Skip to content
Prev 11029 / 29559 Next

regression kriging, newdata parameter

So you have values of 'sm' (target) and 'res' (covariate) at all 
sampling locations. Now you would like to predict at some other 
locations? You obviously have to have values of 'res' at those locations 
(This is why you get an error in gstat)

"krige" function needs either a SpatialGrifDataFrame or a 
SpatialPointsDataFrame, so the prediction locations do not necessarily 
have to be on a regular grid. The issue is that covariates need to be 
available at prediction locations.

Technically speaking, you could also estimate values of predictors using 
some kind of resampling method (e.g. splines), but then you would use 
estimated values (covariate) to estimate target variable, which does 
feel a bit awkward (but is doable).

PS: gstat implements the Kriging with External Drift method 
(http://dx.doi.org/10.1111/j.1467-9671.2006.01015.x), but this should be 
equivalent to RK (regression and kriging separately, then sum the maps).

HTH

T. Hengl
http://www.wewur.wur.nl/popups/vcard.aspx?id=HENGL001

Op 23-2-2011 20:51, giuseppe calamita schreef: