surface_deviations

Helena Mitasova helena at zorro.cecer.army.mil
Tue Apr 12 15:52:29 EDT 1994


>  The raster file and the site file seem to have a
>  variance.  I would like to find the difference
>  between the site and raster file.  If this information
>	can be graphically reference it would be fantastic!
>
>
>Kym M. Lenkus
>Davidson Laboratory
>
>--Boundary (ID biW9oh9w5L+HmQxKPgc2dQ)--


If you use the new s.surf.tps which Irina has just put on moon
for creating the raster file from site file you can use the
option to output deviation site file which is computed simlutaneously
with computation of the resulting raster map. You can then
use s.univar to compute the mean, standard, maximum, minimum
deviations (and more). For visualization you can transform
the site file to a raster file, assign it proper colors and 
display it over the raster map using d.rast -o option.
And if you have SGI you can do even fancier things, like display
the sites as glyphs sized according to the deviation along 
with the 3d view of your surface using SG3d. We will put also s.surf.cr
for small data sets on moon soon so that you can also compute
the site file with predictive error. 
By comparing the deviations (telling how much the
resulting surface differs from given data) and crossval.
error (telling how much you can expect your surface deviates
from the reality between the data points) with the accuracy
of data, you can get a very good evaluation of your results
and use these measures for e.g. error propagation or optimization
of interpolation.

Helena



More information about the grass-user mailing list