[STATSGRASS] unfilled matrices with R and gstat in grass 6.2.0beta3

[Roger Bivand]

On Tue, 13 Feb 2007 roger at spinn.net wrote:

> 
> Folks, 
> 
> This is a user's question, rather than a developer's question. 
> 
> I have R with the gstat library working with grass6.2.0beta3.  On Linux 
> 2.6.13 x86_64 (SuSE 10.0).  The software generally works and I can confirm 
> its results at least approximately against other software.  The problem I'm 
> having is that the resulting map is sometimes not completely filled.  I get 
> calculated values on diagonal lines in the map and the intervening areas are 
> left unfilled -- they have null values.  The same pattern is visible whether 
> I look at the result from within R or export the result to a grass raster 
> file. 
> 
> The pattern is not so simple as a limit to the size of the resulting map.  I 
> have a small set of data (about 60 locations) and need to construct a map of 
> an area of about 20x30 km, using a grid with spacing of 30mX30m.  The 
> resulting map should have 686736 values, but the the result only contains 
> 171684 non-null values.  That is exactly 1/4 of the cells.  If I increase 
> the cell area to 60mX60m so that the grid only contains 171684 cells, then 
> the result contains 85842 non-null values.  That is 1/2 of the cells. 
> 
> I can construct a map using 16 tiles across the same area and patch them 
> together to get a completely filled map, but for most purposes that is an 
> excessive amount of work. 
> 
> With a different data set with about 250 points and trying to construct a 
> map over a larger area using a grid with 169576 cells, each 120mX120m I do 
> get a filled raster map in the result, though the result looks a little odd. 
> 
> I don't understand what is happening and I've found no mention of this 
> problem elsewhere.  Any ideas? 

Without seeing your code, and perhaps save() of the input objects to the 
kriging prediction function call, it's difficult to tell. It feels like 
the newdata= argument for the prediction location being malformed. If you 
can post code and put sample data for replicating the problem on a website 
somewhere (the *.Rdata file from save), we can try to see what it is.

Roger

> 
> Roger Miller
> 
> _______________________________________________
> statsgrass mailing list
> statsgrass at grass.itc.it
> http://grass.itc.it/mailman/listinfo/statsgrass
> 

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no