19. 
1o 
  
sts of 
sts of 
Ires à 
| data 
o the 
ptimal 
sition 
indsat 
main 
tep in 
ta set 
(scene KJ 62/245 from 22.08.1995) was used which had been 
previously rectified on a topographic map (scale 1:25000). The 
average rms error was about 0,5 pixel. According to the spatial 
resolution of the Landsat- TM data the resampling size was set 
to 30 x 30 m2 The next step in processing the data was the 
radiometric correction of the data which is precondition for any 
quantitative comparisons of vegetation state analyses. It was 
applied the atmosperic correction software ATCOR for 
ERDAS-IMAGINE 8.2 (Richter 1990). Some empirical tests 
with the knowledge of the real vegetation types of the study 
area have shown good results in using the calibration model of 
Bolle which obtained high differences in the vital green 
vegetation canopy. For a successful application every data set 
was calibrated with the same reference arcas of dark pine 
forest consisting of trees aged between 70 and 110 years, and 
situated beyond the mining compact line. The visibility used 
for the atmospheric correction of any data set is documented in 
table 1. 
2.2 Vegetation index NDVI and their GIS integration 
The vegetation assessment was carried out with the parameter 
NDVI (Normalised Difference Vegetation Index). Since 
Kriegler et al. (1969) it is not only the most well-known 
parameter featuring the vegetation cover: own investigations 
with other parameters like the MSAVI2 (Qi et al. 1994) or the 
curvature value VM2 (Weichelt & Herr 1987) did not lead to 
better results as compared to the NDVI utilisation. 
Only those habitat types are integrated in the environmental 
long-term monitoring which have relatively stable species 
composition for a long time, i.e. where human impact is 
relatively weak. As suitable for the monitoring 15 out of the 28 
mapped habitat types were selected. They were classified into 
two categories: (1) Forest habitat types with typical tree 
vegetation and (2) Pioneer forest and heath habitat types with 
typical poorly sparsed vegetation cover. As the NDVI values 
were determined by habitat type scaling was carried out in 
relation to the habitat types. Figure 3 shows the eight 
respectively seven NDVI levels of both categories with detailed 
information in the upper level or in the lower level. 
Considering remotely sensed data from Landsat-TM which are 
suitable for average scales, only those habitats were analysed 
with NDVI values that have an area dimension of more than 
one hectare. These requirements were considered in creating 
the GIS. The habitats selected per type and dimension were 
masked out (Fig. 3 and Fig. 5), the NDVI calculated in them 
and the vectorised results transferred into the GIS. 
The technological chain from the georeferenced radiance image 
as input to the output vector data representing the vitality state 
of the habitats. The transformation of the results into vector 
data was a necessary step to meet a central requirement of the 
customer's GIS software. 
  
  
  
  
Habitat related vegetation state 
NDVI levels 
1 2 
-1-0.5 0.5-0.58 0.58-0.630.63-0.68 0.68-0.73 0.73-0.78 0.78-0.83 0.83-1 
1 2 3 4 5 6 7 
EN C7 ESS 
-1-0.1 0.1-0.2 0203 0304 0405 05-058 058-1 
  
  
       
  
  
  
  
Long term monitoring habitat types 
Forest habitat types 
     
[-] EH 
1 2 3 5 6 7 8 9 27 28 
Pioneer forest and heath habitat types 
[1 E33 E EZ 
4 10 18 19 22 
  
  
  
Fig. 3: NDVI scaling in relation to monitoring habitats 
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 153