Constraining 3D Sedimentological Models to Well and Seismic Using Pattern-based Multiple-point Geostatistics

At the late exploration to early appraisal stage, only few wells are often available for making reservoir predictions. Any reservoir modelling should therefore include as much as possible the sedimentological understanding. We propose a new method for realistic geological modelling of complex heterogeneity by hybridizing 3D process modelling of geological deposition with conditioning to wells and seismic data by means of multiple-point geostatistics (MPS).Several process-based sedimentological software have been developed allowing geologists to produce realistic models of the subsurface. However they cannot be properly constrained to observations from well and seismic data. To overcome this issue we applied a new MPS method termed CCSIM (cross-correlation simulation of patterns) that allows the use of process models as training images. In CCSIM 3D geostatistical models are constructed by patching together patterns extracted from training images, allowing for conditioning to well data and properly modelling the non-stationarity. An application to an offshore African reservoir is here presented. Starting from seismic advanced attribute analysis, facies and sedimentological interpretation the 3D sedimentological process based model was built and then used as non-stationary training image for CCSIM simulation.