[R-sig-Geo] spatial correlation structure for GLS (nlme)

[Ing. Agr. Agustín Alesso]

Hello,

I'm trying to include a spatial correlation structure to an ANOVA
one-way GLS model. First, I fitted a simple model asuming uncorrelated
errors and plot the residuals variogram which verified the spatial
autocorrelation.
The corStructure seems to be a mixture between to spherical structures
(2 ranges). So, before adding complexity to the model, I tried to fit a
few models with basic predefined corStructures and then looking at their
residuals variograms. Including correlation results in smallers AICs
than without asuming correlation, but the variograms still showing
spatial correlation of residuals. 

How can I bluid a spatial correlation structure including this
sphericals features?

Thanks in advance,


Agustin