AUTHORS:
TITLE:
DISCUSSION:
P. Frederiksen, 0. Jacobi (Denmark) and K. Kubik
(USA)
OPTIMAL SAMPLE SPACING IN DIGITAL ELEVATION MODELS
Torlegärd (Sweden): You have presented the classification of ter-
Frederiksen:
Förstner (FRG):
Frederiksen:
AUTHOR:
TITLE:
DISCUSSION:
rain and it is all very nice and we understand
that these are the tools that we can use. But how
do they fit to real applications of digital elev-
ation modelling? I mean, have you used this theory
to predict the optimal accuracy and check the result
afterwards and how well does it fit?
Well, that is a different point, naturally. Well,
it was used in Australia as I mentioned and a few
checkings were carried out, but naturally as always
too few. It seemed to be pretty good within 10-15$
of what we estimated. But I would like much more
testing on it. It seems nice and smooth and you can
carry it out without any problems, but more tests
on it should be carried out, I think.
Is it a two-dimensional approach taking the terrain
characteristics in the different directions into
account?
It is a two-dimensional approach and we have only
these two parameters, and observations in profiles.
This is not sufficient really. We should do it in
the surface and obtain a classification with more
or multi-parameter classification. We have not been
working on that. You have to introduce at least
orientation parameters for the classification.
know from geostatistics that normally the variogram
is estimated in two directions perpendicular to
each other and then the variation is within the
difference of the two variograms.
A. Óstman (Sweden)
TERRAIN ANALYSIS BY KARHUNEN - LOEVE EXPANSION
Ligterink (The Netherlands): Terrain classification by parameters
was also what the former speaker Mr. Frederiksen
did. I think it is a useful tool, but how do you
solve these problems, or how do you make your de-
scription by parameters, when you have terrains with
many break lines and how do you solve the problem
of break lines?
62
