[R-sig-Geo] [52N Geostatistics] Space-time spatial variability

[Edzer Pebesma]

On 02/21/2011 10:07 PM, yong.t.li at optusnet.com.au wrote:
> Dear All,
> 
> I plan to use gstat to do some 3D kriging on a spatio-temporal dataset I have collected to look at the spatio-temporal distributions of soluble nutrients in shallow groundwater.
> I did read some papers such as written by Heuvelink and De Inco, quite impressed by the outcomes and curiously want to do my own. When I ran demo(pcb) and demo(wind), found the outputs were slightly different from what I expected. they did not calculate the temporal semivariance (or maybe they did but I just did not know where they are).
> In the literature, people always talk about the combination of spatial variance and temporal variance, such product and product-sum functions, but I did not see them in the demo(pcb) and demo(wind).
> 
> Any experts in the area giving me a full procedure of using gstat to do 3D kriging is highly appreciated.
> 
py
demo(pcb) essentially uses cokriging, and sees the (4) time replicates
as correlated, but different variables. Spatial correlation is then
modelled in the direct variograms, spatio-temporal correlation in the
cross variograms.

demo(wind) in package gstat defines a separable covariance function
(function covfn) similar to the one used by Haslett and Raftery, that is
passed to krige0. You could define your own function, such as a
product-sum model, here.

In package spacetime the same example is worked out (see the vignette),
using spatio-temporal data structures.

In principle you can mimic product-sum variograms by the metric (3D)
variogram model in 3D kriging in gstat by setting the right anisotropy
ranges to values very close to zero.

> Regards,
> 
> Yong Li
> 
> ps. 
> I used to be on the list of r-sig-geo, but was knocked out by unknown reasons. wish this time I can make it.

I'll keep them informed. Maybe your questions were too hard. ;-)

>  
> _______________________________________________
> Geostatistics mailing list
> Geostatistics at 52north.org
> http://list.52north.org/mailman/listinfo/geostatistics
> http://geostatistics.forum.52north.org

-- 
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