No subject

[Helena Mitasova]

Simon writes:

>I have been told that s.surf.tps is the "best" interpolater available
>in grass.  Why is this?  My little experience with it has shown
>that it is certainly not quick!  What are its relative merits,
>in comparison with idw and idw2?

It is a completely different approach, you can find more
when reading the manual page and in GRASSClippings in
the column SURFACES and MODELING. Paper describing
the theory should appear in Mathematical Geology in
August. If you know Mike Hutchinson's
work on interpolation and approximation with splines -
then it is the closest approach to what is used in s.surf.tps.
It is slower, because the equations are more complicated
and it takes significantly larger number of points for
interpolation than idw (this is important when interpolating
from data which have strongly heterogeneous distribution,
like digitized contours in areas with flat terrain with almost
no data, neghbouring to steep slopes with very dense
points, or with climatic stations clustered in densely
populated areas and very sparse in e.g. mountains.)

And Brian is right, the proper parameters are the key
to success, we will be developing some tools which will
provide the user whit at least an interval for the proper
values of  parameters for the given data but I cannot
tell now when it will happen.


>What is the easiest way to convert a raster to a site list, so
>that I can use the s.surf.tps interpolater?  Is it necessary to
>go via an ascii intermediate step?

Through r.stats (you have already got the correct answer in
the list), however,
s.surf.tps has been written for scattered data, and is not efficient
when used for reinterpolating (with smoothing or resampling) the
raster data or for the computation of topographic parameters 
from raster data. We are developing r.surf.tps which uses the
same interpolation method but which takes the advantage of data
being regularly distributed and which is definitely much faster
(and you don't have to go through ascii site file,because
just reading a huge site file and creating the quadtree can 
take a while, and can also cause memory problems). I can give you the program
for testing if you wish sometimes this month.

You can write me directly if you have some more specific questions.

Helena Mitasova
SAS Team, USA CERL
Champaign, IL
helena at zorro.cecer.army.mil