Using data from the PALSAR L-band orbital radar of the Japanese ALOS satellite (Rosenqvist et al., 2007), we are conducting continental-scale mapping of several arid regions on Earth, with an initial objective to evaluate potentials of low frequency radar for planetary exploration (Paillou et al., 2006a). A mosaic of the whole Sahara has been built from PALSAR strips, allowing an efficient detection of subsurface geological features, particularly craters and paleodrainage networks (Paillou et al., 2003a, 2004, 2006b, 2009, 2010), because radar can probe through the first few meters of superficial eolian deposits (Baghdadi et al., 2005; Grandjean et al., 2006; McCauley et al., 1982; Paillou et al., 2003b; Schaber et al., 1986). We thus precisely mapped a 900 km-long paleodrainage system in eastern Libya, termed the Kufrah River (see Fig. 1), which could have linked the Kufrah Basin to the Mediterranean coast through Wadi Sahabi paleochannel in the Sirt Basin, possibly as far back as the Middle Miocene (Paillou et al., 2009). The headwaters of this paleodrainage system are mainly in southern Libya, with detected tributaries arising in three main areas: El Fayoud and El Akdamin hamadas in north-eastern Tibesti, northern Uweinat close to the Sudanese border, and the western Gilf Kebir and Abu Ras plateaux on the Egyptian border (Fig. 1). The Tibesti and Uweinat tributaries, more than 350 km-long, flowed in wide paleovalleys which join in the present-day Kufrah oasis. These paleovalleys had previously been detected by Robinson et al. (2006) using RADARSAT-1 C-band orbital radar. About 80 km north-east of the Kufrah oasis, a shorter (200 km-long) tributary joins from the Gilf Kebir and Abu Ras plateaux. From the Kufrah oasis, the main paleochannel becomes narrower (less than 1 km) and clearly incises the sandstone bedrock. It follows the present-day Wadi Blittah to the northern Jebel Dalmah over a distance of about 230 km. Farther north, in the Sarir Dalmah, the Kufrah River then disperses as a network of small, shallow paleochannels across the surface of a broad alluvial fan that covers more than 15 000 km² (Fig. 1). It is not possible to follow the paleodrainage course to the north, because the large sand dunes of the Calanscio Sand Sea preclude radar mapping of the subsurface. However, about 300 km away to the north-west and emerging from beneath the Calanscio Sand Sea, lies the major, 2 to 4 km-wide, alluvium-filled Wadi Sahabi paleochannel that incised more than 300 meters into Miocene carbonate strata (Barr and Walker, 1973). In a previous study (Paillou et al., 2009), we proposed that the sand sea could hide an ancient pathway between the Sarir Dalmah alluvial fan and the Wadi Sahabi paleochannel (cf. “north path” marked by the red dotted line on Fig. 1).

In addition to our PALSAR mosaic of the Sahara, we have used topographic information derived from SRTM data. The Shuttle Radar Topography Mission consisted of an interferometric radar system that flew on-board the Space Shuttle Endeavour during an 11-day mission in February 2000 (Farr et al., 2007). It produced a high-quality global Digital Elevation Model at a resolution of 3 arc-second (about 90 m), covering all land between latitudes 56°S and 60°N. SRTM data are organized in 1° × 1° cells, and have an absolute vertical height accuracy better than 15 meters. The HydroSHEDS (Hydrological data based on Shuttle Elevation Derivatives at multiple Scales) data set was computed from SRTM topography data by the USGS, and provides hydrographic information such as river networks, watershed boundaries and drainage directions (Lehner et al., 2008). Due to errors and voids in SRTM data, river network products are susceptible to various errors, particularly in very low relief regions. This is unfortunately the case in our region of interest, where low relief combines with numerous voids in SRTM coverage (mainly due to the radar wave penetration and attenuation in sand dunes), and so does not enable the reliable generation of river networks in HydroSHEDS. However, the computing of watershed boundaries is less sensitive to such problems, and watershed boundaries available in HydroSHEDS data can be useful for predicting areas to prospect for paleochannels using radar imagery. In particular, it appears that a large drainage basin exists west of the lower Kufrah River, arising in the Sarir Tibesti and flowing north in the Sarir Calanscio (Fig. 1): A paleodrainage contribution from the western topography (Al Haruj area and northern Tibesti mountains) should then be expected.

Combining topography derived from the SRTM data and the subsurface imaging capabilities of the PALSAR sensor, we mapped seven new paleochannels that are likely to have contributed water and sediment to the Kufrah River paleodrainage system from the west (Fig. 2). The northernmost four paleochannels probably originated from the Al Haruj relief, a Pliocene basaltic intracontinental volcanic field, and potentially connected to the northern Wadi Sahabi paleochannel. The remaining three paleochannels, which include the Wadi Behar Belama paleochannel previously mapped by Pachur and Altmann (2006), are found in the more southerly region of the Sarir Calanscio, located north-east of the Tibesti mountains. They end in the Calanscio Sand Sea, forming alluvial fans that are oriented and slope in the direction of the hypothesized but as yet unmapped 300 km-long “north path” of the Kufrah River (Fig. 1).

3.1 The northern Paleochannels

We mapped four main northern paleochannels (see Fig. 2), which are narrow, single and essentially straight and range from 110 to 190 km-long. They are associated with clear topographic depressions in SRTM data, ranging between 10 and 30 meters in depth, and the corresponding PALSAR images show well-defined and narrow (less than 400 m) dark channels. The paleochannels all appear to start on a limestone plateau that borders the volcanic relief of Al Haruj, a young (∼6 to 0.5 Ma) alkaline basaltic intracontinental volcanic field (Ade-Hall et al., 1974). Although we cannot map the paleochannels as far west as the Al Haruj relief because of sand dunes, they are very likely to have originated there. All the paleochannels would have flowed from the south-west to the north-east and appear to terminate about 30 km to the south of a southerly tributary of the Wadi Sahabi paleochannel, as shown on Fig. 3. Although we have no indication about the age of these paleochannels, it is very likely that their formation is related to the Pliocene volcanic events that led to the formation of the Al Haruj relief (Farahat et al., 2006).

3.2 The southern Paleochannels

We mapped three main southern paleochannels, which are also narrow (less than 200 m) and essentially straight, but longer (200 to 300 km) than the northern ones (Fig. 2). While they barely present a topographic signature in SRTM data, the PALSAR radar sensor clearly detects them. These paleochannels traverse the Sarir Calanscio plain, which descends north-eastwards from the Tibesti mountains (Fig. 2). The three southern paleochannels appear to emerge from the dune fields that cover the Sarir Tibesti: Potentially, they could be extended further southwest towards the Tibesti mountains, but the PALSAR sensor is not able to detect and map them through the thick sand cover. The most southerly paleochannel was previously mapped under the name of Wadi Behar Belama (Fig. 2) by Pachur and Altmann (2006) using LANDSAT-TM images. As with the northernmost paleochannels, the southernmost ones follow a south-west to north-east direction. They terminate at the western margin of the Calanscio Sand Sea, dispersing as networks of small, shallow paleochannels across the surface of alluvial fans (Fig. 4), indicating a decrease in paleoflow competence. Coarse alluvial sand and gravel constituting the fan sediments increase the surface roughness and volume scattering effects, so that the southern paleochannel terminations appear as bright features in the radar images, as shown on Fig. 4.

The seven paleochannels that we have mapped are consistent with the hypothesis of an extensive Kufrah River paleodrainage system that connected the northern Tibesti mountains to the Sirt Basin during the Late Cenozoic as previously proposed by Pachur et al. (Pachur, 1980, 1996; Pachur and Altmann, 2006; Pachur and Hoelzmann, 2000). Using LANDSAT-TM imagery, Pachur and colleagues detected remnants of a paleodrainage system formed by the junction of former great wadis originating in the northern Tibesti mountains, that crossed the Sarir Calanscio in a north-east direction. A main representative of this system is Wadi Behar Belama which contains acid volcanic pebbles originating from the Tibesti. Pachur and colleagues also mapped the upper reaches of a paleoriver system sourced in the Al Haruj volcanic relief, that “probably joined drainage systems that flowed into the Mediterranean” (Pachur, 1996), and suggested that this system is likely to have been active during the Early Holocene. The new paleochannels that we have mapped are also consistent with the location of the “humid corridor” proposed by Rohling et al. (2002) and Osborne et al. (2008), that was connecting southeastern Libya to the Mediterranean Sea at around 120 ka. Osborne et al. (2008) performed a Nd isotopic characterization of Quaternary sediments sampled in Wadi Behar Belama and used the findings to support the interpretation of an uninterrupted sediment transport pathway from the Tibesti mountains to the Mediterranean Sea. The alluvial fans at the termini of Wadi Behar Belama and of the two other southern paleochannels (Fig. 2), might then be transitional features produced during drier periods, when the paleoflow competence decreased and sediment throughput could not be maintained. These paleochannels have certainly contributed to sediment supply of the Calanscio Sand Sea (El Baz et al., 2000; Ghoneim et al., 2012) and so may have also contributed to burial of the through-going “north path” of the main Kufrah River that we hypothesized (Paillou et al., 2009).

While full resolution SRTM topography contains very detailed features, particularly dunes, that preclude detailed mapping of the main paleochannels, one can consider the low-frequency topographic information in order to better infer paleodrainage directions. Craddock et al. (2010) applied this approach to reconstruct a buried fluvial landscape beneath the Simpson Desert eolian dune field in central Australia: while many dunes screen the main drainage directions in full resolution SRTM data, sampling of the low-frequency topography in the topographic lows between the dunes revealed slope trends that trace remnants of paleodrainages. We applied a similar approach here by low-pass filtering the SRTM data in the frequency domain, after a Fourier transform, and by using interpolation techniques to fill SRTM voids: Fig. 5a shows the full resolution SRTM topography with voids filled by interpolation, and Fig. 5b shows the low-frequency topography obtained after low-pass filtering. Applying a simple threshold to the low-frequency topography allows to directly map the local depressions, represented as light blue areas in Fig. 5c. There is no clear evidence of a topographic depression that would extend the alluvial fan terminating the Kufrah River to the northwest: The hypothetical “north path” connecting the Sarir Dalmah to the Wadi Sahabi paleochannel as indicated on Fig. 1 cannot be observed on present-day topography (but still could be buried under the dunes of the Calanscio Sand Sea). However, a local depression (path #1 on Fig. 5c) seems to connect the Kufrah River alluvial fan to a western corridor (path #2 in Fig. 5c). The latter runs from the northern Tibesti, through the Wadi Behar Belama in the Sarir Calanscio, and ends south of the Wadi Sahabi paleochannel (see Fig. 5c). This large paleocorridor, 10 to 20 km wide and about 400 km long, is close to the terminating path of the lower Sahabi River as proposed by Griffin and Drake (see Fig. 1) but again, as stated in the introduction of this study, this cannot be considered as a proof for a continuous river system connecting the Chad basin to the Mediterranean Sea. The observed paleocorricor is more likely to have been fed by the northern Tibesti and Al Haruj reliefs, being an ancient pathway between the paleochannels described in the previous section and the northern Wadi Sahabi paleochannel (Fig. 5c). This paleocorridor, as a depression adjacent to the margins of the aggraded Sarir Dalmah alluvial fan, was possibly re-enforced by local tectonics, since its orientation is comparable to the orientation of the main faults in this area (Ahlbrandt, 2001). In particular, one can observe on Fig. 3 that the course of two of the northernmost paleochannels shows a change in the paleoflow direction of about 90° when approaching the paleocorridor from the southwest, suggesting a possible tectonic control by the horst and graben structures in the region (Ahlbrandt, 2001).

The newly mapped paleochannels and the analysis of the local topography thus strongly support the hypotheses proposed by Pachur (1980, 1996), Rohling et al. (2002) and Osborne et al. (2008) of formerly continuous connections between drainage basins in northern Tibesti and the Mediterranean Sea through the Wadi Sahabi paleochannel. Our findings also support our previous hypothesis that the Sarir Dalmah alluvial fan, located at the terminus of the upper part of the Kufrah River system, could have connected to the Wadi Sahabi paleochannel in the north (Paillou et al., 2009). This connection is, however, not yet obvious, and the Kufrah River is also likely to have terminated as in inland delta in the Sarir Dalmah (Paillou et al., 2009), possibly feeding the northern Al Jaghbub paleolake as proposed by Ghoneim et al. (2012). This re-emphasizes the importance of undertaking some exploratory field work in the Calanscio Sand Sea, using geophysical prospecting techniques such as Ground Penetrating Radar, to detect and map possible paleochannels buried under the sand dunes. The Calanscio Sand Sea is a key area for understanding the history of paleodrainage systems of eastern Libya, and near surface geophysical prospecting could provide new data to help answer outstanding questions regarding the connection of the paleochannels in this area. Exploratory field work is also needed in order to collect samples for geochronological and geochemical analyses that would help determine the ages and histories of the various paleochannels that we have mapped so far: The many tributaries of the Kufrah River system may not necessarily have the same antiquity, some of them possibly being younger impositions on an older paleodrainage network (in particular, the northern paleochannels related to the young volcanic relief of Al Haruj are certainly more recent than the southern ones).

With the plausible assumption that all the seven newly mapped paleochannels at one time were forming part of the Kufrah River paleodrainage system, we can now define a major paleodrainage system in eastern Libya, which at its maximum extent would have drained an area of more than 400 000 km² between the Tibesti, Al Haruj and Gilf Kebir massifs and probably connected to the Mediterranean Sea through the Wadi Sahabi paleochannel in the Sirt Basin. The whole system is likely to have been active at intervals during the Late Cenozoic, possibly discharging at that time a comparable amount of water as does the present-day Nile into the Mediterranean Sea. The Kufrah River system is then clearly a major paleohydrological feature to take into account when studying the past environments and climates of northern Africa from the Middle Miocene to the Holocene. It also represents a likely corridor for fauna and human dispersal in the eastern Sahara, and thus indicates locations where further paleontological, paleo-anthropological and archeological field exploration should be conducted.