Dear Michael
If I can exactly understand you're looking for a point location kriging.
I think your problem can simply be solved in two ways:
- using mask map coords (the way you tried). Gstat uses PCRaster maps and
the resolution you obtain in your output coordinates by using commands
mask:...; set output=...; is due to resolution of your input mask. Your
output coords are rounded to the centers of the cells of your map. You can
improve your resolution near your initial data coords by improving the
resolution of your map. If your map cell sizes are for example 100x100 m2,
you can reduce them to 50x50 m2 (you gain in resolution but loose in disk
space). You'll never obtain predictions and variances at exactly your
sample location coords in this way, unless your samples are located on a
regular grid with cell size identical to your mask map cell size.
- not using mask map coords. You can define dummy data and prediction
locations by commands:
data(a):dummy; data():'your file', x=...,y=...,v=...; method:map;
mask:'your mask'; set output=...;
You'll obtain your original coords and values plus an added column with
your predictions and one with variances.
Your output file format depends on your input data(variable) format:
eas->eas, txt->txt.
I hope this can help you.

Giovanni De Ferrari


Michael Bock wrote:

> Dear list members,
>
> to use universal kriging, I have to add a column in the file where my
> sample datas are stored. I realise that with a certain mask including a
> spatial variable and the command 'set output = 'newfile.eas'.
> Unfortunatly gstat writes in newfile.eas the locations of my samples in
> a rounded format (3679680 5691670 instead of 3679684.49 5691673.77). Is
> there a way in gstat to set the precision of the variable output??
>
> Best regards
>
>
> *******************************************************************
>
> Michael Bock
> scilands GmbH - Gesellschaft zur Bearbeitung digitaler Landschaften
> Goetheallee 11
> D-37073 G�ttingen (Germany)
> Tel: +49 / (0)551 / 531587-0
> Fax: +49 / (0)551 / 531587-3
> e-mail: [EMAIL PROTECTED]
> Internet: http://www.scilands.de


Reply via email to