Integrating geographic information systems (GIS) and modeling: Validating prehistoric site-settlement models for the South Carolina coastal plain using a GIS

Modeling has been a major component of archaeological research for five decades. A type of modeling called predictive modeling has been developed that correlates environmental variables with archaeological site locations to develop maps that depict areas where sites may be located. While these techniques have been valuable for cultural resource management applications, their objective is to better understand prehistoric cultural systematics and explain site location variability over the landscape.; In 1978, a site-settlement model of prehistoric Woodland Period (1000 B.C. to A.D. 1000) utilization of the interior Lower Coastal Plain of South Carolina was developed. This model hypothesized that the interior uplands of the Lower Coastal Plain were occupied by Woodland populations during the late fall and early winter to extract seasonally available resources located in specific environmental zones. Field testing of the hypothesis through probablistic survey indicated that soil drainage characteristics, distance to boundary between well drained and poorly drained soil ecotones, slope, relative elevation and aspect were important to potential site location.; Although the model has been used since 1978 to support numerous cultural resource management and land use planning decisions, the model has never been validated. It also has been applied to a range of temporal and geographic contexts for which the model was not designed. The model was evaluated for a stratified sample of archaeological sites from across the Coastal Plain. Geographic information systems technology was employed to provide data analysis and cartography. Predictive maps were generated from the spatial association of soil drainage, slope, distance to ecotone and aspect. To determine the validity of the model, several hypotheses regarding prehistoric site-settlement in the region were tested using the Chi square (χ2) statistic. Additionally, individual model parameters were evaluated for their contribution to site prediction. Finally, residual analysis was conducted for sites that did not meet model expectations. The model was validated by this study, however, sufficient variability exists in site locations to suggest that additional factors influenced Woodland Period site location. Various methodological, geological and cultural explanations including spatial retrieval procedures and sea-level fluctuation were presented as potential explanations for this variability.