[GRASSLIST:3476] Re: curvature map

Radoslav Bonk bonk at fns.uniba.sk
Tue Apr 9 07:35:34 EDT 2002


Hi Ahmet,

The quality of morphometric terrain paramaters strongly depends on DEM
quality and its vertical disection, interpolation method used, the design
of the input data points , scale, and the type of application you need
them for. To generate acceptable curvature maps is a problem itself.

Basically, the smoother the DEM (the lower the high-frequency artifacts)
the better the curvature map. How to make the DEM soother: general
appraoch is to resample the DEM, or reintreploate it.

The DEM produced by stereo-pairs autocorrelation technique, usualy reveals
high frequency artifacts, although it is of good quality as for its
locational accuracy and precision (the elevation). When smoothing
procedure is applied you lose the locational precision (the elevation),
and for example the terrain features goes to trend, but you can produce
better curvature maps.

Here is the one of the smoothing procedures:
1. Acquire new site data from
your original DEM. Random locations are suitable (r.random, e. g. the
number of sites = 25% for 20m resolution).  Convert them sites
(r.to.sites).
2. Reinterpolation. s.surf.rst. interp. function works fine,
(smoothing - higher values (3-5), tension - lower values (15-25). The
curvatures maps should look better, however the vertical accuracy of such
a DEM was degraded to a certain degree.

Hope this helps,

Rado

PS: Here is the example of resampled and reinterpolated DEM.
http://rbonk.host.sk/example/preprocess_zs.jp

> Concerning the project I work on, I produced a curvature map.
> But it appears too dispersed. how should a correct curvature
> be ? what should I have done so that it does not seem too
> scattered.
> I attached the view of the curvature map.

-- 
 Dipl. geogr. Radoslav Bonk
 Comenius University,Faculty of Natural Sciences
 Department of Physical Geography and Geoecology
 Mlynska Dolina 842 15,Bratislava SLOVAKIA
 fax:07/65429064 phone: 00421 (0)2 602 96 250



More information about the grass-user mailing list