Minnesota Department of Transportation
Jump to main content.
Minnesota Department of Transportation Go to 511
MnDOT A to Z | General Contacts | Search |
Survey Implementation Model Map Cross-section
 

Mn/Model

Final Report Phases 1-3 (2002)

Contact Us   Mn/Model Home | Archaeology | Geomorphology | Geographic Information Systems (GIS) | Implementation

 

Quick Links

Chapters

Appendices

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Chapter 5

The Archaeological Database

 

By Guy E. Gibbon, Craig M. Johnson and Stacey Morris

Statewide Survey Impelmentation Model Map

 

 

Chapter 5 Table of Contents
5.1 Introduction
5.2 State Site Files
5.3 The Minnesota Statewide Archaeological Survey
5.4 Mn./Model Fieldwork
      5.4.1 Comparison of MnSAS and Mn/Model
      5.4.2 Previous Evaluations of MnSAS
      5.4.3 Bias in the Mn/Model Survey
5.5 The Mn/Model Archaeological Database
5.6 Known Archaeological Sites
      5.6.1 SHPO Database
      5.6.2 Chippewa National Forest Database
      5.6.3 Superior National Forest Database
5.7 Survey Literature and Negative Points
       5.7.1 Probabilistic Surveys, Including Mn/Model and the Minnesota Statewide Archaeological Survey
       5.7.2 Negative Survey Points from Probabilistic Surveys Other than Mn/Model
       5.7.3 Negative Survey Points from Mn/Model Surveys
       5.7.4 Non-Probabilistic Surveys
       5.7.5 Negative Survey Points from Non-Probabilistic Surveys
5.8 Assembly and Quality Control of the Archaeological Database
       5.8.1 Assembling and Reviewing the Database
       5.8.2 Submitting the Database to GIS
5.9 Mn/Model Archaeological Database Specifications
      5.9.1 Naming Conventions
      5.9.2 Software and Formatting
      References

 

5.1 INTRODUCTION

The foundation of a predictive model of archaeological site location is the archaeological database. However, archaeological databases for predictive models vary widely for many reasons, including: (1) range and quality of existing data; (2) the methods by which the field information was collected; (3) ease of access of the data; (4) the purpose of the predictive modeling project; and (5) the milieu in which the data is to be managed. For example, it is usual (1) to have mostly the largest and most visible sites in a region recorded in site files; (2) for the information recorded for particular sites to be uneven and unreliable; (3) to have site location data gathered using non-random search procedures; and (4) to have site data only available in a format that makes it difficult to access and manipulate in statistical analyses. Because the accuracy and power of predictive models depend on the representativeness, reliability, and accessibility of its archaeological database, special attention was given in this project to the organization and establishment of the Mn/Model archaeological database (see Altschul and Nagle 1988; Kvamme 1988, for general discussions of many of the conceptual and methodological issues mentioned in this chapter). This chapter discusses these problems as they relate to the database used for Mn/Model.

 

Chapter 5 describes Mn/Model’s archaeological database. It includes sections on: the history and content of state site files (Section 5.2); the Minnesota Statewide Archaeological Survey (MnSAS), which provided the probability-based database required for the Mn/Model project (Section 5.3); fieldwork carried out during the project to evaluate the representativeness and reliability of the MnSAS database and, more broadly, the state site files (Section 5.4); and the archaeological database used to develop the project’s predictive models (Section 5.5).

 

Return to Top

 

5.2 STATE SITE FILES

As archaeology as a discipline became more formalized and professional in the United States in the 1930s and 1940s, interest gradually grew in recording the location and contents of sites in a uniform and systematic manner. A trinomial Smithsonian Institution system was adopted in most states that assigned each site a unique, three-part name composed of the number of the state in alphabetic order (e.g., 21 for Minnesota), a county designation (e.g., FL for Fillmore County), and a number representing the order in which a site was added to a county site database (e.g., 21FL5). Today, these files are usually maintained by the Office of the State Archaeologist (OSA) or by the State Historic Preservation Office (SHPO).

 

The first sites in Minnesota to be added to county site databases were earthworks and larger sites with dense artifact assemblages. A focus on these sites reflects the early interest in mounds stimulated by the widespread belief in the nineteenth century in the existence of civilized, non-Indian Mound Builders and by the dominance in North American archaeology of taxonomic approaches in the 1940s and 1950s (Gibbon 1981). As CRM archaeology accelerated in importance after the mid-1970s, a wider variety of site types were added to the files. With the shift in focus from site-specific investigations to the survey of bounded project areas (Binford 1964; Schiffer et al. 1978, 1982), new search procedures were adopted. Among the new survey methods were shovel-testing (shallow, sub-surface probing with shovels) and patterned transect search procedures. Many smaller sites, especially sparse lithic scatters (small concentrations of culturally-altered stone), were added to site files using these methods. For discussions of the issues involved in this transformation, see Ammerman (1981), Dunnell and Dancey (1983), Lewarch and O’Brien (1981), and Redman (1987).

 

Although most state site files now contain thousands of archaeological sites, there are many problems with these databases for building statewide predictive models. The most serious are alluded to in Section 5.1 above, and include:

  1. wide variations in the comprehensiveness and intensity of survey in different parts of the state;
  2. incomplete and/or unreliable information for particular sites;
  3. a lack of representativeness (since random search procedures were not used not all kinds of sites were given an equal chance of discovery);
  4. limiting surveys to areas where archaeologists expect sites to be, which systematically under represents some kinds of sites in the site files;
  5. the incorporation of data into the site files using different systems of terminology; and
  6. non-computerized data.

 

Together, these problems make existing site files a non-standardized database for the construction of predictive models of site location.

 

The Mn/Model project was fortunate in having available an archaeological database that could be prepared for predictive modeling. The site files were in the process of being standardized and entered in a digital database format by members of the Minnesota State Historic Preservation Office (SHPO) through a Federal Highway Administration (FHWA) Intermodal Surface Transportation Efficiency Act (ISTEA) grant from the Minnesota Department of Transportation (MnDOT). However, this database was not without problems. Aside from the general issues listed above, two particular issues were sources of concern in developing and evaluating the models.

 

The first question concerned the inclusion of lithic scatters in the population of sites modeled. As defined by the Minnesota State Historic Preservation Office Archaeologists, "a lithic scatter is a site that has yielded almost exclusively lithic artifacts and contains no ceramics or obvious features; there are no site area, site integrity, or significance requirements for the definition. Lithic scatters can consist of two flakes or ten thousand stone tools. They can have an area of two square meters or several hectares. Functionally, lithic scatters can be habitation sites, workshops, quarries, kill sites or almost any other type of site where people made or used stone tools. While historic context association is not always immediately evident at a lithic scatter, many lithic scatters are significant because they represent such an extensive part of Minnesota’s past." Moreover, many sites classified as artifact scatters may be lithic scatters, because investigators reporting sites to SHPO were inconsistent. Some classed sites as artifact scatters if they had formed tools, while others used the term only when ceramics were present. Given the broad definition of lithic scatters in the SHPO database, it was not clear whether a site in this classification provided important information, due to the presence of significant sites, or noise, due to the presence of insignificant or perhaps randomly distributed sites. A discussion of the implications of including or excluding lithic scatters is found in Section 6.4.1.1. Section 8.6.2 evaluates the contribution of these sites to Phase 2 model results.

 

A second concern was the accuracy of the locations of site centroids contained in the SHPO files. This problem became apparent in the early stages of the project, when it was observed that sites often occurred within water bodies or across the state borders. There are a number of possible explanations.

  1. The sites could be accurately located, but the GIS data could be poor because of source scale or data entry errors. This is conceivable for errors of a certain size, but could not explain the occurrence of sites hundreds of meters within lakes or outside of Minnesota.
  2. Sites located within water bodies could be accurately located, but the water level when the water body was mapped was higher than when the site was recorded. Again, this can explain only relatively small errors.
  3. The coordinates recorded on the site forms could be in error.

 

The Mn/Model geomorphology project completed an examination of site locations in conjunction with their analysis of the relationship between the temporal placements of Landform Sediment Assemblages (LfSA) and the archaeological sites that occur on or in them (see Section 12.17). This study came to the conclusion that there are a number of location errors in the SHPO database for sites in eight valley/bog project areas. Forty-five out of 880 sites were identified as having "aberrant" locations with respect to LfSA’s. These were traced to seven causes including vague locational data, skeptical site location data, map scale issues, incorrectly calculated UTM coordinates, misinterpreted LfSA age or polygon boundaries, digitizing mistakes, and "unknown" reasons. Based on these results, a larger study was conducted of 778 out of 880 sites to determine if there were any additional errors in site location for the eight valley/bog areas under study by the geomorphologists. This review came to the conclusion that 55 of these sites (7.1 percent) had incorrect UTM locations. These errors are not as important to the archaeological predictive models as they are to the geomorphic models, since the site models calculate distances from sites (points) to polygons.

 

Site records were examined for the 45 flood plain sites having "aberrant" locations. These sites were selected because their ages or locations did not appear to fit the geomorphology model. The results of the investigation revealed important information about the quality of the SHPO site files. These include:

  1. Recorded level of detail varied between sites, depending on the type of survey conducted, type of site located, author, and date of site form.
  2. Locational data for some sites was vague, with site boundaries over-estimated or not known.
  3. Locational data was skeptical for some sites. The artifacts may have been moved from their original location by natural or cultural processes, the site may have been recorded on the basis of an informant’s report with no field verification, or the site location may have been mapped incorrectly.
  4. UTMs were incorrectly calculated for some sites.

 

Return to Top

 

5.3 THE MINNESOTA STATEWIDE ARCHAEOLOGICAL SURVEY

The Minnesota Statewide Archaeological Survey (MnSAS) was a large-scale project funded by the Minnesota Legislature ($400,000) and carried out by the Minnesota Historical Society in 1977-1980 (Minnesota Historical Society 1981). Among the primary charges of MnSAS were: (1) "to properly collect existing data and acquire data by survey on additional significant sites"; (2) to use this data "to eliminate delays in environmental assessments and impact statements"; and (3) to collect and organize the data "in a manner consistent and compatible with the Minnesota land management information system." To implement MnSAS, the Historical Society established four goals: (1) "To formulate models that predict the distribution of archaeological sites throughout the state"; (2) "To locate additional archaeological sites"; (3) "To update as necessary the archaeological site files of the State Archaeologist (a list of currently known archaeological sites)"; and (4) "To design and implement an archaeological site data bank that is compatible with, and takes full advantage of, the Minnesota Land Management Information System (MLMIS)."

 

Because of great differences in Minnesota’s landscape, a division of the state into physiographic areas was adopted to distribute individual surveys. It was assumed that a division based on more permanent features of the landscape - the state’s basic geomorphology - was more valuable for the predictive modeling of the location of prehistoric sites than were features like modern climate or vegetation. The system adopted was Professor Herbert Wright, Jr.’s (1972) division of the state into 27 physiographic areas (Figure 3.9). Because each area is small and formed by a group of similar, related glacial events, there is a degree of internal consistency of soils, topography, climate, and vegetation that suggests an internal consistency in the distribution of prehistoric archaeological sites. The predictive models that were a product of MnSAS were developed for these areas. Of the 27 physiographic areas, eight had sample surveys (Brainerd-Automba Drumlin area, Anoka Sand Plain, Eastern St. Croix Moraine, Owatonna Moraine area, Blue Earth Till Plain, Glacial Lake Agassiz, Coteau Des Prairies [Inner Part], Rochester Till Plain).

 

The design of the MnSAS sampling strategy was determined by: (1) the need to obtain representative samples of data through probabilistic sampling; and (2) the mandate to find additional significant sites, which was felt necessary to justify the cost of the survey. Since a preliminary review of other large-scale surveys and shared personal experience among archaeologists suggested that significant sites (i.e., larger sites that merit mitigation) are close to water, a general "Away-from-Water" and "Adjacent-to-Water" dichotomy was adopted and an emphasis placed on surveying Adjacent-to-Water locations to maximize the number of "significant" sites found. The Adjacent-to-Water stratum was subdivided in all but three counties to explore further the association of sites with particular Adjacent-to-Water landscape features. The exception was the Houston-Fillmore-Winona survey in which a simple dichotomy was made between upland (Away-from-Water) and low river valley settings (Adjacent-to-Water). Because of differences in regional landscape features, the Adjacent-to-Water subdivisions varied widely. Among the landscape features used in various surveys were islands in lakes, lake inlets and outlets, stream confluences, other lakeshore, other streamshore, main stem of major river, and marsh/wetlands.

 

Because of the mandates to survey many areas of the state and to find significant sites, a 50-m "weaving" transect approach was adopted in most surveys. In this approach, surveyors cross a survey area, which was a quarter-quarter section (40 acres) in MnSAS, about 50-m apart but wander from a straight line to inspect promising landscape formations. Although MnSAS archaeologists conceded that a 50-m transect width was too large to discover many of the smaller lithic scatters that occur in the state, they felt a narrower transect interval would not allow them to survey the number of quarter-quarter sections necessary to get a sample size sufficient for statistical tests.

 

By the end of the project, the Historical Society had: (1) surveyed parts of 26 counties (Anoka, Blue Earth, Brown, Carver, Chisago, Clay, Crow Wing, Dakota, Douglas, Faribault, Fillmore, Goodhue, Houston, Isanti, Itasca, Kanabec, Kandiyohi, Koochiching, Nobles, Pine, Pipestone, Redwood, Rock, Sherburne, Washington, Winona) using stratified random or intuitive search procedure; (2) developed predictive models of prehistoric site location; (3) found more than 900 previously unreported precontact archaeological sites and added them to the state site files; (4) field-checked reported but problematic sites in the State Archaeologist’s site files; and (5) created an archaeological data bank compatible with MLMIS, a statewide GIS database at 40 acre resolution. Three surveys were conducted in 1980 after the initial project report was completed. These included surveys in eastern Itasca County, in Kandiyohi County (to explore "various ways of obtaining information about archaeological sites from knowledgeable local residents and collectors"), and along the spine of the Coteau des Prairies in the southwestern corner of the state in a search for boulder effigies and outlines.

 

Five conclusions emerged from the analysis. They were: (1) "Throughout nearly all of the surveyed portions of Minnesota, prehistoric sites were found most frequently on land adjacent to shorelines"; (2) "In areas where lakes are present, prehistoric archaeological sites usually occur with greater density on lakeshores, rather than on river or stream shores"; (3) "In regions where lakes are absent and rivers are deeply incised, sites may be located at greater distances from water"; (4) "In most ‘away-from-water’ areas, prehistoric archaeological sites are often very small and apparently occur at a low density"; and,  (5) "The ‘Driftless Area’ of southeastern Minnesota is apparently different from the rest of the state. Prehistoric archaeological sites located there are not predominantly located near shorelines but are more widely dispersed." Statistical data from the MnSAS surveys were used to support these conclusions in the summary report (Minnesota Historical Society 1981). Information from other large-scale surveys in the state (the Minnesota Trunk Highway

Archaeology Survey between 1973-1980; the Dome Pipeline Right-of-Way survey through southwestern Minnesota; Superior National Forest surveys in 1978 and 1979; the Voyageurs National Park survey in 1976; and, the 1980 Chippewa National Forest survey) were used to support these conclusions. The penultimate conclusion of MnSAS was that in much of Minnesota, sites are near shorelines, and those that are not are usually small and occur at low densities.

 

The results of the MnSAS project provide an unusually large-scale archaeological database gathered using probabilistic search procedures. This kind of large-scale, probability-based database is ideally suited to the creation of statewide predictive models of archaeological site location. Because of the special emphases of MnSAS, however, the database is not without its potential problems as an unbiased sample of Minnesota’s prehistoric archaeological sites (Birk 1979:52-80; Henning 1983). The most important of these problems are: (1) the potential bias inherent in the MnSAS stratified sampling design (Were densities of sites in the Away-from-Water stratum underestimated?); (2) the 50-m transect interval (Were small sites between 50-m transects overlooked?); and, (3) the ‘weaving’ search procedure (Did unpatterned straying from straight transect lines bias transect-line inventories?). A necessary element of the Mn/Model project, then, was a systematic assessment of the effect, if any, of these problems on the content of the MnSAS archaeological database.

 

Return to Top

 

5.4 MN/MODEL FIELDWORK

Archaeological field surveys for the Mn/Model project were conducted in 1995 and 1996 by three field crews. The results of these surveys, besides the procedures and rationale that guided these efforts, are detailed in Appendices C and D. This section presents a preliminary comparison of the MnSAS and 1995 Mn/Model fieldwork databases. As Section 5.3 stressed, it was the project team’s opinion that the MnSAS survey results could not be confidently used as a base for Mn/Model without assessing potential conflicts between MnSAS and contemporary field methods. A particular concern was the 50-m distance between transects, for it seemed likely that the MnSAS database under represented numbers of sites present. If this were the case, the database would have to be "calibrated" to more closely represent actual densities of archaeological sites. The 1995 archaeological field survey was designed to be a comparative study to help in the analysis of the results of the MnSAS surveys. Among the search procedures adopted were a 15-m transect interval, and standardized search and stratification procedures (see Appendix C). The following discussion compares and evaluates the results of the 1995 and MnSAS surveys.

 

5.4.1 Comparison of MnSAS and Mn/Model

There were several major differences in the way each survey was conducted, including consistency of defined strata and field techniques (compare Section 5.3 and Appendices C and D). Although the details of how each county survey was planned and executed for MnSAS will probably never be available, as they are for the 1995 Mn/Model field season, a consistent spacing of pedestrian transects and shovel tests was used in both programs. MnSAS used a 50-m interval for both pedestrian surface scanning and shovel testing. By today’s standard of 15-m intervals, the MnSAS standard interval is considered too large to find many of the smaller sites prevalent in Minnesota. The following analysis is presented to evaluate these differences.

 

Table 5.1 presents the results of an analysis of survey results for seven counties. Portions of three of these counties were surveyed in 1995 (Stearns, Nicollet, Beltrami) and four (Carver, Isanti, Brown, Itasca) as part of the MnSAS project. The table lists the number of 40 acre units or parcels surveyed, the number of sites recorded, and the occurrence or rate of sites per 40 acre unit for three landforms (streamshore, lakeshore, Away-from-Water). These divisions are similar to the ones used in the MnSAS project (Minnesota Historical Society 1981). The actual numbers employed here for MnSAS counties may differ from those in the published version due to pooling of counties or the elimination of sites discovered in intuitive surveys. Three of the counties surveyed during the 1995 Mn/Model field season (Stearns, Nicollet, Beltrami) were paired with one or two nearby counties to control for the effects that different environments may have on site density. These pairings are Stearns-Carver-Isanti, Nicollet-Brown, and Beltrami-Itasca. Landform strata are used in this analysis for two reasons: (1) to control for their effects on site presence; and (2) to explore the relationships between the presence of sites and different environmental factors. The first point is crucial, since the number of units surveyed within each landform varied from county to county. Without these landform divisions, the rates of site occurrence (i.e., number of sites per unit) would be non-comparable. For example, nearly two-thirds of all units in Carver County were classified as "Away-from-Water," while less than one-fourth of all surveyed units in Stearns County were assigned to this stratum.

 

Referring to the series of values at the bottom of the right-hand column in Table 5.1, the number of sites per 40-acre unit is usually higher for Mn/Model counties than for MnSAS counties. The greatest discrepancies occur in streamshore and lakeshore settings in Stearns and Nicollet counties compared with their paired counties of Carver, Isanti, and Brown. Sites either are absent (streamshore) or occur in small numbers (lakeshore) in these strata in Beltrami and Itasca counties. Except for Nicollet County, there is essentially no difference in the rate of site occurrence in the "Away-from-Water" stratum. In fact, the number of sites per parcel in Itasca County is higher than in Beltrami County, which is part of the Mn/Model sample. Several conclusions can be drawn from this comparison: (1) the MnSAS survey under predicted the presence of sites by a magnitude ranging from about four to eight times in those areas of high site potential that required a mix of pedestrian survey and shovel testing (Carver, Isanti, Brown Counties); (2) there was little or no underestimating of site numbers in the MnSAS survey in areas that required intensive shovel testing (Beltrami, Itasca counties); and (3) with a few exceptions, there was little if any underestimation of sites in the MnSAS database in areas classified as "Away-from-Water."

 

These empirical generalizations could be tested by including additional pairs of counties and by determining what field methods (pedestrian survey or shovel testing) were used to find sites, no matter the county they are from. This latter point is important, for it would help determine if interval distance was a crucial factor in finding sites on exposed as compared with non-exposed ground surfaces. A further evaluation of field methods appropriate for surveying areas distant from water is also needed. It could be that sites in these areas are so scattered and consist of such low numbers of artifacts (e.g., isolated finds) that wide intervals are as likely to find them as are narrow ones. This possibility could apply as well to areas, such as the Chippewa Plains-Pine Moraines and Outwash Plains- regions, which require extensive shovel testing.

 

The project statistician conducted a second, more formal analysis of these same three groups of counties. A mathematical model was created that assumed that the number of sites found in Stratum I in survey J (J=1 is Mn/Model, J=2 is Mn/SAS) has a Poisson distribution with a mean equal to A(I, J) exp-->[S(I)+M*(J=1)], where A(I, J) is the number of acres surveyed (the more acres surveyed, the more sites discovered), S(I) the effect of the survey (some strata contain more sites than others), and M the effect of Mn/Model survey methods compared with Mn/SAS survey methods. The basic goal of the analysis was to find the probability of M=0, which equates with no difference in survey methods (the null hypothesis). The results are presented in Table 5.2. For the Nicollet-Brown County pair, the null hypothesis of no difference can be rejected, since p = .004.

 

Table 5.1. Summary Statistics for Seven Counties Surveyed for MnSAS and Mn/Model.*

Measure/County

Landform Strata

Stearns County

42

61

32

135

Carver County

27

33

90

150

Isanti County

28

37

38

103

         

Nicollet County

41

65

29

135

Brown County

18

25

39

82

         

Beltrami County

8

35

20

63

Itasca County

5

23

33

61

         

Stearns County

24

56

2

82

Carver County

4

7

7

18

Isanti County

9

5

3

17

         

Nicollet County

11

27

7

45

Brown County

1

4

5

10

         

Beltrami County

0

6

1

7

Itasca County

0

3

3

6

         

Stearns County

0.571

0.918

0.062

0.607

Carver County

0.148

0.212

0.078

0.12

Isanti County

0.321

0.135

0.079

0.165

         

Nicollet County

0.268

0.415

0.241

0.333

Brown County

0.056

0.16

0.128

0.122

         

Beltrami County

0

0.171

0.05

0.111

Itasca County

0

0.13

0.09

0.098

 

*Stearns, Nicollet, and Beltrami Counties were surveyed during the 1995 Mn/Model field season. Carver, Isanti, Brown, and Itasca Counties were surveyed for the MnSAS.

 

 

The Nicollet County survey found 2.75 more sites than the Brown County survey, which was part of the MnSAS sample. In the Stearns-Carver-Isanti comparison, the null hypothesis of no difference in survey methods was rejected, since p = 0. The Stearns County survey found 5.80 more sites than the combined Carver-Isanti pair. For the Beltrami-Itasca pair, a probability value of .77 means that the null hypothesis of no difference in the number of sites between the Mn/Model and MnSAS survey cannot be rejected. Since the number of sites is so low in these counties, an M effect is unlikely to be discovered even if it does exist.

 

Table 5.2. Summary Statistics for the Comparison of Site Occurrences Between Mn/Model and MnSAS.

 

p-value

M

se

Site Factor

Nicollet-Brown

.004

1.008

.367

2.75

Stearns-Carver-Isanti

.0

1.759

.203

5.80

Beltrami-Itasca

.77

.0

-

-

 

 

Another approach to determining how common sites are in a region is to calculate how many hectares or acres need to be surveyed to find one site. This can be calculated by dividing the actual land area surveyed by the number of sites found. Since the actual amount of land surveyed for MnSAS is not known (i.e., for each 40-acre unit, it is impossible to determine how much was surveyed and where a survey was conducted), figures cannot be calculated for these counties. The amount of surveyed area is; however, available for both the 1995 and 1996 Mn/Model surveyed counties. It is these counties, except Becker County, that formed the basis of the following analysis. It should be kept in mind that since the 1995 and 1996 survey designs were very different (i.e., the 1995 survey used a stratified random sampling procedure weighted toward water and 1996 a simple random sampling procedure), radically different figures resulted. Table 5.3 lists the six Mn/Model counties included in this analysis. The purpose of the analysis was to learn the background occurrence of sites and to compare those figures from region to region. As in the previous analysis, landforms for the 1995 survey were combined into three strata (streamshore, lakeshore, away from water) to increase sample sizes. Since the 1996 field work was a simple random survey not based on landforms, only one set of figures is available for each of these counties. In addition, the number of single versus multiple artifact sites for the 1995 survey is not currently available for the three landform strata.

 

Referring to Table 5.3, there is a wide range of site occurrence values from one landform to another among the three 1995 Mn/Model counties. Lakeshore contains the highest concentration of sites, ranging from one site every 3.9 hectares (9.6 acres) in Beltrami County to one every 27.2 hectares (67.2 acres) in Nicollet County. More survey was required in Stearns (16.1 hectares or 39.7 acres) and Nicollet (39.3 hectares or 97.1 acres) counties to find a single site in streamshore locations. In "Away-from-Water" areas, it took more than 200 hectares (500 acres) of survey to find a single site in Stearns County, while less than 15 hectares (40 acres) of survey was required in Beltrami County.

 

Table 5.3. Summary Site Occurrence Statistics for Six Counties Surveyed for Mn/Model.

Measure/County

Landform Strata

Stearns County

386.04

479.08

420.88

1286

Nicollet County

432.21

733.91

429.95

1596.07

Beltrami County

6.07

23.23

13.31

42.49

Wright County

-

-

-

1259.17

Wabasha County

-

-

-

681.34

Cass County

-

-

-

36.87

         

Stearns County

3.29

22.76

0

26.05

Nicollet County

4.43

21.37

6.75

32.55

Beltrami County

0

0.86

0.02

0.88

Wright County

-

-

-

5.99

Wabasha County

-

-

-

9.71

Cass County

-

-

-

0

         

Stearns County

-

-

-

53

Nicollet County

-

-

-

21

Beltrami County

-

-

-

6

Wright County

-

-

-

8

Wabasha County

-

-

-

19

Cass County

-

-

-

0

         

Stearns County

16.1(39.7)

8.5(21.1)

210.4(520.0)

15.7(38.8)

Nicollet County

39.3(97.1)

27.2(67.2)

61.4(151.8)

35.5(87.7)

Beltrami County

--

3.9(9.6)

13.3(32.9)

6.1(15.0)

Wright County

--

--

--

48.4(119.7)

Wabasha County

--

--

--

20.6(51.0)

Cass County

--

--

--

18.5(45.6)

         

Stearns County

0.0085

0.0475

0

0.0203

Nicollet County

0.0102

0.0291

0.0157

0.0204

Beltrami County

0

0.037

0.0015

0.0207

Wright County

-

-

-

0.0048

Wabasha County

-

-

-

0.0143

Cass County

-

-

-

0

 

 

Based on county totals, the background occurrence of sites is greatest in Beltrami County and smallest in Nicollet County; Stearns County falls between the two. When the simple, unstratified random surveys of the 1996 field season are included, site occurrence drops significantly. Direct comparisons can be made between two sets of counties in similar environmental settings. The occurrence of sites in Wright County (48.4 hectares/site or 119.7 acres/site) is three times less than in Stearns County (15.7 hectares/site or 38.8 acres/site). The Wright County figure may be taken as a close approximation to what is probably occurring in other central Minnesota counties, regardless of landform type, since it is based on a simple random sample. Coincidentally, a similar numerical relationship holds for the Central Lakes Coniferous region (Phase 2 region) or the Chippewa Plains region (Phase 3 region), where the 1995 stratified sampling procedure of Beltrami County overstates the purely random occurrence of sites in Cass County by three times (18.5 hectares/site or 45.6 acres/site versus 6.1 hectares/site or 15.0 acres/site). No direct comparisons can be made between the other two counties, Nicollet and Wabasha, for they have very different environmental settings.

 

The last set of figures in Table 5.3 are a priori probabilities of the occurrence of sites for the six counties used in these analyses. A priori probabilities of site occurrence are the chances of finding archaeological sites without any prior knowledge (e.g., models) of where they are situated across the landscape (Kvamme 1988a:410, 1990:260, 1992:28). Sometimes called base or by-chance rates, a priori probabilities can be calculated in many ways. The most common way is to calculate the ratio between the frequency of survey units containing sites to the total number of units surveyed (see Table 5.2). A more refined technique is to construct a ratio between the summed site areas and the actual area surveyed (Kvamme 1988a:410, 1990:260), as was done in Table 5.3. For a priori probabilities to have their greatest utility, they should be based on probabilistic survey data, preferably obtained using a simple random survey design.

 

According to Kvamme (1992:28, 1990:260-261), a priori probabilities can be used in several ways. First, they provide site occurrence base rates expected by pure chance. These probabilities can be adjusted for any land area. For example, if a simple random survey of a county based on 40-acre parcels yielded the a priori probability of 0.01 of finding a site, then one percent of the land in the county could be expected to contain sites. These base rates yield estimates of the relative sparseness of sites and provide a figure for their occurrence by chance alone. Low a priori probabilities can also be used to justify the selection of randomly dispersed "site absent" locations for comparison on environmental variables with locations where sites are present, as in this project (see Chapter 8), though most of the "site absent" areas have never been surveyed (see Kvamme 1992:28). Second, a priori probabilities provide a measure that can be used to estimate the success of predictive models. That is, models that are successful must partition the landscape into "site presence" zones that have much higher probabilities of site occurrence than the a priori rates and "site absence" zones that have much lower rates.

 

The a priori probabilities of site occurrence for counties surveyed during the 1995 and 1996 Mn/Model field seasons are listed in Table 5.3 along with other data used in their calculation. These probabilities were calculated by dividing the sum of all site areas by the actual area surveyed within each county. Because areas were employed, only those sites containing two or more artifacts (i.e., multiple artifact sites) were considered. The probabilities by sampling strata in the 1995 counties (Stearns, Nicollet, Beltrami) range from 0.0 to .0475. Overall, the lakeshore stratum has the highest rates, while the "Away-from-Water" stratum has the smallest; streamshore generally falls between these two extremes. The situation is very different when probabilities for entire counties are examined. County probabilities for 1995 have a remarkably narrow range of .0204 to .0207. However, the site occurrence rates are not strictly a priori ones, since the 1995 surveys were based on a model of where it was thought sites most commonly occur, which was assumed to be near water. Because the 1996 surveys of Wright and Wabasha counties were simple random surveys, the figures calculated for them are, therefore, the most valid a priori rates presently available for counties (Cass County was excluded from the calculations because no multiple artifact sites were found during the 1996 probabilistic survey). Wabasha County, at .0143, is three times the rate of Wright County (.0048). Multiplying these figures by 100 to convert to percentages suggests that about 1.5 percent of the surveyable land in Wabasha County contains an archaeological site, which is defined in the project as a spatial locus with two or more artifacts, while less than one-half of one percent of the overall landscape in Wright County contains a site. By way of comparison, the stratified survey in Stearns County, which borders Wright County, yielded a site occurrence rate 4.3 times the Wright County figure (.0206), an indicator of the performance of the intuitive model used to design the Stearns County survey.

 

The figures for Wright (.0048) and Wabasha (.0143) counties become more meaningful when they are compared to a priori probabilities for other areas of North America. Studies by Kvamme (1988:410-411), Warren and Asch (1996:5), and Young et al. (1995:Appendix F) yielded a priori rates for project areas of .021 in Colorado, .0053 in a prairie zone of Illinois, and .0015 in southern Ontario, respectively. These site occurrence rates are in the same general range as the Wright and Wabasha county figures, except the southern Ontario figure, which is much lower. This may be in part because sites less than 5 m2 in extent were excluded from consideration in the Ontario study. Other a priori probabilities based on dividing the number of survey units (e.g., 40 acre parcels) containing sites by the total number of units surveyed generally yield higher, but less accurate, rates. Examples include project areas from Colorado (.045, .065), Arkansas (.0725) and Central Montana (0.1) (Carmichael 1990: 221-222; Kvamme 1992:27-28, 1988a:410; Parker 1985:187). Comparable rates can be calculated for multiple artifact sites in Wright (8 parcels/139 parcels = .0576) and Wabasha (16 parcels/85 parcels = .188) counties. The Wright County figure is similar to the other a priori rates, while the Wabasha County figure is much higher. Yet another method of estimating site densities is to calculate the number of sites in a square mile. Otinger et al. (1981:114-115) summarizes these figures for the Southwestern U.S. (2.6 to 65.0) and for Arkansas (1.1 to 7.5). The figures for Wright (8 sites/4.86 miles 2 = 1.65) and Wabasha (19 sites/2.63 miles 2 = 7.22) counties are comparable to those from Arkansas but fall at the lower range of site occurrences elsewhere.

 

5.4.2 Previous Evaluations of MnSAS

Douglas A. Birk (1979) and Elizabeth R. P. Henning (1983: 4.2-4.15) have evaluated the results of MnSAS. Henning’s (1983:4.4) criticism revolves around the fact that the number of units surveyed within each stratum was not proportional to their occurrence in the sampling universe. As a basis for her argument, she cites Black’s (1972:518) discussion of weighting of results when disproportional stratified sampling is used. To summarize a later revised version of Blalock (1979:564-565), if a population is divided into several strata, then the population mean can be estimated by simply summing the numbers where proportional sampling is used (i.e., the numbers in each sample are proportional to their frequencies within the population). Because of this, proportional stratification is self-weighting. When stratification is disproportional, each sample mean must be multiplied by the weight of the stratum within the population. For example, if one stratum is four times larger than another, its mean should receive four times as much weight as the second stratum when adding the samples for an estimate of the population mean.

 

Although Henning (1983) recognized this idea, her application of it to MnSAS was incomplete and incorrect. Henning (1983:4.14) states that the low incidence of sites found in Away-from-Water strata could be due to sampling error, since these strata were under sampled in the MnSAS. Determining sample size is an important topic in statistics, but it cannot be determined by simply examining the relative sizes of the population strata. An informed decision about sample size has to be based on the additional criteria of confidence or significance level of a test, the accuracy within which a parameter is estimated, and estimates of parameters (e.g., mean, standard deviation) (Blalock 1979:216-217). Once these are known, an adequate sample size can be determined. It is clear from the procedures in MnSAS (Minnesota Historical Society 1981:7-11) and Henning’s (1983) reanalysis that this was not done in her analysis. Besides this point, Henning’s (1983) recalculation of site occurrence statistics based on weighting is incorrect because she did not work with the mean number of sites but the actual frequencies of sites, or the ratios between them. The method used to calculate the rate of site occurrence for each stratum by MnSAS is appropriate and has been used by later researchers in other predictive modeling studies (see the discussion of a priori probabilities in Section 5.4.1). However, making inferences about the number of sites in the population based on the samples in MnSAS has not been evaluated. Henning (1983:4.14-4.15) does point to several potential problems in MnSAS, including non-random selection of survey units and the neglect of site variables, such as size, function, and chronology, in the analysis.

 

Birk’s (1979:59-76, 101-104) criticism of MnSAS focuses on the Nokasippi River region in Crow Wing County. His concerns were the following: (1) lack of pre-survey research useful in constructing and testing settlement pattern hypotheses; (2) focus of the survey effort on quadrats near water at the expense of those in Away-from-Water locations; (3) a vague and simplified sample stratification system that reduces the variability in water resources; (4) limited time available for assigning quadrats to strata; (5) inter-observer error in assigning quadrats to strata; (6) failure to establish a ranking system by which quadrats that have features of two or more strata can be consistently assigned to a single stratum (e.g., islands first, inlets/outlets second, lakeshore third); (7) hypothetical stratification problems dealing with precision in mapping and quadrats with multiple water sources; (8) midstream shifts in MnSAS objectives from identifying the cultural context and function of sites useful in settlement pattern studies to determining the spatial distribution of archaeological sites and site frequency differences between strata; (9) no evaluation of the potential differences in coverage or results between shovel testing and surface pedestrian surveys; (10) intuitive placement of shovel tests; and (11) a 50-m pedestrian and shovel test interval that favored the discovery of larger sites (i.e., those near water) at the expense of smaller sites, which were thought to be more common in upland areas.

 

These criticisms of MnSAS are not evaluated in this report. Rather, they were taken into consideration in designing the Mn/Model surveys, with some criticisms rejected and others accepted. For example, Birk’s (1979:66,102) suggestion that a comprehensive pre-survey study of paleo-environmental conditions and cultural-ecological correlations be conducted to construct and test hypotheses of human group/land relationships may not fit into the broader goals of predictive modeling for the entire state or of CRM modeling in general. If a sampling design was constructed for each region of the state based on its own unique set of environmental variables, interregional results would be difficult to compare. Establishing a single sampling strategy is better, as was done during the 1995 and 1996 Mn/Model surveys. Testable hypotheses can be developed later. Results can then be compared from region to region. In addition, Birk’s criticism seems aimed at the construction of settlement models rather than CRM predictive models in which the reasons for site location are not a primary concern (Tomlin 1990:167-225; Warren 1990:90, 94-95; Leusen 1996:181-185).

 

Birk’s observation, which deals with the emphasis on surveying in areas near water, focuses on the fact that areas Away-from-Water are under represented. There is some validity to this argument, and it is one reason the 1996 Mn/Model survey used a simple random sampling design. Since the 1995 Mn/Model survey was designed to evaluate MnSAS, it shares with that survey the emphasis of surveying near water. Although his idea that a simplified stratification system reduces the variability in water resources also has merit, it is difficult to conceive of an objective and cost-effective method for stratifying a sample based on additional hydrological criteria while maintaining adequate sample sizes. The concepts of time limitations and inter-observer error are related in the sense that smaller amounts of training time for landform coding can result in more inter-observer error. There were also time constraints on the 1995 Mn/Model coding of parcels, particularly for Stearns County, where the coding of the hydrologically complex 40-acre parcels was done primarily by individuals working with separate maps. However, the work of the coders was continually checked during a training period until all coders were making the same basic decisions. This helped reduce inter-observer error. The design of the 1996 Mn/Model survey, which was based on a simple random sample of 40-acre parcels, eliminated this potential source of error.

 

To address Birk’s concern that features with characteristics of two or more strata could not be consistently assigned to a single stratum, the 1995 Mn/Model field effort relied on a series of precedence rules, which were designed to assign parcels that included two or more landforms into a single stratum based on a rarity rule (Appendix C). Since the 1996 Mn/Model surveys were based on a simple random sample, no prior landform coding or associated rules were involved. The concerns in Birk's seventh point, mapping precision and multiple water sources, are also addressed in Appendix C. Furthermore, the entire population of parcels was laid out in a grid before using a USGS land locator template before landform coding was initiated. His eighth point, changing objectives, does not apply to Mn/Model, since the objective of recording only site location and size did not change over the two years of field survey.

 

Birk’s assertion that there was no evaluation of potential differences between shovel testing and pedestrian surveys is a valid criticism that has not been formally addressed by Mn/Model. Although it is a common perception among field archaeologists that pedestrian surveys are more likely to encounter cultural materials than shovel testing surveys, the relative advantages of the two techniques have not been evaluated in Minnesota. An extensive amount of research has been carried out, however, to evaluate the way transect spacing, site size, and site density affect the chances of encountering cultural materials (Kintigh 1988; Shott 1985; Wandsnider and Camilli 1992). Birk’s concern, which deals with the placement of shovel tests, was addressed in the Mn/Model surveys as part of the archaeology field survey standards, procedures, and rationale (Appendix C). Birk’s statement that the 50-m transect interval used in MnSAS favors the location of sites near water (due to their large sizes) versus those away from water, remains a valid hypothesis. Although this hypothesis has not been tested with the MnSAS sample, it could be tested by comparing the size of sites near water versus those at further distances from water, with the understanding that the sizes recorded for sites encountered during MnSAS are only estimates.

 

5.4.3 Bias in the Mn/Model Survey

During the 1996 Mn/Model field season in Wright County, many 40 acre parcels could not be surveyed for various reasons (Table D.12). Of the 320 parcels considered for survey, 181 or 57 percent were not surveyed. At the time of landowner contact, most (19 of 24) of the parcels along large lakes could not be included in the survey because of landowner denial, extensive development, or for other reasons. The rate of rejection of these parcels, 79 percent, is much higher than the general rejection rate of 57 percent for the entire survey. If the general rejection rate of 57 percent is calculated for the 24 large lake parcels, 14 parcels would have fallen within the "could not survey" category. This amounts to five fewer parcels than the 19 that actually could not be surveyed. There is evidence that one type of site, mounds, is in proximity to large lakes or waterways (Anfinson 1984:23; Birk 1979:53; Lothson 1967:45). To the extent that mounds and other precontact archaeological sites are more commonly situated along large bodies of water compared with other landforms, the Wright County survey may, then, have underestimated the number of sites in that portion of the county surveyed.

 

Table 5.4 was constructed to find out whether there was a systematic bias in the Wright County survey that resulted in an underestimation of numbers of sites. The table lists the frequencies and column percentages of surveyed and unsurveyed parcels broken down by two landform classes: Near Water and Away-from-Water. This landform dichotomy is the same as that employed in the 1995 Mn/Model survey. That is, the Away-from-Water parcels are on upland locations or along bodies of water less than 40 acres in size. The Near Water parcels are along rivers, streams, and wetlands and lakes larger than 40 acres (see Appendix C for a discussion of the stratification system). All parcels except those completely in water or wetlands are tabulated in the table.

 

Table 5.4. Cross Tabulation of Frequencies of Landform Type and Survey Coverage for the 1996 Mn/Model Field Survey of Wright County (column percentages in parentheses).

 

Near Water

Away from Water

Total

Surveyed

58 (48.7%)

78 (46.4%)

136

Not Surveyed

61 (51.3%)

90 (53.6%)

151

Total

119 (100.0%)

168 (100.0%)

287

Chi square = .15; df = 1; p = .70

 

As the figures suggest, there is little difference in the relative frequencies of surveyed parcels as they relate to water (48.7% vs. 46.4%). While parcels near water (48.7%) had a higher chance of being surveyed than those away from water (46.4%), and parcels away from water have a higher chance of not being surveyed (53.6%) compared with those near water (51.3%), it is apparent from the percentages that there is virtually no relationship between the chance of a parcel being surveyed and its relationship to water. Computing a chi-square statistic (Blalock 1979:279-292) can test the lack of a relationship between landform and survey potential. To test this formally, the null hypothesis of no relationship is established at the .01 level. A value of 6.64 or greater is needed to reject the null hypothesis of no relationship between the two variables. A Chi square of .15 indicates that the relationship between the two variables is significant at the .70 level and the null hypothesis cannot be rejected. In other words, a Chi-square value at the .70 level means that in seven out of 10 times, a Chi-square value of .15 or greater can be expected even if there is no relationship between the chances of survey and landform type. The conclusion is that there was no systematic bias in the Wright County survey that led to an underestimation of numbers of sites during the field survey.

 

Return to Top

 

5.5 THE MN/MODEL ARCHAEOLOGICAL DATABASE

Mn/Model is designed to be a planning tool that predicts high, medium, and low probability areas for the location of archaeological sites from the precontact and contact periods in the State of Minnesota. To predict the location of archaeological sites, the group constructing the model first had to establish a database that recorded where known sites are and where sites have not been found.

 

The information used to compile this archaeological database, which includes both site and non-site data, was taken from several different sources. Chief among them was the recently completed database of known sites in the state from the State Historic Preservation Office (SHPO). Other sources of data were the Chippewa and Superior National Forests, Mn/Model archaeological survey seasons in 1995 and 1996, MnSAS, and a variety of CRM surveys.

 

All known sites were represented in the database by the UTM coordinate of their centroids; there was no attempt to establish or include information about site boundaries. Besides the archaeological sites, non-site points (negative survey points) were required for building an accurate predictive model. These were collected from a variety of survey projects (MnSAS, Mn/Model, and CRM). All known sites and negative survey points were incorporated into a separate database for each of Minnesota’s 87 counties. Because of other considerations, some counties had more than one database (see Appendix B). The completed databases were submitted to the GIS group for further manipulation and analysis.

 

The Mn/Model GIS design required that the same information be gathered for each site or negative survey point in the database, following the same procedures and structured in the same way (detailed in Appendix B). Each record in the Mn/Model archaeological database contains the following fields:

1) SITENUM: for known sites, this was the Smithsonian site number or other unique identifier, but for negative survey points, the chosen non-site identifier was 0000

 

2) EASTING: the UTM easting, taken by SHPO, USFS, or by BRW from USGS quad maps

 

3) NORTHING: the UTM northing, taken by SHPO, USFS, or by BRW from USGS quad maps

 

4) ZONE: the UTM zone (14, 15 or 16), taken by SHPO, USFS, or by BRW from USGS quad maps

 

5) DATUM: the North American Datum year (1927 or 1983), taken by SHPO, USFS, or BRW from the USGS quad

 

6) SITE_ACRES: for known sites only: site acreage, if this information appears in the SHPO database or can be established from the 1995 or 1996 Mn/Model survey data

 

7) DENSITY: a rating of artifact density, an attribute based on SHPO site file information collected only during the earliest stage of the data collection process, and so appearing in few counties

 

8) DIVERSITY: a rating of artifact diversity, an attribute based on SHPO site file information, collected only during the earliest stage of the data collection process, and so appearing in few counties

 

9) SRC_CODE: a field containing one of the following numbers indicating the source of the information:

0 = Phase 2 random points from surveyed areas
1 = the site was first located or the point was examined during a probabilistic survey
2 = the site was first located or the point was examined during a non-probabilistic, CRM survey that met the criteria described in Section 5.7.4 for inclusion in the database. These are referred in this report to as the qualified CRM sites.
3 = the site has undergone data recovery/mitigation or Phase III investigation or the equivalent, or is on or eligible for the National Register of Historic Places (National Register) (known sites only). These are referred to in this report as the significant sites.
4 = the site does not fall into categories 1, 2, or 3, but is known from the SHPO database (known sites only)
5 = the site appears in one of the National Forest databases; these sites are excluded from all versions of the end product but, with permission of the Forest Service, are included for modeling (known sites only)

 

10) SITE_TYPE: a field combining information from two fields (site description and function) in the SHPO database, which briefly describes the site or point according to this list:

0 = Not a site (for negative survey points only)
1 = Single artifact
2 = Temporary camp
3 = Base camp
4 = Quarry
5 = Lithic workshop
6 = Kill/butchering site
7 = Fishing site
8 = Ricing site
9 = Sugaring site
10 = Gathering site
11 = Gardening site
12 = Mortuary/cemetery location, without mounds
13 = Mound/earthwork location, may include human remains
14 = Rock art
15 = Cache location
16 = Trading post, contact period only
17 = Artifact scatter (more than one type of artifact) of unknown function
18 = Lithic scatter (only lithic tools and/or debris) of unknown function
19 = Rock alignment, of known or unknown function
20 = Shell midden

 

Site type served to segregate negative survey points, single artifact sites, and all other sites. No analyses or selections of sites were based on distinctions between sites assigned to codes 2-20. Table 5.5 summarizes the archaeological data incorporated into Mn/Model for the Phase 1 models. Counties with incomplete information had not, at the time of writing, been evaluated for Mn/Model. Table 5.5a updates this information for Phase 3.

 

Table 5.5. Summary of Information Included in the Phase 1 Mn/Model Archaeological Database.

County

Total Number of Acres in County from DNR

Total Number of Known Sites from SHPO

Number of Known Sites Used for Mn/Model

Number of Survey Reports at SHPO

Number of Survey Reports Used for Mn/Model

Number of Negative Survey Points

Aitkin

1,275,882

132

107

69

31

351

Anoka †

285,366

163

132

70

23

230

Becker

925,024

96

68

52

24

95

Beltrami

1,954,851

171

184‡

111

29

361

Benton

264,069

26

19

36

7

71

Big Stone

337,852

61

43

?

*

*

Blue Earth †

489,844

322

250

78

11

160

Brown †

395,749

106

85

?

7

98

Carlton

559,669

32

14

43

14

258

Carver †

240,551

152

111

57

?

293

Cass

1,544,046

315

396‡

177

54

743

Chippewa

376,186

62

46

?

*

*

Chisago †

282,813

118

88

45

19

342

Clay †

674,320

87

62

38

8

74

Clearwater

659,023

69

44

62

19

339

Cook

1,027,871

63

28

70

15

213

Cottonwood

415,260

45

46

?

*

*

Crow Wing

739,691

267

229

76

26

378

Dakota

374,907

98

53

92

19

143

Dodge

281,105

39

6

19

7

94

Douglas

460,613

166

119

58

26

223

Faribault †

461,497

152

126

32

7

202

Fillmore

551,380

188

127

56

16

338

Freeborn

462,093

82

64

?

*

*

Goodhue

498,996

237

205

117

25

394

Grant

368,298

45

33

25

9

34

Hennepin

387,773

297

237

245

61

413

Houston

363,808

243

192

64

15

242

Hubbard

639,401

62

47

50

15

159

Isanti †

288,961

70

53

25

6

135

Itasca †

1,871,189

216

307‡

141

40

636

Jackson

425,831

48

31

?

*

*

Kanabec

341,014

132

121

33

15

139

Kandiyohi

551,512

137

113

31

15

94

Kittson

706,662

44

21

38

7

202

Koochiching

2,016,518

86

70

68

19

257

Lac qui Parle

464,521

37

31

?

*

*

Lake

1,462,187

58

26

48

22

369

Lake of the Woods

1,072,369

33

11

26

8

43

Le Sueur

303,041

90

67

?

*

*

Lincoln

651,291

43

29

33

14

140

Lyon

461,908

123

121

49

13+*

59+*

McLeod

323,428

22

8

?

*

*

Mahnomen

373,191

19

16

22

10

86

Marshall

1,160,962

59

53

62

20

334

Martin

466,699

41

35

?

*

*

Meeker

412,638

31

23

41

10

68

Mille Lacs

435,921

106

74

83

4+*

32+*

Morrison

737,659

176

110

93

11

58

Mower

455,114

43

34

20

9

41

Murray

460,659

82

54

39

11

58

Nicollet †

298,668

157

111

?

18

268

Nobles †

462,362

64

50

22

7

163

Norman

544,564

76

52

32

5

82

Olmsted

418,545

60

22

45

21

157

Otter Tail

1,424,257

159

144

81

19

105

Pennington

395,891

17

6

21

6

81

Pine

917,282

140

70

74

22

319

Pipestone

298,576

42

31

30

6

55

Polk

1,279,543

42

32

66

24

258

Pope †

455,250

43

28

30

3

7

Ramsey

108,790

45

18

70

10

36

Red Lake

276,932

23

9

24

12

108

Redwood †

563,963

79

59

31

7

75

Renville

631,656

63

20

?

1+*

15+*

Rice

330,040

80

62

31

10

89

Rock

309,277

72

67

26

6

101

Roseau

1,074,233

55

46

34

9

63

St. Louis

4,306,973

504

446

138

58

1,942

Scott

235,686

93

54

68

13

81

Sherburne

288,409

96

41

42

10

81

Sibley

384,030

41

25

?

*

*

Stearns †

889,142

171

119

68

37

397

Steele

276,348

19

7

27

2

21

Stevens

361,763

39

29

20

9

68

Swift

481,624

19

13

?

*

*

Todd

626,581

35

24

25

4

63

Traverse

371,897

91

73

18

4

211

Wabasha

351,537

107

92

48

15

147

Wadena

347,421

32

24

31

7

34

Waseca

276,776

67

50

25

*

*

Washington

149,595

102

78

108

45

386

Watonwan

281,419

10

6

?

*

*

Wilkin

467,396

39

31

23

7

54

Winona

410,219

108

56

50

10

29

Wright

456,881

159

125

69

19

271

Yellow Medicine

488,779

108

87

?

*

*

* indicates that the county has not been examined by the data collection team.

† indicates that numbers will be updated for the final model.

‡ indicates that information from the National Forest Service is included with SHPO data.

? indicates that information was not available when the table was constructed.

 

 

Table 5.5a. Summary of Information Included in the Phase 3 Mn/Model Archaeological Database.

County

Total Number of Acres in County from DNR*

Number of Known Sites from SHPO†

Number of Known Sites from USFS‡

Number of Known Sites Used for Mn/Model**

Number of Negative surveys, SHPO† †

Number of Negative surveys, used for Mn/Model

Aitkin

1,275,882

165

 

101

351

351

Anoka

285,366

192

 

127

237

237

Becker

925,024

179

 

67

116

116

Beltrami

1,954,851

186

67

155

363

365

Benton

264,069

34

 

19

82

82

Big Stone

337,852

67

 

44

55

55

Blue Earth

489,844

337

 

247

155

155

Brown

395,749

114

 

83

108

108

Carlton

559,669

36

 

13

258

258

Carver

240,551

165

 

97

357

357

Cass

1,544,046

398

138

359

741

741

Chippewa

376,186

68

 

46

107

107

Chisago

282,813

126

 

79

180

180

Clay

674,320

190

 

49

56

56

Clearwater

659,023

94

 

42

339

339

Cook

1,027,871

159

262

266

212

212

Cottonwood

415,260

49

 

38

13

13

Crow Wing

739,691

293

 

214

378

378

Dakota

374,907

169

 

49

140

140

Dodge

281,105

44

 

6

97

97

Douglas

460,613

182

 

112

241

241

Faribault

461,497

167

 

120

201

201

Fillmore

551,380

341

 

115

338

338

Freeborn

462,093

97

 

64

51

51

Goodhue

498,996

519

 

200

407

407

Grant

368,298

48

 

32

41

41

Hennepin

387,773

340

 

224

426

426

Houston

363,808

462

 

180

242

242

Hubbard

639,401

68

 

0

0

201

Isanti

288,961

79

 

54

143

143

Itasca

1,871,189

253

170

276

633

633

Jackson

425,831

53

 

28

20

20

Kanabec

341,014

149

 

112

139

139

Kandiyohi

551,512

152

 

106

94

94

Kittson

706,662

36

 

18

201

201

Koochiching

2,016,518

115

3

64

252

252

Lac Qui Parle

464,521

43

 

28

42

42

Lake

1,462,187

255

390

340

367

368

Lake of the Woods

1,072,369

57

 

0

54

54

Le Sueur

303,041

114

 

65

34

34

Lincoln

651,291

76

 

28

129

129

Lyon

461,908

149

 

89

79

79

McLeod

323,428

27

 

8

96

96

Mahnomen

373,191

20

 

12

86

86

Marshall

1,160,962

61

 

49

333

333

Martin

466,699

55

 

35

54

54

Meeker

412,638

41

 

22

68

68

Mille Lacs

435,921

122

 

70

82

82

Morrison

737,659

196

 

124

240

240

Mower

455,114

62

 

30

41

41

Murray

460,659

93

 

50

57

57

Nicollet

298,668

178

 

86

305

305

Nobles

462,362

71

 

50

166

166

Norman

544,564

149

 

47

78

78

Olmsted

418,545

65

 

19

157

157

Otter Tail

1,424,257

225

 

132

111

111

Pennington

395,891

7

 

5

81

81

Pine

917,282

407

 

62

318

318

Pipestone

298,576

44

 

30

55

55

Polk

1,279,543

40

 

31

258

258

Pope

455,250

56

 

28

46

46

Ramsey

108,790

51

 

13

39

39

Red Lake

276,932

14

 

7

107

107

Redwood

563,963

90

 

60

76

76

Renville

631,656

66

 

18

61

61

Rice

330,040

105

 

58

89

89

Rock

309,277

78

 

51

101

101

Roseau

1,074,233

82

 

39

30

30

St. Louis

4,306,973

644

205

539

1941

1941

Scott

235,686

104

 

49

93

93

Sherburne

288,409

212

 

40

80

80

Sibley

384,030

45

 

24

82

82

Stearns

889,142

180

 

87

256

256

Steele

276,348

20

 

6

193

193

Stevens

361,763

42

 

24

68

68

Swift

481,624

24

 

13

21

21

Todd

626,581

57

 

22

64

64

Traverse

371,897

87

 

73

211

211

Wabasha

351,537

97

 

71

146

146

Wadena

347,421

41

 

23

34

34

Waseca

276,776

71

 

42

253

253

Washington

149,595

136

 

70

386

386

Watonwan

281,419

16

 

7

10

10

Wilkin

467,396

44

 

31

50

50

Winona

410,219

160

 

54

29

29

Wright

456,881

159

 

102

268

268

Yellow Medicine

488,779

137

 

79

47

47

*Copied from Table 5.5 (Phase 1)

Includes all modeled SHPO sites, plus an additional 108 sites that were obtained in 1998 too late to be included into Phase 3 modeling database
Includes 8 sites outside the state border, three added to Cook Co. and five added to Lake Co. in the table only)
**This excludes single artifacts. 15 of these sites fell outside the state border. They were added to the figures in Lake (5), Cook (3), and St. Louis (2) counties. Errors were discovered in the Mn/Model database - sites from two counties (Hubbard with 47 sites and Lake of the Woods with 11 sites) were not used for modeling because they were erroneously coded as negative survey points (Table 5.5b).
† † Total of 15,848 surveys. Among those, 32 sites are valid sites, not surveys ( [Sitenum] <> "0000" ). The status of sites that were coded with SRC_CODE = 0 as  random points in the Phase 2 Mn/Model database has to be clarified.. There are 201 sites with SRC_CODE = 0 and SITE_TYPE = 0, among them 154 are negative survey points and 81 are sites.

 

 

Table 5.5b. Sites in the Mn/Model Database that are Erroneously Coded as Negative Surveys (SITE_TYPE = 0).

County

Count

Beltrami

2

Carver

11

Hubbard

47

Lake of the Woods

11

Yellow Medicine

1

 

 

 

 

 

 

 

 

 

 

Altogether, there are 72 sites in the Mn/Model database that are erroneously coded as negative surveys (SITE_TYPE = 0), but have valid SITENUM codes (Table 5.5b), values in the DESCRIPT and FUNCTION fields of the SHPO database to allow them to be classified, and a satisfactory location confidence. Of the 16051 negative surveys points in the Mn/Model database, nine are outside the state border: six near St. Louis, one near Lake and one near Cook county borders.

 

Survey reports were used to develop the database in Phase 1, and they are recorded in Table 5.5. BRW archaeologists referred to these reports to check information about sites recorded in the SHPO database and to map negative survey points (which are not part of the SHPO database) using maps of survey boundaries published in the survey reports. Since there may be multiple sites recorded in a single survey report (or no sites recorded if the survey was completely negative), there should not be a correspondence between site number, negative survey point numbers, and number of site reports. Since no further work was done with the survey reports after Phase 1, these columns were not included in the Phase 3 table (Table 5.5a).

 

Single artifacts (or find sites) were not used as sites for Phase 3, but were counted as surveyed places for developing the survey model. In Table 5.5a sites with single artifacts are still included in the Number of Known Sites from SHPO and Number of Known sites from USFS columns.

 

Minor discrepancies in site and negative survey counts summarized in Table 5.5a could occur when comparing county data to the state and ecological subsection data. When the state database was divided into regional and subsection databases for modeling, some sites and surveys near regional and state borders could either be dropped out of the database or be attached to a neighboring region database. Clipping sites by regions in vector or grid formats might give different results for these near-border sites.

 

Return to Top

 

5.6 KNOWN ARCHAEOLOGICAL SITES

5.6.1 SHPO Database

The SHPO database contains information on every known archaeological site in the state of Minnesota and includes, among other information, geographical locations, environmental settings, artifacts and features, degrees of disturbance, published references, and significance according to the criteria of eligibility for the National Register. SHPO data in the Phase 3 Mn/Model archaeological database were current as of January, 1997. Later, when the Phase 3 modeling was in progress, 108 new sites from SHPO were recorded in a separate database that was used only for map display and reference.

 

The SHPO database contained far more information than Mn/Model used; for example, researchers excluded the portions of the database concerning recent, historical archaeological sites. To extract the information desired for Mn/Model and to keep the process consistent, the data collection team followed the procedures below.

  1. All sites identified in the SHPO database’s PERIOD field as R-1 or R-2 (recent period) were deleted. If the PERIOD field indicated one or more components other than recent, the site was retained.
  2. All sites with a locational confidence rating of 4 or 5 in the SHPO database’s LOCCONF field were deleted. These sites could be located or relocated with only minimal accuracy. Sites rated 1, 2, or 3 (very high, high, or moderate), were retained.
  3. All sites with more than one record in the SHPO database were incorporated into a single record. Generally, all records after the first one contained no UTM numbers, so extraction was simple. The extra records in the SHPO database indicate that the amount of locational data recorded for the site was too large or too complex to keep in a single record’s fields.
  4. All sites known solely from historic documentation, identified by the presence of HD (historic description) in the SHPO database’s DESCRIPT field, were deleted.
  5. All transportation sites, such as trails, early military roads, and portages, for which the SHPO database contained a designator such as "Portage" in the FUNCTION field, were also deleted.

 

Information from the SHPO database that indicated the site’s significance according to the criteria for eligibility for the National Register was recorded and transcribed for the Mn/Model database. This information could not be imported directly into the Mn/Model database because of the different field types (text vs. integer). Sites meeting any of the three criteria below were assigned the number 3 in the Mn/Model database’s field identifying source code. This number took precedence over any other number except 5, which indicated a National Forest site.

  1. Sites in which the SHPO database’s WORKTYPE field held a 3 or a 9 had undergone a data recovery/mitigation/Phase III investigation or a combined Phase II/Phase III investigation,
  2. Sites in which the SHPO database’s NRHP field held a Y, indicating that the site is on the National Register of Historic Places,
  3. Sites in which the SHPO database’s CEF field held a Y, indicating that the site is considered eligible for but is not actually listed on the National Register.

 

Information from the SHPO database contained in the DESCRIPT and FUNCTION fields was also recorded and transcribed for the Mn/Model database. Again, because of the variance in field type definitions, the information could not be imported directly. This information could be used to distinguish between, for example, an isolated find and a large, complex assemblage indicating a base camp.

 

See Section 5.8 for further information about assembly and quality control of the database.

 

5.6.2 Chippewa National Forest Database

Information from the Chippewa National Forest was used for portions of Beltrami, Cass, and Itasca counties in accordance with the Memorandum of Agreement signed May 17, 1996. The Agreement was verbally amended on September 24, 1996, when National Forest Archaeologist gave verbal permission to incorporate the SHPO version of the Chippewa National Forest Database. The data were used only for modeling, and will not be distributed to the public or to other agencies with any version of the working or final product without the expressed written consent of the Chippewa National Forest Supervisor. At the request of the National Forest Archaeologist, these data were destroyed at the end of the project.

 

Several problems were encountered with the National Forest’s database. Although Forest Service site numbers have been entered into the Chippewa National Forest database for all sites within the Forest’s borders, no further information exists for the majority of the sites. Site data, particularly the UTM coordinates, are only relatively complete for the last two to three years (an observation supported by Andrea LeVasseur, National Forest Archaeologist, personal communication, 8-5-96). In addition, the Chippewa database appears to contain multiple records for multi-component sites, generally one per component, without necessarily indicating that this is the case.

 

Paper copies of all the Chippewa’s archaeological site files had recently been delivered to the SHPO, and The Survey and Information Management Coordinator for the SHPO had coordinated the assignment of Smithsonian site numbers where appropriate and had determined UTM coordinates for many of the sites which had lacked them. With the National Forest Archaeologist’s permission, the SHPO/Chippewa database was used instead of the original version. This version of the National Forest’s database contained more information, but also had weaknesses. For example, there were many sites which had neither Smithsonian site numbers nor UTM coordinates, and whose names indicated nothing about the site’s period. Although these sites might have been precontact, they could not be included in Mn/Model without more information.

 

The same basic procedures were followed for incorporating the National Forest’s site data as for including the SHPO database. To summarize, recent sites and those with insufficient locational information were deleted. The data were divided along the county borders, incorporated into Mn/Model’s county databases, and identified by entering "5" as the source code. This code took precedence over every other source code.

 

The Chippewa National Forest is in the process of completing a GIS layer of areas within the National Forest boundaries that have been surveyed by Forest Service personnel and outside contractors. As this layer was not completed in time for modeling, it was not incorporated into Mn/Model.

 

5.6.3 Superior National Forest Database

Information from the Superior National Forest for Cook, Lake,and St. Louis counties was used in accordance with the Memorandum of Agreement signed May 17, 1996. The agreement reached by the research team’s database coordinator and the Superior National Forest Archaeologist, allowed the research team to use data compiled by SHPO (Gordon Peters, Superior National Forest Archaeologist, personal communication, 8-5-96). The data were used only for modeling and will not be distributed to the public or to other agencies with any version of the working or final product without the express written consent of the Superior National Forest Supervisor. At the request of the National Forest Service Archaeologist, these data were destroyed at the end of the project.

The information encoded in the Superior National Forest’s original database was inconsistent in terms of both content and organization. Information tended to be entered solely as text, rarely as numerical or letter codes, which made establishing some basic information, such as site type, difficult. The type of information contained in a given field, particularly descriptive ones, was highly variable, and may not even appear in some records. The database format and software were incompatible with the SHPO’s system and required a considerable conversion effort to become useable for Mn/Model.

 

When the original Superior National Forest database was compared with the version compiled by the Survey and Information Management Coordinator for the SHPO from their site files, it became apparent that not all UTM coordinates corresponded. In some cases, there were significant differences that were resolved only by re-examining the copies of the site files and USGS quad maps.

 

Return to Top

 

5.7 SURVEY LITERATURE AND NEGATIVE SURVEY POINTS

5.7.1 Probabilistic Surveys, Including Mn/Model and the Minnesota Statewide Archaeological Survey

A probabilistic survey is one where the parcels chosen for investigation were selected using methods attempting to ensure randomness. The counties presumed to contain probabilistic surveys, particularly those included in MnSAS, were examined during the earliest stage of the Mn/Model archaeological data collection process. Only four ostensibly probabilistic surveys were identified by Mn/Model researchers. The first (Dobbs 1989) could not be used because the areas surveyed could not be established with even the minimum level of accuracy . The second survey (Trow 1984) examined randomly chosen parcels located on the terraces of the Root River and its tributaries. The third and only large-scale survey, MnSAS, was conducted by a variety of archaeology professionals and their crews between 1977 and 1981, and encompassed portions of 26 counties (Minnesota Historical Society 1981). These surveyed areas were mapped in yellow on Mn/Model USGS maps (on file at MnDOT). This survey is discussed in detail in Sections 5.3, 5.4.1, and 5.4.2. MnSAS field crews chose parcels not only based on their various stratification schemes but also by intuition. Mn/Model researchers were not always able to distinguish, even after consulting the actual field maps, the areas that were surveyed based on probabilistic stratification and those which were intuitive. In addition, the survey transect intervals were generally 50 m. At the time of the MnSAS report, standard intervals were 15 m or less, and the maximum acceptable interval for survey reports under consideration for inclusion in the database was 30 m. Therefore, there may actually be sites in the surveyed areas which were not located by the MnSAS field crews; some of the negative survey points chosen based on those surveyed areas may actually be sites. Information from MnSAS was included only where Mn/Model researchers were able to establish, based on a combination of published information, field maps, original field notes and preliminary summaries, that the surveyed parcels were chosen in a probabilistic manner. The final probabilistic survey was Mn/Model itself, in the 1995 and 1996 field seasons. The 1995 were technically probabilistic only within the surveyed areas, which were mapped in green on Mn/Model USGS maps (on file at MnDOT). The 1996 Mn/Model field survey is the only truly probabilistic survey available for inclusion in the Mn/Model database, even though the other surveyes mentioned above will be called "probabilistic" in this report.

 

5.7.2 Negative Survey Points From Probabilistic Surveys Other than Mn/Model

The following procedures were developed in December 1995 and January 1996 for taking random, non-site or negative survey points from areas reported as surveyed in probabilistic surveys other than Mn/Model. A single negative point was chosen for each parcel, typically a 40-acre parcel; if a point could not be chosen by the method described no point was recorded for that parcel.

 

               1) The Smithsonian or trinomial site number for any archaeological site within the borders of a probabilistic survey parcel was recorded. This information was checked against the SHPO database to confirm that the survey had identified the site, sites discovered during non-probabilistic surveys were assigned a "1" in the Mn/Model database’s source code field.

 

               2) Wherever the center point of a surveyed parcel did not lie in or adjacent to a site (within about 2 mm on the USGS map) and did not appear to fall within a wetland or in water, it was used as the random point for that parcel as long as, to the best of the Mn/Model data collectors’ knowledge, that area was actually surveyed by the field crew.

 

For steps 3, 4, and 5: there are occasional dry-land areas within wetlands that are not detailed on the USGS maps. Determining whether a particular area with a point was "dry" required comparing the location of the point on the USGS map with the equivalent on the soil survey maps and comparing that information in turn with the statewide list of hydric (wet) soils. Those soils on the statewide list that fall into “hydric criteria numbers” 2A or 2B3 on the table are intermittently wet and were considered "dry" for Mn/Model’s purposes. In the opinion of Mn/Model’s Phase 1 and 2 Research Director, Dr. Guy Gibbon, 2A or 2B3 soils could have been surveyed, while permanently wet soils were, in all likelihood, not surveyed.

 

               3) If the surveyed parcel did not contain a site, and the center point appeared to fall in or near a wetland area on the USGS quad map, but the soil at the center point fell into hydric criteria number 2A or 2B3 on the statewide hydric soils tables, it remained a viable negative point.

 

               4) If conditions were the same as in 3), above, but the soil at the center point fell in hydric criteria number 1, 2B1, 2B2, 3, or 4 in the statewide hydric soils tables, the negative survey point was moved to the nearest "dry" land within the parcel.

 

               5) If the surveyed parcel contained a site and the center point appeared to fall in or near a wetland area on the USGS topo map, or if the center point fell on or is adjacent to (within about 2 mm on the USGS map) the site, the following approach to establish a negative point and its UTM coordinates was taken. The surveyed parcel was divided into quarters. Any quarter containing a site or a portion of a site, or which contained no "dry" land, was eliminated from consideration. Using a coin toss or other equivalent method, one of the squares was selected and its center point or a point on the closest available "dry" land was recorded as the negative point.

 

5.7.3 Negative Survey Points From Mn/Model Surveys

Mn/Model fieldwork was conducted during the 1995 and 1996 field seasons. The accuracy of information regarding the areas surveyed allowed the development of the following procedures.

 

               1) For parcels surveyed in 1995, the UTM coordinates of each parcel’s center point were recorded on the original parcel form. If, upon mapping and review by the Mn/Model data collectors, the center point had not actually been surveyed, the UTM coordinates in the database were shifted to an area that had been surveyed. Shifting the point proved necessary in most parcels where shovel testing, rather than pedestrian survey, had been the method of investigation.

 

               2) For parcels surveyed in 1996, the UTM coordinate of a randomly generated point from the GIS had determined the placement of parcels. Whenever the point fell inside the area surveyed, it was left unchanged. If the random point was outside the area surveyed, the UTM coordinates of that point were shifted to the closest point within the surveyed area on the USGS quad maps.

 

5.7.4 Non-Probabilistic Surveys

A tremendous amount of literature regarding archaeological undertakings exists for the state of Minnesota. Because the Mn/Model data collectors were concerned with the quality of the data, a table of criteria for survey inclusion was developed in April, 1996, in consultation with National Register Archaeologist Scott Anfinson of the SHPO and with Mn/Model Phase 1 and 2 Research Director Guy Gibbon, and was strictly adhered to for all quality control reviews in the final stages of the project (see Table 5.6). To have a point or points included in the Mn/Model archaeological database, a survey had to meet at least the first four criteria.

 

Most Phase II and Phase III (intensive-level and data recovery/mitigation) reports were excluded from Mn/Model by their site-specific nature; the survey areas generally did not extend far enough beyond the boundaries of the site.

 

Table 5.6. Mn/Model Survey Report Inclusion Criteria

CRITERIA

REASONING

1. Sufficient information about location (either section, township and range, plus any further information to narrow the possibilities, or a USGS-based map with the area actually surveyed marked) to relocate the area actually investigated on a 7.5' USGS map

If the surveyed area cannot be relocated on a USGS map, UTM data cannot be taken

2. Rigorous field methods (for example, comparable to or better than 30m intervals between transects and/or between shovel tests)

If the survey methods were not rigorous, the probability that sites were not recorded rises; taking UTM points from these potentially ‘false negative’ areas could skew the GIS-based predictive model

3. Pedestrian survey or shovel-tested areas larger than 2 acres, or shovel-test survey transects covering more than 60 m (246 feet) (equivalent to 4 shovel tests at 15m intervals)

Areas smaller than these are likely to be difficult to identify or relocate accurately on a USGS map

4. Undisturbed land (for example, a completely new road alignment, or a widening of a road beyond the earlier right-of-way)

Substantial disturbance by modern road, railroad, housing, or industrial or commercial construction increases the likelihood that any sites in the area were destroyed or that artifacts found were transported from their original locations

5. Trunk Highway or Municipal-County Highway annual reports for 1983 and later years

Before 1983, the two programs did not have the funding or time for adequate reporting of field surveys or for surveying anything beyond the highest-probability areas (Scott Anfinson, personal communication, December 1995 and February 1996)

 

 

The information gathered from each survey report included the rigorousness of the field methods and whether or not the survey located any archaeological sites (for the Mn/Model archaeological database’s source code field). This information was recorded on the bibliography of reports for each county, provided by SHPO. Reasons for excluding surveys were also recorded on the bibliography. Whenever possible, the areas actually surveyed, from acceptable reports, were recorded in color-coded pencil (orange) directly on the Mn/Model USGS quad maps (on file at MnDOT) for the county, then labeled with the SHPO survey report identifier.

 

5.7.5 Negative Survey Points from Non-Probabilistic Surveys

To maintain consistency between data collectors, the following procedures were developed for taking the UTM coordinates of negative survey points in non-probabilistic survey areas.

  1. A Land Area and Slope Indicator Template for use on USGS 7.5 and 15-minute series maps was placed over the surveyed area in one of two ways, either aligned with the section lines for relatively large survey areas, or along the road or pipeline for linear survey areas.
  2. For each 40-acre parcel the surveyed area passed through or was in, one UTM was taken from any point in the parcel within the area surveyed.
  3. Points were not taken within wetlands, unless the survey specifically indicated that those areas had been investigated, and were not taken immediately adjacent to a site’s boundaries.
  4. If all possible points within a parcel were in wetlands, outside the survey boundaries, or in or adjacent to sites, the parcel was eliminated.
  5. The UTM coordinates of each point were recorded together with the SHPO survey identification code and the USGS quad name.
  6. During the taking of negative survey points, archaeological sites that appeared to have been discovered during that particular survey were also recorded. This information was checked against the SHPO database to confirm that the survey had identified the site; sites discovered during non-probabilistic surveys were assigned a "2" in the Mn/Model database’s source code field.
  7. When all survey areas for a county were completed, the data, including UTM coordinates, SHPO survey report identifiers, Smithsonian numbers for archaeological sites, and USGS quad names (not for all counties), were entered into the computer.

 

Return to Top

 

5.8 ASSEMBLY AND QUALITY CONTROL OF THE ARCHAEOLOGICAL DATABASE

To summarize the quality control procedures defined in the last two sections, precontact and contact period sites that met certain criteria were extracted from the population of known sites. Negative survey points from a variety of surveys that met certain strict criteria were collected and entered into a database.

 

5.8.1 Assembling and Reviewing the Database

A blank database was created, named according to the conventions described in Section 5.9 and containing the fields mentioned in Section 5.5. The known archaeological site data for a given county were imported into the county’s database, then the information in the source code and site type fields was entered. Negative survey points were either entered directly into the county database or, more often, imported from a separate database after quality control reviews of the survey data were completed. These focused on the source materials for negative survey points, in an attempt to ensure that the survey reports used met the criteria in Section 5.7. Points and, if necessary, archaeological site source codes, were deleted, added, or changed as necessary. Each database was checked to see that all fields were filled appropriately, that no outlier points existed, and that no archaeological site numbers and no sets of UTM coordinates were duplicated. If duplicate negative point UTM coordinates were found, one record deleted. If two sites with different identifiers had the same UTM coordinates, the reviewer checked the original SHPO database and the USGS quad maps and changed any coordinates necessary. These changes were reported to the SHPO.

 

The county’s database, having been thoroughly reviewed, was submitted to the GIS group according to the formatting standards described in Section 5.9.

 

5.8.2 Submitting the Database to GIS

After the initial examination of the data by the GIS group, some county databases needed further corrections. For example, several counties had archaeological site UTM coordinates that were outside the borders of the county as drawn in the GIS. These sites were checked against the USGS quad maps, and the UTM coordinates changed if necessary. If the coordinates in the original SHPO database were incorrect, a correction was reported to the SHPO. If the coordinates were merely shifted slightly within the borders of the site as drawn by the SHPO and copied by BRW, the change was not reported.

 

Other counties’ data showed negative survey points outside the borders of the county. These were checked against the original negative point database and the USGS quad, and shifted or deleted where necessary and/or possible. Occasionally, checked points were located on small islands or peninsulas which did not appear in the borders of the county as drawn in the GIS. Reviewers made a note of these points and left them untouched. After completion of all double-checking, the county database was re-submitted to the GIS group with all corrections incorporated.

 

Return to Top

 

5.9 MN/MODEL ARCHAEOLOGICAL DATABASE SPECIFICATIONS

5.9.1 Naming Conventions

All points from a single county, both archaeological sites and negative survey points, were incorporated into a single database, identified by the letters AR (for archaeology) and the FIPS code for the county. Where the county covered portions of two zones, zones 14 and 15 or 15 and 16, two databases, one for each zone, were compiled. Databases from zones other than 15, including those for counties entirely within zone 14, were identified by the FIPS code for the county followed by the zone. For example, Anoka County’s database was AR003, Rock County’s was AR13314, Cook County’s were AR031 and AR03116, and Marshall County’s were AR089 and AR08914.

 

5.9.2 Software and Formatting

The Mn/Model archaeological database was assembled in Approach, a Lotus-based data management program. The source data were supplied in various formats; the SHPO database was in dBase format, the Chippewa National Forest database was originally constructed in Paradox, and the Superior National Forest was compiled in MINARK and had to be translated into ASCII files to convert any of the data. The final versions of the Mn/Model databases were submitted on disks to the GIS group in dBase IV format.

 

Return to Top

 

REFERENCES

Altschul, J.H., and C.L. Nagle.
   1988. Collecting New Data for the Purpose of Model Development. In Quantifying the Present
        and Predicting the Past: Theory, Method, and Application of Archaeological Predictive
       Modeling
, edited by W. J. Judge and L. Sebastian, pp. 257-299. U. S. Department of the
       Interior, Bureau of Land Management, Denver. U.S. Government Printing Office, Washington,
       D.C.

 

Ammerman, A.J.
   1981 Surveys and Archaeological Research. Annual Review of Anthropology 10:63-88.

 

Anfinson, S.F.
   1984 Cultural and Natural Aspects of Mound Distribution in Minnesota. The Minnesota
       Archaeologist
43(1):3-30.

 

Binford, L. R.
   1964 Consideration of Archaeological Research Design. American Antiquity 29:425-441.

 

Birk, D.A.
   1979 Sampling and Intuition: An Archaeological Survey of the Nokasippi River Valley, Crow
       Wing County, Minnesota
. Report prepared by the Archaeology Department, Minnesota
       Historical Society, St. Paul, Minnesota.

 

Blalock, H.M., Jr.
   1979 Social Statistics (revised second edition). McGraw-Hill Book Company, New York.

 

Carmichael, D.L.
   1990 GIS Predicting Modelling of Prehistoric Site Distributions in Central Montana. In Interpreting
       Space: GIS and Archaeology,
edited by K.M.S. Allen, S.W. Green and E.B.W. Zubrow, pp.
       216-225. Taylor and Francis, London.

 

Dobbs, C. A.
   1989 A Phase One Archaeological Survey of the Cottonwood, Redwood, and Yellow Medicine
       Drainages in Southwestern Minnesota
. Sponsored by the U.S. Army Corps of Engineers in
       cooperation with the U.S. Department of Agriculture Soil Conservation Service. Contract
       Number DACW 37-85-M-1113. Institute of Minnesota Archaeology, Minneapolis.

 

Dunnell, R.C., and W.S. Dancy
   1983 The Siteless Survey: A Regional Scale Data Collection Strategy. In Advances in
       Archaeological Method and Theory,
vol. 6, edited by M.B. Schiffer, pp.:267-287. Academic
       Press, New York.

 

Gibbon, G. E.
   1981 One Hundred Years of Upper Midwest Archaeology. Upper Midwest History 1:41-53.

 

Henning, E. R. P.
   1983 Methodological Approach and Introduction. In Archaeological Reconnaissance Survey of
       the Northern Border Pipeline for the Northern Plains Natural Gas Company - Iowa
       Segment
, Vol. 1, compiled by G. J. Hudak, pp. 4.1-4.15. Report submitted to the Northern
       Plains Natural Gas Company by Archaeological Field Services, Inc., Stillwater, Minnesota.

 

Kintigh, K. W.
   1988 The Effectiveness of Subsurface Testing: A Simulation Approach. American Antiquity
       53:686-707.

 

Kvamme, K. L.
   1988a Development and Testing of Quantitative Models. In Quantifying the Present and
       Predicting the Past: Theory, Method, and Application of Archaeological Predictive
       Modeling
, edited by W. J. Judge and L. Sebastian, pp. 325-348. U. S. Department of the
       Interior, Bureau of Land Management, Denver. U.S. Government Printing Office, Washington,
       D.C.
   1988b Using Existing Archaeological Survey Data for Model Building. In Quantifying the Present
       and Predicting the Past: Theory, Method, and Application of Archaeological Predictive
       Modeling
, edited by W. J. Judge and L. Sebastian, pp. 301-323. U. S. Department of the
       Interior, Bureau of Land Management, Denver. U.S. Government Printing Office, Washington,
       D.C.
   1990 The Fundamental Principles and Practice of Predictive Archaeological Modeling. In
       Mathematics and Information Science in Archaeology: A flexible Framework, edited by A.
       Voorrips, pp. 257-295. Studies in Modern Archaeology, Vol. 3. Holos-Verlag, Bonn, Germany.
   1992 A Predictive Site Location Model on the High Plains: An Example with an Independent Test.
       Plains Anthropologist 37:19-40.

 

Leusen, P. M. van.
   1996 GIS and Locational Modeling in Dutch Archaeology: A Review of Current Approaches. In
       New Methods, Old Problems: Geographic Information Systems in Modern Archaeological
       Research
, edited by H. Maschner, pp. 177-197. Occasional Paper No. 23, Center for
       Archaeological Investigations, Southern Illinois University, Carbondale.

 

Lewarch, D. E., and M. J. O’Brien.
   1981 The Expanding Role of Surface Assemblages in Archaeological Research. Advances in
       Archaeological Method and Theory,
vol. 4, edited by M.B. Schiffer, pp. 297-342. Academic
       Press, New York.

 

Lothson, G. A.
   1967 The Distribution of Burial Mounds in Minnesota. The Minnesota Archaeologist 29(2):29-
       47.

 

Minnesota Historical Society
   1981 Minnesota Statewide Archaeological Survey. Summary: 1977-1980. Minnesota Historical
       Society, St Paul.

 

Otinger, J.L., R.H. Lafferty, and R.D. Jones
   1981 Field Tactics and Laboratory Methods. In Settlement Prediction in Sparta: A Locational
       Analysis and Cultural Resource Assessment in the Uplands of Calhoun County, Arkansas,
      
edited by W.F. Limp, pp. 101-116. Arkansas Archaeological Survey Research Series 14,
       Fayetteville.

 

Parker, S.C.
   1985 Predictive Modeling of Site Settlement Systems Using Multivariate Logistics. In F r
       Concordance in Archaeological Analysis: Bridging Data Structure, Quantitative
       Technique, and Theory
, edited by C. Carr, pp. 173-207. Westport Press, Kansas City.

 

Redman, C. L.
   1987 Surface Collection, Sampling, and Research Design: A Retrospective. American Antiquity
       52:249-265.

 

Schiffer, M. B., A. P. Sullivan, and T. C. Klinger.
   1978 The Design of Archaeological Surveys. World Archaeology 10:1-29.

 

Schiffer, M. B., and S. J. Wells
   1982 Archaeological Surveys: Past and Future. In Hohokam and Patayan Prehistory of
       Southwestern Arizona
, edited by R. H. McGuire and M. B. Schiffer, pp. 345-384. Academic
       Press, New York.

 

Shott, M.J.
   1985 Shovel-test Sampling as a Site Discovery Technique: A Case Study from Michigan. Journal
       of Field Archaeology
12:457-468.

 

Tomlin, C. C.
   1990 Geographic Information Systems and Cartographic Modeling. Prentice-Hall, New York.

 

Trow, T.
   1984 Archaeological Investigations in Southeastern Minnesota: The 1984 Root River Survey.
       Vol. I and II. Minnesota Historical Society, St. Paul.

 

Warren, R.E.
   1990 Predictive Modelling in Archaeology: A Primer. In Interpreting Space: GIS and
       Archaeology
, edited by K.M.S. Allen, S. W. Green, and E.B.W. Zubrow, pp. 90-111. Taylor
       and Francis, London.

 

Warren, R. E., and D. L. Asch
   1996 A Predictive Model of Archaeological Site Location in the Eastern Prairie Peninsula, Illinois.
       Manuscript on file, Illinois State Museum, Springfield.

 

Young, P.M., M.R. Horne, C.D. Varley, P.J. Racher, and A.J. Clish
   1995 A Biophysical Model for Prehistoric Archaeological Sites in Southern Ontario. Research
       and Development Branch, Ministry of Transportation, Ontario.

 

Wandsnider, L. and E.L. Camilli
   1992 The Character of Surface Archaeological Deposits and its Influence on Survey Accuracy.
       Journal of Field Archaeology 19:169-188.

 

Wright, H. E., Jr.
   1972 Physiography of Minnesota. In Geology of Minnesota: A Centennial Volume, edited by P.
       K. Sims and G. B. Morey, pp. 561-580. Minnesota Geological Survey, St. Paul.

 

Return to Top

 

 

 

The Mn/Model Final Report (Phases 1-3) is available on CD-ROM. Copies may be requested by e-mail: mnmodel@state.mn.us

 

MnModel Orange Bar Logo

Acknowledgements

Mn/Model was financed by the Minnesota Department of Transportation using funds set aside by the Federal Highway Administration's Intermodal Surface Transportation Efficiency Act.

 

Copyright Notice

The Mn/Model process and the predictive models it produced are copyrighted by the Minnesota Department of Transportation (MnDOT), 2000. They may not be used without MnDOT's Consent.