Archaeological predictive model of southwestern Kansas

Abstract

Knowledge on the archaeological condition of southwestern Kansas is anomalously low, therefore a high-resolution archaeological predictive model has been constructed for the High Plains region of southwestern Kansas. Using quantitative data about the environment as independent variables, the model was constructed using a combination of Geographic Information Systems (GIS) and statistical software. The location of sites was quantitatively related to the environment through a binary logistic regression analysis. The derived regression equation was used to create a unique probability score for each of the 20 million land parcels in the study area. Analysis indicates the model offers a significant increase (30%) over a random classification. 85% of known site locations and 60% of known non-site locations are accurately predicted. In total, the area predicted as site-present comprises 41% of the total study area; within which, the chances of finding a site are 2.15 times as likely as random.