GeoModels-3D: Three-dimensional spatial coordinates in GeoModels

In GeoModels: Procedures for Gaussian and Non Gaussian Geostatistical (Large) Data Analysis

Three-dimensional spatial coordinates in GeoModels

Description

Conventions and current limitations for purely spatial models whose locations have three Cartesian coordinates.

Details

For an irregular three-dimensional spatial data set, supply either an N \times 3 numeric matrix in coordx, or three equal-length vectors through coordx, coordy, and coordz. Use grid = FALSE and distance = "Eucl".

Euclidean lags use all three coordinates. The principal fitting, covariance, variogram, Cholesky simulation, neighborhood, and global/local kriging paths support this representation.

Regular three-dimensional grids specified by three coordinate axes and grid = TRUE are not implemented. Supply their Cartesian product as an explicit N \times 3 matrix and use grid = FALSE.

For purely spatial univariate fields, spectral turning-bands simulation accepts explicit irregular three-dimensional Euclidean coordinates through GeoSimapprox(..., method = "TB", grid = FALSE). The same TB simulator can be used by GeoVarest() and GeoVarestbootstrap() for such fits. Three-dimensional TB does not support bivariate or spatio-temporal models, and does not accept anisopars; transform coordinates explicitly when needed. CE remains restricted to regular two-dimensional grids. Use method = "cholesky" for unsupported three-dimensional cases.

Distances "Geod" and "Chor" require exactly two columns (longitude and latitude) and cannot be combined with a third coordinate.

For GeoVariogramDir(), three-dimensional lag lengths use all three coordinates, while direction classes are horizontal azimuths computed from the x- and y-coordinate differences. This is not a full three-dimensional directional analysis with azimuth and elevation.

See Also

GeoFit, GeoCovmatrix, GeoVariogram, GeoVariogramDir, GeoSim, GeoSimapprox, GeoKrig, GeoKrigloc

GeoModels documentation built on Sept. 23, 2026, 5:07 p.m.