Comparison of tau-p domain wave-equation inversion with convolutional inversion.

A fast and flexible method for performing wave-equation inversion in the tau-p domain has been developed. The computational kernel is based on the Direct Global Matrix (DGM) method for solving the 1D tau-p domain waveequation for the wave field and based on the adjoint state method for computing the analytic gradient of the wave field. There already exist several pre-processing methods for suppressing surface related long period multiples, hence applying the proposed inversion method is primarily seen as a way to handle short period interbed multiples and converted waves in the context of AVO inversion for which almost no good methods exist presently. To examine the validity of the proposed inversion method a series of tests on the same synthetic data set has been performed. The synthetic tests demonstrate that for reliably suppressing short period multiples and converted waves using a wave-equation method in the context of AVO inversion, it is required that the AVO information is utilized simultaneously for all p (slope) values (1-step inversion) rather than the AVO information is utilized after the wave-equation (2-step inversion).

Link to abstract