[GRASSLIST:5822] Re: s.surf.rst versus the impenetrable wall

[Paul Kelly]

Hello Hamish

On Thu, 20 Mar 2003, H Bowman wrote:

> Hi all -
>
> two questions regarding s.surf.rst:
>
> 1) Is it appropriate to use s.surf.rst for non-topographic data? eg pH
> or some other continuously distributed parameter. I think it's ok, but
> would like to be sure.

I think so, as long as it's continuously distributed; there are some nice
examples relating to engine temperature and electronic structure here:
http://skagit.meas.ncsu.edu/~helena/gmslab/nongeo.html

In theory I think you should be able to use it for non-continously
distributed parameters if smoothing is set to 0 but I couldn't get that to
work properly with image data; there were some strange artefacts and
values outside the 0-255 range resulting.

> 2) Does anyone have any ideas or insights regarding how to stop
> s.surf.rst from interpolating over impenetrable barriers? Unfortunately
> it is necessary for us to interpolate over a large distances with only a
> few data points.

The problem may be an inappropriate distance weighting value being used
in the calculations; I know I am thinking more of how s.surf.idw works
here so it mightn't be appropriate. Instead of using the square root
of the distance in easting and northing to calculate the distance to the
points being interpolated, you would need to find the shortest possible
distance without passing through the impenetrable barrier. Maybe some of
the new vector networking functions in GRASS5.1 could do this but I really
have no idea.

Don't really understand the other suggestions for getting round the
problem so maybe I have the wrong idea. Just a few thoughts though...

Paul