digital elevation model with 25 m raster (DEM 
25), 
2) spot heights; 
trigonometrical geodetic points, 
fundamental geodetic points, 
3) vector lines and polygons digitized from topographic 
maps in scale 1 : 25,000; 
contour lines, 
hydrographic elements (lines of the streams and 
polygons of the lakes and sea) - without height 
attributes, 
railways - without height attributes, 
roads - without height attributes, 
4) other data used only for visual control; 
scanned raster contour lines from maps in scale 
1 : 5000. 
As raster orientated data DEM 100 is available with tested 
and known height accuracy from 3.3 to 16.1 m and 
DEM 25 with predicted height accuracy of 2 m. 
Planimetrical accuracy of DEM 25 and 100 should be 
around 1 m, but more probable it is around 5 m. 
Trigonometrical and fundamental geodetic points have 
theoretical planimetrical and height accuracy up to 1 m, 
and contour lines planimetrical accuracy from 5 to 10 m 
and height accuracy about 10 m. 
4.2. Selection of suitable data for modeling 
For quality control of input data the international data 
standards (CEN), which contain some statistical 
parameters, are used. Unfortunately those parameters are 
not always sufficient for complex data tests. Visual tests 
are also very important for quality control of DTM. Some 
of them could not be replaced with statistical parameters. 
For example statistically one tested DTM could be better 
than other but on the second one could be clearly seen 
river beds, which are unclear on the first. 
Statistical methods of quality control are mostly 
considered as objective while visual as subjective. The 
best choice is combination of both methods. Some of the 
statistical methods can be found in the following groups: 
evaluation of single data layers, 
evaluation with combination of more data layers, 
evaluation of data layers with regard to reference 
points, etc. 
Some of the visual methods of DTM quality control are: 
inspection of characteristic points and lines, 
inspection of course of the hillshaded relief or 
slopes and aspects, 
implementation of Monte Carlo methods for 
example for visibility control, and much more 
(Podobnikar, 1999). 
4.2.1 Implementation the visual quality control: After 
the first visual review of data we decided that lines of 
railways and roads without height attributes can not 
improve the final DTM 25. So we omitted them from 
additional trial. 
clearly seen that in the central part of DEM 25 are some 
flat triangular surfaces. The problem is that this part of 
DEM 25 has been interpolated from contour lines which 
are not presented at the very steep areas. After 
triangulation performed the mentioned holes were 
represented with large triangles. After comparing DEM 25 
with DEM, generated from vector contour lines, we 
decided not to use DEM 25. The reason of such decision 
lies in similarity of the both datasets. DEM 25 was 
obviously generated from the same contour lines. On the 
other side DEM 100 has poor spatial resolution but 
visually it is correct dataset which is independent from 
DEM 25. We decided to use it in interpolation process. 
Figure 3: Comparing hillshaded DEMs: DEM 100 on the 
left and DEM 25 on the right for the Alpine test region (2). 
Next visual control was the comparison of vector contour 
lines (from maps 1 : 25,000) with scanned contour lines 
(from maps in scale 1 : 5,000). With visual overlay of both 
datasets we wanted to perceive difference of two 
(different generalized) sets of contour lines. Because we 
do not have a database of elevation values for both 
datasets, we can comment only the detail differences. It is 
paradoxical that in general the contour lines from scale 
1 : 5000 are not much more detailed than the those from 
scale 1 : 25,000 (figure 4). We even noticed that in some 
cases contour lines in larger scale are more detailed than 
in small ones. The reason probably lies in inhomogeneous 
capturing of data in scale 1 : 5000, while vector contour 
lines are much were captured more “compactly”. 
téiééí 
Figure 4: Comparison of two contour data sets for alpine 
region (2): In black are scanned contour lines in scale 
1 : 5000 which are overplayed with vector contour lines in 
scale 1 : 25,000. 
Further control was done with comparative visual testing 
of both, DEM 100 and DEM 25. Figure 3 shows how much 
more detailed the DEM 25 is than DEM 100. But it is 
We also performed visual methods for elimination of gross 
errors from the contour lines sets. It was done with 
comparison of contour lines derived with interpolation