A sample -oriented cultural resources survey of 80,000 acres in Grand County, east -central Utah, was undertaken and a predictive model was constructed using previously collected data. Fieldwork was conducted in 1983 by Goodson & Associates, Inc. of Lakewood, Colorado, under contract with the Bureau of Land Management. Initially a 10% unstratified random sampling strategy was devised, with the collected data to be used in testing the model. Upon initiation of the fieldwork, it was discovered that site density was much higher than anticipated, based on frequencies reported in the previously surveyed study tracts. Consequently, the sample fraction was reduced to 5%. Data collected from 100 40-acre transects were used to construct a predictive model. An additional 17 transects were surveyed as an independent sample and the resultant data, as well as other information, were used to test the model. A total of 126 sites was recorded with 88 sites used in model construction. Site densities ranged from three to twelve times higher, depending on the sensitivity zone, than was indicated by previous survey.
Several models were constructed using discriminant analysis. Classification accuracy consistently ranged above 80%. In addition to the traditional environmental variables analysis, soil groups were successfully used to discriminate between site and nonsite loci. Relationships among the traditional environmental variables and soil groups were demonstrated, and site sensitivity zones were developed using the soil models. Future research directions are recommended.