137 
from vector ones with vector ones, and visual searching of 
the gross errors (figure 5). 
Figure 5: Computed contour lines (in black) over original 
vector contour lines (in gray). On the left where two sets 
of contours doesn’t fit, it is remarkable gross error. 
4.2.2 Implementation the statistical quality control: 
With visual tests some gross errors were eliminated. 
Further the heights of reference geodetic points 
(trigonometric and fundamental) were tested with 
simultaneous comparing with DEM 100 and 25. Points 
with gross errors were eliminated. 
We wanted to confirm the elimination of DEM 25 as input 
data set by statistical comparison with other DEM, 
generated from contour lines. The results show that the 
RMS errors are almost identical for all three tested 
regions. 
For statistical elimination of attribute gross errors of the 
contour lines many methods were used. Some of the 
effective methods use parameters from comparative 
datasets. We overlaid DEM generated from contour lines 
with DEM 100 or simultaneous compared both DEMs with 
referenced geodetic points. We used also “robust 
estimation” method based on linear prediction 
interpolation method (Pfeifer). We did many statistical 
tests for improvement the datasets. 
The result of data tests were improved datasets and 
parameters of RMS error of each thematic layer with 
regard to reference geodetic points. Table 1 shows RMS 
errors for DEM 100 and (interpolated) contour lines for 
different morphological classes as first parameter and 
average deviation of the reference geodetic points from 
DEM 100 and contour lines as second parameter. We can 
see that in all cases reference geodetic points are in 
average above DEM and contour lines. The reason is that 
geodetic points are mostly on the peaks, where 
interpolated data is always lower because of the missing 
characteristic points for interpolation. 
Morph, classes 
DEM 100 
Contour lines 
Flat surface (1) 
Hills (1) 
Mountainous (2) 
Karst region (3) 
2.0 m / 0.7 m 
10.0 m / 8.5 m 
30.0 m /12.0 m 
7.0 m / 4.8 m 
1.5 m / 0.3 m 
5.0 m / 2.5 m 
10-40 m /3.0 m 
4.0 m / 2.0 m 
Table 1: Morphological classes from three test regions 
(1-3) with parameters: RMS error / average deviation from 
the reference points. 
5. ACQUISITION OF ADDITIONAL DATA FOR 
INTERPOLATION 
With initial quality control we produced a good database 
including parameters for interpolation: 
DEM 100, 
contour lines, 
reference trigonometrical and fundamental 
geodetic points. 
The next step is to produce characteristic lines and points. 
5.1 Extraction of height attributes for streamlines 
From hydrographic elements - lines of streams - we tried 
to acquire elevation attributes by interpolation and 
extrapolation of lines of streams between contour lines 
(Heitzinger and Kager, 1998). The results were generally 
not satisfying (figure 6). The reason is that contour lines, 
digitized from topographic maps were broken on the 
crossings with other topographical features, in our case 
with hydrographic streams. 
Figure 6: Problems in interpolation with hydrographic 
elements (left - the biggest mistakes are marked with