International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004 
  
4. RESULTS 
4.1 Image Processing Results 
For the indication of forest fire risky places, a land use and 
vegetation map was prepared using Landsat TM data. The 
selection of the bands that was used for classification was made 
taking under consideration the spectral profile analysis. 
According to the analysis, Landsat TM bands 4, 5 and 7 were 
selected to make supervised classification. Standard 
topographic maps were used to indicate the control areas. 
Classification accuracy was determined by using 50 random 
pixels. Classification results are shown in the figure 2 and 3. 
The images were classified into eight classes as forestl, forest2, 
forest3, non forest, sea, lake, fog and sea shore using maximum 
like hood algorithm (figure 3 and 4). First, 1992 land use results 
were obtained. Then, the status of the land use in 1998 was 
examined. By subtraction of the two classification results, two 
important changes were indicated. These were acceleration in 
forest and lake areas. In spite of the fact that the classification 
accuracy results were % 93.75 and % 90.26, the cloud class in 
1992 image was misleading in determination of changes in 
forest classes. 
  
Figure 4. The result of 1992 Landsat image classification that 
was integrated with DEM 
Forest 1 
Forest 2 
Forest 3 
Non Forest Area 
Sea Shore 
Lake 
Sea 
Fog 
  
36 
  
Figure 3. The result of 1998 Landsat image classification that 
was integrated with DEM 
Land use information is an important factor in determination of 
forest fire risk. Because of this, land use classes obtained from 
satellite images were converted to vector file and integrated into 
GIS. 
4.2 GIS results 
The information system was formed in Maplnfo software 
having transferred the parameters of vegetation type, 
topography, distance from roads and settlements into a database 
that are important in determination of forest fire risk. For the 
production of the forest fire risk map, five fire rating classes are 
used. These classes are formed according to slope, aspect, 
vegetation type, distance from roads and settlements. Slope and 
aspect image 
were generated using the DEM data (Fig 5 and 6). Aspect and 
slope plays a vital role in spreading of the fire. Fire travels most 
rapidly up-slopes and least rapidly down-slope. Southern slopes 
are more vulnerable to catching fire.