Date of Award
Open Access Dissertation
College of Arts and Sciences
Michael E. Hodgson
Advances in Geographic Information Science (GISci), archaeological databases and statistics enable the development and refinement of spatial applications of multivariate statistics and other quantitative methods for modeling ancient and historical cultural landscapes. Along the Central Savannah River of South Carolina, this research on prehistoric and historic site distributions, their environmental, temporal and cultural context, and geographic modeling explored methodologies for predicting site locations and modeling cultural landscapes to gain a better understanding of the distant and recent past. Methods for testing extant models, detecting changes in land-use through time, and for developing time-sliced and adaptation-based landscape models were demonstrated using archaeological data from the Department of Energy’s Savannah River Site (DOE-SRS), in Aiken, Barnwell and Allendale counties, South Carolina.
Multivariate Analysis of Variance (MANOVA) tests of cultural- and time-sliced datasets using 32 variables surprisingly revealed that only two models were needed to characterize the cultural landscape: a prehistoric and a historic model, respectively. Of the 32 variables available for Binary Logit Model (BLM) development, a knowledge-based approach was used to select seven variables for the prehistoric era plus one additional variable for the historic era. The seven common variables included: elevation, percent slope, profile- and plan-curvature, caloric cost distance to water, relative elevation to streams, and elevation range (plus, caloric cost-distance to 1951 historic roads, for the historic model).
Both the prehistoric and historic BLM models were reclassified into high, moderate and low probability areas and tested with an independent validation sample and nonparametric (X2) statistics for significance. The prehistoric model was highly significant, beyond the 0.001 probability level, and illustrates the importance of ecologically-rich edge environs to prehistoric cultures. Surprisingly, waterways and wetlands, long considered the most significant factors in prehistoric land use, were coincidental to these edge environs.
Conversely, the historic BLM model demonstrates the importance of the rolling hills between the flat upland terraces and bottomland forests to farming and livestock during the historic era. The surprise for the historic era was that historic roads were not the most significant contributor to the model and this was interpreted as a skewed result. With socio-economics likely yielding to the advantages of dispersed farmsteads, roads were crucial to historic settlement and should have made a significant contribution to the model. While significant (p < 0.05), the historic model was comparatively weak statistically and visually it demonstrated low probability values in a few historically-populated areas, particularly in the vicinity of the small town of Dunbarton.
Gillam, J. C.(2016). Multivariate Analysis and Geographic Modeling of Archaeological Landscapes. (Doctoral dissertation). Retrieved from http://scholarcommons.sc.edu/etd/3989