[STATSGRASS] working with lat/long data in R+gstat
Edzer J. Pebesma
e.pebesma at geo.uu.nl
Tue Dec 12 16:03:51 EST 2006
Carlos "Guâno" Grohmann wrote:
> Can I work with gstat with lat/long data?
Ehm, no/yes. If you wait for a few days to get gstat version 0.9-35
(just uploaded to CRAN), it does use great circle distances for distance
evaluations if data are longlat. (i.e., they evaluate FALSE when passed
to is.projected()). Make sure that for e.g. kriging you use a variogram
model that is positive definite on the sphere; gstat does not check
this. It seems that the exponentional model should work. Be aware that
this is a mostly untested domain, for gstat at least. (You may want to
look at package fields as an alternative).
> anyway I can import a
> GeoTiff into R as a dataframe without projection?
I don't know. I think rgdal will import it as a SpatialGridDataFrame,
and I don't know if it retains the projection information in the GeoTIFF
correctly. You may need to specify it yourself after importing.
More information about the grass-stats