Variogram

A variogram (or semivariogram) in petroleum geostatistics is a mathematical function that quantifies the spatial variability of a reservoir property — such as porosity, permeability, net-to-gross ratio, or facies indicator — as a function of the distance and direction between measurement points, by computing half the average squared difference between property values at pairs of sample locations separated by increasing lag distances, producing a curve that starts near zero at short separations (nearby samples are similar) and increases with distance until it plateaus at the sill value (equal to the total variance) at the range distance (beyond which samples are no longer spatially correlated) — the fundamental spatial statistical tool used in kriging interpolation, stochastic reservoir modeling, and uncertainty quantification in petroleum reservoir characterization.

What Is a Variogram?

Between the wells in a reservoir, the distribution of porosity, permeability, and facies is unknown — it must be inferred from the well measurements and constrained by geological understanding of the depositional system. The question "how similar are reservoir properties at two locations separated by some distance?" is fundamental to this inference problem, and the variogram provides the quantitative answer.

The variogram measures how the difference in property values between pairs of sample points increases with increasing separation distance. At short distances, nearby samples tend to have similar values (small differences, low variogram value). At large distances, sample values may be essentially uncorrelated (large differences, high variogram value approaching the total variance of the dataset). The variogram captures this transition, showing at what distance spatial correlation breaks down — the range — and what the total variability is — the sill.

Once the variogram has been estimated from well data and a mathematical model has been fitted to it, the spatial correlation structure is defined in a form that kriging and stochastic simulation algorithms can use to interpolate property values between wells and to generate probabilistic realizations of the full reservoir property distribution. The variogram is therefore the mathematical bridge between the point measurements available at well locations and the volumetric property distribution needed for material balance calculations, simulation model construction, and reserve estimation.

Variogram Applications in Reservoir Modeling

Porosity interpolation for geological model construction uses kriging with a porosity variogram to estimate porosity values on the 3D reservoir grid cells between wells — a spherical variogram model with a range of 2 kilometers in the north-south direction and 500 meters in the east-west direction (reflecting the elongated depositional geometry of a north-south trending fluvial channel system) would assign smoothly varying porosity estimates that follow the channel trend, with the kriging uncertainty increasing systematically with distance from the nearest well and in directions perpendicular to the maximum correlation range.

Uncertainty quantification in reserve estimates uses stochastic simulation with the variogram to generate an ensemble of equally probable porosity and permeability realizations that span the range of reservoir architectures consistent with the well data — the P10 (pessimistic) realization might have low porosity between wells reflecting high variogram uncertainty, while the P90 (optimistic) realization might have higher inter-well porosity where the variogram uncertainty permits; the hydrocarbon volumes computed from each realization provide the probabilistic range from which the P10, P50, and P90 oil or gas in place is estimated. The width of this probabilistic range depends critically on the variogram range — a long range (high well-to-well correlation) gives a narrow probabilistic range (less uncertainty between wells) while a short range (low correlation) gives a wide range (high uncertainty), making the variogram directly responsible for the reserve uncertainty band presented to management and regulatory agencies.

Seismic-constrained variogram modeling uses seismic attribute maps (acoustic impedance, amplitude) as secondary information to improve the variogram estimation between widely spaced wells — in a field with wells spaced 5 to 10 kilometers apart, the experimental variogram computed from well data alone is unreliable at lag distances less than the well spacing because there are few or no data pairs at those short distances; by computing the variogram of the seismic attribute (which provides spatial information at sub-well-spacing scales) and using it to constrain the reservoir property variogram (assuming the seismic attribute and reservoir porosity are correlated), a better-constrained variogram model is obtained that correctly captures the short-range spatial correlation structure that is invisible in widely spaced well data.

Variogram Across International Jurisdictions

Canada (AER / WCSB): WCSB Montney and Cardium reservoir geological models use variogram analysis to quantify the spatial correlation length of porosity and permeability within the formation, with directional variograms computed in the along-depositional-strike and cross-strike directions that reflect the depositional geometry of the Montney turbidite or Cardium fluvial/shoreface systems. AER reserve assessment submissions for WCSB tight oil and gas plays using probabilistic reserve estimates must document the geological model methodology including geostatistical simulation parameters — the variogram model specification is part of the required technical disclosure that allows AER evaluators to assess whether the uncertainty range in the probabilistic estimate is geologically reasonable for the specific play type and well spacing. Canadian reservoir modeling software (CMG's RESULTS, Roxar RMS) implements variogram-based kriging and simulation as the standard geological interpolation approach.

United States (API / BSEE): Gulf of Mexico deepwater turbidite reservoir modeling relies heavily on variogram-based stochastic simulation to characterize the heterogeneity of turbidite channel fill and levee deposits between widely spaced wells (often 2 to 5 kilometers apart in frontier exploration settings), with seismic amplitude attributes providing the secondary spatial information needed to constrain variograms at scales below the well spacing. SEC proved reserve regulations require that probabilistic reserve estimates be technically defensible, and the variogram model specification is a technically auditable component of the geostatistical methodology underlying probabilistic OOIP and GIIP estimates reported to the SEC. Permian Basin unconventional reservoir modeling uses variogram analysis across dense well datasets (wells spaced 200 to 400 meters apart in mature development programs) to quantify small-scale petrophysical heterogeneity and optimize completion designs for specific reservoir quality zones.

Norway (Sodir / NORSOK): NCS PDO (Plan for Development and Operation) reserve estimates must be supported by geological models built using quantitatively defensible methods, and variogram-based kriging and stochastic simulation have been the standard NCS geological modeling approaches since the mid-1990s; Sodir's technical review of PDO submissions includes assessment of the geological model methodology, and variogram analysis documentation is part of the expected technical content. Equinor's reservoir modeling group has contributed significantly to variogram-based geostatistical methodology through publications in the Journal of Petroleum Science and Engineering and SPE conferences, documenting improved variogram estimation methods for NCS reservoirs with complex structural and stratigraphic architectures.