CONCLUSIONS

We have developed a Kirchhoff prestack depth migration method which is consistent with elastic wave propagation through a background model in which one can specify values of P-wave velocity (V_p), S-wave velocity (V_s), density ($\rho$), and the anelastic attenuation factor (Q). The method correctly estimates the $\grave{P}\!\acute{P}$ Zoeppritz elastic reflectivity coefficient R_pp, and the associated specular reflection angles $\theta_{pp}$, in synthetic studies based on an actual field data application with drilling information. This method is a true amplitude depth migration in that the R_pp and $\theta_{pp}$ estimates are valid for laterally heterogeneous velocity models. The algorithmic implementation is computationally efficient in that it takes 6.0 cpu hours on a Sun Sparcstation 1⁺ to migrate 140 marine shot gathers (16,800 traces) into a 2x2x3 km³ target depth image volume (1 million pixels) in a 1-D background model, each of R_pp and $\theta_{pp}$ as a function of depth, horizontal distance and source-receiver offset. However, this time can increase significantly for complex 2-D migration velocity models which require intense raytracing calculations.

A method has also been developed to invert $R_{pp}(\theta)$ gathers in order to produce depth images of subsurface elastic parameter variations. Parameterization choice issues have been investigated, and it has been found that the Elastic Impedance parameter set (I_p, I_s, $\rho$) is the most robust set for our purposes. The other less favorable parameterizations investigated were Velocities, Elastic Moduli, Lamé Parameters, V_p/V_s Ratio, and the industry-standard ``AVO'' Parameterization (A, B, C). In all parameterizations, local relative changes in elastic properties are recovered, and not the elastic property values themselves. A quantitative measure of ``confidence'' has been developed to appraise the quality of the inversion images. Both the elastic parameter inversion and confidence criterion methods have been successfully tested on synthetic data and validated through the field data application. The elastic parameter inversion takes about 10 cpu minutes on a Sun Sparcstation 1⁺ for the field data application discussed in this report.

Further details concerning the migration/inversion method, the confidence estimates, and the field data results are in preparation for journal publication, and are likely to appear in subsequent SEP reports (pending proprietary data release).