Capturing Depositional Processes Using MPS Simulation with Multiple Training Images

Prasenjit Roy¹ and Sebastien B. Strebelle^2
1Earth Sciences R&D, Chevron ETC, Houston, TX
²Earth Sciences R&D, Chevron ETC, San Ramon, CA

Multiple-Point Statistics (MPS) simulation has emerged recently as a practical geostatistical modeling technique to simulate complex depositional facies patterns, such as sinuous channels, that cannot be modeled using conventional (two-point statistics) variogram-based techniques. MPS simulation consists of first extracting patterns from a 3D numerical training image describing the type of geological elements expected in the reservoir under study, and then reproducing similar patterns conditional to well and seismic data in the simulation grid.

However, hydrocarbon fields are very often characterized by multiple depositional processes resulting in the juxtaposition of various types of facies architectures, for example slope valley channels becoming unconfined lobes when reaching the basin floor in deepwater turbidite environments. To capture that spatial variety, a novel extension of the MPS simulation approach is proposed to enable the use of multiple training images representative of different depositional processes. This paper also demonstrates how additional modeling constraints such as facies proportion maps and curves, as well as variable azimuth fields, help build MPS facies models providing a more accurate representation of the underlying geological heterogeneity.

AAPG Search and Discovery Article #90078©2008 AAPG Annual Convention, San Antonio, Texas