Statistics Glossary → Anisotropy
Anisotropyconcept
Direction dependence in how quickly a spatial variable loses similarity with distance; for example, values remaining related over long distances in one direction and only over short distances in another.
In an isotropic process the relationship between two points depends only on the distance between them; under anisotropy it also depends on direction. It is examined with directional variograms: the experimental variogram is computed separately for directions such as 0°, 45°, 90° and 135° within an angular tolerance (e.g. ±22.5°). In geometric anisotropy the sill is the same in every direction but the range changes with direction, which is described by an ellipse; the anisotropy ratio is the shortest range divided by the longest. In zonal anisotropy the sill itself also differs with direction.
Ignoring anisotropy makes kriging underuse information along the direction in which the relationship persists and overuse it along the direction in which it fades quickly, distorting both predictions and prediction errors. Directional variograms rest on fewer point pairs in each direction and are therefore noisier, so with little data the evidence for anisotropy may be weak. An unremoved linear trend can also look like anisotropy, so trends should be examined first.
An everyday example
Soil moisture is measured along a valley. Points 800 m apart along the valley still show similar moisture, while 200 m across the valley moisture changes markedly, because conditions shift quickly from the slopes to the valley floor. The anisotropy ratio is 200/800 = 0.25, and kriging should take this direction dependence into account.
In MerQur
In Variogram + Kriging, the 'Directional variogram (0°, 45°, 90°, 135°; anisotropy)' checkbox computes variograms and a range for each of four directions (±22.5° tolerance, clockwise from north) and reports the anisotropy ratio.
Analyses where this concept appears
Variogram + Kriging (Spatial Interpolation)