200 
aol Fu 
Of 
  
  
— 
T 
1 
  
ooat ti 
d- C7 
aZ Vi 
orti 
00 
  
4 
ec 
DOIL 
00a 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
When the eight image endmembers were processed via the 
SAM algorithm, excellent discrimination between the different 
endmembers was found. The TM results are relatively accurate 
with the geological map shown in Figure l. First of all, the 
basic volcanic rocks endmember (figure 4a), show the major 
basic rocks of the study area such as 4 — K*' Kettera, B — S! 
Mahjoub, C — K* Ahril, D — Gour es Sefra, E — J.Hadit, F — 
Menaa el Kahla and G — J. Ahmar. The acidic volcanic rocks 
endmember also demonstrate good results (figure 4b), as from 
the identification of the major acidic rocks, such as H — K? 
Dalaa, 1 — K" Mirouga, J — K* Aicha, K — K* Hamra and L - 
K" Bouzelfar. The iron cap endmember gave good results with 
the identification of areas where they are known to occur. The 
iron caps were harder to discriminate with SAM because they 
are relatively narrow in thickness (few meters), but they usually 
extend for a few kilometers long. For TM, which has a 
medium spatial resolution, and SAM, which does not consider 
the sub-pixel, the mixture problem for this unit can cause 
problems. Also, there is a similarity between the iron caps and 
the basic intrusions because they are both rich in 
ferromagnesian minerals, thus their altered surfaces (patina) are 
similar due to the dark reddish color caused by the alteration of 
iron oxides and hydroxides. 
As shown from the geological map in figure 1, the schist of 
Sahrlef unit covers most of Central Jebilet. However, when 
comparing with field observations, a thin layer of Quaternary 
deposits, which range from conglomerates to rock pebbles and 
sandy loams, covers most of Central Jebilet. The SAM 
classification map for TM (figure 5d) clearly shows a higher 
distribution of Quaternary deposits relative to the schist unit. 
Because the carbonate endmember was selected in the TM 
image where the unit was observed in the field and is not 
shown on geological maps, the output accuracy of this 
endmember would have to be verified in the field. Finally, the 
vegetation endmember gave an excellent result due to the semi- 
arid environment of the study site and sparse vegetation cover, 
which is very contrast with rocks and bare soils. 
3.2 SAM for Quickbird 
The SAM results for Quickbird (figure 5) show the SWIR 
importance for the identification and differentiation of different 
rock units. The basic and acidic intrusions and iron caps where 
not clearly recognized with the Quickbird image, as they were 
with TM. This part of the spectrum is extremely important for 
mineral and rock discrimination since they have relatively high 
reflectance and numerous absorption bands caused by minerals. 
Quickbird did however identify zones where basic intrusions 
occur such as 4 — K*' Kettera, C — K" Ahril, D — Gour es Sefra, 
E - J.Hadit and F — Menaa el Kahla and where acidic intrusions 
occur such as H — K* Dalaa, J — K* Aicha, K — K* Hamra and 
L - K* Bouzelfar as well as the iron cap of Kettera Mine. Like 
TM, the Quaternary deposits and sandy loams were widely 
spread over the study area compared to the schist unit. 
Vegetation is the endmember that gave the best results because 
of the near infrared band, which has a high reflectance for 
vegetation. The overall results for SAM were scattered and 
most of the geology endmembers were misclassified due to the 
low spectral dimensions of Quickbird. 
603 
  
Output Legend 
HUnclassified 
Basic intrusions 
Iron caps 
M Acidic intrusions 
Vegetation 
Quaternary deposits 
Sandy loams 
Schist 
  
  
Figure 5 — Output classification map for Quickbird. 
4. Conclusions 
In comparing between high and medium spatial resolutions for 
the geological mapping of Central Jebilet using SAM 
algorithm, we have come to the following conclusions. Even if 
TM has a medium spatial resolution and that sub-pixel 
contamination from different rocks or cover material is evident 
while selecting endmembers, it has given good results. On the 
other hand, Quickbird, which has a fine spatial resolution, gave 
inadequate results for the surface geology mapping of Central 
Jebilet. As discussed above, this is caused by the absence of 
spectral bands in the SWIR, which is the key for mineral 
exploration with satellite images. The classification map 
generated with SAM for TM show that this method could 
effectively be used for geological mapping and potentially be 
used for exploration in unexplored areas. Whereas Quickbird 
images would only be helpful for visualizing the study area 
with great details, locate outcrops in the field and facilitate 
structural mapping. This demonstrates that for geological 
mapping, it is much more important to have a high spectral 
resolution rather than a high spatial resolution. Recent advances 
in optical remote sensing sensor technology have led to the 
development of hyperspectral sensors, which acquire images 
data in many narrow, contiguous spectral bands. Accordingly, 
each pixel possesses a detailed spectral signature, permitting a 
more thorough examination than provided by multispectral 
scanners collecting in a few broad and noncontiguous bands. 
Certainly, this new technology would be very valuable for 
geological and mineralogical mapping.