[GRASS-user] Cross-validation output

[Maris Nartiss]

Hello Luca,
first - double check Your data - are all data points OK. I have seem
when in some location are two points with different values (operator
error) and thus interpolated surface looks strange.

second - generate segment map to see which points affect that
"strange" area and probably consider adding additional data points in
bogous area. Sometimes it's clear for human what kind of values one
can expect in specific area but interpolation algorithms can't get it
right.

third - probably You should play around with tension parameter - lover
impact distance of single point and thus reduce interpolated surface
buckling.

Maris.


2008/8/4, Casagrande Luca <thedoktor78 at gmail.com>:
>
> Hello everybody,
>
> I am using v.surf.rst to interpolate a bathymetry survey. Following the
> examples on the 3rd Grass Book, I have
> interpolated the results from the cross-validation procedure to check if
> there are areas that need a more accurate sampling.
> Most of the surface is ok, but there is a small area where the difference is
> very high; you can check it out from this picture:
>
> http://www.thebighouse.biz/data_cv.png
>
> The issue is that there are bigger zones with no points that show a lower
> difference value.
> Why this?
>
> Thank you very much for your help.
> Luca
> --
> View this message in context:
> http://www.nabble.com/Cross-validation-output-tp18808883p18808883.html
> Sent from the Grass - Users mailing list archive at Nabble.com.
>
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/grass-user
>