Any archaeological predictive model must be informed by the history of the environment and culture being modeled. Even though an inductive research design is used in MnModel, expert knowledge and theory are incorporated into the model through the selection of the sites included in the archaeological database and the independent environmental variables used for modeling.
Minnesota's Changing Environment
MnModel's methodology assumes that hunter/gatherers focused their activities in places that had the resources they needed - water, food, and shelter are chief among these. Previous archaeological investigations in Minnesota have shown that sites are often concentrated near lakes and rivers, particularly on landforms such as river terraces and beach ridges. Some predictive models restrict the environmental variables based on previous survey results, for instance using a variable like "within 500 feet of water" or "on a river terrace." In contrast, MnModel uses continuous variables, such as "distance to water" and "height above surroundings," to find combinations of terrain and other factors associated with site locations. This allows the models to predict uncommon site types or sites in locations that might not otherwise have been selected.
Minnesota's Native American Culture History
Minnesota has been occupied for approximately 12,000, since glaciers retreated from the landscape. Chapter 3 of the MnModel Final Report Phases 1-3 (2002) summarizes Minnesota's Native American cultural history, in the context of its changing environment. The Minnesota Statewide Multiple Properties Documentation Form for the Woodland Tradition includes context studies for the following complexes in Minnesota: Brainerd; Southeast Minnesota Early Woodland; Havana Related; Laurel; Fox Lake; Lake Benton; Central Minnesota Transitional Woodland; Southeast Minnesota Late Woodland; Blackduck-Kathio and Rainy River Late Woodland; and Psinomani.
Archaeological data are a necessary input in the development of an archaeological predictive model. In Minnesota, with the exception of the Red Wing area, archaeological sites are rather sparse in the landscape. This provided challenges in developing and testing the models. Where site distributions are particularly sparse, models developed from the database may not be truly representative of archaeological potential. Archaeologists' biases in deciding where to survey have added to this problem. MnModel is the only archaeological predictive model that has made this bias explicit in the final results so that the model's value for any given place can be assessed.
Moreover, the sparseness of the sample and paucity of information on site periods and contexts required that sites from all time periods (12,000 BP through 1836) be combined to create the archaeological database. MnModel's archaeological database included both site and survey locations. Chapter 5 of the MnModel Final Report Phases 1-3 (2002) describes the development of the archaeological database used in Phases 1 through 3 of the project. For Phase 4, 2,787 new sites were added to the database, an increase of 41 percent. The number of surveyed points available for modeling increased by 84 percent. As these sites and surveys were added to the database, a considerable amount of quality control was done on the data. Locations are now more accurate, sites and surveys polygons were digitized, and redundant features were removed.
Archaeological Field Survey
The first two phases of MnModel included archaeological field surveys. In the first summer of surveys (1995), four counties were surveyed using a stratified sampling design to provide data that could be compared to that gathered by the Minnesota Statewide Archaeological Survey (MnSAS), conducted by the Minnesota Historical Society between 1977 and 1980. In the second summer (1996), three additional counties were surveyed using a simple random sample. The intent of this survey was to estimate the a priori probability of finding sites in different regions of Minnesota and to improve model testing. MnModel survey standards and procedures for these two surveys are described in Appendix C of the MnModel Final Report Phases 1-3 (2002). Appendix D details the survey results.
Deeply Buried Sites
Most known archaeological sites in Minnesota are at or near the surface. Because these dominated the archaeological database, and because the environmental variables described surface conditions, the predictive models are used only to assess the potential for surface sites. However, there are many environments in Minnesota, particularly in river valleys, that have the potential to contain sites buried more than one meter below the surface. Subsurface survey is expensive, so MnModel required a component to help determine where such surveys are needed. To achieve that goal, MnModel Phases 1 through 3 included extensive geomorphic mapping of selected parts of the state thought to have high potential for deeply buried sites. From this mapping, models were developed that indicate the suitability of the landform sediment assemblage to have both captured and preserved archaeological artifacts in situ. Details of the mapping project and resulting models can be found in Chapter 12 of the MnModel Final Report (2002).
Subsequently, additional parts of the state have been mapped. These reports can be accessed from the Geomorphology portion of this site. In addition, MnDOT conducted a study comparing several standard methods for locating deeply buried archaeological sites for their cost and effectiveness (Deep Testing Protocol, 2006). Based on this research, a field protocol was developed to guide field work when deeply buried sites might be present.