A Volume Flattening Methodology for Geostatistical Properties Estimation

In the domain of oil exploration, geostatistical methods aim at simulating petrophysical properties in a 3D grid model of reservoir. Generally, only a small amount of cells are populated with properties. Roughly speaking, the question is: which properties to give to cell c, knowing the properties of n cells at a given distance from c? Obviously, the population of the whole reservoir must be computed while respecting the spatial correlation distances of properties. Thus, computing of these correlation distances is a key feature of the geostatistical simulations. In the classical geostatistical simulation workflow, the evaluation of the correlation distance is imprecise. Indeed, they are computed in a Cartesian simulation space which is not representative of the geometry of the reservoir. This induces major deformations in the final generated petrophysical properties. We propose a new methodology based on isometric flattening of sub-surface models. Thanks to the flattening, we accurately reposition the initial populated cells in the simulation space, before computing the correlation distances. In this paper, we introduce our different flattening algorithms depending on the deposit mode of the sub-surface model and present some results.