Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230 242 www.elsevier.

com/locate/palaeo

Evidence for Megalake Chad, north-central Africa, during the late Quaternary from satellite data
Marc J. Leblanc a,*, Christian Leduc b, Frank Stagnitti c, Peter J. van Oevelen d, Chris Jones e, Linus A. Mofor f, Moumtaz Razack g, Guillaume Favreau b
a School of Geosciences, Monash University, Clayton 3800, Australia IRD-UMR Hydrosciences, BP 64501, 34394 Montpellier cedex 5, France c School of Ecology and Environment, Deakin University, Warrnambool 3280, Australia d European Space Agency, Keplerlaan 1, Postbus 299, 2200 AG Noordwijk, The Netherlands School of Computer Science, Cardiff University, Newport Road, PO Box 916, Cardiff CF24 3XF, UK f School of Technology, University of Glamorgan, Pontypridd CF37 1DL, UK g University of Poitiers, 40 avenue du Recteur Pineau, 86022 Poitiers Cedex, France b

Received 22 September 2004; received in revised form 21 July 2005; accepted 27 July 2005

Abstract The existence of a very large Lake Chad during the late Quaternary, Megalake Chad, has long been questioned. A Megalake Chad would present strong evidence for climatic fluctuations of great magnitude during the Holocene in tropical Africa. In this paper we used satellite data from Landsat and Modis sensors to collect and analyse new information on landforms in a 2 000 000 km2 region of the Lake Chad Basin. We detected 2300 km of remains marking the ancient shoreline of Megalake Chad. The satellite data also indicated many Saharan rivers and relict deltas leading to the long paleoshoreline. Large dunefield flattenings were observed and interpreted as the result of wave-cut erosion by the paleolake. Similarities were noticed between the landforms observed along the paleoshoreline of Megalake Chad and that of the former Aral Sea. This finding has significant consequences for reconstructing paleohydrology and paleoenvironments through the Lake Chad basin, and continental climate change. D 2005 Elsevier B.V. All rights reserved.
Keywords: Lake Chad; Paleohydrology; Remote sensing; Geomorphology; Climate change; Ancient shoreline

1. Introduction The Lake Chad Basin in north-central Africa is the worlds largest endorheic basin, covering an area of
* Corresponding author. Fax: +61 399 054 903. E-mail address: Marc.Leblanc@sci.monash.edu.au (M.J. Leblanc). 0031-0182/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.palaeo.2005.07.016

2 500 000 km2 (Fig. 1). The central part of this basin began to fill with sedimentary deposits during the Cretaceous period (Genik, 1992). It is now mainly covered with unconsolidated Quaternary sediments. Presently, most of the northern portion of the basin is dry. Originating in the south, the Chari and the Logone rivers provide most of the inflow to Lake Chada vast and shallow fresh water lake bordering

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

231

Fig. 1. Study area and the remains of Megalake Chad detected with satellite data. Black rectangles indicate the location of the other figures mentioned in the text. Double circles represent the sites whose lithology was determined using spectral signature.

the Sahara Desert. Due to its endorheic nature and because it is located in a region of little relief, climate fluctuations have a great impact on the areal extent of Lake Chad. For example, from 1960 to 1990 the level of the lake fluctuated approximately between 277 and 284 m amsl and its area from 6000 km2 to 25 000 km2 (Olivry et al., 1996). During the last millennium major floods occurred on several occasions. During these events water levels rose above 286 m amsl and the Lake expanded to the north-east, partly flooding the

Bahr el Ghazal valley (Maley, 1981; Olivry et al., 1996). Although large portions of the Bahr el Ghazal were inundated during these major floods, the water never flooded the Lowlandsa vast region of northern Chad about 115 m below the current level of Lake Chad and linked to Lake Chad through the Bahr el Ghazal valley (Fig. 1). The hypothesis of a much larger ancient Lake Chad during the late Quaternary, Megalake Chad, remains controversial. Doubt about the existence of Megalake

232

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

Chad has largely resulted from a paucity of geomorphological mapping, especially at regional scale, due in part to the vast expanse of the Lake Chad Basin and difficulties in accessing sites from the ground. However, remote sensing data provide an excellent new source of information. Satellite imagery also enables very large landforms to be examined entirely and in context with the surrounding environment. Recent studies in other parts of the world have demonstrated the value of remote sensing and GIS to reconstruct paleolakes hydrology (e.g. Pachur and Rottinger, 1997; Komatsu et al., 2001; DeVogel et al., 2004). In the study area, which covers the central and northern part of the Lake Chad Basin (Fig. 1), we used Landsat TM and MSS and Modis data to analyse large geomorphological and hydrological landforms. In the first part of this paper we review the previous research conducted on Megalake Chad. We then detail how we used remotely sensed data to detect landforms in the study area and gather new information. Subsequently, each type of landform identified is detailed and its potential link with Megalake Chad analysed. In the discussion, we compare the evidence gathered in this study with the arguments for a neotectonic hypothesis. We also compare the landforms observed in the Lake Chad Basin with those of the former shoreline of the Aral Sea. Finally the implications of our findings are discussed in regards to the regional paleohydrology and continental climate change.

2. Review of previous work The possible existence of a very large ancient Lake Chad was noted by Tilho (1925). Later, Pias and Guichard (1957) reported that a vast, ancient Lake Chad was responsible for a long sandy ridge that they observed in Cameroon and Chad, and which coincides with an outlet to the Gulf of Guinea through the Mayo Kebbi river. More features of this old, sandy, lacustrine ridge were recognised by geologists, geomorphologists and pedologists during early cartographic surveys in the 1950s and 1960s, in Nigeria, Cameroon, Chad and Niger (Pias and Guichard, 1957; Grove, 1959; Pirard, 1967). Schneider (1967) synthesised these works together with archeological data and topographic cross sections and interpreted the sandy

ridge as the perilacustrine ridge of Megalake Chad with a constant elevation of 325 m amsl. The map published by Schneider (1967) was the first to exhibit rightly the paleoshoreline of Megalake Chad. The study of diatoms and pollens in lacustrine sediments, together with stratigraphical (including 14C dating) and sedimentological methods, confirmed the presence of a large and deep ancient Lake Chad in the Holocene (Maley, 1981; Servant and Servant, 1983). The diatom and pollen data were synthesised by Maley (1981, 2004) and showed that the last main extension of Megalake Chad intervened between 7700 and 5500 14C years BP with 3 successive sub-phases. Recently, an erosion platform and conglomerates were identified around the Hadjer el Hamis inselbergs and were interpreted as the result of wave erosion when the inselbergs formed islands within Megalake Chad (Schuster et al., 2003). However, some researchers argued against the existence of a giant paleolake. Durand (1982, 1995) and Durand et al. (1984) reported that landforms thought to be part of Megalake Chad were in fact of neotectonic origin. Major arguments are as follow. Firstly, it was reported that, across the basin, the distribution of the paleoshoreline is not continuous and shows major gaps, some in wide areas. Secondly, it was assessed that the elevation of the ancient ridge is not constant. Thirdly, it was argued that landforms initially considered to be the remains of the paleoshoreline have limited orientation and therefore are neotectonic features following the direction of major structural faults which were also indicated by gravimetric anomalies (Durand, 1982; Durand et al., 1984; Durand, 1995).

3. Methodology To cover the entire study area, which spans over 2 000 000 km2, five large mosaics of Landsat TM data were obtained from NASA Earth Science Enterprise. Each mosaic was created from 20 to 29 orthorectified and contrast adjusted Landsat TM scenes which were collected between 1986 and 1990 (Table 1). The mosaics provide spectral information in 3 bands (b7MIR, 2.082.35 Am; b4-NIR, 0.760.90 Am; b2-vis, 0.520.60 Am) with a spatial resolution of 30 m. In the south, dense vegetation may mask the underlying

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242 Table 1 List of remote sensing data and maps Data type Remote sensing Data Landsat TM mosaic (Lake Chad Basin) ID N32-15 N33-05 N33-10 N33-15 N14-15 N40-40 N40-45 N41-40 N41-45 P197-R50 P197-R51 P197-R52 P198-R50 P198-R51 P198-R52 P199-R51 P199-R50 P200-R50 P200-R51 MOD09A1.A2003041.h19v07.0 04.2004015043655 Date All images taken between 1986 and 1990 Source NASA Earth Science Enterprise

233

Landsat TM mosaic (Aral Sea)

All images taken between 1986 and 1993 23/01/1975 29/01/1973 15/02/1975 17/02/1973 12/01/1973 07/03/1973 31/01/1973 31/01/1973 01/02/1973 03/12/1975 10/02/2003 to 17/02/2003

NASA Earth Science Enterprise

Landsat MSS (Lake Chad Basin)

USGS

MODIS (surface reflectance) Maps Topographic maps

NASA Topographic maps series of central and west Africa at 1:200 000. IGN, Paris International Geological Map of Africa Sheet no. 2. 1:5 000 000. Choubert, G. Faure-Muret, A. Commission for the Geological Map of the World. UNESCO: Paris. 1987 Carte Geologique de la Republique du Tchad. 1:1 500 000. Wolf, J.P. BRGM. 1964

Geological maps

structures because the Landsat TM scenes of the mosaics were taken at or close to the peak of the growing season. For this reason, in the south TM data were supplemented with Landsat MSS selected from the dry season taken between January 1973 and December 1975 (Table 1). These dry season images avoid interference by cloud and vegetation, thus enhancing the interpretation of geology, geomorphology and pedology. Visual interpretation of the satellite images relied on the analysis of essential surface characteristics within the regional context: image tone, texture, pattern, shape, size and site location. However this process is often subjective. Therefore, digital image processing techniques were also used in order to perform a more rigorous analysis of the Landsat TM and

MSS data. A principal component analysis (PCA) was used to compress the multi-spectral data into black and white images and to highlight parts of the regional spectral variation (Rees, 2001). The bands of PCA data are independent and therefore non-correlated. This leads to greater interpretability compared to the source data. Successive principal components of MSS and TM data were generated by calculating new coordinate systems in which much of the redundant information was removed. A standard deviation contrast stretch was applied to the derived images for visual enhancement. This technique is widely used in remote sensing to improve the visual interpretability of an image. It amplifies the spectral differences between the features of an image to increase their visual distinction. A Sobel edge detection and filtering

234

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

was performed on the Landsat TM and MSS data and on the derived principal components. Edge detection is an effective operation in digital image processing enhancing the identification of features such as lineaments, drainage networks and geological boundaries (Rees, 2001). Using convolution filters, edge detectors smooth out areas of low spatial frequency and create a sharp contrast where spatial frequency is high, which occurs at the edges between homogeneous groups of pixels. Field missions in Chad, Niger, Nigeria and Cameroon were conducted in 2002 and 2004. During these surveys, field observations of the landforms were undertaken to ground-truth satellite image interpretation. A Modis image with 500 m resolution, taken in January 2002, was selected to provide consistent spectral data in seven bands across the entire region (Table 1). Using the soil (Pias, 1970) and geological maps (Choubert and Faure-Muret, 1987; Schneider and Wolff, 1992), a spectral library of the major geological deposits found in the region was built. It was then possible to compare the spectral signature of these materials and determine the lithology of the landforms detected in the Landsat data. A rose diagram was used to analyse the orientation of the detected landforms. Long features may be curved and take several directions. To address this, all the selected features were segmented into sections of iso-direction. The orientation was calculated for each section and all sections were grouped in classes of 58. The cumulated length of all the sections for each 58 class was reported in the rose diagram.

4. Results Inside the Lake Chad Basin, satellite data reveal an extensive series of long features. Bands of geological deposits, up to 3 km in width and hundreds of kilometres in length were detected in Chad, Cameroon, Niger and Nigeria. For example, in northern Chad, the PCA image of Landsat TM data shows a sharply defined, curved structure forming long smooth bends (r in Fig. 2A). It is about 230 km long. The edge detector clearly singles out this feature in the landscape (Fig. 2B). In eastern Chad, the PCA of Landsat TM data shows a sub-continuous group of well-defined bands, which appear in dark tone (r in

Fig. 2C). They form long and adjoining curves, which are recognised by the edge detector (Fig. 2D). Across the study area, these bands of geological deposits were spectrally identified as sand (Fig. 3). In other parts of the region, long sharp breaks in the landscape were detected. For example in western Chad, the Landsat data show a sharp break in the landscape (r in Fig. 2E) marking the limit between the Kanem dunefield, indicated by a mottled texture (d in Fig. 2E), and a nearly featureless plain of unconsolidated sediments, characterised by a homogeneous texture (p in Fig. 2E). In places, both the long bands of sandy deposits and the breaks in the landscape are clearly apparent. For example in Niger, the PCA of Landsat MSS data shows a long sandy band with a smooth curve (r in Fig. 2F), which is also marking the boundary between a dunefield to the west (d in Fig. 2F) and a vast plain to the east (p in Fig. 2F) Similarly in Cameroon, the PCA of Landsat TM data shows that a long sandy band with a smooth curve (r in Fig. 4) separates an old dunefield to the south (d in Fig. 4) and a vast plain to the north (p in Fig. 4). The position of these long features was mapped throughout the entire study area. It appears that they have a particular arrangement defining a group so aligned as to form a very large and closed loop (bold, black lines in Fig. 1). Their shape, curvature, continuity, defined edges and disposition as a loop strongly indicate a relict beach ridge marking the shoreline of a vast paleolake Chad. In places, traces of this putative paleoshoreline are more than 600 km from the current dstandardT Lake Chad. Clearly identifiable shoreline features extend for 2280 km, and linear interpolations between breaks in this paleoshoreline suggest that the paleolake had an approximate circumference of 3100 km and an area of about 340 000 km2. Traces of another, smaller, transgression of Lake Chad, at around 288 m amsl, are evident from the presence of another sandy beach ridge to the west and north-west of the current Lake Chad (Pias, 1970; Maley, 1981, 2004; Schneider and Wolff, 1992; Schneider, 1994). The PCA of the Landsat MSS data shows that the shape, tone and texture of this sandy ridge at 287290 m amsl (rr in Fig. 2F) are similar to those of the sandy bands of the larger beach ridge herein identified.

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

235

Fig. 2. Examples of the remains of Megalake Chad shoreline detected in the Landsat data. A: Landsat TM data, PCA 1st component. The image reveals a very long continuous band of deposits with smooth curvatures (r). Note the Angamma cliff (c) to the south of this feature. B: Sobel edge detector for the PCA 1st component of Landsat TM data. The edge detector identifies this long band of deposits as a major discontinuity in the landscape. C: Landsat TM data, PCA 1st component. The image shows a sub-continuous group of long, aligned bands of deposits (r). Note also contemporary streams (s) cutting through the bands, some of which are feeding Lake Fitri (f). D: Sobel edge detector for the PCA 1st component of Landsat TM data. The edge detector identifies the long bands of deposits as a significant limit in the Landsat data. E: Landsat TM data, PCA 1st component. The image shows a sharp break in the landscape (r) marking the limit between the Kanem dunefield (d) and a nearly featureless, vast plain of unconsolidated sediments (p). F: Landsat MSS data during the dry season, PCA 1st component. The image reveals a continuous sand band (r) associated with a sharp break in the landscape separating dunefields (d) to the west and a vast plain (p) to the east. Note the presence of similar sandy bands (rr) closer to the shoreline of the dmodernT Lake Chad (lc). These later sandy bands have been identified as the remains of the shoreline corresponding to the last transgression of Lake Chad.

In Cameroon, the PCA of Landsat TM data shows Megalake Chad ridge as a regular and slightly curved band, appearing in grey tones over a darker surface (r in Fig. 4). To the east, satellite data indicate that the beach ridge stops abruptly at the edge of the Lake Chad Basin just before adjoining with the Mayo Kebbi river, whose drainage

network is also clearly visible in the image (s in Fig. 4). Therefore, a considerable discontinuity in Megalake Chad shoreline coincides with a substantial spillway, a former outlet to the sea, via the Mayo Kebbi river and the Benue Basin. Satellite images thus confirm that this vast paleolake Chad was no longer a closed lake.

236

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

7000

Spectral signatures
Site 1

6000

Site 2 Site 3

5000

Site 4 Site 5 Site 6

Location of these sites is reported in Fig. 1

4000

Site 7 siliceous sand of dunefields mean basalt

DN
3000

clayey sands ferrallitic sandstone salt deposits

2000

crystalline basement sandy grey clay of oases

1000

urban vertisoils

(620-670nm)

(841-876nm)

(459-479nm)

(545-565nm)

(1230-1250nm) (1628-1652nm)

(2105-2155nm)

0
1 2 3 4 5 6 7

MODIS bands
Fig. 3. Spectral signatures of typical geological deposits in the Lake Chad Basin and of the relict ridge of Megalake Chad.

Fig. 4. Landsat TM data, PCA 1st component. The satellite data confirm that a significant discontinuity in the paleoshoreline (r) coincides with a spillway to the sea, the MayoKebbi (s). Note the current course of the Logone river (l). The image also reveals that the paleoshoreline (r) is associated with a sharp break in the landscape separating a vast plain (p) to the north and a dunefield (d) to the south in the region of Kalfou.

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

237

In the Saharan portion of the Lake Chad Basin, the Landsat TM data reveal large drainage networks. In Niger, the Landsat TM data show that the Dillia landforms, sometimes regarded as corridors of aeolian erosion along neotectonic structures (Durand, 1995), are in fact interconnected channels forming a large drainage network that extends over 350 km. In Fig. 5A, the PCA of Landsat TM data shows dendritic channels upstream in the Dillia (d in Fig. 5A), connecting to parallel channels (p in Fig. 5A) and the main river channel (m in Fig. 5A). Similarly, in northern Chad, the Landsat TM data indicate a series of large drainage networks, of an average length of ~ 250 km, with their dendritic shape upstream (d in Fig. 5B) leading to main channels (m in Fig. 5B). These drainage networks are not compatible with modern climate. The present annual rainfall decreases from about 900 to 15 mm year 1 between latitudes 108N and 188N, respectively (210 mm year 1 at the northern side of the present Lake Chad) and is unable to produce a significant long distance runoff in a sandy flat environment, even when exceptionally heavy localised rainstorms occur. These ancient drainage networks all lead toward the paleoshoreline of Megalake Chad (Fig. 1). In three locations, dense, braided and bifurcated channel networks forming fan shaped structures, which are typical of deltas, are detected in the Landsat

TM data. The largest paleodelta is that of the Chari river (sd1 in Figs. 1 and 6A). In the Sahara, paleodeltas are observed in the Angamma region (sd2 in Fig. 1) and at the terminus of the wadi Achim (sd3 in Figs. 1 and 6B). The Landsat data show that the paleodeltas are positioned along the outer edge of Megalake Chad shoreline. Regions where the shoreline of Megalake Chad sharply limits the extent of old dunefields were presented above (Figs. 2E, 2F and 4). On the Lake Chad side, these old dunefields are replaced by large plains. Given the regional extent, clear layout, and distribution along Megalake Chad shoreline, we interpret these patterns as dunefield flattenings resulting from wave-cut action by Megalake Chad. Further traces of wave erosion by Megalake Chad are revealed in the satellite data. Fig. 7A is a PCA image of Landsat TM data for the Harr dunefield, a region located inside the limits of Megalake Chad. The Landsat data and topographic maps of this area reveal that the higher part of the Harr dunefield, between 300 and 310 m, is homogeneously flat. On the PCA image of the Landsat TM data, this flattened area appears as a homogenous surface forming a nearly featureless sand plain (fd in Fig. 7A). In contrast, in the lower belt surrounding this region, the dunefield regains its typical form and topography. On the PCA of the Landsat data, this is indicated by a series of dark and bright tones (respectively, oases and

Fig. 5. Saharan rivers detected in the Landsat TM data. A: Landsat TM data, PCA 3rd component. The image reveals that the Dillia is a paleoriver. It shows the Dillias large drainage network with dentritic drainage patterns upstream (d), parallel channels (p) following the underlying geology of longitudinal sand dunes (s), and the main channel (m). Note that the long sand dunes (s), with abundant supply of sand, are parallel to the direction of the prevailing wind. B: Landsat TM data, PCA 3rd component. The image shows large Saharan rivers in northeast Chad with dendritic drainage networks upstream (d) leading to main channel (m).

238

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

Fig. 6. Paleodeltas detected in the Landsat data. A: Landsat TM data, PCA 1st component. The image shows the large paleodelta of the Chari (sd1) that coincides with the paleoshoreline of Megalake Chad (r). Note the current course of the Chari (c) and Logone (l) rivers. B: Landsat TM data, PCA 1st component. The image shows the main channel of wadi Achim (m) leading to a paleodelta (sd3) at the border of Megalake Chad shoreline (r).

sand dunes) that are arranged in a regular pattern characteristic of a well-developed dunefield (d in Fig. 7A). We interpret this large dunefield flattening as the result of wave-cut action by Megalake Chad. Fig. 7B shows a series of contemporary streams in central Chad cutting through a 35 km length of the

relict beach ridge of Megalake Chad. Their disposition is particularly revealing. They all flow in the same direction and are perpendicular to the relict ridge. This is an indication that in this area the relict ridge is subhorizontal and that elevation is lower on the Lake Chad side. Just after crossing the beach ridge these

Fig. 7. A: Landsat TM data, PCA 1st component. The image shows the flattening of the Harr dunefield in its higher part (fd) surrounded, at a lower elevation, by typical sand dunes (d). Note the Bahr el Ghazal valley (v) which links Lake Chad to the Lowlands region in north Chad. B: Landsat TM data, PCA 1st component. The image reveals a series of small contemporary streams (s), all cutting the relict beach ridge (r) in the same direction and perpendicularly. Just after the relict beach ridge these small streams generate alluvial fans (af) which indicates that a flat topography is bordering the shoreline of Megalake Chad.

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

239

small streams spawn into alluvial fans (af in Fig. 7B). This indicates that a flat, sub-horizontal area is bordering the inner side of the paleoridge and can be interpreted as a shoreline terrace.

5. Discussion Data presented here show that the major tectonics arguments used to refute the existence of Megalake Chad require re-examination. Firstly, it was reported by Durand (1995) that the traces of Megalake Chad shoreline were discontinuous with areas showing wide gaps. However, many more traces were clearly detected in the satellite data resulting in an increase of 75% in the total length of the paleoshoreline compared to the mapping by Durand (1982). Our study also shows that the remains of the paleoshoreline do not form disconnected straight lines as previously reported in Durand (1982) but rather long continuous curves (Fig. 1). Secondly, the rose diagram (Fig. 8) indicates that the directions of the paleoshoreline are far more numerous than reported in Durand (1982). Neotectonics affecting unconsolidated sediments could not display so many orientations. It is also extremely unlikely that tectonic features with so many orientations would result in such a singular arrangement

Fig. 8. Rose diagram showing the orientations of the remains of Megalake Chad shoreline mapped in this study. Linear sections of the paleoshoreline were grouped in 58 classes. The total length of the sections in each class was then reported on the rose diagram (see methodology).

aligned in a very large closed loop. Neoctectonic features are often secant. However here, none of the adjoining landforms detected in the satellite data extend beyond their junction as faults would. Thirdly, Durand (1995) suggested that on the basis of topographic maps the elongated features forming the ridge are not at constant elevation, at least in Niger. However, close examination of the best topographic maps available across the entire study area reveals that topographic information in the region is still sketchy and provides no clear evidence either way. The Landsat data, on the other hand, clearly demonstrate that all the rivers and paleorivers running through or leading to the ridge drain towards the ridge, indicating that the inside of the loop has a lower elevation than the outside. Finally, Durand (1995) reported that the Dillia and, to the west of Termit, the Dillia Achetinamou were large corridors of aeolian erosion that follow tectonic directions. The extensive parallel features, interpreted as neotectonic evidence (Durand et al., 1984), are in fact ancient parallel river channels that follow the underlying geology which is composed of a series of long parallel sand dunes oriented north-east/ south-west according to prevailing wind of the region, the Harmattan (s in Fig. 5A). In the region of rift systems that is the Lake Chad Basin, tectonics has played an important role since Mesozoic and is probably still active. Therefore, tectonics could have influenced hydrographic forms, present or past. For example, todays direction of the Dillias main channel follows that of a major rift, while the location of Lake Chad coincides with the junction of rift systems (e.g. Burke, 1976; Genik, 1992). However, a very active neotectonics during the Holocene, as required by Durand (1982), was never demonstrated and cannot be at the origin of the variety of forms and directions met in the basin. The hydrological origin for the landforms herein identified is therefore the only pertinent explanation. The limits of the paleoshoreline mapped by the present study are in rather good agreement with the map proposed by Schneider (1967), even if slightly modified in many places. Unlike todays Lake Chad, Megalake Chad was not a closed lake and the stabilisation of its maximum water level by the sill of the Mayo Kebbi allowed the

240

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

development of a well-defined shoreline system that survived in the landscape. Our findings strongly support the view that hydroclimatic fluctuations in tropical Africa during the Holocene were of great magnitude (e.g. Gasse, 2000, 2002). Following the very harsh phases of late Pleistocene, a wetter climate was marked by lacustrine deposits and Sahelian pollens. Satellite data (Figs. 2E, 2F and 4) indicate wave erosion by Megalake Chad of old dunefields that were formed or

reworked during the last major arid phase, 20 000 13 000 14C years BP (Servant and Servant, 1983; Maley, 2004). Megalake Chad reached its maximum filling for the Holocene between 7700 and 5500 14C years BP, with three major phases identified by Maley (2004) as 77007400, 71006600 and 60005500 14 C years BP. Climatic conditions deteriorated progressively after 5500 14C years BP till a relatively short but intense arid event between 4000 and 3800 14 C years BP (Maley, 1997; Gasse, 2000). The last

Fig. 9. Elements of comparison with landforms observed in the Aral Sea. Landsat TM data in the Lake Chad Basin (A and B) and in the Aral Sea (C and D). A: Landsat TM data, PCA 1st component, showing a detail of the paleoshoreline of Megalake Chad with sub-parallel marks (se). Location reported in Fig. 1. B: Landsat TM data, PCA 1st component, showing a detail of the paleoshoreline of Megalake Chad sub-parallel marks (se). Location reported in Fig. 1. C: Landsat TM data Sept. 1989, PCA 1st component, Aral Sea. The image shows sub-parallel marks (se) between the contemporary (1989) and former shorelines of the Aral Sea. These marks are paleostrandlines indicating lake level fluctuations with deposits at successive elevations. Similar sub-parallel marks are observed along Megalake Chad shoreline (A and B). D: Landsat TM data June 1987, PCA 1st component, Aral Sea. The former shoreline of the Aral Sea (r) is marked by a sharp break in the landscape between the dunefield of the Kyzyl Kum desert (d) and costal sand and mud flats (f).

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242

241

important filling of the paleolake Chad occurred between 3700 and 3000 14C years BP but did not reach the levels of the period 77005500 BP: the sandy ridge at about 288 m amsl probably corresponds to this last phase (Maley, 2004). In other close arid regions, especially in neighbouring southern Egypt and northern Sudan, the general pattern is identical (e.g. Pachur and Hoelzmann, 2000; Nicoll, 2004): the return of wetter conditions after the hyper-arid late Pleistocene was in fact a succession of episodes highly variable in rainfall distribution and duration. Large paleorivers were active during wetter periods, significantly more humid than present, and could flow into the Nile. The West Nubian paleolake reached a maximum extent of 5300 km2, for a catchment area of about 78 000 km2 (Hoelzmann et al., 2001). A comparison with another large closed lake of the world is instructive. The Aral Sea, with a known history of recent dramatic shrinkage, has left behind remains of its old shoreline and beach ridge. The remains of the former Aral Sea shoreline have remarkable similarities with that of the paleoshoreline identified from the Lake Chad Basin (Fig. 9). In particular a series of sub-parallel marks (se in Fig. 9C) corresponding to paleostrandlines marking the fluctuations of the Aral Sea with deposits at successive elevations is replicated in several locations along the paleoshoreline of Megalake Chad (se in Fig. 9A and B). A potential next step of this research is the reconstruction of hydrological conditions allowing a Megalake Chad. The climatological regimes during the last wet periods in the Lake Chad Basin are not yet clearly identified and do not correspond with a simple shift to the north of the present rainfall isohyets. Atmospheric circulation and ground conditions were significantly different from present. The estimation of hydrological budget of Megalake Chad and its whole basin in these times is therefore far from an easy task, even when assuming a temporary equilibrium for some centuries. Relative contributions of northern and southern tributaries of the lake varied with time and this variation, obvious between wet and arid phases, also existed between two wet periods comparable in terms of lake levels. Our present assessment of the expanse of Megalake Chad is a first step in the paleohydrological reconstruction of the region during the Holocene.

Acknowledgements We wish to thank the Lake Chad Basin Commission, the DHA (Ministry of Hydraulics and Water Treatment) in Chad and the DRH of Diffa (Ministry of Hydraulics) in Niger for their support. Landsat TM mosaics and Modis data were obtained from NASA. Partial funding by the French national programme ECCO-PNRH is also acknowledged. References
Burke, K., 1976. The Chad basin: an intra-continental basin. Tectonophysics 36, 197 206. Choubert, G., Faure-Muret, A., 1987. International Geological Map of Africa.1:5,000,000. Sheet no. 2. Commission for the Geological Map of the World, UNESCO, Paris. DeVogel, S.B., Magee, J.W., Manley, W.F., Miller, G.H., 2004. A GIS-based reconstruction of late Quaternary paleohydrology: Lake Eyre, arid central Australia. Palaeogeography, Palaeoclimatology, Palaeoecology 204 (12), 1 13. Durand, A., 1982. Oscillations of Lake Chad over the past 50 000 years: new data and new hypothesis. Palaeogeography, Palaeoclimatology, Palaeoecology 39, 37 53. Durand, A., 1995. Geomorphological records of neotectonics in the Lake Chad Basin: the changes in drainage pattern and the pseudo shoreline of Lake Megachad in Kadzell. Comptes Rendus de lAcademie des Sciences Paris IIa 321, 223 229. Durand, A., Fontes, J.-Ch., Gasse, F., Icole, M., Lang, J., 1984. The north-western region of Lake Chad during the Quaternary: alluvial, eolian, palustrine and lacustrine paleoenvironments. Palaeoecology of Africa 16, 215 243. Gasse, F., 2000. Hydrological changes in the African tropics since the Last Glacial Maximum. Quaternary Science Reviews 19, 189 211. Gasse, F., 2002. Diatom-inferred salinity and oxygen isotopes in Holocene waterbodies of the western Sahara and Sahel (Africa). Implications for climate and water resource variability. Quaternary Science Reviews 21, 737 767. Genik, G.J., 1992. Regional framework, structural and petroleum aspects of rift basins in Niger, Chad and the Central African Republic (C.A.R). Tectonophysics 213, 169 185. Grove, A.T., 1959. A note on the former extent of Lake Chad. The Geographical Journal 125 (1), 465 467. Hoelzmann, P., Keding, B., Berke, H., Kropelin, S., Kruse, H.J., 2001. Environmental change and hydrology: lake evolution and human occupation in the Eastern Sahara during the Holocene. Palaeogeography, Palaeoclimatology, Palaeoecology 169, 193 217. Komatsu, G.P., Brantingham, J., Olsen, J.W., Baker, V.R., 2001. Paleoshoreline geomorphology of Boon Tsagaan Nuur, Tsagaan Nuur and Orog Nuur: the Valley of Lakes, Mongolia. Geomorphology 39 (34), 83 98. Maley, J., 1981. Etudes palynologiques dans le bassin du lac Tchad et paleoclimatologie de lAfrique nord-tropicale, de 30 000 ans a `

242

M.J. Leblanc et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 230 (2006) 230242 tchadienne. Comptes Rendus de lAcademie des Sciences Paris 244 (6), 791 793. Pirard, F., 1967. Carte geologique de reconnaissance du Manga (Niger sud-oriental). 1/500 000. Notice explicative. Rapport pour la Direction des Mines et de la Geologie du Niger. BRGM, ORSTOM Publ., Paris. Rees, W.G., 2001. Physical Principles of Remote Sensing. Cambridge University Press, Cambridge. Schneider, J.-L., 1967. Evolution du dernier lacustre et peuplements prehistoriques aux Pays-Bas du Tchad. Bulletin ASEQUA 14/ 15, 18 23. Schneider, J.-L., 1994. Le Tchad depuis 25 000 ans. Geologie, Archeologie, Hydrogeologie. Masson Publ., Paris. 134 pp. Schneider, J.-L., Wolff, J.-P., 1992. Carte geologique et cartes hydrogeologiques au 1/1 500 000 de la Republique du Tchad. Memoire explicatif. BRGM, doc. 209. BRGM, Orleans. Schuster, M., Duringer, P., Ghienne, J.-F., Vignaud, P., Beauvilain, A., Mackaye, H.T., Brunet, M., 2003. Coastal conglomerates around the Hadjer el Khamis inselbergs (western Chad, Central Africa): new evidence for Lake Mega-Chad episodes. Earth Surface Processes and Landforms 28 (10), 1059 1069. Servant, M., Servant, S., 1983. Paleolimnology of an upper Quaternary endorheic lake in Chad basin. In: Carmouze, J.-P., Durand, J.-R., Leveque, C. (Eds.), Lake Chad: Ecology and Productivity of a Shallow Tropical Ecosystem. Junk Publishers. The Hague, The Netherlands, pp. 11 26. Tilho, J., 1925. Sur laire probable dextension maxima de la mer paleotchadienne. Comptes Rendus de lAcademie des Sciences Paris 181, 643 646.

lepoque actuelle. Trav. & Documents ORSTOM, vol. 129. ORSTOM Publ., Paris. Maley, J., 1997. Middle to late Holocene changes in tropical Africa and other continents: Paleomonsoon and sea surface temperature variations. In: Dalfes, H.N., Kukla, G. (Eds.), Third Millenium BC Climate Change and Old World Collapse. NATO ASI Series, Global Environmental Change. Springer-Verlag, Berlin, pp. 611 640. Maley, J., 2004. Le bassin du lac Tchad au Quaternaire recent: formations sedimentaires, paleoenvironnements et prehistoire. La question des Paleotchads. In: Renault-Miskovsky, J., Semah, A.M. (Eds.), Guide de la Prehistoire Mondiale. Publ. Errance, Paris, pp. 179 217. Nicoll, K., 2004. Recent environmental change and prehistoric human activity in Egypt and northern Sudan. Quaternary Science Reviews 23, 561 580. Olivry, J.-C., Chouret, A., Vuillaume, G., Lemoalle, J., Bricquet, J.-P., 1996. Hydrologie du Lac Tchad. Monographies hydrologiques, 12. ORSTOM, Paris. 266 pp. Pachur, H.-J., Hoelzmann, P., 2000. Late Quaternary paloecology and paloclimates of the eastern Sahara. Journal of African Earth Sciences 30 (4), 929 939. Pachur, H.-J., Rottinger, F., 1997. Evidence for a large extended paleolake in the eastern Sahara as revealed by spaceborne radar lab images. Remote Sensing of Environment 61 (3), 437 440. Pias, J., 1970. Les formations sedimentaires tertiaires et quaternaires de la cuvette tchadienne et les sols qui en derivent. Memoire de lORSTOM, 43. ORSTOM Publ., Paris, 411 pp. Pias, J., Guichard, E., 1957. Origine et consequences de lexistence dun cordon sableux dans la partie sud-ouest de la cuvette