og ims 9 d 5 2 
o. v rE TR Y 
Bielovar e pu. 
AT ye Jj 
Sn 
Alio. wi Pr 
. n e ie EON 
Ta 4 
  
  
AE 
erence 
1e oak- 
f NDVI 
pectral 
atures. 
Croatia 
able 1. 
pment 
except 
spread 
widely 
rivers 
in the 
alone, 
y, over 
is area 
(llovski 
orth of 
Wood 
atia is 
  
  
  
TN, m3 B cose 
du un X 
zona” iT ves 
0 HERIEGOVIMA: 
D Ld 
- 
  
Figure 1 Location of five observed oak-woods and three phenological stations. 
2. Satellite images 
In this work 10-day AVHRR composites were used. 
The source of images was Earth Resources 
Observation Systems (EROS) Data Center Distributed 
Active Archive Center (EDC DAAC) [3]. All images are 
S-channel, with 10-bit radiometric resolution and 1.1 km 
spatial resolution. 18-month continuous period was 
available from which we chose 12-month period, 
beginning with April 1992 to the end of March 1993. All 
pre-processing of images and creation of composites 
were conducted at the EDC. For creation of 10-day 
composites afternoon images of NOAA Polar Orbiter 
are used only, beginning with the first day of month. 
From the set of 10 daily images, parts with no cloud 
Coverage were picked up and with these patches of 
image a composite was created with the least possible 
cloud contamination. During pre-processing all bad and 
doubtful pixels were ejected. To speed up the 
processing, we joined all values for particular composite 
to the middle date of period, despite the possibility to 
determine exact date of pixel sampling from data index 
(“10" channel”). 
Pre-processing consists of radiometric calibration, 
atmospheric correction, computing of NDVI, geometric 
registration and composition. Radiometric calibration is 
problematic for first and second channel, because 
sensor technical characteristics are degraded during 
time and there is no automatic sensor calibration 
onboard satellite. For this set of data a method 
developed by Teillet and Holben (1994) [3] was used. 
This method uses reflectance from desert surface for 
monitoring time-dependent characteristics of AVHRR 
sensor. For thermal channels 3, 4 and 5 automatic 
calibration exists using internal blackbody and deep 
space as radiometric references. Atmospheric 
correction extracts influences of ozone and Rayleigh 
scattering. For correction of the Rayleigh scattering 
Teillet'S method (1991) was used [3]. This method uses 
optical depth approximation in the standard atmosphere 
with impact of local terrain to the length of the radiation 
path (global digital elevation data ETOPO5 was used). 
Correction of the influences of ozone was made 
according to climatological values, not actual 
measurements. There is no atmospheric correction for 
water vapour and aerosols. 10-day composites consist 
of spectral reflectance in percent values for first and 
second channel. 
NDVI was computed from calibrated channels 1 
and 2 according to: 
Channel 2- Channel 1 
Channel 2-4 Channel 1 
  
NDVI = 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 751