Environmental changes in a tropical lake (Lake Abiyata, Ethiopia) during recent centuries

Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258 www.elsevier.com/locate/palaeo

Environmental changes in a tropical lake (Lake Abiyata, Ethiopia) during recent centuries Dagnachew Legesse a;b;
, F. Gasse a , O. Radakovitch a , C. Vallet-Coulomb a , R. Bonne¢lle a , D. Verschuren c , E. Gibert d , P. Barker e a

CEREGE^CNRS/UMR 6635, P.O. Box 80, 13545 Aix-en-Provence Cedex 04, France Department of Geology and Geophysics, Addis Ababa University, P.O. Box 1176, Adis Abeba, Ethiopia c Department of Biology, Ghent University, Ledegankstraat 35, 9000 Ghent, Belgium d UMR ORSAYTERRE, Universite¤ Paris-Sud, Ba“timent 504, 91405 Orsay Cedex, France Hysed, Department of Geography, Institute of Environmental and Natural Sciences, Lancaster University, Lancaster LA1 4YB, UK b

e

Received 18 May 2000; received in revised form 26 February 2001; accepted 10 July 2002

Abstract Lake Abiyata is a small, closed, saline^alkaline lake located in the central part of the Ethiopian Rift Valley, East Africa. A multi-proxy study of a sediment core, 116 cm long and with undisturbed mud^water interface, was performed to test the sensitivity of the lake system and of different proxies to the changes in climate and human activities that occurred in the catchment during the past few centuries. The 210 Pb analyses suggest that the upper 80 cm of the core represent the past 200 years. This study complements millennial-scale environmental records available in the region. The main freshwater-climatic and biological features of the modern lake system and their variations over the past decades, as known from observations, are first summarised. Results derived from individual proxies analysed along the core are then presented (successively, major physical and chemical properties of bulk sediments, diatoms and pollen). Uncertainties on the chronological framework are discussed. Major limnological stages are finally identified based on the multi-proxy interpretation of our record. Our record shows large variations in the lake water and salt balances, in the sediment sources, and in the vegetation distribution in the basin. Using our 210 Pb chronology, major changes observed in the core are tentatively compared with environmental events known from instrumental and historical records. The upper 41 cm of the core (210 Pb age: 1940^1998 AD) reveal several fluctuations in diatominferred water depth and salinity which seem to be consistent with known changes in water level. Human impact on vegetation clearly appears since about 30 years. The interval 85^41 cm suggests a period of overall water deficit. Lake Abiyata experienced episodes shallower and more saline than over the past decades, especially around 68^66 cm, 210 Pb dated at ca. 1890 AD. This level may coincide with one of the worst droughts known in the Ethiopian history during 1888^1892. The lower part of the core includes a stage (108^85 cm) of lake level much higher than today and which ended before 1800 AD. Although its base is undated so far, this stage suggests that conditions much wetter than today have prevailed in the region during at least part of the 18th century. Lake Abiyata appears to be a suitable site for a detailed environmental reconstruction over the recent past, although further work is needed to reduce the uncertainties on our record, as discussed in the conclusions. < 2002 Elsevier Science B.V. All rights reserved.

* Corresponding author. Fax: +33 4 42 97 15 95.

E-mail address: [email protected] (D. Legesse).

0031-0182 / 02 / $ ^ see front matter < 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 1 - 0 1 8 2 ( 0 2 ) 0 0 4 7 9 - 0

PALAEO 2916 8-10-02

234

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Keywords: palaeolimnology; past centuries; lake level; salinity; lake sediments; diatoms; pollen; Tropical East Africa

1. Introduction The reconstruction of environment and climate changes over the past few centuries is essential to understand the impact of natural processes and human activities on the ecosystems, and to forecast their evolution in the near future. This is especially relevant in semi-arid regions of the African tropics characterised by large interannual and decadal changes in precipitation (e.g. Hulme, 1992; Nicholson, 1996; Nicholson and Chervin, 1983), and where increasing population pressure makes areas very sensitive to the water resource variability (e.g. Servat et al., 1998). Meteorological and hydrological records can help understand the response of low-latitude regions to global climatic change (e.g. Nicholson, 1989), and can be used to analyse the sensitivity of a lake to environmental £uctuations through hydrological modelling (Bergonzini, 1997; Nicholson and Yin, 2000; Vallet-Coulomb et al., 2001). However, in many parts of tropical Africa, instrumental records cover a few decades only, and may not represent the full range of natural climate variability at longer time scales relevant to society. High-resolution (decadal to annual) proxy records, e.g. speleothems (Holmgren et al., 1999), ice (Thompson, 2000), or lake records (Barry et al., 2002; Muchane, 1996; Verschuren et al., 2000) can provide a means of assessing natural climatic changes over long periods of time and their potential links with cultural development. A crucial prerequisite for the climatic interpretation of lacustrine sediment records is to improve our understanding of the lake system under investigation. In most of the tropical lakes of Africa, high-resolution climatic reconstruction is complicated by a poor knowledge of: (i) the individualistic response of lakes to changes in climate and land use in the catchment; (ii) the sensitivity and response time of individual physical, chemical and biological indicators to changes in lake water properties; and (iii) the site-speci¢c processes governing the incorporation of environmental signals into the sediments. The aim of this paper is to

approach these questions and to test the potential of lakes lying in the Main Rift Valley of Ethiopia (Fig. 1) for a detailed climate reconstruction of the past few centuries that we do not pretend to provide here. This presents preliminary results of a multi-proxy study of a 210 Pb-dated short core taken from one of these lakes, Lake Abiyata (7‡37PN, 38‡37PE, 1578 m above sea level (asl); Fig. 1). Lake Abiyata is a small, closed, saline^ alkaline lake very sensitive to variations in climate (Street, 1979; Street-Perrott, 1982) and human activities in its catchment area (e.g. irrigation, salt exploitation ; Vallet-Coulomb et al., 2001). Instrumental records available for the past few decades have been used for hydrological modelling of the modern system (Ayenew, 1998; ValletCoulomb et al., 2001). Several studies in the region have focussed on millennial-scale variations in the monsoon strength over the past 30 000 years (e.g. Bonne¢lle and Robert, 1986; Chalie¤ and Gasse, this volume; Gasse and Street, 1978; Gillespie et al., 1983; Grove, 1975; Le Turdu et al., 1999; Le¤zine and Bonne¢lle, 1982; Street, 1979). However, little is known about environmental changes over the recent past. The environmental reconstruction presented here is based on the analyses of the major physical and chemical properties of bulk sediments, diatoms, and pollen. Uncertainties on the interpretation of individual proxies are discussed. The multi-proxy record provides evidence for large £uctuations in water depth and salinity as well as changes in vegetation cover in the basin, which can be tentatively compared with instrumental and historical records.

2. Modern setting and interannual variability 2.1. Geology and geomorphology Lake Abiyata lies in a shallow depression about 18 km long within the Ziway^Shala Basin (Fig. 1), an internal drainage basin located in the central part of the Main Ethiopian Rift Valley. The latter is a NNE^SSW structure down-faulted through

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

235

Fig. 1. Location map of the Ziway^Shala Basin in the Main Ethiopian Rift Valley (MER). The E.^M. Holocene Lake corresponds to the Early- to Mid-Holocene highstand (1670 m asl; Gasse and Street, 1978). Figures in parentheses correspond to lake level in m asl. Inset: The bathymetry of Lake Abiyata (m) with location of core E98AB05-CV.

the Ethiopian Highlands, the Somalian Plateau on the east and the Ethiopian Plateau on the west. The region has undergone a number of tectonic and volcanic episodes since the Late Miocene period (Di Paola, 1972; Woldegabriel et al., 1990). The highlands, with a relief of more than 4000 m in elevation, mainly consist of basalt, trachyte and

ignimbrite. The rift £oor (about 1700^1550 m altitude) is covered by lacustrine sediments and acid volcanic rocks (welded and unwelded tu¡, ash £ow tu¡, pumice, and obsidian-rich lava (Di Paola, 1972). Presently, the Ziway^Shala Basin contains four lakes of decreasing elevation and increasing salin-

PALAEO 2916 8-10-02

236

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

ity (data from November^December, 1998): lakes Ziway (1636 m asl; 9 m deep, 0.4 g/l), Langano (1585 m asl, 47 m deep, 1.8 g/l), Abiyata (1577 m asl, 10 m deep, 18 g/l), and Shala (1557 m asl, 266 m deep, 19 g/l). The three former lakes are of tectonic origin, while Lake Shala occupies a deep caldera. Lakes Ziway and Langano are open lakes, and over£ow toward Lake Abiyata through the Bulbula and the Horakelo rivers, respectively. Lakes Abiyata and Shala are closed. During the Early^Mid Holocene and the Late Pleistocene (e.g. 32 000^27 000 cal. yr BP) wet periods, the four lakes were united, forming one large freshwater lake which over£owed to the Awash river to the north (Gasse and Street, 1978; Street, 1979). Sediments of these large lakes mainly consist of thick diatomites which outcrop today in the basin and are intensively eroded by surface runo¡ and wind de£ation.

Fig. 2. Rainfall £uctuations at (a) Kersa station (2800 m asl; on the eastern plateau); (b) Assela station (2400 m asl; on the eastern escarpment); and (c) Langano station (1600 m asl; on the rift £oor). Data from Ethiopian Meteorological Services.

2.2. Regional climate and vegetation The Ziway^Shala Basin is characterised by a semi-arid to sub-humid type of climate with mean annual precipitation and mean annual temperature varying from 600 mm and 25‡C close to the lakes, to 1200 mm and 15‡C on the humid plateaux and escarpments, respectively (data from Ethiopian Meteorological Services Agency). The region is characterised by three main seasons. The long rainy season in the summer (July^September ; locally known as ‘kiremet’) is primarily controlled by the seasonal migration of the Inter Tropical Convergence Zone (ITCZ) which lies to the north of Ethiopia at that time. Due to the intense heating of the high plateau land, the convergence of the wet monsoonal currents from the southern Indian and Atlantic oceans brings much rain to the region (Gri⁄ths, 1972). The ‘kiremet’ rain represents 50^70% of the mean annual total (Degefu, 1987). Highlands £anking the Rift Valley intercept most of the monsoonal rainfall in the region, resulting in a strong moisture de¢cit at the rift £oor in general and near the lakes in particular. The dry period extends between October and February (known as ‘baga’), when the ITCZ lies south of Ethiopia, during which time northeasterly trade winds traversing Arabia dominate the

region. Occasional rains during this period bring 10^20% of the annual mean (Degefu, 1987). The ‘small rain’ season (known as ‘balg’; 20^30% of the annual amount) during March to May coincides with a diminution of the Arabian high as it moves towards the Indian Ocean, causing warm, moist air with a southerly component to £ow over most of the country (Gri⁄ths, 1972). As in most of the arid and semi-arid lands of the African continent (Nicholson and Chervin, 1983), interannual rainfall variability is inherently extreme in the region. Based on statistical analyses of a 100 yr precipitation record at Adis Abeba (2400 m asl, about 150 km north of the Ziway^ Shala Basin) Seleshi and Demare¤e (1995)) presented evidence for a strong interannual variability of the mean annual precipitation on the plateaux and they also observed a decline in the mean annual precipitation trend in the period between 1968 and 1985. A statistical analysis of rainfall time series of the past three decades from 16 stations lying in the Ziway^Shala Basin (to be published in detail elsewhere) shows that the interannual £uctuations in this basin are not in phase with those observed in Adis Abeba. The patterns of rainfall anomalies in the plateaux differ from those in the rift £oor (Fig. 2).

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

The vegetation in the rift valley is mainly characterised by an Acacia combretum open woodland, now extensively overgrazed (Woldu and Tadesse, 1990) whereas deciduous woodlands (Combretum, Olea europaea, Celtis, Dodonaea viscosa and Euclea) occupy the escarpments (Mohammed and Bonne¢lle, 1991). The montane forest exists between 2000 and 3000 m on the eastern Ethiopian plateaux bordering the rift and is dominated by Podocarpus gracilior (Friis, 1986). Since the 1970s, the montane forest has been heavily deforested and exotic species have been introduced (e.g. Cupressus lusitanica, Pinus patula and Eucalyptus globulus ; CADU, 1972). The Ericaceous belt extends up to 3600 m where it grades into Afroalpine vegetation. 2.3. Lake Abiyata hydrology and hydrobiology Hydro-meteorological data available for the past three decades were compiled (Ayenew, 1998; Chernet, 1998; Makin et al., 1976; ValletCoulomb et al., 2001; Ethiopian Ministry of Water Resources) for the lake basin hydrological modelling to be published elsewhere. Only the major hydrological features are summarised below. The main in£ows to Lake Abiyata are the Bulbula and Horakelo rivers, draining from lakes Ziway and Langano, respectively. The upstream lakes Ziway and Langano are mainly fed by rivers emanating from the highlands on either side of the rift. Including the Ziway and Langano subbasins, the catchment area of Lake Abiyata is about 9625 km2 . During the past three decades, the discharge of the Bulbula and Horakelo rivers (mean : 812 and 281 mm/yr, respectively) repre-

237

sented twice the direct precipitation inputs (mean: 620 mm/yr) to the lake (Table 1). Groundwater £ows from the north and intermittent runo¡ from local drainage channels during the rainy seasons are minor contributions. As a closed lake, the only signi¢cant water loss from Lake Abiyata is evaporation, although recently, the loss has been enhanced by development schemes in the catchment such as pumping of water from the lake for soda ash extraction, diversion of feeder rivers and direct use of the Lake Ziway water for irrigation. Evaporation is estimated at about 1800 mm/yr (Vallet-Coulomb et al., 2001; Table 1). The variability of annual evaporation rate is by far less than that of water inputs originating mainly from the highlands. Therefore, due to its terminal position in the drainage area, Lake Abiyata is especially susceptible to changes in rainfall in the surrounding plateaux. The lake level £uctuates at a seasonal scale of several cm with as high as 180 cm (e.g. a rise in March 1984). But these variations do not rival those observed at the interannual scale. Relative lake-level data covering di¡erent periods since 1968 have been combined with bathymetric maps, aerial photographs (Makin et al., 1976; Street-Perrott, 1982), and satellite images. The lake depth reached a maximum of ca. 13 m in 1970^72, and was ca. 7 m in 1989^90 (Fig. 3). These extreme levels correspond to a water volume of 1.575^0.541 km3 , and to a lake surface area of 213^132 km2 , respectively. Before 1968, lake level variations, reconstructed from di¡erent sources (Makin et al., 1976; Street-Perrott, 1982), showed interannual £uctuations of the same order of magnitude, with for example a high level in

Table 1 Mean hydro-meteorological data for Lake Abiyata (data from Ethiopian Ministry of Water Resources and Ethiopian Meteorological Services) Precipitation on the lake (mm) Maximum Minimum Mean Number of years

932 320 620 28

Evaporation from the lake (mm)

1800

Input via Bulbula River Input via Horakelo River (m3 /s) (m3 /s) 338 21 140 14

Evaporation value is the one estimated for the nearby Lake Ziway (Vallet-Coulomb et al., 2001).

PALAEO 2916 8-10-02

154 5 50 15

238

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

1938 comparable to that of 1972, a low level in 1956 comparable to that of 1989^90, and a level even lower in 1967. The ionic concentration of the Lake Abiyata water (Table 2) is typical of East African soda lakes (Talling and Lemoalle, 1998; Talling and Talling, 1965). High conductivity and salinity 23 mainly accounted for by alkalinity (HCO3 3 +CO3 ) þ þþ þþ and Na . Alkaline-earth (Ca and Mg ) concentrations are low because they are removed from the brine solution through the precipitation of Ca- and Mg-carbonates caused by the high alkalinity (Talling and Talling, 1965; Wood and Talling, 1988). Sparse and punctual (both in space and time) chemical data are available since 1926 (Kebede et al., 1994; this study; Table 2). Large variations in ion concentrations are related to changes in the water balance, and thus to climate variability, although water pumping over the last 20 years may also a¡ect the lake chemistry. Since 1961, measured values of the electric conductivity (Ck20 ) of surface water ranged between 30 (May 1961) and 12.4 (April 1983) mS cm31 (Table 2); 23 3 23 the ratios (HCO3 3 +CO3 )/(Cl +SO4 ) and alkaline/alkaline-earths £uctuated between 1.7^3.5 and 200^400 (by meq l31 ), respectively. Most measurements have been made during the dry season, but

a one-year study (Wodajo, 1982) also showed large seasonal £uctuations in the lake ionic concentration: from November 1980 to October 1981, the carbonate^bicarbonate alkalinity has varied from 180 to 297 meq l31 . Due to its shallow depth and intense winds, especially during the dry season, Lake Abiyata is well mixed. Super¢cial strati¢cation is generated daily by solar heating and destroyed by nocturnal cooling and mixing (Baxter et al., 1965). Mixing and windy conditions, especially in winter, commonly result in resuspension of sediments from its shallow depths and increase the non-algal turbidity (Wood and Talling, 1988). The lake has a greenish colour due to the abundance of cyanobacteria (cyanophytes) biomass in the phytoplankton (Wodajo and Belay, 1984). In March 1991, the phytoplankton was dominated by Arthrospira fusiformis (Voronich.) Kom. and Lund ( = Spirulina platensis) and a ¢lamentous blue-green algal species with a high density of heterocysts, indicating nitrogen ¢xation (Kebede et al., 1994). The chlorophyl a content (135.3 Wg. l31 at the surface ; Kebede et al., 1994) re£ected high primary productivity. Cyanophytes, which are actively photosynthesising algae, commonly form thick algal mat covering the lake surface,

Table 2 Hydrochemistry data of Lake Abiyata during the last 60 years Ref. for chemical analyses Omer^Cooper (1930)a Lo¡redo and Maldura (1941)a De Filippis (1940)a Talling and Talling (1965) Wood and Talling (1988) Von Damm and Edmond (1984) Tudorancea and Harrison (1988)a Kebede et al. (1994) 0 m Chernet (1998), p. 81 Gizaw (1996) Ayenew (1998) This work This work a

Collection date

Cations and anions in mmol/l Naþ Kþ

Ca2þ

Mg2þ

6 1940 May 1961 Mar. 1964 Jan. 1976

124.8 130.5 1.9 188.8 12.4 277.3 8.5 222.0 6.5 196.9 4.9

0.25 0.21 0.12 6 0.7 0.05 0.04

0.411 0.247 0.054 6 0.3 6 0.05 0.023

Apr. 1983

196.2

0.02

0.006

1926^

Mar. 1991 Dec. 1992 Nov. 1992 Sep. 1994 Dec. 1998 Aug. 1999

6.0

Cl3

SO32 4

Alkalinity

Salinity

Conductivity

(meq/l)

(mg/l)

(WS/cm)

42.311 42.311 0.702 39.913 91.391 2.186 51.500 11.250 53.930 0.150

81.000 210.000 166.500 138.100

63.466

165.041

1.738

80.000

8 130 8 358 8 288.4 19 380 16 200 12 960

30 000 15 800

12 390

349 430.6 9.6 345.8 10.0 204.1 5.5 161.5 4.2

0.02 0.01 0.05 0.07

0.008 0.012 0.022 0.064

103.1 119.006 4.997 84.057 13.012 61.238 1.707 46.542 1.187

In Kebede et al. (1994).

PALAEO 2916 8-10-02

315.922 34 644 260.426

pH

10.3 9.62 9.6 9.9 10.1

26 000 17 674 14 410

10.0 9.9

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

especially close to the shorelines. The diatom £ora was analysed for 15 samples taken by several limnologists from 1964 to 1998 (Gasse, 1986; Gasse et al., 1983 and unpublished). Samples from the littoral zone and marginal swamps are dominated by the periphytic, benthic taxa Anomoeoneis sphaerophora (Ku«tz.) P¢tzer, A. sculpta (Her.) Cl., A. costata (Ku«tz.) Hust., Rhopalodia gibberula (Ehr.) O. Mu«ller, Navicula elkab O. Mu«ller, N. gawaniensis Gasse, Nitzschia sigma (Ku«tz.) W. Sm, N. vitrea Norman, N. estohensis Cholnoky, N. latens Hust., and N. pusilla Grun. Phytoplankton net samples collected in November 1998 are devoid of diatoms. Other o¡shore samples (plankton and bottom mud collected in 1964, 1971, 1980) are very poor. They contain a signi¢cant number of periphytic taxa probably transported from the lake margins by mixing. Obligate planktonic ( = euplanktonic) and facultative planktonic taxa are Nitzschia sp. af. bacillum Hust., N. sp. af. liebetruthii Rab., N. sp. af. subacicularis Hust., N. sp. af. lacuum Lange-B., and rare Thalassiosira rudol¢ Bachman. Several samples also contain freshwater planktonic taxa, e.g. Aulacoseira granulata (Ehr.) Ralfs, A. agassizii (Ost.) Sim., Stephanodiscus rotula (Ku«tz.) Hendey, Cyclotella ocellata Pant., and Fragilaria Lyngb. spp., representing up to 9% of the diatom assemblages. Although strongly silici¢ed, most valves of these taxa are altered or broken, indicating that they did not live in situ. Rare Fragilaria occur in the modern, dilute water lakes Ziway and Langano and may have been river-transported to Lake Abiyata, but A. agassizii, S. astraea and C. ocellata do not live in these lakes today (Gasse, 1986). All of these taxa predominate in the Late Pleistocene and Holocene diatomites which outcrop around the lake, especially in the Bulbula River Valley. Altered specimens were found in the sediments of the Bulbula River collected in 1998. These allochthonous diatoms are obviously reworked from the nearby ancient diatomites. The fauna is poorly known (Burgis and Symoens, 1987). The zooplankton and zoobenthos are rare and poorly diversi¢ed. The macrofauna is typical of an alkaline, cyanophyte-rich lake. Primary consumers are represented by numerous cichlid ¢shes and Lesser Flamingo colonies. The

239

lake also supports a rich fauna of piscivorous birds, including pelicans.

3. Materials and methods In December 1998, 17 short sediment cores (34^ 128 cm) with undisturbed mud^water interface were collected with a rod-operated single-drive piston corer and a Kajak-corer at o¡shore stations in the four lakes of the Ziway^Shala Basin. This paper presents preliminary results on one piston core, 116 cm long, taken in 6.4 m of water depth from the southeastern part of Lake Abiyata (E98AB05-CV; hereinafter AB05; Fig. 1). Core AB05 was collected from the same location as a longer core (AB95II, 12.6 m long) recovered in 1995 using a Wright’s piston corer. The uppermost, water-saturated sediment was not recovered in this longer core, which was 14 C-dated at about 13 000 cal. yr BP at its base (Chalie¤ and Gasse, in press ; Gibert et al., 1999). The core AB05 was thus ¢rst selected to complement the longer record from core AB95II. The piston core AB05 was extruded in the ¢eld in 1 cm increments with a ¢xed-interval sectioning device (Verschuren, 1993) and transferred to Whirl-Pak1 bags for transport. The core reveals ¢ne-grained sediments throughout. The 210 Pb analyses were done on 27 samples of core AB05 (0^90 cm). The lead chronology and accumulation rate were determined by measuring the 210 Pb activity through its granddaughter product 210 Po, with 209 Po added as an internal yield tracer. Supported 210 Pb was estimated following the method of Binford et al. (1993) and was subtracted from the total 210 Pb to obtain unsupported 210 Pb. The constant rate of supply (CRS) model (Appleby and Old¢eld, 1978), which assumes that the £ux of 210 Pb to the sediment is constant through time, was used to determine the ages and sedimentation rates of the uppermost 80 cm of sediment in the core (24 210 Pb ages). In addition to 210 Pb analysis, accelerator mass spectrometry (AMS) 14 C dating was carried out for three samples. In the absence of well preserved terrestrial biological remains, 14 C dating has been attempted on total organic matter (TOM),

PALAEO 2916 8-10-02

240

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Fig. 3. Relative lake-level variation recorded from 1968 to 1997, Lake Abiyata (Data from Ethiopian Ministry of Water Resources). The dotted line shows the beginning of a sharp decline (in 1985) of the lake level due principally to accentuated human interference: arti¢cial evaporation of the lake water in a pond for soda ash exploitation, diversion of the feeder rivers for irrigation, etc.

after strong acid^alkali^acid treatment on both samples and reference blanks (Gibert et al., 1999). The sedimentary variables measured in this study are all commonly used in core description (Dearing, 1986): lithology, water content, TOM, carbonate content, and magnetic susceptibility. All content values are expressed in percentage of total dry sediment weight. Colour was noted using a Munsell colour chart (Munsell, 1954). Water content (water % by weight) and sediment dry weight were determined by drying overnight at 105‡C. The TOM was estimated from the weight loss after ignition at 550‡C (Dean, 1974). Magnetic susceptibility was measured using a Kappabridge kly-2 susceptibility meter at room temperature. All these analyses were done at 1 cm consecutive intervals. X-ray di¡raction (XDR) analysis was performed on some selected samples. Total organic carbon (TOC), total inorganic carbon (TIC) and organic carbon/nitrogen ratio (C/N) were determined for 31 levels by a CNS elemental analyser. The TIC was multiplied by 8.33 (wt of CaCO3 /wt of C) to convert it to carbonate content, assuming that the carbonate is predominantly composed of calcite as was revealed by the XRD. The C/N ratios are useful to distinguish TOM of phytoplankton origin from that produced by higher plants (e.g. Meyers, 1994; Talbot and Johannessen, 1992): aquatic plants, such as algae, lack cellulose structures

and therefore have elemental compositions that typically have C/N ratios in the range of 4^10, while plants having cellulose, such as grasses, shrubs and trees, contain proportionally more carbon and their C/N ratios range from 20 to 80. Particle-size spectra were determined for 28 selected levels based on TOM and susceptibility variations, using a Malvern automated laser-optical particle-size analyser after removal of TOM by hydrogen peroxide treatment. Each sample was analysed three times and the mean results are given in percentages of sediment dry weight for the major particle-size fractions. The diatom analysis of core AB05 was performed at every 2 cm to infer changes in water depth and water chemistry. The sediments were generally poor in diatoms. Most of the diatom valves are poorly silici¢ed and very delicate. Preliminary observations showed that chemical treatment produces signi¢cant dissolution. Therefore, slides were directly prepared from a suspension of a given weight of dry sediment in demineralised water. A known quantity of polystyrene microspheres was added to the suspension for quantitative estimates of the diatom content per unit of sediment weight (Battarbee and Kneen, 1982). Slides were mounted in Naphrax. Identi¢cation was performed using studies of East African diatoms by (Cocquyt, 1998; Gasse, 1986; Hustedt, 1949; Mu«ller, 1899), works on inland saline water diatoms (e.g. Cumming et al., 1995; Ehrlich, 1995), and general works (e.g. Krammer and Lange-Bertalot, 1986; Krammer and Lange-Bertalot, 1988; Krammer and Lange-Bertalot, 1991). A total of 138 taxa were identi¢ed. Taxonomical di⁄culties were encountered, especially with lanceolate Nitzschia species. For individual samples, 250^500 diatom valves were counted to establish the fossil assemblage composition, this number falling to 150^180 valves for samples extremely poor in diatoms. Percent and absolute abundances were illustrated for the most common taxa. The ecological interpretation was based on the diatom distribution according to physico-chemical gradients in modern East African environments (Gasse, 1986; Gasse et al., 1983; Kilham et al., 1986; Melack, 1976) and on the modern reference samples collected in 1998 in Lake Abiyata and

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

neighbouring lakes. Water electric conductivity (C), pH, cation ((Naþ +Kþ )/(Ca2þ +Mg2þ )) and 23 3 23 anion ratios ((HCO3 3 +CO3 )/(Cl +SO4 )) were inferred from diatom-based transfer functions established using the weighted averaging method applied to a large modern African diatom data set (280 samples; Gasse et al., 1995). When tested on the reference samples of Lake Abiyata (1964^ 1998; including the top of core AB05), reconstructed C-values are underestimated compared to actual values, but the direction and amplitude of changes are reproduced. Therefore, transfer functions were applied to core AB05 to obtain the direction of interannual to decadal trends. Changes in reconstructed chemical variables along core AB05 should be regarded as relative rather than absolute. Pollen analysis was performed on 28 samples collected at about 5 cm intervals, to provide information on the terrestrial vegetation in the catchment and on emergent and aquatic macrophytes in the waterbody. A classical procedure was followed in order to concentrate pollen from the sediment, dissolution with HCl, HF, and KOH, colouring with safranin and addition of glycerine. Pollen identi¢cation was carried out using the morphological description of the most common species from the Ethiopian forests (Bonne¢lle, 1971) and from the sub-desertic savanna (Bonne¢lle and Riollet, 1980). For each pollen sample, total count was above 500 grains. Major results were expressed by percentage diagrams.

4. Analytical results Analytical results on environmental proxies are ¢rst presented as a function of core depth. Because the reliability of the 210 Pb and 14 C ages should be discussed in light of environmental data, results on the chronology are presented later (Section 5). 4.1. Major sediment properties and lithostratigraphy Sediments mainly consist of olive-brown mud (116^110 cm, 104^81 cm, 27^59cm, 20^7 cm).

241

Mottled dark greyish brown organic-rich mud is observed in the interval 60^81 cm. Layers of yellowish to dark brown mud occur at 110^115 cm, 27^20 cm, and 7^0 cm (Fig. 4). The time resolution for analyses performed at every 1 cm is about 1^3 yr according to the 210 Pb chronology. Sediment water content (Fig. 4) generally exceeds 80% and is as high as 90% near the sediment^water interface and falls below 80% at 85^ 76 cm when sedimentation in£ux seems to be low. XRD analyses show that the major mineral components are clay minerals, calcite, quartz and a few amounts of feldspath. A signi¢cant proportion of non-crystallised material is attributed to reworked volcanic glass, tu¡, ash and pumice, and biogenic silica. The Ziway^Shala Basin topography and hydrographic network prevents the coarse detrital materials from the escarpments, which accumulate in the upstream lakes, to enter Lake Abiyata. Therefore, ¢ne, silt-sized particles dominate the sedimentation ( s 70%). However, the proportion of sand-sized and of clay-sized particles varies significantly and in opposite direction along the pro¢le (Fig. 4). Fine sand grains represent more than 10% of the sediment at 115^112, 75, 65, 40, 22 and 10^0 cm, with maximum ( s 20%) at the core base and top. Sandy layers indicate local erosional episodes induced by heavy storms or lake level drops, and which primarily bring unconsolidated ash and pumice fragments that cover the rift £oor around the lake. The TOC and TOM contents are generally high (Fig. 4). The proportion of organic material in the sediment is a¡ected by both input quantity and rates of mineralisation by bacterial activity. TOM inputs may derive from the primary productivity of algae or submerged macrophyte in the lake, from debris of sedges which colonise the £at, swampy areas surrounding the lake, and from terrestrial vegetation brought into the lake mainly by rivers. The highest TOM concentrations ( s 17%) occur around 112, 75^65, 25^23 and 5^ 0 cm, while minimum values (10^12%) characterise the interval 110^84 cm. Low C/N ratios ( 6 12) indicate that TOM is generally dominated by algal-derived material (Meyers, 1994). The phytoplankton origin of TOM is clear in the organic-

PALAEO 2916 8-10-02

242 D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

PALAEO 2916 8-10-02 Fig. 4. Summary diagram of lithological, geochemical, water content, and magnetic susceptibility logs for E98AB05-CV plotted against depth and 210 Pb chronology. TOM: total organic matter obtained from loss-on-ignition at 550‡C; TOC: total organic carbon using a CNS elemental analyser. The 210 Pb axis corresponds to the unsupported 210 Pb activity (j^j is the absolute error bar). On the right, major lithostratigraphic units. Ages indicated below 80 cm depth are extrapolated (see text for discussion).

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

rich top of the sequence and in the TOM-poor interval 110^84 cm (except around 94 cm). Conversely, the highest C/N values observed at 116^ 110 cm, 81^76 cm, and around 40 cm may re£ect some higher plant contribution from the lake margins or reworked terrestrial material. The carbonate concentration (8^24%; Fig. 4) is primarily due to calcite, as revealed by XRD analysis. Lake Abiyata water most likely has always been supersaturated with respect to this mineral for the period represented by the core. Enhanced calcite precipitation may result from evaporative concentration and increased alkalinity during periods of negative water balance. It may also have been favoured by algal and aquatic plant photosynthesis, as suggested by the fairly good correlation between carbonate and TOM contents (r2 = 0.68). A negative correlation (r2 = 0.75) is observed between the magnetic susceptibility and the TOM content (Fig. 4). Low magnetic susceptibility may partly be due to the large contribution of TOM and carbonate to sedimentation, having a dilution e¡ect on the susceptibility signal. At the core top and base, the lowest susceptibility values coincide with the highest sand contents. This is due to the low magnetic behaviour of the locally eroded material (ash, pumice, diatomite). Conversely, ¢ne-grained sediments between 110 and 84 cm exhibit the highest susceptibility. The susceptibility being positively related to magnetic minerogenic inputs (Thompson and Morton, 1979), these low values may re£ect a change in the source of the detrital material, from a local origin to materials having greater magnetic properties from altered basalt and soils from the escarpments. The de¢nition of major sedimentary units is primarily based on the £uctuations in TOM content, as this variable appears to be correlated to most of the other sedimentary proxies. Major changes in sedimentary variables can be summarised as follows (Fig. 4). The short interval 116^110 cm (Unit V) shows, at 112 cm, the highest TOM content and high C/N ratio, and the lowest magnetic susceptibility values observed for the whole of the core, associated with high carbonate and sand contents. This

243

suggests the proximity of the shoreline in an evaporated waterbody, receiving some detrital material from local origin. The transitional zone between Units V and IV is recorded by large and abrupt shifts in all sedimentary indicators (112^ 108 cm), indicating a major change in hydrological conditions. The sedimentary characteristics of Unit IV (110^84 cm) are all opposite to those of Unit V. The relatively ¢ne-grained sediments suggest deeper conditions than before. The high magnetic behaviour of the sediments may indicate the direct in£uence of the escarpment weathering, in contrast to all the other sections of the core. There seems to have been a relatively short-lived interruption of this event at around 100^95 cm, where small £uctuation is recorded by all the proxies. The interval 84^75 cm (Unit IIIb) suggests a return to conditions close to those of the core base and a water level drop. It di¡ers, however, from Unit V in that it has the lowest water content observed in the whole record. This interval re£ects a major change in the lake evolution during the period represented by the core. Indeed, above 75 cm, none of the sedimentary variables show the extreme values observed below this level. A reversal trend is observed in the interval 75^ 57 cm (Unit IIIa). The intervals 57^21cm (Unit II) and 21^0 cm (Unit I) both show a marked decrease and then an overall increase in TOM content. In Unit IIb (57^26 cm), the TOM, the carbonate and the susceptibility remain fairly constant up to 26 cm, above which a small rise in the TOM with a slight decrease in the susceptibility is observed in the yellowish to dark brown, sandy mud in the interval 26^21cm (Unit IIa). Unit I shows a fall in TOM content (21^7 cm, Unit Ib), followed by an interval (7^0 cm, Unit Ia) characterised by a drop in C/N ratio and magnetic susceptibility and an increase in TOM, carbonate and sand content. The very high TOM content in the uppermost centimeters may re£ect a slow rate of organic matter mineralisation. 4.2. The diatom record The diatom analysis was performed at a time

PALAEO 2916 8-10-02

244

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Fig. 5. Summary diagram of diatom analysis. Percent abundance of taxa having percentages v 2% in at least one level of core AB05. On the right, major diatom zones.

resolution of 2^6 210 Pb-years (every 2 cm). Diatom content £uctuates from ca. 7U105 to 8U107 valves/mg of dry sediment (Figs. 5 and 6). The highest concentrations and the best preservation are found at the core base below 90 cm, except at 110^108 cm and 98 cm. Very low concentrations are observed between 80 and 68 cm. Above 68 cm, concentrations are generally higher, except for sections 26^22 cm and 4^0 cm. Peaks of well preserved and concentrated diatom valves occur at 48, 20 and 8 cm, while very poor preservation is observed in samples 40^44, 36, 24 and 4 cm. Diatom concentration depends on diatom productivity, diatom preservation, sedimentation processes and the proportion of other sediment components. In alkaline water, dissolution of biogenic silica in the water column or through diagenetic processes during or after burial may considerably alter the diatom £ora, especially in shallow, £uctuating waterbodies (Barker et al., 1994). This process may have contributed to the low concentrations in certain sections of the core. However, poorly silici¢ed, delicate species such as N. latens and N. estohensis occur in some diatom-poor samples (Figs. 5 and 6), indicating that the taphocenose is still representative of the original £ora.

The dominant taxa of core AB05 (Figs. 5 and 6) were all found in our modern reference samples from Lake Abiyata, and occur in di¡erent proportions in the modern lakes of the Ziway^Shala Basin. According to their distribution in East Africa (Gasse, 1986; Gasse et al., 1983; Melack, 1976), they re£ect generally saline, alkaline conditions. Among the abundant benthic taxa, A. sphaerophora, A. sculpta, A. costata, Rhopalodia gibberula, Nitzschia vitrea, and N. sigma are found in various brine types and tolerate a large range of salinity (Gasse et al., 1995). N. elkab, N. gawaniensis, N. estohensis and N. latens characterised strongly alkaline waterbodies, the two latter being able to enter the plankton. The euplanktonic species T. rudol¢ is typical of alkaline lakes deeper than the modern Lake Abiyata, e.g. Lake Shala or Langano. The distribution of several abundant Nitzschia species is less clear, especially because of taxonomical di⁄culties. Four taxonomical groups (gr.) have been identi¢ed. The ¢rst includes euplanktonic forms close to N. subrostrata Hust. found in the plankton of lakes Shala and Langano. The second includes N. bacillum and N. af. liebetruthii, which also favour alkaline waters, but have been reported from a range of habitats

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

245

Fig. 6. Summary diagram of diatom analysis. Absolute abundance of diatom valves and of diatom taxa having percentages v 2% in at least one level of core AB05. On the right, major diatom zones.

and may best be considered facultative planktonic. N. lacuum is the third important type. The modern distribution of this planktonic species suggests an a⁄nity for fresher water than species of the bacillum group but further studies are needed of its ecological preferences. The ¢nal group, N. sp. af. lacuum, has not been assigned a name although it has formerly been included together with N. lacuum within ‘N. fonticola’ (Gasse, 1986). This taxon was found in the plankton and the benthos of several waterbodies of various salinity, alkalinity, and depth. It occurs in plankton and bottom mud samples of Lake Abiyata and reaches 8% (its highest occurrence) in the surface sample of the present core, indicating a tolerance of strongly alkaline waters. The cumulative percentages of typically benthic and euplanktonic species (Fig. 7) provide rough information on changes in water depth. The large predominance of the former category implies very shallow conditions, although the proportion of periphytic taxa is partly a¡ected by mixing processes. Conversely, very high percentages of euplanktonic taxa imply su⁄cient water depth to suppress benthic diatom productivity and transport of periphytic forms at the core site.

As in the modern reference material, many samples contain freshwater forms mixed with the dominant taxa typical of saline^alkaline waters (Figs. 5 and 6). Their absolute abundance is low, but their percentage is signi¢cant. The centric freshwater species (A. granulata, A. agassizii, S. rotula, C. ocellata) and Fragilaria spp. represent up to 15% and 10% of the taphocenose, respectively. Such unlikely conjunction of species in sedimentary cores has sometimes been explained by spatial or temporal heterogeneity (Gasse et al., 1997). In core AB05, these mixtures are attributed to allochthonous inputs, as in the modern lake. Indeed, these freshwater taxa reach their highest percentages in the uppermost 5 cm, which represents the last few years when the lake salinity^ alkalinity remained too high (Table 2) to allow these taxa to grow. Along the core, their highest percentages coincide with extremely low diatom content, suggesting unsuitable conditions for diatom growth. In the interval 80^70 cm, their altered or broken valves are commonly found in brown, organic-rich aggregates, 200^300 Wm in diameter, which can only be regarded as eroded fragments of the ancient diatomites outcropping nearby. Freshwater centric diatoms are undoubt-

PALAEO 2916 8-10-02

246

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Fig. 7. Ecological interpretation of the diatom analysis. (a) Total diatom concentration and information on diatom preservation. (b^e) Diatoms regarded as allochthonous (reworked from ancient sediments or transported by rivers; see text for discussion) are excluded. (b,c) Diatom habitats. Life form percentages of obligate planktonic (b) and benthic (c) species. The di¡erence at 100% corresponds to facultatively planktonic taxa. (d,e) Diatom-inferred conductivity and pH based on transfer functions using inverse deshrinking established by Gasse et al. (1995). Prediction errors are 0.32 for Logð10Þ conductivity; 0.48 for pH. (f) Major diatom zones.

edly reworked, but the origin of Fragilaria spp. is more questionable. At the top of the sequence, these Fragilaria are sometimes well preserved and occur in chains of up to six associated cells, making the hypothesis of their reworking less likely. Instead, they may be river-transported from lakes Ziway or Langano. However, the overall similarity of the distribution of freshwater centric forms and Fragilaria (Figs. 5 and 6) suggests that allochthonous taxa are all reworked from the ancient diatomites. Whatever their origin, we believe that the typical freshwater diatoms observed in core AB05 did not grow in situ. They have thus been excluded in the reconstruction of ecological conditions in Lake Abiyata. The precision of our diatom-based reconstruction of chemical variables (Fig. 7) is restricted by

numerous factors, besides the standard error inherent to the transfer functions: (i) some resuspension of sur¢cial sediments ; (ii) potential bias by selective dissolution in levels with poorly preserved diatoms; (iii) uncertainties on modern ecology of the numerically important small Nitzschia ; (iv) the lack of information on seasonal cycles in the diatom £ora. Although diatom samples integrate at least one year, the taphocenose rather re£ects conditions during the season of maximum diatom productivity than mean annual values. This may partly explain why, in the upper part of the core, inferred electric conductivity and pH are below their respective values measured during the dry season over the past decades. Diatoms in open water may preferentially develop during the rainy season, when the water becomes more di-

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

lute. Unfortunately, no modern diatom samples are available from Lake Abiyata for the rainy season. Despite these numerous sources of uncertainties, the overall reliability of our reconstruction is supported by the correlation between inferred conductivity and percentages of benthic diatoms (r2 = 0.87 for the interval 0^84 cm), which can be regarded as two independent indicators of changes in the balance: Water inputs minus Evaporation. According to our reconstruction (Fig. 7), Lake Abiyata remained a saline^alkaline lake throughout the analysed period. Conductivity and pH are closely and positively correlated (r2 = 0.73 for the entire pro¢le, and r2 = 0.93 for the interval 0^84 cm). On the basis of abundant taxa (Figs. 5 and 6), life forms and reconstructed chemistry (Fig. 7), the diatom pro¢le shows three major zones, subdivided into sub-zones, as follows. Zone III (116^85 cm) re£ects the highest lake level recorded in core AB05. It is characterised by a Thalassiosira^Nizschia assemblage and the maximum relative and absolute abundances of planktonic forms (T. rudol¢, N. subacicularis gr., N. lacuum). Allochthonous and benthic taxa are quasi-absent. Valves of small delicate diatoms such as N. lacuum, N. latens, N. estohensis and T. rudol¢ are in very good condition. High concentration of well preserved diatoms is probably related to high diatom productivity and rapid burial. The lake was saline^alkaline, but this zone shows the minimum diatom-inferred Cand pH-values observed for the whole core (around 104 and 86 cm). In Zone IIIc (116^107 cm), the dominant taxa (N. bacillum gr., N. subacicularis gr. and T. rudol¢) co-occur with benthic forms which survive very high salt content, e.g. N. latens, N. pusilla, N. elkab, R. gibberula and Anomoeoneis spp. This mixture indicates shortterm £uctuations in water level and ionic concentration. This sub-zone ends with the diatom-poor levels 110^108 cm, when the percentage of obligate planktonic forms shows a minimum. Zone IIIb (107^98 cm) is less diverse (N. subacicularis gr., T. rudol¢, N. bacillum gr. and N. lacuum dominant) and shows the highest diatom concentration. At 98 cm, a minor peak of benthic forms in a level with badly preserved diatoms separates

247

zones IIIb and IIIa. T. rudol¢ and N. bacillum gr. predominate in the latter. The middle section of the pro¢le (Zone II; 85^ 41 cm) is primarily de¢ned by the highest percentages of benthic forms of saline^alkaline water. N. elkab, dominant, is associated with other benthic, salt-tolerant taxa (e.g. R. gibberula, N. sigma, N. vitrea, N. gawaniensis, Anomoeoneis spp.T). This section includes the periods of the shallowest lake depths and highest salinity inferred for the whole pro¢le. The interval 85^68 cm is interpreted as a sharp regression, leading to a shallow alkaline^saline waterbody unfavourable to diatom growth. Zone IId (85^74 cm) begins with a rapid change from Nitzschia^Thalassiosira assemblages to N. elkab and a drastic fall in diatom concentration. The relative abundance of reworked diatoms rises up to a maximum (17%) at 74 cm, indicating that the erosion of ancient diatomites was reactivated. During Zone IIc (75^68 cm), the percentage of benthic forms continues to increase and inferred chemical variables reach the highest values observed in the whole core at 68 cm. The interval 68^41 cm di¡ers from the lower part of Zone II by higher diatom abundance and preservation, and low percentages of allochthonous diatoms. From 68 to 58 cm (Zone IIb), the development of euplanktonic taxa (N. subrostrata gr.) and a slight decrease in inferred conductivity re£ects a positive water-level £uctuation but the lake did not reach the great depths observed during Zone III. In Zone IIa (58^41 cm), the life-form curve and the diatom-inferred chemistry document a new phase of severe hydrological de¢cit centred around 52^46 cm. The upper section of the core (Zone I; 41^0 cm) shows the re-establishment of Nitzschia-rich assemblages. Diatom concentration and preservation tend to decrease, and percentages of allochthonous taxa become signi¢cant again. The percentages of benthic forms £uctuate between ca. 20 and 62%, roughly in phase with changes in inferred conductivity. This zone suggests large £uctuations in water depth and salinity. Zone Id (40^21 cm) starts with the abrupt return of N. bacillum gr. Salt-tolerant benthic forms remain abundant and diversi¢ed. Diatom content

PALAEO 2916 8-10-02

248

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Fig. 8. Pollen diagram for core E98AB05-CV. Percent abundance of NAP grains. On the right, major pollen zones.

and preservation fall and become very low at 26^ 21 cm (Zone Ic), making the environmental reconstruction poorly reliable there. In Zone Ib (21^ 5 cm), N. bacillum gr. is ¢rst associated with well preserved N. lacuum, suggesting an episode of fresher and deeper water, and then with N. sp. af. lacuum. The shallowest and the most concentrated water conditions in this interval are inferred from peaks of N. elkab and N. sigma that are centred around 14 cm. The uppermost subzone of this core (Zone Ia, 5^0 cm) is extremely poor in diatoms. It is characterised by variable percentages of Nitzschia with N. sp. af. lacuum reaching its maximum, the highest percentages of R. gibberula and A. sphaerophora, and of allochthonous diatoms. 4.3. The pollen record A total of 152 taxa, among which 70 belong to trees and shrubs, have been identi¢ed. Among the arboreal pollen (AP), components of three altitude vegetation zones are recognised : Juniperus, Hagenia, Ericaceae belong to the high-elevation forest close to the tree line. Podocarpus, Olea in association with Prunus, Araliaceae, Macaranga, Alchornea, Ekebergia and Myrsine africana belong

to the forest belt proper. Celtis, Euclea, Combretaceae, Acacia, Salvadoraceae, Capparidaceae are deciduous or thorny trees from mid-elevation savanna and woodland occurring today in the rift £oor or on the escarpments. Among the herbs (non-arboreal pollen, NAP), Poaceae (grasses) dominate (30^50% of the total), in agreement with the surrounding environment of the lake, which is dominated by savanna woodland. Flat lake margins occupied by herbaceous cover contain noticeable proportion of Cyperaceae (sedges) that are also expressed in the pollen diagram (10%), whereas fresh-water subaquatic Typha shows alternating phases of small pollen occurrences. The Chenopodiaceae, which normally grows on saline soil (Mohammed et al., 1995), shows here opposite frequencies to the Amaranthaceae (Riollet and Bonne¢lle, 1976), which may represent plants from stream banks such as Achyranthes. Artemisia and Anthospermum come from plants more common in the Ericaceous belt above the forest. The pollen diagrams of the Abiyata core AB05 (Figs. 8 and 9) provide new data for the last few hundred years, completing a longer record (Le¤zine and Bonne¢lle, 1982). In this study, the sampling resolution is much higher (about 5^15 yr accord-

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

249

Fig. 9. Pollen diagram for core E98AB05-CV. Percent abundance of AP grains. On the right, major pollen zones.

ing to the 210 Pb chronology) than in any of the pollen data so far available from Ethiopia (Le¤zine and Bonne¢lle, 1982; Mohammed and Bonne¢lle, 1991; Mohammed and Bonne¢lle, 1998). At such a time scale, the pattern of the pollen diagram suggests little reworking or re-mixing, and variations in the lake level can directly in£uence the relative proportion of sedge pollen. The analysis of modern lacustrine samples shows that Cyperaceae are more abundant in samples close to the shoreline than in those collected in the centre (Bonne¢lle, unpublished data). In the core AB05, the total AP varies, but in small proportion around a mean value of 35%, except between 70 and 72 cm depth, where it reaches almost 50% (Fig. 9). However, the £oristic composition of the arboreal component exhibits signi¢cant changes, based on which three distinct pollen zones are identi¢ed. Pollen Zone I (bottom to 72 cm) is characterised by the highest proportion of Podocarpus (20%), associated with Olea, whereas Ericaceae and Juniperus (5%) show opposite pattern. These are accompanied by diverse trees from savanna/ woodland. On the basis of the relative proportion between Poaceae and Cyperaceae, two subzones can be distinguished: a lower one (115^80 cm)

with higher Poaceae (50%) and decreasing Cyperaceae and Urticaceae (5%); and an upper one (80^72 cm), where Poaceae decreases (30%) and Cyperaceae increases (10%; Fig. 8). Pollen Zone II (72^25 cm) records a decrease in total AP (less than 30%), mainly due to signi¢cant decrease in trees from savanna/woodland such as Celtis, Acacia, Combretaceae, etc., and forest Olea, Podocarpus, with the exception of Juniperus and Dodonaea viscosa, the latter known to be ¢reresistant shrub. Ericaceae is regularly present, Poaceae increases (over 50%), whereas Typha has several repeated occurrences. Pollen Zone III (25^0 cm) is identi¢ed by the joint occurrences of pollen indicating human impact on the vegetation. These indicators are the introduced Plantago lanceolata type, Eucalyptus, and the cultivated Zea mays. In this pollen zone, the decreasing trend in the total AP mainly results from less pollen from montane forest, whereas those from woodland, notably Acacia, Euclea, Acalypha, recover from low frequencies in pollen zone II. 4.4. Chronology The three AMS

PALAEO 2916 8-10-02

14

C ages measured on bulk

250

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Fig. 10. Summary diagram showing some proxies analysed in the core AB05, and units and zones established from sedimentary variables, diatoms and pollen. Major limnological stages are de¢ned from the multi-proxy record. Dates plotted on the Abiyata records are 210 Pb ages (CRS age model) and should be considered as tentative. On the right, some known historical and instrumental records of the region.

organic matter (Table 3) were taken from the core top (2^3 cm) and from the TOM-rich levels at 72^ 73 cm and 112^113 cm (Fig. 4). Lacustrine TOM has been proven to yield poorly reliable 14 C dates (Barry et al., 2002; Old¢eld et al., 1997; Wohlfarth, 1996), especially because of incorporation of dissolved humus products from older soils (Olsson, 1986). However, the reliability of 14 C ages can be discussed in light of other environmental data and should be done on individual sample basis. For the core AB05, the uppermost sample is ‘modern’. It has a 14 C activity of 107.6 W 0.8 pMC, which provides an age of around 1970 AD assuming a 14 C activity of 111.7 pMC for modern atmospheric CO2 at sampling time, i.e. December 1998 (Gibert et al., 1999; Levin et al., 1992; Levin et al., 1980). Presently, the total dissolved carbon of o¡shore surface water is in equilibrium with atmospheric CO2 (Gibert et al., 1999) and thus phytoplankton material is suitable for 14 C dating. The extremely low

C/N ratio in sample 2^3 cm (Fig. 4) indicates that the TOM of this sample mainly derives from phytoplankton. The 14 C age of this sample is thus close to the real age. The other two samples gave 14 C ages of 775 W 55 BP and 1520 W 65 BP, respectively. These ages were calibrated using CALIB 4.2 (Stuiver and Reimer, 1993) and CALIB data sets summarised in Stuiver et al. (1998). Calibrated 14 C ages are about 1260 AD (72^ 73 cm) and 540 AD (112^113 cm; Table 3). The latter two ages are suspected of being biased due to at least two processes. First, reworked material contributed to sedimentation. The sedimentary and diatom remain analyses presented above show that these samples coincide with low lake level and with erosional activity. The core base is rich in sand and level 72^73 cm contains reworked diatoms entrapped in organic matter £ocks, and small, poorly preserved macrophyte debris. The relatively high C/N ratio in these two samples may re£ect some inputs of ter-

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

251

Table 3 AMS radiocarbon age determinations on total organic matter (TOM) of three samples from the Lake Abiyata core AB05 Core

Depth

Type

Analysis No. (Orsay)

(cm) E98AB05 E98AB05 E98AB05

2^3 72^73 112^113

TOM TOM TOM

H2253 H2206 H2213

14

Measured AMS 14 C age

Calibrated AMS 14 C ages

N13 CTOM

(pMC)

(yr BP)

(yr AD)

(x vs PDB)

107.6 W 0.8 90.8 W 0.6 82.7 W 0.6

Modern (3593 W 71) 775 W 60 1520 W 65

V1970 1216 (1265/1286) 434 (542/619)

to be redone 319.1 W 0.02 320.5 W 0.02

C activity

Error bars represent 1c deviation.

restrial organic material. Second, some ageing of the lake water for some sections of the core should be suspected. In the modern lake, some dead carbon is brought to the lake bottom and in interstitial water by groundwater in£ows (Gibert et al., 1999). The relative contribution of groundwater in£ux to the water balance may be enhanced during low stands, and dead carbon should a¡ect the 14 C activity of organic matter from organisms growing close to the lake bottom. This 14 C chronology based on TOM thus appears to be unsuitable for the very recent periods considered in this paper. The 210 Pb sequence covers the upper 90 cm of the core (Fig. 4). The pro¢le shows large £uctuations in the upper 20 cm of the core and then a more regular decay until 80 cm depth, where supported 210 Pb is reached (35.5 W 2 Bq kg31 ; Fig. 4). The CRS model gave 24 dates between 1806 W 24 (80 cm) and 1998 W 0.6 AD (Fig. 4). Generally, errors linked to variations of unsupported 210 Pb become important with depth and 210 Pb dates older than 1900 AD must be considered with caution. Surface sediment reworking and bioturbation may have occurred, although the abruptness of environmental changes recorded by di¡erent proxies suggests that sediment-mixing processes are negligible. Sedimentation gaps, which may explain some irregularities in the 210 Pb decay (e.g. at 40.5 cm), cannot be totally ruled out, although no evidence for desiccation has been detected along the sedimentary pro¢le. The 210 Pb ages are much younger than the 14 C ages, by about 600 yr at 72^ 73 cm. This di¡erence is large but of the same order of magnitude as that observed, for instance, in core samples from Lake Malawi between ages derived from varve counting, which are close to

the actual ages, and 14 C ages, which are ca. 880^ 1500 yr older (Barry et al., 2002). Our 210 Pb chronology, as we shall see later in this article, is in part supported by some known historical and instrumental records. This may, however, be a coincidence and the chronology should be considered with great caution. Dating of other cores correlated by the proxy records are in progress to con¢rm this time scale. Assuming that the mean sedimentation rate obtained for the upper 80 cm is similar to that of the lower section of the core AB05, the latter represents at least the upper part of the 18th century (about 1750 AD at 116 cm). Such an extrapolation only provides a reasonable timeframe for the age of the core base, but cannot be reliable to reconstruct the evolution of the lake with time, especially because sediment facies at the core base di¡ers from the rest of the core. According to the CRS model, mass and linear sedimentation rates are very high, varying between 0.07 g cm32 yr31 and 0.21 g cm32 yr31 (mean : 0.125 g cm32 yr31 ; Fig. 4) and between 0.37 and 1.1 cm yr31 (mean: 0.73 cm yr31 ), respectively.

5. Discussion The accuracy of our record from Lake Abiyata is expected to be intrinsically limited by mixing processes which induce resuspension of material from the shallow depths. However, the high number and the large magnitude of changes observed in the sedimentary and diatom pro¢les indicate that mixing processes did not drastically smooth the signals. All proxies analysed in core AB05 reveal impor-

PALAEO 2916 8-10-02

252

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

tant environmental £uctuations in the lake and in its basin. Here, we ¢rst combine the signals provided by the di¡erent proxies to better understand changes in the functioning of the lake system for the period analysed. A tentative comparison between the environmental changes identi¢ed along the core and climatic or human activity events known from the instrumental and historical records will then be proposed along with the 210 Pb chronology, assuming the latter to be correct. 5.1. Changes in the lake hydrological, biological and sedimentological processes Ages (AD) given below between brackets are based on our 210 Pb chronology and should be regarded as tentative. Although the lake remained saline^alkaline throughout the analysed period, our multi-proxy record allows us to identify major limnological stages (Fig. 10). Stage 5 (116^110; undated) is mainly characterised by the highest value of TOM and the lowest value of susceptibility of the whole record, suggesting relatively shallow but £uctuating hydrological conditions. Further discussion of this stage is limited by its incompleteness. Stage S4 (110^85 cm, except 100^95 cm; before 1800) is one of the most remarkable episodes of the whole record. It is characterised by the lowest TOM ( 6 12%) and carbonate ( 6 10%) contents, relatively ¢ne-grained, clayey sediments and high magnetic susceptibility, the highest diatom concentration (up to 8U107 valves mg31 ) and preservation, and the highest percentages of obligate planktonic forms, a relatively high AP/NAP ratio, and the absence of cyanophyte remains and of reworked diatoms. S4 is regarded as a phase of maximum lake depth which was su⁄cient to suppress benthic diatom productivity and transport of periphytic forms at the core site (Fig. 7). This implies a water level much higher than during the 1970^ 72 highstand, when our reference diatom samples (Gasse, 1986) were dominated by facultative planktonic associated with periphytic taxa. Although £uctuating, reconstructed salinity and pH exhibit the lowest values of the whole record within this interval. A deep lake of moderate sa-

linity/alkalinity is in agreement with abundant and well preserved diatom valves in ¢ne-grained sediments and low carbonate content. A negative shift in the Cyperaceae pollen centred at 100 cm (Fig. 8) indicates an increased distance from the shoreline. High percentages of wooded plant pollen suggest conditions wetter than today. Enhanced rainfall amounts may have reinforced both alteration and weathering of the basaltic escarpments, and generated £ows of the Jido River (Fig. 1), bringing higher amounts of clayey particles having high magnetic behaviour. As a whole, our results agree to conclude that S4 was a wet episode with a water level much higher than that observed over the past three decades. According to the Ethiopian Mapping Authority 1:50 000 topographic maps, a rise of the Lake Abiyata level at 1586 m asl is su⁄cient to connect Lake Abiyata with the oligosaline Lake Langano. A 1586 m level is only 5^6 m above the 1970^72 highstand (Fig. 3). A further rise up to about 1600 m asl leads to its connection with the strongly alkaline Lake Shala. Such connections may explain the speci¢city of stage S4 by inducing major changes in the physical, chemical and biological properties of the lake water. The succession in diatom assemblages suggests linkages between Abiyata and the neighbouring lakes: from 110 cm, we observe successively the predominance of N. subrostrata gr. and of T. rudol¢, abundant today in the plankton of Lake Langano and Lake Shala. The abruptness of the S5^S4 transition as observed in the sedimentary record suggests hydrological threshold mechanisms rather than a climatic change. The establishment of the connections with the neighbouring lakes provides a satisfying explanation for the abrupt shifts in all sedimentary variables, e.g. the magnetic susceptibility. Rivers supplying lakes Langano and Shala directly drain the basaltic escarpments. Their connection to Lake Abiyata may thus have directly linked this lake to the major source of detrital materials having a stronger magnetic behaviour than local materials, inducing an abrupt increase in magnetic susceptibility. If this hypothesis is correct, Lake Abiyata has had a depth of at least 20 m, when it was connected with Lake Langano. Stage S3 (85^75 cm; V1800^1848) shows a re-

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

markable increase in TOM and carbonate content and an opposite trend in magnetic susceptibility. The diatom content falls drastically. The diatom taphocenose shows a mixture of local, benthic, saline^alkaline water forms, and of reworked diatoms. This is interpreted as a drastic fall in lake level, inducing calcite precipitation and inputs of reworked material from nearby ancient diatomites due to deeper incision of the Bulbula valley and gullies. The core site became likely occupied by a swamp, with in situ submerged or emergent plants delivering large amounts of TOM. However, this evidence for a major drying trend during this interval disagrees with the interpretation of the AP/ NAP ratio (Fig. 10) terrestrial pollen record. The latter suggests that level 80^72 cm re£ects the wettest period of the whole analysed period, and that drier conditions appear suddenly at 72 cm. The cause of this discrepancy is not understood yet. Stage S2 (75^41 cm; 1848^1940) starts by a clear cut in the records of all proxies: a decrease in TOM and carbonate contents which peak at 75 cm, a sudden decrease in the percentage of (arboreal) wooded plant pollen at 72 cm, an increase in benthic diatom percentages and diatominferred salinity which both peak at 68^66 cm (ca. 1890). These peaks are interpreted as recording an extremely severe drought. Diatoms suggest periods of increasing water de¢cit from 74 to 68 cm (S2c), and again above 60 cm (S2a), separated by a wetter episode centred around 60 cm (S2b; 1900^1910). As during stage S4, the diatom-inferred deepening and freshening is associated with a decrease in TOM and higher proportion of ¢ne-grained magnetic material in the sediment. The transition S2^S1 coincides with marked changes in the algal biota, while it cannot be clearly identi¢ed in the sedimentary record. In the upper 41 cm of the core, diatoms occur in generally low number. This may re£ect lower diatom production, or lower preservation due to enhanced mixing. Facultative planktonic diatom assemblages alternate with salt-tolerant benthicdominated £ora, re£ecting unstable water and salt budgets. Two intervals, composed of yellowish to dark brown mud, S1c (26^22 cm; 1971^ 1961) and S1a (5^0 cm; 1993^1998) are character-

253

ised by high water and TOM contents, low susceptibility, and extremely low diatom content. In S1a, the predominance of cyanophytes in the modern phytoplankton (Kebede et al., 1994) may account for the high TOM content and for the quasi-absence of diatoms due to the shading e¡ect of £oating cyanophyte mats. Similar relationships between algal population dynamics and TOM content may explain the interval S1c. Human impact on vegetation and environment is clearly expressed only in the uppermost 25 cm of AB05 in the pollen record. In particular, the pollen signal of Eucalyptus, widely introduced in the region in the 1970s following a massive exploitation of the indigenous Podocarpus, Juniperus and Olea trees for timber and other purposes (CADU, 1972), appears at 20 cm. 5.2. Patterns of environmental change with time : Comparison with instrumental and historical records Although further work is needed to ascertain and re¢ne our chronology, the 210 Pb analyses suggest that the upper 80 cm of the core represents the past 200 years. In order to improve our 210 Pb chronology, information derived from the 210 Pbdated section of the core is tentatively compared with instrumental and historical data. Instrumental records in the region are available for the past three decades only. In Ethiopia, the occurrences of droughts and famines, which in many cases were caused by climatic irregularities, were compiled from several historical sources by Degefu (1987). The history of drought and famine goes back to the 11th century. In the upper section of the core AB05 (0^41 cm; 1998^1940), the 210 Pb chronology is supported by the appearance of Eucalyptus in the pollen record around 20 cm (1971). Changes in diatom-inferred water conductivity can hardly be directly compared with the measurements, which were punctual through space and time. But changes in conductivity, and the cumulative percentages of benthic diatoms, can be regarded as surrogate indicators of interannual changes in water volume. As discussed above, environmental reconstruction based on diatoms su¡ers very large uncertainties

PALAEO 2916 8-10-02

254

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

and should be regarded as relative rather than absolute. The diatom-inferred conductivity suggests periods of relatively low ionic concentration around 1998 (0 cm), 1985^88 (8^10 cm), 1971^74 (20^18 cm), 1949^52 (32^34 cm), and 1938^43 (38^40 cm). Periods of higher salinity are centred around 1992^94 (6^4 cm), 1976^79 (14^16 cm), 1968 (22 cm), and 1955^1961 (28^30 cm). The direction of these salinity £uctuations roughly follows that of changes in water level over the past three decades (Fig. 3). In the middle section of the core (85^41 cm; s 1800^1940), one of the major environmental events is a peak in the reconstructed salinity around 68^66 cm, 210 Pb-dated around 1890 using the CRS age model. This event may coincide with the most deadly historical drought, in 1888^1892 (locally remembered as the ‘kefu kan’, literally meaning the bad days) experienced in the whole of Ethiopia over the past centuries (Pankhurst, 1966). Both the big and small rains failed, and about one third of the Ethiopian population perished (Degefu, 1987). Droughts of lower amplitude followed in 1895^96, 1899^1900, and then in 1921^22 and 1932^34, which may account for the high inferred conductivity values observed below 62 cm and around 52^46 cm in core AB05. The latter drought was especially important in southern Ethiopia and induced a signi¢cant drop of the level of Lake Turkana (Degefu, 1987). Between 1800 and 1890, a series of famines/droughts occurred in the whole of Ethiopia in 1800 and 1826^ 27 and in the Shoa region (Central Ethiopia, including a portion of the Ziway^Shala Basin) in 1829 and 1835. The latter has been extremely severe (Degefu, 1987). In our record, this trend of dry years may coincide with Stage S3, interpreted as a regressive phase. No dry events were recorded in Ethiopia between the mid-16th and the beginning of the 19th centuries. The wettest episode for the investigated period is represented by the lower section of the core (Stage S4), when Lake Abiyata reached its maximum level. This wet episode may represent a large part of the 18th century according to the 210 Pb chronology used here. Evidence for increased rainfall from about AD 1670 to 1770

was recently provided by Verschuren et al. (2000) from reconstructed lake-level and salinity £uctuations of Lake Naivasha in equatorial Africa. These authors suggested that higher rainfall in tropical Africa during this period of the Little Ice Age was related with the Maunder minimum in solar radiation. The prominent 18th-century Naivasha highstand was followed by a strong lakelevel decline, and a lowstand maintained during most of the 19th century as at Lake Abiyata. This apparent consistency between the 210 Pb ages of climatic and environmental events observed in core AB05 and dates of events known from instrumental and historical records is not a guarantee to the 210 Pb chronology.

6. Conclusions Our multi-proxy study complements the instrumental records available for the past few decades and the millenial-scale environmental records already established in the region. All environmental indicators analysed in core AB05 show large variations which can be related to changes in water volume and water chemistry, in the sediment source, and in the vegetation distribution in the basin. During the analysed period, Lake Abiyata has experienced successively: (i) an episode of highstand (108^85 cm), (ii) a complex episode with water generally shallower and more saline than today (85^41 cm), including the driest event of the whole record around 68^66 cm, and (iii) a period with several £uctuations in water depth and salinity (41^0 cm) comparable to those known from instrumental records. Assuming that our 210 Pb chronology is correct, Lake Abiyata appears to have recorded known recent environmental events, e.g. the drastic drought of the end of the 19th century (around 1890 AD; 68^ 66 cm), or the deforestation and reforestation of the escarpments in the early 1970s. The episode of highstand much wetter than today ended before 1800 AD and may be coeval with part of the Little Ice Age. From our investigation, Lake Abiyata seems to be a suitable site for a detailed reconstruction of climate £uctuations over the recent past in the

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Ethiopian Rift and surrounding highlands and to analyse human impact on the lake system. Further work is needed, however, to provide a well constrained and quanti¢ed history of environmental changes in the region on the basis of lacustrine records. The main points arising from our study are as follows. (1) Our chronology should be improved and completed. The sequence was dated using 210 Pb and 14 C methods. It is evident that this chronology should be con¢rmed by 210 Pb analyses along other cores from the same lake correlated with ABO5 using environmental indicators, e.g. magnetic susceptibility, as well as from the neighbouring lakes in the basin. Furthermore, the 210 Pb chronology does not cover the entire sequence, and 210 Pb ages older than 1900 should be considered with great caution. The 14 C ages measured on TOM are much older than 210 Pb ages and are poorly reliable because some input of dead carbon to the sediment is suspected. Further 14 C dating could be attempted on isolated organic fractions, e.g. pollen grains, provided that they are not reworked, or phytoplankton material. (2) Among the analysed proxies, both the sedimentary variables and the diatoms appear to be very sensitive to interannual to century changes in limnological conditions. Diatoms are believed to re£ect £uctuations in the lake water and salt budgets, but sedimentary variables are sometimes di⁄cult to interpret alone. Interestingly, most changes recorded by diatoms are in phase with sedimentary changes, leading to a more comprehensive interpretation of the multi-proxy record. The pollen spectrum is more stable. The vegetation, especially the woody species, appears to be less sensitive to interannual £uctuations than short-lived algae. But the pollen record is of utmost importance to document longer-term natural variations and anthropogenic changes in vegetal cover which may act on the water balance of the basin. (3) Diatom-inferred reconstruction of water chemistry should be regarded as indicative rather than accurately quantitative due to the poor preservation of diatoms in some sections and our lack of knowledge on algal population dynamics. More speci¢c diatom analyses are needed on

255

this site. It is also important to use other independent sedimentogical or biological indicators, e.g. other algal remains and chironomids, to con¢rm the salinity changes derived from diatoms. (4) Lake Abiyata was selected because of its sensitivity to environmental changes in its drainage area due to its terminal position and its shallow depth. Nevertheless, comparison with records from the neighbouring lakes remains necessary to re¢ne the interpretation of the major events identi¢ed in core AB05. (5) The causes of the observed hydrological £uctuations should be understood, and the impact of natural processes and human activities on Lake Abiyata should be quanti¢ed. Natural rainfall variability may be related to rainfall anomalies during the summer monsoon, the ‘short rain’ season, or both. Human activity in the region has considerably increased over the past 25 years through river and lake water diversion and agricultural development. Hydrological modelling, constrained over the instrumental period, is in progress to analyse the response of the lake to rainfall anomalies, to estimate the amplitude of rainfall variations which may account for the observed changes in the lake volume and salinity, and to understand the impact of land use change on the overall hydrological balance of the lake. This is essential for water resources and landuse management in the Rift Valley of Ethiopia.

Acknowledgements This work was carried out under a collaboration project between the Adis Abeba University, the French ‘Ministe're des A¡aires Etrange'res’ (MAE) and the ‘Centre National de la Recherche Scienti¢que’ (CNRS). It was supported by the CNRS and a PhD grant (to D.L.) from the MAE. We sincerely acknowledge the Department of Geology and Geophysics (Adis Abeba University), for giving ¢eld work facilities and for allowing D.L. to join this project; the Ethiopian Ministry of Water Resources, and the Ethiopian Meteorological services for providing hydrological and meteorological data, and the French Embassy in Ethiopia. We thank Michel Decobert,

PALAEO 2916 8-10-02

256

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Sophie Bieda, Guy Riollet and Guillaume Buchet for technical assistance. We are grateful to K. Laird, P. Bradbury and A.M. Le¤zine for their constructive comments when reviewing the manuscript.

References Appleby, P.G., Old¢eld, F., 1978. The calculation of lead-210 dates assuming a constant rate of unsupported 210 Pb to the sediment. Catena 5, 1^8. Ayenew, T., 1998. The hydrogeological system of the Lake District basin, Ethiopia. PhD thesis, University of Amsterdam, The Netherlands. Barker, P.A., Roberts, N., Lamb, H.F., van der Kaars, S., Benkaddour, A., 1994. Interpretation of Holocene lake-level changes from diatom assemblages in Lake Sidi Ali, Morocco. J. Paleolimnol. 12, 223^234. Barry, S., Filippi, M., Talbot, M., Johnson, T., 2002. A 20,000-yr sedimentological record from Lake Malawi, East Africa: The Late-Pleistocene - Holocene transition in the southern Tropics. In: Odada E., Olago, D. (Eds.), The East African Great Lakes: Limnology, Palaeolimnology, and Biodiversity. Kluwer, Dordrecht. Battarbee, R.W., Kneen, M.J., 1982. The use of electronically counted microspheres in absolute diatom analysis. Limnol. Oceanogr. 27, 184^188. Baxter, R.M., Prosser, M.V., Talling, J.F., Wood, R.B., 1965. Strati¢cation in tropical African lakes at moderate altitudes (1,500 to 2,000 m). Limnol. Oceanogr. 10, 510^520. Bergonzini, L., 1997. Bilan hydrique de lacs du rift Est-Africain (Kivu, Tanganyka, Rukwa et Nyassa). Approche mensuelle et annuelle, essai d’interpre¤tation de la variabilite¤ inter-annuelle et des £uctuations passe¤es. PhD thesis, Universite¤ Paris XI - Orsay, Paris, France. Binford, M.W., Kahl, J.S., Norton, S.A., 1993. Interpretation of 210 Pb pro¢les and veri¢cation of the CRS dating model in PIRLA project lake sediment cores. J. Paleolimnol. 9, 275^ 296. Bonne¢lle, R., 1971. Atlas des pollens d’Ethiopie: principales espe'ces des fore“ts de montagnes. Pollen et Spores 13, 15^ 72. Bonne¢lle, R., Riollet, G., 1980. Pollens des savannes d’Afrique Orientale. CNRS, Paris. Bonne¢lle, R., Robert, C., 1986. Palaeoenvironment of Lake Abiyata, Ethiopia during the past 2000 years. Geol. Soc. Spec. Publ. 25, 253^265. Burgis, M.J., Symoens, J.J. (Eds.), 1987. African wetlands and shallow water bodies. Travaux et documents no. 211. ORSTOM, Paris. CADU, 1972. Preliminary proposal for the exploitation of the Munessa forest, Assela, Ethiopia. Chalie¤, F., Gasse, F., 2002. Late Glacial-Holocene diatom record of water chemistry and lake level change from the trop-

ical East African Rift Lake Abiyata (Ethiopia). Palaeogeogr. Palaeoclimatol. Palaeoecol. S0031-0182(02)00480-7. Chernet, T., 1998. Etude des me¤canismes de mine¤ralisation en £uorure et e¤le¤ments associe¤s de la re¤gion du Rift Ethiopien. PhD thesis, Universite¤ d’Avignon, Avignon, France. Cocquyt, C., 1998. Diatoms from the N. Basin of Lake Tanganiyka. Bibliotheca Diatomologica 39, Stuttgart. Cumming, B.F., Wilson, S.E., Hall, R.I., Smol, J.P., 1995. Diatoms from British Columbia (Canada) lakes and their relationships to salinity, nutrients, and other limnological variables. Koelz Scienti¢c Publ., Stuttgart. Dearing, J.A., 1986. Core correlation and total sediment in£ux. In: Dearing, J.A. (Ed.), Handbook of Holocene Palaeoecology and Palaeohydrology. Wiley, New York, pp. 247^270. Degefu, W., 1987. Some aspects of meteorological drought in Ethiopia. In: M.H. Glantz (Ed.), Drought and Hunger in Africa. Cambridge University Press, Cambridge, pp. 23^36. Di Paola, G.M., 1972. The Ethiopian Rift Valley (between 7‡00P and 8‡40P lat. North). Bull. Volcanol. 36, 1^44. Ehrlich, A., 1995. Atlas of the Inland-Water Diatom Flora of Israel. The Geological Survey of Israel, Jerusalem. Friis, I.B., 1986. The forest vegetation of Ethiopia. Acta Univ. Ups. Symb. Bot. Ups. 26, 31^47. Gasse, F., Street, F.A., 1978. Late quaternary lake level £uctuations and environments of the northern Rift Valley and Afar Region (Ethiopia and Djibouti). Palaeogeogr. Palaeoclimatol. Palaeoecol. 24, 279^325. Gasse, F., Talling, J.F., Kilham, P., 1983. Diatom assemblages in East Africa: classi¢cation, distribution and ecology. Hydrobiology 16, 3^34. Gasse, F., 1986. East African Diatoms: Taxonomy, Ecological Distribution. Cramer, Berlin. Gasse, F., Juggins, S., Ben Khelifa, L., 1995. Diatom-based transfer functions for inferring past hydrochemical characteristics of African lakes. Palaeogeogr. Palaeoclimatol. Palaeoecol. 117, 31^54. Gasse, F., Barker, P.A., Gell, P.A., Fritz, S.C., Chalie¤, F., 1997. Diatom-inferred salinity in palaeolakes: an indirect tracer of climatic change. Quat. Sci. Rev. 16, 547^563. Gibert, E. et al., 1999. Comparing carbonate and organic AMS-C-14 ages in Lake Abiyata sediments (Ethiopia): Hydrochemistry and paleoenvironmental implications. Radiocarbon 41, 271^286. Gizaw, B., 1996. The origin of high bicarbonate and £uoride concentrations in waters of the main Ethiopian Rift Valley, East African Rift System. J. Afr. Earth Sci. 22, 391^402. Gillespie, R., Street-Perrot, F.A., Switsur, R., 1983. Post-glacial episodes in Ethiopia have implications for climate prediction. Nature 306, 680^683. Gri⁄ths, J.F., 1972. Climates of Africa. World Survey of Climatology, Vol. 10. Elsevier. Grove, A.T., 1975. Former lake levels and climatic change in the rift valley of southern Ethiopia. Geograph. J. 141, 177^ 202. Holmgren, K., Karle¤n, W., Lauritzen, S.E., Lee-Thorp, J.A., Partridge, T.C., Piketh, S., Repinski, P., Stevenson, C., Sva-

PALAEO 2916 8-10-02

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258 nered, O., Tyson, P.D., 1999. A 3000-year high-resolution stalagmite-based record of palaeoclimate for northeastern South Africa. Holocene 9, 295^309. Hulme, M., 1992. Rainfall changes in Africa: 1931-1960 to 1961-1990. Int. J. Climatol. 12, 685^699. Hustedt, F., 1949. Exploration du Lac Albert: Su«sswasser-Diatoen, mission Damas (1935-1936), 8, Inst. des Parcs Nationaux du Congo Belge, Bruxelles, Hayez, 199 pp. Kebede, E., Mariam, Z.G., Ahlgren, I., 1994. The Ethiopian Rift Valley lakes: chemical characteristics of a salinity-alkalinity series. Hydrobiology 288, 1^12. Kilham, P., Kilham, S.S., Hecky, R.E., 1986. Hypothesized resource relationships among African planktonic diatoms. Limnol. Oceanogr. 31, 1169^1181. Krammer, K., Lange-Bertalot, H., 1986. Band 2/1: Bacillariophyceae. 1. Teil: Naviculaceae. Su«Mwasser£ora von Mitteleuropa. Gustav Fischer, Stuttgart. Krammer, K., Lange-Bertalot, H., 1988. Band 2/2: Bacillariophyceae. 2. Teil: Bacillariaceae, Epthemiaceae, Surirellaceae. Su«Mwasser£ora von Mitteleuropa. Gustav Fischer, Stuttgart. Krammer, K., Lange-Bertalot, H., 1991. Band 2/3: Bacillariophyceae. 1. Teil: Centrales, Fragilariaceae, Eunotiaceae. Su«M wasser£ora von Mitteleuropa. Gustav Fischer, Stuttgart. Le Turdu, C., Tiercelin, J.J., Gibert, E., Travi, Y., Lezzar, K.E., Richert, J.P., Massault, M., Gasse, F., Bonne¢lle, R., Decobert, M., Gensous, B., Jeudy, V., Tamrat, E., Mohammed, M.U., Martens, K., Atnafu, B., Chernet, T., Williamson, D., Taieb, M., 1999. The Ziway-Shala lake basin system, Main Ethiopian Rift: In£uence of volcanism, tectonics, and climatic forcing on basin formation and sedimentation. Palaeogeogr. Palaeoclimatol. Palaeoecol. 150, 135^177. Levin, I., Bossinger, R., Bonani, G., Francey, R.J., Kromer, B., Mu«nnich, K.O., Suter, M., Trivett, N.B.A., Wol¢, W., 1992. Radiocarbon in atmospheric carbon dioxide and methane: global distribution and trends. In: Taylor, R.E. et al. (Eds.), Radiocarbon after Four Decades. Springer, Berlin, pp. 503^518. Levin, I., Mu«nnich, K.O., Weiss, W., 1980. The e¡ect of anthropogenic CO2 and 14 C sources on the distribution of 14 C in the atmosphere. Radiocarbon 22, 379^391. Le¤zine, A.-M., Bonne¢lle, R., 1982. Diagramme pollinique holoce'ne d’un sondage du lac Abiyata (Ethiopie, 7‡42PN). Pollen et Spores 24, 463^480. Makin, M.J., Kingham, T.J., Waddams, A.E., Birchall, C.R., Eavis, B.W., 1976. Prospects for irrigation development around Zwai, Ethiopia. Land Resources Study, Land Resources Div., UK Min. Overseas Devel., Tolworth 26. Melack, J.M., 1976. Limnology and dynamics of phytoplankton in equatorial African Lakes. Duke University, Durham. Meyers, P.A., 1994. Preservation of elemental and isotopic source identi¢cation of sedimentary organic matter. Chem. Geol. 114, 289^302. Mohammed, M.U., Bonne¢lle, R., 1991. The recent history of vegetation and climate around Lake Langano. Palaeoecol. Afr. 22, 275^286. Mohammed, M.U., Bonne¢lle, R., 1998. A late glacial/Late

257

Holocene pollen record from a highland peat at Tamsaa, Bale mountains, south Ethiopia. Glob. Planet. Chang. 16^ 17, 121^129. Mohammed, U., Bonne¢lle, R., Johnson, T.C., 1995. Pollen and isotopic records in Late Holocene sediments from Lake Turkana, Kenya. Palaeogr. Palaeoclimatol. Palaeoecol. 119, 371^383. Muchane, M.W., 1996. Comparison of the isotope record in micrite, Lake Turkana, with the historical weather record over the last century. In: Johnson, T., Odada, E. (Eds.), The Limnology, Climatology and Palaeoclimatology of the East African Lakes. Gorden and Breach, pp. 431^442. Mu«ller, O., 1899. Bacillariaceen aus den Natrontha«lern von El Kab (Ober-Aegypten). Hedwigia 38, 274^321. Munsell, 1954. Soil Color Charts. Kollmorgen Ins. Corp., Newburgh. Nicholson, S.E., 1989. African drought characteristics, casual theory and global teleconnections. Int. Union Geodesy Geophys. Am. Geophys. Union 1, 79^100. Nicholson, S.E., 1996. A review of climate dynamics and climate variability in Eastern Africa. In: Johnson, T.C., Odada, E.O. (Eds.), The Limnology, Climatology and Paleoclimatology of the East African Lakes. Gordon and Breach, pp. 25^56. Nicholson, S.E., Chervin, R.M., 1983. Recent rainfall £uctuations in Africa - Interhemispheric teleconnections. In: StreetPerrott, A., Beran, M., Ratcli¡e, R., (Eds.), Variations in the Global Water Budget. D. Reidel, pp. 221^238. Nicholson, S.E., Yin, X., 2000. On the feasibility of using a lake water balance model to infer rainfall: an example from Lake Victoria. Hydrol. Sci. J. 45, 75^95. Old¢eld, F., Crooks, P.R.J., Harkness, D.D., Petterson, G., 1997. AMS radiocarbon dating of organic fractions from varved lake sediments; an empirical test of reliability. J. Paleoclimatol. 18, 87^91. Olsson, I.U., 1986. Radiometric dating. In: Berglund, B. (Ed.), Handbook of Holocene Palaeoecology and Palaeohydrology. Wiley, pp. 273^312. Pankhurst, R., 1966. The great Ethiopian famine of 18881892: A new assessment. J. Hist. Med. Appl. Sci. 21, 39^92. Riollet, G., Bonne¢lle, R., 1976. Pollen des Amaranthace¤es du bassin du lac Rodolphe, Afrique Orientale. Pollen et Spores 18, 69^92. Seleshi, Y., Demare¤e, G.R., 1995. Rainfall variability in the Ethiopian and Eritrean highlands and its links with the southern oscillation. J. Biogeogr. 22, 945^952. Servat, E., Hughes, D., Fritsch, J.M., Hulme, M. (Eds.), 1998. Water resources variability in Africa during the XXth century. IAHS Publication 252, 462 pp. Street, A.F., 1979. Late Quaternary lakes in the Ziway-Shala Basin, southern Ethiopia. PhD thesis, Cambridge University, Cambridge, UK. Street-Perrott, F.A., 1982. Twentieth century £uctuations in lake level in the Ziway-Shala basin, Ethiopia. Palaeoecol. Afr. 14, 99^110. Stuiver, M. et al., 1998. INTCAL98 radiocarbon age calibration, 24,000-0 cal BP. Radiocarbon 40, 1041^1083.

PALAEO 2916 8-10-02

258

D. Legesse et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 187 (2002) 233^258

Stuiver, M., Reimer, P.J., 1993. Extended 14 C data base and revised CALIB 3.0 14 C age calibration program. Radiocarbon 35, 215^230. Talbot, M.R., Johannessen, T., 1992. A high resolution palaeoclimatic record for the last 27,500 years in tropical West Africa from the carbon and nitrogen isotopic composition of the lacustrine organic matter. Earth Planet. Sci. Lett. 110, 23^37. Talling, J.F., Lemoalle, J., 1998. Ecological Dynamics of Tropical Inland Waters. Cambridge University Press. Talling, J.F., Talling, I.B., 1965. The chemical composition of African lake waters. Int. Rev. Ges. Hydrobiol. 50 (3), 421^ 463. Thompson, L.G., 2000. Ice core evidence for climate change in the Tropics: implications for our future. Quat. Sci. Rev. 19, 19^35. Thompson, R., Morton, D.J., 1979. Magnetic susceptibility and particle-size distribution of recent sediments of the Loch Lomond drainage basin, Scotland. J. Sediment. Petrol. 49 (3), 801^812. Vallet-Coulomb, C., Legesse, D., Gasse, F., Travi, Y., Chernet, T., 2001. Lake Evaporation estimates in tropical Africa from limited meteorological data. J. Hydrol. 245, 1^18. Verschuren, D., 1993. A light-weight extruder for accurate sectioning of soft-bottom lake sediment cores in the ¢eld. Limnol. Oceanogr. 38, 1796^1802.

Verschuren, D., Laird, K.R., Cumming, B.F., 2000. Rainfall and drought in equatorial east Africa during the past 1,100 years. Nature 403, 410^414. VonDamm, K.L., Edmond, J.M., 1984. Reverse weathering in the closed basin lakes of the Ethiopian Rift. Am. J. Sci. 284, 835^862. Wodajo, K., 1982. Comparative limnology of Lake Abiyata and Lake Langano in relation to primary and secondary production. M.Sc. Thesis, Adis Abeba University, Adis Abeba, Ethiopia. Wodajo, K., Belay, A., 1984. Species composition and seasonal abundance of zooplankton in two Ethiopian rift valley lakes - Lakes Abiyata and Langano. Hydrobiology 113, 129^136. Wohlfarth, B., 1996. The chronology of the last termination: a review of radiocarbon-dated, high-resolution terrestrial stratigraphies. Quat. Sci. Rev. 15, 267^284. Woldegabriel, G., Aronson, J.L., Walter, R.C., 1990. Geology, geochronology, and rift basin development in the central sector of the Main Ethiopia Rift. Bull. Geol. Soc. Am. 102, 439^458. Woldu, Z., Tadesse, M., 1990. The vegetation in the Lakes region of the rift valley of Ethiopia and the possibility of its recoveries. SINET Ethiopian J. Sci. 13, 97^120. Wood, R.B., Talling, J.F., 1988. Chemical and Algal relationships in a salinity series of Ethiopian inland waters. Hydrobiology 158, 29^67.

PALAEO 2916 8-10-02