INTEGRATING SEISMIC DATA FOR LITHOFACIES MODELING: A COMPARISON OF SEQUENTIAL INDICATOR SIMULATION ALGORITHMS

Abstract

Clastic reservoir characterization starts typically with modeling lithofacies distribution and geometry. The architecture of the reservoir, governed by the lithofacies geometry, is a major source of heterogeneity in such clastic systems. Seismic data provide potentially valuable information about the areal distribution of different lithofacies, such as the averaged prior proportion of each lithofacies. However, seismic data are available only at coarse vertical resolution rather than the fine lithofacies sampling along wells, hence seismic is considered equivalent to 2D data while building 3D geological models. This scale difference between the seismic data and the lithofacies data available along the wells makes direct integration difficult. Different algorithms have been proposed to integrate the seismic data: (1) duplicate seismic data along the vertical line and use the prior proportions provided by the seismic data as prior local means; (2) integrate the 2D seismic data as collocated block averages; and (3) duplicate seismic data along the vertical line and integrate them using a Markov-Bayes algorithm. These three algorithms are applied on a data set originating from a real clastic reservoir. The results are compared with regard to how much kriging weight is applied to the seismic data and how well the information from seismic data is honored.