fitting a surface to widely spaced data

[Eric Small]

Hi.

I am trying to fit a surface to widely spaced data, that is concentrated in 
particular areas.  In other words, i have a ton of points in a cluster, 
but lots of space between clusters.  In general, the distance between 
each data cluster is on the order of several to 10km, while the 
grid spaceing is 50-100m.  

When i use the r.surf.idw (or 2) the results are rather ugly.
In addition to being unattractive, the fitted surfaces does not make 
much sense either -- it has scarps and holes.

Does anyone know a way to fit surfaces not using inverse distance weighting?
How about fitting a 3rd or 4th order polynomial to the data?
I know this can be done with other GIS software, but I am working with
GRASS and have not found a way.

Any help would be great!

Eric Small