| This thesis originated in the Uncompahgre Homesteads Internship (UHI), a summer internship offered by the heritage department of the Grand Mesa, Uncompahgre, and Gunnison Forest Service (GMUG USFS) and designed to re-evaluate known and inventory unknown historic homestead archaeological sites dating approximately from 1890 to 1945 AD on the southern Uncompahgre Plateau in southwestern Colorado, USA. On GMUG lands, historic homestead sites continue to deteriorate as a result of natural decay and human and animal activities. In response, I used ArcGIS software to develop predictive models aimed at assisting in the inventory of homestead sites in the region in pursuit of the National Historic Preservation Act. The predictive models I developed were based on the principles of energy conservation and behavioral optimization advocated by the theory of human behavioral ecology, and each uses a single variable to model historic settlement patterns including: distance to water sources, slope, and aspect. Subsequently, I test the predictive utility of each model using Kvamme's (1988) gain-statistic. My analyses revealed that predictive models based on site distance to water provide a high degree of predictive utility due to the small areas encompassed by these models, while the aspect and slope models offer very little predictive utility, no doubt resulting from the extensive areas covered by these models. In addition, the distance-to-water models present pragmatic discrete and linear areas that would facilitate survey designs. Accordingly, I advocate the use of the distance-to-water predictive models by federal agencies in the pursuit of a more complete inventory of historic homestead sites. |
