jwo at leicester.ac.uk
Fri Jun 18 15:48:31 EDT 1993
> > it is interpolating in a very un-smooth way, *sometimes*.
> > ie for some levels it just "steps" down the contour, other times
> > it makes nice ramps across them.
> On an RS6000 it also takes a very long time... As for the steps, I
> think they are the result of r.surf.contour being just an inter-
> polation routine, ie. the rasterized contour lines themselves are
> not touched by the routine. In some cases, particularly where the
> contours are closely spaced, this means that the elevations for
> which contour lines are available get an inordinate percentage of
> surface area. This is visible as a plateau in 3d representation.
> If you do not want steps, you can mask off the original rasterized
> contours then rerun r.surf.contour or r.surf.idw to fill in the
> holes, or use an approximative method such as s.surf.tps.
I am not sure I entirely agree with the above explanation for the
commonly found plateaux in poorly interpolatied surfaces. In my expereince
the effect is usually more pronounced where contours are widely spaced.
For a discussion of this, and on using visualisation to detect
interpolation error (using GRASS), see
Wood and Fisher (1993), Assessing interpolation accuracy in elevation
models, IEEE Computer Graphics and Applications, March 93, pp.48-56.
It is also worth keeping an eye out for the 'Surfaces' column in GRASS
Department of Geography
University of Leicester
Leicester LE1 7RH
E-mail: jwo at uk.ac.leicester
More information about the grass-user