[R-sig-Geo] uncertainty?, simulations?

Edzer J. Pebesma e.pebesma at geo.uu.nl
Sun Aug 28 21:03:23 CEST 2005



michal wrote:
> Hi List,
> 
> I used ordinary kriging (gstat and R) to interpolate/evaluate spatial
> process (soil respiration rate).I measured points  in regular square
> lattice. Soil respiration rate can be various ( form very variable to
> very homogeneous) depends on many factors like temperature, humidity
> etc. I want to evaluate uncertainty of my scale of lattice. In other
> words I want to know if I chose correct dens for measured points
> (distance space between points) for certain situation.
> I hope I'm more-less clear :).
> 
> How to investigate this? What should I read/use first?
> 
> greetings,
> michal
> 

Michal,

you could look at kriging variances to start with, or at
kriging standard errors, by taking the square root of the variance.

If the standard errors are small enough, you're fine. If they are
too large, try adding points at areas where points are sparse, just
filling in nonsense values, and see how the kriging standard error
decreases--kriging standard errors do not depend on data (measured)
values otherwise than through the variogram.

Packages gstat has a function, ossfim(), which let you vary sampling
density (on a square grid) and block size to see how kriging errors
results from a given combination of them, if you're only interested
in point kriging, use zero block size.

There are a couple of good introductory texts on geostatistics;
you can find more about it on http://www.ai-geostats.org/

Best regards,
-- 
Edzer




More information about the R-sig-Geo mailing list