Euclidian Distance: Variogram Lag Measurement, Search Neighbourhoods, and Anisotropic Rescaling
Euclidian distance is the straight-line distance between two points in Euclidian space, where every dimension is orthogonal to the others, meaning all axes meet at right angles and represent independent directions in physical space. For two points it is computed by the Pythagorean rule: the square root of the sum of the squared differences of their coordinates. In two dimensions that is the square root of the difference in x squared plus the difference in y squared; in three dimensions a depth or z term is added under the same square root. The result is the shortest possible path between the points, the length a taut string would take, and it is the most intuitive and most widely used distance metric in the earth sciences. In petroleum geostatistics, Euclidian distance is the independent variable that drives almost every spatial calculation. When a variogram is built, sample pairs are sorted into lag bins according to the Euclidian distance separating them, and the average squared difference in the measured property within each bin is plotted against that distance to reveal how quickly the property loses correlation with separation. When kriging estimates an unsampled location, it ranks and weights nearby wells largely by their Euclidian distance to the target, and the search neighbourhood that decides which data are used is itself an ellipse or sphere defined in Euclidian terms. The metric also defines the lag vector h, whose length is a Euclidian distance and whose orientation lets the analyst test whether correlation differs by direction. That directional test exposes the main subtlety: real reservoirs are anisotropic, so raw Euclidian distance is often rescaled before use. A property may correlate over 2,000 m along a depositional channel trend but only 400 m across it, so geostatistical software applies an anisotropy transform that stretches or shrinks the coordinate axes before computing distance, converting the physical Euclidian distance into a standardised or reduced distance in which the correlation structure becomes isotropic. This is why two well pairs at the same physical separation can carry very different statistical weight: their orientation relative to the anisotropy axes changes their effective distance. Euclidian distance also has practical limits. It ignores barriers, so two points on opposite sides of a sealing fault are treated as close even though no fluid connects them, which is why some modern workflows replace it with non-Euclidian metrics that respect geological boundaries. It also assumes a flat coordinate frame, so structurally complex or steeply dipping settings are often unfolded into a stratigraphic coordinate system before distances are measured, restoring the layer-parallel geometry that existed at deposition. In Western Canadian Sedimentary Basin modelling, from Cardium shoreface sands to Montney siltstones, the correct use of Euclidian distance, rescaled for anisotropy and measured in stratigraphic rather than present-day structural coordinates, is what keeps a variogram physically meaningful and a kriged map geologically defensible.
Key Takeaways
- Pythagorean straight-line separation: Euclidian distance is the square root of the summed squared coordinate differences between two points, the shortest path through space along orthogonal, right-angled axes. In three dimensions it combines easting, northing, and depth under one square root. This simple, exact formula is the backbone of spatial statistics because it gives every pair of wells a single unambiguous separation value that every downstream geostatistical step can use.
- The independent variable of the variogram: Variograms plot dissimilarity against Euclidian distance. Sample pairs are binned by their separation, and the average squared property difference in each bin defines the experimental variogram. The nugget, range, and sill that describe spatial continuity are all expressed in terms of this distance, so the entire correlation model a geomodeller fits is, at its core, a function of Euclidian separation between data.
- Anisotropy demands rescaling: Reservoir properties rarely correlate equally in all directions, so raw Euclidian distance is transformed before use. The coordinate axes are stretched along the short correlation direction and compressed along the long one, producing a reduced distance in which anisotropic continuity becomes isotropic. Two pairs at identical physical separation thus receive different effective distances depending on their orientation relative to the channel or shoreface trend.
- It ignores geological barriers: Euclidian distance measures through rock indiscriminately, so it treats two points across a sealing fault as near neighbours even when no flow path connects them. Because this can let a kriging estimate borrow from a hydraulically isolated compartment, complex reservoirs increasingly use non-Euclidian or grid-based distances that route around barriers, or restrict the search neighbourhood so the simple metric stays valid only within a connected region.
- Stratigraphic coordinates restore depositional geometry: In folded or steeply dipping WCSB structures, present-day Euclidian distance crosses stratigraphy at an angle and mismeasures true geological proximity. Geomodellers unfold the structure into a layer-parallel stratigraphic frame, often a relative geologic time coordinate, then measure Euclidian distance there. This makes separations follow the bedding that controlled deposition, so the variogram reflects the original sediment continuity rather than later structural overprint.
Computing the Lag Vector in Three Dimensions
A geomodeller working a three-dimensional grid computes the lag vector h between two cells as the difference in easting, northing, and depth. Its Euclidian length, the square root of those three squared differences, places the pair into a distance bin, while its orientation, the azimuth and dip of the vector, places it into a directional class. A pair 800 m apart laterally and 6 m apart vertically has a Euclidian length near 800 m but a nearly horizontal orientation, so it informs the horizontal variogram. Separating pairs this way lets a single dataset yield independent horizontal and vertical ranges, capturing the strong layering anisotropy typical of WCSB clastics.
From Physical Distance to Reduced Distance
When anisotropy is present, software converts physical Euclidian distance into a reduced distance using the anisotropy ratio and the orientation of the principal axes. If a Cardium sand correlates over 2,000 m along depositional strike and 500 m across it, the cross-strike axis is multiplied by four so that 500 m across becomes equivalent to 2,000 m along. After this transform the variogram is isotropic in reduced space and a single model describes continuity in every direction. The kriging engine then weights data by reduced distance, correctly giving along-trend wells more influence than across-trend wells at the same physical separation.
Fast Facts
The metric is named for Euclid of Alexandria, whose Elements, written around 300 BC, formalised the geometry of right angles and straight lines more than two millennia before anyone drilled a well, yet his definition of distance survives unchanged inside every modern reservoir simulator. The deeper irony is that geostatisticians spend much of their effort defeating pure Euclidian distance, rescaling it for anisotropy and unfolding structure, because the subsurface rarely behaves as simply as Euclid's flat, isotropic plane.
Related Terms
Euclidian distance is the input to the variogram, which converts separation into a measure of spatial dissimilarity. It feeds kriging, where it determines both which data enter the search neighbourhood and how heavily each is weighted. Its directional use is inseparable from anisotropy, the directional dependence that forces the metric to be rescaled. And it presumes a stable statistical frame, linking it to stationarity, the decision that the correlation measured at a given distance applies across the modelled domain.
Real-World WCSB Scenario: Faulted Compartments in a Slave Point Pool
A team mapping permeability in a Slave Point carbonate pool near Red Earth used a standard Euclidian search radius of 1,500 m and produced a smooth, optimistic permeability field that implied strong connectivity across the whole pool. Two appraisal wells drilled into the apparently connected eastern flank came in dry of the expected pressure support, costing roughly CAD 6.8 million combined, because a sealing fault the model had measured straight through actually isolated that flank into a separate compartment.
The geomodel was rebuilt with a fault-aware distance that refused to connect cells across the sealing surface, plus a tighter search ellipse aligned to the reef trend. The revised map showed the eastern compartment as the isolated, undrained block it was, and the development plan added a dedicated injector to pressure-support it rather than relying on cross-fault flow, turning a mismeasured Euclidian shortcut into a corrected, compartment-aware design.