MerQuR+Statistics Glossary

Statistics Glossary → Kriging (simple, ordinary, universal, block)

Kriging (simple, ordinary, universal, block)concept

A spatial interpolation method that predicts the value at unsampled locations as a weighted average of nearby measurements, with weights based on the spatial relationship described by the variogram, and that also gives the uncertainty of the prediction.

Kriging weights are computed from the variogram model so as to minimize the variance of the prediction error while keeping the prediction unbiased, which is why kriging is called the best linear unbiased predictor (BLUP). Unlike inverse distance weighting, the weights depend not only on distance but on the actual spatial correlation structure of the data and on how the sample points are clustered relative to each other. A kriging variance is obtained for every prediction point, so an uncertainty map can be drawn.

Simple kriging assumes the mean is known; ordinary kriging estimates an unknown mean that is constant over the area and is the most common type; universal kriging lets the mean follow a trend in the coordinates or in a covariate. Block kriging predicts the average over an area rather than at a point and gives a smaller prediction variance. The quality of the results depends heavily on the variogram model being right and should be checked by (leave-one-out) cross-validation. KDE and hexbin maps, by contrast, show the density of points rather than a value, which sets them apart from kriging.

An everyday example

Soil samples are taken at 40 points in a field and nitrogen is measured. Kriging predicts the nitrogen level at every unsampled point from nearby samples, learning from the variogram how much similarity closeness really brings. In corners far from any sample, the uncertainty of the prediction shows up on the map as clearly larger.

In MerQur

The 'Kriging type' (Ordinary, Simple, Universal), 'SK mean, UK trend and Covariate', 'Block size' and 'Number of neighbours (nmax)' options of Variogram + Kriging.

Analyses where this concept appears

KDE Density MapHexbin MapVariogram + Kriging (Spatial Interpolation)

Related concepts

Variogram (semivariogram)Inverse distance weighting (IDW)Nugget, sill and range