Re: Fw: Gstat- Multiple mode and kriging to non-gridded locations

[Edzer J. Pebesma]

(I Cc: this to gstat-info, as Guillaume tried to send it there; however 
the list is
blocked for non-subscribers)

Guillaume Larocque wrote:

Hi, I have two questions regarding Gstat, hope somebody can help...

1)  What happens if you are in multiple mode (i.e. many variables and
variograms are defined) but each variable has a different search
neighbourhood (i.e.min, max, radius)? Does it just take the neighbourhood
   

of
 

the first variable?

No; gstat follows specifications of each variable separately.

2) I am trying to krige at non-gridded point locations defined in a text
file. So I use something like:
[...]
data(): 'xyz.txt', x=1, y=2, v=3;
set output = 'pred.txt' ;
One of the problems I get is that the predictions in the pred.txt file are
not in the same order
They should be, unless you use sequential simulation (which takes a 
random path by
default, however this default can be overriden -- set rp = 0;)

as the xyz.txt file. Also, my coordinates are in UTM and they are all cut
   

to
 

~6 digits in the pred.txt file. I was also expecting the third column of
   

the
 

pred.txt file to be the third column of the xyz.txt file but it's just a
repeat of the fourth column.
So I cannot use results from the pred.txt file because many of the
coordinates are the same due to the lack of precision and identifiers are
absent.
   

Add  a command like:

set format = '%15.10g";

to let the coordinate printing take up 15 positions, 10 after the 
decimal point, IIRC.
If you can't work it out with "set format", chances are that an 8 byte 
double
cannot represent your coordinates, in which case you should choose a
different origin for your coordinates (e.g. the lower left corner of 
your study
area).
--
Edzer