Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99 www.elsevier.com/locate/palaeo

N18O in mollusk shells from Pliocene Lake Hadar and modern Ethiopian lakes: implications for history of the Ethiopian monsoon Million Hailemichael a; , James L. Aronson b , Samuel Savin a , Michael J.S. Tevesz c , Joseph G. Carter d a

c

Department of Geological Sciences, Case Western Reserve University, Cleveland, OH 44106, USA b Department of Earth Sciences, Dartmouth College, Hanover, NH 03755, USA Department of Biological, Geological and Environmental Sciences, Cleveland State University, Cleveland, OH 44115, USA d Department of Geological Sciences, University of North Carolina, Chapel Hill, NC 27599, USA Received 19 January 2001; accepted 7 June 2002

Abstract Two of the five lacustrine intervals in the largely fluvial Hadar Formation, Afar, Ethiopia, occur in the Sidi Hakoma Member deposited 3.4^3.2 Ma. In a perspective of the N18 O of 11 modern Ethiopian lakes and their shells, the N18 O of the Hadar fossil shells provide a snapshot of the nature of ancient Lake Hadar and Ethiopia’s climate in the Pliocene. Ethiopia’s modern lakes both in the Rift and on the Western Plateau are fed by drainage of Plateau rain with its well established barely negative N18 OSMOW of 31.3x. Except for the man-made Lake Koka reservoir, all other Ethiopian lakes are isotopically quite positive ranging from +5.4 to +16.0x, indicating how significant evaporation is in their water budget. Shells from lakes with extant mollusk populations are mostly in isotopic equilibrium with the N18 O and temperature of their lake water. The upper transgressive interval in the Sidi Hakoma Member is the largest one in the Formation beginning at its base with the ‘Gastropod Beds’ beach deposits. Mollusks from shell beds other than the ‘Gastropod Beds’ show more positive and more variable N18 O between shells, with internal variations within shells as much as 7x. At these times the site must have been underlain by a shallow partially isolated embayment of Lake Hadar which underwent rapid expansions and then contractions by evaporation, within the few year lifetimes of the individual mollusks. The results from the ‘Gastropod Beds’ are of most significance for interpreting the overall paleoclimate at Hadar. Their uniformly negative N18 OPDB shell values that average 36.7x represent a much less evaporated stage of Lake Hadar when its N18 OSMOW was 8x lower than the spectrum of modern lakes in Ethiopia, and indeed even 3x or more lower than average modern Plateau rain. To explain such negative values we hypothesize that the Atlantic-derived air mass component to the Ethiopian monsoon was persistently strengthened during Pliocene summers, which intensified the amount and the negative isotopic character of rainfall onto both the Afar and the Ethiopian Plateaus that drained to Lake Hadar. A similar phenomenon characterized the brief periodic pluvial episodes of the Quaternary, including the latest in the early Holocene, known as the African Humid Period. In contrast to the hot semi-desert steppe conditions of today’s

* Corresponding author. E-mail addresses: [email protected] (M. Hailemichael), [email protected] (J.L. Aronson), [email protected] (S. Savin), [email protected] (M.J.S. Tevesz), [email protected] (J.G. Carter).

0031-0182 / 02 / $ ^ see front matter C 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 1 - 0 1 8 2 ( 0 2 ) 0 0 4 4 5 - 5

PALAEO 2906 29-8-02

82

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

western Afar, the diverse abundant terrestrial fossil fauna at Hadar, including the early hominid Australopithecus afarensis, is explained by the wetter, and probably cooler, summers that persisted throughout the Late Pliocene. C 2002 Elsevier Science B.V. All rights reserved. Keywords: Hadar; mollusk shell; oxygen isotope; Ethiopia; paleoenvironment; hominid site; monsoon

1. Introduction The Pliocene sedimentary strata of the East African Rift System are rich in vertebrate fossils. Among these, the sites of Hadar (Fig. 1) and the Middle Awash in the western Afar of Ethiopia are notable for the remarkable amount and quality of their hominid fossils. The 180-m-thick Hadar Formation has produced over 90% of the known fossils of the early hominid Australopithecus afarensis, including the partial skeleton, ‘Lucy’. The formation accumulated in the late Pliocene 3.4^ 2.3 Ma mostly as the £ood plain and channel deposits of a major meandering river that was ancestral to the modern Awash River. Relative down-dropping of the central Afar since the Pliocene has caused the present Awash River and its tributaries to have cut down through and exposed its ancient deposits at the site of Hadar (Aronson and Taieb, 1981). During the Pliocene, Hadar was the distal and delta plain reach of the river near its entrance to a major lake we refer to as Lake Hadar. Lake Hadar expanded and transgressed over Hadar laying down intervals of laminated lacustrine muds and beach sands with scattered to abundant shells of mollusks (Fig. 2). The last transgression of the lake is well dated at 2.95 Ma and was followed by the sculpting of a major disconformity (Fig. 2) (Aronson et al., 1996). Renewed deposition of the uppermost 15% of the formation includes no record of the existence of the lake, and instead conglomerates appear that were deposited by steep transversely £owing braided rivers from the Western Plateau. Hadar is only 40 km east of the present-day 2-km-high escarpment, along whose faults the Afar has been dropped. In contrast to the hot, dry and mostly sterile setting of today’s Hadar in the western Afar, the very high diversity and abundance of the terrestrial vertebrate fossil fauna of the Hadar Forma-

tion together with much fewer conglomerates beneath the disconformity have led to the hypothesis that the present-day Western Escarpment may have only been at a nascent stage in the late Pliocene, 3 Ma. By this thinking westernmost Afar, including Hadar, may have been a marginal tectonic block of the plateau that only since the Pliocene has descended along the present escarpment fault to become part of the presently hot and arid Afar block (Aronson and Taieb, 1981; Bonne¢lle et al., 1987; Aronson et al., 1996). Alternatively, the climate in the Afar may have been wetter than today. Our study addresses this latter possibility via isotopic study of fossil shells of mollusks that lived in Lake Hadar about 3.2 Ma. In their extensive, but brie£y documented isotopic study of the Hadar Formation, HillaireMarcel et al. (1982) included measurements of shells from most or all of the ¢ve lacustrine intervals, along with carbonate nodules of unspeci¢ed origins. They reported that the shells they analyzed were aragonite and unaltered. Our experience mainly in the central sector of the site shows that diagenesis has extensively a¡ected shells in the beds of the lowermost and the upper three lacustrine intervals by recrystallizing the shells and ¢lling them with spar calcite. This contrasts with the unaltered conditions of the shells in the two lacustrine intervals of the Sidi Hakoma member on which we report here. To better interpret the shell isotopic data in terms of the lake environments represented, we also examined the isotopic relationships of modern shells and waters in some Ethiopian lakes on the Plateau and in the Rift. Many of the fossil and modern shells were analyzed serially via microsamples along the growth direction of the shells to determine how environmental conditions changed seasonally during an individual mollusk’s life.

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

83

Fig. 1. Location of the 11 modern Ethiopian lakes studied, and the Afar site of Hadar. Numbers 1^11 indicate lakes studied in the rift and plateau: (1) Lake Gamari; (2) Lake Hayk; (3) Lake Tana; (4) Lake Hora; (5) Lake Metehara (Beseka); (6) Lake Koka; (7) Lake Zway; (8) Lake Langano; (9) Lake Abiata; (10) Lake Shala; (11) Lake Awasa. The great early Holocene African Humid Period (AHP) expansions of the four lakes of the Zway^Shala Basin in the Main Ethiopian Rift and the four lakes in the Gamari^Abbe series in the Central Afar are indicated. The approximate present-day location of the Inter-Oceanic Con£uence (IOC), summer front between Atlantic and Indian Ocean derived air masses is shown (after Rozanski et al., 1996). This front is proposed to have shifted several 100 km eastward over the western Afar, persistently during the Pliocene and periodically during the Quaternary.

2. Samples and methods Paired samples of lake water and modern mollusk shells from 11 lakes and from the Awash River at Kereyu National Park were analyzed.

The N18 O of water samples from the 11 lakes and other relevant physical and chemical information are presented in Table 1. All of the water samples were collected in glass bottles by wading out from the lakeshores during the rainy months

PALAEO 2906 29-8-02

84

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

Fig. 2. Composite stratigraphic section of the Hadar Formation; and detailed section of the Sidi Hakoma Member. Four of the ¢ve lacustrine intervals are shown, of which the two in the Sidi Hakoma member were examined isotopically in detail here. Just after the 2.95-Myr-old BKT-2 tu¡ was deposited, a major disconformity formed at Hadar, after which there is no evidence of the existence of Lake Hadar. An additional lacustrine interval exists at the top of the Basal Member (BM).

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

of March and June, 1997. The bottles were plastic taped, para⁄n sealed and refrigerated until analysis. The mollusks from lakes Awasa, Zway, Tana and the Awash River were collected alive. Except for two Lymnaea (Radix) peregra gastropods from Lake Zway, all modern and most fossil shells analyzed belong to one of the gastropod species Melanoides tuberculata, Bellamya unicolor, or the bivalve order Unionoida, which are the most commonly used mollusk shells in African lake isotope studies for which no ‘vital e¡ect’ inter-species isotopic fractionation has been noted. Microsamples were obtained by shallow scratching with a drill along the external growth band. For the Pliocene, the N18 O values of 37 microsamples from 14 fossil shells from the Hadar Formation are presented in Table 3. The taxa analyzed are the gastropod Melanoides tuberculata, Bellamya unicolor, Cleopatra bulimoides, and the unionoid and Corbicula bivalve species. Because mollusk shell aragonite is metastable and readily alters to calcite, the absence of calcite indicates the shell’s isotopic composition is likely to be pristine. Powdered samples of all shells underwent X-ray di¡raction, calibrated for calcite sensitivity using mixtures of aragonite and calcite down to 1% calcite where the calcite 104(hkl) peak is still detectable. The absence of this peak in all our shells indicates they are essentially unaltered. The least fresh mollusk shells in the study are fossils from the two of three associated shell coquina marker limestone beds known as the ‘Gastropod Beds’. In those two beds the shells are always naturally bleached white, in contrast to shells from the intermediate bed, from which our HS samples come, which have a thin sur¢cial pinkish brown layer. Because the isotope results of the ‘Gastropod Beds’ shells turn out to be important for interpreting the isotopic nature of paleo-Lake Hadar, we assessed them further for alteration. Broken surfaces were scrutinized under the scanning electron microscope to see if any secondary recrystallization had altered the physical character of the aragonite biostructure. These observations, photographed in Hailemichael (2000), reveal minor areas ( 6 1% of the total area scanned) where textural replacement has oc-

85

curred in the naturally bleached shells. No reorganization was observed at all in the unbleached shells of the intermediate HS layer of the ‘Gastropod Beds’, nor in any of the shells of the 13A bed, all of which preserve a nacreous luster. As ampli¢ed in 5. Discussion, this low degree of alteration of these least fresh samples is within the limits of acceptability for isotopic evidence. All shells were pretreated with sodium hypochlorite, ground and vacuum roasted at 200‡C for 1 h and digested in H3 PO4 at 25‡C according to McCrea (1950). The 13 C/12 C and 18 O/16 O ratios of the evolved CO2 were related to the PDB standard through repeated analyses of the Solenhofen limestone NBS Isotopic Standard 20 with Craig’s (1957) assumed N18 O = 34.14 and N13 C = 31.06x. The N18 O values of water samples were measured according to Epstein and Mayeda (1953) relative to Standard Mean Oceanic Water (SMOW). The equation of Coplen et al. (1983) was used to relate the SMOW and PDB scales. To assess if modern aragonite shells have grown in equilibrium with existing conditions, we used Grossman and Ku’s (1986) isotopic equilibrium equation modi¢ed by Dettman (1994) to relate the measured N18 Oaragonite values to the values of N18 Owater and temperature where the aragonite could have precipitated. The equation used is: 1000 ln Karagonite3water = (166.623T‡C)/ 4.784, where Karagonite3water = (N18 Oaragonite +1000)/ (N18 Owater +1000).

3. Isotope meteorology of modern-day Ethiopia The complex meteorology of Ethiopia, and East Africa in general, is poorly understood because it is related to the seasonal passage of air mass convergence zones across a varied plateau^ rift topography spanning up to 4 km in relief (Gri⁄ths, 1972; Nicholson, 1996). Rainfall on the Western Plateau and most of Ethiopia is monsoonal, with about 75% falling in the main summer rainy season (the Ethiopian monsoon) when the Inter-Tropical Convergence Zone (ITCZ) is north of Ethiopia and the Inter Ocean Con£uence (IOC) is over the Western Plateau

PALAEO 2906 29-8-02

86

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

(Fig. 1). Today these Atlantic- and Indian Oceanderived air masses that are drawn to these convergence zones descend adiabatically 2 or more km of elevation into the Afar where they become hot and dry. The other 25% of Plateau rainfall occurs in the springtime rains when the unstable ITCZ is passing overhead. The mean annual temperature of the Western Plateau is about 15‡C compared to about 28^30‡C in western and central Afar (unpublished data from National Meteorological Services Agency of Ethiopia, 1996; Gri⁄ths, 1972). The potential for evaporation is very high, reaching to s 300 cm in the western Afar (Taieb, 1974). Today’s rain on the Western Plateau is isotopically variable, but well characterized by the 35year isotopic record of rainfall kept by the International Atomic Energy Agency (IAEA) for the Addis Ababa station (IAEA, 1996). The weighted mean N18 OSMOW of 31.3x at Addis is con¢rmed for other areas of the eastern margin of the Western Plateau by our isotopic measurements of individual rains and springs sampled during the spring and summer rains of 1997 (Hailemichael, 2000), and by earlier measurements by Schoell and Faber (1976) on spring and well waters in the same region. Though not well characterized, rainfall in the rift system is isotopically more positive. The IAEA set of six measurements from Awasa and Zway in the Main Ethiopian Rift over the summer rainy season of 1995 averaged 30.9 and 31.9x respectively (IAEA, 1996). In addition to two rain samples collected in 1997 (April and June) from Awasa (5.50 and 32.88x), we measured ¢ve samples from Dilla and Shashemene (June) in the Main Ethiopian Rift (MER). The average N18 OSMOW of rain from the total of these 13 MER rains is 30.4 U 2.3x. Unfortunately the only isotopic data for the Afar are our own spot measurements of intense night-time rains we experienced in the western Afar during the summer of 1997 at Hadar (+1.98x) and Aditu (30.46x) and in the central Afar for two spring rains at Tendaho (+3.47 and +2.19x) and for a single summer night shower at Asaita in the central Afar with a N18 OSMOW of +8.66x. The more positive charac-

ter of the rift system rain compared to Plateau rain is readily explained by: (1) the lower amount of rain in a given storm, the inverse of ‘the amount e¡ect’ (Craig, 1965; Dansgaard, 1953); and (2) more evaporation of the rain while falling through the hotter and drier air. Between the Atlantic-derived and the Indian Ocean-derived air mass sources of rain for the Ethiopian Summer monsoon, the generally held notion has been that the Atlantic’s Gulf of Guinea is the major source, particularly for the Western Plateau, with the moist Atlantic-derived air masses being drawn all the way across Africa by the Indian Sub-continental Low trough in atmospheric pressure (see discussions in Rozanski et al., 1993; Telford and Lamb, 1999; Lamb et al., 2000). But isotopic studies of modern rainfall produce a dilemma for this notion (Joseph et al., 1992). For such a long path over which much rain-out clearly occurs, Rayleigh distillation should deplete considerable amounts of 18 O from Ethiopian summer rain. Yet the well characterized 35-year weighted annual mean N18 O of rain at the IAEA station in Addis Ababa is a barely negative value averaging 31.3x. Joseph et al. (1992) explained this anomaly by proposing that it is Indian Ocean-derived air masses that are the main moisture source for today’s Ethiopian Summer monsoon, not the Atlantic ones. Recent dynamic observations of satellite cloud patterns support this and suggest a high proportion of Ethiopian summer rain, perhaps in some years greater than 50%, comes from air mass sources from the direction of the Indian Ocean (personal communication, 1999, Tesfaye Gissela, Ethiopian Meteorology O⁄ce).

4. Results 4.1. Isotopic compositions of modern Ethiopian lake and river waters The N18 O values of water samples from through-£owing Lake Tana on the Western Plateau (N18 OSMOW = +5.63x in March and +5.79x in April 1997) are higher than those of rain water samples at Bahir Dar (+5.08x in the

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

small rains of March and 33.31x in the large summer rains of July, 1997, Hailemichael, 2000) and indicate evaporative enrichment of 18 O has occurred in Lake Tana. The N18 OSMOW value of a water sample collected from Lake Hayk on the eastern margin of the Western Plateau in July, 1997 was +9.34x (Table 1), close to the value (+8.68x) reported for the same lake by Schoell and Faber, 1976. These values are very much higher than springtime and summer rain from Dese (+1.96 to 34.12x), about 50 km south, and spring water (32.55x) sampled at Tita, between Dese and Hayk (Hailemichael, 2000). This high N18 O value of the closed basin lake water is consistent with intense evaporation during the prominent Ethiopian dry season. Comparably high values (N18 OSMOW = +8.03x) occur in Lake Ashenge, in a similar tectonic setting 100 km to the north (Schoell and Faber, 1976). Eight lakes in the MER were examined in this

87

study. Lake Awasa and Lake Zway are relatively dilute while Lake Abiata is 100 times more concentrated. The N18 O values of lakes in the MER range between +5.44 (Lake Zway) and +7.98x (Lake Abiata) and approximately correlate with major anion concentrations and with conductivity (Table 1). For example, Lake Zway has low conductivity (94 WS) and Cl3 content (11 ppm) and a N18 O value of +5.44x, whereas Lake Abiata, with a conductivity of 20 500 WS and a Cl3 content of 2840 ppm, has a N18 O value of +7.98x. Lakes Langano and Shala have intermediate N18 OSMOW values of +6.84x and +7.66x respectively. Our N18 O value (+7.06x) of the large caldera Lake Awasa in July 1997 matches Leng et al.’s (1999) December 1995 values between +7.3 and +7.4x (n = 12). A divide separates the Zway^Shala basin on the north from the Awash River, where it is dammed to form the large Lake Koka hydroelectric reservoir. North of Lake Koka, in the northern sector

Table 1 Physical, chemical and isotopic character of modern Ethiopian lakes and of the Awash River examined in this study Lake

Plateau lakes Hayk Tana (sample 1) Tana (sample 2) Hayk Ashenge Rift Valley lakes Awasa Shala Langano Abiata Zway Koka Hora Afar lakes Metehara Gamari (sample 1) Gamari (sample 2) Awash River before entering Koka at Kereyu Park at Mile at Asita a b

Elevationa

Deptha

Water temperature (‡C)

F3

Cl3

SO23 4

Conductivity N18 O

(m)

(m)

(ppm)

(ppm)

(ppm)

(WS)

(SMOW)

2030 1785 1785 2150 2300

23 9 9 23 25

9.0 na na na na

26 na na na na

0.9 0.2 na na na

42 3 na na na

0 2 na na na

210 685 na na na

9.34 5.63 5.79 8.68b 8.03b

Jun 97 Mar 97 Apr 97 1977 1977

1675 1540 1580 1580 1637 1590 1770

10 250 46 14 4 9 85

8.4 9.4 8.8 9.6 8.1 8.3 8.7

24 26 25 28 25 27 24

8 213 18 211 15 3 0

27 3080 160 2844 11 25 233

0 119 13 194 2 10 10

184 770 422 20500 95 184 370

7.06 7.66 6.84 7.98 5.44 1.55 7.41

Jul Jul Jul Jul Jul Jul Jul

750 320 320

na na na

na 9.2 na

na 32 29

6.9 5.0 na

114 369 na

480 50 na

3794 685 na

6.74 15.89 16.14

Jul 97 Jun 97 Jul 97

1590 1000 410 320

na na na na

na na na na

na na na na

na na na na

na na na na

na na na na

na na na na

32.41 0.48 4.29 4.19

Jul 97 Jul 97 Jul-97 Jun 97

pH

Ethiopian Mapping Authority, 1988. Data from Schoell and Faber, 1976.

PALAEO 2906 29-8-02

Sampling date

97 97 97 97 97 97 97

88

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

of the MER, are the small closed basin lakes Hora and Metehara (Beseka) fed by drainage from the Western Plateau and with N18 O values of +7.4x and +6.74x. Our 18 O-enriched Ethiopian lake waters match values for lakes in Kenya ranging from +3.6 to +9.2x, except the highly variable Lakes Elmenteita and Magadi (Cerling et al., 1988). The most arid area we were able to visit during this study was Lake Gamari, at 320 m above sea level and the ¢rst of four lakes located in series in the center of the Afar Depression (Fig. 1). We could not reach the other lakes in the series and only gained access to the large partially isolated shallow bay at the northern end of Lake Gamari. The N18 OSMOW values and temperatures of the water samples collected on June 23 and again on July 3 1997 were +16.01x (water T = 32‡C) and +15.87x (water T = 29‡C), the highest values so far reported from any East African lake. Despite its much higher N18 O, the conductivity of Gamari water in June was 685 WS, lower than Lakes Abiata and Shala, but higher than the rest of the lakes in the MER. Because of its large headwater region and through-£owing condition, Lake Koka has the lowest N18 O (+1.55x, July 1997) value of all Ethiopian lakes studied. It has an 18 O enrichment of about 3x above that of the weighted mean Plateau rainwater (31.3x) which falls in the Awash’s headwater region. A sample of the Awash River taken the same day just upstream of Lake Koka has a N18 O of 32.4x (Table 1). Following the Awash River through the Afar in the early 1997 summer, we measured a strong progressive increase in the 18 O from +0.48x below the Awash Falls near the southern juncture of the Afar with the MER to +4.29x at Mile Farms, about 40 km downstream of Hadar; and +4.19x at Asaita, 150 km further downstream near the river’s terminus with lake Gamari in the central Afar.

Fig. 3. Measured N18 OPDB values of modern shells plotted against the N18 OSMOW from the Ethiopian lakes in which they grew. These modern shell values are contrasted with results of fossil shells from the ‘Gastropod Beds’ of the Hadar Formation. For the modern shells the derived temperatures on the equilibrium fractionation curves shown are broadly reasonable. In detail, temperatures are projected to be too warm for Lake Tana, this exception is discussed in the text. Clearly the waters of Pliocene Lake Hadar at the time the shells of the ‘Gastropod Beds’ grew were distinctly more depleted in 18 O than for any present-day lakes in Ethiopia.

4.2. Oxygen isotopic composition of modern mollusk shells

shells of live snails (Bellamya unicolor) and unionoid bivalves are somewhat out of equilibrium for precipitation of aragonite from lake water of the measured N18 OSMOW (+5.7x) at a mean monthly air temperature of 20‡C at Bahir Dar (unpublished data from National Meteorological Services Agency of Ethiopia, 1996). As shown in Fig. 3, a higher temperature of about 30‡C is calculated for the precipitation of these shells from the +5.7x water. Among possible explanations, these mollusks may have resided near a spring outlet of water less enriched in 18 O. Abundant fresh, gray, unbleached shells of Melanoides on the sediment surface of Lake Hayk showed uniformly high N18 O values between +7.58 and +8.04x (Fig. 3) which are in isotopic equilibrium with the measured N18 OSMOW of the lake water (+9.34x) and temperature (26‡C).

4.2.1. Western Plateau Lakes At Lake Tana, the N18 O (+3.3 to +4.2x) of

4.2.2. Main Ethiopian Rift Valley lakes The N18 OPDB values of whole shells of live Me-

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

lanoides collected from Lake Awasa in July, 1997 range from +5.75 to +6.28x, while two microsamples of another Melanoides shell are +6.24 and +6.36x (Table 2). These N18 O values are in equilibrium with the measured water temperature (24‡C) and N18 O (+7.06x) (Fig. 3). These data are also consistent with the mean N18 OPDB

89

value of +5.9x reported for modern Melanoides from Lake Awasa by Leng et al., 1999. At Lake Zway, two whole shells of live Radix, a species not previously reported in isotopic studies, give N18 OPDB values of +5.08 and +5.49x, in equilibrium with the water N18 O (5.44x) and the mean air temperature (21‡C). We found no

Table 2 The N18 O and N13 C values of modern Ethiopian mollusk shells Sampling locality

Genus species

Whole shell Sample #

Western Plateau Lake Tana

B. B. B. B.

unicolor unicolor unicolor unicolor

LTS2-1-W LTS2-2-W LTS2-3-W LTS2-4-W Mean S.D.

Microsamples 18

N O 3.40 3.34 3.38 3.41 3.38 0.03

13

Rift Valley Lake Awasa

location

36.21 33.98 34.70 35.23 35.03 0.94

LTS2-5-A LTS2-5-B LTS2-5-C

apex intermediate aperture Mean S.D.

3.44 3.11 3.58 3.38 0.24

34.02 33.92 32.49 33.48 0.86

LTS1-1-A LTS1-1-B LTS1-1-C LTS1-1-D

umbo mid shell mid shell margin Mean S.D.

3.90 4.83 4.35 3.65 4.18 0.52

33.42 32.84 33.45 31.89 32.90 0.73

LAWS1-5-A LAWS1-5-B

apex aperture Mean S.D.

6.36 6.24 6.30 0.08

0.29 30.92 30.32 0.86

ARS1-3-A ARS1-3-B ARS1-3-C

apex intermediate aperture Mean S.D. umbo mid shell mid shell margin Mean S.D.

0.51 30.06 30.32 0.04 0.42 30.87 30.72 31.84 30.03 30.87 0.75

32.96 33.09 33.35 33.13 0.20 34.08 35.17 35.57 35.27 35.02 0.65

M. M. M. M.

tuberculata tuberculata tuberculata tuberculata

LHS1-1-W LHS1-2-W LHS1-3-W LHS1-4-W Mean S.D.

8.04 7.58 7.56 7.87 7.76 0.23

1.60 31.91 1.68 30.29 0.27 1.71

M. M. M. M.

tuberculata tuberculata tuberculata tuberculata

LAWS1-1-W 5.75 LAWS1-2-W 5.81 LAWS1-3-W 6.25 LAWS1-4-W 6.28 Mean 6.02 S.D. 0.28 LZS1-1-W 5.49 LZS1-2-W 5.08 Mean 5.29 S.D. 0.29 ARS1-2-W 30.16 MICS2-W 30.15

30.27 31.71 30.01 0.26 30.43 0.88 30.50 31.80 31.15 0.92 34.23 34.24

Lake Zway

Radix Radix

Awash River ‘Kereyu Park’

M. tuberculata Bivalve

Awash River ‘Kereyu Park’

unionoid

N13 C

Sample #

unionoid

Lake Hyke

N18 O

N C

ARS1-1-A ARS1-1-B ARS1-1-C ARS1-1-D

Values reported are as measured (aragonite) and are not corrected to calcite values.

PALAEO 2906 29-8-02

90

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

live mollusks at Lakes Langano, Shala, Abiata, and Gamari and no shells at all at Lakes Koka, Hora and Metehara. 4.3. Fossil shells from the Sidi Hakoma Member, Hadar Formation Fig. 2 shows the stratigraphic position of the two lacustrine intervals in the Sidi Hakoma Member of the Hadar Formation whose shells are pristine enough for isotopic analysis. 4.3.1. Lower lacustrine interval The lowermost shell-bearing bed is associated with the thin Kada Meha Tu¡ (KMT), 25 m above the 3.40-Myr-old Sidi Hakoma Tu¡ (Walter and Aronson, 1993). The pelecypod shells with a nacreous luster occur at the base of the tu¡. The isotopic variability of microsamples taken within a given shell was only 0.5x for both 18 O and 13 C, but the N18 O di¡erence between the averages of each shell is +2.92x. This indicates either the two individuals washed in from separate nearby micro-environments, or they lived at di¡erent times in an environment whose isotopic character evolved with time. 4.3.2. Main lacustrine interval In the central sector of Hadar the largest lacustrine interval in the Hadar Formation is 23 m thick at the top of the Sidi Hakoma Member (Fig. 2). It thins to the west and thickens somewhat to the east to about 30 m at Ounda Hadar in the direction of the permanent location of Lake Hadar (Aronson and Taieb, 1981). The base of the interval begins with two to three thin prominent gastropod coquina limestone marker beds known as the ‘Gastropod Beds’, and ends at the top with o¡shore laminated claystones that preserve several thin bentonite tu¡s originally referred to as the ‘Triple Tu¡s’ (TT). Of these, TT-4 has feldspar crystals well dated at 3.22 U 0.04 Myr (Walter, 1993). This tu¡ and its enclosing laminated claystones with abundant ostracods and ¢sh scales are a marker horizon traceable from east to west across the 12 km breadth of the site and represent the largest westward transgression of the lake across Hadar. About

2.5 m beneath TT-4 in the central sector of Hadar, a thin sand with planar shallow cross-bedding and abundant mollusks formed as a local beach deposit between TT-1 and TT-4 (Fig. 2). This thin sand extends for about 1 km across the Kada Hadar Wadi and provided the exquisitely preserved nacreous fossil shells of the 13A sample. 4.3.2.1. Base of main lacustrine interval: the ‘Gastropod Beds’. Samples AT, 25, and HS derive from these prominent beds, which represent a series of two or three closely spaced beach coquinas. During deposition the shells were the coarsest materials available for waves to have swept up and accumulated at the shore. The lowest of the three beds represents the initial area-wide transgressive beach that formed as Lake Hadar expanded over the low £at distal £oodplain. The succeeding beds suggest that the initial transgression wavered back and forth before full lacustrine conditions occupied Hadar. Samples AT, M25 and B25 are collected from the more extensive lower and upper beds where the robust shells are always entirely bleached white. In contrast the unbleached HS shells from the less extensive middle bed are not so packed together in the sand matrix and are not bleached. We excavated the outcrop and hand selected those HS shells which best preserved the pinkish brown exterior of the shell. The mean N18 O values for individual whole shells and microsamples from all of the ‘Gastropod Beds’ are quite uniformly negative only ranging from 34.99 to 38.14x and averaging 36.7 U 1.0x (n = 17). Standard deviations of the N18 O values of microsamples of individual shells only range from 0.34 to 1.09x. This indicates that isotopic conditions were fairly uniform during the life of each shell from the ‘Gastropod Beds’, and that the N18 OSMOW of Lake Hadar was much more negative than when the mollusks below and above this important unit were living. 4.3.2.2. Upper portion of main lacustrine interval (sample 13A). Five mollusk shells (17 microsamples) of 13A were hand picked for analysis over a 100-m outcrop of the regressive beach de-

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

91

Fig. 4. Plot of N18 O of microsamples from individual Hadar fossil shells. The oldest and the youngest stratigraphic units (on the left and on the right, respectively) have mostly isotopically positive N18 O values, with large internal variations within or between shells. These represent a stage when highly varied evaporation condition prevailed at Lake Hadar. By contrast, the middle plot from the ‘Gastropod Beds’ shows shells that recorded uniformly negative N18 O values, and represents a fresh unevaporated stage of Lake Hadar, when more rain from an isotopically depleted air mass source and more cloud cover prevailed.

posit near the top of this lacustrine interval (Fig. 2). The 13A nacreous fossils include whole gastropod shells and large fragments of pelecypod shells. In contrast to the shells from the ‘Gastropod Beds’ the variability of N18 O values within individual 13A shells is large. The N18 O values of all the microsamples from all of the ¢ve individuals span a remarkably large range of 8x (32.02 to +6.33x). A good portion of this range is encompassed within the shell of a single Melanoides, ranging from 30.58 to +6.33x and within a single pelecypod shell from 30.26 to +5.92x (Fig. 4). The gastropod experienced a progressive evaporative concentration of 7x in Lake Hadar all within its approximate one year of life which began in the summer wet season (low N18 O) and ended in the dry season (high N18 O). A practically identical 18 O enrichment is record within the partial shell of the longer-lived pelecypod, but it survived a dry season and its shell continued recording a succeeding wet season. Of the 17 microsamples from the 13A collection, the mean N18 OPDB value is +1.37 U 2.3x. This positive mean value and the mean of the minor transgression of KMT are about 9x more positive than the uniformly negative 36.7 U 1.0x values recorded by the ‘Gastropod Beds’ at the beginning of this main lacustrine interval. The large internal isotopic variability of the 13A fossils

was not observed within any of the microsampled modern shells from today’s large Ethiopian lakes, but such a large internal variation was observed by Abell and Williams (1989) for shells from the small modern ephemeral Lake Lyadu in the Afar. The large cyclic variation in N18 O of the 13A mollusk shells suggests that during the late stage of the main lacustrine interval, the site of Hadar temporarily became a shallow bay that was partially isolated from the main body of Lake Hadar to the east and was strongly isotopically a¡ected by seasonal evaporation. It was probably similar to modern-day Lake Gamari’s NW bay. However, very soon after the regressive 13A beach formed near the end of this major transgressive interval, Lake Hadar brie£y re-transgressed across the entire Hadar site to deposit the ‘Ostracod Beds’ marker claystone and the TT-4 tu¡.

5. Discussion 5.1. Signi¢cance of the low N18 O in the ‘Gastropod Beds’ The rule governing the N18 O of East African lakes is that extensive dry season evaporation results in quite positive values for their waters and the shells which grow in them. Thus the uniformly

PALAEO 2906 29-8-02

92

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

Table 3 The isotopic compositions of fossil shell microsamples from the three stratigraphic lacustrine units of the Pliocene Hadar Formation Age interpolated

Stratigraphic position (m)

Genus species

Sample #

Microsample location

N18 O

N13 C

3.22 Ma

54

B. unicolor

13A1-1 13A1-2 13A1-3 13A1-4

54

M. tuberculata

13A2-5 13A2-6 13A2-7

54

Cleopatra

13A3-8 13A3-9

54

unionoid

13A4-10 13A4-11 13A4-12 13A4-13 13A4-14

54

B. unicolor

13A5-21 13A5-22 13A5-23

42.5

B. unicolor

AT-1-15 AT-1-16 AT-1-17 AT-1-18

42.5

B. unicolor

B25-1-1 B25-1-2 B25-1-3

42.5

M. tuberculata

M25-1-1 M25-1-2

40

M. tuberculata

HS-2-1 HS-2-2 HS-2-3

40 39.5

B. unicolor M. tuberculata

HS-5 HS-3-1 HS-3-2 HS-3-3

apex penultimate penultimate outer lip Mean S.D. apex penultimate outer lip Mean S.D. apex body whorl Mean S.D. outer shell outer shell outer shell outer shell outer shell Mean S.D. penultimate near outer lip body whorl Mean S.D. apex intermediate intermediate aperture Mean S.D. apex mid shell aperture Mean S.D. apex aperture Mean S.D. apex middle whorl aperture Mean S.D. whole shell apex middle whorl aperture Mean

32.02 31.02 30.13 0.81 30.59 1.21 30.58 3.32 6.33 3.02 3.46 1.52 0.78 1.15 0.52 2.13 0.15 5.92 3.60 30.26 2.31 2.55 30.20 1.47 1.49 0.92 0.97 36.15 36.17 35.67 35.49 35.87 0.34 36.36 35.83 37.69 36.63 0.96 36.58 37.70 37.14 0.79 38.14 37.39 36.91 37.48 0.62 36.68 36.29 38.41 37.77 37.49

32.38 31.06 32.69 30.92 31.76 0.90 32.43 0.41 0.90 30.37 1.80 0.89 31.54 30.33 1.72 32.90 33.55 30.65 33.44 35.53 33.21 1.75 32.86 30.87 31.43 31.72 1.03 33.98 34.32 34.02 34.27 34.15 0.17 1.57 0.32 31.30 0.20 1.44 0.75 0.71 0.73 0.03 35.78 34.30 33.60 34.56 1.11 35.79 34.60 32.83 31.46 32.96

3.27 Ma

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

93

Table 3 (Continued). Age interpolated

Stratigraphic position (m)

Genus species

Sample #

3.34 Ma

39 25

B. unicolor bivalve

HS-4 KMT19-2A KMT19-2B

3.34 Ma

25

bivalve

KMT19-2-1 KMT19-2-2 KMT19-2-3

Microsample location

N18 O

N13 C

S.D. whole shell apex margin Mean S.D. ventral margin mid-shell umbonal area Mean S.D.

1.09 34.99 0.88 1.76 1.32 0.62 2.92 2.88 2.16 2.65 0.43

1.57 34.96 34.43 34.92 34.68 0.35 36.85 37.21 37.73 37.26 0.44

Values reported are as measured (aragonite) and are not corrected to calcite values.

quite negative N18 OPDB values of the fossil shells in the ‘Gastropod Beds’ are of utmost signi¢cance for interpreting the paleoclimate of Hadar. They record a relatively rare, least evaporated stage of Lake Hadar, close to, but more positive than the N18 O of rain on the Western Plateau source region during the Late Pliocene. The shells in two of the three layers in the ‘Gastropod Beds’ show slight textural evidence, albeit small, that 6 1% of their aragonite biostructure has been texturally altered. Despite that, the following two reasons make it improbable that their shells have been isotopically reset by pedogenesis, the most common form of diagenesis to have a¡ected the formation. First, being at the base of the thickest lacustrine interval in the Formation, the ‘Gastropod Beds’ were protected from pedogenic in£uence by being 35 meters below the next higher paleosol, beneath the DD-3 sandstone (Fig. 2). Secondly, the distinctly di¡erent N13 C of the shells (32.5 U 2.4x, n = 17; Table 3) compared to that of the soil carbonates (37.3 U 1.1x, n = 19; Hailemichael, 2000) rules out that the two have undergone isotopic exchange. 5.2. The N18 O of Lake Hadar at the time of the ‘Gastropod Beds’ ; and N18 O of Plateau and Afar rain during the Pliocene The least evaporated condition of Lake Hadar recorded by the ‘Gastropod Beds’ was 12x lower in 18 O than that recorded by the most positive

microsample we have observed among the 13A shells (+6x) that grew later during the same lacustrine interval. It is logical that the lowest N18 O values would have occurred in the very initial stage of the largest lacustrine interval in the Formation. These shells grew when the lake was expanding rapidly westward and the rate of input of un-evaporated, low N18 O river water to the lake most exceeded the rate of evaporation. The temperature of lake water is a necessary input for determining the N18 O of the water from the N18 O of the shell, using Dettman’s (1994) equation. More intense cloud cover and rainfall which caused the more negative N18 O of rain during the Pliocene would have also lowered the Pliocene temperatures of the Afar relative to today’s. If we assume the mean temperature for Pliocene Lake Hadar to be about 5‡C less than today’s mean air temperature (30‡C), then the average N18 Oaragonite (36.7x) value in the ‘Gastropod Beds’ would indicate a N18 OSMOW of the lake water to be about 35x; or if the temperature was 30‡C, then the lake water will have a 34x value (Fig. 3). Considering the upstream location of Lake Koka with a N18 O value of 1.55x, we can take its 3x evaporative enrichment relative to the isotopic composition of modern Plateau rain (31.3x) as a minimum which would have prevailed for the much further transport downstream in the Afar to Lake Hadar during the Pliocene. For example, we observe a 6x increase in N18 O for the Awash River’s whole Afar

PALAEO 2906 29-8-02

94

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

route from the Plateau into the central Afar today (Table 1). Thus one may infer that Pliocene rains on the Western Plateau were at least 3x more negative than the 35x value inferred from the ‘Gastropod Beds’ for Lake Hadar during the Pliocene, i.e., a N18 OSMOW of 38x. This value is much less than the 31.3x weighted mean value for Plateau rain today. The inferred N18 O (38x) for rain on the Western Plateau during the Pliocene compares with a N18 O of 35x for Pliocene rain in the Afar itself, calculated with similar assumptions about temperature from the N18 O of paleosol carbonates in the Hadar Formation by Hailemichael (2000). Thus, in a similar fashion as for Plateau rain, Afar rain in the Pliocene (35x) was much lower than today’s value of about +2x as approximated from our few spot measurements of modern Afar rain. That is, both the Plateau and the Afar experienced rain about 6^7x lower during the Pliocene than today in each region.

6. Origin of the low N18 O rain in the Pliocene Ethiopia Only about 1x of the 6^7x lowering of N18 O of Pliocene rain in Ethiopia from today’s value can be accounted by the formation of the low N18 O polar icecaps during the Quaternary. The remaining 5x decrease in the N18 O value of the Pliocene rain both in the Plateau and the Afar settings compared to today’s can be accounted for by some combination of these three factors: (1) increased proportion of rain derived from air mass moisture sources more depleted by Rayleigh distillation in Pliocene than today’s sources; (2) increased amounts of rain per storm, the ‘amount e¡ect’ (Dansgaard, 1953); and (3) reduced evaporation potential. All three of these are argued below to have occurred and to have synergistically interacted, especially once (1) was brought about. To explain the negative isotopic character of the Pliocene rain that supplied Lake Hadar during the time of the ‘Gastropod Beds’, it is helpful to look at the best near-modern analog. This nearmodern analog is in the very center of the Afar,

but not as it is today, rather as it was only 9^6 ka during the early Holocene pluvial period known as the African Humid Period (AHP) (deMenocal et al., 2000). Then the analog area matched not only the sedimentology, but also the climate and the isotopic meteorology of Pliocene Hadar. Its explainable meteorology can in turn be adopted as an explanation for the meteorology of Hadar in the Pliocene. In the early Holocene, summer heating of the Northern Hemisphere maximized to values of 8% more insolation than today due to cycles in Earth’s orbital parameters (Overpeck et al., 1996). More intense summer insolation deepened the East Saharan atmospheric low which in turn strengthened the summer African Southwest monsoon and brought Atlantic-derived moisture much further north than today. The increased rainfall greened the Sahara (Petit-Maire, 1990) and ¢lled the lakes of the Nubian Paleo-lake Basin in what is today’s hyperarid eastern Sahara of northwest Sudan (Hoelzmann et al., 2000). These early Holocene rainwaters of undisputed Atlantic derivation were clearly ¢ngerprinted (in lacustrine and riverine carbonates and in fossil groundwaters) by a distinctly negative N18 O values (Abell and Hoelzmann, 2000; Rodrigues et al., 2000; Thorweihe et al., 1990), as to be expected from their far transport and increased intensity. The great rise in the level of the lakes in the Ethiopian Rift and the Afar (Gasse and Street, 1978) is as equally a de¢ning episode of the AHP as the greening of the Sahara, but because of Ethiopia’s complex meteorology, its meteorological causes have only been addressed peripherally. Our meteorological explanation for the increased early Holocene rainfall of both the Plateaus and Afar in Ethiopia is a simple extension of what happened in the eastern Sahara. Just as a deepened Saharan Low of the early Holocene pulled moist Atlantic-derived air masses northeastward toward the eastern Sahara and shifted the Sahelian rain belts northward (Ritchie and Haynes, 1987), it is logical that the even deeper Tibetan Low (10 millibars lower today), responsible for strengthening the Indian Southwest monsoon (Overpeck et al., 1996), pulled these same moist, isotopically depleted, Atlantic-derived air

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

masses east-northeastward over Ethiopia. As detailed below, the 18 O-depleted early Holocene carbonates in the Afar of various origins suggest the Atlantic-derived component of air masses £owed as far east as the Afar in higher proportions than today. 7. An Early Holocene environmental and isotopic analog of Hadar A depositional analog of the Hadar Formation is today’s low £at distal £oodplain and delta plain of the Awash River at its terminus with Lake Gamari, the ¢rst of the series of four lakes in the central Afar (Aronson and Taieb, 1981). But the hot and dry setting, the low ecological productivity and diversity make it an otherwise inappropriate analog. Gasse (1977), Gasse et al. (1974), and Gasse and Street (1978) have documented that during the early Holocene about 9^6 ka, the level of the central Afar lake system from Lake Gamari to Lake Abbe rose a remarkable 150 m higher in elevation than today’s surface of Abbe. All four lakes coalesced into one great lake that expanded across the Asaita Plain to an area 13-fold that of today (Gasse and Street, 1978) and several times the area of Hadar. Also the level of the Zway^ Shala lakes in the MER rose 80 m, coalesced to over£ow the divide northward into the Awash (Gasse and Street, 1978), and further augment the discharge of Plateau rainfall into the Afar. Not only was the central Afar region an excellent depositional and climatic analog for Hadar during the AHP, but evidence also suggests that the increased rainfall on the Plateau and in the Afar during the AHP had an isotopically negative character similar to that evidenced by the ‘Gastropod Beds’ for the Pliocene. This evidence comes from two previous isotopic studies on early Holocene mollusk shells from the Afar lake systems. Gasse et al. (1974) measured N18 O of 11 14 C-dated shells from various stages of the expanded AHP Abbe lake system. Their N18 OPDB values average 32.1x (n = 11) with one sample as low as 35.0x. Because evaporative concentration of 18 O is so prevalent in East African lakes, it is unlikely that any particular shell sample would

95

catch a paleo-lake at its freshest, least evaporative and lowest N18 O stage. At 25‡C the lowest observation of 35.0x corresponds to a N18 OSMOW value of about 33x for the lake water, far lower than present-day Lake Gamari’s +16x. The negative isotopic compositions show that the increased AHP rainfall on the Ethiopian Plateaus which fostered the lake expansions in the MER and Afar (Gasse and Street, 1978) was isotopically more negative than today. Probably the evaporative potential of the Afar was less, as well. Abell and Williams (1989) also measured shells with negative N18 O from an early Holocene sediments at Lake Beseka in the southern Afar close to the Western Escarpment and from small spring deposits at the base of the Southern Escarpment of the Afar with the Eastern Plateau, both directly fed by Plateau run-o¡. At about 25‡C, the resulting N18 OSMOW of the water when these individual mollusks lived would have been 34 to 31x, much lower than the +6.7x of today’s Lake Beseka (Table 1). Again, the enhanced AHP rainfall on the plateaus whose run-o¡ fed these two small lakes was undoubtedly isotopically more negative than today. Evidence presented elsewhere indicates even the AHP rain in the Afar itself had a much more negative N18 O than today’s approximate +2x value. This evidence comes from modern Afar soils near Hadar whose soil calcite nodules give 14 C dates that indicate the nodules formed during the AHP (Hailemichael, 2000). These soil carbonates have a mean N18 OPDB value of 36.5x. At 25‡C these value would translate to soil waters that was about 34x (i.e., about 5^6x more depleted in 18 O than today). If plateau rain during the AHP was depleted in 18 O by a comparable amount, going from a present-day value of about 31.3x to about 36 or 37x in the early Holocene. This is close to the 38x that we independently deduced for Plateau rain during the Pliocene from the ‘Gastropod Beds’ data.

8. Meteorological hypothesis of the causes for the low N18 O rain in the Pliocene The intensi¢ed summer Indian Sub-continent

PALAEO 2906 29-8-02

96

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

Low of the early Holocene (Overpeck et al., 1996) can be hypothesized to have pulled both components of the Ethiopian monsoon eastward. During today’s summer, the Indian Ocean- and Atlanticderived air masses converge in the northeast trending IOC front, whose location (Fig. 1) is only very approximately known to hover in summer over the western part of the Western Plateau (Nicholson, 1996). It seems logical that by pulling stronger on both components of the Ethiopian monsoon that the intensi¢ed Indian low would have shifted their IOC front eastward. We speculate a shift of perhaps a few hundred kilometers to bring the IOC over the western Afar. The moist, far-traveled, Rayleigh-depleted Atlantic air masses would then have reached the high eastern shoulder of the Western Plateau, the escarpment with the Afar and the western Afar itself. This shift of the IOC would have resulted in dramatically increased amounts of rain both along the eastern shoulder of the Plateau that drained to the ancestral Awash River, and also on the western Afar itself. The con£uence of the Atlantic and Indian Ocean air masses in the Afar would have forced the relatively less stable, more humid, Atlantic air masses (Nicholson, 1996) to have risen over the drier Indian Ocean air masses in storms self-perpetuated by the release of latent heat of precipitation. The increased proportions of the negative N18 O ¢ngerprinted Atlantic component of rain falling in more intense storms on the Plateau headwaters of the ancestral Awash River explains the low N18 O of the ‘Gastropod Beds’. Further, the increased summer cloud cover, rainfall, and vegetation cover in the Afar itself that would have accompanied the eastward shift of the IOC that would have reduced the Afar mean temperature from today’s high mean value of about 30‡C, down perhaps to about 25‡C and reduced the potent ability of evaporation to increase the N18 O of Lake Hadar. Thus, as brie£y suggested by Hillaire-Marcel et al. in their 1982 paper, one only has to go back a few thousand years to the AHP in the central Afar to ¢nd an excellent depositional analog, and also an excellent climatic and isotopic analog for the Hadar Formation. Although indeed it is possible that the tectonic relief created by down-

dropping the Afar may have been less accentuated 3 Ma, there is no need to invoke such accentuated relief to explain the aridi¢cation that has a¡ected Hadar since the Pliocene. Rather, we propose the Pliocene climate at Hadar was very similar to the pluvial climate in the Afar during the early Holocene only 6^9 ka, when the tectonic situation was no di¡erent than today’s. The abundant and diverse terrestrial vertebrate fauna in the main part of the Hadar Formation beneath the disconformity argues that an enhanced monsoon was a persistent feature of the Pliocene. However, one can speculate that the ¢ve transgressions of Lake Hadar may have been caused by the cyclic peaking of the Earth’s orbital factors superimposed upon a persistently strong Ethiopian Monsoon of the Pliocene. They also may have been tectonically induced. Despite the persistent appearance of wetter summer at paleo-Hadar compared to today, the dry season must have been pronounced as shown by the cyclic variation to quite positive N18 O values within the 13A shells, and by the accumulation of carbonate nodules in the formation’s many paleosols. Nevertheless, even in the dry season the environment at Hadar would have been a very habitable refuge because of the year-round presence of river, lake, shore and wetland environments to store the ample summer water.

9. Other possible causes of the low N18 O in Pliocene rain There may have been other causes of the isotopically depleted rain of the enhanced Ethiopian monsoon during the Pliocene and the AHP that acted instead of, or in concert with the increased Atlantic-derived component hypothesized here. These include: (1) Hadar having possibly been at a higher elevation (Bonne¢lle et al., 1989); and/or (2) a strengthened Indian Ocean-derived component to the Ethiopian monsoon. For example, Indian Ocean air mass sources may have been pumped harder over Ethiopia toward an intensi¢ed, relatively closer Saharan Low before being diverted east toward the even stronger atmospheric Tibetan Low. Such Indian Ocean sources of

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

rainfall could have experienced lower N18 O by the ‘amount e¡ect’ and by Rayleigh distillation as these air masses mounted the Western Escarpment with the Afar. This e¡ect would have been reinforced in the Pliocene if the Hadar structural block was an elevated part of the Plateau that had not yet dropped into the Afar. As regards Pliocene Afar, the recent tectonic analysis at the other end of the Indian Ocean by Cane and Molnar (2001) suggests that higher sea surface temperatures (SST) may have existed for the Indian Ocean then. They propose that warm south Paci¢c waters used to come into the Indian Ocean and were closed o¡ about 4^3 Ma by the northward tectonic movement of New Guinea into the Indonesian seaway. Such Indian Ocean warming, should it have occurred, may have been superimposed upon the already globally warmed oceans of the pre-glacial world so as to feed more Indian Ocean-derived, isotopically depleted, storms into the Afar and the plateaus. The warmer SSTs of either one or both of the Atlantic/Indian oceans would account for the long persistence of the wetter summer climate in Ethiopia during the Pliocene, compared to its brief episodic occurrences in the Quaternary.

10. Conclusions The late Pliocene Hadar Formation accumulated in the western Afar mostly as the distal £ood and delta plain sediments of the ancestral Awash River. Lake Hadar transgressed westward across the site ¢ve times. On the basis of the isotopic results of the lacustrine shell zones of the Sidi Hakoma Member in a context of the isotopic hydrology of modern Ethiopia, we conclude the following. (A) Evaporation strongly enriches the 18 O content of the 11 modern lakes, except for the through-£owing Lake Koka. (B) The beach ‘Gastropod Beds’ laid down at the start of the largest lacustrine interval captured a record of the least evaporated stage of Lake Hadar equivalent to a N18 OSMOW value of 35x. This water was derived from Pliocene Plateau rain of at least 38x, much lower than to-

97

day’s average (31.3x). This lower N18 O is comparable to that for rain in the Afar inferred from isotopic studies of Hadar Formation paleosol carbonates to be presented elsewhere. (C) Near the end of this lacustrine interval when the site became a partially isolated shallow bay of the lake, shells have much more positive N18 O value with dramatic cyclic internal variations to values as high as +6x. Evaporation in the Pliocene dry season must have been pronounced like today. But the mosaic ecotone nature of the Pliocene depositional setting included many wet sub-environments for annually storing the much larger summer supply of fresh water through the dry season. This explains the abundant, diverse fossil vertebrates in the formation and the long stability and success there of the hominid Australopithecus afarensis. (D) The best depositional, meteorological and environmental analog to the Pliocene Hadar is the Lake Gamari Plain in the central Afar itself, as it was just 9^6 ka, during the AHP. The early Holocene expansions of the Ethiopian Rift and Afar lakes have been recognized as a de¢ning episode of the AHP comparable to the ‘greening’ of the Sahara. We argue that the cause of the Ethiopian part of the AHP was due to the strengthened Tibetan Low that pulled the moist isotopically depleted Atlantic-derived air mass component of the summer Ethiopian Monsoon as far east as the Afar. By analogy the negative N18 O rainfall of the Plateau and Afar throughout the Pliocene could have originated similarly. However, the high paleoecological productivity throughout the ¢rst half million years of the Hadar Formation means that the Ethiopian Monsoon was persistently strong as opposed to periodically so in the Quaternary.

Acknowledgements We thank the Ethiopian Geological Survey for permits and logistic support in collecting the modern waters and mollusks. We are grateful to the Director, Ketema Tadesse, for logistic support and encouragement and for his dedication to improving knowledge of Ethiopia’s resources. For

PALAEO 2906 29-8-02

98

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99

support at Hadar, we thank all our colleagues at the Institute of Human Origins (IHO), Tesfaye Yemane, Carl Vondra, William Kimbel, Don Johanson, Gerry Eck, and Kay Reid. We are especially indebted to Robert Walter for his generous sharing of his knowledge of the Hadar Formation. Linda Abel gave expert help in the isotope lab at CWRU. Tenesa Mamecha, Mes¢n Dubale and Zerihun Tsegaye helped collect mollusks from Lake Tana, Lake Hayk and Lake Awasa. Financial support for this study was provided by an exploratory grant to J.L.A. at CWRU from the National Science Foundation (Anthropology). M.H. was generously supported as a graduate assistant by Geological Sciences at CWRU. For their role in writing up this paper, J.L.A. and M.H. were supported by Dartmouth’s Earth Sciences Department. Taking full responsibility for the interpretations presented, we are grateful for thorough constructive reviews by FrancXoise Gasse.

References Abell, P.I., Hoelzmann, P., 2000. Holocene Palaeoclimates in northwestern Sudan: stable isotope studies on molluscs. Global Planet. Change 26, 1^12. Abell, P.I., Williams, M.J., 1989. Oxygen and carbon isotope ratios in gastropod shells as indicators of paleoenvironments in the Afar region of Ethiopia. Palaeogeogr. Palaeoclimatol. Palaeoecol. 74, 265^278. Aronson, J.L., Taieb, M., 1981. Geology and Paleogeography of the Hadar Hominid Site, Ethiopia. In: Rapp, G., Vondra, C. (Eds.), Hominid Sites and their Geological Settings. AAAS Selected Symposium 63, pp. 165^195. Aronson, J.L., Vondra, C., Yemane, T., Walter, R., 1996. Characteristics of the disconformity in the upper part of the Hadar Formation. Geol. Soc. Am. Abstracts with Programs (N. Central Section) 28. Bonne¢lle, R., Vincens, A., Buchet, G., 1987. Palynology, stratigraphy and paleoenvironment of a Pliocene hominid site (2.9^3.3 m.y.) at Hadar, Ethiopia. Palaeogeogr. Palaeoclimatol. Palaeoecol. 60, 249^281. Cane, M., Molnar, P., 2001. Closing of the Indonesian seaway as a precursor to east African aridi¢cation around 3^4 million years ago. Nature 411, 157^162. Cerling, T.E., Bowman, J.R., O’Neil, J.R., 1988. An isotopic study of a £uvial-lacustrine sequence: The Plio-Pleistocene Koobi Fora sequence, East Africa. Palaeogeogr. Palaeoclimatol. Palaeoecol. 63, 335^356.

Coplen, T., Kendall, C., Hopple, J., 1983. Comparison of stable isotope samples. Nature 302, 236^238. Craig, H., 1965. The measurement of oxygen isotope paleotemperatures. In: Stable Isotopes in Oceanographic Studies and Paleotemperatures, Spoleto, July 26^30. Craig, H., 1957. Isotopic standards for carbon and oxygen and correction factors mass-spectrometric analysis of carbon dioxide. Geochim. Cosmochim. Acta 12, 133^149. Dansgaard, W., 1953. Comparative measurements of standards for carbon isotopes. Geochim. Cosmochim. Acta 3, 253^256. deMenocal, P., Ortiz, J., Guilderson, T., Adkins, J., Sarnthein, M., Baker, L., Yarusinsky, M., 2000. Abrupt onset and termination of the African Humid Period: rapid climate responses to gradual insolation forcing. Quat. Sci. Rev. 19, 347^361. Dettman, D.L., 1994. Stable Isotope Studies of Fresh-water Bivalves (Unionidea) and Ostracods (Podocopida): Implications for Late Cretaceous/Paleogene and Early Holocene Paleoclimatology and Paleo-hydrology of North America. Ph.D. Dissertation, University of Michigan, Ann Arbor, MI. Epstein, S., Mayeda, T., 1953. Variation of 18 O content of waters from natural sources. Geochim. Cosmochim. Acta 4, 213^224. Ethiopian Mapping Authority, 1988. National Atlas of Ethiopia. Addis Ababa. Gasse, F., 1977. Evolution of Lake Abhe¤ (Ethiopia and T.F.A.I.) from 70,000 B.P.. Nature 2, 42^45. Gasse, F., Fontes, J.C., Rognon, P., 1974. Variations hydrologique et extensions des lacs Holocenes de Desert Danakil. Palaeogeogr. Palaeoclimatol. Palaeoecol. 15, 109^148. Gasse, F., Street, F.A., 1978. Late Quaternary lake-level £uctuation and environments of the northern rift valley and Afar regions (Ethiopia and Djibouti). Palaeogeogr. Palaeoclimatol. Palaeoecol. 24, 279^325. Gri⁄ths, J.F., 1972. Climates of Africa. In: Gri⁄ths, J.F. (Ed.), World Survey of Climatology, V. Elsevier, Amsterdam, p. 10. Grossman, E.L., Ku, T., 1986. Oxygen and carbon isotope fractionation in biogenic aragonite: Temperature e¡ects. Chem. Geol. 59, 59^74. Hailemichael, M., 2000. The Pliocene Environment of Hadar, Ethiopia: A Comparative Isotopic Study of Paleosol Carbonates and Lacustrine MOllusk Shells of the Hadar Formation and of Modern Analogs. Ph.D. Thesis, Case Western Reserve University, Cleveland, OH, 239 pp. Hillaire-Marcel, C., Taieb, M., Tiercelin, J., Page, N., 1982. A 1.2 Mys record of isotopic changes in a late Pliocene Rift Lake, Ethiopia. Nature 296, 640^642. Hoelzmann, P., Kruse, H., Rottinger, F., 2000. Precipitation estimates for the eastern Saharan palaeomonsoon based on a water balance model of the West Nubian Palaeolake Basin. Global Planet. Change 26, 105^120. IAEA, 1996. Database of Global Network of Isotopes in Precipitation (GNIP), available at http://www.iaea.or.at/ programs/ri/gnip/gnipmain.html.

PALAEO 2906 29-8-02

M. Hailemichael et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 186 (2002) 81^99 Joseph, A., Frangi, P., Aranyossy, F., 1992. Isotope characteristics of meteoric water and groundwater in Sahelo-Sudan zone. J. Geophys. Res. 97, 7543^7551. Lamb, A.L., Leng, M.J., Lamb, H.F., Mohammed, M.U., 2000. A 9000 year oxygen and carbon isotope record of hydrological change in a small Ethiopian crater lake. Holocene 10, 167^177. Leng, M.J., Lamb, A.L., Lamb, H.F., Telford, R.J., 1999. Palaeoclimatic implications of isotopic data from modern and early Holocene shells of the freshwater snail Melanoides tuberculata, from lakes in the Ethiopian Rift Valley. J. Paleolimnol. 21, 97^106. McCrea, J.M., 1950. On the isotopic chemistry of carbonates and palaeo-temperature scale. J. Chem. Phys. 18, 849^857. National Meteorological Services Agency of Ethiopia (NMSA), 1996. Unpublished data. 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, Amsterdam, pp. 25^56. Overpeck, J., Anderson, D., Trumbore, S., Prell, W., 1996. The southwest Indian Monsoon over the last 18000 years. Clim. Dyn. 12, 213^225. Petit-Maire, N., 1990. Will Greenhouse green the Sahara? Episodes 13, 103^107. Ritchie, J.C., Haynes, C.V., 1987. Holocene vegetation zonation in the eastern Sahara. Nature 330, 645^647. Rodrigues, D., Abell, P.I., Kro«pelin, S., 2000. Seasonality in the early Holocene climate of northwest sudan: interpreta-

99

tion of Etheria elliptica shell isotopic data. Global Planet. Change 26, 181^187. Rozanski, K., Araguas-Araguas, L., Gon¢antini, R., 1993. Isotopic patterns in modern global precipitation. In: Swart, P.K., Lohman, K.C., McKenzie, J., Savin, S. (Eds.), Climate Change in Continental Isotopic Records, Am. Geophys. Union, Washington, DC, pp. 1^36. Rozanski, K., Araguas-Araguas, L., Gon¢antini, R., 1996. Isotope patterns in the East African Rift. In: Johnson, T.C., Odada, E.O. (Eds.), The Limnology, Climatology and Paleoclimatology of the East African Lakes. Gordon and Breach, Amsterdam, pp. 79^93. Schoell, M., Faber, E., 1976. Survey on the isotopic composition of waters from NE Africa. Geol. Jahrb. D17, 197^ 213. Taieb, M., 1974. Evolution Quaternaire du Bassin de l’Awash. Thesis, Universite¤ de Paris, 290 pp. Telford, R.J., Lamb, H.F., 1999. Groundwater-mediated response to Holocene climatic change recorded by the diatom stratigraphy of an Ethiopian crater lake. Quat. Res. 52, 63^ 75. Thorweihe, U., Brinkman, P.J., Heinl, M., Sonntag, C., 1990. Hydrological and hydrogeological investigations of the Darfur area, western Sudan. Berl. Geowiss. Abh., Reihe A 120, 279^326. Walter, R.C., 1993. Age of ‘Lucy’ and the ‘¢rst family’. Geology 21, 57^61. Walter, R.C., Aronson, J.L., 1993. Age and sources of the Sidi Hakoma Tu¡, Hadar Formation, Ethiopia. J. Hum. Evol. 25, 229^240.

PALAEO 2906 29-8-02