The Mele Havea site is centrally located in Ha`afeva village on an open sandy flat 145 m inland from the present beach. This site was documented in the 1996 survey of central Ha`apai. A systematic augering project and the excavation of a single 1 x 1 m test excavation revealed midden deposits of between 1 and 1.5 m depth. Both plain ware and dentate stamped Lapita pottery sherds were recovered from the excavations, with the latter concentrated in lower levels of the site. Faunal remains from the lower levels also included extinct species of land bird as well as a concentration of sea turtle. Results of the augering project indicated an early Lapita component of approximately 30 x 30 m size. The report on 1996 field activities (Burley 1996b) provides an account of the initial work.
The decision to return to the Mele Havea site in 1997 was predicated upon a number of factors. Preliminary work in 1996 suggested some degree of stratigraphic integrity was present and this would facilitate separation of early Lapita from plain ware components. The site is located in central Ha`apai, and the recovery of a comparative sample would provide a broader geographic representation in subsequent studies of early Lapita ceramic, faunal and other assemblages. Most importantly, the island of Ha`afeva is extremely small (approximately 1.8 km2) and a Lapita settlement on the island hypothetically would have had an immediate and discernible impact on island faunas. Work at the Mele Havea site was conducted between 7 and 19 July.
The site specific field objectives for excavations at the Mele Havea site were defined as:
Based on the above, a 3 x 4 m grid incorporating the 1996 test pit was laid out; all but a single 1 x 1 m unit subsequently was excavated (Figure 10). Field methods included trowel excavation in 10 cm arbitrary levels within larger stratigraphic units. All matrices were screened with nested sieves of 6.4 and 3.2 mm size, and proportionate matrix samples as well as a 100 % shell fish sample were recovered from one unit (No. 4). As field work progressed, it was apparent that the upper 25 cm of site deposit represented a highly disturbed zone in which historic artifacts were mixed with a small degraded assemblage of pottery sherds, the latter presumably transported from below. For the final three excavation units (Nos. 7, 10 and 11), this zone was removed by shovel without sieving. Within the time framework set for the project, this expediency allowed for the completion of one additional 1 x 1 unit.
Stratigraphy of the Mele Havea site is well defined with clear and easily discerned breaks across most of the excavated area. Intrusive cultural features, however, did make stratigraphically-based excavations complex in some areas. From upper to lower, and as illustrated in Figure 11, four strata are present.
Excavation of the block area did expose several well defined cultural features including small and large post holes as well as intentionally excavated pits. Most features appear to originate in Stratum II. One major exception is a large earth oven feature that was dissected by the eastern side of the block excavation. This pit, over 1 m in diameter and 0.8 m deep, appears to originate from the Stratum III occupation zone and extends into Stratum IV sand.
The ceramic assemblage from the 11 excavation units at the Mele Havea site includes 13,378 sherds (Table 10). Of these, 471 are decorated specimens typical of the early eastern Lapita period. The ratio of decorated rim sherds to undecorated rim sherds is 16.3%. The full variety of Lapita and plain ware vessel forms appear represented in the collection.
Compared to the ceramic assemblage, the nonceramic artifact collection is limited, including but 135 specimens (Table 11). Among these are shell valuables, utilitarian items (ie. shell scrapers, adzes, abraders) and several cut, abraded, or pecked specimens without a specifically identified function. Faunal remains are well preserved and abundant.
A total of 46 charcoal samples were collected from the Mele Havea site excavations with four of these submitted for radiocarbon dating. Two of the samples were selected from Stratum II and III contexts in a single excavation unit (No. 10) while the remaining two were individually picked to date Stratum II (Unit 3) and III (Unit 8) elsewhere in the excavation block. No reversals were encountered and the resultant dates indicate a narrow time span of 2640 + 50 BP to 2490 + 50 BP for site occupation (also see Appendix A). With the possible exception of Faleloa, dates on Stratum III at the Mele Havea site indicate an initial occupation slightly more recent than other Lapita sites in northern Ha`apai.
Excavations at the Mele Havea site provide an important comparative sample for an interpretation of the early Lapita period in Ha`apai. Located in central Ha`apai, these collections represent Lapita and plain ware settlements geographically separated from the four other Lapita hamlets at Pukotala, Faleloa, Tongoleleka and Vaipuna. Second, the island of Ha`afeva is extremely small in spatial extent, and one might hypothesize that an initial occupation would have a immediate impact upon the local resource base. Preliminary examination of the faunal collections by Steadman indicate this was indeed the case, with the most severe effect on terrestrial bird faunas. Finally, because of island size and resource potential, it can be hypothesized that Ha`afeva would be among the last of the coral limestone islands in Ha`apai to be settled. Stratum III radiocarbon dates seem to support this hypothesis.
Stratigraphic integrity of the Mele Havea site is among the best in Ha`apai. Analysis of excavated collections should be able to reasonably separate the early Lapita component from plain ware materials. Hopefully this will provide information on the nature of associated subsistence systems as well as insights into the rapid abandonment of decorated ceramic wares. If there is any disappointment with this project, it is a failure to secure more than a small number of nonceramic artifacts. Whether this is a result of sampling bias or is a poverty representative of the site occupation itself remains to be determined.