Irregular data interpolation problem

[Philippe Peeters <philp(at)oma.be>]

Hi all,

Sorry if this has already been answered but I didn't find a solution in
the FAQ, various web site etc...

The problem is quite simple: I have a set of measurements from all over
the Earth. The measurements are irregularly spaced (satellite
measurements) and I want to interpolate the data to a regular grid (1
degree x 1 degree).
I have experimented with TRIANGULATE and TRIGRID with de SPHERE keyword
but the resulting interpolated grid becomes very strange as I increase
de number of measurements.

It seems that, with SPHERE keyword, TRIGRID does not make a linear
interpolation but a quintic polynomial interpolation (even though the
QUINTIC keyword is not used). Second, let's say I feed the
TRIANGULATE+TRIGRID with 10 points, the resulting grid seems reasonable.
With 100 points, it is still ok. With 200, I have very strange
interpolated value (peaks and negative value although the input value is
strictly positive). With higher number of observation points, the result
becomes more and more dramatic (I can have as much as 25000 points).

Maybe I'm doing something wrong, maybe TRIANGULATE+TRIGRID is not the
right tool for the job

-- 
Philippe Peeters
------------------------------------------------------------------------
Institut d'Aeronomie Spatiale de Belgique Tel: +32-2-373.03.81
Belgisch Instituut voor Ruimte-Aëronomie  Fax: +32-2-374.84.23
3 Avenue Circulaire                       Email: philp@oma.be
B-1180 Brussels, Belgium                  http://www.oma.be/BIRA-IASB/