data have been found superior to MSS imagery 
(Horter and Ahern, 1986). 
Latty and Hoffer (1981) studied the utility of TM 
spectral bands for a site in South Carolina using 
TMS data. They analyzed the statistical 
separability of spectral classes using var ious TMS 
spectral band combinations. Their results showed 
high separability between a number of forest 
classes. 
Williams and Nelson (1986) report the results of a 
Nor th Carolina study where substantial 
classification improvements were obtained with TMS 
data (relative to MSS data) for seven forest 
classes. The overall per formance of TMS 
classification was 60 percent as compared to 39 
percent for MSS data. 
and DeGloria (1985) present the results 
actual TM data in forestry 
that the best 
the MSS data 
TM data will 
Benson 
from the use of 
application in Carolina. They stated 
TM band combination was better than 
and the best results indicated that 
provide higher classification. 
The paper of Jones et al. (1988) describes the 
complementary use of digital terrain information 
and SPOT-1 HRV multispectral imagery for the study 
and mapping of semi-natural upland vegetation. A 
digital terrain model was derived for a study area 
in Snowdonia, Wales, and was used to generate 
slope and aspect images. 
the assessment of 
Borry et al. (1990) studied on 
the value of monotemporal SPOT-1 Imagery for 
forestry applications under Flemish conditions. 
Research results revealed that the enhancement 
procedure for visual interpretation is of minor 
compared to the acquisition data. 
visual analysis is lost if 
forest information are 
impor tance when 
The advantage of 
detailed levels of 
required. 
3. MATERIAL AND METHODS 
3.1. Study Area 
The study area chosen for analysis is located 
within the Abant forest and the Alada§ forest area 
in the eastern part of Bolu province. The study 
site is approximately 943 square kilometers (364 
square miles) (Figure 1). 
latitude N 40°42’ and 
This area is situated on 
longitude E 31°28’ with an altitude 738 meters 
above the sea level. The area has highly 
productive, level to rolling terrain and 
intensively vegetated. 
The region has semiarid mesothermal climate with 
dry summers and cold winters. The coldest month is 
January with -4.6 °C mean and the warmest August 
with 39.4 °C Annual average relative humidity is 
74%. Mean annual precipitation is 529.2 
mil imeters, mostly taking place between late 
September and late June. 
Soils of study area are generally within Brown 
Great Soil Group. Soil texture is mainly heavy, 
namely clay and loamy-clay. 
Primary land-cover categories in the study area 
are Forests, Water bodies, Agricultural lands, 
Non-vegetated areas. Almost 70* of the total study 
area is covered by forest. Water bodies, namely 
lakes and reservoirs is 0.3 %. Agricultural lands, 
settlements and other areas constitute almost 29% 
290 
of the study site. 
3.2. Data Acquisition and Preparation 
  
LANDSAT-5 TM data were acquired 
(Path 178 and Row 32) on 16 
July 1984. Center geographic coordinates of the 
image is N 40°20" and E 31°43’. They became 
available on CCTs of BSO format (radiometricaily 
and geometrically calibrated). Reference data used 
to support the analysis was consisted of 
identification by ground observation and recording 
on maps with a scale of 1/25 000 the land-cover 
type of necessary number of fields in the test 
areas. Topographic land-cover maps with a scale of 
1/100 000 were also used for geometric 
rectification of the image and the verification 
and accuracy assessment of the digital natural 
resources classification together with 1/25 000 
scaled maps. 
Cloud-free digital 
over the study area 
Applying special algorithms available on the ERDAS 
image processing system installed at the Remote 
Sensing Laboratory in the Ankara University, 
Agricultural Engineering Department, a subscene 
covering the study area was extracted from TM 
tapes and loaded on to floppies for further 
analysis. 
The extracted subscene was then rectified to a 
state plane map projection. The implementation of 
the rectification process was based on the image 
and map coordinates of 14 control points uniformly 
scattered throughout the study area. The map 
coordinates were acquired by using a digitizing 
table and the appropriate routines of the 
available software. The set of the acquired 
coordinates were used to compute the coefficients 
of the transformation matrix using a least square 
algorithm. The nearest-neighbour interpolation was 
used to rectify the input image. 
3.3. Field Work 
provide indispensable support in the 
remotely sensed data and are 
obtained from actual 
Ground data 
interpretation of 
helpful in the verification 
site of study and in areas of particular interest 
such as agriculture, forestry, water bodies and 
other land-use categories (Gautam and Chennaiah, 
1985) . Therefore ground data collection was 
carried out in the study site. |t involved the 
development of an overall systematic plan so that 
all the selected sites could be visited. Intensive 
study was carried out to collect data from the 
max imum number of representative areas of 
different land-cover type. Combining the 
information of land-cover gathered by field work 
with land-cover map on 1/100 000 scale, 
percentages and acreage of each land-cover 
category related to total area have been obtained. 
3.4. Data Analysis 
Preliminary visual interpretation for the 
delineation of land-cover classes was performed on 
1/50 000 scale screen display of LANDSAT 
multispectral imagery with the help ofa false 
color composite (Bands 4 red, 3 green and 2 blue). 
In this study it was evaluated a number of biomass 
transformations, or vegetation indices, in order 
to emphasize the relative greenness of the land- 
cover classes of interest. This evaluation was 
done on a representative subset of test data, and 
it was determined that the Radiance Ratio TM 4 / 
TM 3 and the Normalized Difference Vegetation 
Index Transformation [(TM 4 - TM 3)/(TM 4 * TM 3)]