International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
Therefore, 1994 Landsat TM images were chosen as reference 
and other images (1974, 1984, and 2001) as predicted and 
linear regression models were set up for each band in multi- 
temporal images. Consequently, the disparities in images 
caused by environmental differences were normalized. 
For the purpose of the classification of the images obtained in 
the years of 1974, 1984, 1994, 2001 and change detection, 4 
different land use classes were determined, namely deciduous 
forest, coniferous forest, settlement areas and other non-forest 
usage. These main classes include many  sub-classes. 
Representing these classes, 42 training areas were selected for 
ages and the same training areas were used in all 
images. Fore§t Management Plans and charts were referred to 
for the determination of the training areas and the terrestrial 
data .was chedked. ) 
  
  
  
The training alleas were first subjected to signature analysis and 
the investigation of the brightness (reflection) values and 
relevant curves of these classes were conducted. Decisions were 
made on the most appropriate band combinations after the 
investigation of the tables and graphics showing the reflectance 
values of the training areas belonging to these classes with 
respect to bands. The appropriate band combinations were 
NDVI, VI (MSS4 - MSS3), Tasselled Cap (Greenness) for the 
year 1974 and 3-4-5" band of TM and ETM', NDVI, VI (TM4- 
TM3) and Tasselled Cap (Greenness) for the years 1984, 1994, 
and 2001. Afterwards, the classification of the sub-classes and 
the band combinations were made. The classification was done 
using the “Maximum Likelihood” method. 
4. RESULT AND DISCUSSION 
To determine land use changes, classificated images belonging 
to the 4 main land use classes were recoded and converted into 
vector layers with ArcInfo software and geographic information 
covers for the years 1974, 1984, 1994 and 2001 were used. 
After the query of the overlay process, the land use changes 
between the years of 1974 and 2001 were examined. 
      
  
[] Nonforest 
LE ] Decidious 
Ed Settlement 
Figure 2. Land use map of 1974 
  
  
  
  
  
  
  
Class Area (ha) Area (96) 
Settlement 3.04 0.13 
Non-forest 142.36 6.23 
Deciduous 2141 93.64 
Coniferous 0.00 0.00 
  
  
  
Table 1. Area distributions of land use class for 1974 
  
|o ; e E 
MIT T An» 
: t T e E 
A o7 PNE SD 
B Coniferous 
  
  
Figure 3. Land use map of 1984 
  
  
  
  
  
  
  
  
Class Area (ha) Area (96) 
Settlement 166.81 7.30 
Non-forest 96.66 4.23 
Deciduous 1337.33 59.39 
Coniferous 665.10 29.09 
  
  
Table 2. Area distributions of land use class for 1984 
      
  
a] Settlement 
Fo Nonforest 
M [|] Decidious 
à M] Coniferous 
xn T ene, 
Figure 4. Land use map of 1994 
760 
Inter 
  
  
Table 
Figure 
  
  
  
  
  
The sp. 
Tables 
1974, 1 
Images 
determi 
the yea 
in plac 
increase 
and set 
1994. [ 
areas of 
Table : 
directio 
Accurac 
underst: 
decision 
the acci 
were ob 
that the 
through