An Investigation of the Distribution of Open Archaeological Sites in the Upper Kickapoo Valley Archaeological District, Vernon County, Wisconsin

An Investigation of the Distribution of Open Archaeological Sites in the Upper Kickapoo Valley Archaeological District, Vernon County, Wisconsin.

Ben Johnston Department of Resource Analysis, Saint Mary’s University of Minnesota, Winona, MN 55987

Keywords: Archaeology, Site Selection, Upper Kickapoo River Valley Archaeology District, Kickapoo River, Wisconsin Archaeology, Spatial Statistics

Abstract

The Upper Kickapoo River Archaeological District contains over 500 archaeological sites. Spatial statistics and site selection factors were explored for 405 open archaeological sites. Nearest Neighbor Index and Ripley’s K-Function analyses show that the sites have a clustering pattern. Hot spot analyses reveal concentrations near a few stream confluences. Topographic variables of the site locations show a preference for lower elevations and low slopes. Furthermore, comparisons of cultural affiliated sites indicate the Paleo-Indian sites were farther from water resources when compared to Archaic and Woodland affiliated sites.

Introduction The ACOE purchased over 8500 acres from local residents for the proposed The Upper Kickapoo Valley Archaeology reservoir and additional adjacent District (UKVAD) is located in Vernon recreation areas. County, Wisconsin, between the villages Most of the archaeological surveys of La Farge and Ontario. With over 500 in the 1960’s and 1970’s were focused on archaeological sites identified throughout the areas within the proposed reservoir the district, the UKVAD was recognized pool level (Holtz and Boszhardt, 1998). by the National Registry of Historical Yet, some surveys covered the adjacent Places in 1999 (www.nps.gov/history/nr/). recreational areas surrounding the lake. The archaeological significance of Later surveys in the 1980’s and 1990 the UKVAD was relatively unknown prior targeted areas for road construction and to 1960, with only a handful of sites highway realignment within the valley reported for the area (Holtz and Boszhardt, (Holtz and Boszhardt, 1998). 1998). However, during the late 1960’s Due to many circumstances, the through the 1970’s, extensive Lake La Farge project was not completed. archaeological surveys were conducted in In 1996, the project was officially anticipation of an Army Corps of decommissioned by the United States Engineers (ACOE) flood control project, Congress. The ACOE acquired land was known as the Lake La Farge Project. The transferred to the State of Wisconsin and project proposed the creation of a 1,780 the Ho-Chunk Nation in 2000, creating the acre lake to help mitigate flooding Kickapoo Valley Reserve. problems in the Kickapoo River Valley. In preparation of the land transfer,

Johnston, Ben. 2010. An Investigation of the Distribution of Open Archaeological Sites in the Upper Kickapoo Valley Archaeological District, Vernon County, Wisconsin. Volume 12, Papers in Resource Analysis. 18 pp. Saint Mary’s University of Minnesota Central Services Press. Winona, MN. Retrieved (date) from http://www.gis.smumn.edu

the Mississippi Valley Archaeological rock flakes or chips associated with the Center conducted a final archaeological shaping of stone for tool production. The survey in 1998 of approximately 3,500 presence of these flakes suggests a tool acres that were not covered in earlier making location and/or short term camp surveys (Holtz and Boszhardt, 1998). (Holtz and Boszhardt, 1998). These areas were mostly upland benches “Isolated finds” are sites where and terraces within the Kickapoo Valley only a few artifacts were identified, but Reserve. not substantial enough to make any definitive conclusions of the site use. Archaeological Cultural Periods in the These could be lost stone tools, short UKVAD stopovers or discarded items (Holtz and Boszhardt, 1998). Diagnostic artifacts such as stone tools of Rock shelters are another type of particular shapes and/or design, the archaeological site found within the presence/absence of pottery, and evidence UKVAD. These are rock overhangs or of wooden tools are used to determine recess caves produced by differential cultural periods, or time intervals when a erosion between sandstone layers. site was occupied (Theler and Boszhardt, Archaeological surveys at several rock 2003). Other cultural specific indicators shelters have yielded evidence indicating include burial practices and evidence of occupancy by prehistoric people, most agriculture. likely during winter (Theler and Many cultural artifacts have been Boszhardt, 2003). A total of 91 rock found in the UKVAD. Three cultural shelters have been identified in the periods represented include the Paleo- UKVAD. Indian Period between 12,000-9,000 B.P (before present), the Archaic Period, Distribution of Archaeology Sites 9,000-2,500 B.P., and the Woodland Period 2,500-1000 B.P (Holtz and A distribution study of archaeological sites Boszhardt, 1998; Theler and Boszhardt, is dependent upon three factors (Guccione, 2003). Cultural periods can be further 2008). The first is whether or not a site is divided into early, middle, or late sub- preserved over time. Natural process such cultures (i.e. Middle Archaic). as sedimentation during a flood can bury a site. Yet stream migration can later expose Archaeological Sites in UKVAD or destroy a site (Guccione, 2008). A second variable in site Archaeological sites are categorized distribution is based on the site selection according to the function of the site and/or process by prehistoric people. Proximity to features found at a given location. Some of available resources such as water (Zhang, the sites include camps, villages, lithic Zhao, and Lui, 2007), hunting/foraging scatter, and isolated finds. Collectively, areas (Carlisle and Friedel, 2009), natural these sites are referred to as “open sites”. shelters, and sources for stone tools The term “village” used in the (Theler and Boszhardt, 1998) are key UKVAD describes larger camp areas of selection factors. Physical environment extensive use over time (Holtz and variables such as topography (elevation, Boszhardt, 1998). Lithic scatter sites are slope, aspect) can influence site selection. characterized by the presence of numerous Political, spiritual, and social factors may

2 also affect the site selection process; UKVAD within the Kickapoo Valley although the role of these factors is not Reserve. In addition, a 500 m buffer was easily determined from the archaeological created surrounding the Kickapoo Valley record (Ebert, 2004). Reserve to incorporate an additional 16 A third factor in site distribution is open sites. This particular study area was the available archaeological information. delineated due to the available The archaeology record is never complete archaeological data. Forty additional sites since sites may not be located and/or beyond the 500 m buffer were excluded significant artifacts may not be found from this study since detailed during a survey (Guiccone, 2008). archaeological information was not Furthermore, sites with more extensive use available. are more likely to be identified during an archaeological survey versus a site used Physical Setting only for a short time. The geology of the UKVAD consists of Purpose of This Study Cambrian sandstone layers at the valley bottoms and along hillsides. These layers This study investigated the distribution of are capped by Ordovician carbonate layers archaeological sites within a portion of the occurring near the ridge top elevations. Upper Kickapoo Valley Archaeological Variations in relative hardness of the District (UKVAD). Spatial statistic tools sedimentary layers have resulted in from ArcGIS 9.3 were used to evaluate differential erosion, producing a dramatic spatial patterns. Four different data sets landscape. were used for the analysis; the entire The Kickapoo River is the largest population of open archaeology sites, and stream within the study area. The total sites affiliated with each of the three length within the study area is 26.6 cultural periods: Paleo-Indian, Archaic and kilometers. The Kickapoo Valley Woodland. watershed is characterized as dendritic Furthermore, the study explored with deeply incised stream channels (Holtz relationships among topographic variables and Boszhardt, 1998). and distances to resources with the open Seven major tributaries join the site locations. Topographic variables Kickapoo River within the study area. In included aspect, slope and elevation. addition, many smaller streams (i.e. Resources considered for this study were unnamed creeks and/or intermittent rock shelters, water, cliffs/steep areas, and streams) are also present. Evidence for stream confluences. Proximity to stream meandering is present along particular stream sizes was also evaluated. portions of the Kickapoo River and some A third part of the study compared of the major tributaries. the cultural affiliated site based on The total relief for the study area locations. Five variables were selected to varies between 90 m (meters) to 152 m. investigate possible differences among the The valley elevation values are between cultural periods. 253 m at the upstream most point to 240 m at the downstream most point of the The Study Area Kickapoo River. The major tributaries have slightly higher elevations of This study focused on the areas of the approximately 270 m.

Ridge elevations within the study locations of archaeological surveys area are between 367 m to 375 m. Just conducted prior to 1997. According to the below the ridge line is an occurrence of metadata, the point features were digitized cliff faces (Holtz and Boszhardt, 1998). in 1997 prior to the land transfer. The data Cliffs also occur along portions of the source was USGS 7.5’ topographic field Kickapoo River and the major tributaries. maps provided by the State Historical The slope at the river bottom and Society. on the ridge top is generally flat. Hillside Also in 1997, the Army Corps of slopes have a wide range and can often Engineers compiled the results of the exceed 50 degrees. archaeological surveys into an electronic Upland terraces or benches occur database. Archaeological information between 275 m to 305 m above sea level included the cultural period affiliated at (Holtz and Boszhardt, 1998). Rock each site as well as the type of site (i.e. shelters are located just below this terrace open, rock shelter, burial mound, etc). The level (Hatfield, 2009). shapefile and database were linked by Alluvial terraces are present along corresponding site numbers. the stream channels and flood plains The Kickapoo Valley Reserve (Holtz and Boszhardt, 1998). The original provided a printed copy of the (middle Holocene age) valley alluvium archaeological results of the 1998 deposits are covered by more recent (+/- Mississippi Valley Archaeological Center 200 years) erosion deposits from post- survey. The report included maps showing European agriculture practices. These archaeological sites identified during this “Post Settlement Alluvium” (PSA) survey. These were plotted on sections of deposits consist of eroded upland top soil 7.5’ USGS 20 ft contour topographic exceeding several feet in some areas maps. (Holtz and Boszhardt, 1998; Johnson, From these maps, point locations 1976). were digitized in ArcGIS for this study. Site locations were referenced with 20 ft Methods contours generated from a 10 m digital elevation model (DEM). Site number, type Spatial statistical analyses and GIS data of site, and cultural period from the 1998 manipulations were conducted using ESRI MVAC report were entered into the ArcGIS 9.3. Tools referenced for analyses attribute table. The 1998 data points were were from the Spatial Statistic Tools, combined with the pre-1997 data into one Spatial Analyst Tools, and Data shapefile. Management tool boxes. Microsoft Excel In compliance with the metadata was used for data organization, provided with the KVR archaeology manipulation, distribution analysis, and shapefile data, a request to use the descriptive statistics calculations. All archaeological data for this study was graphs were created in Excel as well. submitted to the Wisconsin State Historical Society on January 30, 2010. Archaeological Data Sources Permission was granted on February 2, 2010. The Ho-Chunk Nation, Department The Kickapoo Valley Reserve provided a of Cultural Heritage Preservation was also shapefile containing over 450 notified of the project on January 30, archaeological sites. These were the 2010.

Archaeology Database Development An index value was created in order to quantify the popularity of sites across For the purpose of this study, the multiple cultures. This was developed to geographic location, site number, site type, serve as weight for a spatial distribution and cultural period information were analysis. To obtain this index value, each extracted from the compiled archaeology open site was given a value of one, since databases. The ACOE database contained values of zero would not be considered as multiple records for many of the sites, a weight. Thus, all sites without any reflecting different surveys from different cultural components identified (i.e. years. These records were consolidated “undetermined prehistoric”) received a into one record. total value of one. Then, for each sub- Furthermore, since the ACOE culture period listed (i.e. early, middle or compiled database only listed sites as late) an additional tally was given. If only “Open” without any additional a general period was listed (i.e. information, the more specifically “Archaic”), only one tally was recorded, described open sites from Holtz and not three. The tallies were added together Boszhardt (i.e. lithic scatter, camp, village, to reach the popularity index value. isolated find), were all generalized as “Open sites”. Therefore, this study made Spatial Statistical Analysis no distinction between camps, lithic scatter, or isolated finds. Nearest Neighbor Index The cultural periods associated with each archaeological site were tallied Nearest Neighbor Index (NNI) calculates in a spreadsheet. For each of the three the distance between each feature and the major cultural periods (Paleo-Indian, closest neighboring feature (Mitchell, Archaic, Woodland), a tally was recorded 2005). The calculated average minimum if a site had any cultural components distance between all features is compared linking it to a given period. Sites with to an approximation of the expected multiple sub-cultures listed (i.e. early, late, average distance of a random distribution middle), only had one tally recorded to of points (Levine, 2004). The Nearest represent the major period. Neighbor Index is the ratio of observed For sites with cultural periods distribution over the expected distribution. listed as “probable” or with other The ratio provides a generalized questionable qualifiers, the period was distribution pattern of the points within a recorded as definitive. A tally was also given study area. generated for sites from which no cultural A clustered distribution produces a components (i.e. “undetermined ratio of less than one whereas a ratio prehistoric”, “prehistoric”, etc) were greater than one is reflective of a dispersed found. distribution. A ratio value equal to zero is Based on cultural components a perfect random distribution (Wheatley found at each site, three data sets were and Gillings, 2002). developed, one for Paleo-Indian affiliated For a nearest neighbor analysis, a sites, one for Archaic affiliated sites, and Z-score value is calculated by dividing the one for Woodland affiliated sites. difference between the observed and the expected values, with the standard error The Popularity Index (Mitchell, 2005). Positive Z-scores are

5 reflective of dispersed distributions and a confidence envelope. negative values correspond to clustering. Since one random distribution can Statistical significance exists if the vary significantly from the next, multiple absolute Z score value is greater than 1.96 permutations, or versions, are generated. (Mitchell, 2005). The confidence interval values are based Nearest Neighbor Index values on the lowest and highest values from all were calculated for all four sample data permutations for each band (Mitchell, sets using the Average Nearest Neighbor 2005). If observed values are greater than tool. Euclidean distance was chosen for the upper confidence values, the observed the distance calculation method. pattern is statistically significant. Values Since the variations of the total lower than the lower confidence limit extent of the point features can affect the indicates a significant pattern as well. results of the Z score (Mitchell, 2005), the K-function values were calculated entire study area (61,422,251.8568 square for all data sets using the Multi-Distance meters) was consistent for all Nearest Spatial Analysis (Ripley’s K Function) Neighbor Index calculations. tool. Distance intervals were set at 300 m for a total of 20 bands. The band interval Ripley’s K Function was calculated by using half of the greatest x/y dimension distance (12,242 m) of the Ripley’s K function is similar to Nearest study area divided by the number of bands Neighbor Index except it evaluates the (20) (Mitchell, 2005). 99 permutations of dispersion pattern at multiple distances. random points were generated for a 99% Rings, or bands, are drawn at a fixed confidence envelope. interval, around each point. A value (K) is Since features near the study area calculated based on the number of features edges may skew the results due to a or points within each distance band reduction in possible neighboring points (Mitchell, 2005). The process is repeated (Mitchell, 2005), the edge correction for each point in the data set. option “Simulate over boundary values” The observed results are compared was employed in the Ripley K Function to an expected value of a random tool. This correction method “mirrors” distribution of points, based on similar points across the study area boundary to band number and interval (Levine, 2004). reduce edge effects. The study area The K Function describes the distance at parameter was set to the study area which the dispersion pattern exists. polygon. If the observed K values are The output K values were greater than the expected K values, the transformed by the ArcGIS tool. The observed pattern is considered clustered. transformation value, L(d), minimizes the Conversely, if K-observed is less than K- height of the vertical axis (Mitchell, 2005). expected, the pattern is dispersed. As a result of the transformation equation, Upper and lower confidence levels the expected distribution value equals the are used to determine the statistical distance interval, producing a line with a significance of the observed values positive slope of 1 (Mitchell, 2005). (Mitchell, 2005). The confidence level values are generated by calculating K Density/Hot Spot Analysis values for a random distribution of points. Each Ripley’s K Function analysis defines Archaeological hot spots could be

6 considered as areas in which many sites analysis were used to create raster are present (i.e. density), or as the areas surfaces, through the Interpolate to Raster where sites were used most often (i.e. tool, using the Inverse Distance Weight “popular sites”). Both of these possibilities (IDW) Method. were explored in this study. The kernel density function GIS Data and Manipulations calculates the magnitude of point features per unit area based on an output raster Topography (ESRI, 2010). The population parameter within the Kernel density tool serves as a Four 10 m DEMs covering the study area weight value to indicate the number of were obtained from the Natural Resources times to count an individual point. If a Conservation Service (NRCS) Data weight field is not specified, all points are Gateway Website. These were combined considered equally. into one DEM using the Mosaic tool. Two kernel density surfaces were Slope (degree) and aspect raster surfaces created: one based on site density, the were created using the Spatial Analyst other using the popularity index as a toolbar. weight. For each kernel density analysis, the surface search radius was set to 300 m Distance Surfaces and the output was set to a 10 m cell size. The resulting density values from each Distances to numerous resources were raster were extracted at each point using evaluated. All distance surfaces were the Extract Values to Points tool. created using the spatial analyst extension Next, two hot spot analyses were in ArcGIS. Cell values were set to 10 m conducted. The first used site density and Euclidean distances were calculated. values and the other used the popularity Each of the 5 distance surfaces is briefly density values. described below. The Hot Spot Analysis (Getis-Ord Gi*) tool calculates a Z score value for Distance from Water Sources each point feature by subtracting an expected Gi* value of a random Water is a very important resource, across distribution from the calculated Gi* all cultures and time. Shapefiles of the values. The difference is then divided by Kickapoo River and streams within the the square root of the variance (Mitchell, study area were provided by the Kickapoo 2005). If features in close proximity to Valley Reserve. A distance surface from each other have similar high values, the Z the Kickapoo River was created. An score value will be high, identifying a “hot additional distance surface was created spot”. “Cold spots” are reflective of based on distance from any streams, negative Z score values, showing including the Kickapoo River. concentrations of low values. Z score values near zero show no similarity in Distance from Confluences values. Statistically significant values at the 95% confidence level are greater than Archaeological sites are often or less than positive or negative 1.96, concentrated near stream confluences respectively (Mitchell, 2005). (Howard, Brown, Carey, Challis, Cooper, Z score results from each Hot Spot Kincey, and Toms, 2008). The joining of

7 two streams creates a unique environment, open site location by using the Extract possibly providing unique food resources. Values to Points tool. All extracted data Confluences with the Kickapoo River and were exported from ArcGIS and imported the seven major streams were digitized as into Excel. The results from the stream point features. A distance surface was preference spatial join were also exported created based on these points. from ArcGIS and then imported into Excel. Distance from Rock Shelters Statistical Analysis Evidence from rock shelters within UKVAD has shown prehistoric use. Open Descriptive statistics were calculated in site locations may have been influenced by Microsoft Excel for the site selection proximity to rock shelters as possible variables for all four data sets. Histograms refuge from storms or other threats. A were created for each site selection distance surface was generated from rock variable. shelters within the study area. Many of the cultural affiliated sites were shared among the three cultures. Distance to Cliffs/Steep Areas Statistical comparisons of five variables across the cultural sites were conducted The numerous cliffs and overlooks in the using the Kruskal-Wallis test. This UKVAD may have been utilized by past statistical test is a non parametric analysis cultures for hunting, scouting, and of variance able to compare multiple data possibly for wind and sun protection. To sets of different population sizes (Zar, evaluate these areas as a factor for site 1999). The Kruskal-Wallis analysis was selection, areas with a slope greater than conducted using the statistical program 20 degrees were isolated from the slope STATA (Carle, 2010). The variables raster using the reclassification tool of the selected for cultural comparison were Spatial Analyst toolbar. Distances were slope, aspect, elevation, distance from calculated from these areas. streams, and distance from confluences.

Stream Preference Analysis Results

Site selection may be a factor of stream Based on the tally of cultural components size. Larger streams may offer a greater found at the 405 open sites, at least one abundance of food resources; yet a small cultural period was affiliated with 23% stream valley may provide more (n=92 sites); yet 77% (n=319) of the sites protection. did not produce any cultural affiliated To assess a preference of stream artifacts. size for site selection, a spatial join was Many sites had multiple cultural performed based on the proximity of the components identified at a given site open sites to the nearest stream. (Table 1), although some sites were unique to a given period. Raster Surface Data Extraction The Paleo-Indian culture was identified at 12 sites, with only one site Values from topographic rasters and being unique to this period. Archaic distance surfaces were extracted at each cultural artifacts were found at 67 sites,

8 with 36 sites of these unique to this period. distance, the site distribution pattern shifts The Woodland culture was associated with towards more dispersion. None of the data 53 sites; of which 25 were unique to the sets showed statistically significant levels Woodland period. Cultural affiliation from of dispersion. all three periods was found at seven sites.

Table 1. The total number of sites representing a 6 given cultural period is shown as bold. The table 5 also shows number of sites shared between cultures. 4 Paleo Archaic Woodland 3 Paleo 12 10 8 3 Archaic 10 67 27

L (d) 10L 2 Woodland 8 27 53 1 Nearest Neighbor Index Results 0

The results of Near Neighbor Index (NNI) 0 1 2 3 4 5 6 calculations for each data set are shown in Distance 103 (m ) Table 2. All data sets except for the Paleo- Figure 1. Ripley’s K results for open sites (orange line). Expected values are shown as a dashed line; Indian sites showed significant levels of upper and lower confidence limits are shown as clustering. The Paleo-Indian sites had the black solid lines. closest NNI ratio value to zero, an indication of a more random dispersion. 6 The NNI ratios for the Archaic and Woodland period sites were similar. 5

Table 2. Results from Nearest Neighbor Index, 3 showing Observed values (Obsv), Expected values (Exp), Z score, and p-value (p). 3

Paleo Arch Wood Open (d) 10 L 2 n 12 66 53 403 Obsv. 884.6 269.3 287.1 127.2 1 Exp. 1131.2 482.3 538.3 194.7 0 Ratio 0.78 0.56 0.53 0.653 0 1 2 3 4 5 6 Z -1.44 -6.86 -6.50 -13.35 Distance 103 (m) p 0.159 0.000 0.000 0.000 Figure 2. Ripley’s K results for Paleo-Indian sites (blue line). Expected values are shown as a dashed Ripley’s K Function Results line; upper and lower confidence limits are shown as black solid lines. The graphical results for the Ripley K Function analysis for each data set are With respect to the confidence shown in Figures 1, 2, 3, and 4. All four envelopes, only the Paleo-Indian observed sample sets show clustering to exist up to values plotted completely within the upper a range of 2400 to 2700 meters as and lower limits, indicating no statistical significance in the results. indicated where the observed values intersect the expected values. Beyond this Popularity Index Results

Table 3 shows the percent distribution of Z 6 score values for the two hot spot analyses. 5 Approximately 6% of the points produced statistically significant Z score values. 4 3 Maximum values were 5.1 and 4.2 for the 3 density and popularity index analyses,

respectively. L (d) 10 L 2 Table 3. The Z score Hot Spot results as percent of 1 points (n=405) based on the density of sites and based on popularity index. 0 Z Score Range Density PopIndex 0 1 2 3 4 5 6 < 1 83.5% 81.7% Distance 103 (m) Figure 3. Ripley’s K results for Archaic sites (red 1 – 1.96 10.1% 12.3% line). Expected values are shown as a dashed line; >1.96 6.4% 5.9% upper and lower confidence limits are shown as black solid lines. Z score values near zero indicate no concentration of either high or low 6 values (Mitchell, 2005). Thirty percent of the density based hot spot analysis had Z 5 score values between -0.5 and +0.5. The 4 popularity index analysis had 36% point 3 values within this range. 3 Negative values did not exceed the

L (d) 10 L 95 % significance level for either analysis. 2 The minimum values were -1.20 for the 1 density based analysis and -1.36 for the popularity index. 0 Graphical results for the Hot Spot 0 1 2 3 4 5 6 Analyses are shown in Figure 5 and Figure Distance 103 (m) 6. Areas in red are statistically significant Figure 4. Ripley’s K results for Woodland sites (Z score > 1.96), showing high density of (green line). Expected values are shown as a dashed line; upper and lower confidence limits are sites and high popularity index values. shown as black solid lines. Yellow areas are Z score values greater than 1, but less than 1.96. These reflect Of the 405 open sites, twelve percent of areas of high use or density, but the results the sites (n=53) were associated with only are not statistically significant. Z score one cultural sub-period and seven percent values less than 1 are shown in green, (n=27) were affiliated with two sub- reflecting low site densities and/or less periods. Three cultural sub periods were popularity across cultures. identified at 1.2 % (n=5) sites. The percent The density based hot spot analysis of sites with greater than three sub-periods (Figure 5) shows two statistically was 3.5% (n=14), with a maximum of 7 significant areas. The analysis based on cultural sub-periods represented at one the popularity index (Figure 6) shows site. three areas of significance. Both analyses show hot spots occurring near stream Hot Spot Analysis Results confluences.

The locations of yellow areas show between 250 m to 270 m. Less than 1% of some association with confluences, with the sites were located above 330 m. most being located around the statistically significant areas (red). However, a few of these areas appear not to be associated with the major stream confluences.

Stream Size Preference Results

The open sites were assessed for proximity to a particular stream size. Forty-six percent (n=187) of the sites were closer to the Kickapoo River than any other stream. Low flow and/or intermittent streams were the second most popular with 32% of sites (n=128) in closest proximity. Twenty-two percent (n= 90) of sites were located along one of 6 of the Figure 5. Hot spot analysis based on density. Red major streams. The values for these areas are statistically significant (>1.96). Yellow streams were as follows; 8% for Weister areas are greater than 1 but not significant. Green Creek, Warner Creek 5%, Billings Creek shows Z score values less than 1. Black triangles 4%, Indian Creek 3% and the Jug Creek are stream confluences. Scale is 1:160,000. Valley with 2 %. Plum Run, a stream on the south side of the study area had less than 1% (n=3) of the sites in closest proximity. No sites were located closer to Hay Creek, the northern most tributary, than other streams.

Site Selection Variable Results

Eight variables were selected as possible variables for influencing site selection. All were plotted in histograms to visualize the distribution. Most distributions were non- parametric and skewed towards the left. The distribution for aspect was the most uniform.

Elevation Figure 6. Hot spot analysis based on popularity index. Red areas are statistically significant The distribution of sites based on elevation (>1.96). Yellow areas are greater than 1 but not significant. Green shows Z score values less than 1. is shown in Figure 7. The values range Black triangles are stream confluences. Scale is was from 243 m to 364 m. The average 1:160,000. elevation was 269 m. The median value was 265 m, with 53% of the sites located Aspect

Figure 8 shows the distribution of sites Distance to Streams based on aspect. Fifty-eight percent of the sites were located between 1350 and 3150, The distribution of sites based on distance showing a preference of site locations on from any stream is shown in Figure 11. south and west facing slopes. Thirty Values ranged from 0 m to 557 m. The percent of the sites were on south- average value was 128 m and the median southeast and south-southwest facing was 101 m. Ten percent of the sites were slopes (1350-2250). The northeast direction located within 25 meters of any stream (00-450) consisted of 17% of the sites. within the study area.

30 Elevation Aspect 25 20

20 15

15 10

10 5

PercentSitesof 5

250 260 270 280 290 300 310 320 330 340 350 360 Elevation (m) Figure 7. Percent distribution of open site locations Figure 8. Percent distribution of open site locations based on elevation. based on aspect.

Slope Slope 30 The slope value distribution is shown in 25 Figure 9. The average slope and median values were 7.7 degrees and 6 degrees, 20 respectively. The range of values was from 15 0 to 33 degrees. 85% of the sites were

located on slopes less than 15 degrees. 10 PercentSitesof 5 Distance to the Kickapoo River 0 Site distances from the Kickapoo River 3 6 9 12 15 18 21 24 27 30 33 36 Slope (degree) ranged from 0 m to 3387 m. The average distance was 443 m and the median value Figure 9. Percent distribution of open site locations was 206 m. Twelve percent of the sites based on slope. were within 50 m of the Kickapoo River and 11% of sites were located 1200 m or Distance to Confluences greater distances from the river. The distribution of open sites based on distance Figure 12 shows the distribution of sites from the Kickapoo River is shown in based on distance from a stream Figure 10. confluence. The range of values was 81 m

12 to 3657 m. The average value was 907 m 25 Distance to Confluences and the median was 706 m. Four percent of the sites were within 200 meters of a 20 confluence within the study area. 15 20 Distance from 10 Kickapoo River 15 PercentSitesof 5

200 600

1000 1400 1800 2200 2600 3000 3400 3800

PercentSitesof 5 Distance (m) Figure 12. Percent distribution of open site locations based on distance from a major stream 0

confluence with the Kickapoo River.

675 975

375

3375 1275 1575 1875 2175 2475 2775 3075 20 Distance from Distance (m) Figure 10. Percent distribution of open site Rock Shelters locations based on distance from the Kickapoo 15 River.

10 18 Distance from Streams 16 14 PercentSitesof 5 12

10 0

250 650 850

8 450

1050 1250 1450 1650 1850 2050 2250 6 Distance (m)

PercentSitesof 4 Figure 13. Percent distribution of open site 2 locations based on distance from a rock shelter. 0

Distance from Cliffs/Steep Areas

200 300 400 500 100 Distance (m) Figure 11. Percent distribution of open site Figure 14 shows the distribution of sites locations based on distance from any stream. from areas of slopes exceeding 20 degrees. The values ranged from 0 m to 648 m, Distance from Rock Shelters with the average value of 120 m. The median value was 95 m. Nineteen percent The range of distance values of site of the sites were within 25 m of a steep locations from a rock shelter was 0 m to area. 2744 m. The average value was 437 m and the median was 262 m. Twenty-five Comparisons across Cultures percent of the sites were within 100 meters of any rock shelter. The distribution is The results for the Kruskal-Wallis test are shown in Figure 13. shown in Table 4. The results showed that

13 the variation among the distances to cultured affiliated sites showed confluences and the distances to streams statistically significant clustering. Both was significant (p<0.05). The difference data sets showed clustering to occur up to among the cultures was more pronounced 2700 meters. Beyond this distance, the for distance from confluences than the sites became more randomly distributed. distance from streams. The values among The Paleo-Indian sites did not the other tested variables did not show any show significant levels of clustering. This significant difference. was likely the result of a limited number of sites (n=12) with this cultural affiliation. According to the output from 20 Distance from Cliffs 18 the Ripley’s K-Function tool, data sets 16 should contain greater than 30 sample 14 points for the analysis (ESRI, 2010). 12 10 Hot Spot Analysis 8 6 The hot spot analyses results showed a 4 PercentSitesof clustering near stream confluences 2 (Figures 5 and 6). The analysis based only 0

on density revealed two confluences as hot

25 75

375 175 225 275 325 425 475 525 125 spots. Yet for the popularity index Distance (m) weighted analysis, three hot spot areas Figure 14. Percent distribution of open site were revealed. Between the two analyses, locations based on distance from a steep area or one stream confluence was a common hot cliff. spot between the two analyses. Both hot

spot analyses suggest the importance of Table 4. P value results from Kruskal-Wallis test for five selected variables are shown. Significant stream confluences for site selection. (<0.05) values are shown as bold. Areas of high use and density Aspect 0.9134 (yellow) were mostly located near Elevation 0.9596 confluences as well. Generally, these are Slope 0.2552 “buffers” around the statistically Confluence 0.0001 significant areas. This decrease in density Streams 0.0217 and popularity may be a result of an increase in distance from the confluences. Discussion Yellow areas not associated with stream confluences may reflect proximity One of the goals of this study was to to other resources not present in the investigate the distribution of current landscape. These could include archaeological sites within UKVAD. The former ponds or small lakes, or a overall distribution of open archaeological vegetation food sources. sites within the study area showed a pattern of clustering based on the Nearest Stream Size Preference Neighbor Index (NNI) and Ripley’s K- Function analysis. The preference of sites closer to the NNI and Ripley’s K-Function Kickapoo River than other sites was not analysis for the Archaic and Woodland surprising since it is the largest water

14 resource in the study area. The higher within 10 m of an area exceeding 20 percentage along the small intermittent degrees. streams may be a factor of the abundance of these streams rather than an actual Aspect preference. Alternatively, the preference to these smaller streams may reflect the small The distribution of aspect showed a valley size, offering more protection. preference of sites located on south and Plum Run had the fewest number west facing slopes. This was likely due to of sites in close proximity. This stream the greater amount of sun exposure on was located at the study area’s southern these slopes. Such slopes would have been most edge. These results were most likely ideal during colder climates and seasons. a factor of edge effects rather than lack of However, the presence of sites located on resources along this stream compared to northern facing slopes further suggests other streams. prehistoric use of the area through all seasons. Site Selection Variables Distance from Water Resources Elevation Clustering of sites along streams is The elevation distribution of expected (Fletcher, 2008). This was archaeological sites revealed a slight illustrated by the Hot Spot analyses preference to elevations between 250 and (Figures 5 and 6) occurring near 270 m. Comparing range of elevation confluences and through the histograms along the valley (240 m -253 m) with the based on water resources (Figures 10, 11 average (269 m) and median values (265 and 12). 90% of sites were located within m) indicated that most sites were slightly 250 m any stream. above the floodplain, but below the upland Although only a low percentage of terrace level (275 m), as described by sites (4%) were located within 200 m of a Holtz and Boszhardt (1998). These values stream confluence, the hot spot analysis probably reflect that lower elevations were shows the importance of confluences in closer to water sources. site selection. The dynamic nature and high sedimentation rates of confluences Slope (Best 1988; Bradbrook, Lane, Richards, Biron, and Roy, 2001; Howard et al., The majority of the sites were located on 2008) may have destroyed or buried many slopes less than 15 degrees, as suggested of the sites, thus further limiting by Holtz and Boszhardt (1998). archaeological information. Surprisingly 14% of the sites yielded slope In addition, the importance of the values greater than 15 degrees, with a particular four confluences with the higher maximum value of 34 degrees. This upper density and popularity is unclear. range of values was most likely associated Compared to the others, these confluences with the accuracy of the 10 m DEM. Sites may have had unique resources, such as a located on slopes 20 degrees might occur rich vegetation food source or an on small, generally flat areas near, above, abundance of game migration paths. or below a cliff or steep slope. In support of this, 10% of the sites were located Distance from Rock Shelters and Steep

Areas from streams varied more among the three cultural affiliated data set; Paleo-Indian The histograms show site locations were 221 m, Archaic 117 m and Woodland 85 close to rock shelters and steep areas. m. Again, this confirmed that the Paleo- These results may have been influenced by Indian sites were further away from the distribution of these resources rather streams when compared to the other than a site selection process. A Nearest cultures. Yet, the median values for the Neighbor Index evaluation was conducted Archaic sites and the Woodland site did on the 91 rock shelters within the study not show much similarity. area. The results showed a clustered These differences based on water distribution (p=0.0000). Visual inspection resources for the culture sites could be a revealed that many rock shelters were result of changes in stream channel close to streams. Therefore, the close locations over the years. For example, the proximity between rock shelters and open Paleo-Indian sites may have been further sites may not be based on site selection but from streams and confluences, based on rather due to the close proximity of rock current stream locations, but were shelter sites to water sources. significantly closer during the time of site The occurrence of sites close to occupation. Small oxbow lakes and steep areas may result from their meander scars along the current Kickapoo abundance and wide distribution. Cliff River and the larger tributaries indicated faces occur along the river channel as well stream channel migration. These processes as just below the ridge line. Furthermore, undoubtedly have occurred throughout the minimum value (20 degrees) for steep time. areas defined in this study may have been A second possible explanation was to low. Therefore, the steep areas may the stream flow or discharge volume have been too broad or widely distributed. during each of the periods could have A narrower range of slope may have better varied significantly. The climatic record isolated cliffs as a resource. indicates Paleo-Indian period was relatively cooler, having occurred shortly Comparing the Cultured Sites after the Pleistocene Ice Age. The Archaic period was a warm and dry period known The results from the Kruskal-Wallis as the Altithermal (Theler and Boszhardt, analysis indicated that the values based on 2003). Both of these climate trends would distance from confluences and the distance have influenced stream flow and possibly from streams were significantly different channel width. among the cultural affiliated sites. Despite these possible reasons, the Comparing the median values, the Paleo- paucity of Paleo-Indian sites did not Indian sites had a greater median value provide an extensive amount of (3047 m) than either the Archaic (825 m) information to make any definitive or Woodland sites (853 m). This indicated conclusions. the Paleo-Indian sites were further away from confluences than the other two. The Future Studies median values for the Archaic and Woodland sites were similar. This study explored the distribution of However, the median values open sites within a portion of the Upper among the cultured site based on distance Kickapoo Valley Archaeological District,

16 mainly within the Kickapoo Valley abundance and/or the distribution of these Reserve. The study considered each site as resources, rather than an intentional a habitation site and that each site was selection process. chosen by prehistoric people. Furthermore, Comparisons of variables based on the artifacts found at a given site were cultural affiliation show similarity assumed to indicate the site was used, at between the topographic variables. minimum, for a short duration of time (i.e. Distances to water resources revealed one night). some differences based on culture. Future studies should consider using a wider data set extent of the Acknowledgments archaeological sites in the UKVAD. The irregular boundary of the study area may Dr. Connie Arzigian at Mississippi Valley have had an effect on the distribution Archaeology Center (MVAC) at results. University of Wisconsin - La Crosse A more detailed investigation of assisted with project goals and ideas as components found at each site could well as access to archaeology documents. isolate the sites actually used for Also, I thank Dr. Jim Theler at MVAC for habitation. Fire cracked rock is a key considering the project as a “good idea”. I feature suggesting habitation uses (Holtz thank the Wisconsin State Historical and Boszhardt, 1998). Also, statistical Society for permission to use the studies may consider the similarity and archaeological data. Betsy Finlay at the differences among the sites unique to the Historical Society processed the Archaic and Woodland periods, as well as permission request. Dr. Adam Carle at the the sites shared between these periods. Cincinnati Children’s Hospital Medical This study used 20 degrees as the Center in Cincinnati, OH gave valuable minimum value for a steep area. The discussion and assistance with statistics. I resulting areas could have been too broad thank Marcy West and the staff at the and not specific enough to capture the Kickapoo Valley Reserve for GIS data, actual resource of a cliff face. Further archaeology documents, and moral studies could re-evaluate this resource support. Former and current SMU with a narrower range. students, Chip Brown and April Ammann (respectively), reviewed the document and Conclusions offered comments. SMU faculty and fellow students furnished inspiration, This project identified the distribution of education, guidance, and laughter. And I open archaeological sites in UKVAD as thank Gina Rae, of the restaurant formerly having a clustered distribution pattern. known as Gina’s Pies Are Square in Based on physical environmental factors, Wilton, WI, for her patience, trends were shown to favor lower understanding, and support through elevations and gradual slopes. The site another degree. locations also showed a preference for south-southwest facing slopes. References Analysis based on proximity to resources concluded that the sites were Best, J. 1988. Sediment Transport and Bed closer to water resources. The proximity to Morphology at River Channel some resources may be a factor of the Confluences. Sedimentology, 35, 481-98.

Bradbrook, K.E., Lane, S.N., Richards, Center, University of Wisconsin-La K.S., Biron, P.M., and Roy, A.G. 2001. Crosse. Role of Bed Discordance at Howard, A.J, Brown, A.G., Carey, C.J., Asymmetrical River Confluences. Challis, K., Cooper, L.P, Kincey, M., and Journal of Hydraulic Engineering, 127, Toms, P. 2008. Archaeological Resource 351-68. Modelling in Temperate River Valleys: a Carle, A. 2010. Personal Communication. Case Study from the Trent Valley, UK. Cincinnati Children’s Hospital Medical Antiquity, 82, 1040-1054. Retrieved on Center. Cincinnati, OH. March 1, 2010 from EBSCOhost Carlisle, K., and Friedel, R. 2009. database. Reconstructing the Past: A GIS Johnson, W.C. 1976. The Impact of Settlement Pattern Analysis for Environmental Change on the Fluvial Northwest North America. ArcNews Systems: Kickapoo River, Wisconsin. Online, Spring 2009. Retrieved on Ph.D. Dissertation, University of February 16th, 2010 from www.esri.com. Wisconsin, Madison. Ebert, D. 2004. Applications of Levine, N. 2004. CrimeStat: A Spatial Archaeological GIS. Canadian Journal of Statistics Program for the Analysis of Archaeology, 28, 319-314. Retrieved on Crime Incident Locations Manual. Ned January 12, 2010 from EBSCOhost Levine & Associates, Houston, TX and database. National Institute of Justice, Washington, ESRI. 2010. ArcGIS 9.3 Desktop Help. D.C. Retrieved from www.icpsr.umich. Environmental Systems Research edu/icpsrweb/CRIMESTAT/ Institute, Redlands, CA. http://webhelp Mitchell, A. 2005. The ESRI Guide to GIS .esri.com/arcgisdesktop/9.3/index.cfm?id Analysis, Volume 2: Spatial =6190&pid=6188&topicname=Kernel_ Measurements and Statistics. ESRI Density Press, Redlands, CA. 238 pp. Fletcher, R. 2008. Some Spatial Analyses Theler, J. and Boszhardt, R.F. 2003. of Chalcolithic Settlement in Southern Twelve Millennia: Archaeology of the Israel. Journal of Archaeological Upper Mississippi River Valley. Sciences, 35, 2048-2058. Retrieved on University of Iowa Press, Iowa City, IA. February 3rd, 2010 from www.science 254 pp. direct.com. Wheatley, D. and Gillings, M. 2002. Guccione, M.J. 2008. Impact of the Spatial Technology and Archaeology: Alluvial Style on the Geoarcheology of The Archaeological Applications of GIS. Stream Valleys. Geomorphology, 101, Taylor & Francis, Inc. New York, NY 378-401. Retrieved on February 3rd, 2010 269 pp. from www.sciencedirect.com. Zar, J.H. 1999. Biostatistical Analysis 4th Hatfield, C. 2009. Personal Edition. Prentice Hall Inc., Upper Saddle Communication, Kickapoo Valley River, NJ. Reserve. La Farge, WI. Zhang, H.Q, Zhao, W.M., and Liu, B. Holtz, W. K., and Boszhardt, R. F. 1998. 2007. Mathematical Modeling of the Final Inventory of Pre-European Relationships between Neolitic Sites and Archaeological Sites in the La Farge the Rivers in Xi’an (Shaanxi Province, Lake Project Vernon County, Wisconsin. China. Archaeometry, 49, 765-773. Reports of Investigations No. 319, Retrieved on January 22, 2010 from Mississippi Valley Archaeological EBSCOhost database.