Fitted Variogram: Nugget, Sill, and Range Modeling for WCSB Reservoir Porosity and Permeability Mapping

A fitted variogram is an experimental variogram or semivariogram to which a valid mathematical model has been applied, so that the discrete, noisy cloud of spatial-variance points calculated from well data is replaced by a smooth, continuous function that geostatistical algorithms can evaluate at any lag distance. The experimental variogram itself is built by taking every pair of sample points, computing half the squared difference of a property such as porosity or permeability between them, and averaging those semivariance values within distance bins called lags. Plotting average semivariance against lag distance produces the experimental variogram, a series of dots that generally rise from a low value near zero lag toward a plateau at larger separations. Because those raw dots are irregular and undefined between bins, they cannot be used directly by kriging or Gaussian simulation, so an analyst fits a permissible model, most commonly a spherical, exponential, or Gaussian function, that honours three defining parameters. The nugget is the model's y-intercept, the apparent semivariance at zero lag, and it captures measurement error plus real spatial variability occurring at scales finer than the closest sample spacing. The sill is the plateau value the model reaches, representing the total variance where samples become spatially independent, and it is often close to the statistical variance of the data set. The range is the lag distance at which the model first flattens to the sill, and it is the most physically meaningful output because point pairs closer than the range are spatially correlated while pairs farther apart are effectively uncorrelated. Fitting is judged by how well the model tracks the experimental points, weighted toward the well-populated short-lag bins that matter most for interpolation, and it is usually done in multiple directions to capture anisotropy, where the range along depositional strike differs from the range across it. In WCSB reservoir characterization the fitted variogram is the engine of three-dimensional geostatistical modeling used to assess reservoir heterogeneity and connectivity, feeding directly into kriging for smooth best-estimate maps and into sequential Gaussian simulation for multiple equiprobable realizations that quantify uncertainty. A Cardium tight-oil team fitting variograms to core and log porosity will typically find a long range parallel to the northwest-southeast shoreface trend and a much shorter range perpendicular to it, and that anisotropic fitted model then controls how porosity and permeability are propagated between wells across the pool. The quality of the fit propagates straight into volumetric estimates, well-spacing decisions, and waterflood or Enhanced Oil Recovery pattern design, because an over-long fitted range makes a reservoir look more continuous and better connected than it is, while an inflated nugget throws away real short-scale structure and produces overly smooth, pessimistic connectivity. The fitted variogram therefore sits at the heart of geostatistics and links spatial statistics to the reservoir characterization workflows that underpin WCSB development economics.

Fitting Anisotropic Variograms to Cardium Porosity

For a Cardium pool near Pembina, an analyst computes directional experimental variograms of log porosity along and across the northwest-southeast shoreface trend. The major-direction variogram levels off at a range near 900 m while the minor direction reaches its sill by about 300 m, a 3:1 anisotropy ratio reflecting the elongate sand-body geometry. A spherical model is fitted with a nugget of roughly 15 percent of the sill to account for core-to-log scale mismatch. That anisotropic fitted model is then imported into the geomodel so sequential Gaussian simulation stretches correlation along strike, honouring the real depositional fabric rather than an isotropic assumption that would smear porosity equally in all directions.

Choosing Spherical, Exponential, or Gaussian Models

The model shape encodes how continuity decays with distance. A spherical model reaches its sill at a finite range and suits many WCSB clastic reservoirs with clear correlation cutoffs. An exponential model approaches the sill asymptotically and fits properties with gradual transitions such as some carbonate permeability fields. A Gaussian model, with its parabolic behaviour near the origin, represents very smooth, continuous phenomena but can produce unstable kriging if used without a small nugget. Selecting the wrong shape for a Duvernay or Montney property misrepresents short-lag behaviour, which is precisely the region that controls interpolation accuracy between closely spaced horizontal well penetrations.

Related Terms

A fitted variogram is the required input to Kriging, which uses its nugget, sill, and range to compute optimal spatial weights for best-estimate maps. It is the structural model underpinning Geostatistics as a whole, the discipline that treats reservoir properties as spatially correlated random fields. It drives Reservoir Characterization by translating scattered well data into continuous property volumes, and its range and anisotropy connect it to Heterogeneity, the very variability the variogram is built to quantify.

Real-World WCSB Scenario: An Overfitted Range on a Viking Waterflood Near Provost

A Viking light-oil operator near Provost fitted a porosity-permeability variogram using only 11 vertical wells spread across a large lease. With so few pairs at short lags, the experimental variogram was poorly constrained, and the analyst fitted a spherical model with a 1,400 m range that made the thin Viking sand look highly continuous. The geomodel predicted strong interwell connectivity, and a waterflood was designed on 16-hectare spacing. When injection began, tracer surveys showed water breaking through far faster in some directions and not at all in others, revealing the true correlation range was closer to 500 m with strong channel anisotropy the sparse fit had missed.

The operator re-fitted the variogram after drilling six infill wells that populated the short-lag bins, corrected the range and added directional anisotropy, and redesigned the injection pattern. The revised flood improved voidage balance and recovered an estimated additional 4 percent of original oil in place, worth several million CAD over the pool life, a return that hinged entirely on getting the fitted variogram right.