- Management Summary
- Research Design & Time Line
- Environment & Native American Culture
- GIS Design
- Archaeological Database
- Archaeological & Environmental Variables
- Model Development & Evaluation
- Model Results & Interpretation
- Project Applications
- Model Enhancements
- Model Implementation
- Landscape Suitability Models
- Summary & Recommendations
- Archaeological Predictive Modeling: An Overview
- GIS Standards & Procedures
- Archaeology Field Survey Standards, Procedures & Rationale
- Archaeology Field Survey Results
- Geomorphology Survey Profiles, Sections, & Lists
- Building a Macrophysical Climate Model for the State of Minnesota
- Correspondence of Support for Mn/Model
- List of Figures
- List of Tables
by Craig M. Johnson, Connie Arzigian, and James Gallagher
Appendix C Table of Contents
C.2 Site Definition
C.3 Partitioning the Survey Universe
C.3.1 1995 Mn/Model Field Survey
C.3.2 1996 Mn/Model Field Survey
C.4 Survey Rationale
C.5 Landowner Contacts
C.6 Field Methods
C.6.1 Pedestrian Survey
C.6.2 Shovel Testing
C.7 Forms and Documentation
C.7.1 Parcel Form
C.7.2 Cultural Resources Identified Form
C.7.3 Shovel Test Form
C.7.4 Photographic Record
C.7.5 Bag Inventory and Transfer Record
C.7.6 Supervisor's Daily Journal
C.8 Laboratory Procedures
C.9 Quality Assurance and Control
The archaeology field survey program for the Mn/Model project was conducted in two field seasons, 1995 and 1996, by teams from BRW, Inc., Leech Lake Reservation Heritage Sites Program, and Mississippi Valley Archaeology Center (MVAC). Since the goals of each season were different, the sampling design and field methods also varied. There was, however, overlap in certain field and laboratory procedures from one year to the next. These surveys maintained standards that met or exceeded the minimum guidelines for archaeological survey in Minnesota (Minnesota State Historic Preservation Office 1993). Given the uniqueness of the Mn/Model project, a number of additional procedures were also implemented. These are described below. A description of the sampling design, survey rationale, and field methods, in addition to the similarities and differences between the two field seasons, is the focus of the following discussion.
There has been extensive discussion about the management and research implications of different definitions of a site (Zeidler 1994). Zeidler recommends defining sites as only those areas with at least two different artifacts or different classes of items, from which an inference of a purposeful activity can be drawn. In an alternative "non-site" approach, "the archaeological record is most usefully conceived as a more or less continuous distribution of artifacts over the land surface with highly variable density characteristics. Sites in this context represent only a part of the total record, explicitly defined by density characteristics" (Dunnell and Dancey 1983, emphasis original) (quoted in Zeidler 1994). This approach, however, assumes that reliable measurements of artifact density are available at a large number of points across the landscape. Although this information would provide for greater flexibility in predictive modeling, the data is not presently available for Minnesota.
A broadly inclusive definition of a site was adopted by Mn/Model. Like the 1977-81 Minnesota Statewide Archaeological Survey (MnSAS) (Minnesota Historical Society 1981), Mn/Model surveyors recorded both isolated and multiple finds as archaeological sites. Any cultural material at or reasonably close to its original point of deposition was documented. Multiple artifacts less than 100 meters apart, or on the same landscape feature were designated as sites, while single artifacts were designated isolated finds. Both, however, were treated comparably in that uniform field records were kept, MHS site forms were completed, and Smithsonian site numbers were assigned to them.
Site boundaries were defined by the distribution of surface material or positive shovel tests. Wherever appropriate, site boundaries were drawn to correspond with natural features such as knolls or terrace edges. Clusters of material on separate topographic features were generally identified as separate sites, while a continuous scatter of cultural material on one topographic landscape surface was considered one site. Specific site boundaries were created around clusters of material that had less than 100 meters between individual items. Site boundaries were drawn in the field on topographic maps and aerial photographs, and sometimes on sketch maps. Occasionally the distributions of individual artifacts or clusters of artifacts were plotted on sketch maps in 1995. In 1996, the location of each individual cultural item was recorded.
Survey strategy to distinguish isolated finds from multiple artifact sites differed from 1995 to 1996 to reduce the chances that the 1995 survey procedures had inflated the number of isolated finds.. During the pedestrian survey in 1995, archaeologists stayed in transects spaced 15 meters apart and did not go outside them once a single artifact was found. The procedure standardized search techniques among the field crews and ensured that more rather than fewer parcels could be surveyed by reducing parcel search time. In 1996, a modified procedure was introduced. When a single artifact was found in 1996, it was bracketed with survey at 5 and 10 meter intervals around the isolated find to determine if more materials were present.
To consider whether bracketing during 1995 would have identified additional cultural material associated with the single artifacts, the results of a sample of the 1996 survey were examined. In Wabasha County, which had the highest identified site density of the three counties surveyed in 1996, 17 of 19 originally identified isolated finds remained isolated finds despite bracketing. Only two isolated finds (10.5 percent) were converted to sites (two artifacts each) because of bracketing. This result strengthens the argument that many isolated finds are really not associated with other cultural materials and their isolation is not a function of the survey procedures.
Two distinctively different survey designs were employed in 1995 and 1996 with different goals. Both were constructed as random sampling strategies designed to select a relatively small number of 40 acre sample units from a larger survey universe or population of parcels. The 1995 survey was based on a number of landforms thought to have some relationship to the location of precontact archaeological sites while the 1996 survey did not make any assumptions prior to field work. Although the rationale of these two different approaches is outlined later, it can be noted that the 1995 survey design is much more complex than the 1996 one. As a result, the description of the 1995 field survey is relatively long compared to 1996. Despite these differences, the sampling universe consisted of counties or transects within counties. The details of both seasons field work are described below.
The 1995 field survey focused on five counties or parts of counties. The sampling universe in Nicollet County consisted of all lands within its borders. Due to the large size of the remaining three counties (Stearns, Becker, and Beltrami) and the prohibitive amounts of time needed to stratify and code all lands within them, a series of transects were chosen to sample their full range of landscape formations. The following discussion describes how these were divided into survey units and the stratification procedures used to sample the full range of landform variability within them.
C.3.1.1 Survey Units
The basic unit of the 1995 Mn/Model field survey was a 40-acre parcel. Parcels were based on the legal divisions of a section of land (640 acres), each parcel being a quarter of a quarter section. The size and shape of this basic unit of land was chosen to be compatible with MnSAS procedures (Minnesota Historical Society 1981). It was important that the survey units be similar, since an objective of the 1995 survey was to provide for a comparison with the MnSAS results. Another goal was to provide data on site locations and nonsite locations based on sampling for entry in the GIS database.
Each sampling universe, whether an entire county or a transect within a county, was divided into 40-acre quarter-quarter sections based on 7.5 minute USGS topographic quadrangle maps. Sections of irregular shape or size were divided into quarter-quarter sections with the starting point placed at the southeastern corner of the section. A legal land locator template was placed at this corner with its right hand edge aligned along the eastern margin of the section. Parcels that varied from the ideal 40 acres were noted if they were surveyed. Sections with at least an additional ¼ mile of land were divided into additional 40-acre parcels. Any amounts of land less than ¼ mile on a side were combined with the adjacent 40-acre parcel.
To avoid confusion, a unique number or "address" was recorded for each surveyed parcel that indicated township, range, section, and the parcel’s position within the section. Each quarter section, or 160 acres, was given a number from 1-4, starting in the northeast quarter and moving counter-clockwise. Within each quarter section, the four 40-acre parcels were also numbered from 1-4 in a similar fashion. These two numbers were combined into a single two-digit position identifier, with the first number representing the 160-acre parcel location and the second one the 40-acre parcel’s place within the 160 acres. The square below shows the sixteen possible combinations:
22 21 12 11
23 24 13 14
32 31 42 41
33 34 43 44
For example, the address of a parcel in Stearns County in Township 125N, Range 32W, Section 31, NW 1/4, NW1/4 would be 125-32-31-22.
C.3.1.2 Landform Strata Assignments
After all 40-acre parcels within the Mn/Model sampling universe were defined, they were stratified according to a bifurcated or hierarchical system that divided the sampling universe into 19 strata or landforms based primarily on surface hydrology. The purpose of this system of stratification was to examine the relationship between archaeological sites and landscape features, particularly water. Like the choice of 40-acre parcels for the standard survey area, the Mn/Model stratification system was designed to be compatible with that of the MnSAS and to facilitate the use of MnSAS data for Mn/Model. The MnSAS used a simple stratification hierarchy to subdivide the topography of Minnesota. In this system, survey parcels were initially separated into those associated with one or more bodies of water greater than 40 acres in extent and those that were not. Since different MnSAS surveys used various elaborations of this basic system, Mn/Model developed a more elaborate hierarchy that could encompass all of these versions. In Mn/Model, a series of 19 strata were defined for the state and arranged in a hierarchy. Within the "near water" heading, different aspects of riverine and lacustrine topography were defined. Within the category "away from water/water sources <40 acres" a series of small wetlands, intermittent streams, and uplands with no water were defined. The Stratification Outline (Table C.1) summarizes this stratification hierarchy; in the outline, each stratum appears in bold font. Stratum 19 (small lakes) was not defined when Nicollet and Stearns counties were surveyed; instead, parcels with the required characteristics were included in stratum 13 (wetland/marsh). Figure C.1 is a pictorial representation of the stratification system.
The 40-acre parcels for the Mn/Model survey were examined to determine which stratum or strata best represented the landforms present in that parcel. In Stearns and Nicollet counties, landform features were defined based on 7.5 minute topographic quadrangles, in conjunction with early General Land Office (GLO) hydrology records, as recorded on the Trygg maps. This was done because of historic modifications to the hydrology of the counties by draining and damming. Where multiple landforms were present in a parcel, a standard set of precedence rules was established for each county (see Figure C.1); in general, rarer landforms had priority over more common ones to insure they were represented in the sample strategy.
|I. Near Water > 40 acres|
|A. Permanent Stream (solid line on topographic map)|
|1. Confluence (of two permanent streams)||
|2. Stream Trunk|
|a. Small Streams (generally third order or smaller, generally without terraces, etc.)|
|i. Alluvial Fan||
|ii. Other Small Streams (permanent)||
|b. Large Streams (First or second order, generally with terraces, etc.)|
|iii. Upper Bluff Margin||
|B. Lake/Wetlands > 40 acres (on presettlement maps)|
|2. Other Lakeshore|
|c. Straight Shoreline/Other Lakeshore|
|i. Beach Ridges||
|ii. Other Shoreline||
|3. Wetland/Marshes (greater than 40 acres on presettlement map)||
|4. Surveyable Lake Bottom (drained lake)||
|II. Away from Water or Water Less Than 40 Acres|
|A. Small Water Sources|
|1. Slough/Small Marshy Areas (presettlement maps)||
|2. Intermittent Stream/Spring (presettlement maps)||
|3. Small Lakes||
|B. No Water|
|1. Dissected Topography: Overlooks/Heads of Interfluves, etc.||
|2. Generally Level Topography, General Upland Setting||
|III. Not Surveyable (Urbanized, developed, all water, hillsides greater than 15% slope, survey if cultivated)|
Definitions of terms for landforms used in the stratification system are as follows:
- Islands were relatively dry land either surrounded by water or apparently surrounded by water prior to modern land disturbances. The difference in elevation required to distinguish true islands from slightly drier wetlands in lakes differed for each county. In Stearns and Nicollet counties, this was at least 10 feet (one contour line).
- Points or peninsulas were long and relatively narrow, and set off from the mainland at a point where it narrowed to ¼ mile or less. If the landform widened beyond this as it extended out into the body of water, it was still considered a point or peninsula from that point forward.
- Lakeshore, peninsula, island, and cove/bay parcels generally contained standing water. In Nicollet and Stearns counties, where lakes were not abundant, a buffer zone around a large lake or wetland (>40 acres) was also coded as being part of the lake or wetland. Typically, this zone required that some point in the parcel be within 500 feet of the edge of the lake or wetland. In counties with abundant lakes and wetlands, such as Becker and Beltrami, the 500-foot buffer zone was eliminated. For small lakes and wetlands (less than 40 acres), no buffer was defined; points in the parcel had to lie on the margin of the lake or wetland for the parcel to be coded in this stratum. If an area contained a total of more than 40 acres in small lakes or wetlands, the combined lakes and/or wetlands were considered for stratification purposes to be one large body of water.
- Wetlands are areas where there were wetland symbols on topographic or Trygg maps, and where contour lines indicated depressions in the landscape or where soil maps indicated a hydric [wetland] soil.
- River and stream parcels had to include a portion of either a river or a stream. Inlet/outlet parcels contained the junction of a permanent stream and a lake. Confluence referred to the junction of two permanent streams in the parcel, but did not include intermittent streams. In cases where a confluence or inlet/outlet had no habitable land in a parcel, as in wetlands or lowlands, an adjacent parcel with habitable land was substituted.
- At least one-third of each bluff edge parcel was a flat, upland area behind or above the steeper slopes of bluffs.
After assigning all parcels in the survey universe to the appropriate strata, those parcels which fell within each particular stratum were numbered consecutively, from one to however many parcels were in the stratum (see Section C.3.2 on assigning parcels to strata). Each parcel received its own two-number identifier, which was a combination of the stratum number and the consecutive number. For example, the third island parcel was number 8-3 and the 689th general upland parcel was 17-689.
Tables of random numbers, which were provided by the project statistician and differed for each stratum, were used to select 10 parcels within each stratum, the minimum survey goal for the 1995 season. In most cases, 10 parcels were surveyed in each stratum, except where high water or recent alluvium prevented an investigation of all parcels assigned to the unsurveyable floodplain stratum. Random number table ranges doubled starting from 1-10 (i.e. 1-10, 1-20). During this selection process, numbers that were already chosen or out of range were skipped. For example, if a stratum contained 23 parcels, a set of random numbers from 1-40 would be generated. A random number of 20 would indicate that the twentieth parcel in that stratum should be surveyed; a random number higher than 23 (the number of parcels in the example) would be skipped. As the field season progressed and some of the original parcels were eliminated for lack of permission or unsurveyable conditions, the random number tables were used to select additional parcels. Replacement rules used in this process are discussed in Section C.3.1.4.
C.3.1.3 Precedence Rules
In cases where more than a single landform characterized a parcel, a series of rules was used to assign parcels to one stratum or another. The rules used for Nicollet and Stearns counties were also used in Becker and Beltrami counties with some minor modifications. A precedence list based on the rules for Nicollet and Stearns counties appears at the right side of Figure C.1. Two fundamental principles were used to formulate all the precedence rules:
Rule of Rarity: If a stratum was rare in a sampling universe, it took precedence within a parcel over a stratum that was more common. Within each sampling universe, researchers decided which landform strata were rarest, and ensured that an adequate number of parcels within those strata were identified. For example, if a parcel had three strata present, and one was rare and the others relatively common, the parcel was assigned to the rarest stratum.
The "1/3 Rule": Any landform that was relatively common in the survey universe must have covered at least one-third of a parcel for the parcel to have been assigned to that strata. The "1/3 rule" eliminated from the survey any parcel which had less than one-third of its area in a common landform stratum where the remainder of the parcel was unsurveyable. This rule did not apply to parcels with rare landforms. It also did not apply to certain landform types such as alluvial fans, islands, or peninsulas, where the amount of land present was small, as long as the actual feature of interest (e.g. the island) was included.
In the case of particularly small and rare landscape features, if at least one-half of the entire feature was in a parcel, it was coded as being in the feature’s landform stratum, even if its total area was less than one-third of the parcel. For example, if half an alluvial fan was in a parcel, it was coded as a fan, even if it made up only one-tenth of the land area of the parcel. The rule of rarity and "1/3 rule" were used to establish the following sequence of preference (from top to bottom) for Nicollet, Stearns, Beltrami, and Becker counties:
- Surveyable lake bottoms take precedence over all other strata because of their rarity.
- Islands take precedence over all remaining strata.
- Stream landforms in order of precedence: alluvial fan (highest), confluence, terrace, floodplain, bluff margin (lowest).
- Any stream takes priority over a wetland with streams draining wetlands coded as streams. Inlet/outlets take precedence over lakeshore.
- Point/peninsulas take precedence over coves and bays.
In order to achieve the 1995 goal of 10 surveyed units per stratum within the allotted field period, two principles were followed: (1) maximize randomness; and (2) maximize search efficiency (e.g., a cost-effectiveness criterion). The first principle influenced the method used to choose the original survey parcels (discussed above). The second principle governed which parcels were surveyed. A 'surveyable unit' was defined as a parcel within about 15 minutes walking time from a point of access (e.g., a road) for which permission to survey was granted. To increase efficiency and reduce the amount of time needed for shovel testing, the emphasis was on parcels that could be pedestrian surveyed (i.e. those with at least one-third of the entire surface exposed to a minimum of 25 percent visibility). Parcels were defined as ‘unsurveyable’ because permission was denied, the parcel would take more than 15 minutes to reach, more than one-fourth of the surface area was covered by modern construction, or more than one-third of the surface area was covered by wetlands or slopes greater than 15 percent. The only exceptions to these criteria were for landforms that were either very small, such as alluvial fans or most islands, or very rare, such as beach ridges. Finally, parcels for which a landowner contact could not be made in three attempts were also called unsurveyable, except in cases of rarely occurring strata. Reasons for replacing an unsurveyable parcel were documented. Parcels that had been placed in the incorrect strata during the initial strata classification (e.g., some floodplains and terraces in Stearns County) were surveyed as originally coded, assuming all of the above criteria were satisfied.
If the selected original parcel was unsurveyable for any of reasons above, the closest same-stratum parcel within a four-parcel radius of the original unit (as measured from any part of the parcel) was selected as a proximity replacement. If no replacement could be located within four units of the original parcel, a replacement was chosen from the series of random numbers for that stratum. To distinguish these replacements from proximity replacements, they were called random replacements. In anticipation of this scenario, five random replacement parcels for each stratum were selected before the field season began. As the field season progressed, additional blocks of random replacement parcels were selected as potential survey units. The parcels chosen earliest had to be considered for survey before those chosen later, although they did not have to be surveyed in the order selected. Each crew adopted the most efficient order of survey. Both the original and the replacement parcel identifier numbers were documented.
C.3.1.5 Pre-Field Research
The state site files, historic maps and documents, SHPO historic contexts, and other records were examined to identify any known precontact or historic sites in the sample parcels. Areas previously surveyed were automatically excluded from survey if the methodologies were comparable, except in one case in Stearns County where no other suitable replacement parcel could be found. Thus, if a road-side parcel contained five acres surveyed by the 1990 Minnesota Trunk Highway Archaeological Reconnaissance Study, those five acres were not resurveyed. However, the remainder of the parcel was surveyed. If the methods of the previous survey were not comparable or not reliable, that portion of the parcel was considered as "no permission to survey" and was excluded. This procedure was adopted rather than a resurvey of the area because of concerns expressed by some county officials that sites might be found on land cleared by a previous survey.
Unlike the 1995 stratified sampling procedure, the 1996 field survey design was a basic one characterized by a simple random sample of 30 m2 cells within each of three counties: Cass, Wright, and Wabasha. The 30 m2 cell is the smallest geographic unit in the GIS database. Depending on the ground cover present, these cells either represented a point at which shovel testing was conducted, or served to identify for survey the larger 40 acre quarter-quarter section in which they were located. A simple random sample of these cells was generated by GIS. Each randomly chosen 30 m2 cell served as a starting point on the landscape where the field survey was conducted.
This sampling procedure proved to be much faster and involved many fewer manual coding decisions than the 1995 design. GIS generated a series of full sized maps at 1:24,000 depicting the location of generated random cells, known sites by SHPO type, water bodies, major roads, and map edges. No other pre-field background research was conducted. The random cells or points were transferred onto topographic maps and 40 acre quarter-quarter sections drawn around them. Only enough parcels or points needed to complete each survey were plotted. Compared to the 1995 design, which involved gridding the entire county or transects within counties into 40 acre parcels, coding each parcel for the dominant landform based on a series of precedent rules, numbering them, and drawing random parcels from each stratum, the 1996 sampling design saved a great deal of time. It also allowed for the inclusion of all 40 acre parcels in the sampling population or universe, regardless of the size of the county. This proved to be a very important factor in initiating field work, since a relatively short period of time (two months) was available in the spring between authorization to proceed with each survey and ground-obscuring crop cover.
The field surveys started in order of cells or 40 acre parcels drawn, gradually progressing from smaller to larger numbers. Surveying proceeded by completing parcels in the order they were cleared for permission. This order was largely based on geographic proximity; those blocks of parcels near to one another were cleared in order to conserve drive time.
C.3.2.1 Survey Units
Although the method in which survey units were chosen differed from 1995 to 1996, the actual size of the survey units remained largely the same from one season to the next. In Wright and Wabasha counties, where pedestrian surveys were a component of the field work, the basic unit remained the 40-acre parcel. In Cass County, where ground cover required shovel testing of all parcels, the survey unit was the location of the random point, around which were excavated up to 32 shovel tests. Although the configuration and location of the shovel tests differed from 1995, the maximum number of tests remained the same at 32, or 5 percent coverage of a 40-acre parcel. In cases where the random point fell along the edge of two 40-acre parcels, both were plotted. The parcel with the most pedestrian surveyable land was chosen as the one to be surveyed. If the owner of that parcel could not be located, the other parcel was selected. Additional variations in field techniques are discussed later in this appendix.
Unlike the previous year’s survey, there were no replacement procedures in effect during 1996. Any parcel that could not be surveyed for reasons related to permission or the nature of the parcel (e.g. water, wetland, or crop covered, slopes greater than 25 percent, developed), was recorded, and the next sequentially numbered parcel was surveyed. The parcels that could not be surveyed were accounted for on parcel forms, however. This allowed for a tabulation of reasons for not surveying certain parcels and a consideration of the degree to which the survey deviated from a "true" random sample unaffected by unsurveyable parcels. At the conclusion of the Mn/Model project, these figures could help to determine if the unsurveyable parcels or points differed in a systematic way from those that were surveyed. For example, the extensive residential development along the margins of large, deep-basin lakes that made these areas impossible to survey inflated the number of surveyed parcels towards areas away from water or to smaller, shallow-basin lakes. Correcting for these factors would provide a more accurate a priori or baseline figure for the number and density of sites within particular regions. Any major differences may also have an effect on model development.
Other procedures used to define survey units remained the same from 1995. The system of legal divisions was the same, as was the placement of the legal land locator template used to outline each 40 acre parcel. The "address" system used to define each 40 acre parcel was also consistent from one field season to another. These procedures were described in detail under the 1995 field season survey units. During the 1996 field season, FIP codes were added to the random number on the survey form.
Survey rationales and corresponding survey procedures differed for the 1995 and 1996 field seasons. The 1995 field survey was designed primarily to provide data to assess the adequacy of the Minnesota Statewide Archaeological Survey (MnSAS) to discover sites of all sizes. Data from the 1995 surveys also served to develop a predictive pilot model of Nicollet County and provided information used to develop predictive models of the archaeological regions in which Beltrami, Stearns, and Nicollet counties are located. A key component of the sampling procedure was to ensure that all landscape strata, whether they were suspected of containing sites or not, were equally represented in the survey. The 1996 field work, on the other hand, was designed as a simple random survey that could be used to test any one of a series of predictive models for each of the three regions surveyed.
The need to "calibrate" or assess the MnSAS data governed the structure of the 1995 survey procedures. There are many problems with using the MnSAS data for a predictive model. The stratification schemes used for each MnSAS county were inconsistent, often distinguishing only between parcels near current bodies of water larger than 40 acres and those away from such large bodies of water. The field survey methods were also inconsistent, and the survey interval employed was wide, with surveyors usually spaced 50 meters apart (the modern standard is 15 m). In addition, discussions with several MnSAS surveyors suggest that the transects were not always straight lines, but occasionally wove across the landscape, designed to intersect higher probability areas between surveyors and to cover more land between transects. Thus, the actual distance that was not examined between surveyors is not always known. Although it can reasonably be asserted that sites of a certain size and type were present in the MnSAS parcels, additional information was necessary to determine whether MnSAS results needed to be adjusted before they were incorporated into the Mn/Model archaeological database.
The 1995 survey strategy was designed to provide data that would allow for a "calibration" of the large and possibly problematic MnSAS data set. For the 1995 survey, a stratified random sample was drawn from four counties, with up to 19 strata defined in terms comparable to those used for MnSAS. A maximum of ten 40-acre parcels was surveyed from each stratum. The stratification scheme represented a modification of the one used by MnSAS to ensure maximum comparability. The same dichotomy of parcels near and away from water was retained, but it was subdivided into up to 19 separate strata to reflect differences in landscape settings within these two broad categories. The 1995 design attempts to capture variations in stratification procedures used in different MnSAS surveys. The survey unit itself, the 40-acre parcel, was retained for consistency in results. The survey interval was reduced to 15 m, and rigid quality control standards were implemented to ensure that the data collected was accurate and consistent.
To "calibrate" the MnSAS, several statistical approaches were explored. For example, the total area and site density from comparable survey areas (each of the different strata) was calculated to compare the number of sites per unit area surveyed in each stratum. This provided statistical data that allowed evaluation of the different survey intervals and techniques employed in the MnSAS.
Besides providing information that helped in using MnSAS data, the survey data were also used in a pilot study of Nicollet County. This pilot project permitted refinement of the various GIS procedures for data acquisition, manipulation, and statistical analyses within a well-defined and relatively small-scale database, before their application to other regions. Finally, the data from Becker, Beltrami, and Stearns counties were used as probabilistic data in the development of initial models for the Phase 1 and 2 archaeological regions in which they occurred (i.e. Central Deciduous Lakes, Central Coniferous Lakes, Northern Bog).
The results of the 1996 field season were used to test Phase 2 models and to determine a priori probabilities for finding sites in the counties surveyed. This was made possible by employing a simple random sampling design of points, expressed as either 40 acre pedestrian-surveyable or two acre shovel-testable parcels within each of three counties (Cass, Wabasha, and Wright). The simple random survey, one that gives all sample units an equal chance of being selected, was suggested by Mn/Model advisor Dr. Kenneth Kvamme. This sampling design contrasts to the 1995 survey which was weighted towards parcels near water. After testing Phase 2 models, the results of the 1996 surveys were incorporated into developing Phase 3 models of the archaeological regions in which the counties occurred.
Once a parcel was chosen for survey, the landowners and tenants were identified through plat books, phone books, and occasionally county records. In some cases, neighbors were contacted to determine the name of landowners and their place of residency. The limit for obtaining permission was two visits and a telephone call. The first landowner contact generally entailed a brief verbal description of the project, the transmittal of an explanatory form letter and business card, and a request to survey the land and collect any artifacts. During the 1995 season, the standard form letter on BRW letterhead was not developed until most of the Nicollet County survey was completed; consequently, letters and business cards from MVAC, which surveyed that county, were distributed to landowners. In 1996, a standard letter on a BRW letterhead was distributed to landowners in all three surveyed counties. After each survey was completed, thank you letters were sent acknowledging the assistance of each landowner and informing them of the results of the survey on their land. If permission was denied, the replacement rules described for the 1995 season were implemented to find an alternative parcel. If permission was denied during the 1996 field season, no replacements were obtained; the next numbered parcel was surveyed. The archaeologists obtaining landowner permission in 1996 did not visit them in strict numerical order. Rather, they considered blocks of parcels so that driving from one to another could be minimized.
When landowners requested that artifacts be returned to them rather than curated, the person making the initial contact noted the request and indicated that the artifacts would be returned at the end of the project. After permission was granted, any information the landowner could give regarding surface exposure, accessibility, special conditions of access, and identification of the specific sample block in the field was recorded. The appropriate time for the survey and any notification immediately prior to the survey was arranged with the landowner at this time. The parcel form documented the Mn/Model personnel making the contact and the landowner and/or tenant who gave permission for access with their address and phone number. Each landowner was also asked for any information on prehistoric or historic artifacts found in the sample block. Sites identified only through collector information were not included in the predictive model, although the information was gathered as non-probabilistic data. The parcel forms differed somewhat from 1995 to 1996 (Section C.7).
The primary goal of the 1995 field season was to survey ten randomly distributed 40-acre parcels within each of the 19 landform strata as efficiently as possible. Because pedestrian surveying is more cost-effective than shovel testing, the emphasis was on completing as many pedestrian-surveyable parcels as possible while maintaining the overall sampling design within each county. As a consequence, shovel testing was conducted only when there were no alternative or replacement parcels that could be pedestrian surveyed.
The objective of the 1996 field season, on the other hand, was to survey at least 90 parcels or points, taken in order of selection and regardless of landform configuration, surface conditions, or ease of access. The emphasis was on surveying the points or parcels in the order in which they were selected to preserve the random nature of the sample.
The reconnaissance survey for the 1995 and 1996 field seasons excluded areas of water-saturated soil, steep slopes (i.e. greater than 25 percent), and those buried by recent alluvial deposits beyond the range of shovel testing (i.e. deeper than 70 cm). One exception to the steep slope criterion involved draws or coulees, which might be expected to contain exposed flakable raw materials. Since it is relatively common to find lithic reduction sites in these areas and in stream beds, locating them would provide important environmental information on site location strategies. Raw materials, wherever they occurred, were collected in 1995 by all crews; MVAC is the only group that was directed to collect these materials in 1996. Although the focus of both seasons’ field work was on precontact cultural materials, historic archaeological sites were recorded and collected unless they were less than 50 years old.
In areas of adequate surface exposure, such as in cultivated fields, pedestrian survey was conducted at 15 m transect intervals. The two outer transects were placed 15 m in from the margins of a parcel, while the ends of each transect were at the edge of the parcel or at uncultivated soil, as at the margin of a road ditch. Surface visibility of at least 25 percent was required for a pedestrian survey; anything less required subsurface testing (Zeidler 1994). Areas that could not be tested by standard shovel testing procedures (e.g., buried ground surfaces below 70 centimeters in depth) were not examined during the reconnaissance portion of the survey.
To prevent false positive records (i.e., reporting sites where none exist), to ensure that site contexts were correctly identified from recovered artifacts, and to minimize the time in the field spent examining and documenting potential cultural materials, only a sample of artifacts was collected during pedestrian survey. All artifacts observed within a three-meter-wide collecting transect, the equivalent of one corn row on each side of the individual, plus the row that the individual walked in, were flagged, collected, and placed in zip-lock plastic bags with a field tag according to transect. No artifact collecting was allowed between transects, unless a diagnostic artifact was noted. In this case, the diagnostic artifacts were bagged and recorded as part of a non-probabilistic sample. Artifacts and sites not found during actual surveys (i.e., a site found by surveyors walking to a parcel from a vehicle or in a cutbank not part of a pedestrian or shovel test transect) were recorded as non-probabilistic sites. Similarly, sites that extended beyond a parcel were not defined outside the parcel’s boundaries.
If features were encountered, such as an historic building foundation, black and white photographs and color slides were taken. A sample of culturally diagnostic artifacts was collected from the surface of these sites, but no subsurface testing was done. A brief description of the features was noted on parcel forms.
During the 1995 field season, a number of single artifact sites or "isolated finds" were recorded. Since the pedestrian survey procedures did not allow for any systematic examination of areas outside of the 15 m transects as a time-saving procedure, there was some uncertainty whether additional more intensive examinations of the ground surface would produce more artifacts. This problem was rectified for the 1996 field season by adding additional transects when an artifact was found. In this case, two additional parallel transects, 5 m on either side of the transect in which the first artifact was found, were added to the survey. These additional transects were walked for a distance of 15 m in either direction beyond the location of the initial artifact.
Recording locational information also differed somewhat between the 1955 and 1996 field seasons. In 1995, the locations of the sites were recorded on USGS 7.5 minute topographic quadrangles and U.S. Department of Agriculture aerial photographs when available. Areas surveyed and crop cover conditions were also recorded on aerial photographs. In 1996, an enlargement of the USGS 7.5 minute topographic quadrangle for a particular 40 acre parcel was used as a base map to plot survey areas and sites. The scale of this map was 1:3,000, corresponding to 5mm for every 15 m to facilitate field plotting of specific transects, shovel tests, and sites. This scale can be produced by enlarging an original topographic map by 2 x 2 x 2 on a copying machine.
Shovel testing was performed in all areas having less than 25 percent surface visibility. Areas that could not be tested by standard shovel testing procedures (e.g., buried ground surfaces below 70 centimeters in depth) were not examined during the reconnaissance portion of the survey. Although the shovel testing techniques did not vary from season to season, the location and configuration of shovel test transects did. Shovel tests were circles 40 centimeters wide with vertical walls. All soil was dry-screened through ¼- inch hardware cloth; the holes were backfilled. The tests extended to "sterile" deposits, such as glacial till, or to 50 centimeters below the surface, whichever was reached first. In those cases when artifacts were found at 50 centimeters, shovel tests were extended to greater depths to test the nature of the underlying deposits. Cultural materials were bagged by 20 centimeter levels. All artifacts found during shovel testing were collected, bagged, and identified by transect, shovel test number, and depth. During the 1995 season, the depth and soil profile (color, texture of horizons) of all shovel tests were recorded on forms. In 1996, this information was noted for only the first and any positive shovel tests. During both seasons, individual transects and shovel tests were plotted on sketch maps and aerial photographs (1995) or enlarged topographic maps (1996).
Shovel tests were placed at 15-meter intervals along transects separated by 15 meters. The shovel tests between adjacent transects were offset or staggered 7.5 meters so that a grid of tests formed a triangular rather than a square pattern. Although the distance between the nearest shovel tests on adjacent transects is increased from 15 m to 16.8 m, the amount of area between contiguous tests is reduced from a square of 225 m2 to a triangle of 112.5 m2. This improves the probabilities of discovering smaller sites (Zeidler 1994) and sites where there is a strong spatial component in the patterning of artifact locations. The goal was to excavate at least 32 shovel tests in a parcel, or the equivalent of a 5 percent sample of the parcel. Standard information was recorded about each positive shovel test, including location, depth of excavation, soil texture and color, and artifacts found. To increase efficiency, only locational information of negative shovel tests was recorded. The location and number of each shovel test was recorded on a field sketch map.
C.6.2.1 1995 Field Season
To maximize the efficiency of each survey, shovel testing was conducted only when there were no alternative or replacement parcels that could be pedestrian surveyed. Since it was not feasible to shovel test completely all parts of a parcel lacking good surface exposure, the following sampling procedures were developed. Shovel testing procedures differed between those counties where there was extensive agricultural activity (Stearns, Nicollet) and counties where very little tilled ground was present (Becker, Beltrami).
In Nicollet and Stearns counties, where most of the land was in agricultural production, surveyors focused on testing the primary landform identified as the stratum for a 40 acre (quarter-quarter section) parcel. For example, if a wooded parcel was defined as a peninsula but also contained other landforms such as other shoreline and uplands, the peninsula was the focus of testing. Other portions of the parcel with obscuring ground vegetation were generally not shovel tested. If any other portion of the parcel had good surface visibility, it was also pedestrian surveyed. In cases where nearly all of the parcel was pedestrian surveyed, as in Nicollet County, small portions obscured by ground cover were not shovel tested if they did not represent unique landforms.
The goal of shovel testing in a 40-acre parcel was to excavate up to five parallel transects of one-fourth mile in length, up to a maximum of 130 holes or 20 percent of the parcel. In reality, five parallel transects and 130 holes could rarely be excavated given time constraints and the topography of some parcels (e.g., small islands or narrow points and peninsulas). In Stearns County, where most of the early shovel testing occurred, the number of tests per parcel ranged from about 15 to 45 shovel tests per parcel. All shovel testing during the 1995 season was limited to established transects; no bracketing of positive shovel tests was performed.
Due to obscuring ground vegetation or leaf litter, most parcels in Becker County and all in Beltrami County required shovel testing. In order to preserve the integrity of the sampling program and complete the Mn/Model survey in these counties, the procedures described for Stearns and Nicollet counties were modified. In Becker and Beltrami counties, the 40-acre parcel was divided into four ten-acre units (quarter-quarter-quarter section). Of the four, the one that best characterized the parcel’s assigned landform stratum was chosen for shovel testing. For example, if an inlet/outlet parcel contained one 10-acre unit in an upland area and two with less than 10 acres of testable land, they were eliminated in favor of the fourth unit with the most shovel-testable acreage nearest the inlet/outlet. There were no cases where two or more 10-acre units were equal in land area or diversity that would have necessitated a systematic random method of choosing between them.
C.6.2.2 1996 Field Season
Shovel testing during 1996 was based the on modified procedures employed during the 1995 survey of Becker and Beltrami counties. Unlike the landform-based units of 1995, the 1996 shovel testing focused on testing at the 30 m2 random cell, which becomes a point on a 1:24,000 topographic map. Up to 32 shovel tests, the equivalent of 20 percent of the 1995's quarter-quarter-quarter section (10 acres), were excavated at each point. An east-west oriented grid of staggered shovel tests composed of seven transects was centered over the random cell rather than linear transects as in 1995. Shovel tests that could not be excavated for any reason, such as on steep slopes, were not replaced. A minimum of one shovel test had to be excavated at a random cell for it to be included in the survey; random cells unsuitable for shovel testing were recorded and skipped in favor of the next chosen cell. This procedure was used for those parcels that were completely obscured and contained no pedestrian-surveyable areas, such as in Cass County.
A somewhat different series of shovel testing rules was applied to 40 acre parcels containing both walkover and shovel testable lands, as in the Wright and Wabasha county surveys. When the shovel testable land is greater than 25 percent of the total 40 acre parcel, a proportionate number of shovel tests were excavated. For example, if a parcel contained 30 acres of grassland and 10 acres in a corn field, ¾ of a maximum of 32 shovel tests for an entire 40 acre parcel, or 24 holes, were excavated. Shovel testing on these parcels, like those with completely obscured ground surfaces, centered on the randomly generated 30 m2 cell. In cases where the random cell fell within a cultivated field, it was moved to the nearest shovel-testable land. In cases where the exposed and ground-obscured areas were part of the same topographic feature and showed no natural variation, a minimum of eight shovel tests were excavated. If the percentage of obscured ground was equal to or less than 25 percent in this case, no shovel testing was performed. However, if the walkover and shovel testable lands were in different topographic settings, such as a wooded peninsula or point adjacent to an upland agricultural field, then there was no minimum set on the number of shovel tests. The number of shovel tests would be based on a proportion of wooded area to agricultural land. For example, five acres of woods to a total of 40 acres would constitute a ratio of 0.125. Multiplying this figure by the number of shovel tests excavated in a fully obscured parcel (32) would yield a total of four shovel tests.
In those instances where only a single artifact was recovered, the positive shovel test was bracketed at five m intervals in the four cardinal directions up to a maximum of four additional shovel tests. Shovel testing was discontinued whenever a second unequivocable artifact was recovered. The bracketing tests were labeled with their associated shovel test number followed by an N, S, E, or W. The reason for bracketing was to help distinguish between those sites characterized by single artifacts and sites with low numbers of multiple artifacts.
A variety of standardized forms were used to document the survey procedures and results. There were some changes forms from 1995 and 1996, reflecting the usefulness of various types of information after a field application. The amount of information was reduced on the "parcel" and "cultural resources identified" forms, allowing for a consolidation of both onto a single form in 1996. Among the information that was eliminated from 1995 were water characteristics, ground cover conditions, naturally occurring lithic resources, and UTM coordinates from Global Positioning System (GPS) readings. All other forms remained unchanged. Since the 1995 forms were the most detailed, most of the following discussion applies to them. Examples of the most important of the standardized forms used during the 1995 and 1996 field season discussed below can be viewed by clicking the following hyperlinks; the remaining forms varied somewhat from one survey team to another.
This form was the most basic one employed in 1995. It was used to document the location and condition of the parcel, the landowner or tenant’s name and address, the type and amount of field work conducted, and information on ground cover, topographic setting, and identified sites. Some of the environmental variables were derived from the topographic maps and verified or qualified in the field. Other information, such as ground cover and presence of lithic raw materials, became available only through field observation. The parcel form also documented reasons for not surveying any areas, such as evidence of historic disturbance or burial by thick colluvial or alluvial deposits. Aerial photographs marking the location of areas surveyed were attached to the parcel form after the location of areas surveyed, locations of shovel-test transects, ground cover, percent surface visibility, and site location were marked. Shovel tests were drawn on sketch maps and/or aerial photographs during the 1995 field season. In 1996, shovel tests and pedestrian transect locations were drawn on enlarged 1:3,000 copies of 7.5 minute USGS topographic quadrangles.
The 1995 parcel form also included locational information on all four corners of the parcel. Initially, surveyors had planned to use the Global Positioning System (GPS) to provide a detailed location for each parcel. In addition, site locations, shovel tests, and other features were to be located using GPS. GPS readings were taken at most sites and surveyed areas in Beltrami and Nicollet counties. Because of numerous technical difficulties, GPS was not fully incorporated into the project. The readings that were taken were post-processed and will be available for future reference, but were not used to locate parcels and sites and are not contained in this report. The three major problems in using GPS were: (1) the lack of radio signals to allow real-time differential correction of satellite signals; (2) a small number of satellite signals, particularly in wooded areas; (3) time-consuming post-processing of data and conversion into a form useful for GIS. The absence of radio signals for real-time conversion meant that each position had to be read at least five and often ten minutes to achieve the desired level of accuracy. For many of the parcels surveyed, GPS readings at the four corners of the parcel and at the locations of any cultural materials took more time than the actual survey. Since GPS readings were not available for most Mn/Model parcels or sites, conventional techniques were used to generate UTM coordinates from 7.5 minute USGS topographic quadrangle maps.
Archaeological sites, including isolated finds, were recorded on the Cultural Resources Identified form during 1995, which, along with other field documentation, was used to prepare the Minnesota State Site Form. These forms were also completed for non-probabilistic sites, although they were not considered in the probabilistic survey data base. All site forms were archived on computer disk and on paper at BRW, Leech Lake, and MVAC, with paper copies available through the Office of the State Archaeologist (OSA). The Cultural Resources Identified form documented each site’s location, size, type, relation to topographic features, disturbance, diagnostic artifacts, and historic context. All sites were plotted on USGS 7.5 minute topographic quadrangle maps and aerial photographs in 1995, except those in Beltrami County. During the 1996 field season, all sites from Cass, Wright, and Wabasha counties were plotted on enlarged 1:3,000 USGS 7.5 minute topographic quadrangles.
This form was used to record information on positive shovel tests during the 1995 field season. To increase the efficiency of the field work, negative shovel tests were not recorded except as numbers along the transects in each parcel. A somewhat different procedure was followed by MVAC in 1995 in Nicollet County, where all shovel tests, both positive and negative, were recorded in field notebooks rather than on forms.
This form was used to document information for both color slide and black and white photographs. Photographs for all parcels were taken to document environmental conditions. All sites or groups of sites were also photographed.
This form was used to assign a unique lot number to each bag, each of which usually contained either artifacts from a single surface-surveyed site or from a positive shovel test. It was designed to track the transfer of all artifact bags and records to different personnel, so that the location of all materials could be monitored. Each team was assigned a block of lot numbers for this purpose at the beginning of the season.
The supervisor's daily journal included information not provided elsewhere, such as the activities performed each day, reasons for selecting specific parcels, replacement procedures, problems encountered, and the names of amateur collectors. No standard form was used; each supervisor used the most convenient format.
The artifacts from each site were cleaned, described and cataloged, inventoried, and analyzed. The specific procedures used varied from team to team. Artifacts were catalogued and documented for final curation at the Minnesota Historical Society (MHS) using their accessioning system and laboratory protocol. The artifacts returned to the landowners at the end of the project were documented and photographed.
Consistency and comparability in data collection methods are key factors in achieving Mn/Model’s goals. To ensure the quality, reliability, and consistency of all data acquired during the field survey, several quality control procedures were established at the beginning of the project.
Uniform and acceptable survey standards and procedures were established and monitored throughout the survey by the Principal Investigator for Archaeology, James Gallagher, and the Survey Coordinator, Connie Arzigian, both of MVAC. One or both visited each crew in the field at least twice a month to monitor consistency and reliability of field data acquisition. The visits included meetings with field supervisors and reviews of field records for consistency, accuracy and completeness. At the end of the field season, paper copies of field records were sent to MVAC for a final review. Chipped stone artifact verifications and raw material identifications were made by Kent Bakken, a regional lithic specialist.
Minnesota Historical Society
1981 Minnesota Statewide Archaeological Survey. Summary: 1977-1980. Minnesota Historical Society,
1993 SHPO Guidelines for Archaeological Projects in Minnesota. Minnesota Historical Society, St. Paul.
1994 Archaeological Inventory Standards and Cost-Estimation Guidelines for the Department of
Defense. U.S. Department of Defense, Washington, D.C.
The Mn/Model Final Report (Phases 1-3) is available on CD-ROM. Copies may be requested by visiting the contact page.
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.
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.