The geographical situation between the Mediterranean Sea and Sahara makes Tunisia a dry country on the major part of its territory. This aridity, conjugated to the variability of the Mediterranean climate, makes the water a rare resource and unevenly distributed in time and space. Tunisia is classified by the international bodies as being among the least endowed countries in subterranean water resources in the pond of the Mediterranean (World Bank, FAO, OMS, UNESCO). The study of features is fundamental to the research in subterranean waters. Indeed, the main part of water resources is in fractured aquifers [1] -[5] .

The remote sensing thanks to its synoptic vision has been increasingly used for the mapping of the surface and for evaluation of natural resources. Landsat thematic Mapper data in digital format were preferred data due to the availability of seven bands ranging from visible to mid-infrared with 30 m spatial resolution, and one thermal band with 60 m spatial resolution; this permitted a large spectrum of band combinations, useful in visual and automatic interpretation of different features [1] [3] [6] -[10] .

The zone of study is situated in the region of Zahret Median in the North West from Tunisia (Figure 1). The morpho-structural plan in Zahret Median is characterized by impressive reliefs with abrupt slopes and narrow valleys. The sector of study contains two structural units interpreted by Rouvier [11] [12] : The “Kasseb unity” and the “Numidian unity”.

The Kasseb unity is constituted by three series: yprésien carbonated series, clayey lutétienne and of a detrital series of the lower Oligocene. The “numidian unity” is constituted by a material flyschoide silici-clastique surmounting by tangential contact all native and nonnative (native and allochthonous, autochthonous and nonnative) grounds Kasseb unity being in the South [11] [13] . The studied domain contains a dense river system and made an incision with permanent rivers and in numerous sources of water. The rivers main things are river Kasseb draining the pond overturning of the syncline of Kasseb and passing by of the NW towards it up to the plain of Majerda where it makes a junction with river Majerda and river Maaden, draining the valley of river Maaden and passing by known SW towards was born to the zone of Hédils. Of a lesser importance river el Blaaa, river Bou Naîl, river El Brik and river Snain who join in the syncline of Kasseb [11] [12] .

The datasets used include subsets, published geological, topographic maps of the region Zahret median (at scales 1:200,000 and 1/25,000) and Landsat 5 TM image. The Landsat 5 Thematic Mapper (TM) of 04 July 1985 (Figure 1), was downloaded free from the Global Land Cover Facility homepage

(http://earthexplorer.usgs.gov/). This period corresponds to the dry season which is characterized by absence of

clouds, what contributes to the good visibility of the sensors as regards Landsat. According to Youan Ta and al [14] [15] , these satellite images were chosen because of their spectral and spatial characteristics allowing a good structural small-scale mapping. The spatial resolution is 30 m for the reflective bands. These are band 1 (0.45 - 0.52 µm), band 2 (0.52 - 0.60 µm), band 3 (0.63 - 0.69 µm), band 4 (0.76 - 0.90 µm), band 5 (1.55 - 1.75 µm) and band 7 (2.08 - 2.35 µm). The spatial resolution of panchromatic (band 8) and thermal infrared (band 6 (10.4 - 12.5 µm)) bands are 14.25 m and 56 m respectively. The approximate scene size is 170 × 183 km.

Geological Structural investigation such as lineament mapping is normally undertaken based on geomorphological features, such as aligned ridges and valleys, displacement of ridge lines, scarp faces and river passages, straight drainage channel segments, pronounced breaks in crystalline rock masses and aligned surface depression [16] -[20] .

The first step of the methodology is the selection of initial input data for lineament extraction. A Digital image processing for the extraction of linear features involving contrast stretching, Image enhancement, Principal Component Analysis (PCA) were carried out on the satellite data using ENVI 4.7 software. Color composite of bands 1, 2 and 3 of the Landsat 5 Thematic Mapper images bands were made (Figure 2) and (Table 1).

A convolution filtering algorithm involving Directional Gradient-Sobel filters was applied on the enhanced Images in N-S, E-W, NE-SW and NW-SE directions to increase frequency and contrast of the images. Table 2 shows the directional filters in its four principal directions.

The automatic lineament extraction process was carried out with LINE module of PCI Geomatica V9.1 based on automatic detection algorithms [21] .

The analysis of the networks of major fractures of this region obtained after the diverse treatments of the image Landsat TM, proves the existence of the dominant orientations of lineaments. The lineaments were extracted from the Figures 3-6 and Figure 7. Table 3 shows the summary statistics of the extracted lineaments.

The lineament map (Figure 3) represents all the fractures carried out of the filtering the South (filter 0˚), it groups 9 lineaments with a total length of 18 km. The diagram of frequency associated with this map shows an ascendancy of the lineaments of NW-SE trend.

The second map of fracturing obtained by filtering SW shows a homogeneous distribution of lineaments. The statistical analysis of lineaments indicates a preferential direction NW-SE (Figure 4).

The lineament map and therose diagram W (Figure 5), indicates a heterogeneous lineament with an ascen- dancy of the lineaments of NW-SE trend.

The statistical analysis of lineaments obtained by filtering NW (filters 135 lineaments) indicates a preferential direction (NW-SE) (Figure 6).

The map of synthesis of lineaments (Figure 7) represents all the segments resulting from the overlapping of the information contained in four filtered images, it groups 40 lineaments, with a total length of 80 km. The diagram of the frequencies indicates two important families of orientations of lineaments: NE-SW and SW-NE.

The treatments of the satellite image, compared with the result of the geophysics studies realized in this sector, have been recorded. The results of treatment and the exploitation gravimetric realized by Fahem.k [22] proved the existence of two families of lineaments NE-SW and NW-SE. Fahem.k [22] indicated that the Numidien shows an evolution in the time and in the space and indicates that the direction NE-SW is the oldest and the direction NW-SE is more recent.

Comparing the lineament map with the fault lines digitalized from the published geological map (Figure 8), the result shows that the remote sensing approach produced more lineament. Possible reason for this could be the study area is composed of dense vegetation, which obscures the reflectance of the bedrock deformation on the imagery.

The comparison of the faults found in the lineaments map by automatic extraction with the faults of the geologic map of the zone of study asserts the existence of the direction NE-SW, but the direction NW-SE exist

more in lineaments map than the geologic map (Figure 8). This conflict is justified by Rouvier [11] which indicates that the zone of study knew a phase of Neogene comment coats and what the accidents of orientation NW-SE correspond to the fractures of extension of the anté-pliocene compression.

The remote detection constitutes a powerful tool for the researches concerning the prospecting of subterranean waters. It is a method of investigation adapted to the hydrogeologic prospecting of fractured aquifers. The treatment and the analysis of image Landsat 5 TM allowed map of fractures in the region of Zahret Median.

However, the analysis by remote detection did not allow specifying if the identified fractures were hydraulically active, that is if these fractures constituted reservoirs of subterranean waters. The map of fracturing obtained by satellite image must be coupled with the hydrogeological data through a SIG for a better exploitation of water resources. This approach permits to identify, with confidence, more zones favorable to the setting-up of the drillings for the supply of drinking water to the rural populations, even to large cities.

In the future, we intend to couple these results with the data hydrodynamics (debits, transmissivity, permeability, etc.) to optimize the hydrogeological studies and reduce considerably the rate of failure during the setting- up of the drillings. Then, this should allow approaching the problem of subterranean flows on the system fractured through a digital modeling.