[R-sig-Geo] Empirical Variogram from multiple realizations

[Edzer Pebesma]

On 08/09/2012 06:59 PM, Jordan Winkler wrote:
> I am wondering if there is a methodology within R to calculate a single
> variogram based on multiple modeled realizations of the same process.  I am
> performing regressions on residual fields from multiple climate models.  I
> would like to estimate a covariance function common to all models.  I've
> tried using the variog function in geoR, however it does not work with
> multiple values at single coordinates.  Any advice would be appreciated
> 

By "common to all models" I assume you refer to "pooled", or "averaged
over all models". In that case, you can:

library(gstat)
loadMeuse()
v = fit.variogram(variogram(log(zinc)~1,meuse),vgm(1, "Sph", 900, 1))
sim = krige(log(zinc)~1, meuse, meuse.grid, v, nsim=20, nmax=30)
sim.stacked = stack(sim)
coordinates(sim.stacked) = ~x+y
v.pooled = variogram(values~ind, sim.stacked, dX = 0) # wait a while...
plot(v.pooled, v, ylim = c(0, .7))

please look into the dX argument of ?variogram (to get pooling over
different data sets), and look closely what stack() does. Note that in
~ind, in combination with dX ind is not used as predictor, but as
criterium to exclude point pairs coming from different models (those
with different values for ind).

Best regards,
-- 
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