Archaeological and paleontological research in central Flores, east Indonesia: results of fieldwork 1997-98.

Morwood, M.J. ; Aziz, F. ; O'Sullivan, P. 等

Background

Claims for Homo erectus on Flores were first made by Fr. Verhoeven, a Dutch priest and amateur archaeologist, who excavated fossil sites in the region (Verhoeven 1968; Maringer & Verhoeven 1970a; 1970b). He reported stone artefacts at Mata Menge, Boa Lesa and Lembahmenge, and concluded that they were about 750,000 years old on the basis of their association in the fossil deposits with Stegodon, an ancestral elephant species.

The range of evidence presented for his claims was generally judged inconclusive because of doubts about the identification of the stones nominated as artefacts, the uncertain stratigraphic association between these and Stegodon fossils, and the age of the strata (e.g. Allen 1991; Bellwood 1985: 66).

Palaeontological research undertaken by an Indonesian-Dutch team replicated Verhoeven's findings. They reported stone artefacts with the remains of Stegodon at Mata Menge (van den Bergh et al 1996: 32-4) and, on the basis of preliminary work, reported a similar association at Dozu Dhalu (van den Bergh & Aziz 1994: 22; van den Bergh 1997: 249).

Furthermore, their excavation at the stratigraphically older site of Tangi Talo, which lacked associated stone artefacts, revealed that there had been a major turn-over in fauna: Tangi Talo contained the remains of pygmy Stegodon (S. sondaari), giant tortoise (Geochelone sp.) and Komodo dragon (Varanus komodoensis), whereas Mata Menge had large Stegodon (S. trigoncephalus florensis), crocodile and giant rat (Hooijeromis nusatenggara). Sondaar (1987) argued that this faunal turn-over resulted from the arrival of a new predator, H. erectus. Palaeomagnetic determinations from Tangi Talo and Mata Menge suggested that the former was 900,000 years old and the latter 'slightly less than' 730,000 (Sondaar et al. 1994: 1260).

Reaction to these claims was muted, and where published, generally cautious (e.g. Bellwood 1997: 67-8). The major impediments to general acceptance of the conclusions were three-fold: the identification of stone artefacts, the lack of taphonomic detail, and the chronological ambivalence of a palaeo-magnetic transition 3 m below the Mata Menge fossil/artefact deposit.

More recent work has shown that stone artefacts definitely occur in situ at Mata Mange (Morwood et al. 1997), while fission track dates of 880,000-800,000 and 900,000 years for Mata Menge and Tangi Talo, respectively, show that both sites are of Early Pleistocene age (Morwood et al. 1998). These findings have significance for assessing the capabilities of early hominids, for their dispersal and for insular evolutionary processes.

Here we present the results of excavations at two of the sites where stone artefacts were said to be in primary association with Middle Pleistocene faunal remains - Boa Lesa and Dozu Dhalu. These are discussed in the light of regional site-survey data. In both cases our aims included:

a To record the stratigraphy and other information on the context of deposition.

b To ascertain whether stone artefacts occur in situ in the deposits associated with fossils, as claimed by previous researchers.

c To collect geological samples for dating, grain-size analysis and palynology.

d To recover a representative sample of stone artefacts and/or fossils.

Geological context

Evidence for early hominid presence on Flores comes from the Soa Basin on the upper Ae Sissa River [ILLUSTRATION FOR FIGURE 1 OMITTED]. The basin is about 35x22 km in size and is almost entirely surrounded by mountains and active volcanoes. There is one deeply incised river outlet.

The work of Hartono (1961) provided a useful framework for establishing the general and specific geological context of the sites. He identified and mapped three basic stratigraphic units in the study area and our work has added much detail.

The basal material of the Soa Basin is an andesitic breccia termed the Ola Kile Formation. Overlying the Ola Kile Formation is the Ola Bula Formation, a series of tuffaceous sandstones and siltstones up to 120 m thick, which accumulated in the Early and Middle Pleistocene. Within the formation there are very extensive layers of white tuffaceous siltstone, which provide useful chronological markers.

The lateral extent of some siltstone layers indicates that they were deposited in an extensive body of water. For much of its history the Soa Basin would have comprised a large lake or series of lakes, presumably when tectonic activity blocked the outlet. Periodically, however, a new river outlet appears to have been cut and the area drained to become a grassland savanna. Fossil and archaeological sites within the basin only formed during such 'dryland periods'. At such times processes of erosion would have predominated, but localized tuffaceous sediments accumulated in rivers and creeks. Fossils of land animals and stone artefacts occur only in these localized deposits.

The Ola Bula Formation is capped by the 5m thick Gero Limestone, which formed in freshwater conditions (David Haig pers. comm). Its former extent and horizontal distribution across the Soa Basin suggest that it formed in a large lake. Some tuffaceous siltstone deposits occur within and overlying the Gero Limestone, indicating that small-scale volcanic eruptions continued, but these no longer provided a major source of sediments, as occurred in deposition of the Ola Bula Formation. Fossilized algae, fish and freshwater gastropods occur in the Gero Limestone but not those of land animals, which would have been confined to the lake margins. Land animals would have reoccupied the basin after the Ae Sissa River outlet formed and the lake drained.

An important point is that, except for the northern part of the Soa Basin, where a thrust fault has resulted in a dip to the south, the Ola Bula Formation and the Gero Limestone have remained horizontal (Hartono 1961: 15). The relative heights of fossil sites within the Ola Bula Formation are, therefore, generally indicative of their relative ages - although thick deposits formed in large, deeply-incised channels (e.g. Ngamapa) are lower in the sequence than those formed in contemporaneous small channels (e.g. Kobatuwa). On the basis of relative heights, the site of Boa Lesa is younger than Dozu Dhalu.

Excavations at Boa Lesa

Boa Lesa occurs midway along a ridge which runs at 110 [degrees] N between two tributary gorges. Fossils are exposed on the lowermost terrace and scarp of a north-facing slope [ILLUSTRATION FOR FIGURE 2 OMITTED]. The area is grassed with occasional trees in more sheltered areas.

Verhoeven did a major excavation at Boa Lesa in 1963 and his 10x5 m cutting is still evident [ILLUSTRATION FOR FIGURE 3 OMITTED]. He recovered a large assemblage of Stegodon fossils and found many associated in situ stone artefacts (Maringer & Verhoeven 1970a: 230; 1970b). Fossils and stone artefacts on the surface for at least 13 m to the east of Verhoeven's excavation give some idea of the extent of in situ material.

Procedure

We worked at Boa Lesa from 22 July to 16 August 1998. A site datum point was established, and contour plans and cross-sections of the site and environs made. The walls of Verhoeven's excavation were then cleaned and recorded. It was evident that the fossils had been deposited in a channel and that Verhoeven had ceased digging when he encountered the siltstone layer comprising the channel bed and its bank. This layer (D1) is extremely hard and does not contain fossils.

Three areas were selected for excavation as a means of examining spatial differences in deposition. A narrow trench was also dug to clarify the stratigraphic relationship between the channel deposits and those of adjacent areas. Deposits were excavated in 10-20-cm spits with chisels, hammers and geological picks. Where possible all bones and stones were left in place until mapped and photographed. Excavated deposits were broken up, closely examined and passed through 5- and 3-mm sieves. All finds at the sieves were immediately bagged.

The deposits were a maximum of 180 cm deep [ILLUSTRATION FOR FIGURE 4 OMITTED]. They comprised more recent topsoil and slope-wash sediments unconformably overlying a white tuffaceous siltstone (B1), with underlying sandstone layers (C1-C4). The latter had flow structures. The bed of the channel had been incised into a very extensive, light-brown siltstone layer (D1), which also formed the surface of a flanking terrace. In addition, our excavations in Area 3 showed that the original watercourse had cut through a 70-cm thick, coarse sandstone layer (D2) overlying the light-brown siltstone, but a more extensive excavation is required to ascertain the full depth of the original Boa Lesa channel.

Vertical concentrations of bones, stones and occasional artefacts occurred within the sandstone, particularly at interfaces between strata [ILLUSTRATION FOR FIGURE 5 OMITTED]. These seemed to correspond to 'flushes' of material entering the channel, presumably after rain. The bones were disarticulated but many were complete.

The Boa Lesa finds

Only Stegodon bone was found, despite careful examination of excavated rock. Much was fragmented, but notable specimens included a complete femur and tibia, ribs, a pelvis and half a mandible. These were found associated with other evidence for strong water flow, such as water-rolled pebbles. There was a concentration of such material on the channel bed.

Six stone artefacts were found [ILLUSTRATION FOR FIGURES 6, 7 OMITTED], as detailed below.

Area 1

BL1/0/98 Retouched chert flake. Base of Layer C4 on channel bed next to a Stegodon pelvis. (10 g)

BL1/1/98 Basalt flake from a water-rolled pebble. Base of Layer A2. (40 g)

BL1/2/98 Retouched andesite flake. Base of Layer A2. (37 g)

BL1/4/98 Unifacially retouched andesite cobble chopper. Found at the interface between Layers C1 and C3 with a concentration of other water-rolled andesitic and siltstone pebbles, and Stegodon bone, including a complete femur and ribs. (855 g)

Area 2

BL2/1/98 Retouched andesite flake from a pebble. Layer C1. (38 g)

BL2/9/98 Retouched chert flake. Layer C3. (10 g)

Interpretation of Boa Lesa

Stone artefacts and fossil Stegodon bones occur together in tuffaceous sandstones at Boa Lesa. They were deposited in the bed of a water channel cut into older, lacustrine sediments which had already silicified. The extensive strata overlying the fossiliferous channel deposits are also lacustrine - clearly the dryland period during which the fossils were laid down was preceded and followed by lake-full periods.

Low energy conditions prevailed in the water channel, but occasionally strong discharges washed in heavy items such as bones and stone artefacts from adjacent dryland areas. The artefacts showed minimal evidence for water-rolling, indicating that they were not transported far.

The uppermost channel deposit (B1) comprises a homogeneous, white tuffaceous siltstone, which presumably corresponds to a single volcanic eruption. This siltstone contains occasional bones, but no artefacts. It is unconformably overlain by gravel-rich slope-wash deposits.

Overall, Boa Lesa is very similar to the nearby site of Mata Menge in context, nature of the deposits and stratigraphic height. Mata Menge also has stone artefacts and Stegodon fossils deposited in a water channel, and it dates to 800,000+80,000 years BP (Morwood et al. 1998). As expected on the basis of stratigraphic position, Boa Lesa is older: the white tuffaceous siltstone (B1) at the top of the fossil/artefact deposits has yielded a fission track age of 840,000+70,000 years BP (Ref. 96FL03).

Excavations at Dozu Dhalu

Dozu Dhalu occurs on a grassed ridge running down at 335 [degrees] N from a plateau [ILLUSTRATION FOR FIGURES 8, 9 OMITTED]. The ridge, which is about 100 m wide, ascends in a series of steps resulting from differential erosion of underlying sandstone and siltstone layers. Fossils are exposed on the flat and scarp of the second terrace.

Indonesian and Dutch researchers carried out preliminary excavations at Dozu Dhalu in 1994 and recovered Stegodon, Komodo dragon and giant rat bone. They also claimed to have found in situ stone artefacts (van den Bergh 1997: 249).

Procedure

Our excavations were undertaken at Dozu Dhalu from 15 November to 2 December 1997. In addition to the procedures used at Boa Lesa, we transported 200 kg of deposits to the creek below Dozu Dhalu and wet-sieved them.

Three areas (A, B, C) were selected for excavation on the basis of fossil exposures and previous work at the site (van den Bergh & Aziz 1994: 22). Deposits were excavated to a maximum of 1-9 m deep in Area A, where two fossiliferous siltstone layers were separated by hard, sterile sandstones [ILLUSTRATION FOR FIGURE 10 OMITTED].

The Dhozu Dhalu finds

We recovered the remains of large Stegodon, Komodo dragon and giant rat. There was considerable variation in the nature of the deposits and fossils across Dozu Dhalu. For instance, bones in Area A accumulated in low energy conditions, were relatively complete, and often occurred in semi-articulated positions [ILLUSTRATION FOR FIGURES 11, 12 OMITTED]. Each of the major bone concentrations here represented one or two individual Stegodons, with interspersed teeth of Komodo dragon and rat (see also van den Bergh & Aziz 1994: [ILLUSTRATION FOR FIGURE 8 OMITTED]).

In contrast, deposits in Areas B and C formed in high energy conditions. Stegodon tusks and long bones predominated and many were found aligned in previous 'fissures' in the original river bed. Few Komodo dragon or rodent remains were recovered in these areas - most likely because excavated deposits here were not sieved.

Although many stone artefacts occur on the surface at and around Dozu Dhalu [ILLUSTRATION FOR FIGURE 13 OMITTED], no in situ examples were found. Angular fragments of volcanic rock and water-rolled pebbles were excavated from the deposits, especially in Area B, but they lacked striking platforms, bulbs of percussion, well-defined flake scars or retouch [ILLUSTRATION FOR FIGURE 14 OMITTED].

In summary, our work did not substantiate previous claims for in situ stone artefacts at Dozu Dhalu (cf. van den Bergh & Aziz 1994: 22). Where sites contain pieces of rock, which could be misidentified as artefacts, and have scatters of stone artefacts on the surface, the question of exact provenance is crucial.

Two fission track dates are available for the locale: these comprise 850,000[+ or -]90,000 BP for the Dozu Dhalu fossil strata and 920,000[+ or -]80,000 BP for a prominent white 'marker' tuff 20 m below. The apparent absence of stone artefacts is particularly significant in the light of the age of the site and the results of the regional site survey.

Regional site survey

A number of fossil sites previously investigated by Maringer & Verhoeven (1970a; 1970b), Sondaar et al. (1994) and Morwood et al. (1997) were visited and data collected on their contexts, dimensions, contents, locations and relative heights (TABLE 1).

Differences in the height of the Ola Kile/Ola Bula Formation boundary across the Soa Basin show that the fossils were deposited in at least two sub-basins: a relatively shallow sub-basin in the west containing the 'minor channel' sites of Kobatuwa, Mata Menge, Lembahmenge and Boa Lesa, and a deeper sub-basin in the east with the 'major channel' sites, such as Ngamapa, Kopowatu and Pauphadhi.

The complexity of deposition and the large error range for fission track dates mean that much more stratigraphic information is required to work out the chronological sequence for the Soa Basin and to reconstruct site palaeo-environmental contexts in detail. However, on current evidence, there were at least four dryland periods in the Soa Basin, separated by lake-full periods:

Period 1 Associated with deposition of fossils at Tangi Talo 900,000 years ago. Fauna present at the time included pygmy Stegodon, giant tortoise and Komodo dragon. No in situ stone artefacts have been observed in the deposits.

Period 2 Associated with deposition of fossils at Sagala, Dozu Dhalu and Ola Bula more than 850,000 years ago. Fauna present at the time included large Stegodon, Komodo dragon, crocodile and giant rat. No in situ stone artefacts were found in the extensive excavations at Ola Bula (Maringer & Verhoeven 1970a) or during our work at Dozu Dhalu (cf. van den Bergh 1997: 249).

Period 3 Associated with deposition of fossils at Boa Lesa, Mata Menge, Kobatuwa, Ngamapa, Kopowatu and Pauphadhi between 840,000 and 700,000 years ago. Fauna present at the time included large Stegodon, crocodile and giant rat. In situ stone artefacts occur at all these sites.

Period 4 The present dryland period. Age unknown but of sufficient duration for the Ae Sissa River to downcut through the Gero Limestone, Ola Bula Formation and more than 100 m into the Ola Kile Formation in some areas.

In summary, our work has yielded unambiguous and relatively precise dates for the arrival of H. erectus on Flores by 840,000 years BP. Stone artefacts do not seem to occur in deposits slightly older. What are some of the implications?

Implications

Flores now provides the most reliable date for the arrival of H. erectus in Southeast Asia. In contrast, the dating of the well-known fossil hominids in Java is more difficult because the [TABULAR DATA FOR TABLE 1 OMITTED] region is more geologically complex, and many of the deposits and fossils have clearly been redeposited (Ithara et al. 1994).

Arrival of hominids on Flores seems to postdate a major change in the island's fauna, involving the extinction of two endemic species, pygmy Stegodon and giant tortoise. These extinctions about 900,000 years ago most likely relate to natural events, such as volcanic eruption or drought, rather than the impacts of early hominid hunting (cf. Sondaar 1987; Sondaar et al. 1994).

Even at times of low sea level, when Sumatra, Java and Bali were connected to mainland Southeast Asia, at least two sea crossings were required to reach Flores. The first of these deep-water sea barriers, between the islands of Bali and Lombok, is about 25 km wide and constitutes a major biogeographical boundary, the Wallace Line. Prior to human intervention, only animals capable of crossing substantial water barriers by swimming, flying or rafting on flotsam were able to establish populations on Flores (e.g. elephants, rats). In fact, the impoverished nature of the fauna on the island in the Early and Middle Pleistocene rules out the possibility of temporary landbridges from continental Southeast Asia (cf. Groves 1996). The presence of hominids on Flores in the Early Pleistocene therefore provides the oldest inferred date for human maritime technology anywhere in the world. Elsewhere, dates for such capabilities are much more recent (e.g. Cherry 1990). These findings indicate that the intelligence and technological capabilities of il. erectus may have been seriously underestimated. An accumulating body of evidence from elsewhere supports this conclusion (e.g. Thieme 1997).

The complex logistic organization needed for people to build water-craft capable of transporting a biologically and socially viable group across significant water barriers, also implies that people had language (Davidson & Noble 1992: 135). Previously the organizational and linguistic capacity required for sea voyaging was thought to be the prerogative of modern humans and to have only appeared in the late Pleistocene (Bowdler 1993). It now seems that humans had this capacity 840,000 years ago.

A final point is that our increased understanding of local and regional taphonomic processes in the Soa Basin will help identify areas where further archaeological deposits of Early, Middle and Late Pleistocene age may occur. The archaeological record between 300,000 and 50,000 years ago would be of particular interest, given that this is when Homo sapiens first appeared in the region and Greater Australia was colonized.

Acknowledgements. Grants from Australian Research Council and the UNE Vice Chancellor's Fund made the Flores fieldwork possible. We thank Professor Bruce Thom for his generous support. A/Professor David Haig, Dr Victoria Paine, Ir Suminto, Dr Netty Polhaupessy, Tular Sudarmadi, Robert Bednarik and Iwan Kurniarsu undertook specialized aspects of the research. Robert Bednarik also provided useful feedback on this paper. Musa Bali, Ferri Bali, Flori Bali, Cornelius Podhi, Alexander Dadhu, Minggus Siga, Herman Pawo, Ginus Dhenga, Oscar Dapangole, David Adoe, Aba Baly, Willem Lewa Nau, Robert Bago, Deus, Christo Fores, Catherine Marwood, Derek Payne and Esther Komar assisted with fieldwork, while Kathy Marwood drew the stone artefacts.

Fieldwork was conducted tinder LIPI Research Permit Numbers 5835/II/KS/1997 and 3735/II/KS/1998; Travelling Permit Numbers SKJ/POA-1066/X/1997 and SKJ/POA-5408/VII/1998; and Recommendation Numbers 070/1577 and 1450/07/IV/SOSPOL/NTT-98 from the Director of Sospol.

References

ALLEN, H. 1991. Stegodonts and the dating of stone tool assemblages in island S.E. Asia, Asian Perspectives 30: 243-66.

BELLWOOD, P. 1985. Prehistory of the Indo-Malaysian Archipelago. Sydney: Academic Press.

1997. Prehistory of the Indo-Malaysian Archipelago. Honolulu (HI): University of Hawaii Press.

BOWDLER, S. 1993. Sunda and Sahul: a 30 kyr BP culture area, in M.A. Smith et al. (ed.), Sahul in review: 60-70. Canberra: Australian National University, Department of Prehistory/RSPacS.

CHERRY, J.F. 1990. The first colonisation of the Mediterranean Islands, Journal of Mediterranean Archaeology 3(2): 145-221.

DAVIDSON, I. & W. NOBLE. 1992. Why the first colonisation of the Australian region is the earliest evidence of modern human behaviour, Archaeology in Oceania 27(3): 135-42.

GROVES, C.P. 1996. Hovering on the brink: nearly but not quite getting to Australia, in E. Rousham & L. Freedman (ed.), Perspectives in human biology 2: 83-7. Nedlands: University of Western Australia, Centre for Human Biology.

HARTONO, H.M.S. 1961. Geological investigation at Olabula, Flores. Unpublished report. Djawatau Geologi, Bandung.

ITHARA. M., N. WATANABE, D. KADAR & H. KUMAL 1994. Quaternary stratigraphy of the hominid fossil bearing formations in the Sangiran area, central Java, in J.L. Frunzen (ed.), 100 years of Pithecanthropus: the Homo erectus problem, 123-8. Frankfurt: Courier Forschungsinstitut Senckenberg.

MARINGER, J. & T. VERHOEVEN. 1970a. Die steinartefake aus der Stegodon-fossilschicht van Mengeruda auf Flores, Indonesien, Anthropos 65: 229-47.

1970b. Note on some stone artifacts in the National Archaeological institute of Indonesia at Djakarta, collected from the Stegodon-fossil bed at Boaleza in Flores, Anthropos 65: 638-9.

MORWOOD, M.J., F. AZIZ, G. VAN DEN BERGH, P. SONDAAR & J. DE VOS. 1997. Stone artefacts from the 1994 excavations at Mata Menge, west central Flores, Indonesia, Australian Archaeology 44: 26-34.

MORWOOD, M.J., P. O'SULLIVAN, F. AZIZ & A. RAZA. 1998. Fission track age of stone tools and fossils on the east Indonesian island of Flores, Nature 392 (12 March): 173-6.

SONDAAR, P.Y. 1987. Pleistocene Man and extinctions of island endemics, Memoires du Societe Geologique de France (n.s.) 150: 159-63.

SONDAAR, P.Y., G.D. VAN DEN BERGH, B. MUBROTO, F. AZIZ, J. DE VOS & U.L. BATU. 1994. Middle Pleistocene faunal turnover and colonisation of Flores (Indonesia) by Homo erectus, Comptes Residues de la Academie des Sciences Paris 319: 1255-62.

THIEME. H. 1997. Lower Paleolithic hunting spears from Germany, Nature 385: 807-10.

VAN DEN BERGH, G.D. 1997. The late Neogene elephantoid-bearing fauna of indonesia and their palaeozoogeographic implications: a study of the terrestrial fauna succession of Sulawesi, Flores and Java, including evidence for early hominid dispersal east of Wallace's Line. Ph.D thesis, University of Utrecht.

VAN DEN BERGH, G.D. & F. AZIZ. 1994. Sedimentology and fauna of continental deposits of Indonesia. Final report to the Geological Research and Development Centre, Bandung.

VAN DEN BERGH, G.D., B. MUBROTO, F. AZIZ, P.Y SONDAAR & J. DE VOS. 1996. Did Homo erectus reach the island of Flores?, in P. Bellwood (ed.), Indo-Pacific prehistory: the Chiang Mai papers (Proceedings of the 15th IPPA Congress. Chiang Mai, Thailand 1994, 1: 27-36. Canberra: Indo-Pacific Prehistory Association.

VERHOEVEN, T. 1968. Pleistozane Funde auf Flores, Timor and Sumba, in Anthropica Gedenkschrift zum 100. Geburtstag van P.W. Schmidt: 393-403. St Augustin: Verlag des Anthropos-Instituts. Studia Instituti Anthropos 21.