Project Overview

Data  Acquisition

Model Development

Project Home

Model Evaluation

Literature and Link

Project Staff

The assessment results of the 58 known archaeological sites (Table 3) formed the basis of the model testing.  In addition to paper records, the South Carolina State Archaeological Site Files maintains a nascent GIS coverage of known sites within the state.  Currently, this consists of simple site locations, digitized as polygons off of 7.5’ USGS Quadrangles. 

Table 3:  Results of SCIAA site level evaluation and subsequent test against ESRI-USC level I and level II model.

 SCIAA initially intersected these data with coverage of Willamette holdings to determine how many sites occupy Willamette lands.  For model assessment, the resultant site locations were layered over the Level I and Level II High Probability polygon vector layers generated by ESRI-USC.  Each known site location was checked to determine if it fell in the appropriate model area, for example, if a known level II site fell within or intersected the modeled polygon area, the known site was located as predicted by the model (Figure 1).

Figure 1: Level 2 site versus predictive model area.

Discussion

Of the 58 sites on Willamette property, five were evaluated as Level I, five as Level II, and 32 as Level III.  The remaining sites could not be evaluated because there was insufficient data in the State Archaeological Site Files.  Of the five Level I sites, one is a prehistoric site containing earthen mounds while four are historic sites.  Only the mound site and one historic site fall into the High Probability Area for Level I sites.  These results revealed the principal shortcoming of the model: without GIS coverage of historic road networks, it is virtually impossible to model locations of historic sites.  This project explicitly relied on publicly and readily available data for model generation.  As a result, historic road networks could not be modeled.  In contrast to the situation for historic sites, prehistoric sites are closely tied to the river and stream network, which is readily available in GIS format. 

The close relationship between area hydrology and archaeological sites is borne out by the success of the Level II High Probability Modeling effort.  Of the five Level II sites on Willamette property, all are prehistoric and all are located within the Level II area. 

The utility of the model is not only measured by how well it includes appropriate sites, but by how well it excludes other sites.  Only 16% (n=5) of the Level III sites occur within the area defined as high probability for Level I or Level II sites.  The remaining 27 Level III sites fell outside the areas defined by the model parameters. 

Finally, the modeling effort allows some prediction about known sites for which insufficient data was available for assessment.  Of these 16 sites, 8 (50%) fall within the area defined by the Level II predictive model.  Based on the known results of the modeling effort, most of these are likely Level II sites.  In contrast, the 8 known sites which fall outside of the modeled area are, according to the model, unlikely to produce significant archaeological data. 

Conclusions

The model itself appeared to be robust and useful, however some areas will have been omitted by the model.  The model is not able to account for low rises adjacent to the foothills (Figure 2).  This is because the procedure used only selects the area with closed contours as Level I sites. The contours of these low rises may encompass the entire hill and thus would not fall entirely within the buffered flood plain, yet such an area may well be a significant site. Digital hypsographic data with smaller contour intervals (less than 10 ft/contour) would improve resolution and partially solve this problem.  To date, no other automatic solution apart from digitization of each instance is known.

Figure 2: An instance where a high probability area is omitted by the level 1 model.

Archaeological probability models are an important consideration in land-use planning because they provide a cost-effective way to target areas in need of protection or additional work prior to the development of that land.  From a strictly legal perspective, such as that driven by Section 106 of the National Historic Preservation Act, although intensive survey to locate sites is still necessary to ensure they are not unknowingly destroyed, fewer significant sites are likely to be present in low probability areas.  The end result from a management standpoint is twofold.  First, although extensive and detailed survey is still necessary, the associated costs are mitigated by the decreased site density and resultant decrease in survey time devoted to defining site boundaries and contents once they are located.  Second, expenditures for testing or data recovery on significant sites that may be impacted by new or continued land-use practices can be reduced because proportionally fewer sites requiring these measures occur.  In the case of this project, Willamette Industries can continue to operate as usual in areas outside the high probability areas for Level I and Level II sites, allowing a high level of confidence that significant archeological sites will remain undisturbed.  Additionally, they can also avoid or modify their timbering methods within areas that are likely to contain significant sites, and thus effectively fulfill their stewardship role, with the understanding that once these areas have been examined archaeologically, timbering operations may resume in the absence of significant sites.