[R-sig-Geo] custom variogram models

[Edzer Pebesma]

Dylan Beaudette wrote:
> Hi,
>
> Is it possible to create "custom" variogram models that can be used within the 
> gstat framework?
>
> For example, would it be possible to generate a custom exponential model in 
> the form of:
>
> gamma = C [ 1 - exp (-3h/a) ]
>
> instead of the typical:
>
> gamma = C [ 1 - exp (-h/a) ]
>   
Hi Dylan,

variogram function evaluation is done in the gstat C code, to which you 
cannot pass R functions right now. If you need a particular variogram 
function that is not there, it may be easy to implement it (I could help 
here), as the code is set up such that adding functions is close to 
trivial if you've done it once. In that case, fit.variogram, vgm etc 
will also work.

It is possible to pass a custom variogram function to the krige0 
function, which does ordinary point kriging for global neighbourhoods 
(and is pure R). It means that you need the function including its 
parameters, you don't pass on C and a separately, as your example 
suggests. See ?krige0 for an example (and ignore the y argument); in 
addition, demo(wind) ends with an example with a separable product 
space-time covariance defined as R function, which is also not available 
in fit.variogram, krige etc.

This is an active research area for me, now -- feedback & suggestions 
are more than welcome.

-- 
Edzer Pebesma
Institute for Geoinformatics (ifgi), University of Münster 
Weseler Straße 253, 48151 Münster, Germany. Phone: +49 251 
8333081, Fax: +49 251 8339763  http://ifgi.uni-muenster.de 
http://www.52north.org/geostatistics      e.pebesma at wwu.de