Seismic Inversion for Lithology and Reservoir Properties Prediction of Lower-Maykopian Sediments within Subbotina Field

A Bayesian approach to stratigraphic inversion uses prior model constraints such as smoothness as a mean of regularization. Such an approach requires building a model covariance function in a manner that is at least computationally tractable, if not efficient. A general solution involves smoothing, for which we explore the use of an anisotropic diffusion operator. Upon smoothing a 1D solution (which ignores trace-to-trace correlation) with an anisotropic diffusion filter, results were obtained that were very similar to the much more computationally intensive coupled (3D) inversion. We justify this result and demonstrate it on a real seismic dataset.