True amplitude prestack depth migration

Reliable analysis of amplitude variation with offset (or with angle) requires accurate amplitudes from prestack migration. In routine seismic data processing, amplitude balancing and automatic gain control are often used to reduce amplitude lateral variations. However, these methods are empirical and lack a solid physical basis; thus, there are uncertainties that might produce erroneous conclusions, and hence cause economic loss.; During wavefield propagation, geometrical spreading, intrinsic attenuation, transmission losses and the energy conversion significantly distort the wavefield amplitude. Most current true-amplitude migrations usually compensate only for geometrical spreading. A new prestack depth migration based on the framework of reverse-time migration in the time-space domain was developed in this dissertation with the aim of compensating all of the propagation effects in one integrated algorithm. Geometrical spreading is automatically included because of the use of full two-way wave extrapolation. Viscoelastic wave equations are solved to handle the intrinsic attenuation with a priori quality factor. Transmission losses for both up- and down-going waves are compensated using a two-pass, recursive procedure based on extracting the angle-dependent reflection/transmission coefficients from prestack migration. The losses caused by the conversion of energy from one elastic model to another are accounted for through elastic wave extrapolation; the influence of the S wave velocity contrast on the P wave reflection coefficient is implicitly included by using the Zoeppritz equations to describe the reflection and transmission at an elastic interface. Only smooth background models are assumed to be known. The contrasts/ratios of the model parameters can be estimated by fitting the compensated angle-dependent reflection coefficients obtained from data for multiple sources. This is one useful by-product of the algorithm.; Numerical tests on both 2D and 3D scalar synthetic data, and on 2D elastic synthetic data show that the angle-dependent reflection coefficients along a target reflector, after all the compensations applied, are much closer to the theoretical values, and demonstrates the feasibility of the method. However, because of the recursive nature of the algorithm, the computational cost is high. The highly efficient parallel implementations make the task feasible; and this practical limitation will be removed by the fast development of the computer hardware and computational power in the near future.