Nonlinear AVO inversion using geostatistical a priori information

We present a Monte Carlo based strategy for non-linear inversion of seismic amplitude versus offset data. The problem is formulated in a Bayesian framework such that the solution to inverse problem is an a posteriori probability density. A priori information about the problem is defined as a Gaussian probability density. The problem is conditioned by observations of reflected P-waveforms. A nonlinear relation between the model and data based on ray tracing and Zoeppritz equations is considered. As a consequence of these nonlinearities, no closed form expression of the a posteriori probability density can be formulated. Therefore, we apply the Metropolis algorithm in order to obtain an exhaustive characterisation of the a posteriori probability density.