February 14, 2007

The Late Neogene Sahabi Rivers of the Sahara and Their Climatic and Environmental Implications for the Chad Basin

By Griffin, David L


This paper is based on the concept that the drainage system of c. 7.5-4.6 Ma is still reflected in the geomorphology of northeastern Chad and adjacent areas of Libya. During the Messinian and early Pliocene a large lake was present in the Chad Basin that is termed Neogene Lake Chad. It fluctuated in size in response to the precessional cycle and at times overflowed to the east, NE or north, giving rise to the Sahabi rivers. The Eosahabi flowed during the drawdown of the Mediterranean (late Messinian) and eroded the Erdi and part of the East Tibesti Valley. The post-rift Miocene deposits of the Chad Basin, some several hundred metres in thickness, record a transgression over an irregular and faulted terrain with the deposition of coarse and fine clastic material. Fluviatile and lacustrine environments are represented. At least part of the Miocene succession belongs to a phase of late Miocene lake development. Fluctuating climate during the Messinian and early Pliocene led to repeated changes in the environment of the Chad and Eosahabi Basins with over 100 climatic cycles developed. This would favour the development of animal species with high adaptability, for example to littoral, riparian, woodland and savanna habitats.

An examination of the depositional profile of the Miocene-Plio- Pleistocene succession in the offshore Red Sea area led Griffin (1999) to the conclusion that the Messinian sedimentary rocks (Zeit Formation and equivalent) were deposited at a rate greater than other units of the Miocene-Plio-Pleistocene of the same area. This and other aspects of the Zeit Formation were taken to indicate that the Messinian was a time of high rainfall and high sediment yields. This period, termed the Zeit Wet Phase, stands in marked contrast to the arid conditions of the preceding Tortonian Stage, which is represented in the Red Sea area by the halite-bearing South Gharib Formation. The dry period during which the latter formation was deposited was subsequently referred to as the Tortonian Dry Phase by Griffin (2002).

The drainage basins of North Africa were examined by Griffin (2002) and it was demonstrated that evidence exists for rivers originating from a lake in the Chad Basin flowing via the East Tibesti Valley (Fig. 1) to eventually reach the Mediterranean in the Gulf of Sirt. The lake was termed Messinian Lake Chad. Two phases of the rivers can be dated using evidence from the area of the Gulf of Sirt. The early Pliocene Sahabi rivers were recognized on the basis of the Sahabi Formation (Fig. 1), a sequence of deposits of early Pliocene age that record marine, lagoonal, deltaic and fluviatile environments (Boaz et al. 1987; de Heinzelin & ElArnauti 1987). The Sahabi rivers were defined by Griffin (2002) as those of Pliocene age that followed the late Messinian drawdown of the Mediterranean. The late Messinian Eosahabi rivers were recognized on the basis of the Eosahabi Channel (Fig. 1), a channel some 396 m deep and c. 1-5 km wide cut into Miocene limestones near the Gulf of Sirt. Barr & Walker (1973) interpreted this channel as having been cut by a river flowing at the time of the drawdown of the Mediterranean, although they did not trace the source of this river. The Eosahabi rivers were defined by Griffin (2002) as the rivers of the drawdown phase of the Mediterranean. The broad path in which the Sahabi and Eosahabi rivers flowed is shown in Figure 1. The Sahabi and Eosahabi rivers were interpreted by Griffin (2002) to be a product of part of the Zeit Wet Phase. The latter was dated using the dust flux curve of deMenocal & Bloemendal (1995) as covering the period 7.5-4.6 Ma.

As noted, the term Messinian Lake Chad was used by Griffin (2002) for the late Miocene lake of the Chad Basin. Because the history of the lake covered the period from early Messinian to the early Pliocene (possibly including part of the late Tortonian) confusion can arise in using this term in detailed analysis. It is therefore proposed to name this lake Neogene Lake Chad, a name that will refer to any lake that developed in the Chad Basin during this time period.

The magnitude of the river erosion recorded in the East Tibesti Valley (Griffin 2002) indicates that there is a broader history to investigate involving the lake from which the rivers originated and the rivers themselves. This paper pursues these lines of investigation, examining first the Chad Basin and Neogene Lake Chad and then considering the Sahabi rivers that the lake sourced from time to time.

Two main categories of previous work relating to the palaeodrainage of Libya and areas to the south and east are as follows. First, there is the effort that has concentrated on the Quaternary, recent examples being the studies by Pachur & Altmann (1997) and Pachur & Hoelzmann (2000). Williams & Faure (1980) in a wide-ranging work edited a synthesis of then current research relating to the Quaternary geomorphology, climate, environments and archaeology of the Sahara. The geological history of the Nile and Nile Basin was also considered. Second, there is the work concentrating mainly on Egypt and northern Sudan that includes the study of the so-called Radar Rivers. This work was summarized by McCauley et al. (1998). Spaceborne imaging radar (SIR) missions were undertaken by NASA in 1981, 1984 and 1994 that covered parts of southern Egypt and northern Sudan. Interpretation of the data revealed a variety of drainage and channel segments overlain by thin to discontinuous aeolian sand sheet deposits. When less than 1 -1.5 m in thickness these deposits are nearly transparent to SIR signals. McCauley et al. (1998) recognized six main types of drainage features, reaching in length up to several hundred kilometres and varying from several hundred metres to 30 or 40 km in width. Drainage types included broad aggraded valleys to more limited areas with palaeodrainage of dendritic or anastamosing form. These features can be difficult to date and show a variety of flow directions, where these can be determined. They can be seen to form a mosaic of fragments of past drainage systems, which McCauley et al. (1998) interpreted to cover a history of 30 Ma or more.

Issawi & McCauley (1992) interpreted the Tertiary drainage history of Egypt and considered that some of the SW-oriented channels recognized from the 1981 and 1984 SIR missions are remnants of a trans-African drainage system of late Oligocene to Miocene age. It originated on the uplifted edge of the Red Sea rift and flowed to the SW to eventually reach the Atlantic in the Gulf of Guinea. Issawi & McCauley named the system the Qena drainage system. It is noted that this system must largely predate the mid-Miocene formation of the Chad Basin as it is expressed today, the formation of which would block the SW-flowing rivers.

Methods, map sources and satellite imagery

The study of the late Miocene climate of North Africa and the Mediterranean by Griffin (2002) was based on evidence from the sedimentary sequence and fluvial geomorphology. In the current study a major focus is the geomorphology of northeastern Chad and the degree of preservation of evidence for drainage systems of between 8 and 4 Ma. The physiography of the area of interest was examined using Landsat images and topographic maps. The Landsat 4/5 TM scenes used were (Path/Row) 182/046, 182/047, 181/047 and 181/048. The scenes were processed by Geoimage of Perth, Western Australia, to produce a standard 321/RGB composite image, approaching natural colour. The area covered by the composite image is approximately 1730'N-21N and 19E-2230'E. For regional context use was also made of the MrSID Landsat composites (MrSID 2002) available at a NASA website. Russian military maps (Soviet Army General Staff 1954a- 1992a, 1954b-1992b) were the main source of topographic information used in the study. These are issued at a scale of 1:200 000 contoured at 40 m intervals and at a scale of 1:500000 contoured at 50m intervals. Additional information was provided by the map of Chad (Institut Gographique National 1968-1974) and Carte Internationale du Monde, Largeau Sheet, NE-34 (Institut Gographique National 1975). The latter map was used as a guide to nomenclature.

The term Sahabi rivers will be used to refer to the family of rivers that exited from Neogene Lake Chad in northeastern Chad during the history of the lake. More specifically, the Sahabi is used for the river of the postdrawdown Mediterranean, the term Eosahabi for the drawdown phase of the river (c. 5.8-5.3 Ma) and the Palaeosahabi for the early phase of the river (early to mid- Messinian). Part of this study is essentially geomorphological in nature and age assignments for the Sahabi and Eosahabi must depend on data from northern Libya, as noted above. Other dating is by inference using all known relevant information. Even if the dating of the older phases of the river is subsequently revised, it can be argued that the sequence of events presented here is sound. The earliest phase of the river, the Palaeosahabi, is poorly preserved and is therefore the most difficult river to reconstruct.

Geological considerations

The Neogene geology of the Chad Basin is discussed in detail in the following section. The geology of \the other main basins within the Sahabi river system is here briefly summarized. The Kufra Basin (Fig. 1) lies mainly in Libya; that part in northeastern Chad is also known as the Erdis Basin. The sediments of the Kufra Basin are mainly continental sandstones of CambroOrdovician to Early Cretaceous age (Hesse et al. 1987). A marine intercalation predominates in the Lower Silurian sequence (the Tanezzuft shale). A marine intercalation also occurs in the Carboniferous, with marine influences increasing to the south and SW. The Palaeozoic sequence has a thickness of 1300-2000m; Lower Cretaceous sediments are some 10001500 m in thickness and form the surface of most of the basin. Klitzsch (1984) studied the full sequence to the east in Nubia and noted that the typical sediment is a sandstone redeposited several times, in different environments. Almost none of these deposits are windblown sediment.

The Sirt Basin (Fig. 1) is an embayment that opens out northwards into the Mediterranean Basin. The floor of the Sirt Basin is an unconformity above which is a thick sequence of Lower Cretaceous to Recent sediments (Selley 1997). The Miocene Marada Formation, consisting in part of skeletal limestone, is the only part of the Sirt Basin sequence to occur at the surface. The floor of the basin lies at a depth in excess of 5 km.

The late Neogene Chad Basin and Neogene Lake Chad

The northern and southern Chad Basins; the preceding Cretaceous and Palaeogene rifts

The outcrop area of Tertiary and Quaternary deposits in the south- central Sahara reflects the form of the Chad Basin (Fig. 2). The outcrop area in Niger and Chad (and small parts of Nigeria, Cameroon and the Central African Republic) is composed substantially of Quaternary rocks and covers about 1.5 10^sup 6^ km^sup 2^ (Choubert & Faure-Muret 1987). The area resembles a rhombus in shape, with sides of c. 1250 km. The west side of the rhombus is displaced to the north when compared with the east side. The topographic basin is somewhat larger and covers an area of c. 2.3 10^sup 6^ km^sup 2^. Neogene Lake Chad at its maximum size (assumed to be defined by the 400 m contour) lay mainly within Chad (Griffin 2002) and covered about a third of the basin, 700 000 km2 at its maximum size. Outcrop areas of prePliocene Tertiary deposits referred to the Continental Terminal (Wolff 1964) occur within Chad on the north and south flanks of the basin (Fig. 2). The latter term is widely used in West Africa, is not age specific and can include deposits of late Eocene to mid-Pliocene age. Several small outcrop areas of Continental Terminal occur in eastern Niger (Faure 1966).

The Chad Basin is divided into the northern and southern Chad Basins (Chienne et al. 2002; Fig. 2). The northern Chad Basin lies to the NE of Lake Chad and is an area of shallow basement that has only relatively recently (in general terms since the mid-Miocene) become a site of sedimentary deposition following Cretaceous and Palaeogene rift deposition to the south. The NE-trending Bahr el Ghazal and the Bodl Depression are distinctive features. The former is an essentially dry river valley leading from the vicinity of the present Lake Chad to the Bodl Depression some 300 km to the NE. The latter is the lowest part of the Chad Basin, also called the Pays- Bas, with elevations of less than 200 m. The southern Chad Basin is an area of deeper basement and overlies the area of convergence of the following rift basins: the NNE-trending Bornu Basin of northeastern Nigeria, the NW-trending Termit Basin, which extends northwestwards from Lake Chad, and the NE-trending Doha, Doseo and Salamat Basins, which form a linear trend straddling the Chad- Central African Republic border (Figs 2 and 3). The latter three basins and the Bongor Basin will be referred to as the central African subsystem basins, following Genik (1993). The general area in which the three described trends converge is northeastern Nigeria and the drainage basins of the Chari and Logone Rivers (Fig. 2). The rifts share a common history commencing in the Early Cretaceous.

The Termit Basin and central African subsystem basins are Cretaceous-Palaeogene rifts that were probably caused by the stretching and subsidence of African crustal blocks during the breakup of Gondwana in the early Cretaceous and by reactivated plate movements in the early Tertiary (Genik 1993). Genik recognized three phases of rifting (in general, Early Cretaceous, Late Cretaceous and Palaeogene) which terminated in the Late Oligocene with the development of an unconformity. In the midTertiary epeirogenic movements involving rim uplift and basin subsidence (Burke 1976) established the present Chad Basin (Guiraud & Borel 2001; Eocene and Miocene maps). Wilson & Guiraud (1998) noted that Cenozoic magmatism in North Africa was relatively rare before 30 Ma. However, major phases of volcanism occur in the early and late Miocene. They noted that domally uplifted areas within the African plate (Hoggar, Tibesti, Jebel Marra in Darfur and Adamawa in Cameroon; Figs 1 and 2) were volcanically active at this time and may be plume related. It can be reasonably assumed that the Ennedi and Erdi underwent uplift as part of this tectonic cycle.

Rift deposition in the Termit Basin alternated between alluvial, fluviatile, lacustrine and marine, whereas deposition in the central African subsystem basins was mainly fluvial with some alluvial and lacustrine deposition and minor marine deposition in the west. The post-rift phase of deposition extends beyond the rifted area to include the northern Chad Basin and can be divided into two parts. The lower, essentially Miocene, sequence shows the development over a faulted terrain of widespread fluvial and lacustrine deposits of variable thickness. The sediments include medium to coarse clastic deposits, sometimes immature and often showing post-depositional alteration (Schneider & Wolff 1992). The succeeding sequence includes Pliocene sediments, which are mainly fine grained and were deposited in a persistent and apparently slowly contracting lake.

Miocene deposits of the Chad Basin: fluviatile, perilacustrine and lake sediments

Under this heading three areas are considered: the north (the area of Faya and the Bodl Depression), the vicinity of Lake Chad, and the south (the drainage basins of the Chari and Logone Rivers) (Fig. 2). The Neogene stratigraphie succession of the northern Chad Basin is shown in Figure 4. The following discussion concentrates on the Miocene part of the succession: the Chad Baguirmi Series of the Lake Chad area, the Bodl Series in the north (Borkou area) and the Palaeochadian Series of the central African subsystem basins in the south. All these units were referred to as Continental Terminal by Schneider & Wolff (1992).

In the area extending from the southeastern part of Lake Chad to some 200km to the south, Schneider & Wolff (1992) recognized four assemblages in the unit that they termed the Continental Terminal. The unit is probably equivalent to the Chari Baguirmi Series (Servant & Servant-Vildary 1980) of this same area. The lithological assemblages are as follows (starting at the base).

(1) Electrically conductive argillaceous deposits thickening to the north of an east-west basement axis at the latitude of Bongor (1017'N). The deposits rest on basement and towards the north are transgressive on Cretaceous units. They reach 200 m in thickness.

(2) An electrically resistive primarily detrital assemblage composed of more or less argillaceous sandstones with some argillaceous intercalations. Some 150 km south of Lake Chad its thickness exceeds 150 m.

(3) A conductive argillaceous assemblage composed of grey, ochre and variegated clays, moderately sandy, showing some thin sandy intercalations. The assemblage rests on Cretaceous units in the northern part of the lake. It is about 100 m in thickness.

(4) An assemblage of moderately argillaceous sandstones some 75 m in thickness is developed beneath the southern end of Lake Chad.

This sequence appears to correlate with the Bodl Series of the northern Chad Basin. It will be noted in the Conclusions that at least part of the Bodl Series is of late Miocene age. Some 50-70 m of Lower Pliocene sands and clays overlie the Continental Terminal, with the Lower Pliocene over-stepping lower units of the Continental Terminal south from Lake Chad.

The Cretaceous-Tertiary succession of the central African subsystem basins differs from that of the Termit Basin (Fig. 3) in being of dominantly continental origin. Also, the Palaeogene sequence is thin in the central African subsystem basins whereas in the Termit Basin it thickens to several thousand metres to the north of Lake Chad. The central African subsystem basins were emergent or nearly so during this phase, subsiding no more than 200-300 m (Genik 1993). Basin profiles of the central African subsystem basins based on drilled wells, reflection seismic, magnetic and gravity data demonstrate that the post-rift sequence is consistently developed. It is less than 1000 m in thickness (Genik 1993) and probably averages half this thickness or less. In the western half of the central African subsystem trend is an area of about 100000 km^sup 2^ mapped by Wolff (1964) as Continental Terminal (Fig. 2). The area of Continental Terminal extends well beyond the rifts and so is part of the post-rift sequence, but pre-Pliocene. As a consequence of post- Miocene uplift these rocks now lie in the drainage basins of the Chari and Logone Rivers (Fig. 2). During the Quaternary, and possibly the Pliocene, these deposits have contributed sediments to the Chad Basin.

Wacrenier (1961) provided a description of some of these Neogene rocks in connection with an investigation of bauxite deposits of the Doba Trough (Fig. 2). He referred this Tertiary pre-Pliocene sequence to the Series Palotchadiennes (Palaeochadian Series) and d\ivided it into upper, middle and lower parts. Based on proximity and stratigraphie position the Palaeochadian Series can be correlated with the Chari Baguirmi Series; the former probably includes some Palaeogene at the base. Pias (1970) described this sequence (300-700 m in thickness) under the term Continental Terminal and summarized its characteristics over a wide area in southern Chad. The mudstones of the sequence and their variants contain fluctuating amounts of fine sand and appear to have a lacustrine origin. The sandstones, arkoses and sands are of the fluviatile type, little altered in the vicinity of the massifs, whereas to the north subangular sands suggest beach development. The alternation of coarse sandstones or fine sands and mudstones implies several cycles of sedimentation during the deposition of the Palaeochadian Series. Arkosic sediments were deposited in the vicinity of the massifs; beds of white clay occur derived from the decomposition of these feldspathic rocks. After deposition the sediments underwent various stages of pedogenesis including ferruginization, ferralitization and the formation of ferruginous crusts and 'sols rouges' (Pias 1970).

To the north and NE of the Bodl Depression (Pays Bas; see Fig. 2) Tertiary and Quaternary deposits lie unconformably on Palaeozoic sandstones. To the east of the northern part of the Bahr el Ghazal (Fig. 2) they lie directly on crystalline basement; a well has penetrated them in this region revealing a thickness of 525 m (Servant 1983). The mainly Miocene Bodl Series (Fig. 4) to the north of the Bodl Depression was mapped under the term Continental Terminal by Wolff (1964). The area so mapped crops out in an east- west elongated area of about 45 000 km^sup 2^ to the east of Faya. Some 70 km to the SE of Faya seismic information indicates that the Continental Terminal rests on eruptive basement and has an estimated thickness of 250 m; it increases in thickness to the SE to reach 350 m (Schneider & Wolff 1992). The Bodl Series comprises clays, sometimes finely bedded, alternating in an irregular way with fluviatile sands; lenses of gravel and of argillaceous sandstone are also present. Cinerites and volcanic tuffs are widely developed to the SE of Angamma Cliff (Fig. 2).

A mandible of Merycopotamus collected in the Bodl Depression allows at least a part of the Bodl Series to be dated as Late Miocene or Early Pliocene (Servant 1983). On the basis of lithological correlation, Servant suggested that the base of the Bodl Series could be of Oligocene age. He considered the Bodl Series to correspond to fluviatile outpourings interrupted at times by processes of soil development. Fossil wood has been described from the Continental Terminal of the Borkou (Coppens & Koeniguer 1976), and Schneider & Wolff (1992) have drawn attention to two sites, one close to Kirdimi (65 km WNW of Faya) and the other at Ou-Ou (77 km south of Ounianga Kbir; see Fig. 2). The two sites have provided species of Caesalpiniaceae, in particular Afzelioxylon furoni (Kirdimi) and A. tchadense (Ou-Ou). Afzelioxylon is close to the current African genus Afzelia that is found in abundance in wooded savannas, gallery forests and dense forests (Coppens & Koeniguer 1976).

The Sahabi rivers, Neogene Lake Chad and Messinian cyclicity

The scale of the erosional features discussed in the next two sections suggests a long history for the rivers and the lake that sourced them. The reasonable assumption is made that the lake lasted as long as or longer than the Sahabi rivers. The rivers are interpreted to have terminated with the end of the Zeit Wet Phase (4.6 Ma) and it will be suggested that they originated close to the beginning of the Messinian (see Synthesis section; The Palaeosahabi). On this basis a life span of at least 2 Ma will be attributed to the Sahabi rivers. Accepting the existence of a long- lasting lake in the Chad Basin during the Zeit Wet Phase a key question is whether it was continuous or episodic. Several grounds exist for thinking that the lake varied considerably in size in accordance with the precessional cycle, as there is much evidence to indicate that the Messinian was a time of precession-induced depositional cyclicity. For example, Decima & Wezel (1973), Cita & McKenzie (1986) and Keogh & Butler (1999) described and/or made reference to the cyclical nature of the Messinian Upper Evaporites of the Mediterranean. Further, high variability and strong 19 and 23 ka cycles occur in the Messinian and early Pliocene dust flux record at Arabian Sea Ocean Drilling Program (ODP) Sites 721/22 (deMenocal & Bloemendal 1995). Looking at the broader picture, Krijgsman et al. (1999) concluded that the cyclic evaporite deposition of the Messinian Mediterranean is almost entirely related to circum- Mediterranean climate changes driven by changes in the Earth's precession. Other studies (Hilgen et al. 1995; Sprovieri et al. 1996; Sierra et al. 1999) have revealed that the sedimentary cycles of the Messinian pre-evaporites are dominantly controlled by precession-induced changes in the circum-Mediterranean climate. These considerations lead to the view that it is very likely that the climate of the Chad Basin during the Messinian was cyclic and that this will be reflected in the sedimentary record of the basin.

Probable Messinian and early Pliocene cyclicity recently described from the northern Chad Basin

The work of the Mission Paloanthropologique Franco-Tchadienne (MPFT) is beginning to make available some detailed accounts of the Messinian and early Pliocene succession in the northern Chad Basin. Brunet & MPFT (2002) and Vignaud & MPFT (2002) described the palaeontology and geology at Toros-Menalla (Fig. 2), a Miocene (c. 6.5 Ma) hominid-bearing site some 330 km to the SW of the southern end of the Amatinga Valley. At this site at least 4 m of desert sandstone pass upwards into c. 2m of perilacustrine sandstone with a rich vertebrate fauna. The upper part of the section consists of lacustrine clay containing the fossil remains of approximately 10 taxa of freshwater fish. An erosion surface separates the perilacustrine and lacustrine deposits. The lake deposits comprise some 0.5 m of green pelite; their original thickness at the location is unknown. The fauna suggests a biochronological age of between 6 and 7 Ma. The faunal assemblage from Toros-Menalla is compatible with a diversity of habitats including grassland, wooded savanna, fresh water and probably some gallery forest.

The Kossom Bougoudi site (Brunet & MPFT 2000) of the northern Chad Basin lies some 125 km to the east of the TorosMenalla site (Fig. 2) and is dated at c. 5 Ma. The succession at Kossom Bougoudi consists of 3 m of green sandstone, overlain by 1.2 m of green silly clay, in turn overlain by 0.6 m of green sandstone. The sequence is either covered by or episodically interstratified with white diatomites (0.2 m). The sandstones coniain a rich vertebrate fauna including reptiles, birds and especially mammals. The diatomites and clays contain fossil fish. Perennial aquatic conditions are indicated by the occurrence of liana-like leguminous plants and by a fauna with a high biodiversity of fish, some of which usually live in oxygenated water of some depth. The fauna also contains aquatic birds, reptiles and large numbers of specimens of a hippopotamid. In addition to the aquatic habitat, the fauna at Kossom Bougoudi indicates wooded and more open habitats, including grassland. The sequences at Toros-Menalla and Kossom Bougoudi both record lake and near-lake sedimentation. In fact, these sequences could be seen as parts of sedimentary cycles, recording a period of lake growth (and recession at Kossom Bougoudi) associated with increasing (and decreasing) humidity.

Neogene-Quaternary drainage features of the northeastern Borkou- Ennedi area

Strong evidence that a large lake occupied the Chad Basin at times during the late Miocene and early Pliocene comes from the fact that a record is preserved of the rivers that flowed from the lake. This section and the next deal with geomorphological evidence pointing to the occurrence of the late Miocene-early Pliocene Sahabi rivers, and their evolution is also outlined. An index map showing the area of study in the Borkou-Ennedi area of northeastern Chad is shown in Figure 5. A Landsat image showing geographical nomenclature is presented in Figure 6. The geology of the area of study based on Wolff (1964) is illustrated in Figure 7. The area of interest lies on the southern flank of the Kufra Basin (Erdis Basin in Chad) with the unconformable boundary between the Mesozoic Nubian Sandstone and the Palaeozoic rocks forming a south-facing arc, which crosses the area in an east-west direction (Fig. 7). A well-developed scarp of Nubian Sandstone marks the boundary between Palaeozoic and Mesozoic rocks. The general term that will be used for this feature is the Kossamanga Scarp (Fig. 5). A particular part of the Kossamanga Scarp, at its southeastern end, is of interest in relation to the history of the Sahabi rivers; this part of the scarp will be referred to as the Blanga-Bezz Tranga Edge (Fig. 5).

The Amatinga Valley and Amatinga Surface

One of the most striking aspects of the Landsat images of Figures 5 and 8 is the trace of a broad meandering, NE-trending valley in the region 18N-1830'N and 20E-2030'E. The feature is termed Amatinga on Carte Internationale du Monde, Largeau Sheet (Institut Gographique National 1975); the valley will therefore be termed the Amatinga Valley (Fig. 8). The Amatinga Valley is about 40 km in length and varies between about 1 and 3 km in width. It appears to have originated from or entered into a lake at both its northeastern and southwestern ends. Numerous tributaries can be seen approaching the Amatinga Valley and then entering the main valley, either individually or after first coalescing. TheAmatinga Valley is well defined in its southwestern part by the 400 m contour (Fig. 9). In the figure the northeastern part of the valley lying above 400 m is outlined. The angles of convergence of the tributaries of the Amatinga Valley with each other and with the main valley point generally in a northerly direction. Therefore a northeasterly direction of original water flow is indicated, contrary to the current topographic aspect of the valley, which slopes gently to the SW.

The Amatinga Valley more or less separates areas that have different surface textures (Fig. 8). To the NW of the valley the surface texture is closely fluted, a texture widely developed in the Borkou-Ennedi area, which can be attributed to the effect of wind erosion over long periods of time (perhaps of the order of 10 Ma). In the area lying to the SE of the Amatinga Valley, however, the surface has been modified, fluting is less evident and bedding planes can be more clearly observed. There is also a colour difference between the two areas. The area showing these distinctive features lies between the Amatinga Valley in the NW and the fluted terrain of the Mourdi Depression in the SE; it is bounded in the north by the Blanga-Bezz Tranga Edge and fades to the SW into the Chad Basin. It is shown in Figure 5 and is termed the Amatinga Surface. This surface is interpreted to be the area over which predecessors (or successors) of the river that formed the Amatinga Valley migrated during long periods of time. The migrating rivers modified the typical fluted aeolian surface of the Borkou-Ennedi (referred to subsequently as the Borkou Aeolian Surface) rendering it less fluted and more amorphous in texture.

The state of preservation, the direction of past flow and the position of the Amatinga Valley between the uplifted areas of the Tibesti and the Erdi-Ennedi all suggest that the valley is the trace of one of the Sahabi rivers. Because the Amatinga Valley separates a surface eroded by aeolian forces from a surface that has been subject to river erosion, the valley can be seen to be either the first or last expression of a migrating river valley system. The area of uplift closest to the Amatinga Valley is the Erdi-Ennedi, which lies to the east. It is most likely that river migration was a consequence of this easterly uplift and this is the conclusion adopted here. The Sahabi rivers in this area therefore migrated westwards and the Amatinga Valley is the last detailed trace of that great river system. It follows that the Amatinga Valley is of early Pliocene age on the assumption that the Sahabi rivers dried up at the end of the Zeit Wet Phase (Griffin 2002).

The Lakes Area

The Lakes Area is defined as the area between Ounianga Kbir in the west and Ounianga Srir to the east and between 1845'N and 1910'N (Figs 5 and 9). The name Ounianga Kbir refers to the largest lake of the western group of lakes and also to the nearby town. The largest western lake is also called Lac Yoa or Lake Yoan. The western group of lakes near Ounianga Kbir consists of the larger Lake Yoan, which covers an area of about 3 km^sup 2^, and a group of smaller lakes (three or four) lying about 6 km to the SSE with a total area of about 1.5 km^sup 2^. The term Katam is sometimes applied to this group of small lakes. The water of Lake Yoan is strongly saline and contains sodium chloride, sodium carbonate and sodium sulphate (Heseltine 1959). The lake supports significant vegetation and wildlife and is fed by springs of fairly fresh water (Heseltine 1959). Some 40 km to the ESE of Lake Yoan is Ounianga Srir, a group of small lakes, some partly or completely dried up. A small settlement of the same name lies close to the lakes. The lakes have a distinctive NNE orientation resulting from promontories of Nubian Sandstone extended southwestwards by sand spits; these either divide the larger lakes into arms or separate the smaller lakes from one another. The largest lake is just over 4 km^sup 2^ in area and is the largest natural lake in the Sahara (Pesce 1968). A photograph of Ounianga Srir (Pesce 1968, fig. 29) shows the largest lake of the group and gives the impression that when the photograph was taken the smaller lakes were marshy and covered with vegetation. Heseltine (1959) noted springs of 'sweet' water flowing into the lakes and this presumably is their source. In general, however, the lakes are saline, although Heseltine stated that one of the lakes is reputed to be fresh and to contain fish.

The topography of the Lakes Area (Fig. 9) reveals a series of west- and NW-trending channels (swinging to north-trending) defined by the 440 m and 400 m contours. The lowest areas are defined by the 400 m contour, and the lakes of Ounianga Kbir, Katam and Ounianga Srir are located in two of these low areas. The discussion of the previous section has shown that water leaving Neogene Lake Chad flowed initially to the NE via the Amatinga Valley. Eventually this water reached the east flank of the central Tibesti and flowed northwards (Griffin 2002). Therefore, the flow must turn west or NW at some point. During the early Pliocene this turn appears to have been in the eastern Lakes Area and to be brought about by uplifted areas to the east: the Erdi (Figs 6 and 10). This sharp turn in the course of the Sahabi rivers is termed the Elbow of the Sahabi or Eosahabi, as appropriate (Fig. 11). It follows from these considerations that the west-, NW- and north-trending channelling of the Lakes Area in all probability results from the flow of the Sahabi rivers during the late Miocene and early Pliocene.

Attention is drawn to the well-defined c. 50km long (10km wide) NW-trending depression, defined by the 440 m contour, that turns northwards at 19N, 2030'E into a north-trending depression defined by the 400 m contour (Fig. 9). The north-trending part of the above- described depression contains Lake Yoan and the lakes of Katam. This NW- and then north-trending depression will be referred to as the Kbir Channel after Ounianga Kbir. It is interpreted to be a record of the final stage of the early Pliocene Sahabi river in this area. This is based on the fact that it is well preserved for the area and it is the most westerly and southwesterly depression in the Lakes Area. Also, it is in alignment with the other sections of the early Pliocene phase of the river, taking into account the Elbow of the Sahabi.

The Pliocene-Quaternary Tid Berdgui Uplift and the Madadi Valley

The Tid Berdgui area lies between the two previously described areas and is characterized by uplift and a youthful drainage pattern (Figs 5 and 9). The uplift takes the form of two generally east- and NE-trending ridges. The more northern one is made up of the elevated areas of Tid Berdgui and Mayar and the southern one is composed of the high areas of Blanga and Bezz Tranga (Figs 5 and 9). The uplift is named after part of the higher northern feature. The Tid Berdgui Uplift is flanked on its southern side by the Blanga-Bezz Tranga Edge and the Amatinga Surface. Because of uplift and erosion the Amatinga Surface is not well preserved but it is reasonable to assume that the Sahabi rivers crossed from the Palaeozoic to the Mesozoic rocks a short distance to the south of the Blanga-Bezz Tranga Edge. The distinctive area of broad channelling (up to 25 km in an approximately SW-NE direction; see Fig. 5) immediately to the south of the Blanga-Bezz Tranga Edge appears to record Sahabi river erosion on the southern edge of the Nubian Sandstone. Subsequent modification of this channelling by wind erosion is likely. North of the edge the record of the rivers is fragmented and obscured in places by blown sand. Because the Tid Berdgui Uplift disrupts the Messinian or early Pliocene Amatinga erosion surface, it is interpreted to be Pliocene or Quaternary in age.

The broad NE-trending valley that drains the Tid Berdgui Uplift is termed the Madadi Valley (Figs 5 and 9) after the Madadi well. The valley is about 20 km long and varies between 2 and 8 km in width. The Madadi Valley is defined by the 440 m contour and has a youthful, dendritic drainage pattern; the valley drains to the SW. The valley is interpreted to be a reversed segment of the final stage (early Pliocene) of the Sahabi river. It shows a good alignment with the Amatinga Valley and its youthful aspect contrasts sharply with the meandering valley of the almost flat land to the SW of the Blanga-Bezz Tranga Edge. The Madadi Valley is the main focus of drainage for the uplifted area of Blanga and Tid Berdgui. During wet periods water would flow generally to the NW into the broad valley. This supports the idea that Quaternary and contemporary water flow in the area follows pre-existing drainage features.

The Erdi, Mourdi Depression and related features

To the east of the areas so far discussed lie the uplifted and dissected area of the Erdi and the major elongated depressions of the Mourdi Depression, the Derbili-Marhdogoum and Erg Idrisi (Figs 6 and 10). The area occupies the eastern two-thirds of Figure 6. It can be thought of as an area of the three radiating arms of the depressions, which converge to the SW, first on the heavily eroded area of the Erdi and then on the site of the now lost Neogene Lake Chad. It is thought that the Palaeosahabi flowed in the northern part of the Mourdi Depression and in the Derbili-Marhdogoum whereas the Erdi preserves the record of the Eosahabi. These ideas will be developed in the following synthesis section of the paper. The Mourdi Depression is a strongly expressed slightly curved geomorphological feature lying immediately to the north of the Ennedi highlands. The western two-thirds of the depression trends approximately east-erly, whereas the eastern third trends northeasterly. The depression is some 45 km wide and c. 275 km long. It is uplifted towards the east and p\asses by way of a scarp into a more gently eroded plateau (Figs 6 and 10). The Derbili is a 170km long sand-filled depression, some 35 km wide, trending ENE. It passes to the NE into the Marhdogoum. This broad, curved, sand- filled channel, measuring 100km by 20km, passes in turn north- easterly into a narrower more sinuous channel of similar length. The Erg Idrisi, the most northerly of these SW-converging features to be described, measures c. 250 km by 50 km and appears to be a sand- filled graben.

Synthesis: the Sahabi rivers of northern Ennedi and adjacent Borkou

The early Pliocene river and age of the Aeolian Surface; the Blanga-Bezz Tranga Edge

The Sahabi rivers are considered first because they are the easiest to interpret. The best preserved section of the early Pliocene river is the Amatinga Valley, trending for about 40 km to the NE from the approximate place where it left Neogene Lake Chad (Figs 5 and 8). The course is interrupted by the low area defined by the 400 m contour that lies to the SW of the Kossamanga Scarp (Figs 5 and 9). Moving north, in the uplifted area to the NE of the Blanga Edge is the broad NE-trending Madadi Valley. The alignment is good and this is considered to be the continuation of the early Pliocene Sahabi river to the NE. It is noted that drainage in the Madadi Valley is now to the SW. To reach the East Tibesti Valley the Sahabi rivers must have turned to the NW at some place; this change of course is termed the Elbow of the Sahabi (Fig. 11). The next northerly best preserved channel is the Kbir Channel of the Lakes Area. This is interpreted to be the continuation of the early Pliocene Sahabi in the Lakes Area, and taking into account the fact that the river must turn to the NW there is good alignment. A northerly course was eventually established and this set the river on its path towards the western flank of the hamada of the East Tibesti Valley. This outlined course is shown by line aa' in Figure 11.

As discussed above, the Amatinga Surface (Fig. 5) is considered to be the surface over which the Sahabi rivers migrated during the Messinian and early Pliocene. The Amatinga Surface broadens from about 50 km in width in the SW and is cut into the Borkou Aeolian Surface, which can be reasonably interpreted as Tortonian and older in age. The reasons for this conclusion regarding age are as follows. The Borkou Aeolian Surface precedes the Sahabi rivers, which are considered to be mainly Messinian to early Pliocene in age. The surface is therefore older than Messinian. It is younger than the Cretaceous Nubian Sandstone, a formation that shows the aeolian fluting. It is probably not Palaeogene in age because during the Palaeogene the Borkou-Ennedi was at tropical latitudes (Butterlin et al. 1993a, b) and the mid-Miocene cooling that led to drier conditions had not yet begun. Therefore, the Aeolian Surface is probably early to mid-Neogene in age and associated with the above-mentioned cooling. In part at least it can be equated with the Tortonian Dry Phase (Griffin 2002).

The Amatinga Surface is flanked to the NE by the Blanga-Bezz Tranga Edge. The Sahabi rivers flowed over this edge during the early Pliocene and the Messinian. This leads to the conclusion that the place where a Sahabi river initially crossed onto the Nubian Sandstone is at the eastern end of the projection of the Blanga- Bezz Tranga Edge (Fig. 5). There is a conspicuous sand-filled break, or notch, in the Nubian edge some 35 km to the west of the eastern end of this projection; it leads to the Derbili and this could have been the general position of the Eosahabi river at the time of the Mediterranean drawdown. Today, the notch of the Derbili has little topographic expression as it is now filled with blown sand (Figs 5 and 9).

The Messinian Eosahabi and the Erdi

The east-west channels of the Lakes Area (Figs 5 and 9) are less well preserved than the Kbir Channel; they probably predate it and are thought to be Messinian in age. This interpretation is based, first, on the concept that a Messinian record, less well preserved than the Pliocene record, is likely to be present between the East Tibesti Valley and the Amatinga Surface. A second concept envisages the Elbow of the Eosahabi moving westwards over time (Fig. 11) in response to uplift. Taking into account the initial northeasterly flow of water from Neogene Lake Chad, the east-west channels point to water coming from the east. This is the region of the Erdi, which, as noted, is an area of irregular topography and significant erosion, now partly filled with sand. There are suggestions of curving channels, for example between Erdi Dji and Erdi Korko-Erdi Fochimi and also within Erdi Fochimi (Figs 6 and 10), that could represent partial records of the Elbow of the Eosahabi. During the mid- to late Messinian, therefore, the Erdi seems to have been the site of the westward-migrating Elbow of the Eosahabi. This interpretation fits well with the notch of the Derbili being the area in which the Eosahabi passed onto the Nubian Sandstone. The Erdi must originally have been close in elevation to the place of origin of the Sahabi rivers. The uplift of the Erdi that accompanied the migration of the rivers probably continued in the Pliocene and possibly into the Quaternary.

Two stages are recognized in the development of the Eosahabi in northeastern Chad. In Stage 1, the river passed east-northeastwards from Neogene Lake Chad via the Derbili to the area of Erdi Korko, Erdi Fochimi and Erdi Dji, where it turned northwestwards in the Elbow of the Eosahabi (Fig. 11, line cc'). To the west of the Erdi the precise course of the Eosahabi is difficult to follow, largely because of the flat nature and presumed stability of the land surface; it is not unreasonable to assume, however, that in parts at least the river followed a meandering course. The westerly flow continued for 175 km or so, and the river flowing to the north of the southeastern spur of the now extinct volcano, Emi Koussi, was diverted northwards by the mass of the volcano itself. It must be assumed that the downward slope from this point was gentle in a northward direction and on the east flank of the Tibesti the river followed its new course. The gradient here was probably comparable with that of the modern Nile in southern and central Sudan. Between Juba and Khartoum the river flows a distance of about 1800 km and falls only 77 m (Said 1993). In Stage 2, during the late Messinian, the ENE-trending limb of the Eosahabi migrated to the north, whereas the NW-trending limb migrated southwards to pass through the Lakes Area (Fig. 11, line bb').

The Palaeosahabi

It remains now to consider the earliest phase of the Sahabi rivers, of early to mid-Messinian age. This age assignment is made on the assumption that the lake and river correspond approximately to the Zeit Wet Phase (Griffin 2002). It is not possible to say precisely what part of the Zeit Wet Phase the erosional record represents: it may be the early Pliocene and Messinian part only, or it may cover the period from the onset of the Zeit Wet Phase at c. 7.5 Ma. On the basis that the river of the late Messinian drawdown phase of the Mediterranean (the Eosahabi) in all probability had a predecessor, it will be assumed that the Palaeosahabi flowed during the mid- to early Messinian. Although accepting that it might also have flowed during the late Tortonian, for the purpose of discussion in this paper an essentially Messinian to early Pliocene age is postulated for the Sahabi rivers.

The morphological integrity of the Erdi and the northern part of the Mourdi Depression (Fig. 6) suggest that these areas record the mid- to early history of the Sahabi rivers (the Eosahabi and Palaeosahabi). Landsat images reveal the Mourdi Depression to be a broad, shallow, west-trending depression now tilted to the west, that can be divided into northern and southern parts, separated by a broad, disrupted medial channel c. 6.5 km wide (Figs 6 and 10). The eastern third of the Mourdi Depression is substantially sand filled. The northern part of the depression is partially sand filled in its southern part whereas its northwestern part is interpreted to be a projection, now considerably disrupted, of the eastern arm of the Amatinga Surface (Fig. 5). This latter aspect of the depression will be discussed below. The southern half of the Mourdi Depression is relatively flat and can be seen as a pediment of the uplifted Ennedi, less dissected and less vegetated than the more mountainous area to the south. This southern half of the depression shows in the west three westinclined fault blocks that are related to the most elevated part of the Ennedi (Fig. 6). At the Chad-Sudan border the Mourdi Depression passes via a scarp into a plateau area (Fig. 6). In terms of the current topography this is close to the highest part of the Mourdi Depression (c. 700 m).

The eastern arm of the Amatinga Surface is that part of the surface to the east of the notch of the Derbili and south of the sand-filled area of the Derbili (Fig. 5). Like the western part of the Amatinga Surface, the eastern arm is considered to have been subject to fluvial erosion that eliminated the aeolian fluting, and to represent the area over which the Sahabi rivers flowed. It is suggested that the Palaeosahabi flowed eastwards, first over the Carboniferous outcrop, then at a later stage in its evolution over the Nubian Sandstone flanking the Carboniferous to the north. The eastward path of the river is now blocked by the uplifted Erdi Ma (Fig. 6). The approximate place where the rivers passed from the Mourdi Depression to the plateau to the NE can be reasonably interpreted to be at the northeastern end of the Mourdi Depression close to the scarp where the depression passes to the plateau area (Fig. 6). This area currently has an elevation of c. 700 m.

The geomorphology of the Mourdi De\pression-Erdi leads to the interpretation that the Derbili-Marhdogoum served a similar function to the Mourdi Depression in conveying water to the NE from a lake in the Chad Basin. It is suggested that after a period when the river flowed in the northern part of the Mourdi Depression, the river was displaced northwards (in response to the uplift of the Ennedi) to the Derbili-Marhdogoum. This shift may have involved a number of intermediate positions. At the end of the mid-Messinian it appears that uplift of the Erdi-Ennedi reached a point at which the Palaeosahabi was prevented from following its northeasterly course. Instead, it was displaced back from the Marhdogoum so that it flowed from the Derbili to the area of Erdi Fochimi where it was diverted to a general westerly flow. Thus began the history of the Eosahabi.

The actual paths of the Palaeosahabi rivers in northern Sudan and Egypt are currently somewhat conjectural. The paths may be represented by some part of the Radar Rivers of McCauley et al. (1998). These rivers could have reached the Mediterranean by joining the early to mid-Messinian Nile drainage system or by following a more northerly and then northwesterly course to reach the Mediterranean in the Gulf of Sirt (Fig. 1). Both routes might have been utilized but at different times. Further investigation may reveal the most likely paths followed by the Palaeosahabi rivers.

The main evidence for the occurrence of the initially eastflowing Palaeosahabi rivers is as follows. First, the morphological integrity of the Amatinga Surface, including its eastern arm (Fig. 5), suggests a continuous history of development for the rivers extending from the earliest Mourdi Depression phase to the final early Pliocene flow. Also, the eastern arm of the Amatinga Surface has a distinct topographic expression slightly discordant to the rest of the Mourdi Depression (Fig. 6), suggesting a unique origin. Second, the Derbili-Marhdogoum shows morphological and sedimentary evidence of having been a river valley (McCauley et al. 1998). McCauley et al. interpreted the Derbili-Marhdogoum to be a late part of the Trans-African Drainage System. However, the Derbili- Marhdogoum belongs to the mid-Miocene cycle of uplift and basin formation and therefore is most likely to record the later Palaeosahabi drainage system rather than the earlier trans-African system. Also, the 4 Ma Tortonian Dry Phase separates the two drainage systems. Finally, the Elbow of the Eosahabi suggests that an east-north-easterly ancestral river flow (the Palaeosahabi) was subsequently modified to a northwesterly flow (the Eosahabi).

The Amatinga-Mourdi Outlet

The approximately triangular-shaped area where the river flowed over Palaeozoic rocks before crossing onto the Nubian Sandstone was previously referred to as the Amatinga Surface. This feature and the northwestern end of the Mourdi Depression served as the area of exit of the Sahabi rivers from Neogene Lake Chad for over 2 Ma. The place where the early Pliocene, Messinian and possible late Tortonian rivers exited from Neogene Lake Chad will be termed the Amatinga- Mourdi Outlet (Figs 6 and 11) and its significance will be discussed in the next section.

Discussion and conclusions

Tectonic, sedimentary and climatic considerations

The Chad Basin is a composite Neogene structure comprising the northern and southern Chad Basins (Fig. 2). The southern Chad Basin overlies an area of convergence of three rift trends of Cretaceous and Palaeogene age (Fig. 3) and contains post-rift Miocene sediments of fluviatile and lacustrine origin. In the south of the Chad Basin the post-rift deposits are referred to the Continental Terminal or the Palaeochadian Series and include sands and clays, often variegated; kaolin occurs and ferruginous crusts are developed in places. In the Lake Chad area correlative rocks comprise dominantly sandy intervals (100 m or so in thickness) alternating with dominantly clayey intervals of similar thickness referred to the Continental Terminal or the Chad Baguirmi Series. A lacustrine origin is likely for some of these clayey sediments. In the north, 200 m or so of post-rift clays alternating with fluviatile sands are recognized (Bodl Series; Fig. 4) and biostratigraphic dating has established lake deposits at c. 6.5 Ma and c. 5 Ma. Also, Upper Miocene-Lower Pliocene rocks have been recognized on the basis of mammalian remains. Certainly the upper part of this succession is late Miocene in age and represents the record of a widespread transgression in the post-rift Chad Basin, with often immature sediments being deposited over an irregular faulted terrain. This contrasts sharply with the overlying lake deposits of mid- and late Pliocene age that are generally fine grained and reflect a more mature basin setting. In the Chad Basin Quaternary deposits are thin and show evidence of increasing aridity.

The record of the Messinian sediments in the Mediterranean demonstrates clearly that it was a time of climatic cyclicity. This can be reasonably interpreted as the alternation of a dry with a more humid climate in response to the c. 20 ka precessional cycle (Krijgsman et al. 1999; Sierra et al. 1999; Griffin 2002). The dust flux curve from ODP Arabian Sea Sites 721/22 also shows the Messinian climate to be markedly cyclical at the precessional frequency (deMenocal & Bloemendal 1995). This leads to the concept that the climate of the Chad Basin was also significantly cyclical during the Messinian. Recent reports arising from the work of the MPFT have described parts of two sedimentary cycles in the Chad Basin, deposited within the time frame of early Pliocene to Messinian (Brunei & MPFT 2000; Vignaud & MPFT 2002). These cycles indicate the passage from a dry to a wet environment with lake development at the sites studied. The periods of pedogenesis recorded in the Miocene of the south of the Chad Basin and suggested in the north also attest to climate change.

Considerations of geomorphology, river evolution and age

It is concluded that a record of the Messinian to early Pliocene Sahabi rivers is preserved in the northern Ennedi and adjacent Borkou of northeastern Chad (Fig. 6). The geomorphology of northeastern Chad-southern Libya allows the history of the Sahabi rivers to be traced. Their history corresponds approximately to the Zeit Wet Phase (late Tortonian to early Pliocene) although it is not possible to say when the first eastward-flowing Palaeosahabi originated. For the purpose of this discussion the history of the Palaeosahabi is taken to commence in the early Messinian. During the period of the Zeit Wet Phase a large lake, termed Neogene Lake Chad, occupied the Chad Basin (Fig. 2). The lake fluctuated in size in response to the c. 20 ka precessional cycle, giving rise to the Sahabi rivers on an intermittent basis (Figs 1 and 11). The Sahabi rivers flowed in turn eastwards, northeastwards, and northwards from Neogene Lake Chad across the area between the Tibesti and Ennedi highlands, often between gently tilted fault blocks. The earliest river, the Palaeosahabi, flowed approximately eastwards in the northern part of the Mourdi Depression (Figs 6 and 10). It was subsequently displaced to the NE and flowed in the Derbili- Marhdogoum depression. These rivers flowed into the Sudan and Egypt and reached the Mediterranean via the early to mid-Messinain Nile drainage system and/or the Gulf of Sirt (Fig. 1). Tectonic activity at the end of the mid-Messinian resulted in the easterly and northeasterly flow of the river being blocked, so that it turned to the west in the Erdi of northeastern Chad. This marked the beginning of the Eosahabi, which flowed during the drawdown of the Mediterranean (late Messinian). After turning in the Erdi the Eosahabi flowed westwards, reaching the flank of the Tibesti c. 140 km to the east of the volcano Emi Koussi, from where the river flowed generally northwards in the East Tibesti Valley.

During the Messinian a loop of the Eosahabi existed in the Erdi and by migrating to the west contributed to creating the irregular topography of the area (Fig. 11). During part of its history the Eosahabi flowed through the area of the lakes Ounianga Kbir and Ounianga Srir. The latter group of lakes includes the largest lake in the Sahara, only 4 km2 in area (Figs 5 and 9). All these lakes are located in low-lying parts of the old river system. The final stage of the Sahabi rivers, the Sahabi, flowed during the early Pliocene. It had a more north-south course than the previous rivers and was not as incised as the Eosahabi. A particularly well- preserved section of the Sahabi is preserved in the Amatinga Valley of the east-central Borkou (Fig. 8). The valley is mainly 1-2 km wide, meandering for some 40 km in a northeasterly direction, and is cut into sandstone and shales of Palaeozoic age. Although it now has a southwesterly gradient, tributaries to the valley indicate that the original direction of flow was to the NE. The Amatinga Valley defines the western limit of the Amatinga Surface (Fig. 5), an erosional surface of mainly Messinian and early Pliocene age that was cut by the westward-migrating Eosahabi and Sahabi rivers. The Amatinga Surface contrasts with the Borkou Aeolian Surface of east- central Borkou, an older (Tortonian?) fluted surface showing the effects of wind erosion.

Duration of the rivers and lake; cyclicity and environmental considerations

The long history of the Sahabi rivers, demonstrated by the erosional record left in northern Ennedi and adjacent Borkou, allows the conclusion to be drawn that there was also a long history of lake development in the Chad Basin during the period the rivers were in existence. The area at the western end of the Mourdi Depression where the rivers left the Chad Basin is termed the Amatinga-Mourdi Outlet (Figs 6 and 11). It is about 50km wide. The presence of the Amatinga-Mourdi Outlet i\s significant, because were the extreme case to exist in which there was no preserved Messinian sedimentary record in the northern Chad Basin, the existence of the Outlet and associated rivers would testify to a long history of lake development in the Chad Basin. The Miocene of the northern Chad Basin is relatively thin and the Amatinga-Mourdi Outlet therefore suggests that the succession is incomplete.

As noted above, cyclicity has been observed in the Messinian succession of the Chad Basin. Giving consideration to the concept that the Messinian may not be fully developed in the Chad Basin, it is possible that some parts of the late Neogene succession of the northern Chad Basin have been lost as a result of wind erosion during the dry periods. Further study, as demonstrated by the work of Brunei & MPFT (2000) and Vignaud & MPFT (2002), may provide additional evidence of cyclicity. Because of the precessional influence on Messinian climate the Sahabi rivers would develop during some, but not necessarily all, of the wet cycles. These considerations make it possible to envisage the environment of the Chad Basin changing significantly in response to the precessional cycle. In wet times the lake would be large with the lake edge environment at a maximum. During dry periods, with the lake smaller, savanna and woodland would be widely developed. Mammal and other animal species that could adapt to both aspects of these medium- length cycles would be favoured.

I should like to thank the following individuals who helped in various ways in the preparation of this paper: M. Bye and D. Bebbington (Geolmage, Perth, Australia), R. Garrett, K. and S. Griffin, A. Hills, G. Kernick (PageSetter Design, Perth, Australia), M. Martin, K. Murray, D. and C. Taylor, K.-H. Wyrwoll and V Forbes (both of University of Western Australia). A review by M. A. J. Williams (University of Adelaide) led to much improvement in the manuscript. J. Woodward (University of Manchester) provided an editorial review, and a further anonymous constructive review was also provided. These reviews are much appreciated.


BARR, F.T. & WALKER, B.R. 1973. Late Tertiary channel system in Northern Libya and its implications on Mediterranean sea level changes. In: RYAN, W.B.F. & HS, K.J. ET AL. (eds) Initial Reports of the Deep Sea Drilling Project, 13. US Government Printing Office. Washington, DC, 1244-1251.

BOAZ, NT., EL-ARNAUTI, A.. GAZIRY. A.W., DE HEINZELIN, J. & BOAZ, D.D. (EDS) 1987. Neogene Paleontology and Geology of Sahabi. Libya. Alan R. Liss, New York.

BRUNET, M. & MPFT, 2000. Chad: discovery of a vertebrate fauna close to the Mio-Pliocene boundary. Journal of Vertebrate Paleontology, 20(1), 205-209.

BRUNET, M. & MPFT, 2002. A new hominid from the Upper Miocene of Chad, Central Africa. Nature, 418, 145-151.

BURKE, K. 1976. The Chad Basin: an active infra-continental basin. Tectonophysics, 36, 197-206.

BUTTERLIN, J., VRIELYNCK, B. & BIGNOT, G. ET AL. 1993a. Lutetian (46 to 40 Ma). In: DERCOURT. J., Ricou, L.E. & VRIELYNCK, B. (eds) Atlas Tethys Palaeoenvironmental Maps. Explanatory Notes. Gauthier- Villars, Paris, 197-210.

BUTTERLIN, J., VRIELYNCK, B. & BIGNOT, G. ET AL. 1993b. Lutetian palaeoenvironments (46 to 40 Ma). In: DERCOURT, J., Ricou, L.E. & VRIELYNCK, B. (eds) Atlas Tethys Palaeoenvironmental Maps. Maps. BEICIP-FRANLAB, Rueil-Malmaison.

CHOUBERT, G. & FAURE-MURET, A. (co-ords) 1976. Geological World Atlas, Sheets 6 and 8, 1:10000000. International Geological Mapping Bureau. Commission for the Geological Map of the World-UNESCO, Paris.

CHOUBERT, G. & FAURE-MURET. A. (co-ords) 1987. International Geological Map of Africa, 1:5 000 000. International Geological Mapping Bureau. Commission for the Geological Map of the World- UNESCO.

CITA, M.B. & MCKENZIE, J.A. 1986. The terminal Miocene event. In: HS, K.J. (ed.) Mesozoic and Cainozoic Oceans. American Geophysical Union, Geodynamic Series, 15, 123-140.

COPPENS, Y. & KOENIGUER, J.-C. 1976. Signification climatique des paloflores ligneuses du Tertiaire et du Quaternaire du Tchad. Bulletin de la Socit Gologique de France, (7), XVIII(4), 1009-1015.

DECIMA, A. & WEZEL, F.C. 1973. Late Miocene evaporites of the central Sicilian Basin, Italy. In: RYAN, W.B.F. & HsU, K.J. ET AL. (eds) Initial Reports of the Deep Sea Drilling Project, 13. US Government Printing Office, Washington, DC, 1234-1241.

DE HEINZELIN, J. & EL-ARNAUTI, A. 1987. The Sahabi Formation and related deposits. In: BOAZ, N.T., EL-ARNAUTI, A., GAZIRY, A.W., DE HEINZELIN, J. & BOAZ, D.D. (eds) Neogene Paleontology and Geology of Sahabi, Libya. Alan R. Liss, New York, 1-21.

DEMENOCAL, P.B. & BLOEMENDAL, J. 1995. Plio-Pleistocene climatic variability in subtropical Africa and the paleoenvironment of hominid evolution: a combined data-model approach. In: VRBA, E.S., DENTON, G.H., PARTRIDGE, T.C. & BURCKLE, L.H. (eds) Paleoclimate and Evolution, with Emphasis on Human Origins. Yale University Press, New Haven, CT, 262-288.

FAURE, H. 1966. Reconnaissance gologique des formations sdimentaires postpalozoques du Niger oriental. Mmoires du Bureau de Recherches Gologiques et Minires, 47.

GENIK, G.J. 1993. Petroleum geology of Cretaceous-Tertiary rift basins in Niger, Chad and Central African Republic. AAPG Bulletin, 77, 1405-1434.

GHIENNE, J.-F., SCHUSTER, M., BERNARD, A., DURINOER, P. & BRUNET, M. 2002. The Holocene giant Lake Chad revealed by digital elevation models. Quaternary International, 87, 81-85.

GRIFFIN, D.L. 1999. The late Miocene climate of northeastern Africa: unravelling the signals in the sedimentary succession. Journal of the Geological Society, London, 156, 817-826.

GRIFFIN, D.L. 2002. Aridity and humidity: two aspects of the late Miocene climate of North Africa and the Mediterranean. Palaeogeography, Palaeoclimatology, Palaeoecology, 182, 65-91.

GUIRAUD, R. & BOREL, G. 2001. Miocene and Early-Middle Eocene, Northern Africa. In: Stampfli, G., Borel, G., Cavazza. W., Mosar, J. & Ziegler, P.A. (eds) 2001, The Paleoleclonic Atlas of the PeriTethyan Domain. European Geophysical Society (CD-ROM).

HESELTINE, N. 1959. From Libyan Sands to Chad. Museu