An Investigation into the Potential for Geochemical / Geoarchaeological Provenance of Prairie du Chien Cherts
A THESIS SUBMITTED TO THE FACULTY OF
THE GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA
BY
Brian Klawiter
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF MASTER OF SCIENCE
May 2000
Ó
Brian Neil Klawiter 2000
"Keats was quite wrong when he asked, rhetorically, ‘Do not all charms fly... at the mere touch of cold philosophy?’ There is more charm in one cold simple 'mere' fact, confirmed by observation and linked to other facts through coherent theory into a rational system, than in a whole brainful of fancy and fantasy; more poetry in a chunk of quartzite than in a make-believe wood nymph; more beauty in the revelations of a verifiable intellectual construction than in misty empires of mythology."
-Edward Abbey
Abstract
The use of trace-element geochemical analyses to trace archaeological materials to their geologic sources (provenance) has grown immensely diverse in the past few decades. In the upper Midwest, much attention has been focused on distinguishing between various types of chert with annoyingly variable and overlapping visual characteristics (e.g., Luedtke 1976; Hoard et al. 1993; Spielbauer 1984). Instrumental Neutron Activation Analysis (INAA) has been a very effective analytical method for provenance studies of cherts. However, comparatively little effort has been made to geochemically distinguish between chert sources within a single geologic formation. Some preliminary studies conducted by Luedtke (1978) and Luedtke and Meyers (1984), however, showed that this should be a viable avenue of investigation. This paper discusses an investigation into the potential for geochemically distinguishing among separate chert sources within the Prairie du Chien dolomite. To accomplish this, multivariate statistical techniques such as Principle Component Analysis and Discriminant Analysis were applied to a geochemical data set derived by Neutron Activation Analysis of samples collected from 20 locations throughout southeast Minnesota and southwest Wisconsin. The results of the statistical analyses are effective at distinguishing among widely separated sources, but there is a great deal of overlap among geographically adjacent sources. A broader overall geochemical pattern is distinguishable in regional trends of element concentrations; but once again the data show a great amount of overlap, making precise provenance determination difficult.
Table of Contents
INTRODUCTION AND BACKGROUND
Methods of Provenance Neutron Activation Analysis Provenance of Chert Visual Classification Petrogenesis and Geochemistry of Chert Prairie du Chien Group
FIELDWORK
Geologic Sampling
DATA ANALYSIS
Normality of the Data Set Stratigraphic Variations Outcrop vs. Stream Deposits
Principle Component Analysis
Element Selection Element Correlations Plotting Principle Components
Stepwise Discriminant Analysis
More Principle Component Analysis
POSSIBLE CONTROLS ON ELEMENT TRENDS
APPENDIX A: NEUTRON ACTIVATION DATA
APPENDIX B: HAND-SAMPLE DESCRIPTIONS
APPENDIX C: THIN-SECTION DESCRIPTIONS
List of Figures
Figure 1: Simplified stratigraphic column
Figure 2: Approximate outcrop extent of the Prairie du Chien
Figure 3: Sample locations
Figure 4: Models of trace element variation
Figure 5: Examples of stratigraphic trace element plots
Figure 6a: Box-and-whisker plots comparing outcrop and stream deposits
Figure 6b: Box-and-whisker plots comparing outcrop and stream deposits
Figure 7: Plot of PC1 versus PC2
Figure 8: Sample location groups
Figure 9: Plot of DIM1 versus DIM2
Figure 10: Plots of PC scores versus northings
Figure 11: Plots of PC scores versus eastings
Figure 12: Plots of PC scores versus southeastings
Figure 13: Location of the Vosberg Site
Figure 14: PC scores of archaeological samples
List of Tables
Table 1: Elements analyzed by INAA
Table 2: Simple statistics of the data set
Table 3: Correlation of PCA eigenvectors with elements
Table 4: Correlation matrix between elements
Table 5: Correlation of eigenvectors and elements
Table 6: Eigenvalues of the correlation matrix
Table 7: Results of SDA for sample locations
Table 8: Results of SDA for sample groups
Table 9: Cross-validation summary from DA of sample groups
Table 10: Cross-validation summary from DA of locations 5 and 16
Table 11: Cross-validation summary from DA of locations 17 and 19
Acknowledgements
Thanks go to the faculty at the University of Minnesota, Duluth, Department of Geological Sciences. The neutron activation analyses were funded by the DOE grant DE-FG02-95NE 38143 Mod03 and were conducted at the University of Wisconsin, Madison, Reactor Lab under the direction of Richard Cashwell, with the help of Andy Smolinski. Other essential aid to the project came in the form of advice from Rip Rapp (who also provided field work support), Ron Regal, Eric Brown, and Dick Ojakangas. The latter also provided much-needed maps. More maps and invaluable outcrop-hunting advice were provided by John Mossler at the Minnesota State Geological Survey. Archaeological chert samples were thankfully provided by Kent Bakken and Steve Mulholland. Jennifer Kolb, Marlin Hawley, and Leroy Gonsior were very helpful in pointing the way to some archaeological "quarry" sites in Minnesota and Wisconsin, near which geologic samples were collected. I thank God for providing all the materials for this study, as well as beautiful days and wonderful scenery to make collecting the samples so enjoyable. Thanks also belong to my parents for their support; Bob Pearson for his help with the initial SAS analyses; Darcy Hanson for showing me around the Shakopee quarry; and Colleen Wergin for just being such a dang nice person every day.