Current earth modeling practice uses 3D-geocellular grids to represent reservoir volumes. The grid parameters must be entered upfront, and often little thought is given to the appropriateness of the specified cellular dimensions and layering styles. The resulting geocellular topology can be either too fine or too coarse, which poorly represents the subsurface. Consequently, important aspects of the sealed structural framework, such as folds, complex faults, and geometries of key geological features, are not accurately depicted. Geocellular parameterization can become either a trial-and-error process or a compromise, which is time consuming and expensive.
We present a new technology that resolves many common geocellular parameterization and modeling issues by first generating grid-less 3D property models within a sealed structural framework. The geological properties within a volume of subsurface are represented by distributing a plurality of data points in the absence of the grid with the notion of geological continuity and directionality given by a maximum continuity field (MCF).
In many depositional systems, changes exist in the local direction of maximum continuity; however, traditional variogram implementation assumes a single maximum direction of continuity. Property interpolation by means of a variogram only requires the ?correct? distance and local azimuth to compute the covariance matrix. Our method supports defining both a local direction of maximum continuity and the associated correlation distance without a standard grid, offering the user:
? Direct control of local continuity directions and their correlation distances;
? Interaction with ?geologically intuitive? datasets through the notion of MCF;
? Retention of the maximum fidelity of the geological model by designing the optimal geocellular framework as the final modeling step.
The technology was validated and benchmarked by modeling the permeability distribution in a synthetic, complex fluvial system, and combining a set of user-defined MCFs and control data points.