s.surf.tps

[Helena Mitasova]

Martin writes:
>We are trying to use s.surf.tps and keep getting nonsensical results, like
>raster files with all zero or a lot of segments that don't fit together at
>all. The funny thing is that we did use s.surf.tps about half a year ago
>with good results, and to the best of my knowledge we are doing the same
>things now as then.

Martin are you using the new or old s.surf.tps? 
raster files with all zeroes can occur if your data are floats
0-1 and integers make them all zeroes. It is possible that your
new data set needs completely different parameters than what
you were using before and that's why you are getting the visible segments.
(for visible segments increase npmin - with old s.surf.tps you
can go up to about 350, with new one, I have used npmin for some
very bad data as large as 600 with good results).
Also very low tension can cause difficulties with segmentation, because
the influnece of each data point is over a very long distance and 
segmentation assumes that the very distant data points don't influence
the interpolation function in the segment (what is not true for very
low tension).

If you cannot get reasonable results from your data set it would be better
if you send us the data and region information so that we can find
out where the problem is. You can try to run s.surf.tps on your
old data set or any other data you have to see whether your installation 
of program works OK.

Helena

P.S. We are working on making tps programs more intelligent in
finding the proper parameters, new tps programs provide some help
but for an automatic selection of parameters, more work is
needed. We will keep the users posted on any new developments in
this direction.