[R-sig-Geo] Parallelized kriging in R (gstat/intamap)

[Jon Olav Skoien]

Pierre,

There are different ways to call the kriging routines in intamap. The 
simplest is to call interpolate with the necessary data. To see the 
effect of parallelization, you can look at the example below. This uses 
the zinc observations of the Meuse data set, interpolates to 100000 
points on a grid using automap (which uses gstat for fitting and 
interpolation). The speed up effect from parallelization depends on the 
problem, my computer spends half the time using 4 clusters instead of 1 
on the example below.


loadMeuse()
meusegrid = spsample(meuse, 100000, "regular")
system.time(x <- interpolate(meuse, meusegrid, list(mean=TRUE, 
variance=TRUE),
     optList = list(formulaString = zinc~1, nclus = 1), methodName = 
"automap"))

system.time(x <- interpolate(meuse, meusegrid, list(mean=TRUE, 
variance=TRUE),
     optList = list(formulaString = zinc~1, nclus = 4), methodName = 
"automap"))

If you have more specific questions about the use, I will try to answer 
them.

Cheers,
Jon


On 12/6/2010 3:41 AM, Pierre Roudier wrote:
> Dear list,
>
> I've been wanting to parallelize kriging using R - on a cluster.
>
> I am not aware of any parallelized method in gstat (which I'm an avid
> and happy user), and I'm just discovering the intamap package (which
> looks awesome) which seems to have that sort of capability (using the
> nclus parameter). However, I don't really get how to use it.
>
> Any pointers?
>
> Cheers,
>
> Pierre
>
> .
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>