2004
—
'S the
an.
m of
good
ng to
]-5-],
lo not
993)
resent
s that
nd on
n). In
ology
of the
More
the
12
Ca,
OIF)
lation
»ssible
as the
and it
OIF is
olumn
bigger
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
where O , : the standard deviation of band i, i 7 1, 2, 3
,
rK,r ; : the correlation coefficient ob bands i and j
BAND COMBINATION OIF degree
TMI-2-3 TEE 20
TM1-2-4 55,26 10
TMI-2-5 17,26 I5
TMI-2-7 12,79 18
TMI-3-4 86,90 5
TMI-3-5 19.86 12
TM1-3-7 14,79 17
[7 TMI-4--5 90,27 3
TM1-4-7 105,22
TMI-5-7 20,43 1
TM2-3-4 59.65 9
TM2-3-5 16,85 16
TM2-3-7 12,53 19
TM2-4-5 59.91 8
TM2-4-7 67,95 7
TM2-5-7 17,29 14
TM3-4-5 87,73 4
TM3-4-7 101,13
TM3-5-7 19,39 13
TM4-5-7 70,25 6
Table 2. The OIF for all the combinations of bands is presented.
7. PRINCIPAL COMPONENTS ANALYSIS
The principal components were computed of the imagery. The
PC, PC, and PC; contain the 98% of the image variance. The
PC, gets more information from the near and middle infrared
regions (TM4, TMS and less from TM7). The PC; gets more
from the near infrared (TM4) and the middle infrared region
(TM5-7). Consequently it expresses the difference between the
near and medium infrared region, it is presented as indicator of
soll (contrary to the NDVI index). The PC; contains more
information from the visible region (TM1, TM2, TM3), it can
be said that it is a component with poor contrast.
The first three PC,, PC,, PC; new bands present the
Compression of spectral information of multispectral TM
imagery Landsat. The three remainder bands, PC4, PCs kat PC;
express less than 2% of the information. For this reason were
used the three first components, (Tsakiri, et al 2003).
8. VEGETATION INDEX
It is known the importance of application of NDVI index in
multispectral image of forest areas or of vegetation areas. The
NDVI index is computed by the function:
N
22
ND S IRR (2)
IR+R
Where: R = the band TM3 and IR = the band TM4
The image NDVI presents the vegetation cover.
9. INTERPRETATION
Lineaments (faults and faulted zones) mapping was carried out
through supervised navigation on 3D visualization by the
VirtualGIS programme of Imagine 8.5 (Fig. 4, 5) The use of
different combinations (Tsakiri, et al, 2003) for the better color
display of the imagery, was a consequence of the big changing
of the elevations of the area and they facilitated the
interpretation. The full collection of lineaments, as obtained
from the above analyses, was compiled on a geological map at
scale 50.000 (Fig. 7).
Figure 4. 3D Visualization of Landsat TM image, over the
DEM, where: R=band7, G=band4 and B=band2.
d: dolomitic marbles, S1: spring Aggitis, W1: well
The faults and faulted zones are detectable mainly from
pronounced vegetation anomalies, topographic effects or both.
The tensional faults are easer to be recognized on the satellite
image than the shear faults because they are open, wider easily
weathered and with vegetation (Travaglia, et al 2003). In the
present study the most of the faults were localized by vegetation
anomalies, except the faults E-W, which localized by the
supervised navigation (topographic effects-changing slopes).
The human activities are very eliminated in the study area of
marbles because of the elevation and the difficult access. Also
the communities are located at the borders of the marbles..
The rose diagram (Fig.6a) presents the distribution of the
directions of the lineaments (N20°, N50°, N90? and N120°-
N160°) that were mapped (Fig. 7). These lineaments have the
same directions with the faults (Fig.6,b) of the geological map
except in the direction E-W (Fig. 2).
The interpretation key of satellite data is to locate karstic
phenomena and compared with lineaments and that could give
precious indications on which fractures allow higher circulation
of water. The mapping of the sinkholes by the analysis of stereo
