[R-sig-Geo] Local singularity

[José Matos]

Dear All,

I'm testing different interpolation settings and methods for a rainfall
field. From the complete field I take some points, perform the
interpolation and then compute the errors. The variogram was computed
for the original data.
While trying local neighborhoods, I got some unexpected results and am
fairly confident that I'm doing something wrong, although I not sure
about what it is.

It is as if outside the predetermined radius, gstat is computing the 
estimate gathering points up to the predefined minimum but without the 
closest point.

One of the command sets I used (made to enhance this effect) was the 
following:
data(TRMM): 'dataKrig', x=1, y=2, v=3, radius=2, min=4, max=4, force;
predictions(TRMM): 'predKrig';
mask: 'maskKrig.asc';
variogram(TRMM):   0.00000401 Nug(0) + 1232.26185112 Pow(0.54167,0,1);

A more elaborate description and all the files used to run gstat can be 
found at:
https://documents.epfl.ch/users/g/ga/gamito/www/gstat/gstatHelp.zip

All help on this issue is much appreciated.

Thank you in advance.

Kind regards,
José Matos