Ancient tortoise hunting in the southwest Pacific

We report the unprecedented Lapita exploitation and subsequent extinction of large megafauna tortoises (?Meiolania damelipi) on tropical islands during the late Holocene over a 281,000 km2 region of the southwest Pacific spanning from the Vanuatu archipelago to Viti Levu in Fiji. Zooarchaeological analyses have identified seven early archaeological sites with the remains of this distinctive hornless tortoise, unlike the Gondwanan horned meiolaniid radiation to the southwest. These large tortoise radiations in the Pacific may have contributed to the rapid dispersal of early mobile Neolithic hunters throughout southwest Melanesia and on to western Polynesia. Subsequent rapid extinctions of these terrestrial herbivorous megafauna are likely to have led to significant changes in ecosystems that help explain changes in current archaeological patterns from Post-Lapita contexts in the region.


Teouma, Efate Island, Central Vanuatu
The Teouma Lapita cemetery/settlement site is located at Teouma Bay situated on an ancient  (Table S10). Small proportions of burn marks were also observed on the skeletal extremities due to less meat being present on these body portions during either cooking or disposal (Table S10). These include on one mandibular fragment, small numbers of carpals/tarsals, phalanges and a dermal shell fragment. The low proportions of burning on the skeletal material may indicate that steam cooking in earth ovens was the preferred method.

Vao, Northern Vanuatu
The  (Table S10). This quantity of tortoise bone from this single site located on such a small island again infers that people were travelling to hunt tortoises. The big island of Malakula, visible across a channel from Vao, would have provided the ideal hunting ground.

Uripiv, Northern Vanuatu
The Velavi Lapita site is situated on Uripiv island, another small island <2 km² off the northeast coast  Table S5, #SCH 250 collected 18/08/2001) was recovered and identified from the sterile beach sand deposit of Layer 5 immediately underlying the Lapita midden in Test Pit 8. The femur surface was largely encrusted with calcareous beach sand and was heavily gnawed by rats indicating that it was on the beach surface after Lapita people arrived with rats about 3000 BP. That it was gnawed means it was likely fresh bone and so attractive to rats rather than a fossil. We therefore consider it was probably deposited coeval with the first people occupying the site and is likely to have been buried within the upper beach sands by trampling before a discrete midden then buried it further.

Arapus, Efate Island, Central Vanuatu
The  (Table S6) type tortoise bones were identified from basal cultural deposits overlying sterile beach sand. These

Naigani, Viti Levu, Fiji
The VL 21/5 Lapita site on the small island of Naigani (<2 km²) off the Northeast coast of Viti Levu in Fiji was first excavated in 1981 and again in 2000 (7), during which ca. 120 m² were excavated. The site is now approximately 100 m inland, as the shoreline has prograded since initial occupation of the site between 3270-2410 cal BP, when it was situated on a beach dune adjacent to the shore (7).
A single humerus (Fig. S1, Tables S8-S9,  The humerus has been extensively weathered and abraded since deposition, like the majority of bones recovered from the site, due to poor preservation conditions and gardening disturbance.

Yanuca, Fiji
The Yanuca Lapita site (VL 16/81) is located 11 km west of the Sigatoka River on the north coast of value of 9.5 per mil relative to AIR was much higher than the Teouma specimen (Table S14).

Systematic Paleontology
? osteological details in support of the present referral of specimens to ?Meiolania damelipi for the skeletal elements humeri, femora, and scapulae: acromion, glenoid, dorsal scapular process, from the seven Lapita sites from Vanuatu-Fiji discussed herein (Figs. S1-S3). Summary statistics of measurements from tortoise long bones from Vao and Teouma can be found in Tables S11-S12. damelipi. In the latter, it appears as a rectangular groove with crenulations inside and extends up to the proximal lateral process. In the former, it takes a different form as a pronounced narrow ridge protrusion with no groove extending from the distal portion of the dorsal proximal shaft fossa up to the lateral process. The posterior proximal shaft muscle attachment scar is absent from some ?M.
damelipi specimens, but when it is present it appears as a narrow long shallow groove extending diagonally from the medial process to a point approximately one third of the shaft. In some specimens, the groove is shorter and less perceptible. In M. platyceps, it is also not always present (absent in AM F49141), but when it does appear it is a shallow groove extending straight down the shaft from the medial process (AM F18748).  M. platyceps femora vary greatly in size, the longest being AMF: 1203 at 213 mm, but the incomplete AMF: 1788 is estimated to be as large as 274 mm (10). Femora of ?M. damelipi are smaller (Table S11) (Table S12), again reflecting the wide range of age classes hunted. Proximal maximum width is similarly variable for Vao (mean = 38.3 mm) and Teouma (mean = 36.5 mm), ranging between 12.1 mm to 60.9 mm while distal femora width for both sites ranges between 23.3 mm to 57.4 mm (Teouma mean = 37.8 mm; Vao mean = 45.2). Proportionate data (mean proximal humeri width/mean humeri length and mean distal humeri width/mean humeri length) also indicate similarities between Teouma (pw/l = 0.38, dw/l = 0.4) and Vao (pw/l = 0.36, dw/l = 0.43).
The shoulder girdle is triradiate i.e., it has three radiating branches from the glenoid, the dorsal scapular process and the acromion of the scapula, and the coracoid. The scapulae of ?M. damelipi are characterised by the glenoid having no supporting neck, and the dorsal scapular process and the acromion diverging at approximately 105°, compared to 120° for the deep shelled M. platyceps (3,10); suggesting that the Vanuatu tortoises had less deep shells by comparison. Another difference from M. platyceps is that the coracoid remains unfused from the scapular glenoid and is more elongate, which indicates ?M. damelipi had a lower body profile (3). Here (Fig. S3) we identify another distinctive feature for ?M. damelipi, a muscle attachment scar on the lateral acromion surface. This attachment scar appears as a projecting crenulated ridge which often extends from the