[R-sig-Geo] Spatio temporal neighborhoods in Gstat

[Dan Turenne]

Hello R-Sig-Geo,


I was wondering if anyone might be able to shed some light on how the search neighborhoods work in the krigeST function?  I understand the concept of the anisotropy ratio (For example if the ratio=0.5 then an observation 2 time units away is the same "distance" as an observation 1 distance unit away) but the documentation for krigeST in gstat says "In particular, it does not support block kriging or kriging in a distance-based neighbourhood".  Does this mean that the search neighborhood is entirely determined by nMax?  The documentation says nMax describes the maximum number of spatio temporal neighbors to use but is it possible that less then nMax will be used?


In a related question, does the amount of lags included  in the spatio temporal variogram determine the range of the temporal neighborhood?  For example if the variogram model only includes 5 time lags is it possible that an observation that is 6 time units away will be included in the kriging prediction?  Based on what I know about 2D kriging I would assume that it is possible and the semivariance value is simply equal to the sill in this case, but I want to know if there is any downside to including more time lags beyond increasing computation time.  Any help on this subject would be greatly appreciated.


Sincerely,


Daniel Turenne

University of Manitoba

	[[alternative HTML version deleted]]