Probabilistic Seismic Inversion of Facies and Petrophysical Properties Using Gaussian Mixture and Markov Chain prior Models

Prediction of petrophysical properties and elastic attributes has become an important part of the exploration phase in the oil and gas industry to predict the presence of hydrocarbons subsurface. We focus on prediction of lithology/fluid classes, petrophysical properties and elastic attributes given geophysical observations. State of the art techniques are often based on minimization of the error, with respect to a given loss function, between a synthetic forward model and the observed data, either by probabilistic assessment or numerical optimization. We operate in a Bayesian framework where the objective is to assess the posterior probability density/mass function of the variables of interest subsurface.