Geostatistical Seismic AVO Inversion Directly for Facies - Real Case Application

Abstract

We present a new iterative geostatistical seismic inversion algorithm that allows retrieving: density; P-wave velocity; S-wave velocity and facies models. This novel procedure is based on two key main ideas: stochastic sequential simulation and co-simulation as the perturbation technique of the model parameter space; and a genetic algorithm that act as a global optimizer to converge the iterative procedure towards an objective function, the mismatch between recorded and synthetic pre-stack seismic data. At the end of each iteration, the triplet of elastic traces that jointly produce synthetic gathers with the highest correlation coefficient are the basis for generating the new elastic models of the next iteration. The iterative procedure finishes when the global correlation between recorded and inverted seismic data is above a certain threshold. All the elastic models simulated during the iterative procedure honor the marginal and joint distributions of P-wave velocity, S-wave velocity and density as estimated from the available well-log data. We successfully tested this new algorithm on a real pre-stack seismic dataset acquired over a deep-water turbidite oil reservoir. The results show a good convergence between real and synthetic seismic and the retrieved high resolution petro-elastic models agree with previous studies performed with commercial inversion algorithms.