Gridding

Key Takeaways

• Gridding algorithms for geoscience mapping (creating 2D surfaces from point data) include minimum curvature (which produces smooth surfaces honoring all data points with the minimum bending energy), kriging (a geostatistical method that uses the semivariogram of the data to estimate both the interpolated value and the uncertainty at each grid node), inverse distance weighting (which weights nearby data more heavily than distant data in a simple distance-power relationship), and triangulation with linear interpolation (which creates a triangulated irregular network directly from the data points without further smoothing); kriging is generally preferred for reservoir property mapping because it provides uncertainty estimates alongside the property values, enabling probabilistic map analysis, while minimum curvature is preferred for structural mapping of seismic horizons where data density is high and smooth continuous surfaces are geologically expected; the choice among algorithms has practical consequences for reserve estimates and well placement — a kriging map may show a reservoir thinning trend that a minimum curvature map would smooth over, potentially affecting the P10 reserve estimate by 20% or more.
• Reservoir simulation gridding adds a third dimension (depth or time) to the mapping problem and introduces a fundamentally different set of requirements — the simulation grid must honor geological layering (by conforming to stratigraphic surfaces rather than cutting across them), capture faults (as cell boundaries or transmissibility multipliers), be fine enough to resolve the fluid flow heterogeneity that controls recovery (yet not so fine that the simulation becomes computationally intractable), and maintain numerical stability (cells that are too elongated or too thin can cause the finite difference equations to diverge); structured grids (i, j, k Cartesian or corner-point geometry) are most common in commercial reservoir simulation software due to their computational efficiency, while unstructured grids (Voronoi/PEBI cells, tetrahedral elements) offer better geometric flexibility around wells and complex faults at the cost of increased computational complexity; corner-point geometry (where each grid cell is defined by eight corner points that can be displaced vertically to follow geological horizons) is the dominant standard in commercial simulators including ECLIPSE, CMG STARS, and INTERSECT.
• Upscaling — the process of converting a fine-scale geological model (which may contain tens of millions of cells) into a coarser simulation grid (typically 100,000 to 2,000,000 cells) that can be run in reasonable computation time — is one of the most technically challenging aspects of reservoir modeling, because properties like permeability do not upscale linearly and the upscaling method can significantly affect the predicted production behavior; simple arithmetic averaging of permeability in the upscaling direction overestimates effective permeability for flow perpendicular to high-permeability streaks, while harmonic averaging underestimates it for flow parallel to them; power-law averaging with an exponent between -1 (harmonic) and +1 (arithmetic) captures anisotropic permeability upscaling approximately, while numerical upscaling (solving the flow equations on the fine-scale cell to determine the effective transmissibility of the coarse cell) is the most accurate method for complex heterogeneous intervals but is computationally expensive to apply across millions of cells.
• Well-to-well correlation and geocellular model building require gridding decisions that directly control how geological heterogeneity is represented in the reservoir model — the grid layering scheme (proportional, onlap, offlap, or truncation layering that conforms to different geological boundary conditions for stratigraphic sequences) determines whether the model correctly represents the internal architecture of the reservoir; an unconformity-bounded sequence where reservoir sands thin toward a structural high requires a truncation or onlap layering scheme that allows the upper-bounding horizon to cut across the lower layers, while a progradational sequence where sands thicken basinward requires an offlap scheme; using the wrong layering type creates a geological model that misrepresents the actual stratigraphy, which in turn causes the simulation to predict fluid flow paths and sweep efficiency incorrectly, leading to wrong production forecasts and suboptimal injection well placement.
• Local grid refinement (LGR) is a simulation gridding technique that adds fine-scale resolution in specific areas of the reservoir model — typically around well perforations, hydraulic fractures, or high-flow-rate injector-producer pairs — while maintaining a coarser grid in the bulk of the reservoir where fine-scale resolution is not needed for accuracy; LGR allows the near-wellbore coning behavior of water or gas (which occurs at the scale of a few meters around the well), the transient pressure behavior during well testing (which involves radial flow at small scales), and the complex fluid dynamics around hydraulic fractures (where the fracture aperture may be millimeters while the fracture length is hundreds of meters) to be modeled with appropriate resolution without simulating the entire reservoir at that fine scale; commercial simulators implement LGR as nested grids, with the refined region inheriting boundary conditions from the surrounding coarse grid through a transmissibility coupling method that maintains mass conservation across the grid refinement boundary.