The gstat R package and S-Plus library: multivariable geostatistics for S

Introduction

• multivariable geostatistics (direct/cross variogram modelling; cokriging; cosimulation)
• block kriging, for both regular and irregular blocks
• kriging/sequential simulation in a local neighbourhood
• three-dimensional kriging/simulation

Reference

Pebesma, Edzer J, 2004. Multivariable geostatistics in S: the gstat
package. Computers & Geosciences, 30: 683-691 

Interface

# calculate sample variogram:
v <- variogram(log(zinc) ~ 1, locations = ~ x + y, data = meuse)
# fit a model to sample variogram:
v.fit <- fit.variogram(v, model = vgm(1, "Sph", 900, 1))
# plot both:
plot(v, model = v.fit) # view result
# use ordinary kriging on a regular grid, in meuse.grid:
z <- krige(formula = log(zinc) ~ 1, locations = ~ x + y, data = meuse, 
        newdata = meuse.grid, model = v.fit) 
# view ordinary kriging prediction and prediction error
levelplot(var1.pred ~ x + y, z, aspect=mapasp(z), 
        main = "ordinary kriging prediction") # view result
levelplot(sqrt(var1.var) ~ x + y, z, aspect=mapasp(z), 
        main = "ordinary kriging prediction error") # view result

log(zinc) ~ 1

~ x + y

Alternatively,

~ x + y + z

~ x

Documentation

Examples

demo(examples)

data <- function(x) { } # make empty function
meuse <- meuse # make locally modifyable copy

demo(grass)

Feature list

1. uses S formulae for the linear data model and the spatial locations
2. data may be 1-, 2- or 3-dimensional
3. sample variogram, cross variogram, and directional variogram estimation
4. variogram cloud estimation; interactive variogram cloud point pair identification
5. nested variogram model fitting
6. fitting of a linear model of coregionalization using nested models
7. simple kriging, ordinary kriging, or universal kriging given arbitrary base functions (regressors)
8. plotting functions for direct, directional or multivariable variograms and fitted models using lattice/trellis graphics
9. point, regular block, or irregular block kriging
10. univariate, multiple (uncorrelated), or multivariable (correlated) cokriging
11. global kriging or kriging in a local neighbourhood
12. any combination of the above three items (e.g., universal irregular block cokriging)
13. sequential Gaussian or indicator simulation equivalents of any of the above combination chosen
14. n-fold or leave-one-out cross validation for kriging
15. support functions for plotting images using xyplot, image or levelplot
16. efficient (fast) sequential simulation of many realisations in one pass of the random path
17. fast local neighbourhood selection algorithm that uses the bucket PR quadtree algorithm
18. most of the code is in C, most of the C code has been used for a long time
19. optional debug output for many aspects of the process

Installation instructions

the gstat S-Plus library (unix, >= 6.0) requires a few steps:

• unpack the tree in an empty directory,
• in this directory, run:

  ./configure
  

• followed by one of:
  • if you are using ksh or bash:

     SPLUS=Splus6 make library
     

  • if you are using tcsh or csh

     setenv SPLUS Splus6
     make library
     

  where Splus6 is replaced by the script that starts your S-Plus version (assumed to be Splus 6.0 or higher)
• optionally, add the command:

  make veryclean
  

  to got rid of some (now obsolete) files
• and finally, move the gstat directory to the directory where you want to keep S-Plus libraries, say /home/user/S, and load it in an S-Plus session with the command:

  library(gstat, lib.loc="/home/user/S")
  

Install on R for Windows:

Install on S-Plus 6.0 or 6.1 for Windows

• Download the zip file for S-Plus for Windows version 6.0 or 6.1;
• unzip the file, this results in a directory called "gstat";
• move this directory to the library directory in the S-Plus source tree, typically

  c:\Program Files\Insightful\splus61\library

• start S-Plus
• File, Load library, select gstat from list.

Download source tree:

• my R development version
• gstat S-Plus library:
  • 0.9-5,
  • 0.9-6
  • 0.9-7
  • 0.9-9
  • 0.9-12
  • 0.9-17
  • 0.9-18
  • Windows version 0.9-17