A novel workflow for building multiple point statistics training images from virtual outcrops

Outcrop analogues of reservoirs are applied because facies-scale reservoir heterogeneities are frequently unresolvable at seismic-scale and well data provides sparse 1D geometrical data. Traditionally, geocellular models rely on manually measuring variograms or object dimensions from outcrops to define the geometry, size and directionality of facies proportions. Therefore, their ability to capture complex shapes and facies relationships in the subsurface is restricted by the quality of available geological data and the limitations of modelling algorithms. Multiple-point statistics (MPS) is a property modelling technique dependent on representative training images (TIs)- conceptual numerical descriptions of the geology expected in the reservoir under study. Lack of suitable TIs has limited the application of the MPS method to date. Recent advances in digital outcrop mapping methods, including lidar and photogrammetry, permit the rapid acquisition of high-resolution 3D virtual outcrop models. These provide a critical and underused source of qualitative and quantitative information for high quality TI generation. We present a novel approach to apply 3D virtual outcrops as TIs; coupled with the streamlining of lidar integration into subsurface models using examples from the Bolea area, Ebro Basin, northern Spain. This approach will significantly improve prediction of 3D facies heterogeneity and its impact on reservoir performance.