[STATSGRASS] kriging results not within range of original data

[Maciej Sieczka]

Edzer J. Pebesma wrote:
> Hi Carlos,
> 
> that is correct, and due to the fact that kriging weights are not
> constrained to be in [0,1]. In fact, they do become negative and may be
> larger than 1. Consider as alternatives either inverse distance weighted
> interpolation, or some form of constrained kriging (although I wouldn't
> know where to get the latter).

I'd suggest the natural neighbor interpolation. A nice library is here:
http://www.marine.csiro.au/~sakov/

I have written a GRASS script wrapper for the nnbathy interpolation
program that comes with nn lib, r.surf.nnbathy, available at:
http://grass.gdf-hannover.de/wiki/GRASS_AddOns#Raster_add-ons

Maciek