid of triplications and provides a robust estimate of the ``phase domain'' representation of the reflected waves in an anisotropic medium. Thus the traveltime curves in the $\tau-p$ domain are simple curves. From a practical point of view the slant-stack domain offers more advantages such as, dealiasing Nichols (1992), aperture and geometry compensated estimates of the wave field Kostov (1990), and regridding (inteerpolation) capabilities. The events of interest in Figures [*] and [*] (inline) are the uppermost two elliptical curves corresponding to the PP and PS reflection from the shale/chalk interface. The slant-stack difference sections, obtained by subtracting the crossline slant-stack section from the inline counterpart gives Figure [*] and [*]. The differences are obvious at values close to critical horizontal slowness, and at the triplicating event for a large array of horizontal slowness values.

The first step in amplitude extraction is to transform the CMP gathers into the $\tau$-p space using an ``optimum'' least-squares procedure, aiming to give us an unbiased, amplitude preserving estimate of the wave field. The traveltimes in the slant-stack domain are skewed ellipses in the anisotropic case. Since we are interested in PP and PS reflection, it is useful to write the Double Square Root (DSR) equation in the $\tau$-p domain. The DSR equation
\begin{displaymath}
\tau = {z_0\over{v_{down}}} \sqrt{(1 - p^2 v_{down}^2(p))}
 + {z_0\over{v_{up}}} \sqrt{(1 - p^2 v_{up}^2(p))}\end{displaymath} (1)
is a sum of two skewed square roots. Here the velocity depends on the ray parameter, but may not vary laterally. If vdown is not identical to vup, we are calculating the traveltime for a mode-converted wave. Amplitudes can then be extracted along a prescribed corridor given by the $\tau$-p traveltime curve in conjunction with the dominant wavelet period. From those extracted values the amplitude at the centroid frequency is picked. This gives us a robust reflection energy estimate for a given definite wave type.

 
pxseis.f10.0.cmp.SUB
pxseis.f10.0.cmp.SUB
Figure 5
Recorded seismograms from a p-source into an x-component.
view

 
pzseis.f10.0.cmp.SUB
pzseis.f10.0.cmp.SUB
Figure 6
Recorded seismogram from a p-source into an x-component.
view

 
pxseis.f10.0.cmp.SUB.SL
pxseis.f10.0.cmp.SUB.SL
Figure 7
The CMP gather (Fig. [*]) is transformed using a least-squares slant-stack.
view

 
pzseis.f10.0.cmp.SUB.SL
pzseis.f10.0.cmp.SUB.SL
Figure 8
The CMP gather (Fig. [*]) is transformed using a least-squares slant-stack.
view

 
pxseis.f10.SL.DIFF
pxseis.f10.SL.DIFF
Figure 9
Subtraction of inline and crossline slant-stacks.
view

 
pzseis.f10.SL.DIFF
pzseis.f10.SL.DIFF
Figure 10
Subtraction of inline and crossline slant-stacks.
view


previous up next print clean
Next: RESULTS Up: Karrenbach, Nichols & Muir: Previous: FINITE DIFFERENCE METHOD
Stanford Exploration Project
11/17/1997