Ancillary knowledge gical
development and cropping calendars :
obtained from the nan State Production
Center (Figure 1) situated at the Dalaman
District, Dalyan, in the Protected Area (E 28°
45, N 38° 45"). The supporting data was used
to assess the vegetation amount its
distribution. The Producti ente an
area of 3 217 ha and is highly productive. There
is an Intense Soy with respect to harvesting
and regeneration and crop species are
and dude: : corn, wheat, cotton, clover, sun
flower and citrus.
Dalamat ^roduction L enter.
Figur re 1.
S SAR datasets were acquired covering the
growing season from 29 M arch to 21 September
for 1996. These were from ERS 1 and 2 SAR
with specifications of: C-band, V V-polarization,
23.0 incidence angle a mid swath, and
descending o The SAR data vere th
look pr ecision Pinot: with a ground resolution
of 12.5 by n agi nuth and range
directions. The geometry of images with the
same frame number is similar apart from a small
c- and y-shift. A cloud-free Landsat TM scene
for 24 July ; 199€ 5 was also used with combination
Al
with the SAR data. The remotely sensed data
srry 1
specifi icatio ons are g AV en m i able 1.
Fe Satellite Date Orbit / Frame
_ERS1 D 29 March 1996 24601 / 2871
ERS1 D 3 May 1996 25102 / 2871
_ERS2 D 8 June 1996 05930 / 2871
ivi _ERS 2 D 17 August 1996 06932 / 2871
ERS2 D | 21 September 1996 07433 / 2871
Path / Row
[| 24 July 1996 179 / 34-35
1. Specifications of the data used.
Lao
e
DATA ANALYSIS
In this study, SAR data sets were evaluated
through visual interpretation as well as through
selected image processing techniques such as
wultidimensional classification and principal.
compone nt analysis. As a first step, calibration
into sigma-dB values was performed in order to
reference the multitemporal ERS data. Then the
16-bit data were converted to 8-bit and inherent
speckle noise was reduced. For this purpose
various filters, such as Mean, Median, Frost,
Lee and Gamma-Map, were tested with different
window sizes and over two passes on the data
set. The first filtering was done using a 3 x 3
median filter, the second iteration using a 5 x 5
mean filter.
E
The 5 SAR images were co-registered to create
a multitemporal data base. This was achieved by
taking the 17 August data as the reference
coordinate system and resampling to 25 m pixel
size using the nearest-neighbor algorithm in
order to retain the multitemporal information
content of the radar data (Keil et.al., 1995).
Of the Landsat TM data, only channels 2
through 5 were used due to : the susceptiblity of
channel 1 to atmospheric effects, the low
resolution of the thermal band (channel 6), and
the high correlation between channels 7 and 5
(Brisco, 1995).
Finally, the data were geometrically corrected
using 1: 25 000 topographic maps for addition
to the GIS data base. A linear transformation
with a + 0.5 pixel RMS error using 10 and 15
points was performed for co- and geo-
registration process, respectively. An overview
of the data processing and analysis is given in
the flowchart in Figure 2.
560 International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
3.1.
clas
top:
geo
shac
regi
havi
Cla:
alor
Wer
All:
with
Afte
Was
ISO
clust
Max
max
the
norn
to a
prob
to th
othe;
were
