surfaces-sites

[Helena Mitasova]

Andrea writes:
>I have some point measurements over a field and I would like to interpolate them over
>the whole area.
>I am working in a dummy xy location, because I am doing a test work.
>I have created an ascii file containing eastings and northings, and I put the value
>of my measurement as  label.
>
>I  imported the data via s.in.ascii and I tried s.surf.idw, which worked, though the
>output was not very convincing...
>
>Could anybody tell me if I followed the correct procedure ?


Andrea, try s.surf.tps. As you have measured data, they usually have
some error associated with it, so use smoothing (try 0.1, if you
have lots of noise in your measurements you may need more). You can use
the default tension. If you have less than 400 data points, set
segmax=400 (program skips segmentation and runs faster). As it is now,
you will have to put your z-values as #z, your site file should have
 x|y|#z
(Dave Gerdes has already explained why, we will fix it as soon as
possible and follow the suggestions by Dave and Mike Shapiro for
supporting the standard site format)
If you have any troubles with this program, write to me directly
(helena at zorro.cecer.army.mil), we are using this program for almost two
years for a great variety of applications with very good results.
You can find more information on how to use this program in the
manual page and in GRASSClippings (SURFACES and MODELING column).

And then Kym writes:

>When you look at the map you produced did it look like blocks
>and not a smooth line?  If so, I had the same problem

Does this relate to s.surf.tps? If yes,
this sounds like you have used the default npmin and it was too low
for your data, you may need to go as high as 300, ( the default is
150), you can also increase the size of your segments by increasing
segmax.  Visible segments can also occure outside the area where
the points are - to avoid this use mask, so that the surface is
 computed only in the area where the points are located.

And when I am on the line - we are preparing lots of new additions
to multidimensional surface modeling, analysis and visualization tools
 for GRASS - look for the demo (by Bill Brown et al.) and poster (by Irina
Kosinovsky et al.  at the GRASS users conference. Enhancements and 
additions include refined segmentation procedure, output of deviations,
 quadtrees, crossvalidation for finding optimal tension and smoothing, input
from sites or vector maps, splines for resampling, smoothing and 
computation of partial derivatives of raster maps representing surfaces,
sophisticated 2d and 2.5d triangulation, 3D and 4D interpolation and
visualization and more. Some info about these new developments was already
published in GRASSClippings, next issue will bring some news too.
We will be giving the new programs on moon after they pass the inhouse
testing.

If you have problems, comments or suggestions related to surfaces
and modeling with fields in GRASS, don't hesitate to write me

Helena Mitasova
Spatial Modeling and Systems Team
U.S.Army CERL

helena at zorro.cecer.army.mil