Page 1

Remote Sensing in Archaeology

Remote Sensing in Archaeology An Explicitly North American Perspective

Edited by Jay K. Johnson

The University of Alabama Press, Tuscaloosa Published for The Center for Archaeological Research at the University of Mississippi, the University of Mississippi Geoinformatics Center, and NASA Earth Science Applications Directorate at the Stennis Space Center

Copyright © 2006 The University of Alabama Press Tuscaloosa, Alabama 35487-0380 All rights reserved Manufactured in the United States of America ∞ The paper on which this book is printed meets the minimum requirements of American National Standard for Information Science—Permanence of Paper for Printed Library Materials, ANSI Z39.48–1984.

Typefaces: Garamond and Myriad Designer: Kathy Cummins

Library of Congress Cataloging-in-Publication Data Remote sensing in archaeology : an explicitly North American perspective / edited by Jay K. Johnson. p. cm. Based on presentations made at a workshop held in Biloxi, Miss., in 2002, preceding the annual meeting of the Southeastern Archaeological Conference. “Published for the Center for Archaeological Research at the Universtiy of Mississippi, the University of Mississippi Geoinformatics Center, and NASA Earth Sciences Application Directorate at the Stennis Space Center.” Includes bibliographical references. ISBN-13: 978-0-8173-5343-8 (alk. paper) ISBN-10: 0-8173-5343-7 (alk. paper) 1. Archaeology--Remote sensing. 2. Archaeology--North America--Remote sensing. 3. Indians of North America--Antiquities--Remote sensing. 4. Excavations (Archaeology)--North America. 5. North America--Antiquities--Remote sensing. I. Johnson, Jay K. II. University of Mississippi. Center for Archaeological Research. CC76.4.R46 2006 930.1028--dc22 2005054863

For Anne

Contents List of Figures List of Tables Acknowledgments 1. Introduction Jay K. Johnson 2. The Current and Potential Role of Archaeogeophysics in Cultural Resource Management in the United States J. J. Lockhart and Thomas J. Green 3. A Cost-BeneďŹ t Analysis of Remote Sensing Application in Cultural Resource Management Archaeology Jay K. Johnson and Bryan S. Haley 4. Airborne Remote Sensing and Geospatial Analysis Marco Giardino and Bryan S. Haley 5. Conductivity Survey: A Survival Manual R. Berle Clay 6. Resistivity Survey Lewis Somers

ix xv xvii 1


33 47 79 109

7. Ground-Penetrating Radar Lawrence B. Conyers


8. Magnetic Susceptibility Rinita A. Dalan


9. Magnetometry: Nature’s Gift to Archaeology Kenneth L. Kvamme


10. Data Processing and Presentation Kenneth L. Kvamme


11. Multiple Methods Surveys: Case Studies Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley


12. Ground Truthing the Results of Geophysical Surveys Michael L. Hargrave


13. A Comparative Guide to Applications Jay K. Johnson


List of Contributors CD Containing Color Figures

321 inside back cover

Figures 2.1. 2.2. 2.3. 2.4. 2.5. 2.6. 2.7. 2.8. 2.9. 2.10. 2.11. 2.12. 3.1. 3.2.

3.3. 3.4. 3.5. 3.6. 3.7. 4.1. 4.2. 4.3. 4.4. 4.5.

GIS data layer and example database fields for archaeological sites in Arkansas GIS data layer and example database fields for archaeological projects and surveys in Arkansas Gradiometer data for a prehistoric feature Gradiometer data for a nineteenth-century cemetery Comparison highlighting the advantages of using multiple technologies Electrical resistance data and excavation on a prehistoric site in eastern Arkansas Gradiometer data and excavation on a prehistoric site in southwest Arkansas Geophysical signatures for an archaeological feature using multiple technologies Field methodology and results from a prehistoric site in northeast Arkansas Geophysical units of measure Electrical resistance data and georeferenced 2-×-2-m grid for a prehistoric site in Arkansas Archaeogeophysical imagery from four technologies with excavated features Magnetic gradiometer survey of the village portion of the Parchman Place site Survey of buried prehistoric house remnants at Parchman Place with electromagnetics, resistance, ground-penetrating radar, and magnetic gradiometer Ground truth excavation units superimposed on magnetic gradiometer survey Trenches superimposed on magnetic gradiometer survey showing burned floor and charred beams Surface artifact density plot of the Hollywood site Magnetic gradiometer survey of the Hollywood site Magnetic gradiometer survey of the village portion of the Parchman Place site showing houses, pits, and high-density areas Electromagnetic spectrum Diurnal temperature variation of a hypothetical Mississippian house A helium blimp in use as a low-cost, low-altitude remote sensing platform A powered parachute in use as a stable remote sensing platform A black-and-white aerial photograph before and after subsetting and contrast enhancing

20 21 22 23 24 25 25 26 27 28 29 30 36

37 38 38 39 40 41 49 54 58 59 62



4.6. 4.7. 4.8. 4.9. 4.10. 4.11. 5.1. 5.2. 5.3. 5.4. 5.5. 5.6. 5.7. 5.8. 5.9. 5.10. 5.11. 5.12. 5.13. 5.14. 5.15. 5.16. 5.17. 6.1. 6.2.

6.3. 6.4. 6.5. 6.6.

A 1923 Calvin Brown sketch map of the Hollywood site Soil Conservation Service photographs of the Hollywood site from 1938, 1942, 1966, and 1992 The near infrared band from the large-format color infrared photography of the Hollywood site The near infrared band 6 of imagery obtained with the ATLAS sensor, Hollywood site The thermal infrared band 10 of imagery obtained with the ATLAS sensor, Hollywood site Thermal infrared imagery produced by the Agema 570 camera aboard a helium blimp, Hollywood site A conductivity survey in progress with the EM38 conductivity meter A 60-×-60-m view of a “classic” ditch and bank with the EM38 A 40-×-60-m view of an archaeological site showing plow scars Earth conductivity data uncorrected for “digital lag” The conductivity data of Figure 5.4 corrected for digital lag Conductivity data collected on zigzag traverses but uncorrected for digital lag The same data as in Figure 5.6 with digital lag corrected by processing in Geoplot 3.0 Magnetic susceptibility survey, Millstone Bluff, Illinois A 20-×-20-m square centered over a country brick kiln showing effects of walking pace on measurement of ppt with EM38 Conductivity data in contour map form The conductivity data of Figure 5.10 in gray-scale form Gray-scale image of conductivity data produced in Geoplot 3.0 Gray-scale image of conductivity data produced in Surfer 8 Conductivity survey at the Hollywood site Conductivity survey at the Carty site Conductivity survey at the Hopeton Earthworks Conductivity survey at the Hopeton Earthworks A resistivity survey system consisting of a probe array, multiplexer, resistivity meter, and data-processing and display unit Vertical section through uniform soil showing current injection electrode and voltage measuring electrode, with associated current flow and electric field A multidepth survey probe configuration showing current flow and electric field The field configuration for a typical resistivity survey A wheeled square array survey system with automated data logging Schematic representation of zero, low, and high signal-to-noise ratios with probability distribution functions also shown

67 68 69 69 70 71 80 85 87 91 91 92 92 94 95 96 98 98 98 103 104 105 105 110

114 114 115 116 121

Figures ~ xi

6.7. 6.8. 6.9. 6.10. 6.11. 6.12. 6.13. 7.1. 7.2. 7.3.

7.4. 7.5.

7.6. 7.7. 7.8. 7.9. 7.10.


7.12. 7.13. 7.14. 7.15. 7.16.

Large-area resistivity survey at Army City Army City, areas of high resistance emphasized Army City, detail Resistivity survey data from Yucca House Resistivity survey and magnetic field gradient survey at Mission San Marcos Resistivity survey of prehistoric coastal California house pits Resistivity survey and magnetic field gradient survey at a Shields Complex site The Geophysical Survey Systems Subsurface Interface Radar (SIR) system, Model 2000 A 400-MHz profile across a pithouse floor; buried water pipes are visible as reflection hyperbolas A 25-MHz antenna, capable of transmitting radar energy to more than 20 m and of resolving only very large objects of many meters in dimension A 900-MHz antenna, which can transmit energy to about 1 m at most but can resolve features to about 10 cm in diameter A GPR survey of a ground surface that is not flat, in which profiles must be corrected for surface elevation changes in order to produce a more accurate two-dimensional view of the subsurface Example of an amplitude slice-map, showing changes in amplitude in plan view, with each slice representing about 20 cm in the ground Large trenches dug with backhoes to determine the presence or absence of archaeological features A 500-MHz profile from the Valencia site in Tucson, Arizona, showing reflection data obscured by noise The profile in Figure 7.8 processed to remove the interfering frequencies, revealing a pithouse floor Amplitude slice-maps produced from the data in Figure 7.9 showing the location of many pithouse floors and other extramural features Amplitude slice-maps from a pithouse site in Utah, revealing a pithouse floor in a different area of the grid than hypothesized from the concentration of the artifacts Reflections from one 500-MHz profile that crossed the pithouse floor visible in the amplitude slice-maps in Figure 7.11 A 500-MHz reflection profile crossing a great kiva Amplitude slice-maps of the great kiva at Bluff, Utah The convent courtyard at San Marcos Pueblo, New Mexico GPR profiles of historic graves with intact or partially collapsed caskets

123 124 125 125 126 127 128 133 137

138 139

141 142 146 147 147


150 151 152 152 153 154

xii ~ Figures 7.17.


8.1. 8.2. 8.3. 8.4. 8.5. 8.6. 8.7. 8.8.

8.9. 8.10. 8.11. 8.12. 8.13. 8.14.


8.16. 8.17. 8.18. 8.19. 8.20. 8.21.

Amplitude slice-map reflections in a pioneer cemetery in Utah showing many distinct graves, whose locations are rarely coincident with the locations of the extant headstones A three-dimensional rendering of the highest amplitudes in the same grid of data used to make the slices in Figure 7.6, imaging rubble from a historic house Magnetic enhancement of soils at the Cahokia Mounds site Unit susceptibility profile from a basal platform joining two mounds at the Cahokia Mounds site An EM38 in operation The Bartington MS2D sensor The prototype down-hole magnetic susceptibility logger in operation Field evaluation of the Bartington MS2H Depth response of the EM38 Profile across the Grand Plaza at the Cahokia Mounds site showing natural sediments and overlying cultural material, with data gained from soil magnetic studies, soil chemical tests, and core descriptions Magnetic susceptibility survey of a prehistoric structure in southwest Arkansas Down-hole magnetic susceptibility results at the Rustad site Down-hole magnetic susceptibility studies at the Canning site Base map showing the earthworks at Hopeton and the locations of three trenches excavated in 2001 and 2002 Magnetic susceptibility contour map of the north face of Trench 3 at the Hopeton Earthworks Magnetic susceptibility values along a single elevation line on the north face of Trench 3 at the Hopeton Earthworks compared with cesium gradiometer data collected at this location Application of magnetic techniques to the identification of areas of stability, erosion, and sedimentation on a Mississippian period platform mound Mound erosion processes for platform and conical mounds Topographic and soil magnetic data for Mound 36, Cahokia Topographic and soil magnetic data for Mound 62, Cahokia A plot of ARM/χ versus distance from Core A (mound summit), Mound 36 Core locations and topographic profile showing the asymmetrical nature of Mound 56, Cahokia A plot of ARM/χ versus distance from the crest of Mound 56 showing a bimodal version of the pattern observed for Mound 36


156 163 165 168 168 169 171 172

174 182 184 185 186 187


189 191 192 193 194 195 196

Figures ~ xiii



9.3. 9.4. 9.5. 9.6. 9.7. 9.8. 9.9. 9.10. 9.11. 9.12. 9.13. 9.14. 10.1. 10.2.


10.4. 10.5. 10.6.

Principal magnetometers used in archaeology: Geometrics-856 proton precession magnetometer, Geometrics-858 cesium vapor magnetometer, Geoscan Research FM36 fluxgate gradiometer Increasing detail and quality of anomaly definition as a result of greater sample densities over a pair of burned houses at the Menoken Village State Historic Site Magnetometer anomalies created by intensive firing Magnetometer anomalies created by fired artifacts Positive magnetometer anomalies caused by accumulations of topsoil associated with constructed features Negative magnetometer anomalies produced by the removal of magnetically enriched topsoil Significant magnetometer anomalies introduced by imported stone Strong magnetometer anomalies produced by iron and steel artifacts A magnetic survey in the vicinity of granite boulders showing largemagnitude dipolar anomalies stemming from remanent magnetism Massive anomalies caused by large iron or steel bodies on a site Dipolar anomalies representing steel-wire pin flags Total loss of data in one segment of a survey as a result of keys in the operator’s pocket Dipolar anomalies found to be clusters of steel bottle caps and beer cans The spatial organization of the Fort Clark Trading Post and its environs at the Fort Clark State Historic Site as revealed by magnetometry Magnetometry data from Primeau’s Trading Post at the Fort Clark State Historic Site, North Dakota, showing effect of a de-spiking algorithm Magnetometry data from an early Archaic occupation at the Wallace Bottom site, Arkansas, showing use of Fourier methods to remove plow marks A complete magnetic processing sequence: raw data illustrating drift and heading errors; data after “zeroing” the transects; data after application of a “de-staggering” algorithm; data after removal of the gait defect through Fourier methods; data after low pass filtering; data after interpolation Grid imbalances in total field data gathered by a proton precession magnetometer at the Roman city of Empuriés Magnetometry data from Double Ditch State Historic Site, North Dakota, subjected to contrast manipulation Typical modes of graphic display illustrated with magnetometry data from the Great Bear effigy at Effigy Mounds National Monument, Iowa


215 216 217 218 219 220 221 225 226 227 228 229 230 237


240 243 245


xiv ~ Figures 10.7. 11.1. 11.2.

11.3. 11.4.

11.5. 11.6. 11.7. 11.8. 11.9. 11.10. 12.1. 12.2. 12.3. 12.4. 12.5. 12.6. 12.7. 12.8. 12.9. 13.1. 13.2. 13.3.

Magnetometry data from Huff Village State Historic Site, North Dakota, with an interpreted map Geophysical surveys at Whistling Elk village: resistivity, conductivity, magnetic gradiometry Geophysical mappings of circular earthlodge(s) at the Mandan/Arikara village in the Fort Clark State Historic Site: resistivity, GPR time slice, magnetic gradiometry The brick foundation of the Mount Comfort Church as revealed by resistivity, GPR, and magnetic gradiometry Multidimensional geophysics at Army City: resistivity, conductivity, magnetic gradiometry, magnetic susceptibility, and GPR, and RGB color composites of the data Gradiometer image of the Hollywood Mounds site Photographic imagery of the Hollywood Mounds site Airborne digital imagery of the Hollywood Mounds Magnetic gradient image of prehistoric house remains reclassified into three classes of data Airborne imagery used in reclassification Original pixel classification and discriminant function results Results of electrical resistivity survey at the Crying Hawk site The Grossmann site: results of a magnetic field gradient survey and subsequent excavation Results of mechanized stripping, resistance survey, and soil cores at the Hoxie Farm site Electrical resistance map of the Army City site, Fort Riley, Kansas Panoramic photograph of Army City, ca. 1918 Trench excavated to ground truth a fortification ditch at the Double Ditch site Magnetic map of Fort Clark Trading Post, North Dakota Map showing magnetic foundation stones documented in a block of contiguous test units at Fort Clark Trading Post GPR map of Ellis Cemetery showing the location of gravestones Magnetic gradient and magnetic susceptibility images of the Walford site showing pit feature locations A portion of the resistance imagery from the Presidio de Santa Rosa showing the utility of a filter Magnetic gradient image of the Confederate cemetery on campus at the University of Mississippi

248 253

254 256

258 260 261 261 262 263 264 283 287 290 292 294 296 297 298 299 307 309 310

Tables 3.1. 3.2. 4.1. 5.1. 5.2. 11.1. 11.2. 12.1.

Test excavation simulation results Cost simulation of traditional vs. remote sensing–based data recovery Thermal inertia values for common materials Resistivity and conductivity of different soil types Typical data set produced from data logger for processing Standardized canonical discriminant function coefficients for house location analysis Classification results for house location analysis Usefulness of ground truthing techniques at the sites discussed in this chapter

42 43 53 83 95 264 265 282

Acknowledgments This is my third edited volume. Each time I finish one, I vow not to do another. And then the opportunity comes by and it’s too good to pass up. In this case, the quality of the contributions and the timeliness of the collection persuaded me. Also, the prospect of working with such a distinguished and agreeable group of chapter authors made the project attractive. I thank them all. I met most of the contributors to this book when I attended my first National Park Service workshop on remote sensing and archaeology at Chillicothe, Ohio, in 2001. These annual events are sponsored by the Midwest Archeological Center and organized by Steve De Vore. I thank the NPS and Steve for the opportunity to learn so much in such a short period of time. Although Berle Clay is a regular instructor at the NPS workshops, I have known him just about as long as I have been working in the Southeast: a long time. However, he deserves special mention in that he introduced me to the potential of geophysical remote sensing when he showed up at the Hollywood site in 1997 with a conductivity meter. It was an impressive demonstration; test pits in six of the eight possible structures revealed in the resultant imagery came down on house floors. I also met Marco Giardino at the Hollywood site and, as outlined in Chapter 1, he was a coconspirator in organizing the regional workshop in Biloxi that led to this volume. Not only did he help organize it, but he also provided NASA funding for both the workshop and the follow-up meeting in New Orleans at which the contributors got together to work out the details of the volume. Funding for the production of the volume as well as the workshop was provided by the University of Mississippi Geoinformatics Center, a NASA-funded initiative under the direction of Greg Easson at Ole Miss. This volume is the seventh time that I have worked with Kathy Cummins as copy editor and typesetter. As always, it has been a pleasure. I literally could not have done it without her. Finally, I thank Bryan Haley, my coauthor on two of the chapters in this volume, a “first generation” graduate of the remote sensing and archaeology focus of our graduate program, my research associate, and the man who keeps it all working while I attend to administrative and academic matters.

Remote Sensing in Archaeology


Introduction Jay K. Johnson

This book began in a conversation between Marco Giardino and me at the bar in Fitzgerald’s Casino during the summer of 2001. The bar top was embedded with video gaming screens and we had worked out a system whereby it took us nearly two hours to lose $10.00 playing blackjack. All that time we were supplied with “free” beer. Before going any further, I should mitigate this revelation by pointing out that Fitzgerald’s Hotel was the field headquarters for the Ole Miss field school that year. We were working on the Hollywood Mounds, a large, late prehistoric ceremonial center at which geophysical survey techniques, particularly gradiometry and conductivity, have proven remarkably effective. Marco was working with us, wrestling with the much more difficult job of getting informative results from ground-penetrating radar (GPR) in the clays and silts of the Mississippi alluvial valley. We were bemoaning the lack of application of these techniques in Southeastern archaeology in general and cultural resource management (CRM) archaeology in particular. As the chapters that follow will demonstrate, remote sensing, especially the geophysical techniques, has reached the point where it can make a substantial contribution to the dirt archaeology of the Southeast. However you frame the argument, whether in terms of refining the research design or of cost effectiveness, on most sites, the application of remote sensing early on in the fieldwork will lead to better results. However, on some sites you might as well leave the instruments in the truck. One of the goals of this volume is to help CRM administrators integrate remote sensing into their data-recovery programs in an informed way.

2 ~ Jay K. Johnson But, back to Fitzgerald’s. We decided that many of the archaeologists working in the South were not aware of the remarkable advances in remote sensing applications that have occurred during the past 10 years and that what was needed was a workshop on remote sensing applications in archaeology. Marco secured funds through his office, NASA’s Earth Science Applications Directorate at Stennis Space Center; I found additional support from the University of Mississippi Geoinformatics Center; and a workshop was planned for the Wednesday preceding the annual meeting of the Southeastern Archaeological Conference (SEAC), which was held in Biloxi, Mississippi, in 2002. I got on the phone to my friends in remote sensing and in a short time assembled the impressive list of instructors represented in the following chapters. That was followed by the much more demanding task of locating and inviting the state archaeologists, state historic preservation officers, and chief highway archaeologists or their representatives from the 11 states that are traditionally represented in the SEAC membership. We planned to begin the workshop with a field trip to Tullis-Toledano Manor, a historic site in Biloxi, where there would be demonstrations of the various instruments. The afternoon would be devoted to presentations on the several major remote sensing techniques appropriate to archaeology. A reception was planned for the evening, during which we would talk about all that we had done that day. The workshop was a success. Many of the participants expressed an interest in applying the techniques and, in fact, several were from state agencies that were already using some of the instruments. The presentations were all quite good. So good, in fact, that we decided to follow up with a one-day workshop just for the instructors in which we would work on preparing a handbook on remote sensing applications for CRM archaeologists. We met in the French Quarter in New Orleans at the Royal Sonesta Hotel and spent another very successful day talking about the focus of the publication. Then came the hard part: finding the time to fulfill the commitments we had made and actually writing the following chapters. Although the workshop was presented to archaeologists working in the Southeast, the instructors work throughout North America and the volume reflects this broader perspective. For this and other reasons, I am pleased with the results, but, of course, the final judgment will be up to the readers. I would like to address one fundamental question, however. Was such a volume needed? There are, after all, several very good summaries of remote sensing applications in archaeology (Aitken 1961; Bevan 1998; Clark 1996; Gaffney and Gater 2003; Scollar et al. 1990), most of which have the same emphasis on geophysics that is evident in the following pages. However, there are at least three reasons to add one more book to this list. In the first place, we are riding the crest of a technology that is advancing on a daily basis. For members of my generation, who did their dissertation research using punch cards, this is particularly evident. But the rate of advance is accelerating. This is especially true in remote sensing, in which large amounts of data must be processed

Introduction ~ 3 in complex ways and the output is most useful in a graphic format. Driven by applications with much more economic impact than archaeology, computer graphics, memory, and processing time are improving exponentially. If you doubt that, violate the cardinal rule of buying a PC, and see what you could have gotten for the same money a month later. Scollar and his coauthors (1990) published one of the most comprehensive reviews of remote sensing in archaeology to date. Certainly it contains more formulas than any other publication on the subject. And it is still an important source of fundamental concepts. However, it came out more than a decade ago and the discussions of computer hardware and graphic presentation are useful only as a benchmark of where we’ve been. You can effectively date a publication by looking at the pictures. Second, all but one (Bevan 1998) of the major publications on geophysical remote sensing in archaeology use examples drawn from European archaeology, which is at least a decade ahead of us in remote sensing applications. Another interesting thing about the European use of remote sensing is that it is an integral part of their equivalent of CRM archaeology. There is even a popular British television show, “Time Team,” that features applications in archaeology. Why is it that the random person on the street in London is likely to be able to discuss the relative merits of using a magnetometer rather than GPR, while many North American archaeologists are not sure what these instruments do in the first place? Part of the answer lies in the nature of the archaeology. Almost any discussion of the archaeological application of aerial photography will have a long section on crop marks, which appear to be particularly useful in discovering Roman villas and Bronze Age fortresses, site types that are uncommon in North America. Sites that predate the Neolithic are not regularly featured in discussions of remote sensing in Europe, for obvious reasons—the traces left behind are far less structured and much more difficult to detect using remote sensing techniques. This is, of course, the case with most of the prehistoric record in North America. Finally, although most of the archaeologists who attended the workshop in Biloxi came away convinced that remote sensing will make a major contribution to the archaeology that they administer, many also expressed frustration; the successful application of the techniques relies on a great deal of expertise in archaeology, geophysics, digital image processing, and soils. When the right instrument is used on the right kind of archaeological deposit buried in the right kind of soil, the results are often spectacular. However, there is an unfortunate history of inappropriate applications in which a substantial amount of money was spent with no results. A general overview of remote sensing techniques that will guide archaeologists in the selection and application of instruments is badly needed at this stage in the development of the field. That is the major goal of this book. The heart of the book is the applications chapters (Chapters 4 through 9). Each author was asked to cover the following topics:

4 ~ Jay K. Johnson Overview of the technique Discussion of basic principles A brief history of its application in archaeology A summary of currently available and generally used instrumentation A description of the typical field strategy Some idea of the kinds of data-processing software that are most likely to be useful Examples of successful applications Case studies, many of which are drawn from the chapter author’s (or authors’) own research Guidelines for application When to use which combination of instruments Soils Site types Interference Field time Data-processing time The final chapter (Chapter 13) brings together the data included in the concluding section of each of the applications chapters so that archaeologists can make a comprehensive decision about which remote sensing techniques to employ. The major purpose of the New Orleans meeting was to work together on the details of this chapter. No such comprehensive guide to the successful application of remote sensing techniques is currently available, but we judge that the time is right. So, now you have some insight into the origin and justification of this volume, but before concluding this introduction, I would like to do a few more things. First, there is the not-so-trivial question of what exactly we mean by “remote sensing.” There was some discussion of this topic at the meeting of contributors in New Orleans. In fact, most of the geophysical techniques that have provided spectacular results in archaeology are hardly remote. Some remote sensing instruments—gradiometers and conductivity meters—are generally carried back and forth across the site 10–20 cm above the ground. GPR systems must make contact with the surface of the soil in order for the signal to propagate, and conductivity meters can be dragged along the surface. Resistivity readings are taken by inserting probes into the soil, and some susceptibility applications, as being pioneered by Rinita Dalan (Chapter 8, this volume), require that the sensor be inserted into a borehole. Compared with satellite and airborne sensors, these are obviously a different class of readings. However, geophysical techniques are still used to measure phenomena that are remote from the sensor and cannot be seen otherwise. It is just a matter of scale, as suggested by Payson Sheets (1991) in an article entitled “‘Very-to-Barely’ Remote Sensing of Prehistoric Features…,” in which he re-

Introduction ~ 5 ports the results of the application of instruments ranging from airborne multispectral scanners to GPR. More important, there is a fundamental similarity in the way that satellite, airborne, and geophysical data are processed and evaluated. For example, the question of resolution is an important first consideration in all cases. The smaller the unit of observation—pixel size in remote sensing terms—the more likely you are to find small features and the prettier the picture. However, the finer the resolution, the more expensive the data in terms of acquisition, storage, and processing. If you are looking for broad-scale patterns, it is often unnecessary to spend the money on high-resolution images. It may even be a detriment. For example, in GPR, the higher the frequency of the antenna, the smaller the object that can be detected. However, many of the reflections that are recorded in GPR are irrelevant to understanding the cultural deposits at a site. The usual goal of a radar survey in archaeology is to detect buried structures, seen as major reflections that continue across several transects. Much of the data recovered by using a high-frequency antenna is noise. Many of the data-processing techniques used in geophysical analysis were developed for the analysis of satellite data. A high pass filter is a high pass filter whether it is being used on digital data acquired by a sensor orbiting hundreds of kilometers above the earth or on data acquired by a gradiometer carried back and forth across the site at a distance of a few centimeters from the surface. And that filter can be applied using software written specifically for magnetic gradient data, such as Geoplot, or it can be applied using one of several programs written specifically for more traditional remote sensing analysis; ERDAS Imagine, for example. Another conceptual advantage to the more inclusive definition of remote sensing is that many of the standard procedures of satellite image analysis hold tremendous potential in archaeology. For example, anyone who deals with geophysical data uses the basic concepts developed in geographic information systems (GIS) analysis, and it is clear that the integration of data recovered by more than one instrument is likely to increase our understanding of the structure of an archaeological site. But, as Chapter 11 on multiple instrument applications illustrates, we are just beginning to make use of the powerful tools that are available to integrate multiple kinds of spatial data. As I have indicated, this is hardly the first book on remote sensing and archaeology. As the instructions to the chapter authors indicate, it is not meant to be a comprehensive or detailed introduction. For those of you who want to learn more or are interested in the remarkable pace of development in this area of archaeological research, there are several options. The most comprehensive early overview of geophysical survey techniques in archaeology is more than 40 years old but still contains valuable information. Aitken (1961) reviews a field of inquiry that was hardly more than 10 years old at the time of his writing. As the title of the book, Physics and Archaeology, suggests, his topic is broader than just remote sensing, and there are chapters on radiocarbon dating and trace element analysis. However, there are also chapters on magnetic detection and

6 ~ Jay K. Johnson resistivity surveying. Beyond the wonder the book inspires at the determination it took to use instruments that were slow and imprecise by today’s standards and the time it took to record and plot the values by hand, it is also a bit humbling to realize how little we have progressed in the basic understanding of the characteristics of the archaeological record that influence the utility of these techniques. Aitken (1961:1) also makes the distinction between finding archaeological sites and exploring those sites once they are found and notes that different methods and instruments are useful in each case. For example, he includes a brief discussion of aerial photography in his chapter on site discovery, while the chapters on magnetics and resistivity deal mostly with mapping features within sites. The National Park Service has played a lead role in the introduction of remote sensing techniques in North American archaeology. In fact, the first publication with goals that are similar to ours came out in 1977 under the title Remote Sensing: A Handbook for Archeologists and Cultural Resource Managers (Lyons and Avery 1977). Ten supplements were published, the last coming out in 1985. Most deal with regional applications with an emphasis on airborne and satellite sensors. However, Supplement 3 (Lyons et al. 1980) is an extensive bibliography, containing several entries relating to geophysical techniques along with the more numerous citations dealing with airborne and satellite sensors. Supplement 2 (Morain and Budge 1978) presents a discussion of instrumentation and contains the only discussion of traditional geophysical techniques. This short section is introduced with a definition of remote sensing that includes geophysical techniques and makes the observation that “in archaeology, we take considerable interest in buried structures and artifacts, objects that are not visible to the eye and quite probably not directly detectable using space or airborne sensors. This is the area in which ground based remote sensing plays a vital role” (Morain and Budge 1978:24). A summary discussion of magnetometry, resistivity, and radar concludes this section of the supplement. The first comprehensive, generally accessible overview of the application of geophysical techniques in North American archaeology was written by John Weymouth (1986) and published in the Advances in Archaeological Method and Theory series. He concentrates on the exploration of site structure since “these tools are too costly and time consuming to be used for site discovery and identification” (Weymouth 1986:312). Resistivity, one of the oldest methods to be used in archaeology, is discussed in some detail. It is interesting to note that, although Weymouth illustrates several different probe arrays and discusses the disadvantages in using the Wenner array, his examples are drawn from Wenner array data. Because of changes in available instrumentation, this configuration has been almost completely replaced by the twin array. Likewise, although he discusses magnetic gradiometers, a relatively new instrument at the time, most of his examples of magnetic data were recorded using a single instrument and a separate base-station instrument to record and adjust for changes in the magnetic field throughout the day. Finally, electromagnetic conductivity meters are discussed but no case studies are presented. The late 1980s was a period of rapid development

Introduction ~ 7 in instrumentation and, in some ways, this article is located at a watershed. The use of dot density maps to display survey results is another expression of the time at which it was written. Still, the potential of the techniques is clearly expressed and a good deal of insight into the strengths and weaknesses of the various instruments is presented. NASA became active in remote sensing applications in archaeology during the 1980s and, not surprisingly, there was an emphasis on digital remote data, mostly derived from satellite-based instruments (Behrens and Sever 1991; Sever and Wiseman 1985). Limp (1989) provides a useful summary of the use of digital multispectral imagery in archaeology as of 1989. The primary source of these data at that time was sensors mounted on satellites, most of which had a resolution of 20 m or more. SPOT, a French sensor, was launched in 1986 and provided 10-m data. Airborne sensors could do much better, achieving a resolution of 2 m or less. It is not surprising that, of the 68 studies summarized in the report (Limp 1989:table 23), all deal with either site discovery, environmental delineation, or predictive modeling. Only one published study mentioned in the text (O’Brien et al. 1982) was successful in detecting withinsite cultural features. Limp (1989:54) does conclude, however, that “multispectral imagery also has great potential for within-site studies,” particularly with the increasing availability of high-resolution sensors. It is interesting, therefore, that Scollar et al. (1990), in one of the most comprehensive books on remote sensing in archaeology yet to be published, spend three chapters on the use of aerial photography and another on airborne thermography including discussion of sophisticated computer-based image transformations but make no mention of multispectral sensors. This may be partially the result of the tremendous success of the application of aerial photography in Europe (e.g., Wilson 1987) in contrast with its limited use in North America (e.g., Deuel 1969). The treatment of resistivity and magnetic and electromagnetic prospection in Scollar and colleagues’ book contains a great deal of detail, often expressed in equations. Although it is not light reading, it is an essential reference for anyone wanting to truly master the fundamentals of geophysical applications in archaeology. This probably accounts for the fact that it rarely shows up on any of the on-line rare-book dealer sites. As mentioned above, most of the discussions of graphic techniques and nearly all of the details on computer software and hardware have been rendered obsolete by the remarkable advancements in this area. Because it was expensive, hard to find, and tended to be technical, the Scollar text is not nearly as well known as another overview of geophysical survey applications in archaeology published in Great Britain in the same year, that by Clark (1990). Like Scollar, Clark was in on the early growth of geophysical applications in archaeology, having had a part in the mid-1950s in developing one of the first resistivity meters designed especially for archaeologists. The book begins with a thorough review of the history of geophysical prospection in Europe, with an emphasis on Great Britain. Clark (1996:7) notes his conscious decision to avoid the use of equations, and his presentation relies heavily on several rather persuasive images. A second edition, brought out in 1996, was updated with a 17-page supplement that documents some of the notable advances in instru-

8 ~ Jay K. Johnson mentation and processing that took place during the early 1990s. Clark wrote his dissertation on resistivity and, not surprisingly, the chapter covering that technique is the longest in the book. Magnetometry and magnetic susceptibility get their own chapters, but electromagnetic conductivity is treated as a subheading in the resistivity chapter and GPR is included in a chapter titled “Other Methods.” This is an improvement over treatment of the topic by Scollar and his coauthors (1990:575–584), who devote 10 pages (and 12 equations) to GPR. In fact, this coverage reflects the relative importance of the various techniques at the time the books were written. GPR is the one technique that is best known by the public and general archaeological community. The first thing people ask for when they contact me about a survey is GPR. However, it is also the most difficult of the geophysical instruments to use in terms of data processing. It is appropriate, therefore, that there is a book devoted entirely to the archaeological application of GPR (Conyers and Goodman 1997). The timing of its publication is also appropriate because Conyers and Goodman are able to document a major breakthrough in GPR survey, one in which they played a part. The standard output of a GPR survey is a profile along the line of travel of the antenna. This profile shows a series of reflections, most often as horizontal bands that are deflected when the instrument passes over a buried object that has sufficient contrast with the soil matrix in which it is buried. An experienced operator can interpret these profiles, but for the rest of us they are similar to Rorschach tests. And, since most of the features that archaeologists are looking for are easiest to recognize in plan view, the profiles must be considered in aggregate. GPR became much more useful to archaeologists with the development of processing techniques that allow several parallel profiles to be stacked side by side and extrapolated into a data cube. Horizontal slices can then be generated and broad horizontal patterns discovered. As a result of this advance, GPR has become one of the principal tools in geophysical survey in archaeology over the past decade. The National Park Service, particularly the Midwest office, continued to sponsor applications of remote sensing to archaeological research problems throughout the 1980s and 1990s. Two general overview monographs (Bevan 1998; Heimmer and De Vore 1995) as well as a valuable series of annual workshops in remote sensing and archaeology have resulted. The week-long workshops are typically held at a major archaeological site. (My graduate students and I attended our first one in 2001 focusing on the Hopewell earthworks at Chillicothe, Ohio.) Most of the contributors to this volume are regular instructors at the workshops. These workshops are excellent introductions to the field and reflect a shift in emphasis that has occurred in remote sensing applications in archaeology since the handbook by Lyons and Avery (1977) was published by the National Park Service. The primary emphasis at these workshops is geophysics with some discussion of aerial photography but, at least at recent workshops, little on digital airborne or satellite data. Like the earlier publication (Lyons and Avery 1977), there is an explicit emphasis on CRM applications in Heimmer’s (1992; reissued as Heimmer and De Vore 1995) review of geophysical applications to archaeological research. This monograph is a con-

Introduction ~ 9 cise discussion of the several different techniques that are most commonly used in archaeology written by a person trained in geophysics. Appendices include a glossary, a selected bibliography, and a list of providers of equipment and contract surveyors. The final appendix includes a limited selection of images produced by geophysical surveys of archaeological sites. Unfortunately, the discussion of techniques presented in the body of the text is completely divorced from the archaeological examples and the bibliography of archaeological applications. Although the publication is useful, it is difficult to get a feel for the relative merits of the instruments and the way they have been applied on archaeological sites. Bevan’s (1998) monograph goes a long way toward correcting this problem. Bevan has been one of the major practitioners of archaeological geophysics since the mid1970s, and his Geophysical Exploration for Archaeology: An Introduction to Geophysical Exploration reflects this depth of experience. In addition, Bevan has a talent for explaining the physics so that archaeologists can understand them. The only shortcoming of this work is that, as a special publication of the Midwest Archeological Center, it has had limited distribution. Also, Bevan is somewhat conservative when it comes to graphic presentation. In contrast, the latest book-length overview of geophysics and archaeology (Gaffney and Gater 2003) takes full advantage of advances in computer graphics and image analysis that have become available to the archaeologist. More than a hundred mostly gray-scale images and two dozen color plates offer compelling arguments for the importance of geophysical techniques in archaeological research. Alternative methods of data transformation are discussed and illustrated along with side-by-side comparisons of different ways to present the data. Gradiometer images of entire Roman villages look like town maps, and resistance plots showing detailed floor plans of medieval buildings leave little doubt that the long tradition of geophysical research and development in British archaeology has paid off. Beyond the importance of understanding and appreciating the development of archaeological geophysics in Great Britain, there are other important issues that are raised by Gaffney and Gater. The first is the contrast between the application of the techniques here and in England. Although CRM archaeologists are beginning to take advantage of geophysical techniques in the United States, a major portion of the research and most of the publication take place in an academic setting. The reverse is true in England where, of the estimated 450 geophysical surveys performed on archaeological projects per year in that country, the vast majority are done by “independent groups” (the British equivalent of CRM firms) under contract with developers (Gaffney and Gater 2003:fig. 3). A second issue is raised by this (Gaffney and Gater 2003:22) and an earlier review of archaeological geophysics in England (Clark 1996:8). As a result of changes in the guidelines for planning and development that took place in Great Britain in the early 1990s, archaeological geophysics was expanded to become a site-discovery technique. In this, there is another strong contrast between Great Britain and the United States,

10 ~ Jay K. Johnson where there are almost no examples in which geophysical surveys have been used to find sites. In North American archaeology these techniques have largely been used to conduct what Gaffney and Gater (2003:88) call site assessment and investigation. This may be a reflection of the academic emphasis of archaeological geophysics in the United States. It may also be a result of the difference in the archaeological record between the two countries. Gaffney and Gater (2003:120) begin their chapter on case studies dealing with prehistoric sites with the observation that for the Paleolithic the dispersed and ephemeral nature of the surviving archaeology means there is little remaining that the techniques can detect. For the Mesolithic, even though longer lived sites with good evidence for fires exist, finding them provides the sort of challenge that most geophysicists would rather pass by. Likewise, Scollar et al. (1990:4) note that “with the invention of agriculture at the beginning of the neolithic period, man began to perturb this natural sequence [pedogenesis] for the first time. … Before agriculture, tents or small sheds were sometimes constructed, but their remains are very difficult to spot.” The problem for North American CRM archaeologists is that a large portion of the archaeological sites that they deal with fall into the category that European archaeologists find difficult to detect and explore using geophysical techniques. This is partially because in most of the United States, intensive agriculture began little more than 1,000 years ago. There is, however, another factor. When Bryan Haley and I give papers on the results of our geophysical research in the Southeast, we usually show pictures of large, late Mississippian sites on which buried mounds, burned house remains, plazas, and other features are clearly evident in the images. These are exactly the kinds of sites that will be avoided if at all possible by CRM planners because of the large budgets needed to excavate them. Therefore, if we are to make the case for including geophysical survey as a regular part of the CRM planning process in the United States, those of us who do the work must devote more energy to the small hunting camps and horticultural villages that predate the major settlements of the agricultural periods and predominate in the archaeological record. As some of the case studies discussed in the following chapters illustrate, understanding these sites will almost certainly require the application of multiple instruments, the results from which must be evaluated in conjunction with the ground truth excavations. Before I close this literature review with such a decidedly European bias, I should tell you that although there are fewer people doing archaeological geophysics in the United States than in Europe and they have been doing it for a shorter period of time, there are several signs that they are catching up. Ken Kvamme, one of the best of the “second generation” practitioners of archaeological geophysics in the United States, has written a couple of recent review articles (Kvamme 2003a, 2003b) in which he discusses the application of the techniques in terms of their practical and theoretical benefit to New World archaeology and argues persuasively for a central role for archaeological geophysics in CRM archaeology (Kvamme 2003a:452–453). He also maintains the North American Data Base of Archaeological Geophysics website, whose creation was

Introduction ~ 11 funded by the National Center for Preservation Technology and Training, a National Park Service program. This website ( contains links to numerous other websites dealing with archaeology and remote sensing, a comprehensive bibliography, a current list of practitioners, and a project database containing survey results from most of the contributors in this volume. This would be an excellent starting point for anyone interested in learning more about archaeological geophysics in North America. There is one final measure of the boomtown growth of the application of remote sensing, particularly geophysics, in North American archaeology. While I was finishing this introduction, I received a December 2003 monograph authored by Lew Somers and Michael Hargrave entitled Geophysical Surveys in Archaeology: Guidance for Surveyors and Sponsors, published by the Construction Engineering Research Laboratory of the U.S. Army Corps of Engineers. The primary focus of this monograph is a decision support software called ATAGS (Automated Tool for Archaeo-Geophysical Survey), which was written to help CRM archaeologists and novice geophysical researchers in designing survey protocol. The user inputs several different site parameters, soils, integrity, and anticipated kinds of features, and the program makes recommendations about sampling interval and field time for resistivity and gradiometer surveys. An annotated bibliography of major publications on archaeological geophysics along with discussions of field and data-processing procedures and a few case studies are also provided. Clearly, there is a growing awareness of the potential contribution of remote sensing to North American archaeology in general and CRM archaeology in particular. Not only can it be justified in terms of cost effectiveness on large, complex sites (Johnson and Haley, Chapter 3, this volume) but, as Jami Lockhart and Tom Green demonstrate in Chapter 2, geophysical survey fulfills the specific requirements of CRM laws and guidelines in a way that other methods of site assessment cannot match. Although airborne and satellite-based remote sensing has been overshadowed in archaeological applications in recent years, geological, environmental, and surveillance applications have created a market that has resulted in better spectral and spatial resolution at a much reduced price. In many cases, the prehistoric patterns of interest are spread out across the landscape in such a way that airborne imagery is the only reasonable approach. Marco Giardino and Bryan Haley review past and current applications in Chapter 4. In addition, in some instances it has been possible to extend geophysical results by using them as a training set for a multivariate classification of airborne imagery with much broader coverage (see Johnson and Haley’s discussion in Chapter 11). Chapters 5, 6, 7, and 9 cover the four major instruments used in archaeological geophysics in North America. Because we deal with different kinds of sites, electromagnetic conductivity instruments may be more important in the United States than in Europe, and Berle Clay’s chapter does a good job of pointing out the various conditions under which they are likely to be useful. Resistivity, on the other hand, is much more important in British applications than in North America. However, as Lew Somers demonstrates, it is an important technique that should certainly be con-

12 ~ Jay K. Johnson sidered. Larry Conyers’s chapter on GPR deals with data processing a bit more than those of the other contributors because GPR data demand more and greatly different techniques in order for them to be accessible to the average archaeologist. Chapter 8, written by Rinita Dalan, covers magnetic susceptibility, a technique that is not much used in archaeological geophysics in this country, but should be. Kvamme’s chapter on magnetometry makes it clear why this is one of the primary tools for the investigation of large prehistoric and historic sites. Other than the good food and good company, the New Orleans workshop was valuable in that we all got together and made decisions about what we wanted this book to contain. Chapters 10 and 11 are direct results of that meeting. Because data processing is so important in making sense out of remotely sensed data, we decided to devote a separate chapter to it. All participants were invited to contribute but it is fitting that Kvamme ended up writing it since he is always pushing the limits of the data through processing and presentation. We also decided to add a chapter on multiple instrument applications. There are very few sites that would not be easier to understand if two or more instruments were used. Not only do the different instruments detect different things, but often they see the same things differently. Moreover, statistical techniques, many of which have been borrowed from the analysis of satellite-based sensor data, are making “data fusion” more a reality and less a buzz word. One of the common complaints made by people who do archaeological geophysics is that they do the survey and someone else does the ground truth (excavations) using the imagery produced by the remote sensors but failing to get back to the surveyors with the results. This is an unfortunate result of the specialization required to be able to afford and understand these instruments and it has slowed the growth of archaeological geophysics as a science. However, as Mike Hargrave demonstrates in his chapter on ground truth excavation techniques (Chapter 12), the feedback that should occur between those who obtain the imagery and those who do the excavations will allow a much better use of the geophysical survey results and a more comprehensive data recovery. The final chapter is a compendium of the previous chapters with a special emphasis on when remote sensing is likely to be useful and which instruments to use in specific situations. In this it follows a long tradition within the literature, often expressed in tabular form (David 1995). It is a tradition that will continue as the techniques and instruments improve. For example, the first time we tried GPR at the Hollywood site (Johnson et al. 2000) we couldn’t even detect a steel culvert under one of the field roads. This conforms with the general wisdom that GPR is a poor choice of techniques in fine-grained soils. Since various kinds of clays are the major constituent of the Mississippi alluvial valley in northwestern Mississippi, we held little hope for the technique in our research area. However, a newer instrument and better processing techniques allowed a solid Master’s thesis to be written on the use of GPR at the Hollywood Mounds (Peukert 2002).

Introduction ~ 13 And that is one of the major lessons to be learned from this or any review of the use of remote sensing in archaeology. As the number and kind of applications grow, the results are becoming more and more sophisticated. There can be little doubt that in five or fewer years a second edition of this volume will need to be written. The good news is that we have made plans to do just that.

References Cited Aitken, M. J. 1961 Physics and Archaeology. Interscience, New York. Behrens, C. A., and T. L. Sever (editors) 1991 Applications of Space-Age Technology in Anthropology. NASA, John C. Stennis Space Center, Mississippi. Bevan, B. W. 1998 Geophysical Exploration for Archaeology: An Introduction to Geophysical Exploration. Special Report No. 1. U.S. Department of the Interior, National Park Service, Midwest Archeological Center, Lincoln, Nebraska. Clark, A. J. 1990 Seeing Beneath the Soil: Prospecting Methods in Archaeology. B. T. Batsford, London. 1996 Seeing Beneath the Soil: Prospecting Methods in Archaeology, new ed. B. T. Batsford, London. Conyers, L. B., and D. Goodman 1997 Ground-Penetrating Radar: An Introduction for Archaeologists. Altamira, Walnut Creek, California. David, A. 1995 Geophysical Survey in Archaeological Field Evaluation. Ancient Monuments Laboratory, English Heritage Society, London. Deuel, T. 1969 Flights into Yesterday. St. Martin’s Press, New York. Gaffney, C., and J. Gater 2003 Revealing the Buried Past: Geophysics for Archaeologists. Tempus, Gloucestershire, Great Britain. Heimmer, D. H. 1992 Near-Surface, High Resolution Geophysical Methods for Cultural Resource Management and Archeological Investigations. U.S. Department of the Interior,

14 ~ Jay K. Johnson National Park Service, Rocky Mountain Regional Office, Division of Partnerships and Outreach, Interagency Archeological Services, Denver. Heimmer, D. H., and S. L. De Vore 1995 Near-Surface, High Resolution Geophysical Methods for Cultural Resource Management and Archeological Investigations, rev. ed. U.S. Department of the Interior, National Park Service, Rocky Mountain Regional Office, Division of Partnerships and Outreach, Interagency Archeological Services, Denver. Johnson, J. K., R. Stallings, N. Ross-Stallings, R. B. Clay, and V. S. Jones 2000 Remote Sensing and Ground Truth at the Hollywood Mounds Site in Tunica County, Mississippi. Center for Archaeological Research, University of Mississippi, Oxford. Submitted to the Mississippi Department of Archives and History. Kvamme, K. L. 2003a Geophysical Surveys as Landscape Archaeology. American Antiquity 68(3):435–458. 2003b Multidimensional Prospecting in North American Great Plains Village Sites. Archaeological Prospection 10:131–142. Limp, W. F. 1989 The Use of Multispectral Digital Imagery in Archeological Investigations. Research Series 34. Arkansas Archeological Survey, Fayetteville. Lyons, T. R., and T. E. Avery 1977 Remote Sensing: A Handbook for Archeologists and Cultural Resource Managers. Cultural Resources Management Division, National Park Service, Washington, D.C. Lyons, T. R., R. K. Hitchcock, and W. H. Wills 1980 Remote Sensing, Aerial Anthropological Perspectives: A Bibliography of Remote Sensing in Cultural Resource Studies. Remote Sensing: A Handbook for Archeologists and Cultural Resource Managers 3. Cultural Resources Management Division, National Park Service, Washington, D.C. Morain, S. A., and T. K. Budge 1978 Remote Sensing: Instrumentation for Nondestructive Exploration of Cultural Resources. Remote Sensing: A Handbook for Archeologists and Cultural Resource Managers 2. Cultural Resources Management Division, National Park Service, Washington, D.C. O’Brien, M. J., J. L. Beets, R. E. Warren, T. Hotrabhavananda, T. W. Barney, and E. E. Voigt 1982 Digital Enhancement and Grey-Level Slicing of Aerial Photographs: Techniques for Archaeological Analysis of Intrasite Variability. World Archaeology 14(2):173–188.

Introduction ~ 15 Peukert, J. N. 2002 Ground Penetrating Radar at Hollywood. Unpublished Master’s thesis, University of Mississippi, Oxford. Scollar, I., A. Tabbagh, A. Hesse, and I. Herzog 1990 Archaeological Prospecting and Remote Sensing. Topics in Remote Sensing, No. 2, G. Hunt and M. Rycroft, series editors. Cambridge University Press, Cambridge. Sever, T., and J. Wiseman 1985 Remote Sensing and Archaeology: Potential for the Future. NASA, John C. Stennis Space Center, Mississippi. Sheets, P. D. 1991 “Very-to-Barely” Remote Sensing of Prehistoric Features under Tepra in Central America. In Applications of Space-Age Technology in Anthropology, edited by C. A. Behrens and T. L. Sever, pp. 167–180. NASA, John C. Stennis Space Center, Mississippi. Somers, L. E., and M. L. Hargrave 2003 Geophysical Surveys in Archaeology: Guidance for Surveyors and Sponsors. Construction Engineering Research Laboratory, U.S. Army Corps of Engineers, Champaign, Illinois. Weymouth, J. W. 1986 Geophysical Methods of Archaeological Site Surveying. In Advances in Archaeological Method and Theory, vol. 9, edited by M. B. Schiffer, pp. 311–395. Academic Press, New York. Wilson, D. R. 1987 Air Photo Interpretation for Archaeologists. Batsford, London.


The Current and Potential Role of Archaeogeophysics in Cultural Resource Management in the United States J. J. Lockhart and Thomas J. Green

The value of geophysical surveys in archaeological applications is increasingly recognized as project results are disseminated at conferences and in publications within the United States. As a result, use of geophysical techniques is becoming more common—particularly in research applications—as archaeologists come to realize the utility and efficiency of these powerful tools. The effectiveness of archaeogeophysics has not, however, been fully acknowledged in the laws, regulations, and standards that guide cultural resource management practices and excavation strategies. Archaeogeophysical technologies and specific applications are described in detail in subsequent chapters of this volume. As a point of reference, however, the field of archaeogeophysics encompasses a range of noninvasive methods for delineation and analysis of subsurface archaeological and cultural features (Clark 1996; Conyers and Goodman 1997; Kvamme 2001). Generally speaking, archaeological sites are the product of cultural and natural formation processes (Schiffer 1987). Sites are altered by anthropogenic activities and the natural accumulation of sediments. Soils are physically and chemically changed over time, and the archaeological record is transformed both

18 ~ J. J. Lockhart and Thomas J. Green spatially and quantitatively. These site formation processes produce a three-dimensional archaeological matrix or volume composed of topographic and physical properties such as soil texture, soil compaction, stratigraphy, biogenic and biochemical components, differential moisture retention, thermal alteration (burning), and artifact composition. Archaeogeophysical technologies provide the capability to measure the variable strengths and locations of physical properties that make up the archaeological record. The term cultural resource management (CRM), as used in this study, refers to the body of laws, standards, and practices that guide the management of prehistoric and historic properties within the context of modern research, preservation, and land planning. A broader definition of CRM also includes aspects of artifact curation and related document management (King 1998), but this chapter relates specifically to how the practice of archaeogeophysics is relevant to archaeological resource management in land planning activities and the preservation of archaeological properties. In the United States, federal and state laws require the consideration of historic properties in project planning and land management activities and provide for the protection of archaeological sites on public lands. The overarching law is the National Historic Preservation Act of 1966. Section 106 of that act states: The head of any Federal agency having direct or indirect jurisdiction over a proposed Federal or federally assisted undertaking in any State and the head of any Federal department or independent agency having authority to license any undertaking shall, prior to the expenditure of any Federal funds on the undertaking or prior to the issuance of any license, as the case may be, take into account the effect of the undertaking on any district, site, building, structure, or object that is included in or eligible for inclusion in the National Register. The head of any such Federal agency shall afford the Advisory Council on Historic Preservation established under Title II of this Act a reasonable opportunity to comment with regard to such undertaking.

In short, Section 106 requires federally assisted agencies to “take into account” the effects of their projects on historic properties—including archaeological sites—and give the Advisory Council on Historic Preservation the opportunity to comment on them. Section 110 of the same act further requires federal agencies involved in land planning and development activities to implement affirmative management programs designed specifically for the preservation of historic properties. In addition to Section 110, the Archaeological and Historic Preservation Act of 1974 requires federal agencies to document historic and archaeological properties that may be impacted by land management activities. In subsequent legislation, the Archaeological Resource Protection Act of 1979 expressly prohibits the unlawful destruction of archaeological sites located on federal land. The more recent Native American Graves Protection and Repatriation Act (NAGPRA) of 1990 requires federal agencies to consult with Native American tribes prior to the excavation of Native American graves on federal land. Taken as a whole, this body of law and associated regulations governs archaeological research methods and preservation issues associated with federally sanc-

Role of Archaeogeophysics in CRM ~ 19 tioned land management projects in order to mitigate adverse affects to archaeological sites. There are also state laws that govern archaeological research conducted on private and nonfederal public lands. In a 1983 response to the laws outlined above, the U.S. Department of the Interior and the National Park Service published The Secretary of the Interior’s Standards and Guidelines for Federal Historic Preservation Programs Pursuant to the National Historic Preservation Act (1983). This document provides general advice concerning “best practices” for compliance with historic property laws and regulations. Standards and Guidelines includes information about preservation planning, site identification, evaluation, registration, documentation, and professional qualifications. Because the Secretary of the Interior’s publication is intended to provide general guidance for preservation projects across the United States and its territories, many states also have developed more detailed standards and guidelines for archaeological research that address specific regional environmental and cultural situations. In order to assess the extent to which a land-use project might impact the cultural resources of an area, archaeological sites and other historic properties must first be identified and evaluated. One of the first steps in this process is to determine whether sites have been previously located in the project area and whether archaeological surveys have been conducted. To facilitate the initial identification of archaeological sites in a given project area, many states and federal agencies have developed computerized inventories of archaeological and historic properties (Figure 2.1). Integrated geographic information systems (GIS) and database management systems, such as the examples from Arkansas (Hilliard and Riggs 1986), contain attributes relating to site location, site size, National Register status, site features and function, and much more. Archaeological site databases provide cultural resource managers and other land planners information about where sites are and what they are. These statewide databases are continually updated, thereby making up an essential part of an efficient CRM process. However, the potential for discovering previously unrecorded archaeological sites within a project area is also an important consideration. Consequently, it is likewise useful to know where and how people have looked for sites—even if none were identified. Figure 2.2 shows a statewide GIS data layer for archaeological projects and surveys. Many of the areas shown have been systematically surveyed for Section 106 compliance. Attribute databases for archaeological surveys contain information about who did the survey, how they did the survey, and what was found. The dense distribution of known and as yet undiscovered archaeological sites in many parts of the United States—coupled with modern land-use trends, archaeological research, and CRM legal requirements—has driven the development of more powerful tools to effectively manage and protect cultural resources while accommodating construction, agriculture, and other changing land-use patterns. One of these emerging tools in American archaeology is archaeogeophysical survey technology. Geophysical technologies provide the capability to map and analyze subsurface archaeological features (Figure 2.3). Archaeogeophysical surveys are nondestructive by

20 ~ J. J. Lockhart and Thomas J. Green

Figure 2.1. GIS data layer and example database fields for archaeological sites in Arkansas. Source: Arkansas Archeological Survey Automated Management of Archeological Site Data in Arkansas (AMASDA) Database (Hilliard and Riggs 1986).

definition. Generally speaking, each type of device operates at or near the surface of the ground by measuring physical properties that have been created or altered by natural processes and/or past anthropogenic activities. Archaeogeophysical surveys are cost-effective in many situations. Conventional shovel-test methodology at standard intervals (i.e., 30 m) and at comparatively shallow depths may fail to locate even moderately large cultural landscape features. Alternatively, the sampling strategy of a geophysical survey is contiguous and uniform within a study area, commonly ranging from continuous pulses to 1-m sample and traverse intervals, depending on the device. Consequently, the results can provide higher order locational accuracy in locating and often identifying subsurface archaeological features before excavation. Intuitively, geophysical surveys should logically be a routine part of CRM wherever circumstances permit. However, there are still relatively few experts in the archaeological community and correspondingly few applications in the United States, so archaeogeophysical techniques do not currently have any formal status within laws, regulations, or guidelines for assessment of site significance. Moreover, partly because the public, potential clients, and some archaeologists remain largely unaware of geophysical capabilities, these powerful and proven tools are currently underused.

Role of Archaeogeophysics in CRM ~ 21

Figure 2.2. GIS data layer and example database fields for archaeological projects and surveys in Arkansas. Source: Arkansas Archeological Survey AMASDA Database.

Archaeological field methodologies in CRM are formalized, perpetuated, and refined ideally through accepted guidelines. So, how do existing laws, regulations, and standards relate to the use of archaeogeophysics in American CRM? There is little in the way of specific guidance. Interestingly, however, much of the language used in these documents, some of which were written decades ago, seems as though it were intended to outline important advantages of using archaeogeophysical technologies and methods in CRM applications. For example, the Advisory Council on Historic Preservation (ACHP) has published an expanded interpretation of Section 106 in its Protection of Historic Properties (2000), which provides for “nondestructive project planning activities before completing compliance with section 106, provided that such actions do not restrict the subsequent consideration of alternatives to avoid, minimize or mitigate the undertaking’s adverse effects on historic properties.” In many instances, archaeogeophysical methods constitute a nondestructive planning element, which enables archaeologists and land managers to make informed decisions and recommendations. Figure 2.4 shows the results of a geophysical survey within a federally funded project area that contains two nineteenth-century cemeteries (Mainfort 2004). According to law, a survey and mitigation plan was operationalized to locate graves for exhumation and reburial outside

22 ~ J. J. Lockhart and Thomas J. Green

Figure 2.3. Gradiometer data for a prehistoric feature.

the project area. This example illustrates that, added to the knowledge of where sites are (which is what our GIS and relational databases can tell us), archaeologists and land managers can now gather more detailed information about subsurface features at those sites by using geophysical technologies. Archaeological surveys and site mitigations associated with Section 106 compliance are commonly awarded by agencies to private archaeological consulting firms. However, the ACHP publication advises that “the agency official remains legally responsible for all required findings and determinations.” As agency officials become more aware of the utility and efficiency of archaeogeophysical techniques in locating subsurface features—thereby reducing their liability—these techniques will be increasingly integrated into conventional methodological practice in contract archaeology. To help illustrate this point, Figure 2.5 compares the results of two geophysical technologies for a 20-×-40-m area within a site that has prehistoric, historic, and modern components—a situation that is common in CRM applications. The imagery is positioned over a prehistoric earthwork. The magnetic susceptibility data on the left clearly show the location of anomalies that could be tested. Electromagnetic conductivity data for the same area, shown on the right, corroborate many of those same anomalies, with the addition of a linear subsurface feature. The site used in this example is located near a small town, and there were no known utility line maps to consult. The archaeogeophysical survey

Role of Archaeogeophysics in CRM ~ 23 was the only available source of information to suggest the possibility of an underground utility in the study area. Based on the results of the survey, a subsequent search of county records produced a single, hand-drawn map from 1964, which shows an abandoned utility line in this location. This example illustrates the advantage of using more than one technology, as well as the ability to locate subsurface features of multiple types, rather than encountering them without prior knowledge during excavation or construction. One section of the ACHP document under the heading “Consultation” reads, “The agency official shall involve the consulting parties Figure 2.4. Gradiometer data for a nineteenth- [and coordinate] with other requirecentury cemetery. ments of other statutes, as applicable.” These “consulting parties” include State Historic Preservation Offices, Native American tribes, the public, and so on, and “other statutes” include those such as the National Environmental Policy Act (NEPA) of 1969 and the Native American Graves Protection and Repatriation Act of 1990. As shown in Figure 2.6, archaeogeophysical techniques—through skillful and conscientious field methods and data processing and interpretation—can provide information about the location, size, and configuration of archaeological features (Lockhart et al. 2001). This type of detailed site information can be used to more fully communicate the scale and scope of an archaeological project to other concerned parties. In a related section, the ACHP stipulates that “the agency official shall seek and consider the views of the public in a manner that reflects the nature and complexity of the undertaking and its effects on historic properties.” Figure 2.7 illustrates the potential advantages of using archaeogeophysical imagery to communicate the nature and complexity of archaeological projects (Schambach and Lockhart 2003). The example image on the left depicts gradiometer data for a 110-m transect that contains at least five prehistoric structures and other archaeological features at a site in Arkansas. The photographs on the right of the figure show the living floor of one of those structures after test units were positioned based on the archaeogeophysical data. The geophysical survey in this example was completed in two days, and the location and size of these buried prehistoric structural remains were not apparent at ground surface. Traditional

24 ~ J. J. Lockhart and Thomas J. Green

Figure 2.5. Comparison highlighting the advantages of using multiple technologies.

site survey methodologies might have identified features in the area, but not with comparable detail and locational precision while preserving features in place for planned excavation. The Secretary of the Interior’s Standards and Guidelines for archaeology and historic preservation puts forth the following principle: “Important historic properties cannot be replaced if they are destroyed. Preservation planning provides for conservative use of these properties, preserving them in place and avoiding harm when possible.” Without question, the nondestructive nature of archaeogeophysical survey methodology should qualify as a conservative means of preserving and avoiding harm to important archaeological sites. Figure 2.8 compares data collected with four different technologies within the same 20-×-20-m area. Although each device measures various physical properties differently, each also corroborates many of the same anomalies on the site. Intuitively, in any sort of land-altering activity, one of the best methods for avoiding harm to archaeological deposits is to know where they are and what they are. Previous excavations of similar anomalies elsewhere on this site have provided baseline data for recognizing the geophysical signatures of prehistoric structures. Consequently, although the structure imaged in Figure 2.8 has not been excavated, it yields detailed information for intrasite settlement pattern analysis—including structure size and orientation. This example illustrates how archaeogeophysical surveys in combination with expert knowledge of local and regional archaeology can provide

Role of Archaeogeophysics in CRM ~ 25

Figure 2.6. Electrical resistance data and excavation on a prehistoric site in eastern Arkansas.

Figure 2.7. Gradiometer data and excavation on a prehistoric site in southwest Arkansas.

26 ~ J. J. Lockhart and Thomas J. Green

Figure 2.8. Geophysical signatures for an archaeological feature using multiple technologies.

land managers with vital information for project planning while minimizing direct impacts to archaeological features. Under the heading “Identification of Historic Properties Is Undertaken to the Degree Required to Make Decisions,” the Standards and Guidelines recommends “sampling an area to gain a broad understanding of the kinds of properties it contains” and further calls for “a record of the precise location of all properties identified.” As stated earlier, archaeogeophysical surveys are defined by consistent and systematic sampling through the use of measured sample and traverse intervals. Using equally spaced ropes of fixed length (e.g., 20 m) that have equidistant marks along each (e.g., 50-cm intervals), geophysical measurements (z-values) can be collected at prescribed intervals in both the x and y directions on the ground (Figure 2.9). A broad area can be sampled, and every measurement generated has a known location on the ground. Combined with accurate land survey technologies, such as a total station, higher order locational precision can be achieved. The generally raster-formatted data can then be used directly with other spatial data-management technologies, such as GIS. Figure 2.9 illustrates the precision with which test excavations can be placed based on near-surface remote sensing imagery. The example at the upper right, from northeast Arkansas, shows magnetic imagery for an anomaly interpreted as a prehistoric structure that has been bisected by an earthquake-induced liquefaction feature or sand blow (Payne and Lockhart 2002). A 2-×-2-m excavation unit was positioned on the ground based on the georeferenced

Role of Archaeogeophysics in CRM ~ 27

Figure 2.9. Field methodology and results from a prehistoric site in northeast Arkansas.

archaeogeophysical data. The photograph at the bottom right of Figure 2.9 shows the excavated feature, corroborating the inferred prehistoric structure bisected by an early historic or prehistoric sand blow. “Where possible,” the Standards and Guidelines for archaeology and historic preservation goes on, “use of quantitative methods is important because it can produce an estimate, whose reliability may be assessed, of the kinds of historic properties that may be present in the studied area.” Archaeogeophysical methodology is quantitative by definition. The technologies provide the means to measure precise proportions of geophysical properties underground. Also, importantly, in qualified hands the processes and results are objective and repeatable (Figure 2.10). The Secretary of the Interior’s document notes also that “survey methods should be carefully explained so that others using the gathered information can understand how the information was obtained and what its possible limitations or biases are.” This wording could be interpreted as a directive for educating professionals, clients, and the public about geophysical technologies and their capabilities. Standards and Guidelines does offer a broad opinion on all remote sensing technologies under the heading “Special Survey Techniques,” where it states, “Special survey techniques may be needed in certain situations. Remote sensing techniques may be the most effective way to gather background environmental data, plan more detailed field

28 ~ J. J. Lockhart and Thomas J. Green

Figure 2.10. Geophysical units of measure.

investigations, discover certain classes of properties, map sites, locate and confirm the presence of predicted sites, and define features within properties. Remote sensing techniques include aerial, subsurface and underwater techniques. Ordinarily the results of remote sensing should be verified through independent field inspection before making any evaluation or statement regarding frequencies or types of properties.” Also, under “Recommended Sources of Technical Information” there are references for applicable publications (Bevan 1998). In future iterations, the Secretary of the Interior’s Standards and Guidelines for archaeology and historic preservation may be an ideal forum for providing detailed and objective guidance for integrating archaeogeophysical methods into CRM field processes—both as a conventional survey tool and as a primary data source (Kvamme 2003) where applicable. The ACHP’s Recommended Approach for Consultation on Recovery of Significant Information from Archaeological Sites (1999) suggests that “appropriate treatments for affected archaeological sites, or portions of archaeological sites, may include active preservation in place for future study or other use, recovery or partial recovery of archaeological data, public interpretive display, or any combination of these other measures.” The use of geophysical technology to accurately locate and analyze archaeological features is beginning to receive recognition as a useful—even essential—component of “active preservation in place.” Figure 2.11 shows electrical resistance imagery for two confirmed prehistoric structures and other archaeological features. Using a GIS, the imagery has been georeferenced to the on-site coordinate system. Surveying equipment (e.g., a total station) has subsequently been used to accurately stake specific 2-×-2-m excavation units in the field. The use of archaeogeophysical technology to identify and pinpoint the location of subsurface features will eventually become an integral consideration in the “recovery or partial recovery of archaeological data” in all archaeological fieldwork.

Role of Archaeogeophysics in CRM ~ 29

Figure 2.11. Electrical resistance data and georeferenced 2-×-2-m grid for a prehistoric site in Arkansas.

Under the heading “Archaeological Sites and Their Treatment,” the ACHP admonishes that “methods for recovering information from archaeological sites, particularly large-scale excavation, are by their nature destructive. The site is destroyed as it is excavated. Therefore management of archaeological sites should be conducted in a spirit of stewardship for future generations, with full recognition of their non-renewable nature and their potential multiple uses and public value.” In this context, the value of archaeogeophysical methodology with its inherently nondestructive nature speaks for itself. Researchers and cultural resource managers now have the potential to recover detailed archaeological information from sites—small and large—without destroying them. Professionalism, sound field methods, computer skills, and experienced and knowledgeable interpretation can yield images and results that are invaluable in CRM, intrasite planning, and research. So, whether these powerful tools are viewed as a means of site discovery or as an important procedure following site discovery, archaeogeophysics must eventually become an integral part of American CRM. Like other technology-based fields of study, archaeogeophysics is still rapidly evolving in terms of hardware, software, methods, and theory. In the near term, mandating all aspects of its use within CRM law could have the unintended effect of hindering its evolution toward ever-increasing effectiveness and innovation. Moreover, requiring a geophysical survey on every CRM or archaeological project, irrespective of widely varying circumstances, would be counterproductive at this stage. Perhaps one

30 ~ J. J. Lockhart and Thomas J. Green appropriate place to emphasize the use of geophysical remote sensing in CRM while maintaining a level of operational discretion is in the Secretary of the Interior’s Standards and Guidelines for archaeology and historic preservation. As discussed earlier, this document briefly mentions “subsurface” techniques but stops short of emphasizing the effectiveness and efficiency of these nondestructive approaches in the identification and evaluation of archaeological properties. State Historic Preservation Offices and various federal agencies have developed more detailed standards and guidelines for archaeological fieldwork that are specific to regional circumstances and environmental conditions. These standards and guidelines could also be revised to provide objective guidance in the regional use of archaeogeophysical technologies, while maintaining flexibility for a wide range of possible environmental and research design variables. The Secretary of the Interior’s Standards and Guidelines requires “information on the appearance, significance, integrity and boundaries of each property sufficient to permit an evaluation of its significance” (Figure 2.12). Geophysical technologies are important tools that can provide a nondestructive, cost-effective, accurate means of gathering objective information on site appearance, significance, integrity, and boundaries. Although not yet fully integrated explicitly into CRM guidelines, the role of archaeogeophysics—both present and future—is clearly embodied by the spirit and intent of existing laws and standards guiding CRM in the United States.

Figure 2.12. Archaeogeophysical imagery from four technologies with excavated features.

Role of Archaeogeophysics in CRM ~ 31

Acknowledgments We would like to thank Kenneth L. Kvamme for sharing his knowledge and experience, Robert C. Mainfort and Dorothy G. Neely for their suggestions, and the Arkansas Archeological Survey and Society. We also wish to thank the Centers for Archaeological Research and Geoinformatics at the University of Mississippi, the Earth Science Applications Directorate at Stennis Space Center, NASA, and Jay K. Johnson and Marco Giardino for making the 2002 SEAC workshop and this publication possible.

References Cited Advisory Council on Historic Preservation 1999 Recommended Approach for Consultation on Recovery of Significant Information from Archaeological Sites. 64 FR 27085–87. 2000 Protection of Historic Properties; Final Rule. Federal Register. 36 CFR 800. Bevan, B. W. 1998 Geophysical Exploration for Archaeology: An Introduction to Geophysical Exploration. Special Report No. 1. U.S. Department of the Interior, National Park Service, Midwest Archeological Center, Lincoln, Nebraska. Clark, A. 1996 Seeing Beneath the Soil: Prospecting Methods in Archaeology, new ed. B. T. Batsford, London. Conyers, L. B., and D. Goodman 1997 Ground-Penetrating Radar: An Introduction for Archaeologists. AltaMira, Walnut Creek, California. Hilliard, J. E., and J. Riggs 1986 AMASDA Site Encoding Manual. Version 2.0. Technical Paper No. 1, W. Fredrick Limp, ed., Arkansas Archeological Survey, Fayetteville. King, T. F. 1998 Cultural Resource Laws and Practice: An Introductory Guide. AltaMira, Walnut Creek, California. Kvamme, K. L. 2001 Current Practices in Archaeogeophysics: Magnetics, Resistivity, Conductivity, and Ground-Penetrating Radar. In Earth Sciences and Archaeology, edited by P. Goldberg, V. Holliday, and R. Ferring, pp. 353–384. Kluwer/Plenum, New York. 2003 Geophysical Surveys as Landscape Archaeology. American Antiquity 68(3):435–458.

32 ~ J. J. Lockhart and Thomas J. Green Lockhart, J. J., J. M. Mitchem, and T. Mulvihill 2001 Geophysical Investigations at the Late Mississippian Parkin Site, Arkansas. Paper presented at the 58th Annual Meeting of the Southeastern Archaeological Conference, Chattanooga. Mainfort, R. C., Jr. (editor) 2004 Two Nineteenth Century Cemeteries in Crawford County, Arkansas. Report prepared for Burns and McDonnell. Sponsored Research Program, Arkansas Archeological Survey, Fayetteville. Payne, C., and J. J. Lockhart 2002 Cultural and Natural Landscapes at a Late Mississippian Site in the St. Francis Basin, Arkansas. Paper presented at the 59th Annual Meeting of the Southeastern Archaeological Conference, Biloxi. Schambach, F., and J. J. Lockhart 2003 The 2001–2002 Investigations by the Arkansas Archeological Survey and the Arkansas Archeological Society at the Tom Jones Site (3HE40), a Late 14th, Early 15th Century Caddo Mound Group in Southwest Arkansas. Paper presented at the 45th Annual Caddo Conference, Arkadelphia, Arkansas. Schiffer, M. B. 1987 Formation Processes of the Archaeological Record. University of New Mexico Press, Albuquerque. U.S. Department of the Interior 1983 The Secretary of the Interior’s Standards and Guidelines for Federal Historic Preservation Programs Pursuant to the National Historic Preservation Act. 63 Federal Register 20495-20508.


A Cost-Benefit Analysis of Remote Sensing Application in Cultural Resource Management Archaeology Jay K. Johnson and Bryan S. Haley

Contemporary archaeological research in the United States is largely motivated by legislation set in place to protect archaeological resources from modern construction activities. The threat of destruction to these nonrenewable resources is increasing as the population grows and its impact becomes greater. Today, a large segment of North American archaeologists specialize in compliance archaeology. The protection of archaeological resources requires an evaluation of resources within an impact area. This work is often expensive and takes a large amount of time. Remote sensing techniques have been developed that could make this task considerably more efficient. In this chapter we will briefly review the processes involved in cultural resource management (CRM) archaeology. This will be followed by an overview of the most successful of the most promising remote sensing techniques. We will conclude with an example drawn from our own research in the southeastern United States in which we will simulate the costs involved in a standard CRM data-recovery program and contrast that with what it would have cost if remote sensing had been used.

34 ~ Jay K. Johnson and Bryan S. Haley

CRM Archaeology Cultural resource management in the United States is mandated by law, primarily the National Historic Preservation Act (NHPA) of 1966. This requires that the impact to significant archaeological resources from activities conducted with federal money, on federal land, or involving federal permits must be considered. An archaeological resource is deemed significant if it is eligible for inclusion on the National Register of Historic Places. National Register status is determined by a set of guidelines included in Section 101 of the NHPA (U.S. Department of the Interior 1993). Prehistoric sites are usually deemed eligible on the basis of their potential to answer questions of scientific merit. In the southeastern United States, where most large prehistoric sites are located in areas that have been cultivated for a hundred years or more, significance often depends on whether house floors, wall trenches, hearths, pits, or burials remain intact and relatively undisturbed below the plow zone. Traditionally, three phases of work are used in these efforts (Georgia Council of Professional Archaeologists 2001; Sanders 2001). The first of these involves the discovery and assessment of the resources within an impact area, including site boundaries, cultural affiliation, and significance. Following a thorough examination of state site files and general background research, an intensive field survey is conducted to achieve these goals. This can sometimes be accomplished by walking a field at regular intervals while observing ground debris. However, in areas with ground cover, such as are often found in the southeastern United States, intensive survey usually consists of shovel testing along fixed intervals. The accepted interval is usually set by state guidelines but is typically 30 m in high-probability areas. The second phase of work is a further assessment of significant sites, if any, revealed during Phase I. The primary objective of this phase is a determination of National Register eligibility. Work may include closer interval shovel testing to establish the boundaries of the site. Also, a number of test pits are usually dug to gather data on stratigraphic integrity and to recover a larger sample of artifacts. Phase III work is the full-scale data recovery from sites that, on the basis of Phase II testing, are judged to be significant. Often projects are redesigned to avoid large and complex sites. If this is not an option, an effort must be made to recover as much data as possible, usually by means of large-scale excavation. Heavy machinery may be used to strip the plow zone and expose archaeological features. This phase is often an expensive undertaking and it can delay a construction project several months if the site proves to be more complex than suggested by Phase II assessment. The success of the traditional methods in accurately revealing archaeological resources has been debated (Kintigh 1988). While it is unlikely that large sites would be completely missed, it is likely that many features within a site have been missed because the sampling strategy is based on poorly recovered data on the distribution and density of such features. Remote sensing techniques offer several advantages that recommend them for CRM. First, they are time efficient. A magnetic survey can typically cover an area of

Cost-Benefit Analysis of Remote Sensing Application ~ 35

400 m2 or more per day. Resistivity, conductivity, and ground-penetrating radar are somewhat slower than magnetic survey but are still fast compared with most traditional survey techniques. Also, a geophysical survey allows total coverage of an impact area. In contrast, shovel testing and test pit excavation can only reveal data on a very small percentage of the site, and feature density and distribution must be projected from rather incomplete data. Remote sensing also fits well with more recent trends in preservation archaeology in the United States. Newer legislation, such as the Native American Graves Protection and Repatriation Act, provides protection for Native American cultural resources. The nonintrusive nature of remote sensing complies with these goals like no other archaeological data-recovery technique. On the other hand, the instruments and the software needed to process the data are relatively expensive. They demand a good deal more training than the more traditional techniques used by archaeologists. Additionally, we are just beginning to understand the conditions under which the various instruments are likely to produce useful results. Here the existing literature on geophysical prospection is not much use for many of these techniques since they were developed by geologists or environmental protection agency scientists to discover much different things. A cache of steel drums containing toxic waste or oil domes are generally deeper than and different from the house floors and hearths that archaeologists hope to find.

A Cost Simulation of Traditional vs. Remote Sensing–Based Data Recovery A hypothetical CRM project on a real site will illustrate the advantages and disadvantages of remote sensing in CRM archaeology. The Parchman Place site is a large late Mississippian site located in the Mississippi River alluvial valley of northwestern Mississippi. It has been the focus of a research project in remote sensing and archaeology that was begun in 2002 as part of the University of Mississippi archaeological field school. We returned to the site in 2003 and again in 2004. The site contains one very large platform mound and at least one other large mound, as well as an undetermined number of smaller mounds, some of which were incorporated by the larger ones. These traits suggest that the site was the center of a fairly complex political system with substantial differences in the status of its inhabitants. Moreover, the density of artifacts on the surface and the extent of their distribution indicate either a substantial population or a long occupation. Parchman is typical of the sort of large and complicated sites that might be encountered in road or reservoir construction in the southeastern United States. The Parchman Place site is also characteristic of many of the large prehistoric sites in the South in that all but the two largest of the mounds are in cultivation. Although thousands of artifacts can be found on the surface, at least the first 20 cm of the archaeological deposit have been completely churned up by decades of cultivation, most

36 ~ Jay K. Johnson and Bryan S. Haley recently with large and destructive machines. In terms of significance and legally mandated data recovery, it is the features that are located below this plow zone that would have to be located and excavated before the area could be cleared for construction. We began work at Parchman in February 2002 with a broad-scale magnetic survey of the majority of the site’s village area. During five very cold days, two crews completely covered an area of 5.6 ha located immediately to the south of the big mound using a pair of Geoscan FM36 fluxgate gradiometers. The majority of the surface artifacts at the site are located here, as are several areas where heavy concentrations of daub promise the likelihood of subsurface house remains. We did this work during the winter because this area of the site is prime agricultural land and is covered in cotton during the warmer months. The resultant image shows a remarkable number of structures and other prehistoric features (Figure 3.1). For the purpose of this simulation of Phase II and III research, we will assume that only the area that we covered using the gradiometer in February 2002 will be impacted by the hypothetical construction project. Six people spent a total of about 30 persondays collecting the data represented in Figure 3.1. As we are beginning to understand (Clay 2001; Kvamme 2003), the use of more than one geophysical technique is essential in any responsible application of remote sensing. What one instrument is able to detect is likely to be invisible to another. Figure 3.1. Magnetic gradiometer survey of the village portion of the Therefore, additional instruments should Parchman Place site (surveyed with the assistance of Berle Clay).

Cost-Benefit Analysis of Remote Sensing Application ~ 37

be applied to targeted areas of the site. We have done that at Parchman, using a Geonics model EM38B conductivity meter, a Geoscan RM15 resistance meter, and a Geophysical Survey Systems SIR-2000 ground-penetrating radar system, among other instruments. At Parchman, the three named instruments showed much the same thing as the gradiometer (Figure 3.2). However, because we are still in the development stages in geophysical research, where redundancy is a virtue, let us add another 15 person-days to the geophysical prospection phase of the work in order to gather multiple-instrument coverage of areas targeted using the magnetic data. Allow 15 person-days to process the imagery, bringing the total investment to 60 person-days. The result at Parchman is a very clear idea of the general areas where features are located, as well as the specific locations of many houses and pit features. A major focus of our fieldwork at the site has been the ground truth excavations of several features. In July 2002, a 40-×-40-m area to the south of the mound was selected and four excavation units were opened (Figure 3.3). The remains of three houses and one prehistoric pit were uncovered (Figure 3.4). For the sake of this simulation, we will propose the gradiometer image (Figure 3.1) as an intermediary stage in a Phase II project. More ground truth excavations would need to be done in order to produce the data that would be needed in designing a full-scale excavation of the site, but the same excavations would be needed in a Phase II project using traditional techniques. The question, then, is how much would it cost to derive a map of the location of houses and pits with as much detail using the traditional techniques? The first step would be delimiting site boundaries and discovering the areas within those boundaries where the relative density of artifacts indicates the likelihood of subsurface features. Two approaches are traditionally employed in CRM archaeology. On a site that Figure 3.2. Survey of buried prehistoric house remnants at Parchman Place has been cultivated, with electromagnetics (top left), resistance (top right), ground-penetrating a controlled sur- radar (bottom left), and magnetic gradiometer (bottom right).

38 ~ Jay K. Johnson and Bryan S. Haley

Figure 3.3. Ground truth excavation units superimposed on magnetic gradiometer survey.

Figure 3.4. Trenches superimposed on magnetic gradiometer survey showing burned oor (lower right) and charred beams (upper left).

Cost-BeneďŹ t Analysis of Remote Sensing Application ~ 39

face collection is usually used. That is, crop stubble is thoroughly disced under, a grid is laid out, and, following sufficient rain to expose the artifacts, a sample of the artifacts found on the surface is collected and counted and the data fed into a computer-based density contour mapping program. As anyone who has employed this approach will testify, the need for rain following ground preparation is often a major diďŹƒculty, even in a place like the southeastern United States, where thunderclouds are a regular feature of the afternoon sky. Our research in remote sensing and archaeology in the Yazoo Basin began at the Hollywood Mounds (Johnson et al. 2000), a site similar to Parchman located about 50 km to the north. A 10-percent sample controlled surface collection took an eight-person crew three weeks. This was fol- Figure 3.5. Surface artifact density plot of the Hollywood site. lowed by several weeks of washing and counting the artifacts, all to produce a surface density plot in which the relationship between the surface distribution of the artifacts and the subsurface features was blurred by decades of plowing. A comparison of the surface density plot (Figure 3.5) with the gradiometer images of the site (Figure 3.6) supports this statement. A total of about 5 ha at Hollywood took 240 person-days. At the same rate, the controlled surface collection at Parchman would take 269 person-days of work. If the Parchman Place site had been in pasture or woods, a controlled surface collection would not be possible. The standard approach to preliminary Phase II investigation

40 ~ Jay K. Johnson and Bryan S. Haley in this case would be shovel testing. Although the prescribed interval varies, 20 m is fairly representative. It would take 140 shovel tests to cover the same area that we recorded at Parchman in February 2002. Using standard guidelines (25 shovel tests per person per day), this would cost 5.6 person-days. Again, the recovered material would have to be washed, classified, and counted, contributing another 5.6 person-days. The resultant density map would show far less detail than the one based on the controlled surface collection. Nonetheless, all three techniques—magnetometry, controlled surface collection, and shovel testing—would allow the specification of areas of the site where subsurface features are likely to be found. The next step in a traditional Phase II research project would be to excavate 1-m2 Figure 3.6. Magnetic gradiometer survey of the Hollywood site test pits in those areas of (geophysical survey by Berle Clay). relatively high artifact density in order to determine whether subsurface features exist and where and how many there are. These data are critical in designing the Phase III, intensive excavation, portion of the project. The gradiometer image of Parchman shows evidence of at least 30 houses and 42 probable pit features (Figure 3.7). In order to simulate the test pit phase of a traditional project, we have outlined three areas of the site where features congregate, which would likely be reflected by a higher density of surface artifacts. These are the areas that would be discovered using a controlled surface collection or fixed-interval shovel tests. These

Cost-Benefit Analysis of Remote Sensing Application ~ 41

areas, also shown in Figure 3.7, amount to 1.827 ha. A random sample of 1 percent would consist of 183 test pits. Our July 2002 excavations indicated the latest house floors at the site occur at 20 cm and the sterile river sand deposits show up at between 50 and 80 cm. Taking the shallower depth to include those test pits that would come up empty, about 488 persondays would be needed to conduct these excavations. At least 976 days in the lab would be needed to process the artifacts. Using a geographic information system (GIS), we ran a se- Figure 3.7. Magnetic gradiometer survey of the village portion of the ries of 20 simula- Parchman Place site showing houses (red polygons), pits (blue), and hightions (Table 3.1) in density areas (green polygons) (color illustration appears on the CD). which 1-m excavation squares were imposed over a map of features derived from the gradiometer images. The fieldwork and lab work involved in this hypothetical 1-percent sample amounts to an estimated 1,468 person days. The accuracy of the prediction could be improved by increasing the sample size with a corresponding increase in person-hours. Even if the sample were increased to the point where the predicted number of features stabilizes and begins to approach the known number of features, we would still only know how many there were but not where the majority of the features were located. A comprehensive feature documentation usually occurs during Phase III research when large machines, such as road graders or belly pans, are brought in to strip

42 ~ Jay K. Johnson and Bryan S. Haley Table 3.1. Test excavation simulation results Run 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Minimum Maximum Standard deviation Mean Actual features

Area 1 9 6 7 11 9 10 7 10 11 9 6 7 7 12 5 11 10 7 6 6

Houses Area 2 3 4 3 2 4 3 3 4 3 4 4 2 4 1 2 2 3 2 4 3

Area 3 3 4 5 4 3 6 4 4 2 4 4 5 5 4 3 5 3 5 3 4

5 12 2.13 8.3 18

1 4 0.92 3 4

2 6 0.97 4 8

Area 1 3 5 0 2 1 4 2 4 2 2 2 3 1 1 5 4 1 3 2 1

Pits Area 2 3 0 3 3 1 2 1 3 2 1 2 3 0 2 1 1 2 3 2 5

Area 3 1 1 2 1 2 0 0 2 1 0 0 1 1 0 1 1 0 0 2 3

0 5 1.4 2.4 17

0 5 1.2 2 18

0 3 0.76 0.84 7

Note: Areas are those shown in Figure 3.7.

the plow zone. In addition to the cost of the heavy equipment and operators, a large ďŹ eld crew is needed to clean and record the features that are exposed. Using a conservative estimate of three person-days per house and one-half person-day per pit feature, the hypothetical Parchman plow zone removal work would take 111 person days. Add a conservative 20 days of machine time for this operation and the cost becomes quite large. At $60 per hour for machine and operator, this totals $9,600. At this point, under ideal weather conditions, the map produced by traditional techniques would approach or exceed the image we derived during one week in February at the site. In terms of expense, there is no comparison. Converting person-days to hourly wages plus fringe beneďŹ ts so that heavy-equipment costs could be added in (Table 3.2), the project would range from $169,892 to $195,878 using traditional means and $6,048

Cost-Benefit Analysis of Remote Sensing Application ~ 43 Table 3.2. Cost simulation of traditional vs. remote sensing–based data recovery Remote Sensing 60 Person-days × 8 hours × $10.00/hour Fringe benefits ($4,4000 × 26%) Total

$4,800 $1,248 $6,048

Remote sensing total

$6,048 Traditional Survey

Controlled surface collection 269 Person-days × 8 hours × $10.00/hour Fringe benefits ($21,520 × 26%) Total OR Shovel tests 11.2 Person-days × 8 hours × $10.00/hour Fringe benefits($896 × 26%) Total AND Test pits 1,468 Person-days × 8 hours × $10.00/hour Fringe benefits ($117,440 × 26%) Total AND Plow zone removal 111 Person-days × 8 hours × $10.00/hour Fringe benefits ($18,800 × 26%) Heavy equipment Total Traditional survey methods total (with shovel tests) Traditional survey methods total (with controlled surface collection)

$21,520 $5,595 $27,115

$896 $233 $1,129

$117,440 $30,534 $147,974

$8,880 $2,309 $9,600 $20,789 $169,892 $195,878

using remote sensing. There is a hypothetical savings of as much as $189,830, and we haven’t even considered overhead!

Why Aren’t More CRM Archaeologists Using Remote Sensing? Given the above figures and our results at the two late prehistoric sites that we have investigated using remote sensing, it would seem to be amazing that remote sensing is not an essential part of CRM archaeology in the United States. There are some obstacles to the use of remote sensing techniques by archaeologists. Instruments require fairly substantial investments of money for the initial purchase. However, this money is often quickly made up in terms of savings in field time, and equipment is becoming much

44 ~ Jay K. Johnson and Bryan S. Haley less expensive. A substantial amount of training is also required to effectively use the equipment and successfully interpret the data. Remote sensing workshops specifically for archaeologists are becoming much more common, making this more feasible. Attitudes toward the new technology are sometimes also a barrier. Some archaeologists have held the view that remote sensing is a competitor of and possible replacement for conventional excavation methods. However, it is simply a new tool that can make the job of the archaeologist easier. Traditional excavation techniques yield information that, in the foreseeable future, cannot be revealed with remote sensing. Also, remote sensing has often been oversold by the people who use it, creating false expectations. One technique becomes the vogue—it was digital remote sensing, is now ground-penetrating radar, and, can anyone doubt, will soon be gradiometry—and that technique is applied indiscriminately, regardless of the kinds of features that are sought or the kinds of soil in which they are buried. After one or two large projects in which a research design is based on a single remote sensing technique inappropriately applied with few or no results, project managers are more than willing to return to the triedand-true techniques they grew up with. One additional obstacle is the idea that remote sensing should never be used to rule out the existence of archaeological features in a study area. Some constraints, such as weather and modern cultural features, may prevent the collection of good data sets with geophysical instruments. In cases in which clients are eager to begin their construction projects, a negative result may be incorrectly viewed as a license to begin the project. However, we must make it very clear to clients that a negative survey should not be used in this way. Therefore, there is an inherent risk to using remote sensing in archaeological projects. Nevertheless, remote sensing regularly provides information that allows much more comprehensive data-recovery planning.

Conclusions Clearly, remote sensing techniques can be successfully implemented within the context of CRM work. As we have argued, it is efficient in terms of both cost and time. Archaeologists are beginning to be exposed to the new technology and are realizing its merits. Recent state guidelines, in fact, have encouraged the use of remote sensing techniques (Georgia Council of Professional Archaeologists 2001; Sanders 2001). Undoubtedly, remote sensing will play a major role in CRM archaeology in the future. It is, of course, the purpose of this book to promote that end by providing management archaeologists with a more realistic understanding of when and how to use this technology.

Acknowledgments The research at the Parchman Place Mounds and at the Hollywood Mounds was supported through grants and participation by several agencies and individuals. Primary among them are Marco Giardino of the Earth Science Applications Directorate at NASA’s Stennis Space Center, Berle Clay of Cultural Resource Analysts, Inc., and

Cost-Benefit Analysis of Remote Sensing Application ~ 45

John Connaway of the Mississippi Department of Archives and History. Funding was also provided by the University of Mississippi Geoinformatics Center. The Hollywood Mounds were donated to the Mississippi Department of Archives and History by Neal Block and the Parchman Place Mounds are owned by the Archaeological Conservancy. We thank both of these agencies for allowing us to conduct fieldwork on these sites. The village area at Parchman Place is owned by Joe Noe, who allowed our research to take place in the middle of a very hectic growing season.

References Cited Clay, R. B. 2001 Complementary Geophysical Survey Techniques: Why Two Ways Are Always Better than One. Southeastern Archaeology 20:31–43. Georgia Council of Professional Archaeologists 2001 Georgia Standards and Guidelines for Archaeological Surveys. Georgia Council of Professional Archaeologists. Johnson, J. K., R. Stallings, N. Ross-Stallings, R. B. Clay, and V. S. Jones 2000 Remote Sensing and Ground Truth at the Hollywood Mounds Site in Tunica County, Mississippi. Center for Archaeological Research, University of Mississippi, Oxford. Submitted to the Mississippi Department of Archives and History. Kintigh, K. W. 1988 The Effectiveness of Subsurface Testing: A Simulation Approach. American Antiquity 53:686–707. Kvamme, K. L. 2003 Multidimensional Prospecting in North American Great Plains Village Sites. Archaeological Prospection 10:131–142. Sanders, T. N. (editor) 2001 Specifications for Conducting Fieldwork and Preparing Cultural Resource Assessment Reports. Kentucky Historic Preservation Office, Kentucky Heritage Council, Site Protection Program, Frankfort. U.S. Department of the Interior 1993 Federal Preservation Laws. U.S. Department of the Interior, National Park Service, Cultural Resources Programs, Washington, D.C.


Airborne Remote Sensing and Geospatial Analysis Marco Giardino and Bryan S. Haley

Cultural resource management (CRM) consists of research to identify, evaluate, document, and assess cultural resources; planning to assist in decision making; and stewardship to implement the preservation, protection, and interpretation of these decisions and plans. Traditionally, archaeological methods used to accomplish these goals are time consuming, labor intensive, and expensive. Moreover, they rely on sampling strategies that can lead to an inaccurate assessment of cultural resources. One technique that may be useful in CRM archaeology is remote sensing. Remote sensing is generally defined as the acquisition of data and derivative information about objects or materials (targets) located on the earth’s surface or in its atmosphere by using sensors mounted on platforms located at a distance from the targets to make measurements on interactions between the targets and electromagnetic radiation (Ebert and Lyons 1983; Giardino and Thomas 2002; Lyons and Avery 1977; Short 1982). Included in this definition are systems that acquire imagery by photographic methods and digital multispectral sensors, which are the core of the modern remote sensing industry. Today, data collected by digital multispectral sensors on aircraft and satellite platforms play a prominent role in many earth science applications, including land cover mapping, geology, soil science, agriculture, forestry, water resource management, urban and regional planning, and environmental assessments (Lillesand and Kiefer 1994). These systems often employ sensors that record discrete segments of electromagnetic energy well beyond film, such as thermal infrared. Such systems can rapidly accumulate detailed information on ground targets.

48 ~ Marco Giardino and Bryan S. Haley Inherent in the analysis of remotely sensed data is the use of computer-based image processing techniques, which enhance the interpretability of remotely sensed data. Desktop computing power has become less expensive and more powerful, and image processing software has become more accessible, more user friendly, and fully capable of even the most sophisticated processing of digital data, like that collected during remote sensing missions. Geographic information systems (GIS), systems designed for collecting, managing, and analyzing spatial information, are also useful in the analysis of remotely sensed data. A GIS can be used to integrate diverse types of spatially referenced digital data, including remotely sensed data in raster format and supplementary vector map data. In archaeology, these tools have been used in various ways to aid in CRM projects. For example, they have been used to predict the presence of archaeological resources using modern environmental indicators. Remote sensing techniques have also been used to detect the presence of unknown sites based on the impact of past occupation on the earth’s surface. Additionally, remote sensing has been used as a mapping tool aimed at delineating the boundaries of a site or mapping previously unknown features. All of these applications are pertinent to the goals of site discovery and assessment in CRM.

The Beginnings of Aerial Prospection The beginnings of the use of remote sensing in archaeology date to the early twentieth century and were centered in Britain (Wilson 1982:10). In 1906, balloon-based photographs were taken of Stonehenge. A more substantial contribution was made by O. G. S. Crawford beginning in 1921. Crawford recognized that subsurface archaeological features could be detected with aerial photography, and he produced a large series of aerial photographs. At about the same time, Charles Lindberg photographed a number of Mayan sites, including Tikal, Tulum, and Chichén Itzá, from his aircraft (Lindberg 1929). A number of then unknown sites in the eastern Yucatan were first identified using these photographs. The first aerial reconnaissance of a North American site occurred at Cahokia in the early 1920s. Oblique photographs produced by Goddard and Ramey were the first to be published (Crook 1922). The photographs contain a substantial amount of information about the structure of the site and are still a valuable source of information today (Fowler 1977:65). From these early attempts, the advantages of an aerial perspective were apparent (Lillesand and Kiefer 1994:49–50). The improved vantage point of aerial photographs allows us to see ground objects, such as archaeological features, in an expanded spatial context. Moreover, the patterns of archaeological features, which are often geometric and regular but invisible on the ground, may be revealed. In addition, aerial photographs allow a permanent record of ground targets. Perhaps most important, large areas can be surveyed and mapped very rapidly.

Airborne Remote Sensing and Geospatial Analysis ~ 49

Aerial Photography Black and White Standard photographic film is composed of a silver halide emulsion coating that reacts to light intensity (Scollar et al. 1990:89). Light causes a photochemical reaction in the silver halide crystals that produces a latent image (Lillesand and Kiefer 1994:53). Aerial photography is essentially a broad-band, panchromatic remote sensing technique covering the visible portion of the electromagnetic spectrum (Figure 4.1). This spans wavelengths from about 0.4 μm (micrometers or 10–6 meters), which corresponds to violet, to 0.7 μm, which corresponds to red. Within this broad range, electromagnetic energy cannot be differentiated, however. The tradition of aerial photography in archaeology has been carried on most strongly in Britain, where it is a primary method for site discovery and reconnaissance in cultural resource projects. One representative example is offered by Featherstone (1999), who describes a large-scale site survey done in England by the Royal Commission of Historical Monuments. During a particularly dry summer, approximately 415 flight hours were logged spanning a large portion of England and Scotland. A total of 4,570 targets were photographed and identified in crop marks, with about half representing unknown sites and 15 percent contributing new information to known sites. Site types detected included Bronze Age barrows, causeways, Iron Age enclosures, Roman field systems, Roman road systems, Neolithic mortuary enclosures, henges, ring ditches, hill forts, barrows, fortresses, and earthworks covering a multitude of construction types and time periods. Although the method is less common in the United States, there are numerous examples of its use. In the Southwest, Lyons and Hitchcock (1977) describe mapping an Anasazi road system that spanned over 250 miles. In the Eastern Woodlands, Carskadden (1999) describes an application of aerial photography in mapping late Adena and early Hopewell earthworks at the Gilbert site in Muskingum County, Ohio. Other examples include additional work at Cahokia (Fowler 1977), the mapping of lodge depressions along the Knife River in Minnesota and North Dakota (Thiessen 1993), and the identification of earthworks and borrow pits along the Kissimmee River in Florida (W. Johnson 1994).

Figure 4.1. Electromagnetic spectrum (based on Lillesand and Kiefer 1994).

50 ~ Marco Giardino and Bryan S. Haley Color In contrast with black and white, color films employ a subtractive process, which uses several layers of dye that are each responsive to certain wavelengths of energy (Scollar et al. 1990:107). Color film can provide significantly more information than black-and-white film since the human eye can differentiate many more tints of color than shades of gray (Ebert 1984:315). However, it is more difficult to interpret patterns in color photographs. Color photography has traditionally been less often used than black and white in archaeological research, although it is becoming more common. Near Infrared The development of color infrared (CIR) film in the 1940s was a significant advance in aerial photography. CIR film is similar to color films except dyes are sensitive to green, red, and near infrared energy (Scollar et al. 1990:110). CIR film can offer additional valuable information because near infrared wavelengths are sensitive to differences in vegetation health and moisture patterning (Riley 1987:56). One example of the use of CIR is the detection of footpaths in the Arenal region of Costa Rica (Sheets and Sever 1991; McKee and Sever 1994; McKee et al. 1994). The footpaths were first detected as a set of lineaments in 1985 during analysis of a set of large-format CIR photographs acquired by NASA. The anomalies were primarily visible as positive vegetation marks in grassy ground cover. Applications in the United States include work at Fort Mims in southern Alabama (Riccio and Gazzler 1974) and at the Nanticoke village of Chicone in eastern Maryland (Davidson and Hughes 1986).

Multispectral Digital Sensors Like film-based systems, multispectral digital sensors operate by sensing electromagnetic energy, which propagates through space in the form of a wave. All objects reflect and absorb various wavelengths of electromagnetic energy at temperatures above absolute zero. For example, a leaf strongly reflects energy in the infrared area and moderately reflects energy in the green area, while it absorbs energy in the blue and red areas (Limp 1993:186). The human eye is a sensor that detects electromagnetic energy from approximately 0.4 to 0.7 μm in wavelength. Since green energy is within the visible spectrum, we can detect a leaf with our eyes. However, the entire range of electromagnetic energy extends well beyond the visible range. Remote sensing instruments, normally radiometers and scanners, can be designed to sense energy beyond the range of the human eye (Lillesand and Kiefer 1994:9). The electromagnetic energy of a ground target is directed to an array of detectors by some optical device, where it is absorbed. The size of the area sensed is called the instantaneous field of view (IFOV), which is usually expressed as an angle (Lillesand and Kiefer 1994:355). The intensity of the energy received is subsequently converted into a digital value or brightness value (BV). Once in digital form, the brightness value is stored in

Airborne Remote Sensing and Geospatial Analysis ~ 51 a matrix with each value representing an area of the earth’s surface, and these can be viewed as a raster image. Unlike active remote sensors such as radar and lidar, which provide their own energy, passive remote sensors collect energy that is naturally occurring. This energy may be reflected energy resulting from the interaction of solar energy and the earth’s features. Reflected energy makes up the visible and near infrared portion of the electromagnetic spectrum (EMS). Alternatively, the energy may be emitted from a target as thermal infrared energy. Remote sensing systems are often multispectral, which means they detect energy across several discrete segments of the EMS. The particular segment of the EMS sensed is determined by the materials used in each detector in an array. Remotely sensed targets are wavelength dependent, which means that, even within a given feature type, the proportion of reflected, transmitted, and absorbed energy will vary at different wavelengths. Thus, two features that are identical at one wavelength may be different in another area of the EMS (Lillesand and Kiefer 1994:13). Each type of material on the earth has a characteristic response curve that varies when one views the energy along the EMS. Therefore, remote sensing can be extremely useful for determining information on ground cover. Furthermore, multispectral bands are variably sensitive to target phenomena (Lillesand and Kiefer 1994:17–18). The Landsat Thematic Mapper sensor is a good example. Landsat band 1 (0.45–0.50 μm) covers the blue portion of the visible spectrum and can discriminate between soil and vegetation. Band 2 (0.50–0.57 μm) covers the green area and is excellent at assessing plant health. The red band 3 (0.61–0.70 μm) can be used to determine chlorophyll absorption. Band 4 (0.70–0.90 μm) senses near infrared energy and can determine vegetation type, vigor, biomass content, and soil moisture. The mid-infrared band 5 (1.55–1.75 μm) is sensitive to the turgidity or amount of water in plants. Such information is useful in crop drought studies and in plant vigor investigations. In addition, this is one of the few bands that can be used to discriminate between clouds, snow, and ice, which is important in hydrologic research. On the other hand, the second mid-infrared band (2.08–2.35 μm), assigned to band 7 because it was added late in the mission, is an important band for the discrimination of geologic rock formations (Lillesand and Kiefer 1994:468). It has been shown to be particularly effective in identifying zones of hydrothermal alteration in rocks, in vegetation stress analysis, and for soil mapping. The thermal infrared band (10.4–12.5 μm) measures the amount of infrared radiant flux emitted from surfaces and is useful for locating geothermal activity, for thermal inertia mapping for geologic investigations, and for vegetation classification, vegetation stress analysis, and soil moisture studies (Jensen 1986:34). In addition, because the narrow spectral range of multispectral sensors makes each band sensitive to specific target phenomena, they have the potential of detecting much more subtle features. Also, the options are greater both in manipulating the data and in the capability of seeing electromagnetic energy beyond that detected with film. Work

52 ~ Marco Giardino and Bryan S. Haley in the mid-infrared and the shortest segment of the thermal infrared is now showing promise for detecting features of archaeological interest. Digital image processing techniques often vastly enhance the interpretability of remote imagery. Also, the ability of a display system like the computer screen to load a variety of bands in the RGB (redgreen-blue) video guns provides added flexibility for interpretation. Narrow-band imagery, properly calibrated and used in indices, can assess plant vigor and plant stress. Specifically, the mid-infrared region between 1,300 and 2,400 nm offers promise for this task because it is the main absorption band for leaf water. Water-stressed plants have increased reflectance in this wavelength region. The narrow bands of hyperspectral sensors may further increase the utility of remote sensing by aiding in identifying sites through the identification of plant health and stress. Multispectral sensors can also be useful in archaeological applications by typing vegetation. The association of unique vegetation communities with geological, ecological, and archaeological sites is well documented (Eleuterius and Otvos 1979; Penfound 2001). Past occupation can alter the chemical properties of the soil and certain plants may be more adapted to such conditions. When distinct ground cover is consistently associated with archaeological deposits, it may be possible to detect archaeological sites in remote imagery. One excellent example is the shell mounds and middens that are common in coastal Louisiana and Mississippi. Eleuterius and Otvos (1979) report that several species, including red mulberry, coral bean, and buckeyes, found in association with these features are calciophiles, whose presence is “favored and determined by the large amount of calcium” in clam shells. Shell mounds also support a variety of shrubs and woody vines and a number of herbs and grasses that are not found in the marsh. Conversely, the hard substrate formed by buried shells may stunt root development and may result in differences between on-site and off-site plants that are significant enough to allow mapping of buried shell middens from aerial imagery. Also, oak trees may be markers for archaeological sites, particularly in the marshes where sites are often the only ground elevated enough to support these trees. Like coastal Louisiana and Mississippi, other regions that exhibit a number of distinct surface characteristics may be particularly well suited to this approach. Furthermore, hyperspectral sensors may increase the effectiveness of plant species as discriminators of archaeological sites. A series of vegetation variability indices can be determined using image processing techniques and a difference between the site and the surrounding area may be visible. It may be hypothesized that archaeological sites exhibit more variability in plant species than nonarchaeological sites as a reflection of centuries of human activity, including the collection of plants, firewood, and canes. In the absence of drastic ecological changes, it may be postulated that these plants have continued to germinate and to flourish on specific sites. Indices of vegetation variability may provide the evidence for testing such hypotheses, thereby providing practical methods for identifying sites in vegetated areas.

Airborne Remote Sensing and Geospatial Analysis ~ 53

Thermography Thermal infrared energy behaves much differently than reflected energy and therefore represents a unique topic. Thermal infrared energy is emitted from an object, such as the earth, instead of being reflected. The phenomenon that makes thermal bands valuable is that target materials heat and cool variably. More specifically, the thermal behavior of a target is determined by several quantities, which include thermal conductivity, density, and specific heat. These determine how a material stores heat and how readily heat flows through it (Lillesand and Kiefer 1994:381). In a layered earth, the thermal properties of each material and the subsurface thermal gradient are all relevant. A convenient measure, thermal inertia, can be derived from the aforementioned quantities and is inversely proportional to the response of the ground to thermal energy. Thermal inertia values for a number of common substances are shown in Table 4.1. A very basic understanding of how an archaeological target will behave thermally can be gained by considering thermal inertia values. In the morning, as the sun’s heat is focused toward the ground, a subsurface feature may be detected as a positive or negative anomaly. For example, a feature that enhances drying, such as some Mississippian houses, would be visible as a positive anomaly in the morning and a negative anomaly in the evening (Figure 4.2). Conversely, a feature that traps moisture, such as a pit, will result in a negative anomaly in the morning because moisture effectively lowers the thermal inertia of the pit feature. In the evening, this situation will be reversed since Table 4.1. Thermal inertia values for the thermal gradient will be from the ground common materials to the atmosphere. As with any prospection Material Thermal Inertia technique, archaeological features may be de(P) tected with thermal prospecting only if the Basalt 0.053 physical properties of the feature differ enough Clay soil (moist) 0.042 to cause a visible contrast in the imagery. Granite 0.056 Early works involving thermography inGravel 0.033 clude those of Berlin (1977; Berlin et al. 1975) Limestone 0.045 Marble 0.056 and Perisset and Tabbagh (1981), who quantiObsidian 0.035 fied the thermal behavior of buried targets. AnQuartzite 0.074 other example of thermal sensing is the work by Sandstone 0.075 NASA with the Thermal Infrared Multispectral Sandy gravel 0.050 Sensor (TIMS) at the Late Archaic Poverty Sandy soil 0.024 Point site in eastern Louisiana (Gibson 1987; Serpentine 0.059 Sever and Wiseman 1985). The six thermal Shale 0.041 bands of the TIMS revealed several anomalies Slate 0.049 of interest at the site. As determined by later Water 0.036 ground truthing, these were caused by borrow pits, fill episodes, a ramp, and a corridor. Note: Based on Sabins 1997.

54 ~ Marco Giardino and Bryan S. Haley

Resolution When describing remote imagery, it is helpful to characterize several types of resolution. Spatial resolution is the ability of an imaging system to record detail, or the size of the minimum pixel resolved by the sensor. It is also referred to as ground resolution since it describes an area of the earth’s surface. This quantity determines an instrument’s ability to resolve different size parcels (or pixels) of land or water. Since sensors record a fixed number of digital values for an IFOV, spatial resolution is finite. Thus, the IFOV and array size are closely related to the spatial resolution. Low spatial resolution sensors, such as the GOES (Geostationary Operational Environmental Satellites), have high orbits and relatively coarse ground resolutions (about 1-km pixels in the visible bands). They can image an entire hemisphere of the earth and are used widely as weather satellites. Sensors with moderate spatial resolutions, like the Landsat MSS and TM instruments with ground resolutions between 79 m and 30 m, provide regional coverage and have been used extensively in archaeology for landscape analysis and predictive modeling (Custer et al. 1986; Johnson 1991; Limp 1993). High spatial resolution sensors are becoming much more common and collect data useful at the local level. These include the French SPOT-5 (5-m panchromatic and 10-m multispectral), the Indian Remote Sensing Program’s IRS-1D (5.8-m panchromatic), Space Information’s SPIN 2 (2-m panchromatic), formerly classified Russian satellites (approximately 1-m panchromatic), Space Imaging’s IKONOS (1-m panchromatic and 4-m multispectral), and Digital Globe’s QuickBird (0.61-m panchromatic and

Figure 4.2. Diurnal temperature variation of a hypothetical Mississippian house from experimental measurements at the Hollywood Mounds site (based on Haley et al. 2002).

Airborne Remote Sensing and Geospatial Analysis ~ 55 2.44-m multispectral). Paired with new techniques of image analysis, this technology may make the direct detection of archaeological sites a realistic goal. The ability of passive remote sensing instruments to collect energy in specific wavelengths defines the sensor’s spectral resolution and thereby its ability to discriminate between objects based on the materials’ spectral response curves or patterns. Within each band of a sensor, energy is undifferentiated and a target’s spectral properties are indistinguishable. Thus the size and number of bands that a sensor utilizes determines its spectral resolution (Limp 1993:186). A sensor with a higher spectral resolution can differentiate between energy sources better than a sensor with a lower spectral resolution. However, the available energy is a limiting factor on spectral resolution because, as sensor bands become narrower, detectors collect less energy. Radiometers and scanners that are able to record energy in relatively broad bands, normally defined as 10 μm wide, are denoted as multispectral scanners. Landsat MSS, TM, and ETM, the SPOT sensors, the IRS sensors, and those mounted on the newer commercial systems like IKONOS and QuickBird are multispectral. Passive sensors that collect energy in narrow bands, defined normally as about 10 nm (nanometers or 10–9 m) wide, are known as hyperspectral sensors. The newer hyperspectral sensors, such as Hyperion, have only recently been deployed in orbit. However, they have been used on research aircraft for several years. NASA’s Jet Propulsion Laboratory operates an instrument called the Airborne Visible InfraRed Imaging Spectrometer (AVIRIS). This sensor is flown aboard a modified U-2 airplane at an altitude of about 20,000 m. Ground resolution varies with the altitude of the aircraft but is generally l5–20 m; the image swath width is about 11 km. AVIRIS measures surface reflectance in 224 bands in the visible and near infrared portions of the spectrum (from 400–2,500 nm). Each band is approximately 10 nm wide. The amount of data that the AVIRIS produces is prodigious; one flight line covering about a 10-×-11-km area on the ground produces a 140-megabyte image file. But in return, AVIRIS provides an extremely precise record of surface reflectance. The disadvantages include large data sets and platform instability necessitating extensive preprocessing corrections. Where multispectral systems can distinguish broad differences among the earth’s many features, such as broad vegetation classes like hardwood and softwood forest types or tree genera, hyperspectral sensors can identify different tree species as well as more subtle aspects of a plant or soil, such as plant stress or soil mineralogy. Radiometric resolution corresponds to a sensor’s ability to differentiate between amounts of radiation received (Limp 1993:186). Commonly, 8-bit resolution is used that corresponds to 256 values at a range of 0 to 255. However, digital sensors are not limited to this and there are examples of sensors that use 11-bit resolution, corresponding to 2,048 values, or other amounts. Generally, a higher radiometric resolution is advantageous. However, to successfully differentiate between amounts of radiation, more energy is necessary. The newer remote sensors typically have high radiometric resolutions. Temporal resolution is also an important characteristic of a sensor. This refers to the revisit time of a satellite over a particular geographic location. Some sensors in the

56 ~ Marco Giardino and Bryan S. Haley modern fleet of NASA’s Earth Science satellites revisit specific locations twice daily. Others, like Landsat, return to the same locale every 16 days. Sensors mounted on aircraft have variable temporal resolutions since they can be deployed as needed.

Sources of Remote Imagery There are several ways that remote imagery might be acquired by cultural resource managers, including finding existing imagery, hiring a specialist, or producing imagery in-house (Ebert 1984:304). Each of these has advantages and disadvantages for archaeological projects. Finding Existing Imagery Black-and-white aerial photography covering most areas of the United States may be acquired from archives and is increasingly available on-line for very little charge. One example is Digital Orthophoto Quarter Quads (DOQQs) produced by the U.S. Geological Survey’s National Aerial Photography Program (NAPP). DOQQs are 1-m images that are typically in black-and-white form, although color infrared is available for selected areas. DOQQs for the entire state of Mississippi are available free by download from the Mississippi Automated Resource Information System (MARIS) website. Other states have similar archives. Another example is aerial photographs produced by the Soil Conservation Service and the U.S. Geological Survey. These are often purchased inexpensively as hard copy photographs but can be converted to digital form with a high-resolution scanner. One advantage of these images is that they may be available for multiple years dating back as far as the 1930s. Older photographs may provide information that has been lost due to damage from agriculture or other cultural disturbances. Multispectral imagery is now also available at a low cost in on-line archives. The most accessible of these is Landsat imagery, which may be purchased on-line from the EROS Data Center operated by the U.S. Geological Survey. The work of the six Landsat satellites that have been in operation from the early seventies until today allows a nearly continuous temporal coverage of most areas. Similar medium- to high-resolution imagery from numerous other satellite sensors, including NASA’s ASTER and MODIS, EO-1, and the National Oceanographic and Atmospheric Administration’s (NOAA’s) AVHRR, is also available from the EROS Data Center site and from data archives searchable over the internet. Commercial satellite imagery has become more available and often achieves much higher spatial resolution than Landsat. Examples include imagery from the IKONOS satellite, which can be purchased on-line as Carterra digital products, and that from QuickBird. The drawback of the newer high-resolution satellite sensors is that the imagery is relatively expensive to obtain. One can sometimes reduce costs by using the high spatial imagery to study a subsample of statistically relevant sections of larger survey areas that were studied using lower resolution imagery.

Airborne Remote Sensing and Geospatial Analysis ~ 57 Hiring Remote Sensing Specialists Remote imagery may also be obtained by contracting an outside company to conduct a flyover. There are many private companies today that can be hired to acquire remote imagery for specific project areas. These are typically very high quality, but they may be quite expensive. One advantage is that by using an airborne platform, one can control the time and weather characteristics of the mission. Also, one can fly at altitudes that provide various spatial resolutions from sub-meter to dozens of meters and the spectral resolution of airborne sensors is now very advanced. Fairly inexpensive multispectral or even hyperspectral imagery can be collected from fixed-wing aircraft using three or more coregistered digital cameras with charged-coupled device (CCD) arrays and specified interference filters. It is important to understand the spectral response pattern of the features of interest prior to selecting the filters. Producing In-House Imagery In order for archaeologists to produce their own aerial imagery, a substantial commitment is usually required in terms of the purchase of equipment. Necessary equipment primarily consists of a sensor and some platform. The sensor may range from a standard 35-mm film camera to a low-cost multispectral camera. Three-band multispectral cameras designed for agricultural applications are now available for a few thousand dollars or less. Thermal infrared cameras have traditionally been more expensive but are quickly becoming more affordable. The platform may be a kite, balloon (Figure 4.3), unmanned aerial vehicle, or manned aircraft such as a powered parachute (Figure 4.4) or Cessna. Flyovers may also be arranged with local private pilots on aircraft but, over time, this is usually more expensive. Although imagery may be of somewhat lower quality, this method allows the archaeologists greater control over data collection.

Remote Sensing of Archaeological Targets Archaeological features may be apparent in remote imagery as variations in shadowing, soil color, moisture patterning, frost and snow marks, and crop marks (Scollar et al. 1990:37–51). Several works offer detailed explanations of how archaeological resources can be detected in this way (Allen 1984; Hampton 1974; Jones 1979; Riley 1979; Stanjek and Fabinder 1995). Shadow Marks Archaeological features may be visible in shadow marks, which are caused by slight elevation differences (Wilson 1982:78–80). Even small elevation differences can be visible using this technique if the conditions are appropriate. Besides elevation, shadowing is also dependent on the time, date, latitude, view angle, and ground surface color (Scollar et al. 1990:33). In general, photographs taken about an hour after sunrise or

58 ~ Marco Giardino and Bryan S. Haley

Figure 4.3. A helium blimp in use as a low-cost, low-altitude remote sensing platform.

before sunset are fine. However, the best conditions may be difficult to predict and, therefore, photographs should be acquired in several times and seasons. Features that might be enhanced with shadowing include eroded mounds, eroded earthworks, and wall fragments (Lyons and Avery 1977:61). Oblique photography may be particularly useful in helping to enhance shadow marks. Soil Marks The visibility of archaeological features as soil marks can be related to soil chemistry, organic material content, and soil texture (Scollar et al. 1990:37). These characteristics alter the reflectance, or ratio of reflected radiant energy to the irradiant solar energy, of the features (Short 1982:25). One cause for this is that cultural activities will sometimes leave behind an increased amount of chemicals, such as iron oxides. Iron oxides tend to redden the soil color. Organic material also has distinct chemical properties. In this case, the soil color is darkened. In addition, archaeological features that cause a soil texture variation may alter reflectance values. In general, reflectance increases with decreasing particle size (Allen 1984:190). Soil texture differences may be visible for several types of features. Cultural landscape modifications, such as mound construction or fill episodes, may leave a soil layer distinguishable from surrounding soils. Likewise, pits may leave perceivable differences as a result of mixing of topsoil. Buried sites have an effect on soil phenomenology that is observable without the need to penetrate the soil. Any feature that either drains water better than the surrounding area or retains water more than the surrounding area can provide visual evidence on the remote sensing imagery.

Airborne Remote Sensing and Geospatial Analysis ~ 59

Figure 4.4. A powered parachute in use as a stable remote sensing platform.

Soil texture differences can also be developed as damp marks (Allen 1984:68). The diameter of micropores in clays is about 2 μm, while the diameter in sandy soil ranges from 63 to 2,000 μm (Stanjek and Fabinder 1995:95). Therefore, fine-grained soils, such as clays, will drain less moisture than larger grained soils. Thus, if an archaeological feature is visible as a soil texture variation, differential moisture patterning may result. Rainfall levels preceding photograph acquisition are very important in the visibility of damp marks. In some cases, soil marks may be visible for only a few days. In general, they may show best when soils are drying out (Allen 1984:68; Drass 1989:83). The second day after a rain has also been suggested as the best time for soil mark development. Wilson (1982:50) has remarked that, as a general rule, soils should not be excessively wet or dry for best development. Plowing also plays a role in making subsurface features detectable as soil marks. A plowing episode brings up a sample of subsurface features, including archaeological materials, each time the plow passes over an area (Wilson 1982:41). Moreover, the lower materials are usually turned over so that they are most visible (Riley 1987:21). Soil marks may be particularly prominent after fallow fields have been plowed (Drass 1989:84). It should be noted, however, that materials might be transported from their original positions by the plow (Wilson 1982:42). Eventually, repeated plowing may render the ground surface homogeneous and cause marks to disappear. Frost/Snow Marks Frost and snow marks relating to archaeological features may be visible also as a result of soil texture differences (Riley 1987:21). This is primarily due to thermal

60 ~ Marco Giardino and Bryan S. Haley mechanisms. The timing is critical, however, since these marks are often visible for only a few hours after sunrise (Scollar et al. 1990:49). Frost and snow marks, of course, may rarely be applicable in the warmer portions of the United States. Crop Marks Crop marks may also reveal the location of archaeological features when the ground is covered by vegetation. Crop marks are caused by variations in vigor, which may be visible as differences in plant height, leaf area, or plant color (Jones 1979:657). Depending on the type of feature, crop vigor may be enhanced or reduced by buried archaeological features. Features that retain water, such as ditches, will often enhance plant growth. On the other hand, features that inhibit root penetration, such as buried walls, will produce vegetation above them that is less healthy than that in the surrounding area. One factor in the visibility of meaningful crop marks is the type of plant present. Plant species vary widely in their growth cycles, and buried archaeological features may only be apparent at certain stages (Riley 1979:30). For example, a positive mark may result because of increased transpiration of the vegetation, causing early development (Stanjek and Fabinder 1995:100). Later in the cycle, the crop marks may not be visible at all. However, the crops that exhibited enhanced growth will use up water faster and may ripen faster (Riley 1979:31). Thus, the crop marks would once again be visible. Crop marks have often been observed in cereal crops, including barley, wheat, oats, and rye (Allen 1984:75; Jones 1979:656–657; Riley 1987:31). These crops are very responsive to variation in soil moisture. Cereals may reveal archaeological features as variations in development, germination, plant height, and ripening (Riley 1987:33). However, observations of these crops are prevalent because these are the common crops in Great Britain. Grasses have also shown crop marks, but they are generally less responsive to soil differences than cereal crops (Riley 1987:30). Allen (1984:75) has noted that grasses are particularly sensitive to short-term changes in ground moisture and this may cause the disappearance of meaningful crop marks. Indeed, Riley (1979:29) has posited that grasses are not helpful for archaeologists. Other crops have also revealed archaeological features as crop marks, although less commonly. Root crops, such as turnips, potatoes, and beets, vary in their ability to show markings. Generally, those with deep roots are better at developing markings (Wilson 1982:61). Weeds have been observed to show crop marks in some cases (Wilson 1982:64). However, chemical treatments inhibit mark development in weeds (Wilson 1982:30). Regardless of the type of plant present, homogeneous vegetation cover is advantageous (Jones 1979:656). Long-term weather patterns are critical in the visibility of vegetation marks. Droughts often produce the most defined marks (Stanjek and Fabinder 1995:91). Jones’s (1979) experiments have indicated that a soil moisture deficit can trigger crop marks. The available water by volume is a function of soil particle size and thus soil texture is an important factor (Jones 1979:662). For example, Riley (1979:31) has noted that sandy soil frequently produces crop marks, whereas clayey soil does not.

Airborne Remote Sensing and Geospatial Analysis ~ 61 Also, moisture deficits are closely related to the root depth, which is determined by the plant species (Jones 1979:662). In some cases, rapid periods of rainfall have also been observed to cause crop marks (Drass 1989:83). The timing of plowing episodes is an important factor in crop marks, although less than it is with soil marks. In some cases, freshly plowed soils may enhance crop mark visibility (Stanjek and Fabinder 1995:92). However, plowing patterns can sometimes appear similar to archaeological patterns.

Image Processing Techniques Each digital image requires some preprocessing before the needed information can be extracted from the data. One such process involves the rectification of an image either to another image or to a map. The latter process produces images with planimetric characteristics that can be used as maps, similar to DOQQs. The second type of preprocessing that normally is required to properly extract information from remotely sensed data is radiometric correction, commonly referred to as atmospheric correction. Since not all the energy that reaches the sensors can be ascribed solely to the pixel of interest, a radiance measurement at the sensors needs to be converted to a reflectance measurement. The process for doing this is beyond the scope of this chapter, but several references are available that deal in depth with the issue of atmospheric correction. It is important to note, however, that particularly when doing temporal studies (i.e., comparing images from two different periods) or when working in project areas near large bodies of water, it is essential that the imagery be radiometrically corrected from radiance values to reflectance values to ensure proper comparisons and classification of the imagery. Once the images are preprocessed, image processing techniques that are essential for successful interpretation of remotely sensed data can be initiated. These processing techniques can be divided into two types, image enhancements and image classification. The purpose of image enhancement techniques is to more effectively display data for visual interpretation (Lillesand and Kiefer 1994:525). Image enhancements include radiometric enhancement, spatial enhancement, and multiband enhancement (ERDAS 1994:145–146). Radiometric enhancements increase the contrast of certain pixels at the expense of other pixels. This is achieved by altering the intensity value histogram of an image. Contrast stretching is one example. In this technique, the histogram is manipulated in a way to increase contrast between features of interest. This is useful because data rarely extend evenly over the entire intensity range. Thus, stretching the area of the histogram at areas of interest avoids crowding display values into a small range (Lillesand and Kiefer 1994:493). Area of interest subsetting in conjunction with contrast stretching is a valuable tool for archaeological analysis (Figure 4.5). Another frequently used type of spatial enhancement is convolution filtering, which involves the use of a matrix, or kernel, of varying dimensions that is used to manipulate

62 ~ Marco Giardino and Bryan S. Haley

Figure 4.5. A black-and-white aerial photograph before (left) and after (right) subsetting and contrast enhancing.

the digital numbers of the imagery. The kernel is composed of a series of weights that is moved over the image gradually. As it moves, the kernel is multiplied by corresponding values in the image, their products are summed, and the new value replaces the digital number of the center element (Lillesand and Kiefer 1994:555). Low pass filters emphasize low frequencies and deemphasize high frequencies. Therefore, these have a smoothing effect on imagery. High pass filters, on the other hand, emphasize high frequencies and deemphasize low frequencies and thus produce a sharpening effect on imagery. It is important to note that image enhancement techniques like histogram stretching do not alter the digital numbers or brightness values of each cell in the raster grid. Filtering techniques, however, do alter the original data values and therefore

Airborne Remote Sensing and Geospatial Analysis ~ 63 complicate the temporal analysis of imagery particularly when classes of features are being compared. Another group of image enhancement techniques work on multiple images, often various bands of a multispectral digital sensor. The most basic of these is simply multiband viewing. Because the human eye is unable to see beyond the visible spectrum, imaging software allows bands to be assigned to red, green, or blue display colors. Moreover, each of these colors can be viewed simultaneously, allowing multiband viewing. Mathematical operations may be performed on bands of data. For example, subtraction, which reduces common details of bands and enhances contrast, is quite common (Showalter 1993:84). In fact, multiple operations are often performed. One commonly used example is the Normalized Difference Vegetation Index (NDVI), calculated by the equation (near infrared – visible red)/(near infrared + visible red). NDVI is used for vegetation mapping and compensates for illumination conditions, slope, and aspect (Lillesand and Kiefer 1994:506). Change detection is a specialized form of band mathematics that is used to determine differences between two images. In its most basic form, change detection can be accomplished by subtracting the values of a later image from those of an earlier image. Thus, higher values in the resultant image represent a greater amount of change. A more advanced form of change detection results in a thematic map that depicts regions of change beyond a certain threshold. Other multiband image enhancements use statistical operations. One common and useful example is principal components analysis (PCA), which statistically removes redundancy that exists between bands (Cox 1992:260; Lillesand and Kiefer 1994:572; Showalter 1993:84). Here, the correlation between data bands is calculated and used to compress the data. The resultant data set has fewer bands but conveys the same information as the original. Thus, after PCA analysis, the bands are often simpler to interpret visually. Besides the use of PCA as an image enhancement, it is often commonly used as preprocessing to increase the efficiency of image classification and for removal of noise components from the imagery. Image enhancements are designed to aid the user in pattern recognition. Image classification techniques accomplish this by using an automated process. On the basis of user-defined parameters, the image is partitioned into spectral classes. There are two types of classification, unsupervised and supervised, but hybrid techniques can also be used. These types are based on varying degrees of control in selecting the classes into which the image will be partitioned. In unsupervised classification, the computer determines the classes after a number of parameters are chosen by the user. This process is performed by one of several clustering algorithms. One of the most popular is the ISODATA algorithm, which uses a minimum spectral distance to form clusters of data (ERDAS 1994:241). The ISODATA algorithm is iterative with an entire classification performed and new statistics calculated with each iteration.

64 ~ Marco Giardino and Bryan S. Haley In contrast, the significance of the classes is determined in the initial step of supervised classification. The user controls the classes that the image will be partitioned into by specifying training areas for each specific classification algorithm. Then, the machine classifies pixels into the specified classes that they most resemble.

GIS and Remote Sensing Analysis A GIS manages location and attribute data (Lillesand and Kiefer 1994:39), and such a system often includes vector data composed of point, line, and polygon features. These features are linked to a database that may include any of several types of attribute data. The matrix form of raster data can also be included in a GIS. For remotely sensed data, each cell in the matrix contains a reflectance value corresponding to some ground area. Of course, the primary value of a GIS in archaeological research is the ability to examine the relationship between multiple data layers. When registered in a common grid system, diverse data sets, including those from airborne remote sensing and nearsurface geophysics, may be compared and analyzed. Supplementary data, such as those from historic maps, plats, and other spatial documents, may be overlaid with the raster imagery. Other types of data may also be overlaid in vector format. The precision georeferencing, or assigning of map coordinates, to data layers is very significant. Often this is accomplished by referencing one type of digital data to another with a known grid system. This process requires both patience and a good eye for common features. Various rectification algorithms are used then to resample a data set to the new grid system. For imagery with little distortion, a simple first-order polynomial may be used, which only requires three ground control points. For more distorted imagery, higher order polynomials must used. However, in cases of complex, nonlinear distortions, a rubber sheeting model must be used. There are several ways that the analysis of remote imagery and other data layers in a GIS might benefit cultural resource projects. For example, advanced knowledge of terrain features and land cover can assist in the formulation of survey methodology. The total acreage of wetlands, forests, open field, and other ground cover types in the project area can be determined and a plan devised. When arduous field conditions make standard survey methods difficult, such as in the fairly inaccessible coastal wetlands, it can help determine the mode of transportation that will be required and where crews can be dropped off and picked up. Transects can be laid out in advance of a survey as a GIS layer and accurate field positions can be maintained with total station or GPS units. Parcels of land representative of various terrains in the project area can be measured rapidly from digital imagery and quantitative and statistically representative samples can be determined before the crews enter the field. The use of remote sensing as a component of the fieldwork in these areas is most likely to yield positive results. Another situation in which remote sensing is a reasonable option is when time in the field is constrained. Although standard survey techniques are inexpensive, they

Airborne Remote Sensing and Geospatial Analysis ~ 65 can be very time consuming. Again, remote sensing can help make the best of a short field season. Analytically, the landscape classification potential of digital remote sensing data provides information on land cover/land use changes, alternative locations of developments, and high-probability areas for stratified sampling strategies. Finally, remote sensing may be appropriate if there is a long-term research commitment to a particular region. The initial investment in a digital product can provide returns over many seasons of fieldwork.

Integrating Remote Sensing into CRM Projects Remote sensing can be useful in CRM projects in several different ways. Because these applications are diverse, good planning is necessary to integrate remote sensing into a research design. With foresight, remote sensing can be used to address a variety of problems in a standard three-phase CRM approach, increasing both their efficiency and their quality. Applications can be broadly grouped into three categories: predictive modeling, site detection, and site mapping. In archaeology, predictive modeling based on remotely sensed data attempts to connect site location with modern environmental patterning. Although the landscape has, in many cases, changed greatly, large-scale ecological features often remain in place. Analysis of medium-resolution multispectral data, such as those produced by the Landsat satellites, has been demonstrated to be a useful technique in rapidly mapping land cover. Digital data can be manipulated and themes or classes of phenomena on the earth’s surface extracted. Using a GIS, a statistical model can be constructed by comparing known site locations with the environmental zones that have been produced. Predictive modeling is a way to reduce the amount of land included in a survey area and can be useful in the planning stage of cultural resource surveys. A predictive model is never able to account for the location of all sites but can be beneficial in identifying high-probability survey areas. Remote sensing using airborne and orbiting instruments is a useful approach particularly for the detection of sites in Phase I (scoping and surveying) aspects of CRM work as required under the National Environmental Policy Act (NEPA) of 1969, Section 101 (b)4. Site locations may be apparent as lineaments or regularly shaped anomalies in the imagery caused by topographic variations, soil marks, or vegetation marks. Site boundaries may be established by determining the extent of these anomalies. The British have extensively used aerial reconnaissance for site detection, in part because of favorable ground cover conditions. However, as a result of recent developments in sensing technology, it should also be seen as a viable site detection tool in many areas of North America. Site mapping is often performed in the Phase II site assessment or the Phase III data-recovery stages of a CRM project. The mapping of features within a known site can sometimes be accomplished using remotely sensed data. Often, these are subsurface features that are otherwise invisible in ground observations but are visible as subtle

66 ~ Marco Giardino and Bryan S. Haley variations in electromagnetic energy at the surface of the earth. These are primarily visible as soil or vegetation marks caused by the underlying archaeological resources. Throughout a CRM project, data analysis may be aided by the use of remote sensing and GIS techniques. The digital products created during this approach serve as layers in GIS. Coregistration of modern imagery with historic maps, plats, and surveys provides useful information about the location of historic properties.

Ground Truth In most cases, deriving the correct information from the analysis of remotely sensed data requires some ground verification data. Spatial and spectral in situ data are required to georeference or register imagery and to identify the spectral signatures of specific features. Visual identification of vegetation and other features made on the ground is often the best and simplest method to “train” classification algorithms used in supervised classifications of imagery. Spatial ground truth data are normally collected using GPS equipment. For proper registration of high spatial resolution imagery, accurate GPS locations, to within the IFOV or pixel size of the imagery, should be collected using differential GPS. Since the federal government ceased scrambling GPS signals, the ability of most GPS units to provide locations accurate to within a meter or so has been highly enhanced. Spectral ground truth data are collected with spectral radiometers that can be handheld or suspended a few meters above the feature. Spectroradiometers collect energy from the relevant feature along either broad or narrow bands. Since these readings are being collected close to the object, the radiant flux, or energy contributed to airborne and satellite imagery by atmospheric scattering of light or from pixels adjacent to the pixel of interest, is minimized or eliminated. Collected spectral readings enable the remote sensing analyst to radiometrically correct satellite and airborne imagery. Often large placards of known spectral reflectance (large gray scales visible to the airborne sensor) are located along a flight path to allow comparison of the known reflectance with the radiance collected by the sensor over the placard. The difference between these two values can be subtracted from the entire image to produce radiometrically corrected data.

A Case Study: Hollywood Mounds Remote sensing experiments conducted at the Hollywood site provide some measure of the potential that airborne imagery has in mapping archaeological features (Haley 2002; Johnson et al. 2000). Hollywood is a late Mississippian site located in northwest Mississippi a short distance from the modern channel of the Mississippi River. The site contains at least five mounds that are still visible today despite the impact of a century of agricultural activities. A sketch map (Figure 4.6) produced by Calvin Brown in 1923 shows a series of perimeter mounds that are no longer visible

Airborne Remote Sensing and Geospatial Analysis ~ 67 today. In order to locate some of these lost features, the site has been imaged by several geophysical techniques and numerous types of airborne remote sensing. Black-and-white Soil Conservation Service photographs were acquired for the years 1938, 1942, 1966, and 1992 (Figure 4.7). These were scanned using a high-resolution scanner and georeferenced to the site grid system. One valuable aspect of this set of photographs is that they document some of the historic activities that have impacted the site. For example, historic structure are visible on the top of two of the mounds in the earliest two photographs. In addition, numerous high-reflectance patterns are visible in the 1938, 1942, and 1992 photographs in the northern half of the field. These patterns were probably caused by differential vegetation growth, drying variations, or, in the earliest two images, topographic change. Geophysical surveys conducted by Berle Clay in 1998 and excavation by the University of Mississippi in the years that followed revealed the buried remnants of Mississippian houses and plowed-down platform mounds in these areas. Similar patterns are also somewhat visible in large-format, color infrared photographs acquired with a Zeiss camera system by NASA in 1997 (Figure 4.8). Largeformat cameras produce photographs of exceptional sharpness and definition (Riley 1987:55). The Hollywood image was scanned with sufficient resolution to produce a digital image with a ground resolution of 0.39 m. Once the image was in digital format, the area of interest was selected and contrast enhancement performed. The resulting image contains much clearer versions of the anomalies. The same NASA mission also carried the ATLAS sensor, which acquired multispectral imagery at a ground resolution of 2.5 m. ATLAS produces 14 bands of data, including six in the reflected range, two in the mid-infrared range, and six in the thermal infrared range. An image acquired at noon shows the same high-reflectance patterns, particularly in the near infrared, as those in reflected energy bands (Figure 4.9). The thermal infrared bands of the ATLAS sensor contain some different anomalies (Figure 4.10). Several low-emittance ellipses just to the west of the tree-covered Mound A seem to correspond to some of the perimeter mounds. The fill that makes up these mounds contrasts with the surrounding soil, which alters their physical properties and affects their diurnal heating cycle. An artificially filled plaza area to the southwest Figure 4.6. A 1923 Calvin Brown sketch of Mound A is also visible in the ATLAS map of the Hollywood site (Brown 1926).

68 ~ Marco Giardino and Bryan S. Haley

Figure 4.7. Soil Conservation Service photographs of the Hollywood site (22TU500) from 1938 (top left), 1942 (top right), 1966 (bottom left), and 1992 (bottom right). The earliest of these shows the mounds (light marks in the upper half of the image) to be much less impacted by plowing than today. Farm buildings are also visible. The same anomalies are not as clear in the 1942 photograph, probably because vegetation coverage is less ideal. Thick vegetation covers the ďŹ eld in the 1966 image and little can be seen. Even after enhancement (see Figure 4.5), little useful information is apparent. The 1992 photograph, despite a substantial amount of plow impact, oers nearly as much useful information as the older images. Arrows indicate anomalies that coincide with known locations of houses and platform mounds.

Airborne Remote Sensing and Geospatial Analysis ~ 69

Figure 4.8. The near infrared band from the large-format color infrared photography of the Hollywood site. Archaeological anomalies are clearer than in the 1992 black-and-white photograph. Arrows indicate anomalies that coincide with known locations of houses and platform mounds.

Figure 4.9. The near infrared band 6 of imagery obtained with the ATLAS sensor, Hollywood site. Arrows indicate anomalies that coincide with known locations of houses and platform mounds.

thermal infrared image. Other anomalies are suggestive of past cultural activity but have not been tested. Targeted thermal reconnaissance was also performed by the University of Mississippi by suspending a handheld Agema 570 thermal camera from a helium blimp. Three houses and one of the mound patterns were imaged over a six-month period in 1999. The three houses are situated in diering soil matrices ranging from clays to sandy natural levee and thus the thermal behavior of these features varied. Overall, the clearest of the anomalies were produced by the houses situated in the ďŹ ner grained soils at the site (Figure 4.11). In these nighttime images, the houses produce cool anomalies,

70 ~ Marco Giardino and Bryan S. Haley suggesting they have higher thermal inertia values than the surrounding soils.

Conclusions Remote sensing using airborne and satellite imagery is particularly useful during Phase I CRM work. When properly preprocessed and processed, the imagery can serve as planning tools for conducting surveys, including drawing statistically significant samples for random, systematic, and stratified survey strategies. The samples can be drawn on the basis of both the spatial and the spectral attributes of digital data. As a consequence of registering or rectifying an image to a map, the image becomes an accurate map of the project area from which statistically significant samples can be extracted and located in the field. Similarly, after the image has been radiometrically corrected, thematic classification of biophysical features such as vegetation and soils can provide a sound basis for extracting stratified samples that emphasize areas of high site probability, particularly when Figure 4.10. The thermal infrared band 10 various ecosystems are being sampled. of imagery obtained with the ATLAS sensor, When modern imagery is registered Hollywood site. Arrows indicate ellipses that to historic maps and plats, the search for coincide with a line of truncated, plowed-down mounds. To the right of the line of mounds is a structures and features identified on the warm anomaly that is caused by an artificially original documents becomes more effective raised plaza. and less costly. Survey teams should be able to narrow the size of any particular area to be searched. Furthermore, the current land cover classes extracted from modern imagery can provide important information on the probability of finding historic sites. For example, the severity of any river channel migration can be assessed by comparing modern imagery and historic maps to rapidly assess whether a particular site has been irrevocably eroded since its establishment. Satellite and airborne imagery, including updated DOQQs, serve as useful strategic tools to assess the survey methodology of any particular CRM Phase I study. A rapid examination of the survey area using remote sensing data can assist the project manager in determining access to the survey site, the size of the required survey crews, the

Airborne Remote Sensing and Geospatial Analysis ~ 71

Figure 4.11. Thermal infrared imagery produced by the Agema 570 camera aboard a helium blimp, Hollywood site. The image is a composite of an image acquired at 10:51 P.M. on September 30, 1999 (right side) and one acquired at 5:36 P.M. on December 8, 1999 (left side). The circled cool anomalies correspond to known locations of two houses on the western edge of the site. Other anomalies were produced by surface microtopography, such as that caused by the tracks of a truck.

possible spacing of transect lines, the equipment required, and the amount of surface visible in any particular area. Under the best conditions, passive remote sensing done from aircraft and satellites serves to discover sites, delineate their extent, and accurately map their features. Spectral analysis of well-calibrated digital data using predetermined spectral bands has identified trenches, moats, wells, earthworks, pits, and organic soils. Hyperspectral data hold great promise for refining the use of crop marks for identifying subsurface deposits. Plants may show added vigor as a result of organic matter or, conversely, show stunting as a result of a hard substrate that hinders root growth. Modern digital imagery filtered by narrow-band spectral interference lenses advances this traditional method of site identification. Federal Distributed Active Archive Centers (DAACs) provide greater access to digital remote sensing data, often at nominal costs. In addition, agencies like the U.S. Geological Survey provide digital line graphs (DLGs) and digital elevation models (DEMs) available on CDs or through direct downloading over the internet. These data are excellent for registering images to maps and therefore deriving planimetrically accurate

72 ~ Marco Giardino and Bryan S. Haley products in a variety of scales and projections. Commercial firms that operate satellites and aircraft for collecting remote sensing data are much more common than they have been and are becoming more affordable, particularly for applications after the spectral response curves of specific features of interest have been identified in the laboratory, allowing the proper choice of spectral filters for CCD cameras that can be mounted on inexpensive platforms such as fixed-wing aircraft, blimps, or large kites. Recent technological advances have significantly improved the tools available for remote sensing data processing and analysis. In just the past 5 to 10 years, computers have been created with vastly improved RAM, storage, and speed, making the use of laptops and desktops for image processing a very viable alternative. Advancement in hardware has been matched by similar progress in remote sensing and GIS software. Windows-based systems, with simple graphical user interfaces (GUIs) and drop-down menus, provide even the beginning analyst with all the preprocessing and processing tools to derive planimetric and thematic information from all types of digital remote sensing data. In summary, remote sensing data from airborne and orbiting platforms can save significant resources during all aspects of CRM, particularly in Phase I surveys. Even more important, these data improve the accuracy and thoroughness of surveys, particularly those conducted in relatively inaccessible areas like coastal wetlands and marshes.

References Cited Allen, G. W. G. 1984 Discovery from the Air. Aerial Archaeology 10:37–92. Berlin, G. L. 1977 Archaeological Field Patterns Revealed in North-Central Arizona by Aerial Thermography, Soil Chemistry, Pollen/Plant Analysis, and Archaeology. American Antiquity 42:588–600. Berlin, G. L., J. R. Ambler, R. H. Hevly, and G. G. Schaber 1975 Archaeological Field Patterns Revealed in North-Central Arizona by Aerial Thermography. Proceedings of the American Society of Photogrammetry October 26:231–243. Falls Church, Virginia. Brown, C. S. 1926 Archaeology of Mississippi. Mississippi Geological Survey, University. Carskadden, J. 1999 The Gilbert Mound and Earthwork Complex, Muskingum County, Ohio. Ohio Archaeologist 49(4):4–9.

Airborne Remote Sensing and Geospatial Analysis ~ 73 Cox, C. 1992 Satellite Imagery, Aerial Photography, and Wetland Archaeology, an Interim Report on an Application of Remote Sensing to Wetland Archaeology: The Pilot Study in Cumbria, England. World Archaeology 24(2):249–267. Crook, A. R. 1922 The Origin of the Cahokia Mounds. Bulletin of the Illinois State Museum, May 1922. Springfield. Custer, J. F., T. Eveleigh, V. Klemas, and I. Wells 1986 Application of LANDSAT Data and Synoptic Remote Sensing to Predictive Modeling for Prehistoric Archaeological Sites: An Example from the Delaware Coastal Plain. American Antiquity 51(3):572–588. Davidson, T. E., and R. Hughes 1986 Aerial Photography and the Search for Chicone Indian Town. Archaeology 39:58– 59, 76. Drass, R. R. 1989 Application of Remote Sensing to Archaeology. Bulletin of the Oklahoma Anthropological Society 38:79–97. Ebert, J. I. 1984 Remote Sensing Applications in Archaeology. In Advances in Archaeological Method and Theory, vol. 7, edited by M. B. Schiffer, pp. 293–357. Academic Press, Orlando, Florida. Ebert, J., and T. Lyons 1983 Archaeology, Anthropology and Cultural Resources Management. Manual of Remote Sensing II. American Society of Photogrammetry, Falls Church, Virginia. Eleuterius, L. N., and E. G. Otvos 1979 Floristic and Geologic Aspects of Indian Middens in the Salt Marshes of Hancock County, Mississippi. SIDA Contributions to Botany 8:102–112. ERDAS, Inc. 1994 Erdas Imagine Field Guide. 3rd ed. Erdas, Inc., Atlanta. Featherstone, R. 1999 Aerial Reconnaissance over England in Summer 1996. Archaeological Prospection 6(2):47–62.

74 ~ Marco Giardino and Bryan S. Haley Fowler, M. L. 1977 Aerial Archaeology at the Cahokia Site. In Aerial Remote Sensing Techniques in Archaeology, edited by T. R. Lyons and R. K. Hitchcock, pp. 65–80. Reports of the Chaco Center, No. 2. National Park Service, Albuquerque. Giardino, M. J., and M. Thomas 2002 NASA Remote Sensing Research as Applied to Archaeology. The SAA Archaeological Record 2(3):15–19. Gibson, J. L. 1987 The Ground Truth about Poverty Point: The Second Season, 1985. University of Southwest Louisiana Center for Archaeological Studies Report No. 7. University of Southwestern Louisiana, Lafayette. Haley, B. S. 2002 Airborne Remote Sensing, Image Processing, and Multisensor Data Fusion at the Hollywood Site, a Large Late Mississippian Mound Center. Unpublished Master’s thesis, Department of Sociology and Anthropology, University of Mississippi, Oxford. Haley, B. S., J. K. Johnson, and R. Stallings 2002 The Utility of Low Cost Thermal Sensors in Archaeological Research. Center for Archaeological Research, University of Mississippi, Oxford. Report prepared for the Office of Naval Research, NASA grant NAG5-7671. Hampton, J. N. 1974 An Experiment in Multispectral Air Photography for Archaeological Research. The Photogrammetric Record 8:37–64. Jensen, J. R. 1986 Digital Image Processing: A Remote Sensing Perspective. Prentice Hall, Englewood Cliffs, New Jersey. Johnson, J. K. 1991 Settlement Patterns, GIS, Remote Sensing, and the Late Prehistory of the Black Prairie in East Central Mississippi. In Applications of Space-Age Technology in Anthropology, edited by C. A. Behrens and T. L. Sever, pp. 111–119. NASA, John C. Stennis Space Center, Mississippi. Johnson, J. K., R. Stallings, N. Ross-Stallings, R. B. Clay, and V. S. Jones 2000 Remote Sensing and Ground Truth at the Hollywood Mounds Site in Tunica County, Mississippi. Center for Archaeological Research, University of Mississippi, Oxford. Submitted to the Mississippi Department of Archives and History.

Airborne Remote Sensing and Geospatial Analysis ~ 75 Johnson, W. G. 1994 Early Aerial Photography: A Remote Sensing Technique Used to Detect Prehistoric Earthworks in the Kissimmee River Basin. Florida Archaeologist 47(3):269–279. Jones, R. J. A. 1979 Crop Marks Caused by Soil Moisture Stress at an Iron Age Site in Midland England, U.K. Archaeo-Physika 10:656–668. Lillesand, T. M., and R. W. Kiefer 1994 Remote Sensing and Image Interpretation. 3rd ed. John Wiley & Sons, New York. Limp, W. F. 1993 Multispectral Digital Imagery. In The Development of Southeastern Archaeology, edited by J. K. Johnson, pp. 184–206. University of Alabama Press, Tuscaloosa. Lindbergh, C. A. 1929 The Discovery of the Ruined Maya Cities. Science 70:12–13. Lyons, T. R., and T. E. Avery 1977 Remote Sensing: A Handbook for Archeologists and Cultural Resource Managers. Cultural Resources Management Division, National Park Service, Washington, D.C. Lyons, T. R., and R. K. Hitchcock 1977 Remote Sensing Interpretation of an Anasazi Land Route System. In Aerial Remote Sensing Techniques in Archaeology, edited by T. R. Lyons and R. K. Hitchcock, pp. 111–134. Reports of the Chaco Center, No. 2. National Park Service, Albuquerque. McKee, B. K., and T. L. Sever 1994 Remote Sensing in the Arenal Region. In Archaeology, Volcanism, and Remote Sensing in the Arenal Region, Costa Rica, edited by P. D. Sheets and B. R. McKee, pp. 136–141. University of Texas Press, Austin. McKee, B. K., T. L. Sever, and P. D. Sheets 1994 Prehistoric Footpaths in Costa Rica: Remote Sensing and Field Verification. In Archaeology, Volcanism, and Remote Sensing in the Arenal Region, Costa Rica, edited by P. D. Sheets and B. R. McKee, pp. 142–157. University of Texas Press, Austin. Penfound, W. T. 2001 Plant Distribution in Relation to the Geology of Louisiana. Proceedings of the Louisiana Academy of Science 8:25–34.

76 ~ Marco Giardino and Bryan S. Haley Perisset, M. C., and A. Tabbagh 1981 Interpretation of Thermal Prospection on Bare Soils. Archaeometry 23(2):169–187. Riccio, J. F., and C. Gazzler 1974 Infrared Color Photography of the Fort Mims Site, Alabama. Journal of Alabama Archaeology 20(2):217–221. Riley, D. N. 1979 Factors in the Development of Crop Marks. Aerial Archaeology 4:28–32. 1987 Air Photography and Archaeology. University of Pennsylvania Press, Philadelphia. Sabins, F. F. 1997 Remote Sensing, Principles and Interpretation. W. H. Freeman, New York. Scollar, I., A. Tabbagh, A. Hesse, and I. Herzog 1990 Archaeological Prospecting and Remote Sensing. Topics in Remote Sensing, No. 2, G. Hunt and M. Rycroft, series editors. Cambridge University Press, Cambridge. Sever, T., and J. Wiseman 1985 Remote Sensing and Archaeology: Potential for the Future. NASA, Stennis Space Center. Sheets, P., and T. L. Sever 1991 Prehistoric Footpaths in Costa Rica: Transportation and Communication in a Tropical Rainforest. In Ancient Road Networks and Settlement Hierarchies in the New World, edited by C. D. Trombold, pp. 653–658. Cambridge University Press, Cambridge. Short, N. 1982 The Landsat Tutorial Handbook: Basics of Satellite Remote Sensing. NASA Reference Publication 1078. NASA Scientific and Technical Information Branch, Washington, D.C. Showalter, P. S. 1993 A Thematic Mapper Analysis of the Prehistoric Hohokam Canal System, Phoenix, Arizona. Journal of Field Archaeology 20:77–90. Stanjek, H., and J. W. E. Fabinder 1995 Soil Aspects Affecting Archaeological Details in Aerial Photographs. Archaeological Prospection 2(1):91–101.

Airborne Remote Sensing and Geospatial Analysis ~ 77 Thiessen, T. D. 1993 Aerial Photography and Mapping at the KNRI. In The Phase I Archaeological Research Program for the Knife River Indian Villages, National Historic Site, edited by T. D. Thiessen, pp. 167–176. National Park Service, Lincoln, Nebraska. Wilson, D. R. 1982 Air Photo Interpretation for Archaeologists. St. Martin’s Press, New York.


Conductivity Survey: A Survival Manual R. Berle Clay

Earth conductivity survey, also known as electromagnetic (EM) survey, measures the ability of the soil to conduct an electric current. The value, measured in siemens, is the reciprocal of resistivity (to convert to resistivity in ohm meters, divide the conductivity, in millisiemens per meter [mS/m], into one thousand [Bevan 1983:51]). This said, there is considerable difference in the way earth conductivity and earth resistivity are measured. Although the theory behind EM survey is considerably more complex than the theory behind resistivity, there are fortunately a number of lucid published explanations aimed specifically at archaeologists (see Bevan 1983, 1998:29–43; Frohlich and Lancaster 1986), as well as more technical discussions (McNeill 1980), that should be helpful to both the user and the manager. The following discussion builds on these and focuses on my personal experiences from almost 20 years of use with one particular EM survey instrument (Figure 5.1). This brief introduction is designed to get the first user or discouraged archaeological user (I find there are many of these) into (or back into) the field collecting useful EM data on archaeological sites. In the United States, one of the problems with doing EM survey is that the technology is used most extensively by nonarchaeologists for a variety of geological, environmental, and agricultural applications. Because of the wide availability of EM technology in colleges and universities, in many instances an earth conductivity meter, generally wielded by a nonarchaeologist, may be the archaeologist’s

80 ~ R. Berle Clay hands-on introduction to near-surface geophysical survey. Because of different data-collecting goals, which are reflected in field techniques, there tends to be little understanding between archaeological and nonarchaeological users (with the exception of communication with the geophysicists themselves, who may or may not have archaeological interests). Again, and as another reflection of the many users, EM technology has not been built specifically for archaeology but has remained generalized and hence applicable to a wide range of geophysical interests. These problems should not stand in the way of the widespread use of EM survey in archaeology, although they do, as a product of initial Figure 5.1. A conductivity survey in progress. The field efforts that did not really seem to EM38 is carried in my right hand, ca. 15 cm off the ground. It is contained in a sheath of foam accomplish useful archaeological goals. Finally, I always think in terms of usinsulation held together with duct tape. The data logger is carried in the left hand and manipulated ing EM survey in concert with magnetic with the left thumb. The operator’s attention is fo- survey, exploiting the specific advantages cused on the marked rope below the EM38. of the different survey methodologies (Clay 2001). Therefore I am less concerned with the strengths or weaknesses of EM survey in contrast to another form of survey technology in archaeology than with how the specific qualities of EM survey data may be incorporated in a larger strategy for collecting multiple sets of complementary geophysical data on archaeological sites.

Summary Comments To begin, it is useful to outline what I see as the strong and weak points (from an archaeologist’s standpoint) of EM survey technology and the specific survey problems I wish to discuss here. Most of these points I specifically mention; all are implicit throughout my discussion. Strong Points of EM Survey 1. Fast because it involves no electrical contact with the ground 2. Can be used in a variety of ground conditions (grass/brush/tree cover, ridged ground, and so on) where other techniques may be more difficult to deploy

Conductivity Survey ~ 81 3. Works well in team with magnetic gradient survey 4. Can be used in dry periods as well as wet (an edge here on resistivity) 5. Can measure magnetic susceptibility (in parts per thousand), as well as earth conductivity (in mS/m), a mixed blessing 6. Can do a certain degree of vertical separation of geophysical phenomena Cautions These must be managed with field technique: 1. Temperature drift 2. Digital lag 3. Appropriate visual output and data processing 4. Metal on the operator Weak Points 1. Sensitive to a wide range of metals (from flip tops to water mains!) 2. Depends upon the presence of soil contrasts: if they don’t exist, don’t take the machine out of the box 3. Much greater sensitivity of near-surface targets (can be seen as positive or negative) 4. Difficulty in making depth discrimination between targets 5. Directions of technological development making the meter less and less satisfactory for archaeology (a multipronged problem) 6. Subject to electrical interference of certain electromagnetic frequencies (overhead power lines)

Historical Background EM techniques in archaeology seem to have been developed in Europe in the late 1960s and early 1970s (see Tite and Mullins 1970). Their use in North American archaeology largely grew out of the commercial development in the 1970s of earth induction meters by Geonics Limited, a Canadian firm specializing in geophysical equipment. Its work made available off-the-shelf instruments that could be used for a variety of applications including archaeology (also geology, soil science, and environmental monitoring). One of the earlier published archaeological applications for the United States was by Bruce Bevan (1983), a geophysicist who received his training in geophysics and electrical engineering and his contacts with archaeology at the University of Pennsylvania. In the article, he discussed the use of an EM31 earth induction meter at Fort de Chartres, an eighteenth-century French fort in Illinois, at the Deer Creek historic Wichita site in Oklahoma, and at La Ciudad, a Hohokam settlement near Phoenix against a more general background of the field methods he used to collect and analyze the data. In the

82 ~ R. Berle Clay three field trials Bevan produced plausible EM evidence for the existence of a buried fortification trench at Fort de Chartres, plowed-down mounds at Deer Creek, and a feature at La Ciudad that may have been a prehistoric sedimentation basin. Dr. Bevan has continued to be the principal geophysicist in the United States using EM techniques in archaeological prospection and has up to the present reported on a worldwide selection of sites emphasizing the use of multiple, complementary techniques. However, the techniques have been adopted by a number of archaeologists. A parallel interest in EM survey techniques has continued in continental Europe, particularly in France. Interestingly, although the first EM earth induction meter made specifically for archaeological applications seems to have been built in England (Howell 1966)—a device known vernacularly as “the banjo” and a product of a fertile tradition of British geoprospection—British field methods since then have generally not consistently involved EM techniques (David 1995:20). This may reflect the fact that the technology has commercially developed in Canada, and not in Great Britain, where the widely used field instruments, stressing magnetometry and resistivity, have been developed locally in close cooperation with the Ancient Monuments Laboratory of English Heritage.

Theoretical Basis: Mind Your Soils EM surveys use an instrument called an electromagnetic low-frequency induction meter that induces an electromagnetic signal into the ground and measures how well it is conducted by the soil. The frequency of the signal may vary (the instrument I use generates an audio frequency signal of 14.6 kHz), but critical to the design of the induction meter is the feature that there is no electrical connection between the survey instrument and the ground, unlike instruments measuring soil resistivity, which use a variable array of metal probes inserted in the ground and wired to the resistivity meter—probes that must all or in part be shifted to a new location to take a new reading. The induction meter uses a coil near ground surface to “broadcast” its low-frequency signal, which is received by another coil, also near ground surface. The transmitted signal causes the conductive material below the meter to generate its own faint signal, which is detected by the receiver coil. Both coils are built into the meter and the spacing between them governs the effective depth to which the meter can measure earth conductivity. The instrument is not designed to be a metal detector, but high-conductivity metals also generate a strong signal in response to the meter and their response tends to overload the circuitry. The meter is designed, rather, to measure the much smaller signals generated by the conductivity properties of soils. The electronics of the meter convert the signal into the measure of conductivity (mS/m, because of its small size). In general, soil conductivity meters measure differences in the conductivity of soils that are a product of their composition and formation. A typical “spread” of soil types produces the ranges of mS/m shown in Table 5.1 (Bevan 1998:8) (resistivity measurements are also listed).

Conductivity Survey ~ 83 When soils have been moved around in an archaeological site by its occupants, horizontal conductivity contrasts can be created that a conductivity meter might record. In most cases this will be true if there already exists vertical variation in soil composition in the local soil column. In addition, the composition and texture of the soil may be changed by cultural activity (and natural forces as well). An EM earth induction meter may record all of these events. However, it is quite important before doing a conductivity survey to have some idea of the local soil column, most directly, but on a theoretical level, by consulting the local soil map and the description of the soil type. Often systematic shovel tests (which may have been used to initially locate an archaeological site) may have also collected valuable information on the local soil column and perhaps also its variation across a site. Fortunately, earth conductivity meters have also been designed to readily estimate vertical changes in soil conductivity. Most may be held in two orientations relative to the ground surface. In what is known as the “vertical dipole” position, that is, upright or at right angles to the ground surface, they measure mS/m to a greater depth than they do when held parallel to the ground surface in what is known as the “horizontal dipole” position. For example, the popular EM38 earth conductivity meter, which is 1 m long, measures mS/m in the vertical dipole position to approximately 1.5 m whereas in the horizontal dipole position it measures only to a depth of ca. 50 cm. Any field use of an earth conductivity meter should begin with an informal attempt to estimate the vertical properties of the soil at a number of points in the survey field. If conductivity contrasts do not exist or if it cannot be demonstrated by this simple technique that the near-surface soil is more (or less) conductive than the subsoil, the chances are the conductivity survey may have rather limited useful archaeological results. As a rule, in “upland” soils (my stamping ground in Kentucky), conductivity will vary with depth below surface. For example, in an upland South silt/clay loam (which I am well acquainted with from central Kentucky), conductivity tends to increase with depth, from a loam at ground surface to sticky clay at ca. –1 m (see Table 5.1 for the implications of this). On the other hand, in a southern Ohio glacial outwash context, variable loams may overlie glacial gravels, in which case conductivity will decrease with depth. In some contexts, for example, on an alluvial floodplain, there may be relatively little variability in conductivity with depth. Still, where there has been active channel development (cutting and filling) along high- Table 5.1. Resistivity and conductivity of different soil types velocity rivers (e.g., the Mississippi), there may Soil Resistivity (ohm-m) Conductivity (mS/m) be real contrasts (sand to Sand, gravel 1,000–10,000 0.1–1 heavy clays), reflecting Silty sand 200–1,000 1–5 channel cutting, natural Loam 80–200 5–25 levee development, and Silt 40–80 12.5–25 10–40 25–100 periodic overflows. The Clay 5–10 100–200 more one knows of this Saline soil

84 ~ R. Berle Clay sort of information, the easier it may be to interpret collected EM data. Most archaeologists with a regional focus already have a general idea of the types of soil they may encounter, and the EM implications of this should always be considered before, not after, an EM survey is attempted. At the same time, the most precise (and useful) information on vertical soil structure is usually contained in a county soil survey. Bear in mind, however, that the soil column reported in the county soil survey has generally been modified by a century or more of commercial agriculture. For example, at the top of a hill, sheet erosion has often removed a meter (or more) of the theoretical column, depositing it at the bottom of the hill, effectively burying the theoretical column. Soil scientists are well aware of these potential changes and treat the published soil survey as a reference point against which to evaluate their field data: archaeologists would do well to emulate this methodology if they do not already. When pits and ditches are dug and refilled and mounds created and worn away, soils become redistributed on an archaeological site. It is the contrast between redistributed soils and those still in place that is usefully read by a conductivity meter. Not surprisingly, an earth conductivity meter tends to do very well in recording earthworks of all sorts, including mounds, ditches, banks, and the like, if there is significant contrast between fill and nonfill (Figure 5.2). At the same time, earth conductivity can record low-conductivity features, like buried stone and masonry foundations and fired clay features such as central hearths in late prehistoric Native American structures. Figure 5.2 nicely illustrates how an understanding of local soils, their formation processes, and contemporary agriculture is really essential to an interpretation of an EM survey beyond superficialities (i.e., if it looks like an archaeological feature, the survey must have “worked”). The Little Spanish Fort earthwork is located on a flat alluvial plain in the active floodplain of the lower Yazoo River in the Mississippi delta country. The soils of this plain have fairly high clay content and are therefore conductive; they were probably laid down as a “back swamp” some distance from the active river channel. Subsequently, the Yazoo River has cut back toward the site and, in fact, this segment of the earthwork is quite close to the current river bank. The river now overflows directly onto this portion of the earthwork and the coarser, less conductive soil carried in solution by the flood is deposited along the bank when the water velocity rapidly decreases as the floodwater disperses to the floodplain. In effect, the site is being alluviated by flood action, which, here at least, is depositing a less conductive layer of silt over a more conductive clayey subsoil. The conductivity meter has registered the tops of the exterior berm and interior bank created by excavation of the ditch and the small mound apparently built as a blind at the entrance through the bank, because they were built of the more conductive, clayey subsoil. Finally, modern farming has used a subsoil “ripper” to enhance the local growing properties of the bottomland. Because EM technology uses “open” signal collectors, it can also pick up other environmental noises that can seriously degrade the measurement. Depending upon their

Conductivity Survey ~ 85 transmission frequency, overhead power lines can cause fluctuation in recorded mS/m. Importantly, earth spherics, notably lightning discharges, can cause unwanted noise that often persists for a time after the discharges. Finally, EM meters register the presence of various forms of metal (not simply ferrous material) both incorporated in the soil (pipes and other utilities) and above ground (e.g., wire fences and metal structures), which tends to overload the recording instrument (depending upon the size of, and characteristics of, the metal). The instruments are similar to metal detectors; Figure 5.2. A 60-×-60-m view of a “classic” ditch and bank with the EM38 (higher conductivity = darker; however, in the latter the coils are lower conductivity = lighter). From the lower right, “coaxial” (as opposed to laterally there is a slight exterior berm, paralleled by a sandy siltseparated), providing a much more filled ditch on left (light, low mS/m); then a high mS/m precise location of small metal ob- (dark) bank with a break through it; then the interior jects. Because of these factors, it is of the earthwork and a small mound (round, dark, high mS/m) opposite the entrance through the bank (diagomy experience that the technology nal subsoiling marks and possible external feature paraldoes not work particularly well for leling the berm) (Little Spanish Fort, Mississippi, data me as an archaeologist in urban or collected for Dr. Edwin Jackson, University of Southern densely built-up modern environ- Mississippi). ments, although it may be used with good effect in archaeological historic environments, particularly those that became archaeological in the nineteenth century or at least before ca. 1940 (after which the amount of metal junk and modernizing utilities seems to create a quantum jump in the environmental noises that can affect EM surveys).

Technological Choices While new suppliers of EM technology have recently entered the market, it is still dominated by Geonics Limited, a Canadian firm that also has had a long-standing interest in archaeological applications (Geonics n.d.; McNeill 1980). Geonics produces a variety of instruments, all operating on the same principle but varying in their intended application. The principal variation among machines is the spacing between transmitting and receiving coils, hence depth of sensitivity, reflecting the fact that a wide variety

86 ~ R. Berle Clay of users use their technology. Two of their instruments, the EM31 and the EM38, have been used in archaeology, principally to measure earth conductivity using the Q phase or out-of-phase component of the recorded signal. With an intercoil spacing of 3.66 m, the EM31 effectively measures earth conductivity to ca. 6 m, whereas the EM38, with an intercoil spacing of 1 m, measures effective conductivity to ca. 1.5 m. The “cost” in ease of use for the depth sensitivity of the EM31 is high. Physically it is a boom 4 m long attached to an electronics unit. Both weigh 12.4 kg (a data logger adds an additional 1.5 kg). It is a tedious instrument to carry through a long day and often difficult to thread through natural obstacles (e.g., trees) without changing its orientation (which can affect the measurements it takes). By way of contrast, the EM38 is 1 m long and weighs only 2.5 kg (with 1.5 kg for a data logger). As a result it is principally the EM38 that has been used for more recent archaeological applications. New instruments are being developed (although not by Geonics as yet) that fall between the EM31 and the EM38 in coil spacing, hence depth sensitivity. These promise to offer additional depth sensitivity over the shallow-depth EM38 and the depth sensitivity of the EM31 without the size problems involved with the latter. They are something to watch. Depth Discrimination Both Geonics instruments may be rotated 90 degrees (effectively laying each instrument over on its side) to take measurements in the “horizontal” mode at approximately one-half the depth sensitivity of the “vertical” mode and thus to explore vertical change in soils (an internal mercury switch signals the data logger of the change in orientation). However, it is not very handy to do this and it considerably slows down area survey so that both machines are effectively used in the “vertical” mode. Nevertheless, this ability of earth induction meters does permit the surveyor to get a “feel” for vertical changes in soils, for example, at the beginning of the survey as a check against published soil descriptions, without the necessity of collecting extensive data sets in the horizontal orientation, and this can be important. In EM meters the measure of conductivity is “generalized” over the depth of sensitivity. This means that any archaeological stratigraphy is also generalized and neither instrument effectively discriminates complex stratigraphic differences. This can be a problem given the depth sensitivity of the EM31 but is less so with the EM38 that, with its 1.5-m depth sensitivity, is admirably suited to a wide range of minimally stratified, near-surface archaeological sites. Recently Geophysical Survey Systems, Inc., has offered an earth induction meter called the Gem 300 that purports to discriminate depth in conductivity by varying the frequency of the transmitter (Won et al. 1996) under the general theory that lower frequencies penetrate to a greater depth than do higher ones. It is the opinion of Geonics Limited (McNeill 1996) that the Gem 300 does not perform as advertised and there have been, to my knowledge, no attempts to test depth discrimination with archaeological applications using it.

Conductivity Survey ~ 87 The following discussion is built around the EM38 that I have used extensively in archaeological survey. It is important to recognize that the technology is most sensitive to objects near ground surface (Bevan 1993:52–53, fig. 7); in fact, the EM38 conductivity signal primarily reflects mS/m within the top 50 cm of the soil below the instrument. This is dramatically illustrated in Figure 5.3, a conductivity survey of 2,400 m of an archaeological site in central Kentucky on an eroded hilltop in an area of eroded silt/clay loam. The EM38 has done a fairly good job of picking out the plow scars in the clay subsoil, which have created fairly subtle conductivity differences (see also Figure 5.2). It has done this because this hilltop has been heavily eroded by sheet erosion with the result that the clay is closer to the surface than otherwise, increasing the contrast between the plow zone and the subsoil. With the published soil column in one hand, and EM survey results in the other, it is possible to estimate the percentage of the column that has been lost to sheet erosion, and this is important in the evaluation of archaeological site integrity. The particular sensitivity of the EM38 also means that it reacts strongly to metal objects that are near or at ground surface, such that historic “trash” tends to overwhelm the machine, making the machine difficult to use in recently quitted, metal-laden historic contexts. These “nearer surface” targets can often be somewhat muted by carrying the EM38 15–30 cm above ground surface (a good technique to use where there is unimportant historic trash).

Figure 5.3. A 40-×-60-m view of an archaeological site showing plow scars (higher conductivity = darker; lower conductivity = lighter). Although a surface collection was made from this area and items were recovered from shovel tests, no other features were identified in further testing, probably because they had been destroyed by sheet erosion.

88 ~ R. Berle Clay Temperature Drift As Bevan (1998:42–43) has pointed out in detail, the earth conductivity meter is affected by changes in temperature and this can be severe inasmuch as the temperature of the instrument changes during the day (recorded conductivity in mS/m rises as the meter warms up). Some attempt should be made to control these changes during the course of measurement, or they should be corrected afterwards with appropriate software because the resultant gradients in mS/m can obscure archaeological features. While the EM38 may be re-zeroed during the course of survey in response to this drift, this does not really solve the problem. The best approach is to turn the meter on at the beginning of the day (well before the beginning of the survey; 9-volt batteries don’t cost very much), allowing it to adjust to air temperature (for winter surveys I leave my EM38 in the garage overnight). It is most affected by the contrast between sunlight and shade and if it is left lying in full sunlight the changes can be dramatic. As a rule, if a day is overcast, the temperature remains fairly constant, and the EM38 has been warmed up beforehand, drift is minimized. The effects of the sun particularly, but temperature in general, can be minimized, as Bevan has indicated, by carrying the EM38 in an envelope constructed of sheet foam one-half inch thick (Figure 5.1), and it is probably good advice to always insulate it and, during data dumping and field breaks, to keep the instrument shielded from the sun. Needless to say, none of these suggestions come from a Geonics manual but rather have been developed by Bevan (1998) building on extensive work with earth conductivity meters. Recently manufactured EM38 meters incorporate circuitry designed to minimize temperature drift, but I have not had an opportunity to test its effectiveness. Moisture Unlike instruments that measure resistivity, conductivity meters seem to function well over a wide range of soil moisture levels, and this is one of their strong points. The meters may not function adequately over ice, although they may be used over frozen ground (e.g., permafrost) with good results (Bevan, personal communication 2003). Even in the summer when the ground is quite dry, I have been able to get measurable variation, though slight, with the EM38 in conditions that a resistivity meter might find overwhelming (i.e., such high resistance that it obscures any variation). In addition, because the EM38 requires no electrical connection to the ground, resistances created by poor probe/ground contact (exacerbated by low-moisture, high-resistance soil conditions) do not occur. This said, increased moisture causes conductivity to rise, and a high water table, for example, should obscure the measurement of significant variation in mS/m. At the same time, if the general level of soil moisture changes dramatically during a survey (after a rainfall, for example) the values of mS/m will also change, and, if the rain has been excessive, the ability to discriminate low-level variation in mS/m will decrease. In other words, a rain in the middle of a survey causes a problem (besides that of any lightning that might occur with it).

Conductivity Survey ~ 89 Data Logger In operation the EM38 is most efficiently connected to an external data logger (it is tedious to record data manually and later enter it into a computer). This is carried in one hand while the EM38 is carried in the other hand, either suspended from a strap or on the end of the Geonics patented nonmagnetic handle, or is dragged behind on a skid made from PVC pipe. I carry my EM38 because I like to see the point I am measuring parallel with my legs (Figure 5.1). My logger is the venerable Polycorder 722 (Omnidata International), which has seen duty on all fronts, from grocery stores to geological expeditions. I bought it secondhand from the Geonics rental pool 10 years ago. It is simple and rugged, and far more sophisticated recorders are currently available. There is room for considerable experimentation in selecting or designing a recorder for your earth conductivity meter: think of feeding the signal directly to a small hand-held computer. The operator should be careful to wear nonmetallic shoes and no metal from approximately the waist down. In addition, the data logger and the cable connecting it with the meter, because they are metal, should be carried as much as possible in a constant orientation relative to the EM38 sensors. Readings may be taken with the instrument resting on the ground, carried above it, or pulled behind on a nonmagnetic sled. The handle supplied by Geonics results in a carrying height for me of ca. 15 cm, which, while it does reduce the depth sensitivity, tends to mitigate somewhat the powerful effect of near conductivity (see above). I usually set the logger to take readings automatically every 0.5 second, moving forward over a marked rope at approximately 1 m per second, giving a reading every 50 cm. In these cases I carry the meter about 10–15 cm above the ground. Digital/Analog Output The EM38 produces a continuous variable voltage output measuring mS/m. Earlier versions have an analog output (they have an analog meter with a needle read against a scale). It is difficult to interpolate a value and manually and rapidly record mS/m from such a meter although a common digital multimeter may be wired to the output port to produce a readable signal (which is not mS/m, although it can be converted to mS/m). Later versions produce digital output (they have a digital LCD readout). This is very easy to read if one is recording mS/m manually. It is important to remember that, while the analog output responds directly and rapidly to the measurement of conductivity below the meter, the digital output performs a running average of measurements of mS/m over approximately the past 0.5 second. It is common practice to survey in a zigzag pattern (out on one transect and back on the next, reducing the walking time by 50 percent), flipping the EM38 endto-end at the end of each line (by simply reversing the direction the operator is facing; the EM38 is generally bidirectional in this sense). If the ground is traversed at approximately 1 m per second using an EM38 with digital output, there is a 50-cm lag in

90 ~ R. Berle Clay the direction traveled between the point being measured on the ground and the value being recorded in the data logger with specific x/y coordinates. This does not occur in the EM38 with analog output. If the operator returns on the adjacent transect going in the opposite direction, the lag of 50 cm occurs again, but in the opposite direction, producing a 1-m offset between measurements for side-by-side transects. If there are significant linear features, an offset of this magnitude will effectively obscure them and the lag must be corrected either as the data are collected or by postprocessing. (It was only when I was able to fully correct this problem in zigzag data sets that I realized that in some soil types the EM38 does an excellent job of recording the low-amplitude conductivity variation indicating cultivation scars like the subsoil scars at the Little Spanish Fort earthwork [Figure 5.2].) The most direct way to avoid the problem altogether is to survey only on unidirectional or parallel traverses when using a digital EM38 and traveling at 1 m per second (also true for the EM31; however, it is difficult to carry that machine at a rate of 1 m per second because of its size). Remember, however, that there will be a 50-cm offset between features on the ground and as recorded (Figure 5.4). Alternatively, the speed of ground coverage can be reduced; however, this increases the cost of survey time. I have found one software package (Geoplot 3.0) that can be used to easily correct the offset from using zigzag traverses, but it also imposes strictures on file size (see below). My own data collecting with the EM38 tends to evolve through time as I work with the little meter (it is a good idea to remember that all archaeologists using nearsurface geophysical survey instruments tend to modify their field technique through time in an effort to improve it and that use of the instruments should be viewed as a continuous learning exercise and instructions such as these are not a cookbook to be followed). In a search for denser data sets I now try to take four readings per meter which, given my good ol’ Polycorder 722 and its 1980s vintage software, means that I travel at 1 m per two seconds. This probably eliminates most of the data lag problem. However, at the same time I am getting away from zigzag survey traverses altogether because I have experienced slight variations in voltage output when the EM38 is flipped end-to-end, which can produce objectionable “grooving” in the data. My current solution is simply to walk the additional distance and stick to parallel traverses. The problem of digital data lag is clearly illustrated in Figures 5.4 and 5.5. These illustrations show a portion of a survey of the Short site in the Mississippi delta that I did for the University of Mississippi. A site of Archaic period vintage, Short is confined to a sandy knoll (low conductivity) in the otherwise flat floodplain of a tributary draining into the delta from the east. The knoll is considerably earlier than the more conductive alluvium that has been deposited around it. This survey recorded principally the soil contrasts and probably very little of the archaeology. The area was surveyed in zigzag fashion along transects 1 m apart with readings every 0.5 m. The ground speed of the survey was 1 m per second. All these factors combine to

Conductivity Survey ~ 91

Figure 5.4. Earth conductivity data uncorrected for “digital lag.” The data are from a portion of a survey of the Short site in the Mississippi delta done for the University of Mississippi. The lighter areas are lower mS/m; the darker, higher (note scale in mS/m).

Figure 5.5. The conductivity data of Figure 5.4 corrected for digital lag.

produce the frenzied product of Figure 5.4. The blunt point of this demonstration, as evidenced by Figure 5.5, is this is not the way the data are supposed to look and the effect can be corrected. Figures 5.6 and 5.7 show the effects of digital lag correction as applied to a grayscale raster image (an image I tend to prefer for conductivity data). Here there is some archaeology involved—clay features below a plowed-down mound at the Pinson site in

92 ~ R. Berle Clay

Figure 5.6. Conductivity data (higher conductivity = darker; lower conductivity = lighter) collected on zigzag traverses but uncorrected for digital lag (Pinson Mounds, Tennessee).

Figure 5.7. The same data as in Figure 5.6 with digital lag corrected by processing in Geoplot 3.0.

west Tennessee. This is another demonstration of how correcting for lag considerably clarifies the graphic result. Despite this, I have seen friends persist in presenting uncorrected data as the final product. You can imagine all sorts of ways to correct for digital lag, including inserting dummy measurements into endless data lists and starting “ahead” of yourself to make things come out right. It can be done very simply with a geophysical software package made explicitly for archaeology, Geoplot 3.0 (Geoscan Research) using a routine therein called “destagger” designed to correct for pacing inaccuracies when using a zigzag traverse pattern, principally in the collection of magnetic data. Furthermore, in this program you can gauge the immediate effects of the destaggering process and

Conductivity Survey ~ 93 make sure that you do not destagger too far. Getting your earth conductivity data into Geoplot 3.0 is another problem, discussed below. Despite the complications it produces, the signal averaging of digital meters reduces the noise in the output (here, random variation in signal caused either by the electronics or by nearby conductivity effects) and therefore it is a welcome feature. However, if you have a chance to purchase or use an analog EM38 do not avoid it simply because it is not the latest model (with an LCD display) and by no means, if you have an analog EM38, allow Geonics to modify the meter to digital output simply to keep up with things! Be prepared, however, to invest in a data logger for ease of data recording. With it you can zigzag to your heart’s content, ignoring the whole problem of digital lag and producing excellent data sets. However, if there is metal around, for example, in the case of historic archaeological sites, the metal will cause somewhat more variation in output as the meter valiantly tries to keep up with the complex conductivity signals it encounters. Magnetic Susceptibility Like other earth conductivity meters, the EM38 can measure magnetic susceptibility in the I mode (this term refers to the in-phase part of the self-generated signal that the EM38 is expressing). Magnetic susceptibility, measured in parts per thousand (ppt), in this sense is not an absolute but a relative measure, unlike the value produced by other magnetic susceptibility meters, but, in the case of the EM38 as with dedicated susceptibility meters, it seems to record the effects of midden formation and burning. Although I have used this I mode, I tend not to for three reasons. First, the EM38 measures ppt only to an effective depth of about 50 cm (Geonics 1992:15). When the unit is carried slightly above the ground this means that it is essentially plow zone that is being measured. Many sites I have surveyed are only momentarily out of cultivation with the result that the susceptibility data may be somewhat obscured. In one case I was privileged to survey an archaeological site that had never been farmed (Millstone Bluff, Illinois, courtesy of Southern Illinois University). Using the in-phase measurement I was able to delineate late prehistoric structures with some success, I believe, because the near surface of the soil had not been extensively disturbed (Figure 5.8). Millstone Bluff was a small, late prehistoric community consisting of houses clustered around a small plaza. Because of survey conditions (the site was, and still is, an obstacle course of downed trees, vines, and forest trash where it is difficult to deploy any geophysical survey equipment), measurements of ppt were taken at 0.5-m intervals by manually triggering the EM38 as it rested on the ground. The survey effort is a nice example of the reality that if you take your time you will increase the quality of the geophysical data you collect, a point I often forget. The plaza at the site stands out as a square marked by low susceptibility because of the nature of its prehistoric use being less intense than that outside the plaza or because

94 ~ R. Berle Clay

Figure 5.8. Magnetic susceptibility survey, Millstone Bluff, Illinois (lower susceptibility = darker; higher susceptibility = lighter). In the lower left hand (southwest) a house is clearly outlined by a higher susceptibility floor (burned) surrounded by walls. In general, the areas of higher susceptibility seem to reflect the rebuilding and burning of houses around the open plaza located on top of the bluff (courtesy of Southern Illinois University).

of the possibility that this central area was systematically cleaned. In the southwestern corner of the surveyed area there is the distinct outline of a rectangular house whose walls are marked by low susceptibility and floor by higher susceptibility. There is a suggestion of a similar structure at the middle of the top edge of the survey. Structures at both of these spots were also suggested by the topography. Around the plaza the susceptibility is higher and the patterning is confused because here the houses have been rebuilt multiple times (Brian Butler, personal communication); in fact, the point of highest ppt at the northwest corner of the plaza suggests a specialized structure of considerable occupational complexity. As for the second reason I tend not to use the I mode, it becomes apparent from balancing the EM38 in the in-phase that the height it is carried above the ground is critical and that variation in height is translated into variation in recorded ppt. When the meter is carried as I would normally carry the EM38 for a conductivity survey (ca. 15 cm above ground surface), this can produce some strange output if you do not watch your pace (Figure 5.9). Translated into action, this means that the EM38 should

Conductivity Survey ~ 95 probably be placed on the ground for each measurement of ppt or, if not, then carefully carried at a constant elevation (which was no problem at Millstone Bluff, where I actually set the EM38 on the ground). This slows down survey speed. Finally, because I generally combine conductivity survey with magnetic gradient survey (using a Geoscan Research FM36), I tend not to use the EM38 to survey magnetic susceptibility because the two survey techniques appear to produce somewhat redundant results. The magnetometer would seem to do the better job of recording susceptibility. In short, do not assume that you can do the work of a magnetometer with an earth conductivity meter. Data Collecting/Processing At this point begins a discussion of what may be one of the more important questions involved in the routine use of earth conductivity meters. I use a Polycorder 722 data logger generally supplied with the EM38 (and other models) by Geonics until quite recently. This is loaded with a simple data-recording program (DL720/38 Data Logging System version 1.00) developed by Geonics (Geonics 1992). This program requires that the operator set up a file typically for each square being surveyed (e.g., a 20-m square) and, at the beginning of each transect, specify the number of the transect, the starting point, the direction of travel, and the automatic recording interval. This generates a data set (Table 5.2) with x and y coordinates for each value of mS/m (the z-value). At the beginning of the transect the operator starts the recorder, proceeds forward, timing the pace to marked intervals on a rope, and then turns off the recorder at the end of the transect. A new transect (either zigzag or parallel) must be initiated with the logger (which can either increment or decrement the location, Table 5.2. Typical data set produced from data logger for processing

Figure 5.9. A 20-×-20-m square centered over a country brick kiln (high ppt = dark), ca. 1836, showing effects of walking pace on measurement of ppt with EM38 carried at about 15 cm above the ground surface (Auvergne Farm, Kentucky).



z (mS/m)






















96 ~ R. Berle Clay meaning you can go either way) before it is restarted. Alternatively, the logger may be set so that it is triggered by pushing a button on the handle rather than automatically, ultimately wearing out your thumb in the process. This produces an x, y, z ASCII file that can be downloaded using a companion program in a laptop called DAT 38 (version 3.22a) (Geonics 1994) and converted into a file that can be read by a graphics program such as Surfer 8 (Golden Software). DAT 38 runs only in a pure DOS environment (not under Microsoft Windows), so I must keep Windows 98 loaded in my laptop (in addition to later versions) because only Windows 98, and not later versions, gives one the opportunity to run a program in a “real” DOS environment (thanks to Microsoft). My good friends remind me that I can get a Windows version of the data-transfer program so the problem is not as dismal as I used to think. Data Processing Traditionally, EM data have been presented as contoured maps (Figure 5.10). These are easily produced with a graphic program like Surfer 8. However, unless contoured with a very close interval (at which point they become graphically quite cluttered), contour maps may not be the best way to display very small variations in conductivity that may indicate archaeological features. I have tended to go for raster images gener-

Figure 5.10. Conductivity data in contour map form (40 × 60 m, Boone’s Station, Kentucky).

Conductivity Survey ~ 97 ally in gray scale (Figure 5.11) (most mapping programs can add color to both contour and raster graphics to enhance the output, sometimes making them easier to interpret). In the example, a survey of a portion of Boone’s Station for the University of Kentucky, I believe the gray-scale image is more “revealing,” but judge for yourselves. In addition, because I also routinely use a Geoscan Research FM36 fluxgate gradiometer, I generally process EM38 graphics in Geoscan Research’s Geoplot 3.0, a full-service graphics package specifically designed for use with Geoscan instruments (magnetometers and resistivity meters). This program gives me access to a number of sophisticated filters that are not available in Surfer and that work on raster images, not contoured maps. In addition, though this may be debated, I believe that the gridding algorithm of Geoplot 3.0 (Figure 5.12) produces a better gray-scale image (more contrast) than the “image” option of Surfer 8 (Figure 5.13). Equipment Selection Geonics has continually upgraded its line of earth conductivity meters. The EM31 and EM38 have become of interest importantly in soil science (Davis et al. 1997; Doerge et al. n.d.) and have been updated to meet this demand. Two important changes have been made to the EM38, producing the EM38B and the EM38DD. The first measures mS/s and ppt concurrently. The second measures concurrently in both the vertical and horizontal modes, effectively recording two depth sensitivities simultaneously. These modifications seem to reflect the soil scientists’ interests in essentially the topsoil (near, near surface). The company has been very conservative in its developments and accomplished these changes merely by doubling the electronic components. Both instruments are considerably less handy to carry, being both heavier and off balance, and are not recommended to archaeologists for this reason. As mentioned above, the measurement of ppt is of questionable importance—certainly if use of the EM38 is combined with magnetometer survey. Again, the advantages of having two depth sensitivities with the EM38 is also questionable and hardly worth the cost of doubling the cost of the meter and doubling its weight. Fortunately, the straight EM38 is still being made and, for the interested archaeologist, all may be rented from the company and a number of independent suppliers for trial prior to purchase or normal use (strongly recommended). These changes reflect the fact that electromagnetic earth conductivity meters (and other forms of near-surface geophysical instrumentation) are increasingly being seen as passive data-collecting sensors. Mounted on a sled or cart, singly or in arrays georeferenced with a GPS antenna, and pulled across a site with a tug (e.g., an ATV), they may be used to cover large areas rapidly and at low cost. However, it should be understood that there is a considerable difference in measurement density between accepted soil science practice and archaeology. A recent soil science technical brief (Doerge et al. n.d.:3) suggests that 50 measurements of mS/m per acre is an acceptable density of readings for soil analysis. A normal density of mS/m readings in ar-

98 ~ R. Berle Clay

Figure 5.11. The conductivity data of Figure 5.10 in gray-scale form; note square house foundation with somewhat less distinct foundations to either side (40 × 60 m, Boone’s Station, Kentucky). Small low/high/low anomalies represent metal targets registered as the EM38 passed over them. Note the pipeline that strays across the lower right-hand corner, which is not as obvious in the contoured presentation.

Figure 5.12. Gray-scale image of conductivity data produced in Geoplot 3.0 (brick kiln, Auvergne Farm, Kentucky; see Figure 5.9 for ppt data of same feature).

Figure 5.13. Gray-scale image of conductivity data produced in Surfer 8 (same data as in Figure 5.12).

Conductivity Survey ~ 99 chaeology is 8,000 per acre and, using some types of magnetometers, 16,000 or more readings per acre. American practice in archaeological geophysical applications, furthermore, seems to be moving toward denser and denser data sets as hardware and software (memory and download and processing speed) improve. This is because of the nature of North American archaeological sites, which generally involve a wide and variable range of small archaeological features, and the belief, which seems reasonable, that finer calibrated readings will image these. Two things are important, denser data sets and finer calibration—that is, closer control over what is being measured on the ground. This said, there is also a marked and increasing divergence in software theory and development between archaeologists and soil scientists (and others who use geophysical survey instrumentation). Most software packages used with the EM38 were not made specifically with archaeological applications in mind but rather to serve, more generally, nonarchaeological users. The problem comes not in the dataprocessing software but rather in the data-collecting software. For example, the latest version of the popular Surfer program (Surfer 8) remains a highly flexible and useful package for the processing of geophysical data collected either by soil scientists or by archaeologists. To begin with, for hand-carried applications, nonarchaeological use has moved away from the sort of simple data-collecting procedure I have described above that requires careful attention to the operator’s pace. Many data-collecting programs, including that currently distributed by Geonics with the EM38, get around pace accuracy by substituting a different approach in which a “fiducial marker” is inserted by the operator in the data stream when a marked interval is crossed (say every 10 m). When the data are downloaded and prepared for processing, the program will then “average out” the relative locations of the measurements on the y-axis between recorded fiducial markers (or between transect beginnings and ends). However, this means that in the same data set (from a given square you are surveying) adjacent transects may have different numbers of readings, reflecting variable traverse pace, yet the data sets remain acceptable to the processing program. Still, the problem of variable pace remains and this can reduce the accuracy of the data despite the averaging, which can make them less suitable for recognizing relatively subtle features. In addition, for towed data collecting, the EM38 is now being linked to GPS technology to provide real-time x and y coordinates. While these systems are touted as sub-meter in accuracy, they still introduce problems in the accuracy of the data that can affect interpretation. A processing program I use represents a divergent theory. I use a British program, Geoplot 3.0, to process conductivity data sets. This is one of the few software packages produced explicitly for archaeogeophysics. Geoplot 3.0 is a well-designed program that has a series of processing features that allow it to handle resistivity and conductivity data (which, understandably, are processed in much the same manner)

100 ~ R. Berle Clay as well as magnetometer data. There is a method to this, because I generally use both the FM36 and the FM256 fluxgate gradiometers manufactured by Geoscan Research in the same survey and I can therefore use the same program to process my geophysical data. This package, however, is very precise about the nature of the files it processes. A file is set up so that each traverse has the same number of data values. The Geoscan data collectors also are instrumented so that they collect a precise and constant number of values per transect and turn themselves off when that total is reached. This is because they place a high premium on precise pacing (the more precise the better). Furthermore, this particular approach is moving toward denser data sets (more readings per meter) in an attempt to improve resolution of small archaeological features. Precise pacing remains an essential aspect of this search for finer resolution. If, and only if, earth conductivity files meet these data criteria, they may be easily imported into the Geoplot environment for processing (some processing packages permit the user to resample a data set to get around variable numbers of values per transect, but these may introduce additional complications). It is a small wonder, therefore, that there is little understanding between these two divergent trends in data collecting; in short, the data needs of archaeologists (at least as defined by our British workers) as opposed to the rest of the geophysical survey world seem to be quite different. Geoplot 3.0 is not, however, nearly as well developed as Surfer 8 in producing graphic output for published images (e.g., putting in all forms of labels, marginal scales, colors, and so on), and most workers export processed data from Geoplot to Surfer to produce these graphics. I go one step further because I like the output of the Geoplot gridding algorithm better than the output of the Surfer gridding algorithm. I export the gray-scale image from Geoplot as a bit-map file from Geoplot into Didger 3 (Golden Software, in one sense a mini–geographic information system [GIS] program), preserving its contrast, then move it into Surfer 8 to add the nice graphic details. All of this suggests that the processing of EM data for archaeology is more complex than the processing of other types of geophysical data. Really this is not the case; rather, the user needs to be fully aware of what is being done. Certainly there is room for continued experimentation in the collecting and processing of earth conductivity data. I have taken one path and find the results rewarding. There are surely others, but perhaps the lack of a clear “cookbook” approach in this case deters some from trying.

Data-Collecting Strategy I use both an earth conductivity meter and a form of magnetometer (fluxgate gradiometer) in my work and both are well adapted to the fast-paced and variable, unexpected conditions of cultural resource management (CRM) as well as the more leisurely and “planned” fieldwork of non-CRM research. Furthermore, I view them as an integrated pair of survey tools, not alternative choices (Clay 2001). As a rule, I do not expect much from the pair in built-over and busy urban contexts (with lots

Conductivity Survey ~ 101 of metal in the form of utilities, construction debris, and just plain junk). Perhaps the most useful instrument here would be resistivity. Also, after having surveyed many of them, I tend to stay away from historic cemeteries. While I am able to locate graves at times, I have been in far too many situations where I am unable to sufficiently define the limits of, complexity of, and population of historic cemeteries with any accuracy. Because my firm has subsequently excavated these cemeteries I am made painfully aware of my failures (Bybee 2004). Because of this, I tend to waste clients’ time and money with the geophysical survey of historic cemeteries. If a cemetery must be moved, the only positive way to define the graves is to strip the topsoil and outline the grave shafts. Discussing this with an archaeologist friend during the course of a survey, we both concluded that a historic act of burial tends to be a rather brief geophysical event: dig the grave, place the coffin, and fill it in (often in the space of an hour or two). This may not even produce major soil contrasts and perhaps the main contrast produced by the event is the difference in compactness between the grave filling and the surrounding undisturbed ground (remember also, any metal in the coffin may be 6 feet down, pretty deep for a fluxgate gradiometer to detect). It is, of course, this geophysical aspect that ground-penetrating radar can detect best, but even here, the technique is not foolproof. As a rule, if there is any suggestion of large-scale earthworks at a site (mounds, banks, ditches, earthwork alignments, and so on), present, plowed-down, or potential, my first choice is an earth conductivity survey. This will generally indicate the nature of the earth moving that has been involved in the site, for example, revealing plowed-down mounds where they may not be visible today. Unless these structures also contain features that have a magnetic signature, they will not be detected by the magnetometer. I always, in these cases, follow the conductivity survey with magnetometer coverage. On many sites, or perhaps most, my first choice as a survey instrument is the fluxgate gradiometer simply because the instruments I use (Geoscan Research FM256 fluxgate gradiometers) have the proven ability to reveal a wide range of possible archaeological features, are fast, and are well “integrated” with a software package specifically designed for them and for archaeological applications (Geoplot 3.0). This makes them easy to use although it takes some skill, training, and experience to effectively use a fluxgate magnetometer in gradiometer configuration. I then follow with the EM38, using the conductivity meter to further “inform” areas of the magnetometer survey that appear archaeologically interesting. This generally means that I do not completely resurvey the same area with the EM38, only parts of it. The two instruments effectively complement each other (Clay 2001). In one sense the two complement each other quite nicely. Burned clay features (like prehistoric hearths and burned house floors) can look to the magnetometer quite similar to anomalies created by ferrous (iron) targets. On the other hand, the EM38 conductivity meter will not respond to the burned clay feature per se. If it is sufficiently burned, like a brick mass, to reduce earth conductivity, that will be reflected in a

102 ~ R. Berle Clay lower value of mS/m. Thus a conductivity resurvey of a site in which several possible burned clay anomalies have been identified with a magnetometer will quickly establish whether the anomalies are in fact metal targets. I download data from the data logger in the field and generally check during the day on data quality, perhaps with a quick Surfer map (strongly recommended, especially when you are covering a lot of ground). It is possible to make mistakes in setup (metal on one’s person, bad connections between the data logger and the EM38, or bad connections between the logger and the computer) that could potentially blight an entire day’s efforts if not detected. In addition, a new field situation is always somewhat of a leap of faith. That is, there may be environmental conditions (e.g., an overhead power line) that create unacceptable noise. Again, it is nice to know early on whether the conductivity results you are getting seem to have any archaeological implications at all. As I have mentioned in passing, the processing of conductivity data can get somewhat involved, and I have not gone into it in detail here. Needless to say, processing may be 50 percent of the effort involved. I happen to use a combination of Geoplot 3.0, Didger 3, and Surfer 8. There are many other possibilities; in fact, the processing and presenting of geophysical data on archaeological sites is a creative endeavor and I rely increasingly on computer-aided design (CAD) and GIS software packages. In my CRM firm, geophysical data collecting is generally tightly tied to a research design that integrates my data collecting with other forms of more traditional data collecting (metal detection, shovel tests, coring, strip plowing or scraping, test units, and so on). The geophysical survey generally precedes other forms of data collecting, most often at a Phase II, “site evaluation” point in site treatment under Section 106 of the National Historic Preservation Act. Importantly, the results of the geophysical data collecting do not dictate the traditional forms of data collecting that may be used, rather they inform the techniques that have been scheduled, generally in response to statewide guidelines for conducting site evaluation. Many managers rightly fear that geophysical data that have not been adequately identified will drive the evaluation of archaeological resources without adequate checks. Rather, the geophysical survey techniques should be used as yet another archaeological field research “stage” in larger multistage research designs. For example, Phase II evaluations may call for shovel testing (ST) at a systematic interval. If the ST has been informed by a prior earth conductivity survey, then the systematic ST design may be supplemented if necessary with additional ST to adequately sample anomalies identified by the sensing device. In a stroke, the systematic ST becomes also “smart” ST and certainly the mapped conductivity anomalies will aid in the interpretation of any and all shovel tests. While it may seem that geophysical survey merely adds to the cost of doing fieldwork, we have found that it reduces the cost mainly by helping to provide more adequate evaluations of archaeological contexts: for every field context where the geophysical data may indicate much more than might have been detected by conventional means (increasing the overall cost of fieldwork), many more sites are revealed as less than might have been supposed (reducing the overall cost).

Conductivity Survey ~ 103 One of the nice side effects of working in the CRM context is that, given such a field strategy, I rapidly find out what the conductivity data mean in archaeological terms. Communication back from the archaeologist who digs to the geophysical archaeologist who did a survey at some earlier point is a continuing problem. All too often a geophysical survey of an archaeological site is performed by an itinerant specialist who never finds out what he or she was surveying or, worse yet, is not on hand to adequately explain the implications of the findings to the archaeologists doing the excavation. This said, it also cannot be stressed too strongly that, to effectively use geophysical survey techniques in archaeology (EM survey and others), they must be used. This is a cryptic way of saying that it takes a lot of field experience to gain confidence in one’s results so that they can become a really effective adjunct of the archaeologist’s tool kit. They are not techniques to be taken out of the box once a year merely to demonstrate for students or to apply to one’s pet project. All too often this sort of approach to their use has generated less than informative results, as well as the discouragement that I mentioned at the beginning of this chapter.

Examples Hollywood Site, Tunica County, Mississippi (Figure 5.14) This survey, covering quite a large area, shows a series of plowed-down house platforms and house floors. The plowed-down house platforms are the roughly circular anomalies. In one of them, at roughly 250 m S, the square high-conductivity shape of the fill of the platform is quite visible. The houses are circular, low-conductivity anomalies north of this feature. (Graphics in Surfer 8, data courtesy of Jay Johnson, University of Mississippi.)

Figure 5.14. Conductivity survey at the Hollywood site.

104 ~ R. Berle Clay Carty Site, near Columbus, Ohio (Figure 5.15) This example shows the complementary nature of earth conductivity and magnetometry. In this case, both survey techniques indicate the portion of the earthwork, and the magnetometry suggests that there is a burned structure below or in it, not indicated in the conductivity data set. (Data courtesy of the Ohio Historical Society.) Hopeton Earthworks, Ohio (Figure 5.16) This is a conductivity survey of a small portion of a very large earthwork. Elements that are visible here are an earthwork bank in the lower left-hand corner (a conductivity low) and in the west 60 m the faint outline of a circular earthwork that has been completely plowed down. There are cultivation scars over most of the area of this survey. This graphic was processed in Surfer 8.

Figure 5.15. Conductivity survey at the Carty site.

Conductivity Survey ~ 105 Hopeton Earthworks, Ohio (Figure 5.17)

In this example, further processing has been used to enhance the unprocessed conductivity data (top image). Here, a high pass ďŹ lter is used (central image) followed by interpolation of values (bottom image). The circular earthwork on the left becomes somewhat more visible and a somewhat rectangular conductivity low (light rectangle) in the middle of it suggests a feature (in fact, a shovel test at this point produced ďŹ recracked rock). (Data courtesy of the National Park Service and Jennifer Pederson.)

Figure 5.16. Conductivity survey at the Hopeton Earthworks.

Figure 5.17. Conductivity survey at the Hopeton Earthworks.

106 ~ R. Berle Clay

References Cited Bevan, B. 1983 Electromagnetics for Mapping Buried Earth Features. Journal of Field Archaeology 10:47–54. 1998 Geophysical Exploration for Archaeology: An Introduction to Geophysical Exploration. Special Report No. 1. U.S. Department of the Interior, National Park Service, Midwest Archeological Center, Lincoln, Nebraska. Bybee, A. D. 2004 Bioanthropological Investigations of the Burning Spring Branch and Reynolds Cemeteries, Kanawha County, West Virginia. West Virginia Archaeologist 52(1 and 2):1–51. Clay, R. B. 2001 Complementary Geophysical Survey Techniques: Why Two Ways Are Always Better than One. Southeastern Archaeology 20:31–43. David, A. 1995 Geophysical Survey in Archaeological Field Evaluation. Research and Professional Services Guideline No. 1. English Heritage Society, London. Davis, J. G., N. R. Kitchen, K. A. Sudduth, and S. T. Drummond 1997 Using Electromagnetic Induction to Characterize Soils (Missouri). Better Crops with Plant Food 81(4):6–8. Potash and Phosphate Institute, Norcross, Georgia. Doerge, T., N. R. Kitchen, and E. D. Lund n.d. Soil Electrical Conductivity Mapping. Site Specific Management Guidelines, SSMG-30. Potash and Phosphate Institute, Norcross, Georgia. Frohlich, B., and W. J. Lancaster 1986 Electromagnetic Surveying in Current Middle Eastern Archaeology: Application and Evaluation. Geophysics 51(7):1414–1425. Geonics Limited 1992 DL720/38 Data Logging System Operating Instructions for EM38 Ground Conductivity Meter with Polycorder Series 720 (Version 1.00). Geonics Ltd., Mississauga, Ontario, Canada. 1994 Computer Program Manual (Survey Data Reduction Manual) DAT 38, Version 3.22a. Geonics Ltd., Mississauga, Ontario, Canada. n.d. Selected Papers on the Application of Geophysical Instruments for Archaeology. Geonics Ltd., Mississauga, Ontario, Canada.

Conductivity Survey ~ 107 Howell, M. 1966 A Coil Conductivity Meter. Archaeometry 9(20):20–23. McNeill, J. D. 1980 Electrical Conductivity of Soils and Rocks. Technical Note TN-5. Geonics Ltd., Mississauga, Ontario, Canada. 1996 Why Doesn’t Geonics Limited Build a Multi-Frequency EM31 or EM38? Technical Note TN-30. Geonics Ltd., Mississauga, Ontario, Canada. Tite, M. S., and C. Mullins 1970 Electromagnetic Prospecting on Archaeological Sites Using a Soil Conductivity Meter. Archaeometry 12(1):97–104. Won, I. J., D. A. Keiswetter, G. R. A. Fields, and L. C. Sutton 1996 Gem-2: A New Multifrequency Electromagnetic Sensor. Journal of Environmental & Engineering Geophysics 1(2, August):129–138.


Resistivity Survey Lewis Somers

“The soil is an historic document which, like a written record, must be deciphered, translated and interpreted before it can be used” (Barker 1995:12). Resistivity survey offers one means of “reading” the archaeological record. The “ink” on the page is the resistivity contrast between the archaeological record and the surrounding soil matrix. Reading is performed by scanning the site with a resistivity survey system and viewing the results on a computer screen by means of analysis and display software. Resistivity survey is an active survey method. In this respect it is similar to groundpenetrating radar because both probe the subsurface by recording the response to the active injection of electromagnetic energy. These methods contrast with magnetic survey, which is a passive survey method. In magnetic survey one simply measures the soil- and feature-associated magnetic fields as they appear on the site surface. Resistivity surveys are implemented by scanning the site with a probe array that is connected to a resistance meter. The probes are used to inject current into the site and measure the local resistance. Since resistance data are collected at discrete intervals, the data sample density, an important survey design parameter, must be appropriate for the anticipated archaeological feature size and contrast. The resistance meter consists of a calibrated current source operating at a low (adjustable) frequency, a synchronized high-impedance voltmeter, and digital data logger (to record data value, grid number, line number, and position). The remainder of the survey system includes appropriate data-processing and display software. Figure 6.1 is a schematic representation.

110 ~ Lewis Somers

Figure 6.1. A resistivity survey system consisting of a probe array, multiplexer, resistivity meter, and data-processing and display unit.

Resistivity survey has application at almost all archaeological sites. In addition, it has application where many other survey methods cannot be used. Examples include conductivity (electromagnetic) and magnetic survey, which can be corrupted by scattered iron or other metal on a site (metal buildings, fences, wire, pin flags, and rebar), and ground-penetrating radar, which can be limited by site surface obstructions preventing continuous antenna contact with the soil (wooded sites, dead fall), as well as by low soil resistivity (clay, high salinity), which attenuates the radar wave and prevents data collection. The discussion of geophysical survey techniques in a recent textbook on field archaeology in Great Britain begins with resistivity, observing that this is the first of the instrument-based techniques to be used on an archaeological site and that it “remains perhaps the most important technique available to archaeologists, closely followed by magnetometry” (Drewett 1999:51). Bevan (1998:7) recommends resistivity as an excellent first choice because it is relatively simple to use, the instruments are inexpensive, and “resistivity surveys give good results at more sites than will any other type of survey.” Weymouth (1986) recites several early applications of resistivity in the United States. The implication is, and I agree, that resistivity survey is generally applicable at a broad spectrum of historic and prehistoric sites. This said, historically resistivity survey (and magnetic survey) has been less used in North America than on other continents. The reasons revolve primarily about the North American archaeological record, the early data-sampling strategies, the absence

Resistivity Survey ~ 111 of automated spatial registration of collected data, and the absence of appropriate dataprocessing software. Simply stated, the North American archaeological record is dominated by small, low-contrast features of irregular geometry, a combination that requires spatially registered high-density data to be collected with large dynamic-range instruments and impeccable field techniques. In addition, these large data arrays must be processed and interpreted with image (geospatial) processing software capable of both quantitative and statistical analysis plus be displayed with modern graphic formats. It is only in recent years that the need for adequate sampling strategies, automated resistivity meters with large dynamic range, and suitable processing and display software has been fully appreciated. Thanks in part to modern digital technology and in part to geophysical survey workshops sponsored by the National Park Service, these requirements are now routinely satisfied in cost-effective commercial and research surveys. The remainder of this chapter presents an introduction to resistivity survey, the basic concepts involved, the need for appropriate survey design and data processing, and finally a few examples of recent North American surveys.

Principles of Resistivity What is resistivity? Simply stated, it is the resistance of the soil to the passage of an electrical current. Soils with low resistivity typically have a high concentration of highly mobile ions (i.e., free water molecules and hydrated ions). It is the drifting motion of these ionic electric-charge carriers that is the flowing electrical current. That is to say, the soil grains per se contribute very little to the electrical current; rather, it is the motion of the soluble ions through the soil-particle matrix that constitutes the current flow. Soils with low resistivity have an abundance of highly mobile ions that are capable of carrying an electrical current, thus a low resistance to the flow of current. Soils with high resistivity typically lack an abundance of highly mobile ions and are less able to carry an electrical current (Carr 1982; Clark 1996; Klein and Santamarina 2003). But there is more to the story. At a more detailed level, drawing upon semiconductor resistivity models and descriptions (Grove 1967), resistivity is governed by both the number and the mobility of free charge carriers. The number of mobile ions is primarily governed by the number of water-soluble compounds available in the soil, as well as the presence/absence of soil moisture. The mobility of the ions is governed by a complex combination of soil moisture, soil grain size, temperature, and soil compaction, as well as the surface electrochemistry of the soil grains. All these processes govern soil resistivity for the low-frequency electrical currents used in these surveys. Features, activity areas, and settlement and occupation patterns are typical archaeological objects of interest in resistivity survey. They fall into two distinct categories. Many contain brick, stone, cement, or highly compacted soils, features that have their own “intrinsic” resistivity and can be considered an intrusive feature, that is, one that intrudes into the surrounding soil matrix. Intrusive features tend to be relatively high-

112 ~ Lewis Somers contrast features. There are also earthen or “disturbed-soil” features such as filled pits, house pits, middens, post molds, field boundaries, and even footpaths and animal tracks. These have a developed resistivity—developed in the sense of being a soil-matrix resistivity feature developed in situ by the myriad of physical and chemical processes associated with feature construction, use, and abandonment, as well as postoccupation processes. Earthen or disturbed-soil features are usually low-contrast features of undetermined sign. They are low contrast because they tend to be alterations of the soil matrix as opposed to a “manufactured” feature. Their resistivity can be higher or lower than the surrounding soil-matrix resistivity and the corresponding contrast positive or negative. Carr discusses these processes and issues extensively in Chapters 3 and 4 of his book (Carr 1982). The typical prehistoric North American site is dominated by these lowcontrast earthen/disturbed-soil features. At an archaeological site, we are interested in detecting and mapping the contrast between the soil-matrix resistivity and the archaeological-feature resistivity. When the resistivity contrast is high, the features of interest are easy to detect and recognize. When it is low, a characteristic of most North American prehistoric sites, features may be difficult to detect and recognize. Fortunately for the latter case there is a dataprocessing procedure that can both enhance feature detection and separate the highresistivity features (positive contrast) from the low-resistivity features (negative contrast). This procedure is discussed in the data-processing section later in this chapter. Archaeologists with a good conceptual understanding of these issues and detailed knowledge of their site stratigraphy, soils, and features are in an excellent position to develop useful archaeological interpretations from resistivity as well as other geophysical survey data.

Instrumentation One measure of the important role that resistivity has played in geophysical prospection in Great Britain is the fact that the first geophysical instrument commercially produced specifically for archaeologists, the Martin-Clark resistivity meter, appeared over 40 years ago (Clark 1996:19). This instrument relied on five individual probes that were inserted in the soil at fixed intervals in transects across the site. Four probes were used to measure soil resistance while the fifth probe was used to set up the next reading, leapfrog fashion. Although a vast improvement over early instrumentation, this approach was still time consuming. Improvements in electronics and experiments with different probe configurations led to the development of a new line of resistivity meters for archaeology by Geoscan Research. In its standard configuration, Geoscan’s RM15 uses a pair of stationary remote probes connected to two or more mobile probes mounted on a frame, along with a control box including a data logger. While there are other commercially produced resistivity systems, they have been designed with other applications in mind, primarily geological exploration, and they make use of multiple,

Resistivity Survey ~ 113 individually set probes in order to look deep into the surface of the earth. Since the RM15 dominates the field of archaeology, this presentation is based on that instrument. Resistivity surveys introduce a known electrical current into the soil and measure the ease (resistivity) with which this current flows within the soil by means of a voltage measurement. In practice, the current is introduced into the soil with a metal probe and the voltage is measured by connecting a voltmeter to a second nearby probe. While there are several possible probe arrangements (see Clark 1996 and Gaffney and Gater 2003 for extended discussions), the twin-electrode array is the primary configuration used today. It consists of a mobile current-voltage probe array (Figure 6.2) that is scanned over the site to collect the resistivity data. Associated stationary current and voltage probes are located a long distance from the survey area. These provide a return path for the injected current and a reference voltage for the voltmeter, respectively. These stationary probes are connected to the scanning probe array by wire. For the twin-electrode array configuration, the distance, a, between the current probe and the voltage probe is an approximate measure of the depth of the resistivity survey. The reason for this relationship can be seen in Figure 6.2, where the voltage probe is in contact with the constant voltage contour of radius a. An important feature of probe resistivity survey is the ability to control the depth of survey by increasing or decreasing the probe separation distance, a. In addition, by introducing a number of voltage electrodes, multidepth surveys can be performed (Figure 6.3). By virtue of the probe geometry and the physics of electrical current flow within the soil, the recorded data can be thought of as an “average” made up of contributions from both the soil matrix and the archaeological record. Consider three different examples: (1) as the survey scan passes over a stone wall whose resistivity is much higher than the soil resistivity, the recorded data will show an increase in measured resistance; (2) as the survey scan passes over a stone wall whose resistivity is much lower than the soil resistivity, the recorded data will show a decrease in measured resistance; and (3) as the survey scan passes over a stone wall whose resistivity is approximately equal to that of the soil matrix (i.e., zero contrast), the recorded data will show no change and the feature will not be mapped. The measured resistance, calculated within the resistance meter, is R = V/I and the associated soil resistivity for the twin-probe configuration is ρ = πaV/I, where R is the measured resistance in ohms, V is the measured voltage in volts, I is the known current in amps, ρ is the resistivity in ohm-meters, π is 3.141592---, and a is the probe separation in meters. This description of resistivity survey has concentrated on the details of current flow and associated electric fields in the vicinity of the scanning probe array. The remainder of the electrical circuit, typical of a site survey, can be seen in Figure 6.4, where stationary remote probes are shown 100 m removed from the scanning probes and the survey grid area. As noted earlier, the remote probes and the connecting wires are required to complete the electrical circuits for both the injected current and the voltmeter.

114 ~ Lewis Somers

Figure 6.2. Vertical section through uniform soil showing current (I) injection electrode and voltage (V) measuring electrode, with associated current flow (arrows) and electric field contours (circles). The significance of the distance between the current probe and the voltage probe (a) is discussed in the text.

Figure 6.3. A multidepth survey probe configuration showing current flow (arrows) and electric field contours (circles) at various depths.

An interesting variation on the twin array is the square array. In this array the four electrodes (two for the current injection and two for the voltmeter) are arranged in a square geometry, typically 1 m on a side. Various wiring configurations are possible and are referred to as the alpha, beta, and gamma probe configurations. Each configuration has a unique directional sensitivity and when data from each are properly combined this array offers improved resistivity map quality. This probe configuration is well known in the semiconductor industry and was explored by Clark (1996) some

Resistivity Survey ~ 115

Figure 6.4. The field configuration for a typical resistivity survey.

decades ago for archaeological applications. A wheeled version with automated data collection using the RM15 resistance meter has recently been developed (Figure 6.5). The wheeled configuration offers faster data collection and improved image quality, speed being of particular value in reducing the field time and cost of resistivity survey.

Resistivity Survey Design and Field Procedures The keys to successful resistivity survey are adequate data sample density, highquality data, meticulous field procedures, and appropriate data processing. Adequate data sample density is implemented by collecting as many (2, 4, 8, or 16) data samples per square meter as required to satisfy the survey objectives. High-quality data (i.e., large dynamic range data) are obtained by appropriate instrument configuration (voltage, current, averaging time) combined with impeccable field procedures. Data quality must be monitored continuously in the field by the surveyor throughout the survey. Appropriate data processing includes a number of linear and nonlinear operations applied in the correct sequence. After the appropriate data sample density is identified, the field survey begins by deployment of the remote electrodes to an arbitrary, but centrally located, position on the site and randomly sampling site resistivity on a large scale. This will provide some

116 ~ Lewis Somers

Figure 6.5. A wheeled square array survey system with automated data logging. An RM15 resistance meter can be seen on the instrument platform (copyright Geoscan Research [USA] 2001).

perspective on site uniformity and thus some guidance on instrument configuration. Following this site assessment the resistivity meter is configured to meet the survey design requirements. Data dynamic range is the principal issue. The RM15 resistivity meter is capable of delivering resistivity measurements with a dynamic range greater than 1:2,000. To achieve this dynamic range it is essential to configure the output voltage, current, gain, and auto-log speed (averaging) appropriately. This is achieved by placing the probe array in the ground at the configuration station (see below) and adjusting these instrument parameters until the resistance reading is in the appropriate range and is stable to 1:1,000. For large-area surveys it is useful to establish a configuration/drift/calibration station. This station consists of four arbitrary but precisely maintained probe positions: two for the remote probes and two for the mobile probes. Probe position must be maintained within <±3.0 mm in both depth and location. At this station it will be possible to confirm instrument stability and performance throughout the survey. It will also be possible to monitor site resistivity change during the survey. In addition, if the four probes are arranged in a Wenner configuration, an absolute soil resistivity measurement can be recorded at this station.

Resistivity Survey ~ 117 By convention and in order to be compatible with data-processing software and data file management protocols, it is best to start surveying each grid in the southwest corner. Data collection proceeds from the southwest corner to the north edge of the grid (first traverse). Subsequent scans can be performed in the parallel or zigzag mode. Surveys are usually performed in rectangular grids or units (i.e., 20 × 20 m, 30 × 30 m, 5 × 20 m, and so on). The site is divided into these units by land survey methods. Grid subdivision for data sample location is implemented by means of fiberglass survey tapes combined with the automated data logging features in the RM15 resistivity meter. Data in each grid are referenced to the south edge (a grid baseline) and the southwest corner (an individual grid datum) of the grid unit. By using this concept and preserving the grid corner locations, it is possible to relocate an anomaly for testing in the field to within a few centimeters. Acquisition of low-quality data is probably the single most important cause of failed surveys. Data quality control is a field activity and the surveyor must monitor the stability and “reasonableness” of each reading. Stability is governed by instrument configuration, as discussed above. “Reasonableness” refers to monitoring data values as collected, sample by sample, for continuity. For example, readings that differ significantly (factor of two) from adjacent readings are either very significant or very erroneous. In either case they should be confirmed by resampling.

Resistivity Survey Data Processing There are three principal stages in resistivity data processing. The first is concerned with mapping the large features and the high-contrast features. The second is concerned with mapping the small features and the low-contrast features. The former does not generally require high pass filtering of the data; the latter does. Stage three identifies and separates the high-resistivity features from the lowresistivity features, a valuable archaeological interpretation concept and tool. Additional data-processing operations depend on the nature of the data and the goals of the survey. The following discussion uses terminology associated with Geoplot 3.0 dataprocessing, analysis, and display software; however, the concepts and individual processes are generic and can be implemented with other image-based data-processing software if it is capable of handling data files containing both positive and negative values (bipolar data). Stage I: Large Features and High-Contrast Features Raw Data Merge

This operation combines data from individual survey grid files into a composite file. This creates a graphic for the entire survey and enables all analysis and dataprocessing algorithms.

118 ~ Lewis Somers Edit/Remove Defective Data

Data obviously inconsistent with soils, sediments, geology, and archaeological features should be removed. This is implemented with the search and replace processing function. Defective data are usually replaced with the dummy-data value, 2047.5. Display and Initial Interpretation

The composite map is displayed and visually examined for anomalies and archaeological features of interest. This process is also used to distinguish between archaeological and geological features and anomalies. Interpolate Data

The composite file data are interpolated to achieve the same data sample density in the north–south and east–west directions. This may or may not be necessary depending on the data sample densities used during the survey. The interpolation operation is used to expand the lower density direction data in steps of ×2. A final data sample density of 2 × 2 per meter is useful for display, production of hard copy, and export to other software packages. Display and Interpretation

The uniformly sampled data file is displayed and visually examined by means of gray scale, relief, trace, contour, or false color plots as required. At this point in the processing sequence, both the high-contrast features and the large features present in the survey data will be apparent. Variations in the site soils, sediments, and geology will also be apparent. Archaeologically interesting features and anomalies are identified for further work (e.g., testing). Stage II: Small Features and Low-Contrast Features High Pass Filtering

A high pass filter should be applied to the uniformly sampled data file (end of Stage I processing). This is implemented with a large, uniformly weighted high pass filter window (the default high pass filter parameters in Geoplot). After filtering, the small features and the low-contrast features in the survey will be more visible. The high pass filter operation has effectively subtracted the local average background resistivity from the survey data, creating a new (filtered) data file with an average value of zero. In addition to increasing the visibility of small low-contrast features, the high pass filter has also created a file with both positive and negative data values. The positive values are located wherever the survey data are higher in resistivity than the local mean resistivity, and the negative values are located wherever the survey data are lower than the local mean resistivity. Conceptually, the positive data now represent the high (higher

Resistivity Survey ~ 119 than the local mean) resistivity features and the negative data now represent the low (lower than the local mean) resistivity features. Positive values can be confidently interpreted as areas/features/anomalies with higher resistivity than that of the immediately surrounding soils. Conversely, negative values can be confidently interpreted as areas/ features/anomalies with resistivity that is lower than that of the surrounding soils. This bipolar segregation provides a convenient and valuable interpretative opportunity because resistivity features with similar use, formation processes, and material components tend to have similar contrast polarity. It also provides an opportunity to separate the high- and low-resistivity features into separate files and maps. Stage III: Separating High- and Low-Resistivity Features Clip Operation

To isolate the high-resistivity features, the high pass filtered data from 0 to 9,999 are clipped and the high-resistivity data are saved in a new file (positive/high). The lowresistivity features are isolated by reloading the high pass filtered data, clipping from –9,999 to 0, and saving the low-resistivity data to a new file (negative/low). At this point in the processing sequence (positive) high-resistivity data will be assembled in one file and (negative) low-resistivity data will be assembled in another file. Archaeological features with similar materials and formation histories will be found in each of these files—for example, post molds in one file and hearths in the other. Additional processing might include noise reduction to improve feature visibility. Low Pass Filtering

A low pass filter can be applied to the uniformly sampled data file (end of Stage I processing) to reduce the random noise (random instrument, soil, and operator defects in the data). This is best done with a low pass filter window diameter approximately equal to the size of the features of interest. That is, the low pass filter window length and width should be chosen to be approximately matched to the features of interest. If the window is too small, the full benefit of filtering will not be realized. If the window is too large, the feature of interest will be smeared and less “legible.” Both uniform and Gaussian weighted filters should be tried; one may be slightly better than the other. At this point in the processing sequence, the random noise has been reduced and low-contrast feature and anomaly visibility improved. The modest reduction in spatial resolution (detail) that accompanies low pass filtering can be a small price to pay for improved anomaly detection. Archaeologists typically use geophysical results in terms of pattern recognition: if it looks like a wall trench house, then it likely is one. However, much more attention needs to be directed toward statistical analyses of the data. For example, when the signal-to-noise ratio is high, anomalies are easily recognized. As the signal-to-noise ratio approaches zero, it becomes increasingly difficult to recognize anomalies with any con-

120 ~ Lewis Somers fidence, because their probability distributions overlap with the random background probability distribution. Figure 6.6 depicts data with three different signal-to-noise ratios and their associated probability distribution functions. The overlap is evident. Under these circumstances it is possible to choose a threshold that can be used to statistically select anomalies with a known confidence level. By examining an anomaly-free region of the survey, the standard deviation (SD) of the random background noise can be obtained. With this, various threshold values (1 SD, 2 SD, 3 SD) and their associated levels of statistical confidence can be chosen. It will then be possible to select all anomalies greater than the desired threshold, resulting in a map of statistically significant anomalies.

Test Unit Design and Placement Test unit placement is typically concerned with the edges of recognizable features, architectural or otherwise. The interior and exterior of the feature are also of interest. When edge features are of concern, a linear trench perpendicular to the feature is recommended. The test trench should be placed perpendicular to the edge and overlap into the low- and high-resistivity areas, a distance approximately equal to three twinelectrode separation distances (a in Figure 6.2). This test trench design will sample a minimum, yet meaningful, portion of the interior and exterior, as well as the transition region. Excavation results from a test trench meeting this design should provide access to elements of the archaeological record. Excavation to a depth of approximately three times the twin-electrode spacing or to a sterile horizon will reveal the subsurface components contributing to the measured resistivity values. On occasion, resistivity features may not be visible in the test trench or the profile. For example, a high salinity moist soil may be visually identical to a low salinity moist soil, but the measured resistivity could easily differ by a factor of 100 (consider a typical salt lick in an abandoned pasture). On these occasions careful attention must be paid to local variations in soluble ion concentration, physical soil particle size, and moisture gradients. Not all resistivity features are necessarily visible features.

Archaeological Interpretation of Resistivity Survey Data There are two phases to data interpretation: geophysical interpretation and archaeological interpretation. The geophysical interpretation is straightforward. High-resistivity anomalies are associated with high-resistivity soils/features, and low-resistivity anomalies are associated with low-resistivity soils/features. There is little more to say. An archaeological interpretation is more complex because there is not necessarily a simple 1:1 mapping from a high- or low-resistivity anomaly to a specific archaeological feature. For example, a small-diameter, low-resistivity anomaly could be a storage pit. Equally, a small-diameter, high-resistivity anomaly could be a storage pit; the sign of the feature contrast depends on the pit contents—for example, organic rich material versus stone/rocks or other high-resistivity material. A small-diameter anomaly could

Resistivity Survey ~ 121

Figure 6.6. Schematic representation of (a) zero, (b) low, and (c) high signal-to-noise ratios with probability distribution functions shown on the left (copyright Geoscan Research [USA] 2001).

122 ~ Lewis Somers also be a post mold. The point is that the archaeological interpretation must be performed within the context of the anticipated archaeological record. Experience teaches that archaeological interpretation of geophysical survey data is best performed as a joint effort between the surveyor and the archaeologist.

Example Surveys Fort Laramie (Figure 6.7) A twin electrode resistivity survey (data sample density 1 per square meter) at Fort Laramie performed by students at the 1992 National Park Service Geophysical Survey Workshop organized by Steve De Vore. Note the in situ footings and foundations as well as the wagon track running between building foundations. The ill-defined amorphous features in the north of the survey are said to be the result of excavation and backfill. (Credits: S. De Vore; sponsor: National Park Service.) Army City (Figures 6.8, 6.9) A large-area resistivity survey (data sample density 1 per square meter) of a burned, abandoned, and salvaged recreation “city” near Fort Riley, Kansas. Note building foundations, stream bed (now buried), utility lines, and the vehicle tracks associated with salvage operation. (Credits: M. Hargrave, L. Somers; sponsor: U.S. Army Corps of Engineers.) Yucca House, Chaco Canyon (Figure 6.10) A map of the standard deviation (SD) of resistivity survey data collected at Yucca House, a Chacoan outlier. The SD was calculated for a 1.5-square-meter area from 4 sample per square meter data. Note the high SD associated with the stone structure collapse in the west and in the area of kiva and other walls. Note the intermediate values in the open spaces and note the extremely low values in the east, which are possibly associated with a clay resource. Finally, note the intermediate values, possible activity areas surrounding the possible clay resource. (Credits: D. Gowacki, L. Somers; sponsor: National Park Service, Crow Canyon.) Mission San Marcos, Spanish Mission (Figure 6.11) Resistivity survey (data sample density 1 and 2 per square meter) and magnetic field gradient survey (data sample density 8 and 16 per square meter) of a Spanish Mission church and convento. The dark gray linear features are low-resistivity interior and exterior adobe walls. The light gray area surrounding the walls is adobe melt. The contour lines are the positive magnetic field gradient contours associated with weakly

Resistivity Survey ~ 123

Figure 6.7. Resistivity survey at Fort Laramie.

124 ~ Lewis Somers

Figure 6.8. Large-area resistivity survey at Army City.

Resistivity Survey ~ 125

Figure 6.9. Army City, detail (color illustration appears on the CD).

Figure 6.10. Standard deviation of resistivity survey data from Yucca House (color illustration appears on the CD).

126 ~ Lewis Somers

Figure 6.11. Resistivity survey and magnetic field gradient survey at Mission San Marcos (color illustration appears on the CD).

magnetic foundation stones supporting the exterior walls. Note the large dark (adobe collapse) area in the southwest corner (lower left), which is associated with a large rectangular magnetic foundation stone feature. (Credits: K. de Dufour, B. Chada, L. Somers; sponsor: American Museum of Natural History, D. H. Thomas.) Prehistoric House Pits, Prehistoric Coastal California Occupation (Figure 6.12) Resistivity survey (data sample density 2 per square meter) of prehistoric house pits. Dark “circular” features are high-resistivity house pits. House pits exhibit high resistivity because postoccupation fill has higher resistivity than the deeper horizons into which the pits were cut. The highlighted features are all confirmed by excavation. The random dark features are recent heavy-machine surface disturbance. The very small (1 m) black rectangles are random test units placed on the site prior to the resistivity survey. Six are circled at the top of the map. Note the difference in site assessment that results from the six to eight random test units compared with that from the resistivity survey. The costs of the random test units and the resistivity survey were comparable. (Credits: L. Somers; sponsor: Statistical Research.)

Resistivity Survey ~ 127

Figure 6.12. Resistivity survey of prehistoric coastal California house pits (color illustration appears on the CD).

128 ~ Lewis Somers Shields Complex, Early Southwest Occupation (Figure 6.13) Resistivity survey (data sample density 1 per square meter) and magnetic field gradient survey (data sample density 8 per square meter) of the Shields site. Dark “circular” features are high-resistivity house pits aligned along a known road. Red and green contour lines (see the color illustration on the CD), when co-located with dark resistivity features, are strong magnetic anomalies associated with a burned house. House pits exhibit high resistivity because postoccupation fill has higher resistivity than the deeper horizons into which the pits were cut. The east–west stripes in the northern portion of the map are recent agriculture. (Credits: Crow Canyon Interns, L. Somers; sponsor: National Geographic, Crow Canyon, M. Varian.)

Figure 6.13. Resistivity survey at the Shields Complex site (color illustration appears on the CD).

Resistivity Survey ~ 129

References Cited Barker, P. 1995 Techniques of Archaeological Excavation. 3rd ed. Batsford, London. Bevan, B. W. 1998 Geophysical Exploration for Archaeology: An Introduction to Geophysical Exploration. Special Report No. 1. U.S. Department of the Interior, National Park Service, Midwest Archeological Center, Lincoln, Nebraska. Carr, C. 1982 Handbook on Soil Resistivity Surveying: Interpretation of Data from Earthen Archeological Sites. Center for American Archeology, Research Series, Vol. 2. Evanston, Illinois. Clark, A. J. 1996 Seeing Beneath the Soil: Prospecting Methods in Archaeology, new ed. B. T. Batsford, London. Drewett, P. 1999 Field Archaeology: An Introduction. Taylor & Frencis, London. Gaffney, C., and J. Gater 2003 Revealing the Buried Past: Geophysics for Archaeologists. Tempus, Gloucestershire, Great Britain. Grove, A. S. 1967 Physics and Technology of Semiconductor Devices. John Wiley and Sons, New York. Klein, K., and J. C. Santamarina 2003 Electrical Conductivity in Soils: Underlying Phenomena. Journal of Environmental Engineering Geophysics 8(4):263–273. Weymouth, J. W. 1986 Geophysical Methods of Archaeological Site Surveying. In Advances in Archaeological Method and Theory, vol. 9, edited by M. B. Schiffer, pp. 311–395. Academic Press, New York.


Ground-Penetrating Radar Lawrence B. Conyers

Ground-penetrating radar (GPR) has recently gained a wide acceptance in the archaeological community as a method that can quickly and accurately locate buried archaeological features, artifacts, and important cultural strata in the nearsurface. The GPR method has been especially effective in certain sediments and soils between about 20 cm and 5 m below the ground surface, where the targets to be imaged are fairly large, hollow, or linear or have significant physical and chemical properties that contrast with the surrounding medium. Features as diverse as Mayan house platforms and plazas (Conyers 1995), burial tombs (Goodman and Nishimura 1993), historic cellars, privies, and graves (Bevan and Kenyon 1975), camp sites (Vaughan 1986), and pit dwellings and kivas (Conyers and Cameron 1998) have been discovered and mapped using the method. The archaeological community has also recently seen the need for near-surface mapping using GPR in order to identify buried cultural remains for protection and future preservation and as a planning tool for selective excavation. Ground-penetrating radar has a reputation as one of the more complex of archaeological geophysical methods because it collects large amounts of reflection data from numerous transects within grids, oftentimes producing massive three-dimensional databases. The ability to detect multiple interfaces at different depths below the surface, the interpretation of these numerous reflections, and the difficulty in correlating the abundance of reflections between many profiles within a grid can make GPR data collection and processing a somewhat intimidating venture for the uninitiated. However,

132 ~ Lawrence B. Conyers with modern data acquisition and processing and a knowledge of how radar energy travels and reflects from interfaces in the ground, GPR mapping in archaeology need not be as daunting as its reputation suggests. Some of the earliest model GPR systems recorded raw subsurface reflection data on paper printouts that allowed little postacquisition processing. Although these radar systems, a few of which are still in use, can many times yield valuable subsurface information, modern digital systems record reflection data on a computer hard drive for later filtering, processing, and sophisticated data analysis. Most important, when the data are digital, a computer can process, filter, and enhance raw field data almost immediately after they are collected. Computer manipulation of the digital data, which removes unwanted noise and enhances the portions of the signal that are important, allows for rapid data processing and dramatically increases subsurface resolution and interpretation of complex data sets. Accompanied by a trend in equipment miniaturization, computer processing of the acquired GPR data can now occur immediately after they are acquired and interpretation can often begin while the operators are still in the field. The recently acquired speed at which data can be filtered, processed, and interpreted can often allow archaeologists to produce three-dimensional images of buried features just hours after data are acquired. When this is done, further data acquisition or the planning of excavations to confirm features of interest that have been discovered can begin almost immediately, making geophysical data collection, interpretation, and excavation an iterative process. Modern GPR systems are quite compact and easy to use. The typical system consists of surface antennas, a radar system to produce pulses, a computer to process and save the data, a video monitor, a keyboard, and a power source (Figure 7.1). This system can be easily transported to the field by plane, car, and backpack. Processing of data can be done in the laboratory or while in the field using a portable laptop computer. A complete system can be purchased for about $20,000 and used systems for substantially less. GPR systems with multiple antennas can also be rented from a number of vendors for about $200 per day. With careful planning and the ability to work in areas that are not cluttered or topographically complex, GPR data for grids of 50 × 50 m or more, with a 50-cm profile separation, can be collected in a day. If transects within grids must be shortened or lengthened to avoid obstacles or many small grids must be constructed to cover the targeted areas, the amount of ground that can be covered can be substantially less. It is also often desirable to collect and process data from an area one day, process them that night, and then re-collect them again the next day using different antenna configurations, grid orientations, or other collection parameters depending on the results. Processing programs that produce images very quickly from complex databases now allow these kinds of immediate results in the form of processed profiles and plan maps of reflections at various depths in the ground. These types of data-analysis products will be discussed more below.

Ground-Penetrating Radar ~ 133

Figure 7.1. The Geophysical Survey Systems Subsurface Interface Radar (SIR) system, Model 2000. The radar control box contains a computer for data collection, processing, and data storage and a computer screen. The radar signal is transmitted to and from the computer by cables, which are attached to the 400-MHz antennas in this photo. Reflection data are visible on the computer monitor during collection and processing.

History of GPR in Archaeology The first attempt at what would today be called ground-penetrating radar was conducted in Austria in 1929 to determine the depth of ice in a glacier (Stern 1929). This pioneering work got little attention at the time but demonstrated that electromagnetic energy could be transmitted in media other than air. The first large-scale application of radar was during World War II when the British, and later the Americans, used crude but effective systems to detect reflections of radar pulses from airplanes in the sky. The word radar was coined just prior to that time and is an acronym for RAdio Detection And Ranging (Buderi 1996). Little work was done with radar transmission in solid media until 1972 when a prototype GPR system was built by NASA and sent on Apollo 17 to the moon to study the electrical and geological properties of the crust. The archaeological community was quick to grasp the potential of using GPR to both locate and map buried archaeological features and associated sediment and soil layers. One of the first applications to archaeology was conducted at Chaco Canyon, New Mexico (Vickers and Dolphin 1975), where buried walls at depths of up to 1 m were imaged. These rudimentary studies at Chaco Canyon were soon followed by a number of applications in historical archaeology in which GPR was successfully used

134 ~ Lawrence B. Conyers to search for buried barn walls, stone walls, and underground storage cellars (Bevan and Kenyon 1975; Kenyon 1977). In these early studies, what were described as “radar echoes” were recognized in paper printouts as being generated from the tops of buried walls, and depth estimates were made, using approximate velocity measurements from local soil characteristics. Initial successes in historical archaeology applications were followed in 1979 by work at the Hala Sultan Tekke site in Cyprus (Fischer et al. 1980) and the Ceren site in El Salvador (Sheets et al. 1985). Both of these GPR surveys produced unprocessed reflection profiles that were successful in delineating buried walls, house platforms, and other archaeological features as radar anomalies. Most of the initial GPR successes can be attributed to the very dry material that covered the archaeological remains, which was almost “transparent” to radar energy propagation, allowing for deep energy penetration and the production of relatively uncomplicated reflection records that were easy to interpret in the field. A comprehensive series of GPR surveys were conducted in Japan in the mid-1980s to locate buried sixth-century houses, burial mounds, and what were called “cultural layers” (Imai et al. 1987). These studies were successful in identifying ancient pit dwellings with clay floors, which were buried in some cases by as much as 2 m of volcanic pumice and loamy soil. The interface of the house floors with the overlying pumice produced very distinctive reflections that were easily recognizable on GPR profiles. Most important, much of the site discovered by GPR was excavated to confirm the results. Other ancient features were discovered and found to be burial mounds and associated trenches. Three distinct “cultural layers,” visible in reflection profiles, were found to be buried soil horizons containing many stone artifacts, with each layer delineating a different occupational period. This important conclusion allowed GPR profiles to be used to map the ancient landscape of each distinct living surface throughout portions of the site that had not been excavated. Throughout the late 1980s and early 1990s, GPR continued to be used successfully in a number of archaeological contexts, with a growing usage in cultural resource management (CRM) projects. In most cases these studies were little more than “anomaly hunting” exercises. Usually unprocessed or partially processed GPR profiles were viewed as paper records or on a computer screen as they were acquired, and interesting reflections, which could possibly have archaeological meaning, were targeted for excavation. These types of “on the fly” acquisition and interpretation methods led to mixed results, with a few successes but sometimes leaving field archaeologists with the impression that GPR was a “hit or miss” method at best. Prior to 1993 the most encompassing and successful archaeological applications of GPR were those employed in the mapping of the houses and burial mounds in Japan, discussed above (Imai et al. 1987). These successes were soon followed by numerous other Japanese GPR surveys, conducted by Dean Goodman and his colleagues (Goodman 1994; Goodman and Nishimura 1993; Goodman et al. 1994;

Ground-Penetrating Radar ~ 135 Goodman et al. 1995). These advancements were possible because about the same time GPR manufacturers began to produce systems that could store reflection data as digital files, allowing large amounts of data to be collected for later processing. Also, inexpensive and increasingly powerful personal computers had become available that could process digital data in ways that were not previously possible, at least on the typically small archaeological budgets. The pioneering studies of Goodman and his collaborators led to many important GPR acquisition and data-processing techniques, including amplitude slice maps, computer-simulated two-dimensional models, and three-dimensional reconstructions of buried features (Conyers 2004; Conyers and Goodman 1997; Goodman et al. 1994; Goodman et al. 1995; Goodman et al. 1998). The development of the amplitude slice-map method was one of the single most important developments that made GPR technology more understandable to the archaeological community. This data-processing and mapping technique allows huge amounts of data from many tens of reflection profiles in a grid to be processed simultaneously, producing maps of the areal distribution of reflections analogous to arbitrary levels in standard archaeological excavations. Another important development was computer programs that have the ability to produce synthetic computer models of buried archaeological features and associated stratigraphy, which were also used as an aid in interpretation during data analysis (Conyers 1995; Goodman 1994). These advancements and the application of many more advanced data-processing techniques in the 1990s (Goodman et al. 1995) demonstrated that even radar data that do not yield immediately visible reflections can still contain valuable reflection data when computer processed (Conyers and Cameron 1998). Recent research has demonstrated (and quantitatively assessed) the differences in data quality among numerous antenna frequencies and the differences in data quality that can occur because of line spacing and the density of reflections along transects (Neubauer et al. 2002). This type of research has demonstrated how variable field acquisition parameters can greatly influence the final product (Conyers et al. 2002a, 2002b). The ability of GPR to collect data in a three-dimensional block has recently led some researchers to begin analyzing reflected wave amplitudes in complex ways (Goodman et al. 1998; Moran et al. 1998). If higher amplitudes can be shown to denote the location of important buried archaeological features, then their locations in three dimensions can be analyzed and visualized using a number of software programs. In this way, the lower amplitude reflections are effectively removed from the data set and only those of importance remain. The locations in space of certain radar amplitudes, which are proxies for the actual location of features in the ground, are then rendered in a number of fashions to produce “virtual reality” images of what lies below the surface, much like CT scans are used in the medical profession (Conyers et al. 2002b). This has been done by cutting through the block of data (Neubauer et al. 2002) or rendering out only the higher amplitudes and presenting images of buried archaeological remains in a three-dimensional, rotating image (Conyers et al. 2002b; Leckebusch and Peikert 2001).

136 ~ Lawrence B. Conyers

The GPR Method GPR is a geophysical method that can accurately map the spatial extent of nearsurface objects and archaeological features or changes in soil media and ultimately produce images of those materials. Radar waves are propagated in distinct pulses from a surface antenna, reflected off buried objects, features, bedding contacts, or soil units, and detected back at the source by a receiving antenna. As radar pulses are transmitted through various materials on their way to the buried targets, their velocity changes depending on the physical and chemical properties of the material through which they travel (Conyers 2004; Conyers and Goodman 1997). The greater the contrast in electrical and to some extent magnetic properties between two materials at a subsurface interface, the greater the strength of the reflected signal and therefore the greater the amplitude of the reflected waves. When the travel times of energy pulses are measured and their velocity through the ground is known, distance (or depth in the ground) can be accurately measured to produce a three-dimensional data set (Conyers and Lucius 1996). Each time a radar pulse traverses a material with a different composition or water saturation, the velocity changes and a portion of the radar energy is reflected back to the surface to be recorded at the receiving antenna. The remaining energy continues to pass into the ground to be further reflected, until it finally spreads and dissipates with depth.

Recording Radar Reflections In the GPR method, radar antennas are moved along the ground in linear transects and two-dimensional profiles of a large number of periodic reflections are created, producing a profile of subsurface stratigraphy and buried features along each line (Figure 7.2). When data are acquired in a series of transects within a grid, and the reflections are correlated and processed, an accurate three-dimensional picture of buried features and associated stratigraphy can be constructed. Reflections occur at buried discontinuities where there are changes in the electrical properties of the sediment or soil, variations in water content, lithologic changes, or changes in bulk density. Reflections can also be produced at interfaces between anomalous archaeological features and the surrounding soil or sediment. Void spaces in the ground such as caskets in cemeteries, tunnels, and buried pipes or conduits made of either metal or plastic will also generate strong radar reflections as a result of a significant change in radar-wave velocity. These features tend to produce reflection hyperbolas generated from a distinct “point feature” in the subsurface (Figure 7.2), which could be archaeological in origin but also produced from buried stones, tree roots, or tunnels created by burrowing animals. These point source hyperbolas are produced because GPR antennas generate a transmitted radar beam that propagates from the surface down into the ground in a conical pattern, radiating outward as it travels deeper in the ground (Conyers 2004; Conyers and Goodman 1997:35). Radar energy leaving the surface antenna will therefore spread out with depth and be reflected

Ground-Penetrating Radar ~ 137

Figure 7.2. A 400-MHz profile across a pithouse floor. Buried water pipes are visible as reflection hyperbolas.

from buried objects that are not directly below the antenna. Although much of the energy is still radiated directly downward, some wave travel paths that move outward in the cone of projection are longer and the reflections generated from objects not located directly below the antennas will be recorded as being deeper in the ground. The hyperbolas visible in reflection profiles are generated because energy will be recorded from a point source prior to the antenna being directly on top of it, and the antenna will continue to “see” the object after it has passed. The hyperbola is generated because the time it takes for the energy to move from the antenna to the object along an oblique path is greater the farther the antenna is away from the source of the reflection, and when time is converted to depth, the reflection is recorded deeper in the ground. As the buried source comes closer, the reflection is recorded closer in time until the antenna is directly on top of the buried source of the reflection. The same phenomenon repeats in reverse as the antenna passes away from the source, resulting in a hyperbola in which only the apex denotes the actual location of the buried source, with the arms of the hyperbola creating a record of reflections from radar pulses that traveled the oblique wave paths. The success of GPR surveys is to a great extent dependent on soil and sediment mineralogy, clay content, ground moisture, depth of burial, surface topography, and vegetation. It is not a geophysical method that can be immediately applied to any subsurface problem, although with thoughtful modifications in acquisition and dataprocessing methodology, GPR can be adapted to many differing site conditions. Although radar-wave penetration and the ability to reflect energy back to the surface are enhanced in a dry environment, moist soils can still transmit and reflect radar energy and GPR surveys in these conditions can yield meaningful data.

138 ~ Lawrence B. Conyers

Depth of Penetration and Resolution The depth to which radar energy can penetrate and the amount of definition that can be expected in the subsurface is partially controlled by the frequency of the radar energy transmitted. The frequency controls both the wavelength of the propagating wave and the amount of spreading and attenuation of the energy in the ground. One of the most important variables in GPR surveys is the selection of an antenna with the correct operating frequency for the desired depth and resolution of target features. Commercial GPR antennas range from about 10 to 1,200 megahertz (MHz) center frequency. Proper antenna frequency selection can in most cases make the difference between success and failure in a GPR survey and must be planned for in advance. In general, the greater the necessary depth of investigation, the lower the antenna frequency that should be used. Lower frequency antennas (below 100 MHz) are large, heavy, and difficult to transport to and within the field (Figure 7.3). They must be either towed behind a vehicle on a trailer or sled or carried manually. In contrast, a 900-MHz antenna or other higher frequency antennas are quite small and can easily fit into a suitcase (Figure 7.4). Subsurface feature resolution varies with radar energy frequency. Low-frequency antennas (10–120 MHz) generate energy with wavelengths of many meters that can

Figure 7.3. The 25-MHz antenna capable of transmitting radar energy to more than 20 m is difficult to transport to the field and within grids. It is capable of resolving only very large objects of many meters in dimension.

Ground-Penetrating Radar ~ 139 penetrate up to 50 m in certain conditions but are capable of resolving only very large subsurface features. In contrast, the maximum depth of penetration of a 900-MHz antenna is about 1 m or less in typical soils, but its generated reflections can resolve features down to a few centimeters. A tradeoff therefore exists between depth of penetration and subsurface resolution. These factors are highly variable, depending on many site-specific factors such as overburden composition and porosity and the amount of moisture retained in the soil.

How Materials in the Ground Affect the GPR Signal The primary goal of most archaeological GPR investigations is to differentiate subsurface interfaces, whether they are archaeological or geological. All sediment and soil layers, as well as ar- Figure 7.4. The 900-MHz antenna chaeological features, have particular electrical and can transmit energy to about 1 m at magnetic properties that affect radar-wave velocity, most but can resolve features to about reflection, and dissipation in the ground (Conyers 10 cm in diameter. 2004; Conyers and Goodman 1997). The propagation velocity of radar waves depends on a number of factors, the most important ones being the electrical and chemical properties of the material through which they pass (Olhoeft 1981). Radar waves in air travel at the speed of light, which is about 30 cm per nanosecond (one nanosecond is one billionth of a second). When radar energy travels through dry sand, its velocity slows to about 15 cm per nanosecond (Conyers 2004). If the radar energy were then to pass through water-saturated sand or a clay layer, its velocity would slow further to about 5 cm per nanosecond or less. Radar energy both disperses and attenuates as it radiates into the ground. When portions of the original transmitted signal reflect back toward the surface, they suffer additional attenuation before finally being recorded at the surface. Therefore, to be detected as reflections, important subsurface interfaces not only must have a sufficient electrical (or magnetic) contrast at their boundary but also must be located at a shallow enough depth where sufficient radar energy is still available for reflection and reception back at the surface. As radar energy propagates to increasing depths, the signal becomes weaker and is spread out over more surface area. Less energy is then available for reflection and it is likely only very low-amplitude waves will be recorded. It is therefore usually necessary to enhance reflections that come from deeper in the ground, using an amplitude adjustment method called range gaining. The gain factors to be applied

140 ~ Lawrence B. Conyers to the reflections are usually adjusted specifically for each site and are unique for those ground materials present at the time the survey is performed. Sometimes these gain functions are applied prior to recording the data in the field, or they may be applied afterward when the data set is processed. The velocity of radar travel in the ground can also be defined by a particular constant for each material called its relative dielectric permittivity (RDP). The higher the RDP, the slower radar waves travel in that material. The ability of the ground to transmit radar energy to depth is not affected by the RDP, which is only a measurement of the velocity of energy movement. Whether energy will be able to penetrate into the ground is purely a function of the electrical conductivity (and to a lesser extent the magnetic permeability) of that material. Ground that has a higher electrical conductivity, such as some wet clays, will remove the electrical portion of the propagating electromagnetic radar waves, effectively attenuating all radar propagation. Low-conductivity ground, such as dry quartz sand, will readily allow radar energy propagation to depth, usually at a fairly high velocity. But some materials could have a high RDP (radar will travel through them at a relatively slow rate) but also a low electrical conductivity, and even though the energy is moving relatively slowly, it will still propagate fairly deeply into the ground. In ground with a high electrical conductivity, no radar energy will travel to any great depth no matter what its RDP might be. Reflection of radar energy in the ground occurs at interfaces between materials with differing RDP. Any change in velocity of propagating radar energy will generate a reflected wave that travels back to the surface and will be detected at the surface antenna. Only a portion of the radar energy is reflected, while some continues to depth to be reflected further until it is finally absorbed by the ground. Changes in RDP at buried interfaces are primarily a function of the magnitude of the difference in electrical properties between two materials on either side of that interface, which also affects velocity (and RDP). The greater the change in RDP at those interfaces, the stronger the reflected wave (Sellman et al. 1983). Higher amplitude reflected waves recorded at the surface antenna are a product of reflection from the contact between highly different materials such as sand and clay or, for example, the boundary between a house floor and the surrounding matrix. The ability to discern radar reflections in GPR data is related to the amplitude of those reflected waves. Higher amplitude waves produce more visible reflections, which can be denoting the presence of buried archaeological materials or natural stratigraphic interfaces. Lower amplitude reflections, such as those generated from subtle soil changes, usually occur when there are only small differences in the electrical properties between layers. Subtle changes in the nature of buried soil or sediment layers are usually all but invisible to the human eye but are still recorded in GPR profiles as small digital changes in their amplitudes. In order to enhance these changes, so they may be mapped and viewed, sophisticated amplitude analyses discussed below must be applied to the data set.

Ground-Penetrating Radar ~ 141

Computer Processing to Produce Images of Features in the Ground The standard image for most GPR reflection data is a two-dimensional profile, with depth on the x-axis and distance along the ground on the y-axis (Figure 7.2). Profiles are constructed by stacking together many reflection traces, obtained as the antennas are moved along a transect. Reflection profiles are most often displayed in gray scale, with variations in the reflection amplitudes measured by the depth of the shade of gray. Color palettes can also be applied to amplitudes in this format, but it is usually easier for the human brain to process gray-scale changes than color differences. Often two-dimensional profiles must be corrected to reflect changes in ground elevation. Only after this is done will images correctly represent the real world when the ground surface is irregular. This process, which is usually only important when topographic changes are great, requires detailed surface mapping of each transect within the data grid and then reprocessing the data from each transect by adjusting all reflection traces for surface elevation changes (Figure 7.5). Standard two-dimensional images can be used for most basic reflection data interpretation, but analysis of tens or hundreds of images within a grid can be tedious. In addition, the origins of each reflection in each profile must sometimes be determined before accurate and meaningful subsurface maps can be produced. Detailed and accurate profile interpretation comes only with a good deal of experience. A more sophisticated and faster type of GPR data manipulation of large data sets is amplitude slice-mapping, which creates images of the spatial distribution of reflected wave amplitude differences within a grid. The result can be a series of maps that illustrate

Figure 7.5. If the ground surface is not flat, profiles must be corrected for surface elevation changes in order to produce a more accurate twodimensional view of the subsurface. In this profile the high-amplitude reflection at about 65 cm on the left of the profile becomes difficult to see on the right as it becomes more deeply buried. The surface soils in this area absorbed much of the radar energy and very little was available for transmission below about 90 cm.

142 ~ Lawrence B. Conyers the three-dimensional location of reflection anomalies derived from a computer analysis of many two-dimensional profiles (Figure 7.6). This method of data processing can only be accomplished with a computer using GPR data that are stored digitally. Raw reflection data are nothing more than a collection of many individual traces along two-dimensional transects, each of which contains a series of waves that vary in amplitude depending on the amount and intensity of energy reflection that occurred at buried interfaces. An analysis of the spatial distribution of reflected wave amplitudes is important because it is an indicator of potentially meaningful subsurface changes in lithology or other physical properties of materials in the ground. If amplitude changes can be related to the presence of important buried features, the location of those changes can be used to reconstruct the subsurface in three dimensions. Areas of low-amplitude waves usually indicate uniform matrix material or soils while those of high amplitude denote areas of high subsurface contrast such as buried archaeological features, voids, or important stratigraphic changes. In order to be interpreted, amplitude differences must be analyzed in slices that examine only changes within specified depths in the ground. Each amplitude slice consists of the spatial distribution of all reflected wave amplitudes, which are indicative of these changes in sediments, soils, and buried materials. Amplitude slices need not be constructed horizontally or even in equal depth intervals. They can vary in thickness and orientation, depending on the questions being asked.

Figure 7.6. Example of an amplitude slice-map, showing changes in amplitude in plan view, with each slice representing about 20 cm in the ground. These slices are imaging rubble within a historic building foundation in the 100–140 cm depth slices (color illustration appears on the CD).

Ground-Penetrating Radar ~ 143 To produce horizontal amplitude slice-maps the computer compares amplitude variations within traces that were recorded within a defined time window. For instance, if data were recorded to a maximum of 30 nanoseconds in the ground, six slices of 5 nanoseconds in thickness would be analyzed and the spatial distribution of amplitudes in each slice would be produced. When this is done both positive and negative amplitudes of reflections are compared to the norm of all amplitudes within that window. No differentiation is usually made between positive or negative amplitudes in these analyses; only the magnitude of amplitude deviation from the norm is expressed. An abrupt change between an area of low and high amplitude can be very significant and may indicate the presence of a major buried interface between two media. Degrees of amplitude variation in each slice can be assigned arbitrary colors or shades of gray along a nominal scale. Usually there are no specific amplitude units assigned to these color or tonal changes. Slices that are produced in thicknesses based on radar travel time can readily be converted to depth slices, if the velocity of energy movement through the material (or its RDP) is calculated. This is the preferred format for most archaeological applications. There are a number of computer programs available that can estimate velocity of radar travel times from individual reflection profiles; alternatively, direct measurements can be made in the field if open excavations are present (Conyers and Lucius 1996). Direct velocity measurements can be obtained by inserting a metal object (a pipe or other object of this sort) horizontally into the face of an excavation and measuring its depth in centimeters. Metal is a perfect radar reflector, and when antennas are pulled over the buried pipe, it will often be visible as distinct reflection hyperbolas. Radar-wave travel times to the object can then be measured and velocity easily calculated. This average velocity can then be used to convert all time measurements to approximate depth in the ground for final data presentation.

Methods of Ground Truthing GPR Maps and Profiles Two-dimensional reflection profiles, once processed and corrected spatially, give an accurate representation of what lies below the surface. However, it is always important to recognize that reflection records do not necessarily mimic exactly what is in the subsurface. This is because radar energy travels not only in a vertical line from the surface antenna to the object or surface of interest and then directly back to the receiving antenna but also in other complex paths. It spreads out from the antenna, and therefore reflections are recorded that were generated from outside the plane of the transect and from in front of and behind the surface antenna. In addition, radar waves often reflect multiple times from buried objects, as they “bounce around” between layers in the ground and other large objects, before finally being recorded back at the surface. Waves of radar energy can also refract at boundaries between distinct layers, further creating a sometimes confusing picture of the subsurface. An understanding of the complexity of GPR reflection profiles comes with experience, as well as from directly comparing

144 ~ Lawrence B. Conyers reflection profiles to the “real world,” using ground truthing methods (comparing geophysical images with what is known to be in the ground). The same holds true for three-dimensional maps produced from the spatial analysis of the amplitudes of GPR reflections. These maps are produced from many hundreds or thousands of reflections in a grid. Often in ground truthing the amplitude slice-maps it is necessary first to compare the amplitudes mapped with the twodimensional profiles in order to make sure the origin of the mapped images in the slices is known. At that point, if subsurface confirmation of at least some of the reflections can be made, the overall confidence of the remaining portions of the maps increases. But in all cases some kind of subsurface confirmation of features imaged with GPR is important. This kind of verification can be accomplished by standard archaeological excavations including shovel test pits, square excavations, and trenches. If the mapped features visible in amplitude slice-maps are found to be strata of interest (for instance, a layer that could be a buried soil), cores or auger samples can be taken and their depth compared with the GPR reflections in individual profiles. Sometimes, if mapped features are fairly close to the surface, hard layers (perhaps rocks or buried walls) can be discerned by soil probes, which give less direct confirmation but are easy to use, and this testing is very quickly accomplished. In all cases interpretations based only on GPR reflections can be prone to errors. While many very distinct features, such as standing walls or hard-packed and dense stratigraphic surfaces, are easily recognizable and can be interpreted with some confidence, more subtle features such as soil composition changes are often difficult to discover and interpret. For this reason, integration of good subsurface information from cores, excavations, probes, or augers with GPR amplitude slice-maps and profiles is always a necessity.

Limits of GPR in Archaeology Although GPR is a powerful tool for imaging and mapping the subsurface, there are some limitations in its applicability to CRM archaeology, the most obvious limitation being depth of investigation. The trade-off that exists between depth of investigation and resolution can be important if buried features and stratigraphic interfaces of interest are buried too deeply. Below about 2–3 m, low-frequency antennas (300 MHz and lower) are necessary for the transmission of radar energy. With those antennas, resolution is diminished, making many subtle changes in beds and archaeological features all but invisible in GPR profiles and maps. If high resolution is necessary to map smaller features of interest, with the present technology, they must be fairly close to the surface. The chemical and physical properties of the medium through which radar energy must pass can also be a limiting factor in GPR studies. Any medium that is electrically conductive, such as some wet clays, or any sediment or soil with a high electrolyte

Ground-Penetrating Radar ~ 145 content (those high in salts or carbonate, for instance) will attenuate radar energy quite rapidly and often the resulting GPR data can be unusable. The same can hold true for sediment or soil that is highly magnetic, but materials of this sort are relatively rare. Soil moisture differences can often severely disrupt radar energy, producing reflections that are difficult to interpret and obscuring those that are potentially meaningful. If an area has been recently irrigated or there has been a recent heavy rain, pools of water can be differentially preserved in buried sediments and soils. When this happens, radar reflections may occur from both the pools of water and the zones or objects of interest, complicating the data. It is often difficult to know in advance whether ground conditions are conducive for GPR studies. Some have tried to predict GPR success based on soil survey maps or gross generalizations about the geology of an area (Collins and Kurtz 1998). While these types of analyses can be a useful guide in a general sense, GPR success in a specific area can usually only be determined by actually collecting and processing data. One of the greatest limitations to the GPR method in CRM archaeology is the timing of its use. Usually GPR surveys are conducted prior to excavations, which is only natural because archaeologists would always like to know in advance what is under the ground before they dig. When surveys are done in this way, there are usually anxious excavators waiting for results, with often unrealistic expectations that GPR surveys will tell them everything they want to know about the subsurface (Conyers 1999). Sometimes this approach works well and exciting archaeological features just jump out of processed maps, leaving little ambiguity about their origin. These types of features are usually those that are most distinct, such as house floors, walls, and other architecture that would be hard to miss by even the most inexperienced GPR interpreter. In much CRM archaeology, when the target features are often much more subtle, they can be difficult to find, and it is often challenging to make a definitive interpretation. In these cases interpretations that would please the excavator, the client, and the geophysical archaeologist can be arrived at only by merging and integrating information from excavations, GPR data, and other sources. Unfortunately, the timing of many CRM projects precludes this iterative process of give and take, making many GPR maps less useful to many archaeological projects than they should be. The correct way to use GPR in archaeological mapping would be to first collect the reflection data and interpret them, with the knowledge that the results will often reveal only a little about the subsurface. Those maps and profiles can then be used as a guide to test interesting features and horizons using excavations or coring and augering. The data from this ground truthing should then be integrated back into the GPR data so that horizons and features of interest can be remapped, using stratigraphic information obtained from the ground. This type of timing necessitates what amounts to a “first look” at the GPR, then a reinterpretation of it, and often a third round of data analysis, as new information comes to light. CRM archaeologists who expect this type of prolonged analysis and budget for it in terms of time and expenses will be much more satisfied with the final results.

146 ~ Lawrence B. Conyers

Examples of Successful GPR Surveys In some areas of North America the standard method of discovering buried archaeological features in CRM archaeology is with random excavations or trenches, which often discover buried materials but only after they are partially destroyed. Any features that might lie between the excavations remain undiscovered. One of the more egregious and destructive methods is to trench with backhoes, sometimes in a methodological way within areas to be surveyed (Figure 7.7). The picture in Figure 7.7 was taken near Phoenix, Arizona, in a planned shopping center where a buried Hohokam village was expected. In southern Arizona and New Mexico CRM firms typically trench huge areas of ground, discovering pithouse floors and other features in the soil and sediment that are recovered or visible in the walls of the trenches. In a test to determine whether GPR could identify these features nondestructively, a test survey was conducted at the Valencia site near Tucson, Arizona. A small grid of data (30 × 40 m) was collected where some backhoe trenching had already occurred, so it was known that pithouses and other cultural features were buried in the area. Much of the test area was also scheduled to be totally stripped to expose all buried remains, so it made for a very good test case for GPR mapping. The reflection data visible on the computer screen during collection were very poor in quality, containing mostly “noise” from nearby radio and television antennas and from cell phone and other electromagnetic signals. Each individual profile appeared to be totally worthless, at least on

Figure 7.7. Often large trenches are dug with backhoes to determine the presence or absence of archaeological features, but when this method is used nothing is known about the areas lying between the trenches.

Ground-Penetrating Radar ~ 147 first glance (Figure 7.8). In the hope of obtaining useful data, all profiles were computer filtered to remove many of the frequencies that appeared to be producing the noise, leaving only about 10 percent of the frequency data to interpret. In this processing step all frequencies above 600 MHz, which were creating much of the interfering noise from nearby radio transmitters and cell phones, were removed, which cleaned up the data considerably. The remaining noise consisted of horizontal lines on profiles, which was likely generated from antenna “ringing” and internal system noise, and it was then arithmetically removed from each profile by applying a background removal filter. After this step only the nonhorizontal reflections that were presumably generated from within the ground remained. When those processing steps were finished, Figure 7.8. A 500-MHz profile from the Valencia site in Tucson, Arithe floors of pit struc- zona. Many of the reflection data from the subsurface were obscured tures became imme- by noise from many different electromagnetic sources nearby, making this profile worthless in its present state. diately visible in the remnant reflections (Figure 7.9). Amplitude time slices were then produced from the filtered profiles, which showed many oval and rectangular features, which were the floors of pithouses (Figure 7.10). When the site was later stripped during excavations, 85 percent of the archaeological features that were discovered were vis- Figure 7.9. When the profile in Figure 7.8 was processed to remove the ible in the GPR maps interfering frequencies, a pithouse floor became visible.

148 ~ Lawrence B. Conyers

Figure 7.10. Hundreds of reflections from many profiles such as those in Figure 7.9 were compared, correlated, and gridded on a computer to produce amplitude slice-maps. This slice from 50–75 cm in depth shows the location of many pithouse floors and other extramural features (color illustration appears on the CD).

(Conyers and Wallace 2003). The remaining 15 percent not found in the GPR maps were subtle features composed of local clay that had not been burned or compacted, producing little in the way of lithological or chemical changes to reflect radar energy. A survey within the city limits of Alamagordo, New Mexico, demonstrated how even in severely disturbed areas, GPR is capable of producing images of buried features. A GPR survey was conducted within a water line right-of-way where three 10-inchdiameter pipes had been laid in trenches during the 1950s and 1960s. The city was unsure exactly where the pipes were located within the right-of-way and completely unaware of the presence of any buried archaeological features. During data collection it became immediately apparent in profiles where the pipes were located, as they produced very strong hyperbolic reflections when crossed by the antennas (Figure 7.2). Although the stratigraphy was quite complex because of the number of times the area had been trenched and backfilled, floors of a number of pithouses could be seen adjacent to and between the trenches. When all the profiles were processed into amplitude slice-maps it was apparent that earlier pipeline construction had cut directly through a Mogollon period village, destroying much of it but preserving portions of pithouses and other features in many places. The city engineers were quite disturbed by the results of the GPR surveys, but not because of the ancient village that had been partially destroyed. They were more concerned that this new technology could quickly produce images of the wealth of buried archaeological materials in their right-of-way (none of

Ground-Penetrating Radar ~ 149 which was detectible by traditional methods), necessitating additional cultural analysis before their proposed construction could begin. In the Four Corners region of the American Southwest, pit structures are commonly buried by floodplain silt and windblown sand. One Basketmaker period site was discovered using GPR in the floodplain of the San Juan River near Bluff, Utah (Conyers and Cameron 1998). The area was seen as prospective because there were numerous ceramic sherds and lithic flakes on the surface that had been brought up from depth by burrowing animals. Much of this material had been washed around over the years when the river flooded or during torrential summer thundershowers, concentrating it in a few places on the ground. Areas of concentration were assumed by the field archaeologists to be the loci of pit structures that were buried below, and a GPR grid was set up to cover the prospective area where the artifact concentrations were located (Figure 7.11). As the GPR data were being collected the profiles visible on the computer screen produced images of pithouse floors (Figure 7.12), none of which were in the areas where the surface artifacts were located. Amplitude slice-maps showed the location of one pithouse floor and a portion of a smaller one to the west, at about 1 m below the surface (Figure 7.11). A rodent burrow, not visible on the ground surface, was also visible on the slice-maps and may have been the conduit for the transfer of many of the artifacts from the pithouse floor to the surface sometime in the past. This GPR survey illustrates how even when there are good surface indications of buried materials, the location of artifacts may not necessarily be correlative to their buried source. Near the pithouse site, an above-ground pueblo was in the process of being excavated by the University of Colorado on a terrace overlooking the San Juan River floodplain. About 50 m from the Chaco period Great House was a shallow depression that was postulated to be the “great kiva” associated with that structure. In the American Southwest, kivas are especially important structures for study, as they were the religious and social focus of much prehistoric activity. But because they often contain ceremonial remains and because of their religious significance, Native Americans are very reluctant to see them excavated extensively without good reason. Understanding the sensitive nature of this possible feature, a GPR survey was conducted over the depression prior to excavation to initially determine whether it was in fact a kiva and the results were then used as a guide to excavation. The GPR profiles showed a bowl-like depression surrounded by standing walls, filled with what appeared to be windblown sand. It appeared, at least in cross section, to be much like other kivas found in the area (Figure 7.13). The GPR slice-maps generated from the reflection data unambiguously illustrated the circular nature of the outside walls of the kiva in the shallowest slices (Figure 7.14). What was intriguing was that there appeared to be walls within the kiva in the deeper slices. The amplitude slices were indicating the presence of features that were totally unexpected, as double-walled structures of this sort were unrecorded in the vicinity. An excavation trench was placed on the east edge of the kiva, in an area that would encounter the outer wall at about 30 cm below the surface and the deeper feature at

150 ~ Lawrence B. Conyers

Figure 7.11. Amplitude slice-maps from a pithouse site in Utah. The surface scatters of artifacts led archaeologists to this site. GPR mapping revealed a pithouse floor in a different area of the grid than hypothesized from the concentration of the artifacts (color illustration appears on the CD).

about 90 cm. Both were revealed at exactly these depths. The deeper wall reflections were found to have been generated from an interior wall of the kiva, with the shallower outside wall enclosing a series of antechambers. Exactly how this unusual architectural style of kiva fits in with what is known about the Chaco and post–Chaco period structures in the area remains in question. What is not in question is the usefulness of GPR to quickly map these architectural features, determine their probable origin, delineate

Ground-Penetrating Radar ~ 151 features within the kiva, and guide excavations precisely so that important elements could be studied with limited intrusion into the ceremonial structure. The utility of employing multiple geophysical methods at the same archaeological site is just being realized and will play an increasingly important role in archaeo- Figure 7.12. Reflections from one 500-MHz profile that crossed the logical geophysics. At the pithouse floor visible in the amplitude slice-maps in Figure 7.11. San Marcos Pueblo in New Mexico, both GPR and magnetic gradiometry data were collected, which provided different but complementary images of the subsurface. The grids were collected at this huge site just to the south of a Spanish colonial church, which had been burned during the Pueblo revolt of the late seventeenth century. The magnetic gradiometry map indicated a series of linear features in an area that previous work has postulated was the location of the convent courtyard, as it was adjacent to the church (Figure 7.15). Of equal interest was the discovery of two strong dipole features within these walls that suggested strong burning episodes or perhaps large pieces of metal. A grid of GPR data was collected directly on top of this feature, and the amplitude slice-maps showed a roughly circular feature in the same location as the large magnetic anomalies (Figure 7.15). Many different working hypotheses were proposed about the origin of the circular feature within the walls of the convent courtyard that was visible by GPR and that the magnetic maps showed to have been burned. One of the more exciting, but also somewhat outlandish, hypotheses was that the circular anomaly was a kiva that had been constructed within the convent courtyard during mission times as a way to attract Native Americans to the new Roman Catholic faith. To test this idea, as well as other hypotheses, a large excavation trench was placed directly on the edge of the circular feature. At about 50 cm below the surface a partially burned layer of what appeared to be cow manure was discovered, lying on a sloping surface of silty clay. Further digging exposed more clay layers beneath, interbedded with fine sand, which were interpreted to have been periodic depositional episodes of adobe wall melt alternating with aeolian sand deposition. It now appears that the GPR and magnetic features discovered within the walls of the courtyard are the product of events that occurred long after the burning and abandonment of the church and its associated buildings. Once abandoned, the adobe walls surrounding the convent court-

152 ~ Lawrence B. Conyers

Figure 7.13. A 500-MHz reflection profile crossing a great kiva. The outside wall of the kiva is apparent as high-amplitude reflections on either side of the profile. The kiva depression was filled with homogeneous aeolian sand, producing very low-amplitude reflections.

Figure 7.14. Amplitude slice-maps of the great kiva at Bluff, Utah. The outside wall is visible in the upper slices and a deeper interior wall below about 90 cm in depth. These data were collected after a test excavation had been placed on the eastern side of the kiva and backfill material is visible as a high-amplitude reflection in most of the slices (color illustration appears on the CD).

Ground-Penetrating Radar ~ 153

Figure 7.15. The convent courtyard at San Marcos Pueblo, New Mexico. On the magnetic gradiometry map the remnant adobe walls can be seen, as well as some extensive burned areas within the walls. The GPR amplitude slice from 9.5–19 nanoseconds indicates a circular bowl-shaped anomaly in the area of burning (color illustration appears on the CD).

yard began to erode, or “melt,” as rainstorms washed the clay into the surrounding open spaces. Over many years this clay deposition filled in the corners of the courtyard and the areas immediately adjacent to the walls first, gradually leaving a bowl-shaped depression within the walls. Periodic windstorms also deposited fine-grained sand in this newly formed basin-like depression. It then appears that during historic times ranchers living nearby used the depression as a natural corral for their cattle, which left

154 ~ Lawrence B. Conyers a thick layer of manure on the clay and sand surface. This manure was then set on fire, naturally or intentionally, and partially burned, creating the strong magnetic feature visible on the magnetic gradiometry map. The San Marcos geophysical mapping demonstrates how multiple tools can indicate very different aspects of a site’s nature, allowing for multiple working hypotheses to be developed and then tested. In this case excavations supported none of the working hypotheses, and the origin of the geophysical features visible with both tools was very different from any that had been considered. Although the exciting archaeological materials that the excavators were hoping for were not present, the use of multiple geophysical methods proved to be a very useful tool for feature definition and analysis. Determining the location of historic graves is often a problem when cemeteries are old and headstones have been moved or have disappeared. Graves are often challenging to locate with geophysics because they are fairly small targets and often old caskets have collapsed and bodies decomposed, leaving little in the way of a geological anomaly to measure geophysically. Sometimes all that remains of some graves are truncated soil or sediment layers where the shafts were dug, which can be discovered with GPR by manually analyzing each individual reflection profile within a grid and hand mapping the areas where vertical truncation is located. However, if there are still void spaces present in the subsurface, even if caskets have partially collapsed, GPR can readily image the grave locations. Caskets or void spaces in partially collapsed caskets are visible in profiles as distinct hyperbolic reflections (Figure 7.16).

Figure 7.16. Historic graves with intact or partially collapsed caskets are visible when viewed in profile as distinct hyperbolic reflections derived from their upper surfaces.

Ground-Penetrating Radar ~ 155 When many profiles in a grid are processed into amplitude slice-maps, the locations of graves at numerous depths are readily apparent. A GPR survey conducted at a Mormon pioneer cemetery in Utah showed that the extant headstones were for the most part in very different locations than the actual graves mapped with GPR (Figure 7.17), which is often the case in older cemeteries. Many other cemeteries have been studied with GPR, each of which exhibits different characteristics depending on grave ages, types of burial materials, associated stratig- Figure 7.17. Amplitude slice-map reflections in a pioneer cemetery in Utah show many distinct graves, whose locations are raphy, and soil types. rarely coincident with the locations of the extant headstones, Often a series of two- most of which have been moved or have disappeared (color ildimensional slice-maps is not lustration appears on the CD). sufficient to produce an adequate image of the materials below the surface. The slices in Figure 7.6 illustrate rubble from the walls of a historic building that was dumped in its foundation before the site was leveled and paved. In general these slices show high-amplitude reflections from about 100 to 160 cm in the ground. When a series of very thin slices 2 cm thick is produced for this same grid of GPR data and only the highest amplitude reflections are placed into a three-dimensional matrix of digital values, a “cube” of values will represent the spatial placement of values in the ground. Those high-amplitude values can then be rendered in a three-dimensional imaging program to produce a much more realistic view of the rubble pile (Figure 7.18).

Conclusion GPR is an extremely useful tool for CRM archaeology because fairly large areas of ground can be surveyed quickly, producing a three-dimensional reflection data set of buried features and related stratigraphy. If reflection profiles are separated by 50 cm or a meter, areas of more than 50 × 50 m can be collected in a day and processed that evening. Each reflection profile in a grid of this sort has the ability to produce images of buried materials in profile that can be interpreted singularly or all in a batch

156 ~ Lawrence B. Conyers for three-dimensional analysis. The production of complex maps in the form of amplitude slices or rendered three-dimensional visualizations can occur quickly, making GPR one of the most accurate methods for determining Figure 7.18. A three-dimensional rendering of the highest amplitudes in depth and dimenthe same grid of data used to make the slices in Figure 7.6. This image is sion of archaeologic rubble from a historic house that was pushed into its basement before the features while opersite was leveled and paved over. ators are still in the field. When this is accomplished, GPR results from one day can be used the next to plan excavations or further GPR data collection. If little is known about what lies under the surface in a prospective area, the spatial geometry and orientation of reflection anomalies in GPR amplitude slice-maps can be studied to see whether features of interest are discovered. If the velocity of radar waves in the grid is known, not only can the orientation of the features be determined but also their depth. Areas of greater interest can then be resurveyed in smaller grids with more closely spaced profiles for greater feature definition. Lower or higher frequency antennas can also be used for either greater depth penetration or higher resolution, depending on what is discovered in the first survey. The GPR maps can also be compared and integrated with maps from other geophysical tools as a way to measure different properties of the subsurface. For instance, magnetic maps can determine burned areas or areas of more highly magnetic materials associated with certain archaeological features. GPR maps show only compositional or water saturation differences and therefore are imaging very different properties than are measured with other methods. If multiple GPR surveys are combined with subsurface information gathered by excavation, auger, or core testing, the origin of important reflections can be determined, giving the geophysical maps even more utility. In areas where subsurface testing has already uncovered features of interest, GPR data can be used to project that information into areas of a site where nothing is known. In this way a great deal of knowledge about a small excavation area can be used to determine a great deal about potentially large areas in the immediate vicinity. Most important, that information can be obtained without expensive and potentially destructive digging.

Ground-Penetrating Radar ~ 157 The GPR tools discussed here can be employed for benefits beyond just the discovery and mapping of unknown areas, although that has been GPR’s greatest historic contribution to CRM archaeology. If enough is learned about the subsurface using both excavations and geophysical mapping, working hypotheses about the spatial distribution of features that have cultural significance can be formulated and tested. In this way GPR maps can be used as a primary database for the analysis of anthropological ideas about past lifeways.

References Cited Bevan, B., and J. Kenyon 1975 Ground-Penetrating Radar for Historical Archaeology. MASCA Newsletter 11(2):2–7. Buderi, R. 1996 The Invention That Changed the World. Simon & Schuster, New York. Collins, M. E., and J. L. Kurtz 1998 Assessing GPR Performance in Three Soil Geographic Regions. In Proceedings of the Seventh International Conference on Ground-Penetrating Radar, Radar Systems and Remote Sensing Laboratory, pp. 171–175. Radar Systems and Remote Sensing Laboratory, University of Kansas, Lawrence. Conyers, L. B. 1995 The Use of Ground-Penetrating Radar to Map the Buried Structures and Landscape of the Ceren Site, El Salvador. Geoarchaeology 10(4):275–299. 1999 Geophysics, Ground-Penetrating Radar and Archaeology. Society for American Archaeology Bulletin 17(4):26–29. 2004 Ground-Penetrating Radar for Archaeology. AltaMira, Walnut Creek, California. Conyers, L. B., and C. Cameron 1998 Ground-Penetrating Radar Techniques and Three-Dimensional Computer Mapping in the American Southwest. Journal of Field Archaeology 25(4):417–430. Conyers, L. B., E. G. Ernenwein, and L.-A. Bedal 2002a Ground-Penetrating Radar Discovery at Petra, Jordan. Antiquity 76:339–340. 2002b Ground Penetrating Radar (GPR) Mapping as a Method for Planning Excavation Strategies, Petra, Jordan. E-tiquity 1 ( html). Conyers, L. B., and D. Goodman 1997 Ground-Penetrating Radar: An Introduction for Archaeologists. AltaMira, Walnut Creek, California.

158 ~ Lawrence B. Conyers Conyers, L. B., and J. E. Lucius 1996 Velocity Analysis in Archaeological Ground-Penetrating Radar Studies. Archaeological Prospection 3(1):25–38. Conyers, L. B., and H. D. Wallace 2003 Ground-Penetrating Radar. In Archaeological Excavations at Valencia Vieja, edited by H. D. Wallace and M. W. Lindeman. Technical Report No. 2001-11. Desert Archaeology, Inc., Tucson. Fischer, P. M., S. G. W. Follin, and P. Ulriksen 1980 Subsurface Interface Radar Survey at Hala Sultan Tekke, Cyprus. In Applications of Technical Devices in Archaeology, edited by Peter M. Fischer. Studies in Mediterranean Archaeology 63:48–51. Goodman, D. 1994 Ground-Penetrating Radar Simulation in Engineering and Archaeology. Geophysics 59:224–232. Goodman, D., and Y. Nishimura 1993 A Ground-Radar View of Japanese Burial Mounds. Antiquity 67:349–354. Goodman, D., Y. Nishimura, H. Hongo, and O. Maasaki 1998 GPR Amplitude Rendering in Archaeology. In Proceedings of the Seventh International Conference on Ground-Penetrating Radar, pp. 91–92. Radar Systems and Remote Sensing Laboratory, University of Kansas, Lawrence. Goodman, D., Y. Nishimura, and J. D. Rogers 1995 GPR Time-Slices in Archaeological Prospection. Archaeological Prospection 2:85–89. Goodman, D., Y. Nishimura, R. Uno, and T. Yamamoto 1994 A Ground Radar Survey of Medieval Kiln Sites in Suzu City, Western Japan. Archaeometry 36(2):317–326. Imai, T., T. Sakayama, and T. Kanemori 1987 Use of Ground-Probing Radar and Resistivity Surveys for Archaeological Investigations. Geophysics 52:137–150. Kenyon, J. L. 1977 Ground-Penetrating Radar and Its Application to a Historical Archaeological Site. Historical Archaeology 11:48–55.

Ground-Penetrating Radar ~ 159 Leckebusch, J., and R. Peikert 2001 Investigating the True Resolution and Three-Dimensional Capabilities of GroundPenetrating Radar Data in Archaeological Surveys: Measurements in a Sand Box. Archaeological Prospection 8:29–40. Moran, M., S. A. Arcone, A. J. Delaney, and R. Greenfield 1998 3-D Migration/Array Processing Using GPR Data. In Proceedings of the Seventh International Conference on Ground-Penetrating Radar, pp. 225–228. Radar Systems and Remote Sensing Laboratory, University of Kansas, Lawrence. Neubauer, W., A. Eder-Hinterleitner, S. Seren, and P. Melichar 2002 Georadar in the Roman Civil Town Carnuntum, Austria: An Approach for Archaeological Interpretation of GPR Data. Archaeological Prospection 9:135–156. Olhoeft, G. R. 1981 Electrical Properties of Rocks. In Physical Properties of Rocks and Minerals, edited by Y. S. Touloukian, W. R. Judd, and R. F. Roy, pp. 257–330. McGrawHill, New York. Sellman, P. V., S. A. Arcone, and A. J. Delaney 1983 Radar Profiling of Buried Reflectors and the Ground Water Table. Cold Regions Research and Engineering Laboratory Report 83-11:1–10. Sheets, P. D., W. M. Loker, H. A. W. Spetzler, and R. W. Ware 1985 Geophysical Exploration for Ancient Maya Housing at Ceren, El Salvador. National Geographic Research Reports 20:645–656. Stern, W. 1929 Versuch einer elektrodynamischen Dickenmessung von Gletschereis. Gerlands Beitrage zur Geophysik 23:292–333. Vaughan, C. J. 1986 Ground-Penetrating Radar Surveys Used in Archaeological Investigations. Geophysics 51(3):595–604. Vickers, R. S., and L. T. Dolphin 1975 A Communication on an Archaeological Radar Experiment at Chaco Canyon, New Mexico. MASCA Newsletter 11(1):6–8.


Magnetic Susceptibility Rinita A. Dalan

Magnetic susceptibility surveys occupy a unique niche in archaeological research distinct from other near-surface geophysical methods. Historically, susceptibility surveys have not been as widely employed as magnetometry, resistivity, or ground-penetrating radar surveys; however, recent field and laboratory-based applications together with advances in instrumentation have resulted in an increase in interest in the application of susceptibility techniques. A susceptibility study can provide information that is unavailable through more traditional geophysical methods. For example, when applied in conjunction with laboratory soil magnetic techniques, not only may archaeological features be located and defined but also questions regarding formation and postdepositional processes may be approached. Susceptibility studies are applicable to a broad range of archaeological sites, features, and environments and, through the use of various sensors, are capable of resolving contrasts in susceptibility over large and small scales. Though they can be time consuming compared with other geophysical methods, the complementary information they provide and the ease with which these techniques can be mastered render them worth consideration for archaeological research.

Overview of the Technique Basic Principles Magnetic susceptibility provides a measure of a material’s ability to be magnetized. As its full name, low field magnetic susceptibility, suggests, this property quantifies the response of a material to a weak magnetic field (i.e., one on the order of the earth’s

162 ~ Rinita A. Dalan field). It is defined as the ratio of the magnetization induced in a sample to the inducing (magnetizing) field. Magnetic susceptibility is measured in the presence of the magnetizing field and thus is distinct from a magnetic remanence, or “permanent” magnetization, that is measurable in the absence of a magnetic field. Magnetometer surveys record spatial variations or anomalies in the earth’s magnetic field caused by local changes in magnetization. These magnetic anomalies may arise from localized deposits of materials with a higher or lower susceptibility or from materials possessing a magnetic remanence. Magnetometer surveys make no distinction between these; the observed anomaly expresses only the net effect of any induced and remanent magnetizations. Magnetic susceptibility surveys differ from magnetometer surveys in that they measure only the induced component of this signal. As magnetic susceptibility surveys can be used to distinguish features resulting from susceptibility contrasts from those carrying a magnetic remanence, they can serve as a useful complement to magnetometer surveys. Magnetic susceptibility can be expressed either as a susceptibility per unit volume (κ) or as a mass normalized susceptibility (χ ). Volume susceptibility (κ) is a ratio of the volume magnetization induced in a material of susceptibility κ by an applied weak magnetic field. In the SI system of units (i.e., International System of Units, abbreviated from the French Le Système International d’Unités), this is a dimensionless quantity. Mass susceptibility is equal to the volume susceptibility divided by density and has units of cubic meters per kilogram in the SI system. As part of soil development, surface soil layers typically become magnetically “enhanced” in comparison to subsoil layers. This results in greater values of magnetic susceptibility for topsoil as opposed to subsoil layers. The general process that produces these susceptibility contrasts is the conversion of weakly magnetic oxides and hydroxides to more strongly magnetic forms within the surface soil layers. Figure 8.1 presents susceptibility profiles measured from two cores at the Cahokia Mounds State Historic Site, Illinois. In both profiles, enhancement of topsoil over subsoil layers is apparent, with relatively high mass magnetic susceptibility values (χ , second column) at the surface, which decrease with depth and then stabilize in the underlying subsoil. Surface soils for both cores also show increases in another magnetic property, anhysteretic remanent magnetization (ARM), as indicated in the first column of Figure 8.1. ARM is an artificial remanence given to the sample in the laboratory. Like low field susceptibility, it increases with an increasing concentration of magnetic grains. Mechanisms responsible for this process of magnetic enhancement include naturally occurring or human-generated fires, as well as pedogenic enhancement through various inorganic and organic pathways. When explanations for magnetic enhancement were first proposed (LeBorgne 1955, 1960a, 1960b), firing was suggested as the primary mechanism. Since that time, however, pedogenic enhancement has been documented as a widespread phenomenon (Dearing et al. 1996; Maher and Taylor 1988) that is equally important in producing the enhanced magnetic properties of topsoil. Pedogenic enhancement occurs as part of soil development through low-temperature

Magnetic Susceptibility ~ 163

Figure 8.1. Magnetic enhancement of soils at the Cahokia Mounds site. The bottom panel represents “natural” pedogenic enhancement. The top panel represents soils from a more intensely occupied portion of the site. Here magnetic enhancement for surface horizons is approximately four times subsoil values (as opposed to the doubling of subsoil values observed in the bottom panel). The peak in susceptibility at approximately 70 cm below the surface on the top panel is a buried soil. (The CGS system was used for this research.)

chemical reactions (i.e., inorganically) as well as organically via magnetotactic bacteria, iron-reducing bacteria, and bacterial-induced chemical reactions (Evans and Heller 2003). Typically, it is a very fine-grained magnetite or maghemite that is produced within these surface soil layers. These would be magnetic grains in the superparamagnetic to stable single-domain size range (i.e., smaller than 0.1 μm) (Hunt et al. 1995). In Figure 8.1, the third column presents data on the frequency dependence of magnetic susceptibility (χ fd) for the two cores from the Cahokia Mounds site. As with χ and ARM, χ fd values are in general greatest for the surface soil layers of these two cores. Frequency dependence of susceptibility is the percent difference in susceptibility measured at two frequencies (the two frequencies measured by the Bartington MS2B dual-frequency

164 ~ Rinita A. Dalan susceptibility sensor are 470 Hz and 4,700 Hz). Variation in susceptibility with frequency is due to a delay between the application of the magnetic field and the magnetization of the sample. Measurement of χ fd is used to investigate the contribution of ultrafine magnetic grains, as these show the most pronounced frequency dependence of susceptibility. Because of the delayed response of these ultrafine grains to the magnetizing field, a decrease in susceptibility occurs at higher frequencies. An increase in frequency dependence suggests an increase in the percentage of ultrafine magnetic grains. An increase in frequency dependence in conjunction with an increase in susceptibility is potentially indicative of a developed soil. The magnetic enhancement process is both conservative and environmentally sensitive (Maher 1986). Because the process of magnetic enhancement is conservative, the enhanced magnetic signal of these soils will persist unless they are gleyed or the iron minerals otherwise reduced. For example, topsoil will retain an enhanced susceptibility signal even if it is deeply buried. This, of course, has critical implications in the search for buried archaeological sites and layers. As the end products of the magnetic enhancement process are environmentally sensitive, they can be used to identify and to characterize different soil types and layers and even, to some degree, to understand the processes responsible for these contrasts (Evans and Heller 2003; Maher 1986; Maher and Thompson 1999; Rummery et al. 1979; Thompson and Oldfield 1986). This latter type of study involves a more thorough magnetic characterization in addition to the measurement of magnetic susceptibility and other low field properties. Soil forms in response to a given climate and suite of living organisms as these act on parent materials in a particular landscape or topographic setting over a period of time. Soil magnetic studies can be used to understand variation in these soil-forming factors. Magnetic characteristics of soils will be dependent on the type of material in which the soils form, the time over which enhancement has been allowed to proceed, climatic variables such as precipitation and temperature, the landform in which the soil develops, and the plant and animal life on and in the soil. Humans are one of the living organisms that may impact soil development and thus human activities may be investigated via soil magnetic techniques. Archaeologists have long recognized (Aitken 1970; Mullins 1974; Tite and Mullins 1971) that soils from archaeological sites are often additionally enhanced in terms of their magnetic properties over surrounding “noncultural” or nonsite topsoils. The degree of this enhancement has been shown to depend not only on the climate and on the geologic strata from which the soil derived but also on the extent of a site’s occupation, its firing history, the quantity of organic material, and other human-controlled variables (Tite and Mullins 1970). A closer look at Figure 8.1 indicates that though both cores from the Cahokia Mounds site show magnetic enhancement of surface layers, the level of enhancement is not the same for the two cores. The bottom core was collected in a lightly impacted area of the site with no evidence of prehistoric occupation. The top core was taken from a more intensely occupied portion of the site. Here, magnetic enhancement has

Magnetic Susceptibility ~ 165 proceeded to a much greater degree. Although these cores by no means encapsulate the variety of soils at the Cahokia Mounds site, they do typify the contrast between more strongly enhanced cultural soils and naturally developed soils that are not as strongly enhanced. Higher total phosphorus levels track the enhanced magnetism of the core taken in the occupation area. The peak in susceptibility (and in ARM and χ fd) at approximately 70 cm on the top panel represents a buried soil. The level of magnetic enhancement depends on climate and on iron present in parent materials but also on a number of other variables that may be directly impacted by humans. These include factors such as soil temperature, soil chemistry, and soil porosity. On an archaeological site, humans may increase organic matter, alter porosity, and increase temperatures through firing. These changes will impact the pedogenic enhancement process and hence susceptibility contrasts. People may also incorporate fired or high-susceptibility materials into the site matrix. As a result, sites and features often exhibit susceptibility contrasts with surrounding soils. In addition, humans continually redistribute these soil layers, digging ditches and pits into low-susceptibility subsoil, for example, which may then be filled with magnetically enhanced topsoil. Magnetic susceptibility surveys therefore provide an easily measured signal that is thus keyed directly to cultural as well as natural processes. Figure 8.2 provides an illustration of the importance of this information for archaeological studies, demonstrating how magnetic susceptibility and other soil magnetic parameters can be used to distinguish natural from culturally loaded soils, identify buried soils, and understand archaeological landforms. The soil magnetic measurements were gained from a test excavation unit within the West Borrow Pit Mound Group at the Cahokia Mounds site. This unit was placed within a postulated earthen platform joining the two main mounds of the group

Figure 8.2. Unit profile from a basal platform joining two mounds at the Cahokia Mounds site. Postconstruction soil development is indicated by enhanced magnetic values in Strata I and II. Below pedogenically altered soils, variation in fill layers is apparent. Enhanced values and relatively fine magnetic grains (as indicated by an increase in ARM/ χ ) at the base of the unit (Stratum VIII) represent an intact buried soil.

166 ~ Rinita A. Dalan to assess whether this basal platform was a created feature or simply a natural accumulation of sediments at the base of the mounds (Holley et al. 1996). Postformation soil development is indicated by enhanced susceptibility values (χ ) in Strata I and II. ARM values are similarly high. Once below pedogenically altered topsoil, excavators were able to recognize layering of fill deposits and to observe evidence of prehistoric basket loading. What was interpreted as a stable A horizon was documented at a depth of approximately 120 cm below the surface. An increase in χ and ARM was noted for this level (Stratum VIII) and for Stratum VI. An evaluation of changing ratios of these two magnetic properties (i.e., ARM/χ ) allows an understanding of the genesis of these high magnetic values. The ratio of ARM to χ provides information on relative changes in the grain size of the magnetic carrier. In general, an increase in the ARM/χ ratio indicates relatively finer magnetic grains and a decrease indicates relatively coarser magnetic grains (Banerjee et al. 1981; King et al. 1982). A marked increase in ARM/χ for Stratum VIII is characteristic of the fine magnetic grains expected for a developed soil. In contrast, a decrease in the ARM/χ ratio and hence an increase in average magnetic grain size is indicated for Stratum VI. Charcoal and other products of burning noted during excavation are probably responsible for the increased ARM and χ values exhibited by this stratum. Soil magnetic studies in tandem with archaeological and pedological investigations provided documentation of a cultural origin for the basal platform. They also supplied information on the construction of this feature. Construction of the platform was initiated by the deposition of Stratum VII on an intact surface. Stratum VII deposits were capped by deposits formed by in situ burning and midden (Stratum VI). The second stage of construction involved the addition of a thicker layer of fill (Stratum V) and a subsequent cap (Stratum IV). Construction was renewed again in Stratum III. The uppermost strata (Strata I and II) represent the combined effects of pedogenesis, prehistoric construction, colluvium from nearby mounds, and trash from historic park activities. History of Application Because archaeologists recognized that the success of magnetometer surveys depended on susceptibility contrasts, susceptibility studies followed quickly upon pioneering magnetometer applications. These studies were directed toward understanding factors controlling enhancement at a given site (Tite 1972; Tite and Linington 1975; Tite and Mullins 1970, 1971). The ultimate focus of these investigations was to explore the link between magnetic properties and magnetometer surveys, that is, to understand the nature of magnetic anomalies and to assess the potential for magnetometer surveys at a particular site. As these studies were not directed toward developing the magnetic susceptibility method as a geophysical prospecting technique in its own right, susceptibility surveys have not advanced as much as other methods described in this volume. Soil magnetic studies have remained limited in archaeology up to the present (Dalan and Banerjee

Magnetic Susceptibility ~ 167 1998), despite their successful application to a broad range of geological and environmental problems (Maher and Thompson 1999; Thompson and Oldfield 1986). In fact, even basic research on soil magnetic properties in archaeology stalled after an initial period of research. Susceptibility surveys in the United States have been largely restricted to lateral surveys using the Geonics EM38 instrument in the in-phase mode (see Clay, this volume) and these in-phase surveys are far less common than quadrature-phase applications that measure changes in conductivity. Magnetic susceptibility studies are more common in the United Kingdom. There, the Bartington MS2 suite of sensors is used in applications ranging from topsoil surveys, to surveys of stratigraphic sections, to laboratory measurements of collected cores and samples. Despite an increased frequency of application, however, susceptibility studies in the United Kingdom tend to be applied as a complementary method aimed at gathering data to aid in the interpretation of other geophysical data sets and not as a survey technique on par with these other methods. This has also limited the development of the magnetic susceptibility method. The good news is that experimental and applied research on soil magnetic properties is well developed within the fields of rock magnetism and environmental magnetism. This research has led to a much-refined understanding of the processes and products of pedogenic enhancement and provides an invaluable resource that can be used in developing soil magnetic approaches for archaeological applications. Instrumentation Perhaps another reason for the relatively infrequent application of susceptibility studies in archaeology relates to limitations imposed by currently available susceptibility instruments. Instruments that measure magnetic susceptibility can be divided into dual- and single-coil types. Dual-coil instruments are also known as slingram instruments. Slingram instruments, such as the Geonics EM38 (Figure 8.3), are composed of separate transmitter and receiver coils. When operated in the in-phase mode, the EM38 effectively investigates the susceptibility of the top 0.5 m of soil. Other slingram-type instruments that have been employed in archaeological investigations in Europe are the SH3, the CS60, and the CS150 (Benech and Marmet 1999; Parchas and Tabbagh 1978; Tabbagh 1986). Slingram instruments are often preferred because they are less influenced by soil conductivity and they provide deeper penetration depths than single-coil instruments (Benech and Marmet 1999). Effective depths of investigation for current slingram instruments used by archaeologists, however, are relatively shallow. All slingram instruments listed above have maximum investigation depths of less than 70–80 cm. Single-coil instruments, as the name implies, are composed of just one coil. Also known as coincident loop instruments, they require direct contact with the soil or sediment and their depth of investigation is related to the size (i.e., diameter) of this coil. For this reason, field surveys with single-coil instruments are generally restricted

168 ~ Rinita A. Dalan to topsoil measurements or applied to stratigraphic exposures. Bartington Instruments has long marketed two field coils, the MS2D and the MS2F, with effective penetration depths of ca. 10 cm and 1 cm, respectively. The MS2D (Figure 8.4) is generally employed in topsoil surveys. The Bartington MS2F is a smaller coil with a shouldered housing that provides an immersion limit in soft soils and sediments. A newly developed surface scanning sensor, the Bartington MS2K, was designed for use on moderately smooth surfaces and thus is much more appropriate than the MS2F for use on exposed sections. Response is to 2.5 cm. Other single-coil instruments include the Exploranium KT9 (Scully 1997) and the Geofyzika KT-5 Kappameter, which has an effective penetration depth of ca. 2–3 cm (Lecoanet et al. 1999). For surface susceptibility surveys, as for all surface geophysical methods, resolution decreases with increasing depth of penetration. Lowering a geophysical sensor down an open borehole, however, provides an opportunity to maintain resolution with depth. Magnetic susceptibility borehole instruments are particularly appropriate for investigations of deposits deeper than 1 m. They can be employed to document vertical contrasts in susceptibility from the surface to depths approaching 20 m or even greater. Surface surveys may more quickly document lateral changes in susceptibility; however, they do not provide the critical depth information that is necessary for understanding the effects of pedogenesis or for resolving stratigraphic layering. The only way to resolve fine-scale depth variations in susceptibility is to conduct a borehole survey or to collect samples for laboratory measurement. Susceptibility soundings using the EM38 (Bevan and Dalan 2003; Dalan and Bevan 2002) can be

Figure 8.3. An EM38 in operation.

Figure 8.4. The Bartington MS2D sensor.

Magnetic Susceptibility ~ 169 employed for a quick estimation of the magnetic stratigraphy of the soil, but only a single magnetic layer is likely to be detected. Borehole instruments include both dual- and single-coil types. Unfortunately, equipment options for archaeological applications are limited. Geonics markets the EM39S, a dual-coil magnetic susceptibility instrument for logging down a 2- to 3-inch borehole (McNeill et al. 1996). Instruments like the EM39S, however, are not optimal for archaeological applications because of limitations in vertical resolution. The vertical resolution of the EM39S is on the order of a few tenths of a meter. Logistically, these instruments may also be difficult and expensive to apply because of the required largediameter boreholes. As an alternative, I constructed a prototype susceptibility down-hole instrument using Bartington Instruments components (Dalan 2001) to be operated down a core hole made with a commonly available 1-inch (2.54-cm) push-tube corer (Figure 8.5). This prototype instrument was unique in that fine-scale vertical changes in susceptibility could be measured with minimal impact to the archaeological record. The instrument resolved layers on the order of 2–4 cm. The prototype logger was shown to be time and cost effective, allowing measurements of susceptibility to be obtained approximately 10 times faster than collecting samples for laboratory measurement from an exposed section or through coring. Improvements to the prototype model were supported by the National Science Foundation through a Major Research Instrumentation grant. Decreased drift, changes in sensor geometry resulting in increased sensitivity and resolution, increased depth of

Figure 8.5. The prototype down-hole magnetic susceptibility logger in operation.

170 ~ Rinita A. Dalan application, and improved portability and flexibility of operation were accomplished through a partnership with Bartington Instruments, who assumed responsibility for engineering and producing a commercial product. I conducted laboratory and field tests of design prototypes directed toward evaluating performance (Figure 8.6). Performance was evaluated over a number of variables including repeatability (noise), sensitivity, stability, resolution, and speed and ease of operation. The commercial product, the Bartington MS2H sensor, is planned for release in 2005. This sensor produces a highly isotropic radial field pattern with a penetration depth of 12 mm. Measurements are completed in one second. A low operating frequency (approximately 2 kHz) eliminates response to conductivity effects and temperature stabilization reduces instrument drift. Measurement resolution is 1 × 10–5 SI. Extensions to reach 3 m in depth will be standard but additional sections may be added to extend to greater depths. If down-hole instruments are not employed, susceptibility depth information may be gained through laboratory measurements on samples collected by coring or from exposed sections. Measurements can be taken over whole cores or individual samples. Bartington Instruments makes a series of loop sensors (MS2C sensors, ranging from 45 to 160 mm in diameter) that can be used to obtain volume susceptibility measurements of various diameters of whole cores. Surface scanning sensors may be used on split cores. Alternatively, samples may be packed in containers of various shapes and sizes. I use a 5.28-cc plastic (nonmagnetic) box that holds roughly 10 g of soil (the Althor P15 box). These boxes are packed with bulk soil samples recovered from excavation or exposures or from soils extracted through coring. Packing time can be reduced in soft sediments by filling containers by pushing them directly into an exposed section, or samplers can be employed to extract a volume of soil that is then extruded directly into the sample container. These methods can minimize disturbance of soil structure and density. Field Survey Investigations of areal variations in magnetic susceptibility may be accomplished with either dual- or single-coil instruments. The Geonics EM38 is the most commonly employed instrument in area surveys of magnetic susceptibility in the United States. Measurement of magnetic susceptibility is accomplished in the in-phase mode. Either transects of data or gridded measurements may be recorded. For further detail on archaeological magnetic susceptibility surveys using this instrument see Bevan (1996) and Clay (this volume). Because the drift of the EM38 is higher in the in-phase mode than in the quadrature-phase mode, increased concern with drift correction must be part of a magnetic susceptibility survey using this instrument. Typically, an initial reading is taken with the instrument held high above the ground (at a height of ca. 1.5 m), where the response of the instrument (to changes in both susceptibility and conductivity) is essentially zero. The survey then commences, but periodically throughout the survey the

Magnetic Susceptibility ~ 171 instrument is again lifted to this height to check the drift of this reference value. The reference checks can be used in post-fieldwork processing to correct for instrument drift. The time or distance interval between reference values will be fixed by the operator, keeping in mind that more frequent reference measurements will allow more accurate drift corrections but will also increase survey time. As a minimum, instrument drift should be evaluated at the end of each survey line. As described by Clay (this volume), there are a number of procedures that can be used in the field to minimize instrument drift. Another complication with the EM38 is that when it is operated in the normal, upright position (magnetic dipoles vertical and coplanar), the response of the Figure 8.6. Field evaluation of the Bartington instrument changes sign with depth (Fig- MS2H. ure 8.7). A positive response is confined largely to the top 0.5 m of the soil, with the maximum positive response at approximately 20 cm. Below 0.6 m, the response is weakly negative. This variation in the polarity of response can cause uncertainty in the interpretation of susceptibility survey data. With the dipoles horizontal (with the instrument on its side), the response is always positive, but Bevan (1996:C-21) suggests that there is greater noise interference with the instrument in this position. He recommends using the instrument in its upright position but following certain field procedures to deal with the changing polarity of response. Bevan suggests holding the instrument at a height of 0.6–0.7 m above the ground surface so that instrument response to subsurface features will be entirely negative. A simple correction for elevation and sign, as described in the data-processing section below, must follow. Although this procedure ensures that the response of the instrument maintains a constant sign, this comes at the expense of sensitivity. Area surveys may also be accomplished with the Bartington MS2F and MS2D sensors, which have been used in topsoil surveys conducted in the United Kingdom. The work of Anthony Clark, a pioneer of susceptibility studies in Great Britain, can be referred to for further detail on these and other types of magnetic susceptibility surveys (Clark 1996). A guide to using the Bartington MS2 system for environmental magnetic applications by Dearing (1994) is also a valuable source for those wishing to conduct field and laboratory measurements.

172 ~ Rinita A. Dalan

Figure 8.7. Depth response of the EM38. With the instrument on its side (dipoles horizontal), the response is always positive. With the instrument in the upright position (dipoles vertical and coplanar), the response of the instrument changes sign depending on the depth of the feature.

In general, the density of measurements achieved using the EM38 is not matched by surveys using MS2 sensors. This is due to the slower speed of the MS2 instruments. Drift is also an issue with the MS2 instruments. Commonly, these sensors are zeroed in air prior to each reading. Alternatively, the readings taken in air before and after each measurement can be used for drift correction. The MS2 meter and appropriate sensor should be turned on and allowed to equilibrate to site conditions for ca. 30–60 minutes prior to the survey. In addition to area surveys, the MS2F, MS2D, and new MS2K sensors may also be employed on exposed sections and excavation walls and floors. On vertical surfaces, depth profiles of susceptibility are usually recorded, although gridded maps of excavation walls (and floors) can also produce very useful information. Unlike the EM38, the MS2 system does not provide a dedicated data logger. The MS2 meter

Magnetic Susceptibility ~ 173 does have an RS232 interface, however, and connection to a laptop loaded with the Multisus 2 program (which can be downloaded free of charge from the Bartington website) may be used to record data that can be output to a printer or downloaded to a spreadsheet program. The EM38 may be employed at the surface to obtain a quick estimation of the variation of magnetic susceptibility with depth. This may be particularly appropriate in areas where it is difficult to core. This procedure is referred to as an electromagnetic sounding and it involves taking a series of readings as the instrument is lowered from an elevation of approximately 2 m above the ground to the ground surface. Resulting information on the variation of susceptibility with height can be used to approximate the magnetic stratigraphy of the soil. This is done by comparing the measured sounding to model calculations. For specifics regarding measurement, data processing, and analytical procedures, see Dalan and Bevan (2002) and Bevan and Dalan (2003). To obtain fine-scale resolution of layering with depth, however, down-hole instruments are necessary. Down-hole instruments appropriate for archaeological applications are just becoming available. Drift can be a particular problem for these instruments as a result of changing temperature conditions from air into and down the core hole. There are several options for dealing with this temperature-induced drift. One is to first zero the sensor in air, next take a series of measurements down the core hole, and finish with a final reading in air. The difference between the initial zero reading in air and the final air reading (instrument drift) may then be linearly distributed along the readings down the core hole, assuming a constant drift. More accurate, although also more time-consuming, options would involve more frequent drift checks in the course of logging the core hole. Instrument drift could be evaluated by bringing the sensor up and taking a reading in air a number of times in the course of logging the core hole. For example, if a 2-cm recording interval is employed, instrument drift might be checked every 5 or 10 readings (every 10 or 20 cm). Optimum measurement precision would be accomplished by removing the sensor from the core hole and zeroing it in air prior to each measurement. Even better, an initial air reading, a reading down the core hole, and a final air reading could be taken, allowing the estimated air value at the time of the measurement in the core hole to be subtracted from the measured value. Any of these measurement sequences could be effected using the Bartington MS2H sensor and MS2 meter connected to a laptop with the Multisus program. The Multisus program provides options for automatic measurements at 1.5-second intervals or for any of the measurement sequences mentioned above. Starting depths and depth increments may be chosen and modified as necessary, drift corrections can be made (or this can be done separately during processing), and a screen graph of the raw or drift-corrected data then provides immediate feedback in the field on susceptibility changes with depth. Resolution of fine-scale variation in susceptibility with depth and across space is also possible through laboratory measurement of collected samples. Field instruments may be preferred as they can be used to rapidly characterize susceptibility over a site. With the exception of down-hole instruments, however, they are limited in

174 ~ Rinita A. Dalan the depths that can be investigated. Laboratory studies also provide the option to normalize measurements by the mass of the sample, allowing correction for variable packing density and direct comparison with other samples and studies. Additionally, inclusions in the soil, such as pieces of metal, pottery, or charcoal, may be removed so that the susceptibility of the soil matrix and contained artifacts may be separately determined. This is not an option for field studies. It is only through sample preparation that factors such as sample density, moisture, and inclusions and their effects can be controlled. Once samples have been packed for magnetic susceptibility studies, they can then be used in a number of other soil magnetic tests directed toward understanding not only magnetic contrasts but also the processes that have generated these contrasts. Susceptibility surveys enjoy a highly variable scale of application ranging from EM38 surveys to coarse topsoil surveys to surveys aimed at resolving fine-scale lateral or vertical variations in magnetic susceptibility. The Bartington MS2F and MS2K field coils and MS2B laboratory sensor allow relatively small volumes of material to be investigated. These sensors can be applied in small-scale investigations (e.g., to resolve stratigraphic layering within a pit feature) or used to map large-scale features or landscape elements. Figure 8.8 provides an example of the latter. Susceptibility and other soil magnetic studies were conducted on cores taken from the eastern section of the Grand Plaza at the Cahokia Mounds site. These studies, in conjunction with soil physical descriptions, allowed the identification of preoccupation topography (including natural sand ridges and swales) and provided evidence of earthmoving activities involved in mound construction and creation of the Grand Plaza.

Figure 8.8. Profile across the Grand Plaza at the Cahokia Mounds site showing natural sediments and overlying cultural material. Data gained from soil magnetic studies, soil chemical tests, and core descriptions. Locations of the cores are indicated in the top inset.

Magnetic Susceptibility ~ 175 Data Processing and Interpretation Data treatment for magnetic susceptibility surveys conducted with the EM38 will involve drift correction and conversion of field measurements to magnetic susceptibility values. The drift of the instrument between reference value positions (see the field survey section, above) should be calculated and a linear progression used to apply a drift correction to measurement stations located between reference value positions. To convert the in-phase measurements to values of magnetic susceptibility (McNeill 1986:15), the following formula is used: volume magnetic susceptibility (κ) = 58 × 10–6 (Δσa), where Δσa is the difference between the drift-corrected reference reading and the instrument reading on the ground in units of millisiemens per meter. In the SI system, the units of κ are dimensionless. If the procedure suggested by Bevan (1996:C-21) has been followed (i.e., lifting the instrument above the ground so that the polarity of instrument response will not switch for underground features), then an additional correction must be made. This can be done at the same time that the data are converted to volume susceptibilities. For example, if the instrument has been carried at a height of 0.6–0.7 m, volume susceptibility in SI units should be calculated as follows: κ = (58 × 10–6 (Δσa)) × (–5). As outlined by Bevan (1996:C-21), the positive part of the response curve for the in-phase measurements with the dipoles vertical has an area of 1.202. The negative part of the response curve has an area of 0.202. These differ by a factor of 5 and thus multiplying each value by a negative 5 reverses the sign so that the response is positive and corrects for the elevation of the instrument. Once these procedures have been accomplished, other types of data treatment may follow. See Clay (this volume) for data-processing tips specific to EM38 surveys and the chapters by Kvamme in this volume for excellent discussions concerning computer processing of digital data sets, including methods for dealing with instrument noise, operator noise, and noise from features not of interest to the survey, as well as guidance on enhancing and presenting geophysical data. This information will not be repeated here. Data processing for surveys employing the Bartington sensors is minimal. Drift corrections must be made. Values must also be converted to volume or mass susceptibilities (the latter is only possible for collected samples). The MS2B laboratory sensor is absolutely calibrated using a standard of a stable iron oxide. Provided a 10 cm3 sample is used, measured values are simply multiplied by a factor of 1 × 10–5 to arrive at volume susceptibilities in SI units. If other sample volumes are employed, simply multiply the meter readings by 1 × 10–4 and divide by the sample volume in cubic centimeters. To get mass susceptibility, divide volume susceptibility by the bulk density of the sample (mass/volume). Intercalibration between sensors is possible using the response of the various sensors when measuring the volume susceptibility of water (Dearing 1994). For example, the MS2F would yield a value approximately equal to 0.5 of the volume susceptibility measured by the MS2B for tip contact and approximately the same quantity if the sensor is immersed up to its shoulder. The MS2D would yield a

176 ~ Rinita A. Dalan quantity of approximately 0.75 of the volume susceptibility measured by the MS2B (0.5 × κ on rough surfaces). Thus, to estimate volume susceptibility using the MS2F (with tip contact) one would multiply the measured value by 2 × 10–5 SI. If samples have been collected in addition to field measurements, a direct calibration can be accomplished using volume susceptibilities measured on the MS2B sensor. The MS2K sensor is supplied with a calibration standard that can be used to convert field values to volume susceptibilities. Gridded measurements obtained using the MS2D, MS2F, and MS2K sensors can be treated the same way as data gained with the EM38 and other geophysical instruments. Data may be migrated into programs such as Surfer and Geoplot for manipulation and presentation. In contrast, susceptibility data obtained along transects or down vertical profiles are usually processed and graphed within a spreadsheet program (e.g., Excel). Drift correction may be accomplished in these spreadsheet programs or this may be done in the Multisus program. Data treatment is minimal, involving at the most smoothing the data or taking an average of repeated measurements. Interpretation is relatively straightforward although a basic familiarity with soil magnetic theory is helpful. Once plots have been produced, they are inspected for contrasts in susceptibility and for conformity to the model of magnetic enhancement. This is similar to the approach taken for any other soil physical or soil chemical property and thus should be a familiar process for archaeologists. One of the distinct advantages of a magnetic susceptibility survey is that a full interpretation of the data is possible. This is accomplished through additional soil magnetic studies. The magnetic susceptibility of a material will depend not only on the concentration of magnetic grains but also on their grain size and mineralogy. If samples have been collected then contrasts in magnetic susceptibility can be fully resolved in terms of these parameters. With the help of applied and theoretical studies available in the literature, it is possible to relate this information to soil-forming factors and human impacts and hence to understand past environments and cultural activities. A discussion of the application of soil magnetic studies to archaeological problems is presented in Dalan and Banerjee (1998).

Guidelines for Application Limitations to Application As pedogenic enhancement is a ubiquitous process, magnetic susceptibility studies are potentially more broadly applicable than any other geophysical method. Humans, as one of the soil-forming factors, will impact the susceptibility signal. People also tend to purposefully or inadvertently redistribute enhanced topsoil on a site. Theoretically, then, susceptibility surveys would be useful for investigating many types of archaeological sites and features. Several caveats to this statement must be raised, however. First, susceptibility studies should not be considered in situations where iron oxides have

Magnetic Susceptibility ~ 177 been reduced, such as in gleyed (waterlogged) and podsolized soils. Other limitations relate to particular instruments applied and these concern the depth of investigation, resolution, external sources of geophysical noise, instrument drift, and vegetation. Current instruments provide measurement of magnetic susceptibility for a limited range of depths. For single-coil instruments, the maximum penetration depth is perhaps 10 cm. Most single-coil instruments measure to a depth of only 1–3 cm. In the United Kingdom, single-coil instruments are employed in topsoil surveys. Interestingly, these studies have indicated that deeper features may be located as a result of bioturbation and other means of moving magnetic materials upwards through the soil column. Dual-coil instruments such as the EM38 extend the effective depth of penetration only an order of magnitude (to less than 1 m). If features are deeply buried, dual-coil instruments such as the EM31 may be used although resolution will suffer. Horizontal resolution for the EM38 is related to intercoil spacing. The minimum width of a feature that can be resolved is a fraction (ca. one-fourth to one-third) of this intercoil spacing. For the EM38, with an intercoil spacing of 1 m, horizontal resolution is approximately 0.25–0.30 m. For the Bartington MS2 sensors, resolution is related to sensor diameter and the corresponding volume measured. The MS2D measures a surface area of approximately 270 cm2. The MS2F measures a surface area of approximately 2 cm2. For the newly developed MS2H down-hole sensor, resolution of layers on the order of 1 cm is possible. Sensitivity to metallic noise sources such as fences, pipelines, and historic metal trash (if this is not of interest) may limit the use of magnetic susceptibility surveys with electromagnetic instruments. The electromagnetic instruments are also sensitive to power lines and to atmospheric electricity (spherics). As single-coil instruments may not experience such problems, Clark (1996) suggests their use where magnetic interference is serious. Drift has been mentioned as a perennial problem. If one is looking for small changes in susceptibility, drift can easily swamp the signal. In such cases, stringent field procedures and close attention to drift correction are necessary. With the single-coil instruments, rough ground or thick vegetation can affect the accuracy of the readings. A relatively smooth surface is necessary for even contact with the sensor. Plowed fields or thin vegetation are best. Thick vegetation or brushy conditions are not a problem for the EM38. Soil compaction (e.g., as in paths, roads, and so on) may produce variability in measured values, particularly for the single-coil instruments with their relatively shallow penetration depths. All field sensors measure volume susceptibility and will reflect, but not correct for, variable density. If information on the variation of susceptibility with depth is of interest, then options for instruments become even more limited. Samples may be collected, exposures may be measured, or EM38 depth soundings may even be employed for limited resolution of shallow depths. Borehole measurements provide another option. Development of archaeologically appropriate instruments has been lacking but the potential for this

178 ~ Rinita A. Dalan type of data collection is improving. These instruments will be limited in hard, gravelly, or rocky soils that are difficult to core. Types of features particularly appropriate for magnetic susceptibility surveys include magnetic features (e.g., burnt features), excavated features (e.g., ditches or pits filled with contrasting soils), and other earthen features. Susceptibility surveys have been successfully used to interpret complex earthen features and to characterize archaeological landscapes. They may be used to identify the focus of settlement, use areas, areas of industrial activity, and arable or managed soils. In general, they are not as useful for investigations of building remains made out of stone, masonry, and the like. Susceptibility surveys can identify subtle features; high-contrast boundaries required by certain other geophysical methods are not necessary. Clark (1996) has recommended magnetic susceptibility surveys for areas where magnetic interference or igneous bedrock may limit the use of magnetometer surveys. The bottom line in deciding whether or not to employ a susceptibility survey is the susceptibility contrast. The beauty of this method is that this is a very easy question to answer. It is a simple process to see whether susceptibility surveys are going to work. Take samples to reflect background values, site matrix, and feature types. Measure the susceptibility of these samples (susceptibility meters are quite common and widely available) or send them to someone else to measure. Compare contrasts and the sensitivity of your instruments and you will easily be able to evaluate the potential of this method in a given situation. Instrument Comparison It is difficult to directly compare magnetic susceptibility surveys to other geophysical methods in terms of cost. In general, magnetic susceptibility surveys are not conducted using the standard 20-×-20-m or 30-×-30-m grids that have become common with other geophysical methods. Field procedures and technique of application vary widely. Several parameters, however, allow cost to be compared indirectly. These include the cost of instruments and measurement times. The purchase price for an EM38 with a data-logging system and an extender arm for ease in surveying is slightly over $12,000. The cost for an EM38B (a “piggy-back” model that joins two instruments together and allows for the simultaneous measurement of conductivity and susceptibility) increases to over $15,000. The single-coil systems, such as the Bartington meter and sensors, are relatively low cost. A Bartington MS2 meter may be purchased for approximately $3,000. The MS2F and MS2D sensors are approximately $1,000 each. A probe handle that works interchangeably with either sensor adds an additional $1,000. The MS2B dual-frequency laboratory coil may be purchased for just under $3,000. An MS2K sensor is priced at around $2,500 as are the MS2C core-logging sensors. The down-hole sensor is priced at just over $3,000. Magnetic susceptibility surveys are certainly not as rapid as other surface methods. The EM38 provides the fastest option for data collection in area surveys. As mentioned

Magnetic Susceptibility ~ 179 by Clay (this volume), the data logger can be programmed to take a reading every 0.5 second. Walking a 1 m per second pace will yield two readings every meter and relatively efficient coverage of a large area. In-phase surveys tend to be slightly slower than conductivity (quadrature-phase) surveys because of the attention that must be paid to drift correction. For the Bartington field coils, using the 1.0 sensitivity range, measurement time is approximately 1 second. Measurement time must be doubled if the sensors are zeroed first in air before every measurement (or alternatively tripled if the air reading, measurement, air reading sequence is employed). Laboratory measurements are more time consuming as a result of the time required for collecting and packing the samples, although the actual measurement time for laboratory coils is similar (ca. 1 second). If the 0.1 sensitivity setting is used, this increases measurement time by an order of magnitude (to ca. 10 seconds). Procedures for drift correction (zeroing in air or taking bracketing air readings) will double or triple this time. Time to insert the sample within the coil must also be included. Therefore, an estimate of approximately 30 seconds per sample for laboratory measurements would be reasonable. Measurement time will be increased if several readings per sample are taken (used in averaging to increase precision). Measurement time for the down-hole sensor is on par with that of other field and laboratory coils. Using automatic measurements at 1.5-second intervals and moving the sensor at 2-cm increments, a 1.5-m hole can be logged in two minutes. If the sensor is removed from the core hole and zeroed before each reading, a 1- to 1.5-m hole can be logged in approximately 20 minutes (using a 2-cm recording interval). A 10-fold increase in time would be required to collect samples using a corer and to pack and measure the same depth section in the laboratory. An EM38 survey, using the data-logging system, is a one-person job. As the MS2 system does not have a dedicated data logger, it is probably more effective to use two people for field surveys using this equipment. A laptop computer, loaded with the Multisus program, can electronically record data; however, it is difficult for one person to carry a laptop as well as the MS2 meter and sensor. If data are hand recorded, one person may conduct a field survey with an MS2F or MS2D sensor, but it is probably quicker and easier to have one person measuring and one person recording. Clark (1996) believes that magnetic susceptibility instruments are not suitable for fine-interval area surveys because they are so labor intensive. Indeed, if speed is an issue, then other methods allow coverage of wider areas or increased density of readings. Combined with the depth limitations of current instruments, this may be a powerful incentive to first try other geophysical methods. It is very easy to learn to use susceptibility equipment, however, especially the single-coil sensors. Also, interpretation of the data is relatively simple and straightforward. Thus, if there is not enough money to buy or rent expensive geophysical equipment or to pay a professional crew to do a geophysical survey but there are an abundance of crew members willing to record measurements, then a susceptibility survey may be a smart choice.

180 ~ Rinita A. Dalan Scale of Application There is no standard for application of magnetic susceptibility surveys in spatial terms. The measurement intervals employed will necessarily depend on problem orientation; however, a spacing of 0.5–1 m would be reasonable in area investigations of magnetic susceptibility using the Geonics EM38. Long lines or large grids will often be more efficient than smaller (ca. 20 × 20 m) grids. For single-coil instruments such as the Bartington sensors, variable approaches and scales of application have also been employed. In the United Kingdom, the MS2D is most commonly used for what are termed “topsoil surveys.” These can be conveniently divided into coarse and fine sampling schemes. A coarse-sample topsoil survey may involve transects or grids. In the guidelines entitled Geophysical Survey in Archaeological Field Evaluation (David 1995), English Heritage suggests that intervals should not exceed 20 m and 10 m is preferable. As part of a coarse sampling survey, English Heritage recommends collecting representative samples of both topsoil and underlying subsoil and also samples from archaeological features. This will allow a comparison of topsoil susceptibility with the susceptibility of the subsoil and also an evaluation of the contrast with archaeological features. The number of samples that should be collected will depend on the size of the area, time constraints, and other variables; however, as a general guideline English Heritage suggests collection at each grid intersection (every 20–30 m) if a magnetometer survey is planned. A coarse-sample topsoil survey would give a general idea of susceptibility levels and contrasts. Approximate site limits and other areas of enhancement would be identified. This type of survey would not be appropriate for defining individual features, except possibly very large ones. Appropriate goals would include delimiting use areas, identifying where features might be concentrated, and mapping large natural and cultural features in the landscape. If intervals of less than 10 m are employed, then definition of smaller features is possible. A general rule of thumb in designing this type of survey would be to use the expected size of features and to employ a sampling interval such that at least two survey points would be located over this area. Challands (1992) suggests a measurement interval of 0.5–1 m and recommends that fine-sample topsoil surveys also involve collecting a representative number of soil samples. The size of the survey area and the variability of susceptibility values over this area can be used to design the collection strategy, although Challands provides a general suggestion of every fiftieth field reading. Down-hole surveys are relatively uncommon. They may also be employed at a wide variety of horizontal scales. In terms of measurement interval down the core hole, a 2-cm interval proved effective with the prototype magnetic susceptibility logger. With the increased resolution of the Bartington MS2H sensor, a 1-cm measurement interval would be reasonable, although a 2-cm interval will probably suffice for most applications. Samples should be collected from a representative number of cores, not only to

Magnetic Susceptibility ~ 181 check mass susceptibility values in the laboratory but also for use in magnetic studies directed toward providing information on magnetic mineralogy, concentration, and grain size. Integration with Other Geophysical Methods In its guidelines entitled Geophysical Survey in Archaeological Field Evaluation (David 1995), English Heritage lists two roles for the magnetic susceptibility method. The first is a supportive role in conjunction with magnetometer surveys. The second role is as a prospecting technique in its own right. In the view of English Heritage, the primary contribution of this method is its complementary role. Challands (1992) provides a contrasting view, arguing for the potential of magnetic susceptibility as a primary survey method. Magnetic susceptibility data may be used to complement other geophysical methods in various ways. English Heritage (David 1995) cites its role in prediction and preliminary assessment when it is applied prior to other geophysical surveys such as magnetometer surveys. Magnetic susceptibility surveys may be used to evaluate whether a magnetometer survey will work well in a given situation. Magnetic susceptibility surveys may be employed to target areas for subsequent geophysical work using other methods—in essence, to track broad trends in the landscape. Another role would be the use of magnetic susceptibility data in interpreting magnetometer surveys. Susceptibility data might even be used to generate models of the depths and dimensions of magnetic anomalies. English Heritage (David 1995) is quite strong in its recommendation that magnetic susceptibility surveys and magnetometer surveys be combined. In its opinion, one should always measure magnetic susceptibility when doing magnetometer surveys. Combining susceptibility surveys with conductivity surveys would also appear reasonable, especially if using an instrument like the EM38, with which both sets of data can be collected with one instrument. Problem Orientation Magnetic susceptibility surveys can serve predictive or interpretive roles in conjunction with other geophysical methods, but they may also be employed as a geophysical prospecting method in their own right. Because of the broad range of scales at which they may be applied and the well-developed suite of laboratory measurements within which they may be understood, they are appropriate to consider for solving a wide range of archaeological problems. These archaeological applications include (1) locating and defining sites, activity areas, and features; (2) locating and mapping buried sites; (3) delineating natural and cultural stratigraphy; (4) understanding formation processes; and (5) documenting the processes of erosion and deposition as they affect the archaeological record. A brief illustration of problem orientation follows in the case studies section below. The reader is referred to Dalan and Banerjee (1998) for a more complete treatment of this subject.

182 ~ Rinita A. Dalan

Case Studies EM38 Survey Surface surveys of magnetic susceptibility accomplished with the Geonics EM38 can provide an effective means of exploring archaeological sites and defining archaeological features. Figure 8.9 presents the results of a Geonics EM38B survey on a prehistoric Caddo site in southwest Arkansas. Jami J. Lockhart of the Arkansas Archeological Survey conducted this survey. The 20-×-20-m area was surveyed using a 50-cm transect interval and a 50-cm measurement interval along transects. The data were processed using Geoplot software. For Figure 8.9, the data have been interpolated to 25-cm2 cells. Test excavations were used to assess the large anomaly in the southwest corner of the survey area shown in Figure 8.9. These excavations confirmed that the large anomaly represented a structure and uncovered large amounts of fired clay (daub) and approximately 15 ceramic vessels. Post molds documented by excavation are overlain on the geophysical data. Radiocarbon dates derived from collected material place the structure within the early fifteenth century a.d.

Figure 8.9. Magnetic susceptibility survey of a prehistoric structure in southwest Arkansas. The unit scale is SI volume susceptibility in parts per thousand. Post molds documented by excavation are shown atop the magnetic susceptibility high produced by the structure (figure courtesy of Jami J. Lockhart, Arkansas Archeological Survey).

Magnetic Susceptibility ~ 183 Buried Site Location Buried site location is critical in many areas where landscape change has impacted and perhaps also preserved, through burial, archaeological deposits. In cases of planned construction, in areas where the potential for buried deposits is high, effective and costefficient methods that can be used to locate such deposits are critical. Protection and management of these resources is only possible when their presence is known. Traditional approaches to buried site location have emphasized the use of locational models based upon physical and cultural data to target potential sites. Labor-intensive coring or excavation is the usual approach to assessing these locations. Surface geophysical methods have been occasionally used to test locational models but often cannot provide the resolution required to discriminate thin archaeological horizons at depth. A general rule of thumb for geophysical prospecting is that location is possible for layers or objects that are as large as they are deep. For example, geophysical methods would likely identify a 1-m-wide feature at a depth of 1 m, but they would probably not identify this feature at depths significantly greater than 1 m. This rule of thumb may be modified, of course, depending on the contrast between the layer or object of interest and the surrounding soils or sediments. In general, however, it comes down to an inverse relationship between resolution and depth. Down-hole instruments and laboratory magnetic studies of collected samples provide a means of circumventing this limitation and thus may be employed to locate and explore deeply buried archaeological deposits. In a down-hole application, detection is no longer dependent on the depth of the layer or feature below the surface because the sensor is lowered down a core hole. Down-hole magnetic susceptibility techniques thus can allow fine-scale resolution of magnetic changes through the soil column and underlying sedimentary section. As susceptibility contrasts may result from soil development and cultural occupation, susceptibility surveys provide a relatively inexpensive means of locating or delimiting deeply buried stable surfaces and archaeological sites. Down-hole techniques are relatively noninvasive, requiring just a core hole and not a large-scale excavation. Down-hole magnetic susceptibility techniques are being tested and developed for buried site location within the Red River Valley region of Minnesota/North Dakota on deposits dating to the early to middle Holocene. These deposits lie within the former lakebed and margins of the southern extension of Lake Agassiz, an immense proglacial lake that formed more than 12,000 years ago. The early prehistory of this region is not well known. Only a handful of sites dating before 2000 b.p. have been documented. Paleoindian (ca. 9500–8000 b.p.), Plains Archaic (ca. 8000–2500 b.p.), and even Early and Middle Woodland (ca. 2500–1100 b.p.) sites are rare. Research at sites like Rustad (Michlovic 1996; Running 1995) and Canning (Michlovic 1986), however, indicates that people were present in the region during the early to middle Holocene. Thus, it is likely that the paucity of recognized sites is due to our inability to locate deeply buried deposits. The results of geoarchaeological studies underscore

184 ~ Rinita A. Dalan the potential for buried archaeological sites in the Upper Midwest and Northern Plains (Artz 1995; Bettis and Hajic 1995). Figures 8.10 and 8.11 present down-hole magnetic susceptibility results at two buried archaeological sites in the Red River Valley. The Rustad site is an Early Archaic (ca. 7550â&#x20AC;&#x201C;7180 b.p.) site located in southeastern North Dakota (Michlovic 1996; Running 1995). This site was preserved in alluvial fan sediments sandwiched between lacustrine (Lake Agassiz) and overlying eolian sediments. The buried soil that makes up the occupation layer at the site was clearly distinguished within the sedimentary section by enhanced susceptibility values (Dalan 2001). Down-hole susceptibility studies were also conducted at the Canning site (Michlovic 1986), an Archaic period occupation situated on a terrace remnant along the Sheyenne River, North Dakota. The buried occupation layer at this site was similarly identiďŹ ed by high susceptibility values. Buried soils that make up these sites were distinguished by virtue of their enhanced magnetic signal. In addition, increased values of magnetic susceptibility were noted for noncultural buried soils identiďŹ ed at these sites. At both sites, susceptibilities of buried cultural soils were double those measured on buried noncultural soils, indicating that the magnitude of the susceptibility signal may aid in determining whether ancient land surfaces were occupied by humans.

Figure 8.10. Down-hole magnetic susceptibility results at the Rustad site (34RI775). Approximately 80 cm of surface soils were removed prior to measurement. The buried cultural soil comprising the occupation is clearly indicated by an increase in susceptibility values.

Magnetic Susceptibility ~ 185 Hopeton Earthworks In addition to deďŹ ning sites or features, magnetic susceptibility surveys can also be employed to explore susceptibility contrasts within and between cultural features. These data may be used to understand formation processes or to plan or interpret magnetometer or other geophysical surveys. Both of these purposes have guided magnetic susceptibility investigations at the Middle Woodland period Hopeton Earthworks (33RO26), Ross County, Ohio. The objectives of the soil susceptibility studies dovetail with those proposed for research at Hopeton in general (Lynott and Weymouth 2001), namely, to provide information on the construction history of the earthworks as well as on postconstruction, or postdepositional, processes. Soil susceptibility investigations have been applied within a broader suite of soil magnetic studies and coordinated with excavation, surface geophysical surveys, and investigations of micromorphology. This research is being directed and coordinated by Mark Lynott, National Park Service, Midwest Archeological Center. Soil magnetic investigations at Hopeton have focused on a large square enclosure (Figure 8.12). Cesium gradiometer surveys (Weymouth 2002) indicate that the wall of the square enclosure can be deďŹ ned by two parallel magnetic highs (with a strength 5â&#x20AC;&#x201C;10 nT/m over background values) that extend along the length of the wall.

Figure 8.11. Down-hole magnetic susceptibility studies at the Canning site (21NR9). An even greater increase in magnetic susceptibility than that measured at the Rustad site (Figure 8.10) was observed for the buried soil comprising the Canning site.

186 ~ Rinita A. Dalan

Figure 8.12. Base map (adapted from Squier and Davis 1848) showing the earthworks at Hopeton and the locations of three trenches excavated in 2001 and 2002 that crosscut the large square enclosure. Data in Figure 8.13 were gained from Trench 3.

Figure 8.13 is a magnetic susceptibility contour map of the north face of Trench 3, one of three machine-excavated trenches that crosscut the large square earthwork (Figure 8.12). Thirty-five meters of the north wall of Trench 3 were surveyed with a Bartington Instruments MS2 susceptibility meter and an MSF field sensor using a 20cm measurement interval. Field values were first converted to volume susceptibilities in SI units and then contoured at a 0.0005 contour interval using the Golden Software Surfer program, version 7. Volume susceptibilities were estimated by multiplying field values by 1.71 × 10–5. This calibration factor was calculated through a comparison of field values recorded with the MS2F and laboratory values measured using the MS2B sensor for samples collected at this location. Though these studies were conducted within excavated trenches, equivalent data sets could be gained in unexcavated areas using a down-hole instrument.

Magnetic Susceptibility ~ 187

Figure 8.13. Magnetic susceptibility contour map of the north face of Trench 3 at the Hopeton Earthworks. Volume susceptibility measurements were accomplished using a 20-cm grid spacing with a Bartington MS2F sensor (color illustration appears on the CD).

As shown in Figure 8.13, zones of increased magnetic susceptibility are centered at approximately 2845E, 2850E, and 2857E. The zones at 2845E and 2850E comprise the outside (in this case western) of the two magnetic highs documented by magnetometer surveys along the length of the wall. The magnetic zones at 2845E and 2850E are bifurcated by a distinct silt deposit characterized by a relatively low magnetic susceptibility. Soils at 2857E form the inner (in this case eastern) magnetic zone. The area between the two magnetic zones, which makes up the core of the earthen wall, is characterized by low magnetic susceptibility, as is the subsoil underneath the wall fill. The soil magnetic data provide an explanation for the source of the two parallel magnetic highs recorded by magnetometer surveys and allow these magnetic patterns to be correlated with soil units identified in profile walls. Weymouth (2002) has compared the cesium gradiometer magnetic profile obtained along what would later be the north wall of Trench 3 with the magnetic susceptibility data obtained along the 99-m elevation line on the north wall of Trench 3 (Figure 8.14) and has shown that there is agreement both in the locations of these magnetic highs and lows and also in their relative magnitudes. As suggested from initial field studies by Bevan (2002) and confirmed through further laboratory studies conducted by Dalan et al. (2003), the magnetometer anomalies result not only from increased susceptibilities of soils comprising the magnetic zones but also from a remanent component. This remanent magnetization is approximately equal to the induced magnetization with a strength of approximately 1 × 10–1 A/m and a declination near zero degrees. Comparative work within several trenches that crosscut the enclosure wall (Figure 8.12) has indicated a general pattern consisting of a low-susceptibility core and more highly magnetic flanking soils. No developed soils have been located within the earthwork that would indicate significant interruptions of the construction sequence. There is also no evidence for developed soils at the base of the earthwork and this indicates that natural soils were truncated prior to construction. Despite these general patterns, however, differences in soils and magnetic signatures among trenches have also been

188 ~ Rinita A. Dalan noted and these are being used to provide information on construction patterns along various segments of the enclosure. Multiple fill sources have been suggested. Continuing soil magnetic studies are being directed toward understanding the sequencing and sourcing of earthen fills at various points along the enclosure and tracking postdepositional processes. Mound Form Susceptibility and other soil magnetic studies can also be used to assess postformation processes that may impact archaeological deposits. Because susceptibility contrasts are generated through pedogenic enhancement, susceptibility studies can provide information both on the effects of pedogenesis and also on processes of erosion and deposition. Erosional processes, such as soil creep and soil wash, alter spatial patterns of topsoil depth and particle size distributions and hence the depth and character of magnetic enhancement (Dearing et al. 1986). If pedogenic enhancement of the upper horizons has occurred, magnetic techniques can be used to trace relative topsoil-subsoil movements along a slope. Thus, they can be extremely useful for exploring processes of hillslope erosion (Dearing et al. 1985; Dearing et al. 1986; Thompson and Oldfield 1986).

Figure 8.14. Magnetic susceptibility values along a single elevation line (99 m) on the north face of Trench 3 at the Hopeton Earthworks compared with cesium gradiometer data collected at this location before Trench 3 was excavated. The source of the gradiometer anomalies is clearly indicated by the susceptibility contrasts.

Magnetic Susceptibility ~ 189 Information on the depth of magnetic enhancement can be used as a ďŹ ngerprint to track processes of erosion and deposition. In this way, susceptibility studies can provide an avenue for assessing the integrity of archaeological properties; they can reveal how landscapes containing cultural materials have evolved. Thus, susceptibility surveys can serve as valuable management and assessment tools. For example, Figure 8.15 presents susceptibility proďŹ les from three locations along the slope of a Mississippian period platform mound. If we assume that enhancement mechanisms operate similarly across a mound, then the depth of enhanced soils (topsoil) will be thicker in areas of net sedimentation and thinner or even depleted in erosion areas. Location I represents a stable position on the mound where magnetic enhancement is evident. Location II is on the sideslope of the mound and only a relatively thin enhanced layer is apparent. Here, erosion has exceeded the rate of enhancement. Location III reveals where soils eroded from the sideslope of the mound have been deposited. High susceptibility and ARM values throughout the depth sampled (60 cm) indicate a thick wedge of enhanced soils at this location. Susceptibility studies can be used to identify areas of stability, erosion, and deposition, and this information can even be used to determine the original form of earthworks. This type of study was initiated on Mississippian period mounds at the Cahokia Mounds State Historic Site (Dalan and Watters 1994; Dalan et al. 1996). At the Cahokia Mounds site, erosion and sedimentation, exacerbated by modern cultural activities, have worked to blur and sometimes completely eradicate the original shapes

Figure 8.15. The depth of enhanced soils can provide a clue to erosional histories and thus magnetic techniques may be used to identify areas of stability, erosion, and sedimentation. Enhanced soils will be thicker in areas of net sedimentation (Location III, footslope of a mound). Enhanced soils will be thinner or even depleted in areas where the rate of magnetic enhancement is slower than the rate of topsoil removal (Location II, sideslope of a mound). (This research was conducted using the older CGS [centimeter-gram-second] system of units for mass susceptibility and ARM.)

190 ~ Rinita A. Dalan of the mounds. A methodology consisting of a combination of fine-scale topographic mapping and soil magnetic techniques was developed and tested to determine the original form of prehistoric earthworks. This methodology has been used to distinguish platform and conical mounds and to identify less common forms such as multiterraced mounds, shared platforms, and causeways. The initial step in this process involves the preparation of a fine-scale topographic map of the mound or earthwork in question. Based upon an analysis of this map, an interpretation of the original form of the earthwork is forwarded. (Alternatively, a hypothesis regarding the original form of a mound might arise from the analysis of historic records or maps.) Susceptibility studies and other magnetic techniques are next applied to evaluate the topographic interpretation. This evaluation requires the identification of areas of stability, erosion, and sedimentation, and this information must be combined with models concerning the evolution of different forms of earthen mounds. The shape of a mound will affect the rate of erosion at different locations along its slope, as shown in Figure 8.16 for a platform mound and a conical mound. Erosion of both types of mounds would involve concurrent processes of contraction and basal spreading (slopewash). Both mounds would approach the convex/concave shape typical of hillslopes in humid climates (Carson and Kirkby 1972; Daniels and Hammer 1992; Kirkby 1971). For both platform and conical mounds, the basal concavity is formed by deposition of eroded material. But there are essential differences in terms of the erosive process as it affects these two mound types. The upper convexity for a platform mound is produced by overall downwearing in combination with the rapid erosion of steep platform margins. A conical mound already approaches the convex/concave shape typical of hillslopes in humid climates and thus erosion of a conical mound would progress more uniformly from the summit down the slopes of the mound. For a platform mound, the initial removal of high-gradient, sharply defined platform margins would produce a distinctive signature potentially measurable by soil magnetic techniques. This would include enhanced topsoils across the summit, depletion of enhanced soils along the shoulder and sideslopes, and a thick sequence of enhanced soils at the base of the mound. For a conical mound, similar depths of enhanced soils would be expected from the summit down the slope of the mound, with a thick sequence of enhanced soils at the base. Gathering the information to identify these erosional processes is a simple matter. For the study at Cahokia Mounds, samples were collected using a 1-inch (2.54-cm) push-tube corer. Samples were collected at 10-cm intervals to a depth of 60 cm below the surface. Previous research at the site indicated that this depth would encompass the transition from enhanced topsoil to less magnetic subsoil in stable situations. A down-hole susceptibility instrument, which was not available at the time of this study, could now be employed to collect these data much more quickly. All slope positions (i.e., summit, shoulder, sideslope, and footslope) were repetitively sampled. In general, sampling was accomplished along transects across two faces of each mound. Samples

Magnetic Susceptibility ~ 191

Figure 8.16. Mound erosion processes for platform and conical mounds. Erosion of sediments is much more uniform for a conical mound. Highgradient platform margins on a platform mound would be expected to exhibit an increased rate of erosion.

were packed into Althor P15 plastic (nonmagnetic) boxes (5.28-cc volume) and low field mass magnetic susceptibility (χ ) and ARM were measured on each sample. All data were measured using the older CGS system of units. The topographic analysis indicated that Mound 36 was a square, single platform mound. Areas of relative stability, erosion, and sedimentation, identified through magnetic analyses (Figure 8.17), are consistent with the model pattern expected for a platform mound. In contrast to Mound 36, Mound 62 was interpreted as a conical mound on the basis of the topographic data and historic records. Although layering of sediments within the near surface complicated susceptibility studies at this mound, the soil magnetic data indicated an erosive process very different from that observed at Mound 36. A stable summit, as observed at Mound 36, is not evident in the χ and ARM graphs for Mound 62 (Figure 8.18). As expected for a conical mound, erosion of enhanced topsoils appeared relatively uniform down slope, in contrast to the markedly different rates of erosion expected for a platform mound.

192 ~ Rinita A. Dalan

Figure 8.17. Topographic and soil magnetic data for Mound 36, Cahokia. At the Core A location, susceptibility and ARM values are relatively high at the surface and smoothly decrease with depth. This represents a relatively stable platform where magnetic enhancement has exceeded the rate of erosion. The Core B and Core G locations define the extent of the platform remnant. Relatively low susceptibilities and ARM values, characteristic of truncated soils where erosion has exceeded enhancement, are apparent in Cores C, D, H, I, and J. In contrast, susceptibility and ARM values remain relatively constant or even rise with depth at Cores E, F, K, and L, indicating a deep sequence of enhanced topsoils characteristic of a depositional situation. (The CGS system was used for this research.)

The depth of magnetic enhancement provides information about processes of erosion and deposition, but the character of magnetic enhancement can also be explored. As discussed earlier, the combination of χ and ARM can be used to understand relative changes in the concentration and grain size of the magnetic carrier. For magnetite and maghemite, an increase in the ARM/χ ratio indicates relatively finer grains and a decrease indicates relatively coarser magnetic grains. Figure 8.19 plots ARM/χ against distance from the mound summit and allows an inspection of spatial changes in magnetic grain size. ARM/χ increases at the footslope positions where relatively finegrained magnetite, produced through pedogenesis on the mound, has been deposited. A decrease in this ratio at the sideslope positions results from the erosion of magnetically fine-grained material. This methodology used to determine the form of platform and conical mounds can also be adapted to investigate other and more complex earthen forms. Speculations as to the original form of Mound 56 at the Cahokia Mounds site have ranged from ridgetop to conical to platform. Historic cultivation has been suggested as the reason that the northern slope of Mound 56 is longer and gentler than the southern slope (Figure 8.20). Fine-scale topographic maps and soil magnetic studies indicate that Mound 56 is a terraced platform (two-tiered platform). This interpretation is supported by a plot

Magnetic Susceptibility ~ 193

Figure 8.18. Topographic and soil magnetic data for Mound 62, Cahokia. At this mound, as expected for a conical mound, there is evidence of erosion from the crest of the mound down slope. At Core G, only 10 cm of enhanced soils were observed. A thin enhanced layer was similarly indicated at adjacent cores F, H, and I. Moving down slope, susceptibility values were either constant or increased with depth, indicating erosion of enhanced topsoils to expose an underlying layer characterized by higher concentrations of magnetite and finer magnetic grains. (The CGS system was used for this research.)

of ARM/χ versus distance from the crest of the mound (Figure 8.21) that shows a bimodal version of the pattern observed for Mound 36. Figure 8.21 indicates an upper platform remnant, a highly eroded upper platform edge, deposition at the base of the upper platform on a lower terrace, a lower linear slope, and a depositional footslope at the base of the mound.

Conclusion Susceptibility studies look different from the other types of geophysical surveys detailed in this volume. Guidelines from English Heritage (David 1995) recognize these contrasts and place susceptibility studies at a lower level, not on par with mainstream geophysical methods like resistivity and magnetometry. This assessment is mirrored in the general use of the Geonics instruments in North American archaeological investigations: the EM38 is employed much more commonly to measure conductivity than susceptibility. This view, however, does much to inhibit the application of susceptibility studies and fails to recognize the distinct capabilities that can be provided by this type of research. Though it is important to recognize that susceptibility studies do differ from other geophysical techniques, it is also important to be aware of critical information that these studies provide that is not supplied by other types of geophysical surveys.

194 ~ Rinita A. Dalan

Figure 8.19. A plot of ARM/χ versus distance from Core A (mound summit), Mound 36. An increase in ARM/χ is apparent for Cores E and F (the toeslope positions) and this is due to an increased concentration of relatively fine-grained magnetite deposited at these positions. A slight decrease in this ratio at Cores C and D is the result of the erosion of magnetically fine-grained material from the sideslope of the mound (color illustration appears on the CD).

First and foremost, susceptibility studies have the potential to yield information regarding the dynamics of archaeological terrains. They can be used to distinguish cultural from natural surfaces (through enhanced susceptibility values) and even to understand the cultural and natural processes that have produced the observed magnetic signatures. Because susceptibility contrasts have been linked to soil formation, investigations of susceptibility and other magnetic properties provide an avenue for understanding the factors that influence soil formation, including parent material, climate, topography, time, and living organisms. Humans are one of the living organisms that affect the soil environment and thus there is a direct link between human activities and susceptibility. Susceptibility studies can be employed to understand formation processes and they provide an avenue for examining the impacts of pedogenesis and erosion on the archaeological record. Thus, they provide an effective means for understanding both how archaeological landscapes form and how they change. The interpretive potential of magnetic susceptibility studies has made them a popular complement to magnetometer surveys. They may be used to assess the potential for magnetometry at a given site, to separate induced and remanent components, to interpret magnetic anomalies, and to generate magnetic models. Magnetic susceptibility surveys, however, can also be used to provide basic information that transcends this complementary role. To achieve a full interpretation in terms of formation and postformation processes, susceptibility data will need to be combined with other soil magnetic techniques. Labo-

Magnetic Susceptibility ~ 195

Figure 8.20. Core locations and topographic proďŹ le showing the asymmetrical nature of Mound 56, Cahokia.

ratory soil magnetic techniques allow the susceptibility signal to be resolved into components of magnetic mineralogy, concentration, and grain size. Theoretical and applied studies can be accessed to interpret these magnetic contrasts. Laboratory soil magnetic techniques are well developed and broadly applicable. They are of high sensitivity and are, in general, rapid and economical. Relatively small samples are required. Depending on the container chosen, perhaps only 10 g of soil or less is needed. These are nondirectional samples and they can be gained using a variety of sampling procedures (even subsampled from bulk soil samples). (I have even used soil samples archived from past excavations dating back multiple decades.) Provided a protocol for the sequence of measurement is followed (Maher 1986), the same samples can be used for all magnetic tests. Equipment for basic studies is relatively low cost and widely available. More specialized equipment may be accessed at a number of university laboratories, such as the Institute for Rock Magnetism at the University of Minnesota. Translation of the geophysical, in this case susceptibility, data into archaeological terms is perhaps conceptually an easier process than it is with other methods. This is due partially to the relatively small volumes sampled, as with the single-coil instruments. Susceptibility can be viewed as a property of a particular soil or sediment and archaeologists commonly work with a variety of soil properties in their interpretations of archaeological features and stratigraphic sections. Susceptibility can be treated as one of a number of properties aimed at distinguishing diďŹ&#x20AC;erent soils and sediments

196 ~ Rinita A. Dalan

Figure 8.21. A plot of ARM/χ versus distance from the crest of Mound 56 shows a bimodal version of the pattern observed for Mound 36 (Figure 8.19). Cores A and B represent relatively stable summit platform positions. Cores C, D, and E, with lower ARM/χ ratios, represent more highly eroded platform edges. Higher ARM/χ ratios at Cores F and G mark the deposition of magnetically fine-grained material at the base of this upper platform. The Core F position represents the start of an accumulation point on a lower terrace. The relatively stable second platform is represented by Core H; a more highly eroded linear sideslope by Cores I, J, and K; and a depositional toeslope for the second platform by Cores L and M (color illustration appears on the CD).

and understanding the genesis of these contrasts. In terms of basic soil susceptibility studies, measurement is straightforward, with little opportunity for field and operator error, and data processing is minimal. Information on the variation of susceptibility with depth is critical to the interpretation of susceptibility surveys. Just as one would not attempt to characterize a particular soil by examining only one horizon, susceptibility data cannot be fully interpreted without this depth information. Lateral surveys of magnetic susceptibility are possible, but it is recommended that these surveys include an analysis of representative samples collected at depth. While this is an added requirement compared with other geophysical surveys, susceptibility studies provide greater potential for understanding fine-scale site stratification than any other method with the exception of radar. Though down-hole instruments allow much quicker data acquisition than traditional sampling techniques and laboratory studies, they are still relatively time consuming. Because the process of magnetic enhancement occurs as part of soil development, susceptibility studies have the potential for broad application in archaeology. Of course,

Magnetic Susceptibility ~ 197 the degree and type of magnetic contrasts will be unique to the soil-forming factors operating in each area. Certainly contrasts will be greater in some areas and for some features, while in others they will be so slight as to be unusable. Gleyed or podsolized soil environments are not conducive to magnetic susceptibility surveys. Magnetic susceptibility surveys are potentially applicable to a wide range of archaeological sites, features, and layers. As with magnetometer surveys, areas of developed soils or areas with soils capable of significant enhancement are good bets for susceptibility surveys. Relatively young soils, for example, the alluvial soils at the Cahokia Mounds site with their contrasting textures, may also produce useful data, however. Direct measurement of representative samples from topsoils, subsoils, features, and cultural layers allows a quick and accurate determination of the potential of magnetic susceptibility studies at a particular site. Susceptibility studies have been applied at highly variable horizontal and vertical scales. They can recognize diffuse or graded boundaries; a high contrast interface is not necessary. Area surveys covering broad regions are possible with the Geonics instruments. Coarse-sample topsoil surveys with the Bartington sensors may also be used to characterize area variations in susceptibility. The use of widely spaced down-hole measurements or the analysis of collected depth samples may allow three-dimensional changes in susceptibility to be effectively imaged. Small-scale horizontal and vertical changes in susceptibility may also be investigated using appropriate sensors and collecting strategies. The MS2F and MS2K sensors, for example, allow contrasts over 1–2 cm to be resolved. The development of down-hole instruments allows this resolution to be maintained at depth. Even given changing attitudes toward the magnetic susceptibility method, it is unlikely that magnetic susceptibility surveys will ever surpass or even equal other methods of geophysical exploration such as magnetometry or resistivity. The reason for this relates largely to the issue of survey speed, although there are other concerns. Electromagnetic instruments are not as fast as magnetometers. Single-coil instruments are even slower. The collection of depth information is time consuming. Drift is a perennial problem that must be addressed with careful field procedures. Sources of magnetic interference may prohibit the use of electromagnetic instruments. Down-hole applications will be limited in areas that are difficult to core. The relatively shallow depths of investigation provided by current instrumentation are also a critical limitation, although down-hole instruments will expand capabilities for collecting this information and result in increased applications of magnetic susceptibility surveys in archaeology. Despite these concerns, however, magnetic susceptibility surveys provide distinct opportunities for solving imponderable archaeological questions that cannot be approached with other geophysical methods. Magnetic susceptibility instruments, like other geophysical equipment, can be employed to identify and define archaeological sites and features. But they can also be applied to address questions for which these other methods probably would not be used. For example, magnetic susceptibility studies can locate deeply buried archaeological layers, even those containing relatively few

198 ~ Rinita A. Dalan archaeological features. Susceptibility instruments allow application at relatively small scales. They could be used to define the limits of archaeological features that are not immediately apparent in plan or profile, or they might be employed to resolve fine-scale layering within an archaeological feature. Importantly, in combination with other soil magnetic tests, they can be directed toward understanding the genesis of archaeological features and even to track erosion and other postformation processes. Archaeologists have just begun to explore the potential of these data. Future applications of magnetic susceptibility studies appear unlimited.

Acknowledgments The case studies presented in this chapter were made possible through funds provided by various agencies. The mound form studies were supported by a grant from the Cahokia Mounds Museum Society. The National Park Service, Midwest Archeological Center, supported soil magnetic work at the Hopeton Earthworks. The National Park Service and the National Center for Preservation Technology and Training have funded efforts directed toward the development and application of down-hole susceptibility instruments. The contents of this publication are solely the responsibility of the author and do not necessarily represent the official positions or policies of the National Park Service or the National Center for Preservation Technology and Training. This material is also based upon development work supported by the National Science Foundation under Grant No. 021572. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science Foundation. Jami J. Lockhart of the Arkansas Archeological Survey supplied the illustration and information used in the EM38 case study. John Weymouth provided the magnetic profile that was used in comparison with the magnetic susceptibility data from the Hopeton Earthworks. Aaron Fogel prepared the magnetic susceptibility contour map for Trench 3 at the Hopeton Earthworks and the figure comparing the magnetometer and magnetic susceptibility data. The soil magnetic work at Hopeton has been a joint effort by George R. Holley, Kelsey Lowe, Aaron Fogel, and me. The mound form study was conducted by Harold W. Watters, Jr., and me. Harold W. Watters, Jr., prepared all figures used in that case study. Many thanks are due NASA and the Geoinformatics Center at the University of Mississippi for support of the workshop that generated this publication and for their efforts in producing this volume.

References Cited Aitken, M. J. 1970 Magnetic Location. In Science in Archaeology, edited by D. Brothwell and E. Higgs, pp. 681–694. Praeger, New York.

Magnetic Susceptibility ~ 199 Artz, J. A. 1995 Geological Contexts of the Early and Middle Holocene Archaeological Record in North Dakota and Adjoining Areas of the Northern Plains. In Archaeological Geology of the Archaic Period in North America, edited by E. A. Bettis III, pp. 67–86. Geological Society of America Special Paper 297. Boulder, Colorado. Banerjee, S. K., J. W. King, and J. A. Marvin 1981 A Rapid Method for Magnetic Granulometry with Application to Environmental Studies. Geophysical Research Letters 8:333–336. Benech, C., and E. Marmet 1999 Optimum Depth of Investigation and Conductivity Response Rejection of the Different Electromagnetic Devices Measuring Apparent Magnetic Susceptibility. Archaeological Prospection 6:31–45. Bettis, E. A. III, and E. R. Hajic 1995 Landscape Development and the Location of Evidence of Archaic Cultures in the Upper Midwest. In Archaeological Geology of the Archaic Period in North America, edited by E. A. Bettis III, pp. 87–113. Geological Society of America Special Paper 297. Boulder, Colorado. Bevan, B. 1996 Geophysical Exploration for Archaeology. Geosight Technical Report No. 4. Geosight, Weems, Virginia. 2002 Geophysical Tests in the Hopeton Excavations. Geosight, Weems, Virginia. Submitted to the Midwest Archeological Center, National Park Service, Lincoln, Nebraska. Bevan, B., and R. A. Dalan 2003 Magnetic Susceptibility Sounding. Electronic publication. Copies available from Geosight, Weems, Virginia. Carson, M. A., and M. J. Kirkby 1972 Hillslope Form and Process. Cambridge University Press, Cambridge. Challands, A. 1992 Field Magnetic Susceptibility Measurement for Prospection and Excavation. In Geoprospection in the Archaeological Landscape, edited by P. Spoerry, pp. 33–41. Oxbow Monograph 18. Oxbow Books, Oxford. Clark, A. J. 1996 Seeing Beneath the Soil: Prospecting Methods in Archaeology, new ed. B. T. Batsford, London.

200 ~ Rinita A. Dalan Dalan, R. A. 2001 A Magnetic Susceptibility Logger for Archaeological Application. Geoarchaeology 16:263–273. Dalan, R. A., and S. K. Banerjee 1998 Solving Archaeological Problems Using Techniques of Soil Magnetism. Geoarchaeology 13:3–36. Dalan, R. A., and B. Bevan 2002 Geophysical Indicators of Culturally Emplaced Soils and Sediments. Geoarchaeology 17:779–810. Dalan, R. A., K. Lowe, and A. Fogel 2003 Soil Magnetic Studies of the Hopeton Earthwork 2002 Excavations. Minnesota State University Moorhead. Submitted to the Midwest Archeological Center, National Park Service, Lincoln, Nebraska. Dalan, R. A., and H. W. Watters, Jr. 1994 Determining the Form of Mississippian Mounds. Office of Contract Archaeology, Southern Illinois University Edwardsville. Submitted to the Cahokia Mounds Museum Society, Collinsville, Illinois. Dalan, R. A., H. W. Watters, Jr., and S. K. Banerjee 1996 The Application of Soil Magnetic Techniques for Determining the Original Form of Prehistoric Earthworks. Proceedings of the Sixty-Sixth Annual Meeting of the Society of Exploration Geophysicists, vol. 1, pp. 790–793. Tulsa, Oklahoma. Daniels, R. B., and R. D. Hammer 1992 Soil Geomorphology. John Wiley and Sons, New York. David, A. 1995 Geophysical Survey in Archaeological Field Evaluation. Ancient Monuments Laboratory, English Heritage Society, London. Dearing, J. A. 1994 Environmental Magnetic Susceptibility Using the Bartington MS2 System. Chi, Kenilworth, Great Britain. Dearing, J. A., K. L. Hay, S. M. J. Baban, A. S. Hudleston, E. M. H. Wellington, and P. J. Loveland 1996 Magnetic Susceptibility of Soil: An Evaluation of Conflicting Theories Using a National Data Set. Geophysical Journal International 127:728–734.

Magnetic Susceptibility ~ 201 Dearing, J. A., B. A. Maher, and F. Oldfield 1985 Geomorphological Linkages between Soils and Sediments: The Role of Magnetic Measurements. In Geomorphology and Soils, edited by K. S. Richards, R. R. Arnett, and S. Ellis, pp. 245–266. Allen and Unwin, London. Dearing, J. A., R. I. Morton, T. W. Price, and I. D. L. Foster 1986 Tracing Movements of Topsoil by Magnetic Measurements: Two Case Studies. Physics of the Earth and Planetary Interiors 42:93–104. Evans, M. E., and F. Heller 2003 Environmental Magnetism: Principles and Applications of Enviromagnetics. Academic Press, New York. Holley, G. R., R. A. Dalan, and H. W. Watters, Jr. 1996 Investigations at the West Borrow Pit Mound Group, Cahokia Mounds State Historic Site. Office of Contract Archaeology, Southern Illinois University Edwardsville. Submitted to the Illinois Historic Preservation Agency, Springfield, and the Cahokia Mounds Museum Society, Collinsville, Illinois. Hunt, C. P., B. M. Moskowitz, and S. K. Banerjee 1995 Magnetic Properties of Rocks and Minerals. In Rock Physics and Phase Relations: A Handbook of Physical Constants, edited by T. J. Ahrens, pp. 189–204. AGU Reference Shelf 3. American Geophysical Union, Washington, D.C. King, J. W., S. K. Banerjee, J. Marvin, and O. Ozdemir 1982 A Comparison of Different Magnetic Methods for Determining the Relative Grain Size of Magnetite in Natural Materials: Some Results in Lake Sediments. Earth and Planetary Science Letters 59:404–419. Kirkby, M. J. 1971 Hillslope Process-Response Models Based on the Continuity Equation. Institute of British Geographers, Special Publications Series 3:15–30. LeBorgne, E. 1955 Susceptibilité magnétique anormale du sol superficiel. Annales de Géophysique 11:399–419. 1960a Etude expérimentale du trainage magnétique dans le cas d’un esemble de grains magnétiques très fins disperses dans une substance normagnétique. Annales de Géophysique 16:445–494. 1960b Influence du feu sur les propriétés magnétiques du sol et sur celles du schist et du granite. Annales de Géophysique 16:159–195.

202 ~ Rinita A. Dalan Lecoanet, H., F. Lévêque, and S. Segura 1999 Magnetic Susceptibility in Environmental Applications: Comparison of Field Probes. Physics of the Earth and Planetary Interiors 115:191–204. Lynott, M. J., and J. W. Weymouth 2001 Investigations at the Hopeton Earthwork, Ross County, Ohio in the 2001 Season. National Park Service and the University of Nebraska, Lincoln. McNeill, J. D. 1986 Geonics EM38 Ground Conductivity Meter: Operating Instructions and Survey Interpretation Techniques. Technical Note TN-21. Geonics Ltd., Mississauga, Ontario, Canada. McNeill, J. D., J. A. Hunter, and M. Bosnar 1996 Application of a Bore-Hole Induction Magnetic Susceptibility Logger to Shallow Lithological Mapping. Journal of Environmental and Engineering Geophysics 0(2):77–90. Maher, B. A. 1986 Characterisation of Soils by Mineral Magnetic Measurements. Physics of the Earth and Planetary Interiors 42:76–92. Maher, B. A., and R. M. Taylor 1988 Formation of Ultrafine-Grained Magnetite in Soils. Nature 336:368–371. Maher, B. A., and R. Thompson (editors) 1999 Quaternary Climates, Environments and Magnetism. Cambridge University Press, Cambridge. Michlovic, M. G. 1986 The Archaeology of the Canning Site. The Minnesota Archaeologist 45:3–36. 1996 Archaeological Excavations at the Rustad Site. In Quaternary Geology of the Southern Lake Agassiz Basin, edited by K. L. Harris, M. R. Luther, and J. R. Reids, pp. 127–135. North Dakota Geological Survey Miscellaneous Series 82. Bismarck, North Dakota. Mullins, C. E. 1974 The Magnetic Properties of the Soil and Their Application to Archaeological Prospecting. Archaeo-Physika 5:143–347. Parchas, A., and A. Tabbagh 1978 Simultaneous Measurement of Electrical Conductivity and Magnetic Susceptibility of the Ground in EM Prospecting. Archaeo-Physika 10:682–691.

Magnetic Susceptibility ~ 203 Rummery, T. A., J. Bloemendal, J. Dearing, and F. Oldfield 1979 The Persistence of Fire-Induced Magnetic Oxides in Soils and Lake Sediments. Annales de Géophysique 35:103–107. Running, G. L. IV 1995 Archaeological Geology of the Rustad Quarry Site (32RI775): An Early Archaic Site in Southeastern North Dakota. Geoarchaeology 10:183–204. Scully, F. M. 1997 An Evaluation of the Use of In Situ Magnetic Susceptibility during the Course of Excavation. MSc Dissertation, Department of Archaeological Sciences, University of Bradford, Bradford, Great Britain. Squier, E. G., and E. H. Davis 1848 Ancient Monuments of the Mississippi Valley. Smithsonian Contributions to Knowledge No. 1. Smithsonian Institution, Washington, D.C. Tabbagh, A. 1986 Applications and Advantages of the Slingram Electromagnetic Method for Archaeological Prospecting. Geophysics 51:576–584. Thompson, R., and F. Oldfield 1986 Environmental Magnetism. Allen and Unwin, London. Tite, M. S. 1972 The Influence of Geology on the Magnetic Susceptibility of Soils on Archaeological Sites. Archaeometry 14:229–236. Tite, M. S., and R. E. Linington 1975 Effect of Climate on Magnetic Susceptibility of Soils. Nature 256:565–566. Tite, M. S., and C. E. Mullins 1970 Magnetic Properties of Soils. Prospezioni Archeologiche 5:111–112. 1971 Enhancement of the Magnetic Susceptibility of Soils on Archaeological Sites. Archaeometry 13:209–219. Weymouth, J. W. 2002 Geophysical Investigations at the Hopeton Earthwork, Ross County, Ohio: The 2002 Season. Submitted to the Midwest Archeological Center, National Park Service, Lincoln, Nebraska.


Magnetometry: Natureâ&#x20AC;&#x2122;s Gift to Archaeology Kenneth L. Kvamme

Magnetometry is one of the most productive prospecting methods employed in archaeology. It is a method that responds particularly well to the archaeological record because a variety of natural and cultural processes combine to generate numerous magnetic variations that point to subsurface features. It is almost as if nature designed the components of archaeological sites to be made visible by the magnetic variations they exhibit. Additionally, the rapid data acquisition rates of current instrumentation allow large areas to be surveyed in relatively small amounts of time. High spatial sampling densities are also achieved, well into the sub-meter range, oďŹ&#x20AC;ering good spatial detail. This combination of large-area coverage and high spatial resolution surpasses the capabilities of all other ground-based geophysical survey methods, and this is not a trivial factor. The imaging of large contiguous areas increases the likelihood that whole cultural features with regular, interpretable geometric shapes will be encountered and recognized. This phenomenon forms the basis of a fundamental recognition element in air photo interpretation and satellite remote sensing (Avery and Berlin 1992:52): circles, ellipses, squares, rectangles, and straight lines in imagery are generally of human origin (such geometric forms occur much less frequently as products of nature). Surveys of small areas might reveal cultural features, but without seeing them in their entirety and context it is diďŹ&#x192;cult or impossible to interpret what they might represent or to ascribe

206 ~ Kenneth L. Kvamme significance to them. A small linear feature could represent part of a room, a house wall, a ditch edge, pavement, a trail, or a road, for example. Surveys of large contiguous areas increase the likelihood that sense can be made of patterns in cultural landscapes, and magnetometry is one of the best ways to obtain such surveys in archaeological geophysics. The combination of high spatial resolution, wide-area coverage, the increased pattern recognition that large-area surveys allow, and the many characteristics of archaeological deposits that cause magnetic variations gives magnetometry one of the highest detection probabilities in remote sensing for small and varied cultural features over broad landscapes. The net result is that magnetometry has become the workhorse of archaeological geophysics. Successful case studies number in the thousands from around the globe, and entire villages and even urban areas of the past have been surveyed and mapped within areas approaching a square kilometer (e.g., Gaffney et al. 2000; Summers et al. 1996). Magnetometry is a prospecting method that maps local variations of the earth’s magnetic field in the near-surface. It is a rare passive method of remote sensing because it employs the earth’s magnetic field, rather than generating its own artificial one (the active technologies of radar, resistivity, and electromagnetic induction all introduce artificial fields to measure responses in the ground). Useful results in any form of remote sensing are obtained from contrasts between archaeological features and the natural background. In other words, if archaeological deposits or features possess physical properties different from those of the surrounding matrix, a distinction may be noticed between them. A buried stone foundation, for example, might be more magnetic than the surrounding earth. Such contrasts are referred to as “anomalies” until they can be identified, a task that often requires excavation. Yet, anomalies frequently can illustrate a sufficiently clear pattern for direct interpretation, as when the rectangle of a house foundation is unambiguously expressed. This phenomenon forms the basis of the pattern-recognition approach, which becomes increasingly possible with surveys of large areas because whole features are better understood than partially revealed ones. These ideas are more fully illustrated below. The basic sources for archaeological magnetometry include Aitken (1974), Bevan (1998), Clark (2000), Tite (1972), and Weymouth (1986), with more advanced material presented by Scollar et al. (1990).

History of Archaeological Magnetometry Magnetometry is concerned with the strength of the earth’s magnetic field. Although a common compass is designed to measure the direction of this field, it also responds to the field’s strength by its oscillation period. This phenomenon may be employed as a crude detector of large anomalies. Bevan (1998:20) reports that the National Park Service used this method to locate a sunken Civil War ironclad and illustrates changes in a simple compass needle by as much as two degrees in the vicinity of an iron-filled well at a Civil War fort in Petersburg, Virginia. Magnetometers are instruments specifically designed to measure the strength of a magnetic field. Early work in archaeomagnetic dating in the late 1950s noted not only that pottery kilns gave evi-

Magnetometry ~ 207 dence of the direction of the earth’s magnetic field at the time of their use but also that a magnetometer should theoretically be able to detect their appreciable magnetic signal. The discovery of the proton free precession principle allowed Aitken et al. (1958) of Oxford University to construct an instrument to test this very idea for the first time in archaeology. With it, a relatively precise measurement of magnetic field strength could be acquired in an astonishing 5 seconds; by sampling a large area every 5 feet, a Romano-British pottery kiln was ultimately located by anomalously large readings in its vicinity (Clark 2000:17). Subsequent use demonstrated that topsoil-filled pits and ditch features could also be detected, which came as a surprise despite Le Borgne’s (1955) earlier work that had demonstrated that topsoil possesses a greater magnetic susceptibility than subsoil. Additionally, highly burned pits resulted in astonishingly high measurements, which we now realize was due to the different phenomenon of thermoremanence (see below). So was born the modern field of magnetic prospecting in archaeology (Clark 2000:17). Other developments soon followed. Alldred (1964), also from Oxford, developed the continuously reading fluxgate gradiometer primarily for its great advantage of speed over the widely accepted proton systems (see below), demonstrating that a survey of 2,700 m2 could be accomplished in an hour (exceeding even contemporary survey rates, but transect separation was high and few anomalies that required investigation were located). At about the same time, Scollar and Krückeberg (1966), in Germany, demonstrated the utility of collecting data automatically onto punch tapes (an early form of data logging) and the benefits of computer processing and display of magnetometry data. In the United States, Weymouth (1976, 1986) conducted numerous tests and applications of magnetometry, particularly in the Great Plains.

Types of Magnetism Thermoremanent Magnetism In the absence of a magnetic field few materials exhibit magnetism. Any magnetism that does exist is termed remanent because it remains after the process that generated it. Most soils, clays, and rocks contain between 1 and 10 percent iron oxides that form small magnetic domains pointing in random directions. Their net magnetic effect is small owing to mutual annulment. At high temperatures, beyond what is known as the Curie point (about 600° C), the domains line up parallel to the orientation of the earth’s magnetic field at the time of the firing. Upon cooling, they remain “frozen” in this situation, forming the basis of archaeomagnetic dating (see Eighmy and Sternberg 1990). Although the magnetization of each domain returns to its previous state, the total effect is now appreciable, however, because the magnetic fields add together instead of canceling each other out, making fired features readily detectable by magnetometry (Aitken 1970:683; Clark 2000:64). This form of magnetism is known as thermoremanent magnetism, which is permanent. Although its maximum effect occurs

208 ~ Kenneth L. Kvamme when clay is heated above the Curie point, any elevated temperature produces some thermoremanence (Tite 1972:11). It is carried by all igneous rocks but is strongest in recently formed basalts and weakest in granites. Induced Magnetism and Magnetic Susceptibility Obviously, all soils, sediments, and rocks exist within the earth’s magnetic field, which introduces a second form of magnetism known as induced magnetism that exists only in the presence of that field. The ability of a material to become magnetized is a function of its magnetic susceptibility, which depends on the presence of magnetizable minerals. In soils these are essentially the three oxides of iron: hematite, magnetite, and maghemite, but only the last two are significantly magnetic (Clark 2000:100). Most soils and rocks contain only small proportions of these minerals that make magnetic detection possible. In general, soils possess a magnetic susceptibility larger than that of the rocks from which they came (except for those that develop from very magnetic volcanic rocks). This arises from several factors. First, there is a natural tendency for iron minerals to accumulate in topsoil; they are relatively insoluble and remain while there may be a net loss of less magnetic materials (calcites, silicates; Aitken 1970). A natural “fermentation process” also occurs with alternating periods of wetness and dryness that may have some effects on oxides and the transmutation of hematites to more magnetic maghemites. Fires, whether natural or anthropogenic, reduce hematite to magnetite, which upon cooling is partially reoxidized to maghemite, both of which have greatly increased magnetic susceptibility (Dabas and Tabbagh 2000). Additionally, magnetotactic and other bacteria concentrate magnetic compounds in topsoil layers (Fassbinder et al. 1990). It is of special relevance to archaeology that extended human occupations tend to exacerbate some of these effects through the introduction of organic and fired materials to the topsoil (see below). It is also of some importance that paleosols tend to retain this effect (see Dalan, this volume, for further details). It is emphasized that thermoremanent and induced magnetism in materials of high magnetic susceptibility look the same to a magnetometer. These instruments measure the sum of induced magnetism due to susceptibility and all forms of remanent magnetism. The magnetic contrast between a feature and adjacent deposits must be large enough to be detected given instrument resolution and sampling densities, however. Electromagnetic induction meters that actively generate artificial magnetic fields are able to directly measure the magnetic susceptibility of materials, but they constitute a different form of prospecting that frequently requires laboratory measurement of samples or field instruments confined to very shallow depths (see Dalan, this volume).

The Earth’s Magnetic Field Magnetic field strength is measured in nanoteslas (nT; 10–⁹ tesla; formerly known as the gamma). The earth’s magnetic field, arising from currents produced by the molten iron core, has a strength ranging from about 30,000 nT at the magnetic equator to

Magnetometry ~ 209 about 60,000 nT at the magnetic poles (the magnetic equator and poles are different from their geographic counterparts; Weymouth 1986:341). This is noteworthy because magnetic anomalies of potential archaeological interest often lie well within ±5 nT of the background, and soil unit differences can be as subtle as 0.5 nT and less (recent work by Becker [1995] shows anomalies in the picotesla [0.001 nT] range). Magnetic survey instrumentation must therefore be incredibly sensitive, capable of detection on the order of one part in a half million. In effect, the earth’s magnetic field acts as if a giant bar magnet is at its center. That its alignment does not coincide exactly with the axis of rotation means that magnetic north is spatially offset from geographic north by a small amount. The difference in the angle between geographic north and geomagnetic north measured with a compass at some point on the globe is known as declination, a characteristic that has some significance in magnetometry surveys. The lines of magnetic force that cause a compass needle to point to the poles also intersect the earth’s surface at various angles, depending on location. At the magnetic poles this inclination angle is vertical, at the magnetic equator it is horizontal, and at the midlatitudes of the 48 contiguous United States and of much of Europe it varies from about 55 to 75 degrees (Mussett and Kahn 2000:139–142). This angle, too, has some effect on magnetic interpretations and causes a characteristic phenomenon in magnetometry data sets. Magnetic features possess their own local magnetic fields that are added to that of the earth’s, with north and south poles generally aligned on the current magnetic axis (although deviations can occur depending on feature shape and other factors). A separation between these poles occurs as a result of the inclination angle. The two apparent poles are referred to as dipolar, with a positive peak generally lying somewhat south of the true locus of the anomaly and a depression (or negative) in the magnetic field occurring to the north at mid-northern latitudes. Secular variation in the earth’s magnetic field occurs over long spans of time, causing the magnetic poles to migrate over the globe. This phenomenon makes the archaeomagnetic dating of hearths and other fired features possible because anomalies caused by thermoremanence yield a magnetic declination matching that at the time of their firing, which allows an estimate of their age if the location of the geomagnetic poles is approximately known (Eighmy and Sternberg 1990). It also means that dipolar anomalies associated with hearths and other burned areas may not align with the present magnetic axis but with that at the time of firing. There are other temporal variations in the earth’s magnetic field with periods that range from seconds to hours that have a more immediate effect on magnetometry surveys. The solar wind interacts with the earth’s magnetosphere producing a diurnal variation, or regular daily cycle, owing to the planet’s rotation. On a typical day this variation might be 40 to 100 nT (Weymouth and Lessard 1986), but occasional magnetic “storms” (that tend to correlate with sunspot activity) produce changes in the field that range over hundreds of nanoteslas in the course of hours. Superimposed on top of this are micropulsations on the order of fractions of a second to minutes, with

210 ~ Kenneth L. Kvamme amplitudes as high as tens of nanoteslas (Campbell 1997:156–160). Magnetometry survey methods and instrument design specifically allow for these phenomena because in recording magnetic field strength one must be able to separate actual variations in space from changes that occur through time.

Instrumentation There are three basic types of magnetometers that are commonly employed in archaeological surveys: the proton precession magnetometer, the fluxgate gradiometer, and the cesium (or alkali vapor) magnetometer. Their names merely denote some fundamental characteristic of their physical principle (see Scollar et al. 1990 for details). Each offers advantages and disadvantages, and the proton instrument now represents an older technology that is seldom employed outside of teaching (or as an inexpensive starter system) owing to its slow rate of data acquisition. All can be configured in two different modes: as a total field instrument with a single sensor that measures the actual magnitude of the magnetic field at any locus or as a gradiometer that measures a difference between two sensors. Fluxgate sensors are almost always configured as gradiometers, but the other instruments can be utilized in either mode. Both modes must deal with the diurnal variation of the earth’s magnetic field. For a total field survey, two instruments should be utilized (an added expense). One instrument is maintained at a fixed locus as a base station that records only temporal magnetic changes. A second, roving unit simultaneously measures spatial and temporal variations (Weymouth and Lessard 1986). By differencing the two data sets, measurements are derived that represent only the spatial variation obtained by the roving instrument. A gradiometer survey, on the other hand, records the difference between measurements made simultaneously by two vertically separated (usually by 0.5–1 m) sensors in a single instrument package. It makes use of the fact that magnetic field strength falls off with the third power of distance from a target. If a sensor 1 m from a target yields a measurement of 1 nT, a second sensor at a distance of 2 m will record a value of only 1/23 = 1/8 nT. The simultaneous measurements in a gradiometer eliminate temporal variations and the need for two instruments but at the cost of somewhat reduced sensitivity. In the foregoing example, while a total field instrument will give a value of 1 nT at 1 m from the target, a gradiometer with 1 m sensor separation and the bottom sensor also at 1 m from the target will yield a measurement of only 1 nT – 1/8 nT = 7/8 nT. Most instruments now in use are configured as gradiometers, partially because fluxgate systems are so popular in archaeology (owing to their generally lower cost) and the reduced expense of a single instrument. They offer other advantages as well. Owing to the effects of differencing between two vertically separated sensors, gradiometers are less affected by lateral changes than total field instruments. Consequently, the proximity of large magnetic fields and iron objects on a site is less of an issue, and the instruments also are better able to distinguish between two closely spaced anomalies (Bevan 1998:19). On the other hand, two sensors can be much heavier to carry than one, and

Magnetometry ~ 211 the two must be maintained in a vertical alignment—any tilt and incorrect data will result. As noted, because a difference in measurements is computed, the magnitude of the anomalies detected will be smaller than that obtained with a total field instrument, meaning that the latter can locate anomalies that are smaller and deeper. Finally, the differencing has another effect. Broad changes in the magnetic field tend to be cancelled out because the differencing, in effect, acts much like a high pass filter such that low-frequency or slowly changing spatial variations are removed (see Kvamme, this volume, Chapter 10). Like other geophysical instruments, most magnetometers are manufactured by companies that have a primary focus in geological, engineering, environmental protection, or military applications. Archaeology represents a very small portion of their income. Most commercial, off-the-shelf instruments are therefore designed for general-purpose applications outside of archaeology, and one must bear this in mind not only when purchasing an instrument but also when configuring one for use in prospecting at archaeological sites. The instrument’s instruction manual may offer advice, such as sensor heights above the ground surface, more appropriate for detecting geological, as opposed to archaeological, features, for example. Other considerations in purchasing or renting an instrument relate to ease of use and handling, data acquisition speed, sensitivity, accuracy, and factors such as instrument noise, drift, and the need for tuning. Detection probabilities are affected by instrument sensitivity. A magnetometer accurate to 0.01 nT can detect smaller, deeper, and more subtle anomalies than one that measures only to the 0.1-nT level, for example. Instrument noise refers to variation in measurements when it is held stationary over a target (this variation largely results from the quality of the electrical components as well as sensor type). If variation is on the order of ±0.3 nT, then targets must have considerably larger values than this to be detected. Instrument speed relates to the size of the areas that can be surveyed in a given amount of time. Proton Precession Magnetometers Beginning with Aitken et al. (1958), early work utilized proton precession magnetometers that were capable of 0.1-nT resolution but were very slow, requiring about 5 seconds for a measurement (Clark 2000; Weymouth 1986). The Geometrics-856 series became one of the predominant instruments in use through the 1970s and 1980s and included an integrated data logger (an internal memory device for the storage of measurements) capable of storing 1,400 readings (2,500 in base station mode), with later models more than quadrupling those numbers. It can be configured in total field or gradiometer modes with variable sensor heights and separations. The requirement of eight D-cell batteries for about two days of operation brings an added expense to a project, especially if two instruments are in use. Moreover, the magnetic fields generated by the batteries, especially modern steel-jacketed ones, can introduce artifacts to the data if a large and constant distance between the instrument and the sensor heads

212 ~ Kenneth L. Kvamme is not maintained. Obtaining quality data usually requires two operators linked by a cable: one to carry the sensor head(s) and a second to operate the instrument/battery package. The latter individual should maintain a large and constant distance and direction from the sensors (dictated by the cable length, about 2 m) in order to eliminate the possibility of introducing data artifacts. These robust instruments are still in use and new ones may be purchased for about $5,000 (Figure 9.1a). Cesium-Vapor Magnetometers Cesium-vapor magnetometers are popular instruments in archaeology, capable of yielding 10 measurements per second at 0.1-nT resolution, with 0.01 nT available at a much slower rate (about 1 measurement/second). A number of companies manufacture these instruments. The Geometrics-858 is a popular model in the United States, available with two sensor heads (for gradiometer configuration) for about $25,000. Owing to sensor sensitivity to electromagnetic fields, they must be separated from the instrument and battery pack. Geometrics’s solution is to use a 2-m horizontal boom with sensor(s) on one end and a counterweight on the other (Figure 9.1b), which is not only heavy but may also exacerbate vertical sensor height changes over the soil arising from the operator’s gait that can lead to minor survey defects (see Kvamme, this volume, Chapter 10). Geometrics’s design philosophy, common to American geologyoriented companies, is to let the instrument run at a specific rate (e.g., 10 measurements/second) through the course of a survey transect; the operator may insert fiducial markers at regular intervals (e.g., 1 m or 5 m) to indicate spatial position. The result is a somewhat irregular distribution of measurements from meter to meter, with variation between, for example, 7–14 measurements/meter depending on the operator’s walking speed. While this is not a problem with modern computer software, subsequent data processing requires resampling and interpolation of the data to a regular grid (with a consistent number of measurements/meter) for handling by raster displays, which some may view as a drawback because one does not work with actual measurements but with estimated ones. For improved data quality one may eliminate use of the horizontal boom and configure a two-person arrangement much like that described for the proton precession magnetometer (Figure 9.1a). One person carries the sensor package linked by a cable to the other operator holding the instrument and batteries, and these two people maintain a large and constant separation during data acquisition. Fluxgate Grdiometers Fluxgate magnetometers are directionally sensitive and must be configured as gradiometers for practical application. Unlike other instruments that record the difference between two total field measurements in a gradiometer mode, a fluxgate gradiometer is able to measure only the vertical component of the magnetic field. Fluxgate gradiometers are subject to drift (a systematic change in the instrument’s zero point) with

Magnetometry ~ 213

Figure 9.1. Principal magnetometers used in archaeology. (a) The Geometrics-856 proton precession magnetometer showing typical field handling with sensor (right) separated from the data logger/battery package (left). Note the stationary base station with sensor to the rear. (b) The Geometrics-858 cesium vapor magnetometer in gradiometer mode with counterweighted boom assembly, instrument package at waistline, and batteries at small of back. (c) The Geoscan Research FM36 fluxgate gradiometer with twin sensors encased in vertical post and data logger/controls to the front.

temperature changes and must be frequently tuned. They also must consistently face a single direction during a survey. The Geoscan Research FM256 is an unusual instrument because it is specifically designed for archaeological applications, allowing rapid, high-density data acquisition over broad areas to a resolution of 0.1 nT. A British product, it and its predecessor (the FM36) dominate applications in Europe and, in recent years, in the United States. Unlike instruments from other manufacturers that allow sensor separation distances to be varied in gradiometer mode, the Geoscan FM series has fixed top and bottom sensors with a vertical separation of 0.5 m within the instrument (Figure 9.1c). The FM256 may be placed in an automatic recording mode that acquires up to 16 measurements per time unit, with the interval allowed to vary between about 1 and 3 seconds. Much like a metronome, an audible signal is given at the end of each interval. During survey, the instrument is moved at a uniform pace along each transect with great care to ensure that it is aligned with meter marks as each signal sounds, the speed of which may be regulated according to the user’s walking pace. The matching of the audible signal with the meter marks ensures that the data are properly located spatially. Although fluxgate gradiometers are capable of recording data very rapidly, the quality of information acquired is partially a function of how well the instrument is currently tuned (it must be zeroed frequently to minimize instrument noise and drift) and how steady the instrument’s heading is maintained across a transect—any wobbling or wiggling by the operator can introduce errors to the data (see Kvamme, this volume, Chapter 10). In practice, the acquisition of quality data at the sub-nanotesla level demands a slow walking pace and a rate of data logging much less than the instrument’s capabilities. An internal data logger is capable of storing 250,000 measurements for later downloading and processing. The current price is about $20,000.

214 ~ Kenneth L. Kvamme

Field Methods Magnetic surveys are typically performed in area surveys composed of grids that range from 20 × 20 m to 50 × 50 m in size. Most practitioners utilize ropes or tapes with meter marks placed parallel to each other on the ground to form the grids. The instruments are then moved along these guides to accurately locate the measurements (Figure 9.1). Spatial resolution is controlled by the separation between transects, the number of samples taken per meter, and the sampling capabilities of the instrument (see Clark 2000:158–164). Early surveys typically utilized 1-m transect separations and 1-m samples (1 measurement/m2). Current high-speed instrumentation allows greater sampling densities, with 1-m transect separation and 4 samples/m (4 measurements/m2) common for reconnaissance surveys and 0.5-m or even 0.25-m transect separations and 8–16 samples/m (16–64 measurements/m2) for high-precision surveys. A comparison of different sampling densities and their effects on data quality and interpretation is illustrated in Figure 9.2.1 Spatial sampling density determines the size of archaeological features that can be resolved, with a rule of thumb being that the interval between measurements should be no greater than half the size of the smallest feature to be detected in order that multiple measurements may delineate it (Weymouth 1986:347). It should be obvious, given this requirement, why resolving small portable artifacts and features like post holes is usually not possible (even if they exhibit a magnetic susceptibility contrast), while larger features like houses or ditches are frequently detected.

Causes of Magnetic Variation in the Archaeological Record Natural Processes From the foregoing we have seen that three natural processes generally are responsible for magnetic variations (once the effects of diurnal and related temporal changes are factored out). They are the following: 1. Differences in magnetic susceptibilities exist between various materials, deposits, and soils. 2. Topsoil becomes magnetically enriched owing to physical and chemical processes that include weathering and as a result of biogenic processes that include magnetotactic bacteria. 3. Firing increases soil magnetism, producing a thermoremanent effect. People live within their environment, generally on topsoil, and modify it and other deposits. Consequently, various human behaviors interact with these natural processes in multiple ways to produce a predictable set of culturally induced magnetic variations commonly seen in archaeological sites.

Magnetometry ~ 215

Figure 9.2. Increasing detail and quality of anomaly deďŹ nition occurs with greater sample densities over a pair of burned houses about 60 cm below the surface at the Menoken Village State Historic Site (32BL2), a Late Woodland (ca. A.D. 1240) fortiďŹ ed village in North Dakota. (a) 2-m samples; (b) 1-m samples; (c) 0.5-m samples; (d) 0.25-Ă&#x2014;-0.125-m samples (source: Kvamme 2003a).

Cultural Processes The formation of magnetic anomalies within archaeological sites can be attributed to seven cultural processes that result from the interaction between common human behaviors and the foregoing natural processes (Clark 2000; Scollar et al. 1990; Weymouth 1986).

216 ~ Kenneth L. Kvamme 1. People create fires. In sites of human occupation firing occurs in several forms. Fires for cooking and warmth occur in hearths, fireplaces, and ovens built within and outside of buildings (Figure 9.3). Technological production requires fires for creating ceramics and other crafts (glass, iron), generally in the form of specialized kilns or furnaces that typically are larger in size and in their magnitude of firing. Accidental firing also occurs in human occupations, as when a house or other structure burns; destructive fires have a similar effect, as when a village is razed in warfare. In all cases, the high temperatures that are produced introduce thermoremanent anomalies to the archaeological record (Figure 9.3). Additionally, fired materials become dispersed through settlements through time owing to hearth cleaning or to subsequent constructions that redistribute materials from earlier hearths or burned structures over a site. Fires are also frequently employed for clearing fields, forests, and brush and for controlling vermin (e.g., see Boyd 1999). All of these factors serve to increase the magnetic susceptibility of deposits (see point 3, below). 2. People make constructions and artifacts composed of fired materials. Brick, made of fired clay, is a common construction material that is employed in large amounts, creating prominent anomalies owing to the relatively large mass of fired material (Figure 9.4a). Ceramic vessels are also fired artifacts. Individual broken sherds may not be detectable owing to their small masses and the relatively low magnetic contrasts they

Figure 9.3. Intensive firing can create pronounced anomalies. (a) Burned houses about 60 cm below the surface at the Menoken Village State Historic Site (32BL2), a Late Woodland (ca. A.D. 1240) fortified village in North Dakota (source: Kvamme 2003a). (b) House with burned features verified by excavation about 1 m below the surface at Whistling Elk village (39HU242), a fortified settlement of the Initial Coalescent variant (ca. A.D. 1300) in South Dakota (source: Toom and Kvamme 2002). Key: 1, partially burned house; 2, completely burned house; 3, burned house perimeter; 4, centrally placed hearth; 5, burned post.

Magnetometry ~ 217 produce, but accumulations of ceramic sherds or whole vessels may sometimes be detected magnetically (Figure 9.4b). 3. Human occupations exacerbate magnetic enrichment of surface soils. As noted above (point 1), people introduce fired materials to surface soils through dispersal of hearth and other fired materials. Additionally, occupations introduce organic materials to the topsoil (food, waste products) that promote bacterial growth, including magnetotactic and other bacteria that concentrate magnetic compounds like magnetite (Fassbinder et al. 1990). The subtle rise in magnetic susceptibility that occurs is small but significant. Yet, because it is a settlement-wide phenomenon, it typically cannot be detected through conventional magnetometry, which is based on magnetic contrasts between a feature and the adjacent matrix (Clark 2000:101; Dabas and Tabbagh 2000:339). An active instrument such as a magnetic susceptibility meter, however, can directly measure the size of this effect (see Dalan, this volume, for further details). Nevertheless, magnetically enriched settlement soils are noteworthy in magnetometry surveys because concentrations that fill storage or house pits, for example (see point 4, below), can yield very large magnetic contrasts. 4. Human constructions accumulate topsoil. People build many constructed features using topsoil, which creates local increases in the magnetic field owing to the concentration of magnetically enriched material, particularly when settlement soils of higher

Figure 9.4. Fired artifacts can produce detectable anomalies. (a) The brick foundation of the Mount Comfort Church (3WA880; 1840s–1863), Arkansas, is visible among numerous iron-produced anomalies from stove parts, nails, and other artifacts about 35 cm below the surface, as revealed by excavation (Hilliard and Perigo 1994). (b) Excavations 25 cm below the surface in the indicated 2-m square revealed four adjacent ceramic vessels (each about 50 cm in diameter) that may be associated with strong magnetic monopoles in a heavily burned Caddo house (ca. A.D. 1400) at the Grandview site (3HE40), Arkansas (Source: Jami Lockhart, Arkansas Archeological Survey). Key: B, brick; I, iron; P, ceramic pots.

218 ~ Kenneth L. Kvamme magnetic susceptibility are employed (point 3). Small mounds for burials, platforms, or effigies may be mere accumulations of nearby topsoil, yet they may exhibit large magnetic contrasts (Figure 9.5a). Some cultures employ sod for building materials, such that rectangles of the grassy topsoil are employed to form components of walls, as did the Euroamerican pioneers in the American Great Plains in the nineteenth century. Similarly, the prehistoric occupants of Plains Village earthlodges in the same region mounded nearly a half meter of topsoil over the wooden frames of their dwellings; as this material eroded through the action of wind and rain a low ring of increased soil magnetism accrued about their perimeters (Figure 9.5b). When ditches are excavated, whether for simple drainage or as components of fortifications, the removed topsoil is commonly mounded adjacent to the trench to form raised berms and an enlarged magnetic contrast (Figure 9.5c). In turn, the ditches themselves may

Figure 9.5. Accumulations of topsoil associated with constructed features can cause positive anomalies. (a) Topsoil mounded to a height of about 1 m forming the Great Bear effigy (Mound 31; ca. A.D. 650–1300) at Effigy Mounds National Monument, Iowa (source: Kvamme 1999). (b) Eroded roof soil forms circular rings about the loci of three former earthlodges (with centrally placed hearths) about 50 cm below the surface at the Mandan/Arikara village (ca. 1822–1861) in the Fort Clark State Historic Site (32ME2), North Dakota (source: Kvamme 2003a). (c) Excavation showed topsoil mounded adjacent to the open fortification ditch about 10 cm below the surface at the Menoken Village State Historic Site (32BL2), a Late Woodland (ca. A.D. 1240) fortified village in North Dakota (source: Kvamme 2003a). (d) Excavation showed that surface soils completely fill a fortification ditch and bastions, obscuring them visually but not magnetically at the Double Ditch State Historic Site (32BL8), an ancestral Mandan village in North Dakota (ca. A.D. 1500–1780; source: Kvamme 2003c). (e) Soil coring verified the presence of food storage pit features filled with topsoil (with bottoms about 1.5 m below the surface) surrounding part of a rounded rectangular house at Huff Village State Historic Site (32MO11), a village of the Terminal Middle Missouri Variant (ca. A.D. 1450) in North Dakota (source: Kvamme 2003a, 2003b). Key: H, hearth.

Magnetometry ~ 219 eventually become filled with surface soils, forming a concentrated zone of high magnetism owing to the large volume of magnetically enriched material composing the fill (Figure 9.5d). Finally, small storage or other pit features generally become packed with topsoil upon their abandonment, making them readily detectable because the magnetically enriched topsoil in a vertical pit feature acts much like a bar magnet (Figure 9.5e). 5. Human constructions remove topsoil. Many types of constructions or constructed features require the excavation and removal of topsoil. The net result is typically a local lowering of the magnetic field over those features, or a “negative” anomaly, because the amount of overlying magnetic material is smaller than that in nearby areas. The digging of ditches, recessed house floors, subterranean storage pits, cellars, and even looters’ “pot holes” effectively removes small to large areas of magnetically enriched topsoil, causing negative contrasts. Additionally, simple incisions in the ground caused by foot or vehicular traffic along trails or roads can produce the same effect. These phenomena are well illustrated in Figure 9.6a, in which an unfilled excavation, a number of pot holes, incised cattle and road tracks, and the sides of an unfilled fortification ditch are expressed as negative anomalies. Occasionally, one sees a similar result in cemeteries where the sediments and soils were not replaced in their original order within a grave. In these situations the more magnetic topsoil might become

Figure 9.6. The removal of magnetically enriched topsoil can produce negative anomalies. (a) At the Menoken Village State Historic Site (32BL2), a Late Woodland (ca. A.D. 1240) fortified village in North Dakota, negative anomalies are seen at the loci of (1) an unfilled 1930s circular excavation, (2) unfilled pot holes, (3) incised vehicle tracks, (4) incised cattle trails, and (5) the slopes of a mostly unfilled fortification ditch (the bottom of the ditch is somewhat magnetic owing to its partial in-filling with eroded topsoil or pedogenesis) (source: Kvamme 2003a). (b) At the Confederate cemetery (founded in 1862) at the University of Mississippi, numerous interments are indicated as negative anomalies, most likely because topsoil was not replaced at the top of each grave (source: courtesy of Jay Johnson, University of Mississippi).

220 ~ Kenneth L. Kvamme deeply buried, causing the loci of graves to be expressed as negative anomalies relative to adjacent undisturbed ground with topsoil in place (Figure 9.6b). 6. People import stone and other materials for constructions. Buildings, building foundations, pavements, and floors are frequently constructed of stone. Some rocks might be more magnetic (e.g., igneous rocks) whereas others might be less magnetic (e.g., many limestones) than surrounding soils, thereby creating magnetic contrasts (see a common table of rock magnetic susceptibilities, as in Mussett and Kahn 2000:173). Foundation blocks made of a somewhat magnetic sandstone yielded the positive anomalies seen in Figure 9.7a, while buried stone walls of a nonmagnetic limestone are expressed as negative anomalies in Figure 9.7b. Specialized sediments might also be imported to a site for certain desirable characteristics, such as clay for prepared floors or a pit lining or sand or gravel for trails, walkways, or use as a base substrate for larger constructions. These materials, too, can produce detectable contrasts depending on their inherent magnetic susceptibilities, volume, and depth.

Figure 9.7. Imported stone can be more or less magnetic than surrounding soil, introducing significant anomalies in either case. (a) The outlines of principal rooms, room blocks, and corner bastions (in the lower left and upper right) are revealed by a highly magnetic sandstone used as foundation blocks, about 50 cm below the surface as indicated by excavations at the Fort Clark Trading Post (1831–1861), an outpost of the American fur trade at the Fort Clark State Historic Site (32ME2), North Dakota (source: Kvamme 2003b). (b) Iron-free limestone blocks used for construction, shown by excavation to occur nearly a meter below the surface, produce negative anomalies compared to the sandy soils about them at the Roman city of Empuriés (ca. A.D. 100–300), in northeastern Spain (these total field data were acquired at a sampling density of only 1 m using a proton precession magnetometer and corrected by differencing from base station readings).

Magnetometry ~ 221 7. People make iron artifacts. Iron artifacts were first introduced to the Americas by Europeans in the historic period (aside from a few specimens of meteoric origin), although an Iron Age exists much earlier in other parts of the globe. Iron artifacts tend to be readily detected by magnetometry depending on their size, shape, orientation, mass, and depth below the surface (Figure 9.8). Iron markedly alters the earth’s magnetic field, producing large measurements commonly expressed as dipolar anomalies (see above). While the dipolar “peaks” will frequently line up along the angle of declination, pointing to magnetic north (with the negative pole lying to the north in the northern hemisphere), this is not necessarily the case if the source is irregularly shaped (nonspherical), possessing a principal axis that can cause the poles to align in any direction. It should be noted that iron or steel artifacts of recent origin—the debris that litters so much of the earth (e.g., tractor parts, trash)—generally constitute a nuisance or noise factor in prehistoric sites, because subtle soil and other changes (points 1–6) tend to be obscured by the stronger magnetic anomalies that these artifacts generate (see below). On the other hand, that very iron litter may represent targets of interest at a historic period site, pointing to significant artifacts, dumping areas, or the loci of former wooden structures that employed nails in their construction.

Limits of Magnetometry Archaeological Detection Probabilities The likelihood of a buried archaeological feature being detected by magnetometry depends on a complex interaction among several factors: (1) the amount of magnetic contrast between the feature and surrounding deposits, (2) the size of the feature rela-

Figure 9.8. Iron and steel artifacts produce strong anomalies readily detected by magnetometry. (a) Characteristic dipolar anomalies generated largely by Civil War artillery projectile fragments lying immediately below the surface as shown by excavations at the Prairie Grove Battlefield State Park, Arkansas (December 7, 1862). (b) A ring of dipolar anomalies indicative of many large iron artifacts about 50 cm below the surface surrounding the perimeter of an earthlodge at the Mandan/ Arikara village (ca. 1822–1861) in the Fort Clark State Historic Site (32ME2), North Dakota (source: Kvamme 2003a). Key: H, hearth.

222 ~ Kenneth L. Kvamme tive to the sampling density of the measurements, (3) the depth of the feature below the surface, (4) the level of confounding noise factors (soil disturbances, metallic debris) that might overlie it, (5) the degree of regular pattern that the feature exhibits, and (6) instrument sensitivity and the quality of data acquisition. In general, small portable artifacts are difficult to detect. An exception is iron or steel artifacts, which are readily sensed (Figure 9.8), but these pertain only to historic period occupations in the Americas. Large masses of ceramic artifacts might occasionally be revealed by magnetometry, as when a concentration of sherds or a large pot immediately below the surface subtly raises the local magnetic field (owing to their firing; Figure 9.4b). Buildings and other constructions are much more readily located because their larger sizes and masses of contrasting physical properties increase detection probabilities relative to common instrument resolutions, sampling densities, and field methods. Moreover, their regular geometric shapes (lines, circles, rectangles) make them more easily recognized as cultural anomalies than unpatterned ones that can result from biogenic activity (tree throws, animal dens). Deeply buried archaeological features are more difficult to detect than shallowly buried ones because the soils above may act like a filter that degrades the signal of what lies beneath and because of inherent depth limitations to magnetometry (see below). The density of metallic litter on the surface, agricultural practices (plowing), modern constructions (fence and power lines, pipes), and biological phenomena (rodent burrows) all introduce noise to remote sensing data sets that complicates recognition and interpretation of possible cultural anomalies. Furthermore, the complexity of archaeological deposits can make anomaly identification and interpretation difficult: intensive occupations with dense cultural stratigraphy or superimposed constructions can “jumble” the signals that might be obtained, making patterns unclear. Maximum Depth of Detection The depth of anomaly detection in magnetometry depends on the magnetic susceptibility contrast of the materials being sensed. A typical rule of thumb is that magnetometry is usually confined to the uppermost 1–2 m for most soil features in archaeological sites, with a practical limit of about 3 m (Clark 1990:78–80; large burned or iron masses may be detected at much greater depths). The depth criterion can be examined more exactly utilizing the falloff with the third power of distance rule. If a small anomaly measures 1 nT at 1 m below the sensor, an identical source at 2 m below the sensor would measure 1/23 = 1/8 = 0.125 nT, which would be barely detectable with an instrument that offers a conventional 0.1-nT precision. Similarly, a series of identical hearths lying at 1-m intervals below the instrument might yield 10 nT at 1 m, with predictable measurements of 1.25 nT at 2 m, 0.37 nT at 3 m, and an almost undetectable 0.15 nT at 4 m. It is emphasized that many anomalies stemming from small variations in their magnetic susceptibilities lie within the 1–3 nT range and so become undetectable at large depths. The reader is reminded that

Magnetometry ~ 223 magnetic gradiometers are somewhat less sensitive than total field instruments and therefore cannot detect anomalies as small or as deep as those detectable by the latter instruments (see above). Estimating Anomaly Depth There are a number of ways to estimate the depth of individual anomalies seen in magnetometry data sets. The most prevalent in archaeology makes use of the halfwidth rule, which states that the diameter of an anomaly at half its maximum value is approximately equal to its distance below the sensor (bear in mind that the sensor may be 30–50 cm above the ground surface). Bevan (1998:24) notes that such estimates may be somewhat large because the method assumes a compact spherical source when in fact an actual target may be spread out or flattened. Aitken (1970:687) asserts that at midlatitudes (with inclination angles of 60–70 degrees) the reverse or negative pole of an anomaly will lie to the north of the source at a distance approximately equal to its depth. Estimating Anomaly Location At the midlatitudes of the sub-Canadian United States and of much of Europe the inclination of the earth’s magnetic field varies from about 55 to 75 degrees (Mussett and Kahn 2000:139–142). At these latitudes the maximum of a normal dipolar anomaly will actually lie to the south of the target’s true locus by about one-third of the source-to-sensor distance (Weymouth 1986:344). Obviously, this discrepancy will be small for shallowly buried objects, but it can be a significant factor when locating more deeply buried features. This normal north–south relationship is disturbed by highly magnetic sources with elongated axes oriented in some other direction.

Problems and Issues in Magnetometry Surveys One is faced with a number of issues when contemplating conduct of a magnetometry survey. The largest is the nature of the site. Variables to consider include the nature of the soils (whether there will be sufficient magnetic contrasts); the presence of highly magnetic igneous outcrops or bedrock; the presence of many iron artifacts at a historic site (where results might simply present a mix of overlapping dipolar anomalies of extreme size); location of the site in an urban setting with a general litter of iron or steel; proximity of electromagnetic fields and passing or parked automobiles; the presence of tourists and dogs (both bearers of significant metal); whether the ground is in flat pasture or parkland or is heavily vegetated, forested, undulating, or dissected (affecting survey capabilities); and whether previous archaeological work has been conducted (archaeologists are notorious depositors of ferrous metals). Given these potential site variables, the practitioner has more control over the specific details of the magnetometry survey itself, but variation can also be intro-

224 ~ Kenneth L. Kvamme duced here. Most of this variation is derived from the operator, relating to magnetic cleanliness and instrument handling skills. Other variation comes from the instrument itself in the inherent noise it introduces to the data from its components or tendency to drift. These issues are discussed in the following sections and in Chapter 10 of this volume. Level of Magnetic Contrast Anomalies in magnetometry can occur only when there is a significant magnetic contrast between a feature and the surrounding matrix. In some soils the presence of iron compounds may be extremely low or nonexistent such that detectable magnetic contrasts may not form. Soil development may be insufficient on a site to allow some of the natural and culturally induced variations described earlier. An A horizon may be absent in some areas or soils may be poorly developed, as in certain regions of the Southwest where magnetic anomalies tend to be of extremely low magnitude, for example. In volcanic soils magnetism may be uniformly high such that contrasts cannot be seen. Some archaeological features may be no more magnetic than the soils in which they lie; for example, adobe blocks possess the same magnetic susceptibility as the soils from which they came and sandstone blocks may be no more magnetic than surrounding sands. Some practitioners advocate testing the magnetic susceptibilities of soil samples (see Dalan, this volume, for further details) prior to a survey to ensure that significant contrasts exist (e.g., David 1995), but aside from the rather unusual settings listed here my experience suggests that adequate contrasts exist in most regions of North America for anomalies to be detected with instrumentation capable of recording at the sub-nanotesla level. Forested Landscapes and Rough Terrain Heavily vegetated or wooded landscapes, as well as rough, sloping, or undulating terrain, make instrument passage difficult and can greatly slow down the progress of a magnetometry survey. While current instrumentation is able to collect data very rapidly, one must be able to carry the instrument at a sufficient pace to make use of that speed. Moreover, to ensure correct spatial positioning, instruments must be moved along linear transects, a clear difficulty in wooded settings, for example. All instruments can record data in a “manual mode,” however, whereby the operator presses a button to acquire readings at desired locations, so given enough time instruments may be properly placed and readings acquired in nearly any setting, but survey speed and area of coverage will suffer commensurately. Igneous Rocks Native igneous rocks, particularly recent basalts, can be extremely magnetic owing to remanent magnetism that can persist for millions of years (Mussett and Kahn

Magnetometry ~ 225 2000:172). They can profoundly affect the quality of magnetometry surveys or even preclude them. Large dipolar anomalies can result with orientations significantly different from that of the earth’s current field (Figure 9.9). Nevertheless, if sufficient overlying sediments exist, magnetometry surveys and useful results may be possible, as long as the surveys remain some distance from outcrops of such rocks. The overlying sediments and soils serve to mask the bedrock’s effect by increasing the sensor distance such that variations in the overburden can be recorded. Iron Bodies and Electromagnetic Fields Buried pipes, electrical lines, metal fences, automobiles, or visitors with excessive iron or steel artifacts on their bodies (e.g., cameras, glasses, shoes with steel shanks) negatively impact magnetometry surveys owing to the large anomalies they introduce, some of which can encompasses many meters. One should avoid working in proximity to iron objects at all costs owing to their large impact on magnetic data sets (Figure 9.10). Signage, trash cans, fencing, and the like may sometimes be temporarily removed during a survey. Urban Settings Urban settings are plagued with larger concentrations of metallic debris and rubbish, iron and steel fencing, buried pipes, lampposts, signage, electromagnetic fields, people with their metallic adornments and cell phones, passing and parked automobiles, and the like, all to the point where useful survey results might be impossible. On top of these factors, given that urban landscapes frequently have undergone extensive and intensive reworking, discerning patterns associated with culturally produced anomalies can be extremely challenging and even unproductive in the complex deposits that can occur in these settings. Archaeological Practices The common archaeological field practice of using steel nails, rebar, and particularly pin flags as semipermanent or temporary markers within archaeological sites produces a large and lasting negative impact

Figure 9.9. A magnetic survey in the vicinity of granite boulders in the Rocky Mountains of southern Colorado produced large-magnitude dipolar anomalies stemming from remanent magnetism. Note the variable dipolar alignments.

226 ~ Kenneth L. Kvamme

Figure 9.10. Large iron or steel bodies on a site can introduce massive anomalies. (a) A buried steel pipe with each joint yielding a dipolar anomaly. (b) The effects of a steel post fence line lying 2 m to the west of the indicated anomalies; each positive-negative pair represents a single post.

for magnetometry and other geophysical surveys (Figure 9.11). Some pin flags and nails are invariably lost or even purposely left behind after a project, precluding subsequent magnetic surveys within a meter or more of their locations, depending on their sizes (a 2-m diameter can be affected by steel pin flags and as much as 8 m by rebar). Their use must be discouraged and considered unethical conduct by the profession (discussion of this issue is currently under way among ethics committees in the Society for American Archaeology and the Society for Historical Archaeology). Nylon-post pin flags are commercially available, for example, and wood, plastic, nylon, aluminum, or brass posts or stakes can work equally well as datums. All are invisible in magnetometry data sets, leaving no detectable anomalies (it is emphasized that, while such nonferrous metals as aluminum and brass are invisible to magnetometry, metals of any kind impact other geophysical instruments, such as electromagnetic induction meters and ground-penetrating radar, so metal markers of any kind should be avoided). Magnetic Cleanliness Operator

Because magnetometers are so sensitive, the instrument operator must be absolutely free of ferrous metals of any kind (jewelry, glasses, zippers, rivets in jeans, underwire bras, grommets, buttons on hats, and shanks in shoes), of electrical fields (battery-powered wristwatches), and of magnetic fields (magnetic strips exist on

Magnetometry ~ 227

Figure 9.11. Most of the dipolar anomalies at this important historic site in South Carolina represent steel-wire pin flags cut after a mowing, present despite previous “cleaning” of the area with a metal detector (some are indicated with an arrow). Their large number served to reduce the utility of this survey, as did multiple rebar datums (labeled R), each of which obscured other magnetic variations within a 3–5 m radius.

credit cards). This is not a trivial concern, because many a magnetic survey has been compromised by the inadvertent presence of iron or steel artifacts on the operator’s body (Figure 9.12). The necessary safe course of action is to require each operator, even one assumed to be magnetically “clean,” to undergo a simple field test (one is frequently surprised by the hidden grommet, paper clip, or zipper). With the magnetometer parked in a stationary position, the operator should pass feet, legs, hips, arms, torso, and head adjacent to a sensor. Only when no variation in readings occurs should the survey commence. An added precaution is that site visitors, including dogs, should be kept from approaching the operator during a survey owing to the invariable presence of ferrous metals on their bodies; a “safe” distance depends on the instrument type,

228 ~ Kenneth L. Kvamme with 5 m probably appropriate for vertical gradiometers and 10 m for total field instruments. Site

Just as an instrument operator can be “cleaned,” so too can a site. Recent metallic litter of iron or steel can degrade or even destroy the quality of magnetometry surveys. A large nail will produce a pronounced dipolar anomaly 0.5 m in width; a piece of steel wire a half meter in length may create a dipolar anomaly 3 m in diameter. A large number of such anomalies can greatly obscure more subtle ones stemming from Figure 9.12. The instrument operator forgot potential archaeological features of interest to remove keys from his pocket, resulting in a total loss of data in one segment of a survey. (Figure 9.13). Weymouth (1986:346) adThe striping is a result of the keys passing by vocates identifying and removing offending materials prior to a survey through use of a the sensors with each step. metal detector; the magnetometer itself may also be employed for this purpose in a quick “scan” mode. Of course, on sites of the historic period, this very litter may represent targets of interest, so caution must be employed to remove only recent surface iron.

Magnetic Surveys as Landscape Archaeology Magnetometry surveys are relatively fast, have good sensitivity and sampling density, and respond to a myriad of anthropogenic conditions in the archaeological record, which makes magnetometry an ideal prospecting tool. It can produce imagery of subsurface cultural features over large areas, often in great detail. From them, accurate plan maps of many aspects of past settlements can be generated, with arrangements of dwellings or other structures, rooms within them, internal room features, streets, lanes, trails, public spaces, dumping areas, middens, mounds, gardens, storage features, fortification systems, and other elements indicated. This capability has profound implications for archaeology (Kvamme 2003b). Besides serving as a mere discovery tool that can point archaeologists to significant cultural features, thereby reducing fieldwork costs, magnetometry surveys allow the direct study of settlement form and content through the analysis of the imagery alone, as primary data. Such information can be suitable for the study of site structure, content, and organization; for examining spatial patterns and relationships; and for directly confronting specific questions about a site and the past. With large-area surveys, space can be viewed in terms of hectares as opposed to the square meters typical of archaeological excavations, potentially allowing

Magnetometry ~ 229

Figure 9.13. Nearly all of the indicated dipolar anomalies at Menoken Village State Historic Site (32BL2), a Late Woodland (ca. A.D. 1240) fortified village in North Dakota, were found to be clusters of steel bottle caps and beer cans. A large group of them obscured other magnetic patterns within the house labeled 1, which turned out after excavation to look much like the magnetically outlined house labeled 2 (source: Kvamme 2003a).

entire cultural landscapes to be visualized and analyzed. Large surveys are now routinely conducted, including complete coverage (78 ha) of the Roman city of Wroxeter, England (Gaffney et al. 2000), and an even larger region at the Iron Age citadel of Kerkenes Dag, Turkey (Summers et al. 1996), for example. Compared with traditional archaeological results confined to a small number of test pits, block excavations, or surface collections, more information can potentially be produced from magnetic surveys about sitewide aspects of layout, content, and organization, simply because areas of the subsurface “exposed” can be increased by several orders of magnitude (Figure 9.14). We might ultimately imagine the compilation of “libraries” of settlement plans derived from magnetometry surveys that will allow systematic and comparative studies of settlement form, structure, and content, for example. Such capabilities will potentially engen-

230 ~ Kenneth L. Kvamme

Figure 9.14. The spatial organization of the Fort Clark Trading Post (1831–1861) and its environs at the Fort Clark State Historic Site (32ME2), North Dakota, as revealed by magnetometry (source: Kvamme 2003b). (a) The trading post showing room blocks, corner bastions, and interior compound; (b) the circular earthlodge and surrounding garden compound of Pierre Garreau, a native Arikara (who adopted his stepfather’s name) and trading post interpreter; (c) footprint of outlying structure; (d) dumping ground; (e) earth-borrowing pits; (f ) Euroamerican cemetery.

der new and richer understandings of human uses of space and past interactions with the environment, compared with mindsets now constrained to the few square meters typical of excavation projects. Magnetometry and other remote sensing surveys capable of imaging large areas may ultimately promote development of an entirely new landscape perspective for American archaeology.

Note 1. Unless otherwise specified, all illustrated data sets were acquired with a Geoscan Research FM36 fluxgate gradiometer at sampling densities of 8 or 16 measurements/m2.

Acknowledgments I wish to thank Jay Johnson for inviting me to participate in this volume and the associated workshops; all is much appreciated. This chapter benefited enormously from conversations with and outright instruction from Lew Somers, Bruce Bevan, and Rinita Dalan. I am particularly grateful to Jay Johnson and Jami Lockhart for allowing me use of data sets that helped to best illustrate a number of magnetic principles. Work at sev-

Magnetometry ~ 231 eral of the sites illustrated here (Double Ditch, Fort Clark, Huff, Menoken) was supported by grants from Stan Ahler and the PaleoCultural Research Group of Flagstaff, Arizona, and Fern Swenson of the State Historical Society of North Dakota. Research at Whistling Elk was supported by a grant from the National Center for Preservation Technology and Training, National Park Service. Results at Effigy Mounds were acquired during a National Park Service training workshop. Findings at Empuriés, Spain, were obtained in the Boston University archaeological field school of 1996. Finally, I wish to thank all of the students who assisted in these projects and particularly Jo Ann, Charles, Kristina, and Emily Kvamme who participated in most of them.

References Cited Aitken, M. J. 1970 Magnetic Location. In Science in Archaeology, edited by D. Brothwell and E. Higgs, pp. 681–694. Praeger, New York. 1974 Physics and Archaeology. 2nd ed. Clarendon Press, Oxford. Aitken, M. J., G. Webster, and A. Rees 1958 Magnetic Prospecting. Antiquity 32:270–271. Alldred, J. C. 1964 A Fluxgate Gradiometer for Archaeological Surveying. Archaeometry 7:14–19. Avery, T. E., and G. L. Berlin 1992 Fundamentals of Remote Sensing and Airphoto Interpretation. 5th ed. Macmillan, New York. Becker, H. 1995 From Nanotesla to Picotesla—New Window for Magnetic Prospecting in Archaeology. Archaeological Prospection 2:217–228. Bevan, B. W. 1998 Geophysical Exploration for Archaeology: An Introduction to Geophysical Exploration. Special Report No. 1. U.S. Department of the Interior, National Park Service, Midwest Archeological Center, Lincoln, Nebraska. Boyd, R. T. (editor) 1999 Indians, Fire, and the Land. Oregon State University Press, Corvallis. Campbell, W. H. 1997 Introduction to Geomagnetic Fields. Cambridge University Press, Cambridge.

232 ~ Kenneth L. Kvamme Clark, A. J. 2000 Seeing Beneath the Soil: Prospecting Methods in Archaeology. Reprinted. Routledge, London. Originally published 1990, B. T. Batsford, London. Dabas, M., and A. Tabbagh 2000 Magnetic Prospecting. In Archaeological Method and Theory: An Encyclopedia, edited by Linda Ellis, pp. 335–339. Garland, New York. David, A. 1995 Geophysical Survey in Archaeological Field Evaluation. Ancient Monuments Laboratory, English Heritage Society, London. Eighmy, J. L., and R. S. Sternberg (editors) 1990 Archaeomagnetic Dating. University of Arizona Press, Tucson. Fassbinder, J., H. Stanjek, and H. Vali 1990 Occurrence of Magnetic Bacteria in Soil. Nature 343:161–163. Gaffney, C., J. A. Gater, P. Linford, V. Gaffney, and R. White 2000 Large-Scale Systematic Fluxgate Gradiometry at the Roman City of Wroxeter. Archaeological Prospection 7:81–99. Hilliard, J., and M. Perigo 1994 Archeology Reveals Brick Church at Mount Comfort. Flashback 44:24–38. Published by the Washington County Historical Society, Fayetteville, Arkansas. Kvamme, K. L. 1999 Archaeo-Geophysical Surveys at Effigy Mounds National Monument, Iowa. Submitted to the Midwest Archeological Center, National Park Service, Lincoln, Nebraska. 2003a Multidimensional Prospecting in North American Great Plains Village Sites. Archaeological Prospection 10:131–142. 2003b Geophysical Surveys as Landscape Archaeology. American Antiquity 68(3):435–458. 2003c Geophysical Findings at the Double Ditch State Historic Site (32BL8), North Dakota, 2002. Submitted to the State Historical Society of North Dakota, Bismarck. Le Borgne, E. 1955 Susceptibilité magnétique anormale du sol superficiel. Annales de Géophysique 11:399–419. Mussett, A. E., and M. A. Khan 2000 Looking into the Earth: An Introduction to Geological Geophysics. Cambridge University Press, Cambridge.

Magnetometry ~ 233 Scollar, I., and F. Krückeberg 1966 Computer Treatment of Magnetic Measurements from Archaeological Sites. Archaeometry 9:61–71. Scollar, I., A. Tabbagh, A. Hesse, and I. Herzog 1990 Archaeological Prospecting and Remote Sensing. Topics in Remote Sensing, No. 2, G. Hunt and M. Rycroft, series editors. Cambridge University Press, Cambridge. Summers, G. D., M. E. F. Summers, N. Baturayoglu, Ö. Harmansah, and E. McIntosh 1996 The Kerkenes Dag Survey: An Interim Report. Anatolian Studies 46:201–234. Tite, M. S. 1972 Methods of Physical Examination in Archaeology. Seminar Press, New York. Toom, D. L., and K. L. Kvamme 2002 The “Big House” at Whistling Elk Village (39HU242): Geophysical Findings and Archaeological Truths. Plains Anthropologist 47:5–16. Weymouth, J. W. 1976 A Magnetic Survey of the Walth Bay Site. Occasional Studies in Anthropology No. 3. Midwest Archeological Center, Lincoln, Nebraska. 1986 Geophysical Methods of Archaeological Site Surveying. In Advances in Archaeological Method and Theory, vol. 9, edited by M. B. Schiffer, pp. 311–395. Academic Press, New York. Weymouth, J. W., and Lessard, Y. A. 1986 Simulation Studies of Diurnal Corrections for Magnetic Prospection. Prospezioni Archeologiche 10:37–47.


Data Processing and Presentation Kenneth L. Kvamme

The computer processing of magnetometry data is an activity that is nearly as important as collecting the raw data. It is also an essential activity given the volumes of information collected. Surveys with current instrumentation routinely approach coverage of from one-half to 1 ha per day, at sampling densities ranging from 4 to 60 measurements per square meter, meaning that tens of thousands of measurements are commonly acquired. The only way to manage and use this volume of information is through computers. Moreover, correct data processing allows subtle patterns in data sets to be enhanced and made clear, while defects and other forms of noise can be removed or reduced (Ciminale and Loddo 2001; Kvamme 2001; Scollar et al. 1990:126–204, 488–506). It is frequently possible to produce imagery of cultural patterns with sufficient clarity that the specialist and nonspecialist alike can instantly understand and interpret results, as when the floor plan of a house or the layout of an entire settlement is clearly expressed (e.g., as in Clay 2001; Gaffney et al. 2000; Toom and Kvamme 2002). In other words, the results of magnetometry surveys can now look much like the buried archaeology thanks to computer graphics (e.g., see Figure 9.14 in Kvamme, this volume, Chapter 9). This outcome is probably the single factor most responsible for the large acceptance of magnetometry surveys in Europe and their growing use in North America (Kvamme 2003a).

236 ~ Kenneth L. Kvamme Ironically, these very changes have led to fundamentally new goals in geophysical surveys and the realization of new benefits. Early work, which acquired comparatively few measurements in relatively small areas, had a tendency to be much more deductive with a focus on explaining and interpreting each anomaly. Modern work simply yields too much information to examine it in this manner; specific “deductions” can only be applied to a limited number of representative anomalies or ones of special interest. Conclusions reached might then be applied to other anomalies of similar form, a very inductive process (see Gaffney et al. 2000). A more common approach today typically (1) collects high-density data over as large an area as possible, (2) computer processes the data to clarify regular, culturally formed patterns, and (3) utilizes pattern-recognition principles as an aid in the interpretation process. In general, anomalies exhibiting regular geometric shapes (lines, circles, squares, rectangles) tend to be of human origin (see Chapter 9). The following sections discuss data-processing methods commonly applied to magnetometry data sets. The computer processing of magnetometry data is an essential activity that carries many benefits; it can also be dangerous. Insufficient processing can leave important features unseen while improper methods might actually remove significant features from the data. On the other end of the continuum, data can be overprocessed such that spurious features are introduced. Any data manipulation should always proceed with great care. An excellent precaution is that after each processing step the result should be differenced from the previous state of the data to allow examination of any changes that might have been made, including the removal of important features. Procedurally, data from various survey blocks in a project are concatenated to make a single composite image. Noise and defects are then removed, followed by application of enhancements that might highlight or exaggerate features of interest. One benefit of the ongoing computer revolution is that initial data processing may now be pursued on-site with portable field computers. This gives an important and nearly immediate link with the data that can influence survey decisions by guiding work to areas of greater archaeological interest and potential. To understand the nature of the anomalies in the following examples the reader should review Chapter 9 of this volume. Much of the presentation is influenced by familiarity with algorithms offered in Geoplot software (Geoscan Research 2000), a prominent archaeogeophysical data-processing software package, but other techniques common to geographic information systems (GIS) and image processing are also examined (Kvamme 1999a, 2001; Lillesand and Kiefer 1994). The examples were produced using Geoplot 3.0 (Geoscan Research), Idrisi-Kilimanjaro (Clark Labs, Clark University), and Surfer 8 (Golden Software). Unless otherwise specified, all magnetic data sets were acquired with a Geoscan Research FM36 fluxgate gradiometer.

Noise and Defects Results that define cultural features in geophysical surveys are the object of the survey, referred to as the “signal,” while “noise” refers to everything else that is measured

Data Processing and Presentation ~ 237 and that obscures the targeted features. Frequently, one project’s signal may be another’s noise, making this process somewhat confusing. For example, iron debris may be a nuisance to the measurement of subtle soil changes in prehistoric sites, but that very litter may represent targets of interest at a historic period site, pointing to significant artifacts. Noise and signal therefore become relative concerns dependent on survey goals. Noise derives from three principle sources: from within the soils and deposits of the site surveyed, from the instrument itself, and from “survey defects” that stem from how a magnetometer is handled in the field by the operator. Noise Due to the Nature of a Site’s Soils, Deposits, and Context Site noise can arise from such natural disturbances as excessive rodent work, badger or coyote dens, tree throws, and the like. Geological disturbances such as buried paleochannels can produce extensive anomalies as can the archaeological record itself, as when numerous and dense cultural features in higher levels make anomaly detection at deeper targeted levels difficult to visualize. Modern cultural disturbances such as pavements, landscaping, buried pipelines, nearby electrical fields, metallic debris, passing or parked automobiles, and plowing also introduce numerous unwanted anomalies and can be particularly bad in urban contexts, where more noise factors are likely. In general, it is difficult or impossible to remove the effects of large features like paleochannels, pipes, pavements, or tree throws from imagery; they are part of the record. Steps may be taken to eliminate or reduce the effects of smaller or periodic noise factors, however. Data spikes refer to isolated extreme measurements that typically stem from the presence of ferrous metal artifacts in magnetic data sets (Figure 10.1a). De-spiking al-

Figure 10.1. Magnetometry data from Primeau’s Trading Post (1850s–1861) at the Fort Clark State Historic Site (32ME2), North Dakota. (a) The site contains a littering of iron objects characteristic of historic sites of this period. (b) A de-spiking algorithm removes many of the magnetic “point” anomalies caused by iron artifacts, somewhat improving visualization of the principal room blocks and interior compound. (c) Difference between images a and b showing features removed (source: Kvamme 2002).

238 ~ Kenneth L. Kvamme gorithms remove isolated extreme measurements, typically by replacing them with the average of neighboring measurements (Figure 10.1b, c). Plow marks are frequently visible in magnetic data because the ridges and furrows represent greater and lesser amounts, respectively, of magnetically enriched topsoil (Figure 10.2a). This type of noise may be removed using Fourier methods (Figure 10.2b, c) owing to their regular orientations and periodic nature (Kvamme 2003b). Instrument Noise Instrument noise is inherent to any electronic device and arises from the quality of its electrical components, the nature of the sensors, and how well it might be calibrated or tuned. Such noise is manifested in a number of ways, and a variety of processing steps are designed to address them. Drift

Some instruments “drift,” meaning that their zero point fluctuates through time owing to perturbations in their electronics and other factors such as temperature (Figure 10.3a, particularly grids 4–6). The slope of the trend in the measurements in each transect can be removed by fitting a least-squares line (after data spikes are removed) and working with the computed residuals (Figure 10.3b). The latter are zero-centered (i.e., their mean is zero), which effectively removes any drift or change in mean value between the individual transects (illustrated in Figure 10.3c). Heading Errors

Some instruments, such as fluxgate gradiometers, yield minor “heading errors” that create a slight shift in the average measured value with minor changes in the direction of their facing (seen as stripes in Figure 10.3a). This invariably occurs as a survey zigzags through a site, despite operator attempts at constant orientation. By normalizing the data such that the mean values of each transect are made equal (they are usually zeroed), such heading errors can be eliminated (Figure 10.3b, c; note that the mean of gradiometry data is theoretically close to zero, so this procedure makes theoretical “sense”). Obviously, this form of processing is closely related to drift corrections. A danger in transect normalization occurs when linear archaeological features coincide with the transect direction (e.g., walls, trails). When the data are normalized, such features can be removed or reduced in the outcome. Care must therefore be taken to ensure that significant features are not removed; if they are, such normalizations should be avoided (or other more specialized methods carried out). ”Random” Perturbations

Ideally, if a magnetometer is “parked” in one place it should return the same measurement again and again (assuming a gradiometer configuration or that diurnal

Data Processing and Presentation ~ 239

Figure 10.2. Magnetometry data from an early Archaic occupation (ca. 7000â&#x20AC;&#x201C;10,000 B.P.) at the Wallace Bottom site (3NW50), Arkansas. (a) Magnetically revealed plow marks from mule-drawn plowing of the nineteenth century partially obscure anomalies associated with the Archaic occupation. (b) Fourier methods successfully remove or reduce their presence, allowing improved visualization of other anomalies. (c) DiďŹ&#x20AC;erence between images a and b showing features removed.

240 ~ Kenneth L. Kvamme

Figure 10.3. A complete magnetic processing sequence illustrating many required operations. The data portray multiple bastions, fortification trenches, trails, subterranean storage pits, and other features from the Double Ditch State Historic Site (32BL8; ca. A.D. 1500–1780), a fortified ancestral Mandan earthlodge village in North Dakota (source: Kvamme 2003b). (a) Raw data from six 20-m grid squares illustrating drift and heading errors. (b) Data after “zeroing” the transects; note the staggering defect resulting from operator errors in grids 3–6. (c) Difference between images a and b showing magnitude of drift and heading errors. (d) Data after application of a “de-staggering” algorithm; note the subtle horizontal “bands” now visible, particularly in grids 3–6, that result from sensor height variations caused by the operator’s gait. (e) Data after removal of the gait defect through Fourier methods; note that some high-frequency noise remains. (f ) Difference between images d and e showing the subtle and periodic nature of the operator’s gait imbedded within the data. (g) Data after low pass filtering to further reduce image noise. (h) Difference between images e and g showing high-frequency components removed. (i) Data after interpolation from the 0.5 × 0.25 m of image g to 0.25 × 0.25 m, yielding a smoother, more continuous-looking result.

changes in the magnetic field are removed). This is not usually the case, however, because small variations in measurements will inevitably occur. Fluxgate gradiometers can be “tuned” periodically to reduce the magnitude of this effect, but it nevertheless remains (in Figure 10.3b it is obvious that the instrument was better tuned in grids 1 and 2 than in grids 3–6, which appear noisier). The high sampling rates of current instrumentation can sometimes be employed to average or “stack” measurements to

Data Processing and Presentation ~ 241 reduce this source of variation, owing to the well-known statistical theorem that the standard error of the data is a function of the number of measurements: s.e. = σ/√n, where σ refers to the standard deviation of the instrument error distribution (generally unknown) and n is the number of averaged measurements (obviously, with only four stacked measurements this error can be halved; see Hays 1994:214). Stacking should ideally be applied during data collection, but it may also be undertaken in the dataprocessing phase, provided that a large number of samples per meter are acquired. If, for example, 16 samples/meter are obtained in the field, then one may average up to 4 measurements and still retain a respectable 0.25-m spatial resolution, while reducing the standard error by half. The result can be a smoother, less noisy data set. Noise Due to Instrument Handling An experienced surveyor is particularly important in magnetic surveys. The operator must be absolutely free of ferrous metals of any kind. Moreover, he or she must be cognizant of heading errors, drift, and random perturbations that can be partially corrected by frequent tuning with some instruments. A uniform walking speed is necessary to ensure a constant number of samples per meter and accurate spatial location of the measurements. A constant instrument height above the ground surface must also be maintained owing to the falloff of the magnetic field with the third power of distance. Yet, even with experienced surveyors, defects commonly arise from instrument handling, and a variety of data-processing algorithms exist for their correction. Staggering

Some instruments work on timing systems, in which a fixed number of measurements are acquired each second (e.g., 4, 8, 10). The operator must match a metronome-like signal produced by the instrument with meter marks on the ground (controlled by tapes placed adjacent to transects, see Chapter 9), or simply walk at a constant pace, to ensure that the measurements are accurately located. If timing is off by even a small amount in zig-zag surveys, linear and other anomalies become staggered, causing a characteristic “herringbone” effect (Figure 10.3b). This defect is normally corrected by digitally “sliding” every other transect a small amount until the stagger is removed (Figure 10.3d). Gait

With high instrument sensitivity and sampling rates, a surveyor’s gait can become visible in the data as a periodic defect. This phenomenon is due to regular variations in instrument distance from the surface that correlate with the operator’s walking movements. Because magnetic field strength falls off with the third power of distance, changes in instrument height of only a few centimeters with each step can introduce a regular sine wave to the data on the order of 0.1–1.0 nT, unless great care is taken. While invisible in data sets of high dynamic range, this defect can become pronounced in magnetically quiet sites with anomalies at the sub-nanotesla level. This is the case in

242 ~ Kenneth L. Kvamme Figure 10.3d, in which the operator’s gait can be seen as a subtle horizontal banding in grids 3–6. Fourier methods are employed for the removal of this subtle defect (Figure 10.3e; Kvamme 2003b). The difference between the images in Figure 10.3d and Figure 10.3e reveals the periodic nature of this phenomenon (Figure 10.3f ), with the horizontal striping corresponding to height variations associated with each step of the operator; it also indicates its generally low magnitude (the data range is –0.8 to +1.1 nT; mean positive amplitude is only 0.4 nT). Grid Matching

In total field data sets, in which each grid is differenced from a base station to remove diurnal change effects, large differences in measurement magnitudes can occur if the base station is placed in different locations during survey of each grid (Figure 10.4a, b). In these contexts the mean difference is determined between the edge rows or columns of adjacent grids and its value is then added to or subtracted from one of the grids to eradicate the difference and “balance” the measurements (Figure 10.4c).

Enhancements Once obvious defects and other forms of noise are removed from magnetic data sets, a variety of enhancements may be applied and a suitable presentation developed (Ciminale and Loddo 2001; Scollar et al. 1990:488–506). Some methods, such as low pass filtering, achieve enhancement by further reducing image noise. These tasks are summarized within several categories. Low Pass Filtering The technique of low pass filtering is designed to block high-frequency information in an image and “pass” low-frequency data, where “frequency” can be thought of as the spatial dimension of image components. For example, random instrument perturbations might occur at a scale of 10–20 cm; rodent holes in a prairie dog town might typically be 30 cm in size; prehistoric ditches are 2 m in width; houses are 8 m in diameter; and broad geological trends are dozens to hundreds of meters long. Low pass filtering is designed to reduce the contribution of image features small in size (e.g., sub-meter elements), a dimension that corresponds with many sources of noise. It can be accomplished through Fourier methods (where the amplitudes of highfrequency components are suppressed or eliminated; see above), but generally convolution techniques are employed (Lillesand and Kiefer 1994:553). In the latter, a simple average or Gaussian weighted average of the measurements within a narrow radius or “window” of each measurement is computed (Gaussian weights assign more influence to measurements near the center of the window and less influence to more distant ones following the pattern of a Gaussian or “normal” curve). Simple averaging and large-radius windows result in greater smoothing than Gaussian weights and windows

Data Processing and Presentation ~ 243

Figure 10.4. Many grid imbalances are seen in these total field data gathered by a proton precession magnetometer at the Roman city of Empuriés (ca. A.D. 100–300) in northeastern Spain, owing to differencing from base stations placed at several locations (these data were acquired at a sampling density of only 1 measurement/meter). (a) Orthographic view and (b) gray-scale view showing grid differences. (c) Balanced data after grid edge matching of mean values (color illustration appears on the CD).

of smaller width. A mild amount of smoothing can decrease high-frequency image components (Figure 10.3g) and reduce image noise components that might remain after other forms of processing (Figure 10.3h), making cultural features of somewhat lower frequency more apparent. Interpolation Interpolation is employed to improve image continuity and eliminate rectangular pixels that result from unequal sampling rates typically caused by there being more measurements per meter along transects than in the between-transect direction. Image continuity is an important visualization property in gray- or color-scale imagery. With low sampling densities and relatively few “blocky” pixels, the eye is drawn primarily to the pixel edges rather than the patterns of the grays or colors represented. That is why one typically “steps back” to view such imagery, because the apparent discontinuities between pixels then become blurred and data patterns are better recognized. Interpolation is an alternative that allows many more measurements to be estimated, reducing pixel size, their overall blocky appearance, and the apparent discontinuities that result. It is based on the realization that the actual measurements represent a systematic sampling of a continuous magnetic field; because it is continuous, estimates of additional values on that surface can be made between extant samples using simple spline or even linear functions. Scollar et al. (1990:502) argue that interpolation of new measurements between extant ones may be carried out as many as four times without introduction of spurious information. The data in Figure 10.3g, at a sampling density of 0.5 × 0.25 m, are interpolated to a uniform 0.25 × 0.25 m in Figure 10.3i, resulting in a smoother, more continuous-looking image.

244 ~ Kenneth L. Kvamme Contrast Manipulation Magnetic data sets are typically leptokurtic, a statistical term meaning that the distribution is highly peaked with long straggly tails (Figure 10.5a). In other words, 95 percent of the data might lie between ±1 nT, but the remaining 5 percent can vary between ±200 nT or more owing to the presence of relatively few extreme measurements. Obviously, if one “maps” 256 color or gray values (the common number in most computer displays) to the latter range, only the most prominent anomalies will be illustrated and few of the colors or grays will be assigned to the bulk of the data, resulting in poor image contrast and the invisibility of most features in the data (Figure 10.5a). Contrast improvement is typically achieved through a “linear stretch” with “saturation” of extreme values. That is, the 256 grays or colors are mapped to the range of the central bulk of the data, for example, the 95 percent of the measurements lying between ±1 nT, while the more extreme values are saturated to the lowest and highest grays or colors in the palette (Figure 10.5b). The result is good contrast and image detail because subtle changes of only fractions of a nanotesla are represented by multiple gray or color values. Shadowing Anomalies of extreme subtlety, owing to low magnetic susceptibility contrasts or greater depth, frequently represent targets of interest. Artificial shadowing of the magnetic surface can sometimes exaggerate and enhance these features. This is accomplished by computing image brightness and darkness values on the magnetic surface resulting from an imaginary light source placed at a low angle above it such that “shadows” result from even the smallest of anomalies. Moreover, by varying the light source direction, further enhancement can be achieved, because a linear feature is best revealed when the light source is perpendicular rather than parallel to it, for example. Shadowing is illustrated from four directions in Figure 10.5c–f. In each case, different anomalies are emphasized. A northwest–southeast trending trail, partially seen in the raw data (Figure 10.5b), is shown to cross-cut the entire study area by the shadow image with lighting from an orthogonal angle (indicated by arrows in Figure 10.5d).

Presentation The potential of even the “best” magnetometry data set may not be realized without a properly designed display that maximizes conveyance of the information it contains. Modern computer graphics offer a host of ways in which information may be presented. Some more effectively portray the results of magnetometry surveys than others. Choice of display form also depends on the goals of the presentation. Certain forms are best for illustrating relative or absolute differences between the magnitudes of anomalies. Other modes are best suited for illustrating features of extreme subtlety or regular cultural patterns that might occur across an area. Presentation of results to the media or the general public may require yet other kinds of displays.

Data Processing and Presentation ~ 245

Figure 10.5. Magnetometry data from Double Ditch State Historic Site (32BL8), North Dakota (ca. A.D. 1500–1780). (a) A pair of iron-caused dipolar anomalies (with values ranging from –4 to +12 nT, indicated by arrows) cause poor image contrast when available display values are associated with the full data range (source: Kvamme 2003c). (b) A linear contrast stretch that “maps” available display values to the –1, +1 interval, improving contrast and visualization of image details; more extreme data values become “saturated” with the minimum and maximum display values (here, white and black, respectively). (c) Shadow image with light source from the north. (d) Shadow image with light source from the northeast; arrows point to a subtle linear feature interpreted as a trail. (e) Shadow image with light source from the east. (f ) Shadow image with light source from the southeast.

246 ~ Kenneth L. Kvamme Continuous Gray or Color Scales In general, monochrome imagery best reveals subtle details, and cultural anomalies in much of North America are frequently expressed by very small changes of only fractions of a nanotesla. With a gray scale of 256 values, extremely faint anomalies can be illustrated and systematic patterns across a region more easily discerned and recognized (Figure 10.6a). Varied color palettes, on the other hand, can be misleading because the eye views transitions between different colors as high contrast changes when, in fact, the actual measurements behind them may vary little. The use of varied color palettes can therefore obscure subtle anomalies and create false impressions of transitions in the data, and they should be employed with caution during data exploration. Yet, a welldesigned color scheme can enhance a presentation when anomalies of relatively strong magnitude occur in distinct ranges and specific colors can be assigned to them. Color is also preferred by the media and the lay public, is pleasing to view, and can therefore be essential to project promotion. (See various Figure 10.6 portrayals in color on the accompanying CD.) Other Modes of Portrayal: Contouring and Pseudo-Three-Dimensional Views Contouring and pseudo-three-dimensional views have the advantage that the absolute or relative magnitudes of anomalies can be determined by the sizes or densities of the contours (Figure 10.6b) or by how pronounced the “peaks” might be in three-

Figure 10.6. Typical modes of graphic display illustrated with magnetometry data from the Great Bear effigy (Mound 31; ca. A.D. 650–1300) at Effigy Mounds National Monument, Iowa (source: Kvamme 1999b): (a) gray scale, (b) contouring, (c) pseudo-three-dimensional wire frame, (d) pseudo-three-dimensional shaded surface.

Data Processing and Presentation ~ 247 dimensional views (Figure 10.6c). Such information is typically lost in a gray scale because extreme measurements may lie within the “saturated” portions of the scales (e.g., the head or leg areas in Figure 10.6a; compare also Figure 10.4a, b). On the other hand, small and subtle anomalies generally become “lost” in contour or wire frame views, and they are frequently the ones of most interest (see Figure 10.4a, in which few details of Roman room blocks can be seen). Furthermore, because the eye is drawn to discrete edges, it tends to follow the individual contour lines rather than the overall pattern that may be presented. One alternative is to combine the gray or color scale with a three-dimensional surface representation (Figure 10.6d). Such portrayals offer the advantage of showing subtle details and the relative magnitudes of large anomalies at the same time. A Goal: Interpreted and Transcribed Maps The production of a final “end-image” that clearly portrays subsurface features of possible cultural origin through extensive data processing and enhancement is not necessarily the ultimate goal of magnetometry surveys or of remote sensing generally. The utility of results is greatly benefited by an analytic phase that interprets and transcribes features seen in the data accurately onto maps (Schmidt 2001:23). The process by which information locked in a magnetic image is transformed into a record of the archaeology is complicated but important. It requires knowledge of the kinds of archaeological features that might be present, some of which might be learned through archaeological testing, and experience with the magnetic signatures of those kinds of features, which can vary regionally; it is affected by the amount of noise remaining in the data and the clarity of regular or geometric patterns the features express; and technical knowledge of cartography and mapping is essential. Transcribed maps are composed of points, lines, and polygons that are built up and cumulate, through inspection of the data under a variety of filters, application of shadowing effects, or use of color palettes. If data from other remote sensing techniques are available, comparisons between them are useful to corroborate findings, allowing significant anomalies to be verified, modified, and ultimately codified (see Kvamme, Johnson, and Haley, this volume). GIS technology is particularly useful because features recognized in imagery may be digitized on-screen directly into a map coordinate system. With interpreted vectors entered into regional GIS-driven databases, a permanent record of the anomalies and the potential cultural resource base is developed. Such an inventory of known or suspected archaeological features is essential to guide management and planning decisions and future archaeological work at a site. A processed magnetic image is portrayed alongside an interpreted map containing polygons representing significant archaeological features in Figure 10.7. Raw geophysical imagery may not be familiar to and easily interpreted by many viewers with little background or training in this area. It is interpreted maps similar to that in Figure 10.7b that are most useful to archaeologists as an aid in targeting excavations and to site managers for documenting site content.

248 ~ Kenneth L. Kvamme

Figure 10.7. Magnetometry data from Huff Village State Historic Site (32MO11), a village of the Terminal Middle Missouri Variant (ca. A.D. 1450) in North Dakota. (a) Numerous anomalies surround and lie within rounded rectangular houses at the site. (b) An interpreted map based on anomalies >0.7 nT and an intensive soil coring program that led to the identification of many anomalies as subterranean food storage pits or hearths (soil-cored features are indicated with small dots); the remainder of the small anomalies are interpreted from these results, while house outlines can be discerned by faint lineations in a (supported by slight depressions in the surface and resistivity survey findings), and the “plaza” (a characteristic element in villages of this type) is denoted by an absence of anomalies (source: Kvamme 2003a, 2003b).

Acknowledgments Work at several of the sites illustrated here (Double Ditch, Huff, Primeau’s Trading Post) was supported by grants from Stan Ahler and the PaleoCultural Research Group of Flagstaff, Arizona, and Fern Swenson of the State Historical Society of North Dakota. Results at Effigy Mounds were acquired during a National Park Service training workshop. Findings at Empuriés, Spain, were obtained in the Boston University archaeological field school of 1996.

Data Processing and Presentation ~ 249

References Cited Ciminale, M., and M. Loddo 2001 Aspects of Magnetic Data Processing. Archaeological Prospection 8:239–246. Clay, R. B. 2001 Complementary Geophysical Survey Techniques: Why Two Ways Are Always Better than One. Southeastern Archaeology 20:31–43. Gaffney, C., J. A. Gater, P. Linford, V. Gaffney, and R. White 2000 Large-Scale Systematic Fluxgate Gradiometry at the Roman City of Wroxeter. Archaeological Prospection 7:81–99. Geoscan Research 2000 Geoplot Version 3.00 for Windows, Instruction Manual. Geoscan Research, Bradford, England. Hays, W. L. 1994 Statistics, 5th ed. Harcourt Brace, Fort Worth, Texas. Kvamme, K. L. 1999a Recent Directions and Developments in Geographical Information Systems. Journal of Archaeological Research 7:153–201. 1999b Archaeo-Geophysical Surveys at Effigy Mounds National Monument, Iowa. Submitted to the Midwest Archeological Center, National Park Service, Lincoln, Nebraska. 2001 Current Practices in Archaeogeophysics: Magnetics, Resistivity, Conductivity, and Ground-Penetrating Radar. In Earth Sciences and Archaeology, edited by P. Goldberg, V. Holliday, and R. Ferring, pp. 353–384, Kluwer/Plenum, New York. 2002 Final Report of Geophysical Investigations at the Fort Clark and Primeau’s Trading Posts, Fort Clark State Historic Site (32ME2): 2000–2001 Investigations. Submitted to the State Historical Society of North Dakota, Bismarck. 2003a Geophysical Surveys as Landscape Archaeology. American Antiquity 68(3):435–458. 2003b Multidimensional Prospecting in North American Great Plains Village Sites. Archaeological Prospection 10:131–142. 2003c Geophysical Findings at the Double Ditch State Historic Site (32BL8), North Dakota, 2002. Submitted to the State Historical Society of North Dakota, Bismarck. Lillesand, T. M., and R. W. Kiefer 1994 Remote Sensing and Image Interpretation. 3rd ed. John Wiley & Sons, New York.

250 ~ Kenneth L. Kvamme Schmidt, A. 2001 Geophysical Data in Archaeology: A Guide to Good Practice. Oxbow Books, Oxford. Scollar, I., A. Tabbagh, A. Hesse, and I. Herzog 1990 Archaeological Prospecting and Remote Sensing. Topics in Remote Sensing, No. 2, G. Hunt and M. Rycroft, series editors. Cambridge University Press, Cambridge. Toom, D. L., and K. L. Kvamme 2002 The “Big House” at Whistling Elk Village (39HU242): Geophysical Findings and Archaeological Truths. Plains Anthropologist 47:5–16.


Multiple Methods Surveys: Case Studies Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley

The foregoing chapters have demonstrated that a geophysical survey with a single instrument can provide much insight. Yet, most instruments respond primarily to a single physical property of the earth: magnetometry to soil magnetism, resistivity and electromagnetic induction to soil conductivity, and ground-penetrating radar (GPR) primarily to soil dielectric properties (Weymouth 1986:371). It is easy to conclude that surveys with multiple methods must offer greater insights because different dimensions of the subsurface are being quantified. Buried cultural features not revealed by one method might be made visible by another (Clay 2001; Piro et al. 2000). Additionally, subsurface features indicated by one method may be very different from and add complementary information to patterns revealed by another, as when a hearth and burned elements of a house are shown by magnetometry while soil changes that make up the house floor are portrayed by resistivity, conductivity, or radar methods. Recent advances in computer graphics and geographic information system (GIS) technology allow various data sets to be overlain or combined, making possible simultaneous visualization of features in each. Moreover, new GIS-based “data fusion” methods seek ways to combine the multidimensional imagery into composite data sets that enhance interpretability through the use of principal components analysis, statistical, or context-based methods (Johnson and Haley 2004; Kvamme 2001; Piro et al. 2000).

252 ~ Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley

Case Study 1 Prehistoric Earthlodges at Whistling Elk Village, South Dakota Kenneth L. Kvamme Whistling Elk (39HU242), located on the Missouri River in South Dakota, is a fortified earthlodge village assigned to the Initial Coalescent variant of the Plains Village tradition (ca. a.d. 1300). This settlement lies beneath nearly a meter of Missouri River sediments and lacks any surface expression aside from artifacts that occasionally spill out of the embankment and vegetation markings indicating the locus of a fortification ditch in a 1968 aerial photograph. This site was completely surveyed by electrical resistivity using a Geoscan Research RM15 twin-probe array with 1-m probe separation (Figure 11.1a), electromagnetic conductivity using a Geonics Limited EM38 conductivity meter (Figure 11.1b), and magnetic gradiometry using a Geoscan Research FM36 fluxgate gradiometer (Figure 11.1c). Sampling density was 1 × 0.5 m for the first two methods and 1 × 0.25 m for the last. Resistivity probe separation was 1 m (see Kvamme 2003; Toom and Kvamme 2002). Electrical contrasts between the occupation surface and more resistant sediments filling the abandoned ditches and houses produced the best definition of the site’s structure in the resistivity data (Figure 11.1a). An outer fortification ditch containing five uniformly spaced bastions encompassing approximately 1.45 ha is readily seen, and approximately 64 anomalies within the area enclosed by this ditch are interpreted as likely houses, based on the co-occurrence of the various forms of geophysical evidence. Although similar results were obtained by the conductivity survey (Figure 11.1b), cultural features were less well defined owing to peak instrument sensitivity at 0.4 m depth, somewhat above the cultural features (the conductivity data also contained a large plow mark response owing to its shallowdepth focus, signs of which were removed by Fourier methods as described in Kvamme, this volume, Chapter 10). The magnetic gradiometry data appear very noisy at the global scale of Figure 11.1c but nevertheless at a large map scale contain much detail about internal components of the village and individual houses. For example, the presence of large-magnitude positive anomalies suggests that as many as 34 (53 percent) of the houses may have been burned, quite possibly by a prehistoric sack of the village (see Kvamme 2003). The foregoing is best illustrated by focusing on one house, referred to as the “Big House” because of its unusual size (measuring 100 m2, three to four times the area of other houses). The resistivity data unambiguously define its square form and southeast-facing entryway, characteristic of houses of the Initial Coalescent variant (Figure 11.1d). The magnetometry data, on the other hand, provide information about such internal features as the locus of the centrally placed

Multiple Methods Surveys: Case Studies ~ 253 hearth and four roof support posts characteristic of this variant (Figure 11.1e; these magnetic data are from a second high-density survey placed over the house that obtained samples every 0.5 × 0.125 m for improved detail). The magnetic data also strongly suggest that this house was burned owing to large-magnitude measurements along much of its perimeter, particularly near the entryway (a principal source of fresh air and therefore a place where a fire would have burned hotter). These hypotheses were archaeologically tested and here, too, the accuracy of the geophysical mappings proved useful. With the time and personnel available it was only possible to open a 2-×-6-m trench. The geophysical mapping guided the placement of this trench to extend exactly beyond the central hearth on one end to just beyond the house wall on the other, including one of the predicted central roof support posts. The excavation also revealed the house floor at about 98 cm below the surface, that it was burned, and that it contained many utilitarian artifacts including a complete pot and evidence of food debris (Toom and Kvamme 2002). Validation of the geophysical signatures seen in this house lends greater confidence to the interpretation of other anomalies seen within the village (Figure 11.1a–c).

Figure 11.1. Geophysical surveys at Whistling Elk village (39HU242): (a) resistivity; (b) conductivity; (c) magnetic gradiometry. Close-up views in (d) resistivity and (e) magnetic gradiometry images reveal great detail about house form and content. P = locations of known and inferred roof support posts (color illustration appears on the CD).

254 ~ Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley

Case Study 2 Historic Earthlodge at Mitu’tahakto’s Village, North Dakota Kenneth L. Kvamme Mitu’tahakto’s Village was founded by the Mitutanika Mandan about 1822 but was later usurped by the Arikara after the disastrous smallpox epidemic of 1837 decimated the Mandan; the Arikara occupied it until 1861 (Wood 1993). It is a large 7.1-ha village located on the Missouri River in the Fort Clark State Historic Site (32ME2). During this time earthlodges were circular, measuring 14–18 m in diameter, and contained long, linear entryways. Many are depicted by the artists George Catlin and Karl Bodmer, who visited this settlement in 1832 and 1833–1834, respectively. Magnetic gradiometry and electrical resistivity surveys were conducted within a 400-×-20-m transect through the core of this village that indicated numerous circular earthlodges and many of their internal features with floors lying only 0.4–0.7 m below the surface. Most of these houses corresponded with depressions visible in the surface that varied between 0.3 and 0.8 m in depth (see Kvamme 2003). One house that was not indicated by surface evidence was discovered by the magnetometry and resistivity surveys; a GPR survey of this 20-×-25-m area was subsequently conducted using a Geophysical Survey Systems, Inc. (GSSI), SIR-2000 GPR system with a 400-MHz antenna. Each of these data sets offers a very different view of this earthlodge. The resistivity data were sampled every 0.5 × 0.5 m with a 0.5-m probe separation; magnetic data every 0.5 × 0.25 m; and GPR data every 0.5 × 0.02 m, with 512 samples at each point in a 20-nS time window. As with other houses at this site and at different earthlodge villages, the resistivity data indicate a circular ring of high resistivity surrounding a floor of somewhat lower resistivity (Figure 11.2a). This phenomenon is interpreted as a zone of less compact sediments eroded from lodge roofs by the actions of wind and rain. The resistivity image of this area is particularly interesting because it suggests the possibility of overlap-

Figure 11.2. Geophysical mappings of circular earthlodge(s) at the Mandan/Arikara village in the Fort Clark State Historic Site (32ME2): (a) resistivity; (b) GPR time slice (11–16 nS); (c) magnetic gradiometry (color illustration appears on the CD).

Multiple Methods Surveys: Case Studies ~ 255 ping houses that may indicate a later Arikara dwelling superimposed over a previous Mandan one. The GPR data present almost a mirror image of the resistivity findings in an 11– 16-nS time slice (about 42–63 cm below the surface; Figure 11.2b). The earthlodge floor is indicated by high-amplitude reflections and a southeast-facing entryway is unambiguously defined. These anomalies may be due to greater compaction of the floor areas. The circular outer zone of high resistivity, however, yielded few reflections, emphasizing very different characteristics between electrical conductivity and dielectric properties. No indication of a second earthlodge was seen in any of the GPR time slices, although it may be that the time window was too narrow to encounter a house that might exist at a somewhat lower depth. The magnetometry data reveal a plethora of anomalies associated with the interior of this house (Figure 11.2c), but there is little clear pattern compared with features in other earthlodges that have been surveyed by magnetometry (see elsewhere in this volume). This is partially because most of the anomalies are large in magnitude, resulting from the presence of iron artifacts, and therefore obscure more subtle soil anomalies (most of the iron-caused anomalies are dipolar). It is evident that this household had ready access to iron trade goods. The ubiquitous centrally placed hearth can also be discerned, and other anomalies probably correspond to subterranean corn storage pits and auxiliary hearths. While it is difficult to distinguish among the types of features these anomalies may represent in a simple graphic, a detailed analysis of each anomaly that considers its form, areal extent, and magnitude is able to do so for a large percentage of them.

Case Study 3 A Historic Church Foundation in Arkansas Kenneth L. Kvamme The Mount Comfort Church (3WA880) was constructed of brick in the 1840s, serving in its history as a church, meetinghouse, female seminary, and hospital for both sides during the Civil War. The building was burned in 1863 and many of its bricks were salvaged by Union troops to build barracks. Limited archaeological excavations in the early 1990s revealed that much of the brick foundation remains only 35 cm below the surface and that the area is littered with numerous iron artifacts, principally nails and stove parts (Hilliard and Perigo 1994). Resistivity, GPR, and magnetometry surveys each give new insights about this structure. The resistivity survey employed a Geoscan Research RM15 twin-probe array with 50-cm probe separation and 0.5-×-0.5-m samples. The highly resistant brick clearly reveals the outline of the rectangular foundation, measuring approximately 10 × 12 m, as well as several interior building support piers (Figure 11.3a). The GPR survey utilized a GSSI SIR-2000 system with a 400-MHz antenna, acquiring data every 0.5 × 0.04 m,

256 ~ Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley with 512 samples at each point in a 50-nS time window. An 8–16-nS time slice (about 24–48 cm below surface) yields anomalies that correspond almost exactly with the resistivity findings (Figure 11.3c). Yet, a higher slice shows features not seen in the resistivity results (possibly a walkway, room partition, patio/connecting room, or debris pile; Figure 11.3b). In deeper GPR slices, the foundation begins to disintegrate (Figure 11.3d), until only bedrock variations are ultimately seen in lower slices (not illustrated). The magnetometry survey employed a Geoscan Research FM36 fluxgate gradiometer with sampling at 0.5 × 0.25 m. In part, the magnetic results parallel findings in the other data sets because the remanent magnetism in the brick foundation produces pronounced anomalies, making many of the foundation lines and support piers visible (Figure 11.3e). The magnetic data also indicate numerous robust dipolar anomalies that indicate the presence of much ferrous metal below the surface, including pieces of large size. As suggested by the earlier excavations, many of them are probably iron stove parts.

Figure 11.3. The brick foundation of the Mount Comfort Church (3WA880) as revealed by (a) resistivity, (b) GPR in a 2–8-nS slice (to 24 cm below surface), (c) GPR in a 8–16-nS slice (24–48 cm below surface), (d) GPR in a 16–24-nS slice (48–72 cm below surface), and (e) magnetic gradiometry (color illustration appears on the CD).

Multiple Methods Surveys: Case Studies ~ 257

Case Study 4 Five-Dimensional Data Fusion at Army City, Kansas Kenneth L. Kvamme Army City was a privately owned commercial complex established in 1917 by local entrepreneurs to service troops stationed at nearby Camp Funston (both are now part of Fort Riley). It included hotels, restaurants, barbershops, churches, a bank, theaters, clothing and jewelry stores, and other shops. Owing to the lack of troops after the end of World War I in 1918 and a fire that burned much of the town in 1920, Army City soon declined and was ultimately abandoned. It now rests under a hay field (Hargrave et al. 2002). A resistivity survey of the site in the mid-1990s (reported in Hargrave et al. 2002) revealed the potential of Army City for further geophysical investigations because a whole village lies buried, complete with cement foundations, water and sewage pipes, burned structures, streets, and many other features, many less than a half meter below the surface. As part of a project to investigate the potential of “fusing” multiple remote sensing data types, several geophysical surveys were carried out within a 100-×-160-m block at the heart of Army City in 2002. An electrical resistivity survey used a Geoscan Research RM15 twin-probe array with 0.5-m probe separation to capture data every 1 × 0.5 m. These data emphasize a number of resistant building foundations and building footers, probably of concrete, and give hints of the outlines of other buildings and streets (Figure 11.4a). An electromagnetic conductivity survey using the quadrature phase of a Geonics EM38 conductivity meter reveals primarily buried pipelines, also with 1-×-0.5-m samples (Figure 11.4b). The magnetic gradiometry survey with a Geoscan Research FM36 fluxgate gradiometer acquiring samples every 1 × 0.25 m paralleled the conductivity survey by showing iron pipelines, but a number of individual iron artifacts and buildings are also indicated, the last possibly from burning (Figure 11.4c). The in-phase component of a Geonics EM38 conductivity meter and the shallow depth of many of the cultural features allowed magnetic susceptibility data to be obtained every 1 × 0.5 m. These data offer some of the best detail for many of the village’s structures, including individual rooms, most likely from magnetic enrichment of the soil, perhaps in part stemming from its firing (Figure 11.4d). Finally, a GPR survey using a GSSI SIR-2000 system with a 400-MHz antenna obtained data every 0.5 × 0.04 m, with 512 samples at each point in a 60-nS time window. Multiple time slices were generated, each giving a somewhat different view of this town. One slice at 8–12 nS (36–54 cm below surface) shows particularly good detail about structures and even rooms within them (Figure 11.4e). Although viewing the individual data layers side by side is informative (as in Figure 11.4a–e), observing multiple layers simultaneously in single images can yield additional and new insights because complementary and different anomalies can be seen at the same time. Several such methods were investigated from a group of techniques collectively referred to as data fusion (Kvamme 2001). The most common, and per-

258 ~ Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley

Figure 11.4. Multidimensional geophysics at Army City: (a) resistivity; (b) conductivity; (c) magnetic gradiometry; (d) magnetic susceptibility; (e) GPR time slice at 8–12 nS (36–54 cm below surface); (f ) RGB color composite (R = magnetic susceptibility, G = resistivity, B = GPR); (g) overlay of translucent maps a–e; (h) first principal component; (i) RGB color composite of principal components (R = component 2, G = component 3, B = component 1) (color illustration appears on the CD).

haps simplest, technique of data combination is known as RGB (red-green-blue) color compositing, which simply assigns a primary color (red, green, blue) to each of three map layers. When these are viewed simultaneously a false-color image is produced containing components of all three data sets (Figure 11.4f; note that this image and others are best viewed in color on the accompanying CD). A related method employs a computer graphic technique in which each remote sensing layer is assigned a different color, made translucent, and overlaid, allowing simultaneous visualization of all layers (Figure 11.4g). Principal components analysis (PCA) is a mathematical-statistical technique that linearly combines multivariate measurements in input layers based on intercorrelations between them (Dunteman 1989). The resulting components are ordered in such a way that the first contains more of the total variance (information content) than the second, which contains more than the third, and so on. The first three components will typically contain the bulk of the total variance in the input data; by color compositing them, most of the variation may therefore be presented in a single image.

Multiple Methods Surveys: Case Studies ~ 259 Running a PCA on geophysical data poses certain problems. While resistivity (Figure 11.4a), magnetic susceptibility (Figure 11.4d), and GPR (Figure 11.4e) yield results that increase monotonically with the phenomenon being sensed (soil resistivity, magnetic susceptibility, and dielectric contrast, respectively), the conductivity (Figure 11.4b) and magnetometry (Figure 11.4c) results are dipolar at Army City owing to the presence of large amounts of metal in the site. Dipolar results mean that (1) positive and negative measurements are associated with a single anomaly, greatly upsetting correlation structures with other layers, and (2) the poles are somewhat offset from the true locations of the target. Working with the absolute values of these data sets gives a partial solution, which is employed here. The correlation matrix between the variables revealed low levels of correlation, with the highest, that between magnetometry and resistivity, being only r = .33 (indicating that, at best, the layers have only 100r 2 = 100(.33)2 = 10.9 percent of their variance in common). Yet, the first three components of a standardized PCA accounted for 74 percent of the total data variance, with the first component accounting for 36 percent by itself. Significantly, that component had roughly equal loadings (correlations between the input maps and the component) of moderate size with each of the data sets (ranging from .44 with magnetic susceptibility to .76 with magnetometry), indicating that the first component, alone, forms a suitable composite of the five data sets (Figure 11.4h). The second and third components were less mixed, representing essentially conductivity and magnetic susceptibility, respectively. An RGB color composite of the first three components, containing more than double the information in the first component, is given in Figure 11.4i.

Case Study 5 Using Geophysical Results in a Multivariate Exploration of Airborne Imagery at the Hollywood Mounds, Mississippi Jay K. Johnson and Bryan S. Haley A five-year program of archaeological and remote sensing research focusing on the Hollywood mound site in northeastern Mississippi was concluded in 2002. Two contract reports (Haley et al. 2002; Johnson et al. 2000) and several theses have presented results from the program (Edwards 2003; Haley 2002; Peukert 2002; Reynolds 2002). This late prehistoric ceremonial center consists of a rectangular ring of low mounds over the corner of which a larger platform mound with a ramp was constructed late in the occupation of the site. All but the platform mound and three of the low mounds have been leveled by more than a century of agricultural activity, and we would have had a great deal of difficulty finding them if Berle Clay had not done a gradiometer survey of the site (Figure 11.5). Mounds are evident in the remote imagery as oval rings that mark the interface between the base of a mound and the fill surrounding the mounds. The

260 ~ Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley remains of several prehistoric houses are evident as rectangular patterns in the northern third of the image. The remote sensing portion of the Hollywood program analyzed data layers derived from standard aerial photo graphy, airborne multispectral sensors, and near-surface geophysical methods. Several sets of blackand-white aerial photographs ranging in date from 1938 to 1992 were acquired from the Soil Conservation Service Program. These were scanned at a high resolution to be used as digital data layers. The 1938 and 1992 photographs were included as data layers Figure 11.5. Gradiometer image of the Hollywood Mounds site (geofor this experiment. physical survey by Berle Clay). Another set of photographic data was produced during a NASA flyover of the site using a large-format camera loaded with color infrared film. Ground conditions were bare earth at that time. These photographs were also scanned using a high-resolution scanner and stored as three bands of data to be compared with other data layers (Figure 11.6). The NASA flyover also carried an ATLAS sensor, which produced 15 bands of data in visible, near-infrared, mid-infrared, and thermal infrared wavelengths with a ground resolution of 2.5 m. Two other multispectral data sets were collected while vegetation (low weeds) covered the site. These were obtained by ADAR, an airborne sensor flown by Positive Systems, and by IKONOS, a satellite-based sensor operated by Space

Multiple Methods Surveys: Case Studies ~ 261 Imaging. Both sensors produce four bands of data covering visible and near-infrared wavelengths. Ground resolution for these data sets was 0.7 m for the ADAR and 1 m for the IKONOS (Figure 11.7). All photographic, digital multispectral, and geophysical data layers were georeferenced to the established Hollywood site grid system using ERDAS Imagine 8.5 software. An area centering on one of the burned structures, an area including a house pattern, and an area including the interface of two mound rings were chosen for this data fusion

Figure 11.6. Photographic imagery of the Hollywood Mounds site. CIR = Color infrared (color illustration appears on the CD).

Figure 11.7. Airborne digital imagery of the Hollywood Mounds site. N = near infrared; R = red; G = green; TIR = thermal infrared (color illustration appears on the CD).

262 ~ Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley experiment. Each of these 20-×-20-m areas was subset from each data set and resampling was performed using standard Kriging techniques with 0.5-m cell sizes. Since burned structures and the fired clay rubble that results are a consistent feature of late prehistoric sites in the southeastern United States, magnetometry is particularly useful in mapping buried structures. Therefore, we began this analysis by using the gradiometry imagery to classify the pixels in the test cases into areas of high, medium, and low values (Figure 11.8). Low was defined as all the pixels with values less than one standard deviation from the mean for that image. The medium class contained all pixels within one standard deviation of the mean and high referred to those with values that were greater than one standard deviation from the mean. Because areas with strong remanent magnetism like the burned clay remains of the houses at Hollywood were expressed as a dipole, these features were represented by high positive and negative values. The low and high groups of pixels marked the locations of wall remnants or burnt clay talus deposits at the site. Each of the 20-m2 test areas was made up of 1,681 pixels for which, not counting the gradiometry data, 27 values were recorded (ATLAS, 14 bands; ADAR, 4 bands; IKONOS, 4 bands; 1938 panchromatic; 1992 panchromatic; large-format color infrared, 3 bands) (Figure 11.9). Given the goal of the analysis and the need to evaluate the relative importance of each of the sensors in duplicating the gradiometry results, discriminant analysis was selected. A stepwise consideration of each variable based on the Wilks’s lambda criterion was done in order to reduce the final set of variables used in the analysis, simplifying the interpretation. The house site and mound ring test areas were analyzed separately but similar results were produced. The stepwise procedure reduced the number of variables from 27 to 18 for the test area including the house remains (Table 11.1). All but four of the variables used were recorded when the ground was bare, suggesting that patterns in the growth of the ground cover do not reflect the buried structure well. Because of the nature of the computation, a three-class analysis produces two discriminant functions. In this case, the first

Figure 11.8. Magnetic gradient image of prehistoric house remains reclassified into three classes of data.

Multiple Methods Surveys: Case Studies ~ 263 function accounted for 81.0 percent of the variation in the data matrix. Three variables showed particularly high coefficients on this function. One was the green band of the scanned large-format color infrared film image and the other two were the blue and red bands of the IKONOS image. The IKONOS image was recorded with vegetation

Figure 11.9. Airborne imagery used in reclassification. LF CIR, Large-format color infrared film.

264 ~ Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley Table 11.1. Standardized canonical discriminant function coefficients for house location analysis


Function 1 2 –.178 .302 .247 –.559 –.425 –.960 .325 .781 –.615 –.053 .390 .540 .380 –.443 –.305 .549 .511 .047 –.366 –.332 –.370 –.224 –1.881 .508 1.910 –.469 .527 .320 –.204 .904 .523 .019 –1.034 –.069 .576 –.067

LFCIR = Large-format color infrared film.

covering the ground. The statistical package that was used (SPSS for Windows) allows a comparison between the original class to which each pixel was assigned on the basis of the gradiometer data and the reclassification on the basis of the two discriminant functions (Table 11.2). A 65.7-percent correct reclassification resulted, about twice what would be expected by chance alone. However, when the reclassified values are displayed, much of the patterning in the image that made it possible to identify the likely location of a burned structure is lost (Figure 11.10). There are two spatial issues that may have reduced the effectiveness of the data fusion technique in this project. The most obvious of these is the accuracy of the data registration procedure. Because the ground targets are relatively small and pixel values are being compared directly, a small amount of referencing error

Figure 11.10. Original pixel classification and discriminant function results.

Multiple Methods Surveys: Case Studies ~ 265 Table 11.2. Classification results for house location analysis

1.00 Original group membership

Predicted Group Membership 1.00 2.00 3.00 114 27 44

Total 185
















could negatively impact the data fusion. Another problem could be in the nature of the gradiometer data. Magnetic gradient survey typically has a north-south offset by as much as 50 cm due to the way the earth’s magnetic field is oriented. A pole reduction is sometimes performed to counter this problem and this might have helped in this case. Still, given these considerations, a greater than 60-percent correct reclassification would seem to suggest that the experiment was a moderate success. It is possible to use airborne imagery to duplicate gradiometry results. As expected, there is a relationship between the various data sets that might be useful. Since, like most geophysical techniques, gradiometry survey is a labor-intensive technique, demanding a good deal of fieldwork and image preparation, the use of airborne sensors could be important in developing a general area research design.

Acknowledgments Research at Whistling Elk was supported by a grant from the National Center for Preservation Technology and Training, National Park Service. Work at the Fort Clark State Historic Site was supported by grants from Stan Ahler and the PaleoCultural Research Group of Flagstaff, Arizona, and Fern Swenson of the State Historical Society of North Dakota. Students at the University of Arkansas assisted with the collection of data at the Mount Comfort Church. Work at Fort Riley was funded by the Strategic Environmental Research and Development Program, Department of Defense, W. F. Limp and K. L. Kvamme, co-principal investigators. The work at Hollywood was funded by grants from the Mississippi Department of Archives and History (MDAH), the National Aeronautics and Space Agency (NASA), and the University of Mississippi Geoinformatics Center (UMGC) and the Mississippi Space Commerce Initiative (MSCI). UMGC and MSCI are NASAfunded initiatives. In addition, Marco Giardino, a NASA scientist with the Earth Science Applications Directorate at Stennis Space Center has been a constant collaborator in this research. Berle Clay at Cultural Research Analysts conducted the gradiometer survey at the Hollywood site. The Hollywood Mounds were donated to MDAH by Neal Block.

266 ~ Kenneth L. Kvamme, Jay K. Johnson, and Bryan S. Haley

References Cited Clay, R. B. 2001 Complementary Geophysical Survey Techniques: Why Two Ways Are Always Better than One. Southeastern Archaeology 20:31–43. Dunteman, G. H. 1989 Principal Components Analysis. Sage, Newbury Park, California. Edwards, P. A. 2003 An Analysis of Late Prehistoric Ceramics from the Hollywood Site (22CO500) in Tunica County, Mississippi. Unpublished Master’s thesis, University of Mississippi, Oxford. Haley, B. S. 2002 Airborne Remote Sensing, Image Processing, and Multisensor Data Fusion at the Hollywood Site, a Large Late Mississippian Mound Center. Unpublished Master’s thesis, Department of Sociology and Anthropology, University of Mississippi, Oxford. Haley, B. S., J. K. Johnson, and R. Stallings 2002 The Utility of Low Cost Thermal Sensors in Archaeological Research. Center for Archaeological Research, University of Mississippi, Oxford. Report prepared for the Office of Naval Research, NASA grant NAG5-7671. Hargrave, M. L., L. E. Somers, T. K. Larson, R. Shields, and J. Dendy 2002 The Role of Resistivity Survey in Historic Site Assessment and Management: An Example from Fort Riley, Kansas. Historical Archaeology 36(4):89–110. Hilliard, J., and M. Perigo 1994 Archeology Reveals Brick Church at Mount Comfort. Flashback 44:24–38. Published by the Washington County Historical Society, Fayetteville, Arkansas. Johnson, J. K., and B. S. Haley 2004 Multiple Sensor Applications in Archaeological Geophysics. In Sensors, Systems, and Next-Generation Satellites VII, edited by R. Meynart, S. P. Neeck, H. Simoda, J. B. Lurie, and M. L. Aten, pp. 688–697. Proceedings of SPIE, vol. 5234. SPIE, Bellingham, Washington. Johnson, J. K., R. Stallings, N. Ross-Stallings, R. Berle Clay, and V. Stephen Jones 2000 Remote Sensing and Ground Truth at the Hollywood Mounds Site in Tunica County, Mississippi. Center for Archaeological Research, University of Mississippi, Oxford. Submitted to the Mississippi Department of Archives and History.

Multiple Methods Surveys: Case Studies ~ 267 Kvamme, K. L. 2001 Archaeological Prospection in Fortified Great Plains Villages: New Insights through Data Fusion, Visualization and Testing. In Archaeological Prospection: 4th International Conference on Archaeological Prospection, edited by P. M. Doneus, A. Eder-Hinterleitner, and W. Neubauer, pp. 141–143. Austrian Academy of Sciences Press, Vienna. 2003 Multidimensional Prospecting in North American Great Plains Village Sites. Archaeological Prospection 10:131–142. Peukert, J. N. 2002 Ground Penetrating Radar at Hollywood. Unpublished Master’s thesis, University of Mississippi, Oxford. Piro, S., P. Mauriello, and F. Cammarano 2000 Quantitative Integration of Geophysical Methods for Archaeological Prospection. Archaeological Prospection 7:203–213. Reynolds, M. D. 2002 Magnetic Remote Sensing and Ground Truth: Some Examples from the Hollywood Site, Tunica County, Mississippi. Unpublished Master’s thesis, University of Mississippi, Oxford. Toom, D. L., and K. L. Kvamme 2002 The “Big House” at Whistling Elk Village (39HU242): Geophysical Findings and Archaeological Truths. Plains Anthropologist 47:5–16. Weymouth, J. W. 1986 Geophysical Methods of Archaeological Site Surveying. In Advances in Archaeological Method and Theory, vol. 9, edited by M. B. Schiffer, pp. 311–395. Academic Press, New York. Wood, W. R. 1993 Integrating Ethnohistory and Archaeology at Fort Clark State Historic Site, North Dakota. American Antiquity 58:544–559.


Ground Truthing the Results of Geophysical Surveys Michael L. Hargrave

Spectacular images of submound structures, Plains pithouse villages, and Spanish missions may convey the impression that most “good” geophysical surveys, like latenight television, leave little to the imagination. This impression is a result of an understandable tendency for geophysicists to distribute images from their most dramatic surveys. In fact, well-executed surveys often yield useful results that are not immediately interpretable by many archaeologists. Effective ground truthing is often the key that unlocks the information content of a geophysical map. Ground truthing is an effort to verify and enhance the results of a remote sensing study through the use of independent evidence. Note that the word truthing refers to the interpretation of the remote sensing data; it does not imply that the actual data may be spurious. If a remote sensing study is properly executed, variation in the data other than statistical noise will have some cultural or geological source or sources. The origins of ground truthing are, of course, intertwined with those of remote sensing. Remote sensing began with, or was at least greatly stimulated by, the development of photography. In the form of aerial photography, remote sensing was first used systematically during World War I (Scollar et al. 1990:26). Archaeological features were sometimes detected during the course of military reconnaissance missions. The recognition that remotely sensed military information could be extremely useful must have led to an almost immediate concern with verifying the reliability of that information. Soon after the war, aerial remote sensing for archaeological purposes began in earnest (Crawford 1924).

270 ~ Michael L. Hargrave In archaeogeophysics, ground truthing generally focuses on determining the subsurface sources of geophysical anomalies. Nonarchaeological information that can be used to ground truth geophysical data includes current and historic maps and photographs (particularly aerial photographs), other historical documents, and anecdotal information provided by local informants. These sources are generally most useful in identifying relatively recent phenomena that may account for geophysical anomalies: historic buildings, roads, drainage features, and other landscape modifications. Archaeologists generally view excavation as the preferred means of ground truthing a geophysical survey. Unfortunately, most of the available overviews of geophysics that are directed at archaeologists provide little or no discussion of the merits of various approaches to archaeological ground truthing. This may reflect or at least perpetuate the assumption that effective ground truthing is a straightforward matter that requires no particular consideration. In fact, the potential information return of a geophysical study is far too commonly greatly limited as a result of insufficient or ineffective ground truthing. The goal of this chapter is to identify some of the important issues in ground truthing and to discuss a number of approaches that have been found useful by the authors of this volume.

Why Ground Truth? To explain why ground truthing is so important, it is useful to review the goals of a geophysical survey. At the most basic level, the archaeologist is generally interested in identifying subsurface cultural features. In remote sensing terms, this goal involves both detection and classification. One must first detect anomalies that may be associated with discrete subsurface phenomena. Classification begins with an attempt to differentiate anomalies associated with cultural features from clutter, i.e., anomalies associated with other phenomena. Clutter can include anomalies related to tree roots, rodent burrows, rocks, plow furrows, recent metallic debris, and so forth. Under favorable circumstances, the goals of classification can be more ambitious. Here one may attempt to classify anomalies into useful archaeological categories such as pits, house basins, wall trenches, and graves. The preceding chapters in this volume have presented a number of maps in which this type of classification can be achieved with minimal excavation. These maps result from surveys that were well executed under favorable conditions. At many sites, however, classification is seriously complicated by the issue of equifinality. Very different phenomena can be manifested in a geophysical map by very similar anomalies. Consequently, a geophysical map cannot necessarily be interpreted as if it were an aerial photograph of a site from which the A horizon has been removed to expose the subsurface features. The size and shape of a geophysical anomaly may or may not resemble the dimensions of the underlying subsurface deposit. The factors that influence the spatial relationship between an anomaly and its source include the geometry and material composition of the source, the nature

Ground Truthing the Results of Geophysical Surveys ~ 271 of the contrast between the source object and its surroundings, and the geophysical sensor and survey design. Magnetic anomalies can be particularly difficult to interpret by means of a simple visual inspection. Depending upon their depth and orientation, small pieces of iron or steel can be manifested as weak circular monopoles similar in appearance to those associated with earth-filled pits (Bevan 1998:25). Natural phenomena such as rodent burrows, burned and unburned tree roots (particularly large taproots), and iron-rich noncultural rock may also be manifested by anomalies similar to those of pit features (Somers and Hargrave 2001). Highly magnetic materials (particularly metal objects) are often manifested by a dipole anomaly. Strong, potentially large magnetic anomalies will obscure weaker anomalies such as those commonly associated with prehistoric archaeological features. Uncertainties about the depth and horizontal position of the underground sources of anomalies also make it highly inadvisable to treat a geophysical map as if it were simply an archaeological map of features. For example, the magnetic object associated with a dipole anomaly is likely to be offset somewhat from the location of the magnetic high. Typically, the buried object or source will be found a short distance from the magnetic high along an axis that connects the high with the associated magnetic low (Bevan 1998:24). The depth of a buried object can be estimated using the half-width rule and a contour map of the anomaly. The objects’ depth below the sensor (not the ground surface) is approximately equal to the mean width of the anomaly at 50 percent of the anomaly’s maximum value. This estimate can be made using data collected by any type of magnetometer (Bevan 1998:25; Clark 1996:83). At intensively occupied sites, the palimpsest effect can make it difficult to differentiate anomalies associated with individual features. Localized areas of particularly rich midden may, for example, be mistaken for discrete features. For some techniques, the same feature may be manifested in variable ways, depending upon soil moisture (Clark 1996:125; Kvamme 2001a:361). This is particularly true for electrical resistance. Pits may be positive anomalies one season but appear as negative anomalies at other times of the year. On balance, the weak contrast exhibited by many prehistoric features, the presence of clutter, uncertainties about depth and exact horizontal position, and the palimpsest effect at intensively occupied sites are factors that contribute to the need for ground truthing information to maximize the archaeological value of a geophysical survey.

Key Issues in Ground Truthing At least five issues should be considered in selecting the general approach to be used in archaeological ground truthing: (1) information return, (2) cost, (3) invasiveness, (4) social and political issues, and (5) risks to personnel. These factors are listed here more or less in order of their likelihood of playing a key role in how ground truthing excavation should be executed.

272 ~ Michael L. Hargrave 1. Information return refers to the nature and quality of information that will result from ground truthing. For example, if a geophysical survey is conducted in the context of a National Register of Historic Places (NRHP) eligibility assessment of a site, the focus of ground truthing may simply be to verify whether at least some subsurface features (pits, hearths, and so on) are present at the site. The mere presence of such features does not necessarily mean that a site can be considered eligible, but knowledge of their presence or absence certainly represents an important element in an eligibility assessment. In contrast, the ground truthing strategy for a large-area survey of a late prehistoric mound center may focus on the classification and quantification of a diverse array of anomalies. Here ground truthing may require a rather elaborate sampling strategy in order to provide the data needed for inferences about community plans, population estimates, and so on. 2. Cost is nearly always a primary consideration in ground truthing. In the context of cultural resource management (CRM) projects, it may be particularly important to minimize the costs of ground truthing in order to offset the costs of the geophysical survey (Hargrave, Somers et al. 2002). In contrast, geophysics may be used in a site mitigation program to reduce the amount (and total costs) of subsequent large-scale excavations. Here it may be desirable to accept higher costs for a ground truthing program designed to sample the site deposits. Savings in the cost of the overall data recovery project may be achieved through effective placement of large excavation blocks. Savings may also be achieved in that the use of geophysics may obviate the need for the systematic excavation of shovel test pits and/or the controlled surface collection of artifacts (Kvamme 2003b). 3. Invasiveness is the degree to which the ground truthing destroys some of the archaeological deposits at the site. By its very nature, archaeological excavation destroys or at least damages the deposits being investigated. This is often not a concern in CRM contexts where site destruction is a likely outcome of construction activities. At scientifically and/or culturally important sites that are well protected (e.g., state historic sites), it is generally important to minimize the invasiveness of ground truthing excavations. 4. Social and political factors are occasionally important considerations in ground truthing. Recent legislation suggests that these factors may become increasingly important in the future. Sites with human remains represent obvious examples. The possible presence of Native American burials will, in most cases, favor a minimally invasive approach to ground truthing. This is particularly true for sites on federal land and in situations in which the site will not necessarily be destroyed by construction activities. Similarly, ground truthing may not be permitted in geophysical surveys of historic cemeteries, at least until decisions are made as to whether impacts of construction projects can be avoided. 5. In some cases, health and safety considerations may play an important role in determining the nature of ground truthing. On military installations there is, in some areas, a risk of unexploded ordinance (UXO). Areas known to contain UXO are gen-

Ground Truthing the Results of Geophysical Surveys ~ 273 erally omitted from CRM investigations, but the possibility of encountering UXO should always be considered. In urban and industrial areas there may be a risk of encountering toxic waste deposits, either in sealed containers or as a component of the soil. Such areas often represent poor candidates for geophysical survey for a variety of reasons (dense scatters of metal debris, a complex history of land disturbance, and so on). Nevertheless, the possibility of health and safety factors must be considered whenever excavation is planned in urban or industrial areas. In most situations, information return, cost, and invasiveness will be the key factors in developing a ground truthing strategy. Information return tends to be positively correlated with cost and invasiveness. That is, more extensive excavations will be more invasive but should (if reasonably well planned and well executed) result in more information return. With careful planning and experience, however, one can develop ground truthing strategies that maximize information return while minimizing cost and invasiveness. There is no single approach to ground truthing that will be the best choice for all sites and situations. In developing a strategy for ground truthing, one should first consider the relative importance of the key factors: information return, cost, invasiveness, social and political issues, and risks to personnel. One should then identify and use excavation techniques that are appropriate to the site conditions and that are feasible given the relative importance of the key factors discussed above. Each archaeological site represents a unique combination of factors that influence the potential success of a geophysical survey. Sites vary greatly in terms of soil, moisture, bedrock, vegetation, rodent activity, disturbance of the upper strata, and so on. A key issue determining the success of a geophysical survey is the nature of the contrast between the archaeological features of interest and the surrounding soil (Kvamme 2001a:356). Archaeologists are also aware that there may be a wide range of variation in the nature of the archaeological deposits present at a site. In some situations, features may be relatively easy to define. The fill in pit features may be enriched by organic content, particles of carbon and burned soil, and abundant artifacts. Such features often exhibit a sharp contrast in color and texture with the relatively sterile surrounding soil. At other sites that were more intensively occupied, features may be intrusive into rich midden deposits. In extreme cases, it may be impossible to identify the outlines of the uppermost portions of many features. Features may also be exceedingly difficult to identify as discrete deposits at sites characterized by sandy soils. Here the organic content of features rapidly leaches away, and the artifact contents of pits may disperse through the sandy soil. The remains of such features can be very difficult to identify, particularly in a soil core or small shovel test. On balance, the archaeologist may not know what excavation techniques will be necessary to identify features unless he or she has previously excavated at a particular site.

274 ~ Michael L. Hargrave

Approaches to Ground Truthing Categorizing and Prioritizing Anomalies It is a common occurrence for geophysical surveys to identify far more anomalies than can be thoroughly investigated by means of formal test units. One is then faced with the practical problem of determining which anomalies to investigate. Two courses of action are recommended here. In some situations, it may be useful to categorize or prioritize the anomalies. In many situations, it will be wise to use a multistaged approach to ground truthing to systematically delete the less promising anomalies and to identify those that warrant investigation using the most invasive and expensive investigation techniques—typically, hand excavation of test units. The problem of too many anomalies is particularly troublesome and likely to occur in the context of NRHP eligibility assessments. Here the objective is generally to determine whether intact deposits are present and to assess their integrity and significance. Most NRHP assessments are conducted within the confines of a tight budget and schedule. What can the archaeologist do when his or her geophysical consultant provides a map that includes dozens or even hundreds of anomalies? Perhaps most important, the archaeologist should work with the geophysicist in selecting anomalies for ground truthing. The approach suggested here is most likely to be successful when there is an effective dialogue between the two specialists. It is unfortunate that this dialogue generally occurs via e-mail or telephone conversations, but distance does not preclude an effective exchange of information and ideas. Let us use as an example a situation in which approximately 100 anomalies are detected but the project budget will only support the excavation of 8 or 10 small (1-×-1-m) test units or a larger number of less expensive tests (e.g., shovel test pits). Some approach to sampling the anomalies is obviously necessary. If the archaeologist has no familiarity with geophysics, he or she might subdivide the site on the basis of the distribution of artifacts on the surface or in shovel tests, of topography, and/or of vegetation cover and then excavate a few of the anomalies that happen to be located in each division. In many cases, stratifying the anomalies instead of the site itself will be more productive. Here input from the geophysicist is extremely important, particularly if he or she is familiar with the local archaeological record. Criteria that should be considered in stratifying the anomalies include, for each anomaly, (1) dimensions (size and shape); (2) amplitude (magnitude of the data value); (3) discreteness; (4) sign (positive or negative); (5) location relative to other anomalies, site topography, or archaeological factors such as artifact distributions, test units, and so on; and (6) detection by multiple sensors. 1. Dimensions

As mentioned previously, geophysical anomalies are not necessarily coterminous with their subsurface sources. This is particularly true in magnetic data. Small, highly

Ground Truthing the Results of Geophysical Surveys ~ 275 magnetic objects such as iron or steel and, to a lesser extent, intensively burned soils can be manifested by strong anomalies. Depth and position are also important determinates of anomaly dimensions. One should not automatically give priority to testing anomalies whose circular shapes and diameters are consistent with pit features. Magnetic anomalies with these characteristics may be associated with small pieces of metal (Bevan 1998:25). In the absence of metal and other relatively magnetic artifacts, however, anomaly size can be informative about pit feature volume (Clark 1996:128). 2. Amplitude

Geophysicists frequently use the term amplitude in reference to the magnitude of a geophysical data value. Contrast is perhaps the key concept here. By definition, anomalies are loci characterized by data values that contrast with their immediate surroundings. At prehistoric sites in North America, cultural features may exhibit very modest contrasts with their surroundings. Thus, small-amplitude anomalies cannot be ignored. The ideal situation occurs at sites on which the cultural features exhibit a strong contrast with their surroundings. In these cases, the highest amplitude anomalies (or at least, in the case of magnetic anomalies, those that do not exhibit the dipoles often associated with metal) may be the best candidates to be cultural features. At many sites there may be a more or less continuous distribution of anomaly amplitudes. Here the type of map provided by the geophysicist is important. Gray-tone image maps are preferred by many geophysicists, but these can make it difficult to differentiate anomaly amplitudes. A proper use of color can be useful in categorizing anomalies on the basis of amplitude. Statistical threshold maps can be very useful at sites on which cultural features tend to be associated with the higher amplitude anomalies. Here one uses color to distinguish those anomalies whose amplitude exceeds some specified threshold. For example, features that exceed two standard deviations above the mean may be displayed in one color, whereas those that exceed three standard deviations are displayed in a second color. 3. Discreteness

By definition, an anomaly is a localized area that differs from its surroundings. If an anomaly is associated with a pit that was excavated into a relatively homogeneous soil matrix and then filled with culturally enriched soil, it may exhibit rather discrete boundaries. In contrast, anomalies that resemble the bull’s-eye of a target (i.e., that are surrounded by concentric bands of gradually decreasing amplitude) are less discrete than those that represent highly localized areas of higher amplitude in a field of relatively homogeneous lower values. 4. Sign

Some of the data-processing routines commonly used in preparing geophysical maps result in a data set with a mean of approximately zero. Thus, the sign of an

276 ~ Michael L. Hargrave anomaly can be useful in anomaly categorization and interpretation. For example, in an electrical resistivity survey, a pit whose fill holds moisture somewhat better than its surroundings may appear as a negative resistance anomaly. However, if the pit is characterized by coarse fill, it may hold moisture less and be manifested by a positive resistance anomaly. Sign can also be useful in interpreting and categorizing magnetic anomalies. For example, pit features often appear as weakly magnetic positive anomalies. The magnetic susceptibility of their fill has been increased through the addition of organic material, fragments of burned soil, and so on. At sites characterized by cultural midden or, at least, an intact A horizon, old (filled-in) test units often appear as negative magnetic anomalies. Here the organically and artifactually enriched soil in the unit has been partially removed. Stronger magnetic anomalies can result from remanent magnetism, which occurs when objects are heated above their Curie temperature. Remanent magnetism in the archaeological record may be associated with hearths, earth ovens, burned houses, the use of rock in cooking, and so on. At many sites, the factors that underlie an anomaly’s sign may be complex, and interpretations may be more difficult than in the examples presented here. Nevertheless, sign represents a useful criterion for categorizing anomalies. 5. Location

One should bring to the categorization process any archaeological information that may help categorize the anomalies in a meaningful way. For example, anomalies that occur in an area characterized by relatively abundant surface artifacts, near features identified in previous test units, or in favorable topographic settings may well be more promising than those for which such supportive evidence is absent. At many sites, there is a tendency for cultural features to be somewhat clustered. In single occupations, this clustering of features probably reflects the human tendency to conduct many activities in convenient proximity to one another. Through the course of multiple, temporally distinct components, features may cluster because of subtle factors that are difficult to perceive archaeologically: shade, wind, localized drainage, or simply the tendency to take advantage of previously cleared areas. In any case, distinct clusters of anomalies may be more promising than those that are isolated. Similarly, anomalies located in areas characterized by large trees or rodent activity may be viewed as less promising. The reader can, no doubt, imagine many valid exceptions to these broad generalizations. The point is to use all relevant information in categorizing anomalies. However, one should not lend disproportionate weight to archaeological criteria simply because one is more comfortable with them than with geophysical factors. 6. Detection by Multiple Sensors

One of the most useful criteria for categorizing anomalies is detection by multiple sensors. Those locations where anomalies are detected by more than one sensor are

Ground Truthing the Results of Geophysical Surveys ~ 277 particularly promising, at least in terms of the likelihood of the presence of a distinct subsurface source for the anomaly. For example, a weakly positive magnetic anomaly could be a pit or a piece of metal that is very small or buried relatively deeply. If a resistance anomaly is detected at the same location, one can feel somewhat more conďŹ dant that a subsurface feature may be present. It remains possible, of course, that the subsurface deposit in question may be nonarchaeological. A thorough discussion of surveying using multiple sensors is found in Kvamme, Johnson, and Haley, this volume. At any given site, some of these categorization criteria will prove to be more useful than others. It is not feasible to identify a single scheme for weighting the criteria. One should simply attempt to use these factors to assign the anomalies to a number of categories. It is not necessary, and in many cases may not be desirable, to divide the anomalies into a large number of subtly deďŹ ned categories. Anomaly categorization is easiest at sites that are not characterized by a scatter of nonarchaeological metal debris or other common sources of clutter such as tree roots or rodent activity. Ironically, categorization of anomalies is probably most important at the sites that do exhibit clutter yet where features are likely to be present. A Multistaged Approach to Ground Truthing The basic idea of a multistaged approach is to systematically evaluate the anomalies using a series of increasingly invasive and expensive techniques. The goal is to eliminate some of the anomalies at each stage. Consequently, the more expensive and invasive techniques are applied only to a minority of the anomalies that have survived the earlier stages of the screening process. Any of the excavation techniques used by professional archaeologists can potentially play a role in ground truthing geophysical surveys. The techniques to be used will depend upon the conditions encountered at the site, as well as the desired balance of the key factors (information return, cost, invasiveness, and so on). The multistaged approach described here has proven useful at prehistoric sites at Fort Leonard Wood, Missouri (Ahler, Asch et al. 1999; Ahler et al. 2003), Fort Campbell, Kentucky (Ahler, Schroeder et al. 1999), and Fort Bragg, North Carolina (Idol 2003; Idol and Pullins 2001). 1. Visual Inspection

With the geophysical map in hand, the archaeologist systematically inspects each anomaly selected for possible ground truthing. Here the goal is to identify and reject any anomaly that exhibits obvious evidence for a nonarchaeological origin. Common sources for such anomalies include vehicle ruts, localized depressions, localized areas of greater soil moisture, localized concentrations of gravel or apparently noncultural rock, tree roots, and so on. Note that coincidence may lead to the inappropriate rejection of an otherwise valid anomaly. For example, a rodent burrow may happen to occur at the location of an anomaly that is actually associated with a prehistoric pit. Such co-

278 ~ Michael L. Hargrave incidences will sometimes occur, but they do not offset the importance of a systematic visual inspection to ameliorate the common situation of having far more anomalies than can be investigated by means of controlled excavation. 2. Metal Detection

Next, a metal detector is used to investigate all remaining magnetic anomalies. At some sites, virtually all of the magnetic anomalies can be explained by the apparent presence of modern metal (a scatter of metallic debris is common on military installations). Note that the archaeologist does not excavate the anomaly in order to verify the presence of metal. It is assumed that most metal detectors will not detect prehistoric artifacts such as fire-cracked rock. If the metal detector indicates that metal is present and the focus is on prehistoric occupations, the anomaly is deleted from further consideration. It is recognized that in some cases a prehistoric feature could be present, with the presence of the metal being coincidental. Again, the goal of the multistage screening is to quickly eliminate anomalies unlikely to be associated with prehistoric features, not to resolve every possible co-occurrence of prehistoric and recent materials. 3. Soil Coring

Small-diameter soil (e.g., Oakfield) cores are next used to investigate the remaining anomalies. Ideally, cores should be taken from within the anomaly as well as outside of its apparent limits. Here the idea is to compare the two cores to better ascertain whether the core from within the anomaly exhibits any evidence for the presence of a feature. Evidence could include small flecks of carbon, burned soil, bone, a relative concentration of organic content, or small artifacts. Evidence for a feature could also be a very subtle difference in soil compaction, texture, or moisture. In many cases, one will not be able to determine whether a feature is present solely on the basis of the soil core. Thus, any anomaly in which there appears to be a contrast, however subtle, between the cores from within and from outside the anomaly should be retained for further investigation. If no contrast is noted, however, the anomaly should be deleted. 4. Shovel Tests

Shovel tests are simply small holes excavated to sterile soil or to a depth of 50 to 75 cm below the surface. Shovel tests used to identify possible features should be excavated in levels and all soil should be screened. This allows one to note in the field an increase in the abundance of artifacts or other cultural materials. As with soil cores, the shovel tests should be excavated in pairs, one within the anomaly and one outside its apparent limits. One side of the shovel test should be profiled and carefully inspected for any evidence of a cultural feature or concentration of artifacts or noncultural materials that could account for the anomaly.

Ground Truthing the Results of Geophysical Surveys ~ 279 5. Test Units

Finally, a small test unit is excavated in those locations where a shovel test has indicated the possible presence of a feature. These are typically 1 × 1 m and excavated in 10-cm levels. Some of these stages will need to be modified or omitted at some sites. The visual inspection and use of a metal detector are quick and easy and will almost always be productive (except in situations where nonarchaeological metal objects are very common). Test units will almost always be useful to verify the presence of a feature. However, test units may not be possible in situations where social and political factors require minimally invasive ground truthing. At many sites it may be desirable to excavate trenches rather than square units. The longer profiles provided by trenches will often increase the potential for detecting very subtle pits or other features. Soil cores and shovel tests are the most problematic components of multistaged ground truthing (Hargrave, Somers et al. 2002). They should be used if possible because of their low cost and minimal invasiveness. Unfortunately, soil cores and shovel tests are not likely to be reliable at sites characterized by very ephemeral features or at intensively occupied sites with deep, rich midden. In such situations it may be very difficult to detect features in small excavation units. Soil cores and shovel tests will prove most effective when the archaeologist is very familiar with the types of soils and archaeological features that are present at the site. Mechanized Excavations for Ground Truthing At first blush, geophysical instruments and backhoes might appear to be strange bedfellows. To explain the role of mechanized excavation in ground truthing, it is useful to consider reasons for the limited use of geophysics in U.S. archaeology. Most professional archaeologists have no formal training or previous experience in the use of geophysics. CRM in the United States is a highly competitive arena, and most CRM companies are hesitant to make the substantial investment in equipment purchase and training required by geophysics. In many cases, the State Historic Preservation Office (SHPO) staff is similarly unfamiliar with geophysics and thus does not advocate its use, even in those situations where it could be highly effective (Hargrave, Somers et al. 2002). One problem confronting SHPO staff is the difficulty of evaluating negative evidence in geophysical surveys. If a survey has been conducted and no evidence for features has been revealed, can it safely be assumed that no features are present? Was the survey properly conducted, using instruments and field methods appropriate to the site? Were the data properly processed and interpreted? Such questions will continue to delay the integration of geophysics into U.S. archaeology (particularly CRM) until SHPO staff, CRM practitioners, and university-based researchers become more familiar with the results of surveys conducted in their own regions. One way to familiarize archaeologists with geophysics would be to develop regional databases of geophysical surveys and ground truthing results. Perhaps the most com-

280 ~ Michael L. Hargrave pelling and persuasive type of ground truthing is the excavation of large, continuous areas. In many states, heavy equipment such as backhoes, track hoes, and even pan scrapers are used to remove the plow-disturbed strata. Subsurface features can then be mapped and excavated. In Illinois this approach has become standard practice in projects conducted by the Illinois Department of Transportation. Mechanized excavation has proven to be a cost-effective way to deal with the hundreds and in some cases thousands of cultural features at late prehistoric habitation sites in the American Bottom region and adjacent uplands. Mechanized excavation has permitted the exposure of entire settlements, providing opportunities to document and explain the evolution of large, relatively complex societies (Bareis and Porter 1983; Pauketat 2003). The use of mechanized stripping will not, in most situations, be a viable option for ground truthing geophysical surveys. Many archaeologists who employ mechanized stripping to excavate sites will ask, “Why use geophysics when we intend to expose and excavate all of the features?” However, to expedite the adoption of geophysics in the United States, departments of transportation and other state and federal agencies should consider the use of geophysics at sites slated for mechanized excavation. These geophysical surveys would, in some ways, represent an additional (albeit modest) expense. However, these surveys would provide valuable opportunities for archaeologists working in a region to quantify the degree to which various geophysical techniques can detect a wide variety of feature types. Such demonstration projects would, of course, have to be conducted in many regions because of the highly variable nature of site conditions and archaeological deposits. Only after seeing the results of many geophysical surveys followed by well-planned ground truthing will archaeologists in academic and CRM settings and SHPO offices be willing and able to effectively integrate geophysics into archaeology in the United States. Mistakes to Be Avoided Archaeologists who are not experienced in working with a geophysicist are likely to make a number of mistakes. Fortunately, most of these can easily be avoided. 1. Do not assume the geophysicist’s job is done when he or she provides you with a map. The geophysicist can and in many cases should provide substantial input into decisions about which anomalies to test. He or she should also be available (at least via telephone or e-mail) to discuss the ongoing results of ground truthing. A requirement for such consultation should be stipulated in the geophysicist’s Scope of Work so that the required time can be included in his or her budget. 2. Do not expect the subsurface phenomena manifested as geophysical anomalies to necessarily exhibit a sharp visual contrast with the surrounding soils or to be spatially coterminous with the anomalies. In some cases, geophysical sensors can detect evidence for subsurface phenomena that are exceedingly difficult to see, particularly in very small excavation units. When ground truthing a geophysical anomaly, one should

Ground Truthing the Results of Geophysical Surveys ~ 281 not merely look for an archaeological feature but should also look for any indications of localized differences in soil texture, compactness, color, mixing, and so on. Such observations should be recorded in some detail, as this may assist in the interpretation of similar anomalies. 3. Do not simply use your standard approach to excavation when ground truthing a geophysical survey. For example, noncompliance investigations of large habitation sites may often consist primarily of controlled surface collections followed by large block excavations. Incorporating some aspects of a multistaged approach to ground truthing is likely to make far better use of the information provided by a geophysical survey. 4. Do not focus all of your ground truthing efforts on anomalies. One of the biggest obstacles to a wider use of geophysics in CRM is the issue of negative evidence. Does the failure to detect anomalies mean that no features are present? It is very important to excavate in some areas where anomalies are absent and features are not expected to be present. 5. Do not limit your own role in geophysical investigations by failing to learn as much as possible from the geophysicist. By working with the geophysicist in the field and by discussing the basis for his/her interpretations of anomalies, the archaeologist can, through the course of several projects, learn enough to prioritize anomalies and conduct highly effective ground truthing. 6. Do not fail to provide the geophysicist with the draft and final reports. By reviewing pertinent sections of the draft, the geophysicist can help ensure a proper presentation of geophysical methods and findings. The results of ground truthing are important to the geophysicist, as they allow him or her to better understand how features are manifested under particular soil conditions. A geophysicist who is very familiar with a region’s archaeological record, soils, bedrock, and other characteristics will be able to do a better job of interpreting and prioritizing anomalies.

Ground Truthing Case Studies The remainder of this chapter presents a number of ground truthing case studies. Because the multistaged approach is advocated here, a fairly detailed example is provided of its use at Fort Leonard Wood, Missouri. Other briefer case studies focus on the successful use of a more narrow range of excavation techniques. See Table 12.1 for a summary of the usefulness of the ground truthing methods discussed above in relation to the sites in these case studies. Crying Hawk Site, Pulaski County, Missouri Fort Leonard Wood is located in a portion of the Ozark uplands that was somewhat marginal to major developments throughout prehistory. The installation includes many rock shelters and caves, some of them containing rich prehistoric deposits. Numerous open habitation sites have been evaluated for NRHP eligibility, but their role

282 ~ Michael L. Hargrave Table 12.1. Usefulness of ground truthing techniques at the sites discussed in this chapter Visual Inspection Crying Hawk Useful Grossmann Would not Hoxie Farm Would not Army City Would be Double Ditch Useful Fort Clark Useful Ellis Cemetery Useful Site

Metal Detection Useful Would be Would not Would be Would be Would be Would not

Oakfield Probe Useful Would be Useful Not Useful Useful Useful Unknown

Shovel Tests

Test Units

Useful Would be Would be Problematic Useful Useful Not appropriate

Useful Would be Would be Useful Useful Useful Not appropriate

Mechanized Stripping Would be Useful Would be Would be Would be Would be Not appropriate

Note: “Would be” means the technique was not used but would have been useful. “Would not” means the technique was not used and presumably would not have been useful.

in the evolving settlement and subsistence systems remains unclear. One factor making it difficult to interpret these open sites is the apparent paucity of discrete features such as pits, hearths, and architectural remains. It is unclear whether the scarcity of features reflects the ephemeral nature of prehistoric occupation or is simply a bias resulting from a site evaluation strategy based on shovel tests and a small number of 1-m2 test units. Since 1997, the Construction Engineering Research Laboratory (CERL) and Fort Leonard Wood have used geophysics and ground truthing to better document the occurrence of features at open habitation sites (Hargrave 1999; Hargrave, ed. 1999; Hargrave, Somers et al. 2002; Mathys and Maki 1997). Crying Hawk site (23PU556) is situated on an eroded Pleistocene terrace remnant overlooking the Big Piney River. Small oak and cedar trees cover most of the site. From the 1940s into the 1960s, the site area was used for military training. Indications of this use include a scatter of rusted military food tins and other metallic debris as well as fighting positions (“foxholes”) and vehicle ruts. Approximately one-third of the site has been severely damaged by topsoil borrowing (Ahler, Asch et al. 1999:84–85). Magnetic field gradient and electrical resistivity surveys conducted in 1997 by the Institute for Minnesota Archaeology Consulting (IMAC) each covered an area of 3,100 m2 (Mathys and Maki 1997) (Figure 12.1). Ground-penetrating radar (GPR) was used in a limited way to further investigate several magnetic and resistivity anomalies. Mathys and Maki numbered 57 magnetic and 60 resistivity anomalies (additional anomalies were detected but viewed as less promising and were not numbered). Prioritization of the magnetic anomalies emphasized amplitude, sign, and distribution. Many of the magnetic anomalies were relatively strong and dipolar, suggesting the presence of metal objects. However, samples of sandstone collected at the site were magnetically enhanced, suggesting that culturally or naturally heated rock could account for some of the magnetic anomalies (Mathys and Maki 1997).

Ground Truthing the Results of Geophysical Surveys ~ 283

Figure 12.1. Results of electrical resistivity survey at the Crying Hawk site.

284 ~ Michael L. Hargrave The resistivity anomalies were prioritized on the basis of the strength of their contrast with their surroundings, size, sign, and distribution (clustering). The smaller resistivity anomalies, particularly those exhibiting a low (negative) resistivity, were viewed as possible pits, whereas some of the larger anomalies were described as possible midden areas. Clustering was viewed as a possible indication that the anomalies were associated with pits created during particular occupational components. Five magnetic and five resistivity anomalies were identified as representing the top priority for ground truthing (Mathys and Maki 1997). The Illinois State Museum Society (ISMS) conducted archaeological ground truthing of the geophysical survey in the context of an NRHP eligibility assessment of the site. Prior to ground truthing, the ISMS project director inspected the IMAC geophysical maps and added an additional 24 (3 magnetic and 21 resistivity) anomalies that had not been numbered or recommended for ground truthing. This action increased the total number of magnetic anomalies to 60 and the total number of resistivity anomalies to 81 (Ahler, Asch et al. 1999:95). A multistaged approach (described above) was used to ground truth the geophysical anomalies at Crying Hawk. Forty-nine of the magnetic anomalies (including the five high-priority anomalies) were eliminated as a result of visual inspection and use of a metal detector. Each of the remaining 11 magnetic anomalies was investigated using a transect of three soil cores. In each case, the central core was positioned near the mapped center point of the anomaly and the two flanking cores were located outside the anomaly. Four of the magnetic anomalies were eliminated as a result of disturbed or truncated soil horizons. The remaining seven magnetic anomalies exhibited normal soil profiles in the soil cores. Two of these anomalies were deleted because they were very near a tree tip-up or vehicle ruts, and two others were eliminated because of proximity to bedrock thought to contain ferrimagnetic oxides. The remaining three magnetic anomalies warranted further investigation using shovel tests (Ahler, Asch et al. 1999:93–94). Of the 81 resistivity anomalies, 46 were eliminated following visual inspection because of proximity to vehicle ruts, large tree roots or tree tip-ups, military foxholes, large rodent burrows, or significant erosion. The remaining 35 resistivity anomalies were then examined using transects of three soil cores. Fifteen of the resistivity anomalies were eliminated on the basis of very thin or absent cultural horizons. The 20 remaining anomalies exhibited normal soil profiles and warranted additional investigation using paired shovel tests (Ahler, Asch et al. 1999:98). Unfortunately, the funds allocated to ISMS for the archaeological investigations were not sufficient to support so many additional shovel tests. The Scope of Work and project budget permitted the shovel-test investigation of nine of the remaining anomalies. Given this, ISMS and IMAC conferred as to which 9 of the remaining 3 magnetic and 20 resistivity anomalies most warranted further investigation. ISMS had noted evidence for vehicle tracks that had not been visible at the time of the geophysical survey and felt that these tracks could have compressed the soil strata and thus cre-

Ground Truthing the Results of Geophysical Surveys ~ 285 ated resistivity anomalies. The geophysicists were still optimistic about the potential significance of clustering among the resistivity anomalies. They were more skeptical as to whether the vehicle tracks could have created the type of discrete anomalies under consideration. A compromise was achieved wherein ISMS agreed to investigate one anomaly from each cluster, even if it was located near vehicle tracks and had thus been previously eliminated (Ahler, Asch et al. 1999:98). Two shovel tests were excavated to investigate each of the nine selected anomalies. In each case, one shovel test was positioned within the anomaly and the other was located outside its mapped limits. No definite features were identified in any of the shovel tests. In fact, the density of tree roots made it difficult for the archaeologists to feel confident that very subtle variation in soil color, texture, and so on could be reliably detected in the shovel test profiles. In retrospect, the ISMS archaeologist would have preferred to use very small (e.g., 0.75 × 0.75 or 0.5 × 1 m2) test units rather than paired shovel tests. Despite these difficulties, four of the anomalies were selected for further investigation using 1-m2 test units (Ahler, Asch et al. 1999:100–101). Criteria for investigation using test units included stronger soil structure, greater artifact density, and the presence of diagnostic artifacts in the shovel test located inside the anomaly. One of the four test units (TU6) encountered a very shallow depression at the BE/Bt horizon boundary. While this could conceivably have been the remains of a shallow pit basin, its fill did not exhibit any of the characteristics of a cultural feature. A second test unit (TU7) encountered a fairly large (ca. 35 cm long) taproot that probably accounted for the resistivity anomaly. The third unit (TU8) produced no feature or discrete disturbed area that could account for the resistivity anomaly. However, tree roots were particularly abundant in this unit, making it difficult to achieve good floors for the four excavation levels (Ahler, Asch et al. 1999:105–107). The remains of a disturbed feature were documented in the fourth test unit (TU5). No plow zone or midden was present in this unit. The A, EB, and BE horizons were characterized by relatively abundant artifacts. Diffuse areas of relatively compact silt loam sediments (designated as Feature 1) were detected and mapped at the base of Level 1. In contrast to the unit as a whole, no fire-cracked rock or other artifact concentrations were associated with the feature. To verify whether Feature 1 genuinely represented a discrete cultural feature, a 100-g sediment sample was subjected to a full suite of soil particle size and quantitative chemical analyses at the University of Missouri Soil Characterization Laboratory. The Feature 1 sample was characterized by concentrations of calcium, magnesium, potassium, and phosphorous that were not high in absolute terms but were nevertheless 150 to 300 percent greater than the concentrations found in the surrounding soil matrix. As stated in the report, “These findings support the field interpretation of Feature 1 as the remains of a hearth or living area that has been enriched by the by-products of human activities” (Ahler, Asch et al. 1999:105). By most archaeological standards, Feature 1 does not represent a particularly interesting feature. It does not provide a sample of directly dateable material; does not

286 ~ Michael L. Hargrave represent a dramatic concentration of artifacts, floral, or faunal remains; and is very poorly preserved. The Crying Hawk feature is perhaps most important in methodological terms. Its discovery demonstrates that the multistaged approach does provide an effective way to weed out the less promising anomalies, systematically allocating the more expensive and invasive excavation units to those anomalies that are more likely to be associated with cultural features. Do the results of ground truthing suggest that Feature 1 is the only cultural feature present in the surveyed area at Crying Hawk? This would be implied if we could assume that each decision about eliminating anomalies from further consideration was correct and that all features present were manifested by geophysical anomalies. Realistically, these assumptions are not justified. The presence of at least one feature suggests that at least some others may also be present in the surveyed area. The systematic ground truthing also seems to indicate, however, that features are clearly not abundant at the site. Thus, the program of geophysical survey and ground truthing at Crying Hawk was not exciting, but it was successful in addressing the issue of the occurrence of features at this open site. Grossmann Site, St. Clair County, Illinois The Grossmann site exemplifies the use of mechanized removal of the plow zone and hand excavation of all exposed features to quantify the success of a geophysical survey. Grossmann is an early Mississippian period settlement located in the uplands about 18 km east of Cahokia. In 1998, Illinois Transportation Archaeological Research Program (ITARP) archaeologists excavated 17 structures located in the right-of-way of a new road. A controlled surface collection of artifacts suggested that the main portion of the site was located outside the right-of-way. I conducted a magnetic field gradient survey of the site in 2001. Susan Alt (University of Illinois at Urbana-Champaign) then directed large-scale excavations at Grossmann, under the aegis of Tim Pauketat’s University of Illinois Field School in Archaeology and the National Science Foundation–funded Richland Archaeological Project (Pauketat 2003). An area of 3,188 m2 was mechanically stripped, exposing 42 early Mississippian period structure complexes (55 building episodes) and 58 other features (pits, post pits, truss trenches, hearths, and so on) (Alt 2002; Hargrave, Alt et al. 2002). It was known when the magnetic survey was done that a good portion of the Grossmann site would be excavated. The project was viewed as a useful opportunity to quantify the extent to which various feature types would be detected in a geophysical survey. Numerous rectangular house basins, a few wall trench patterns, and a considerable number of probable pits can easily be identified on the magnetic map (Figure 12.2a). When one overlays onto the geophysical data a map showing the outlines of excavated features (Figure 12.2b), it is easy to “see” anomalies that might not be identified if the excavation data were not available. One of the challenges confronting those who conduct geophysical surveys of very

Ground Truthing the Results of Geophysical Surveys ~ 287

Figure 12.2. The Grossmann site: (a) the results of a magnetic ďŹ eld gradient survey; (b) the limits of subsequent excavation shown in green, prehistoric houses in yellow, and pits and other features in red (color illustration shown on the CD) (excavation data courtesy of Susan Alt, University of Illinois).

288 ~ Michael L. Hargrave low-contrast sites (such as Grossmann) is the need to detect anomalies in an objective, replicable manner. A statistical approach to feature detection provided the objectivity needed to quantify rates of feature detection in the Grossmann magnetic survey. First, the standard deviation (SD) was calculated for a portion of the site where few anomalies were present. Several maps were then produced using different thresholds: 1, 2, 3, and 4 SDs. For example, on one of the maps, all values greater than 4 SDs above the mean were plotted in a distinctive color. A feature was assumed to have been detected if it overlapped to any degree with one of the >4 SD anomalies. Use of lower thresholds resulted in a larger number of anomalies whereas use of higher thresholds reduced the anomaly count as well as anomaly size. The objective of the exercise was to identify which threshold would maximize the percentage of features detected while minimizing the occurrence of false positives (i.e., areas where anomalies were detected but no feature was identified during excavation) (Hargrave, Alt et al. 2002). About 60 percent of the structures and 40 percent of the other features were detected using the best (4 SD) threshold model. Fewer than 20 percent of the anomalies were false positives. One would, of course, hope for an even higher rate of feature detection. This might well have been achieved if a second geophysical technique had also been used across large portions of the site. The detection rates that were achieved, however, would have permitted reasonably accurate inferences about site organization and feature density. Had the site been investigated using hand-excavated test units rather than mechanized stripping, the geophysical survey would have provided an excellent basis for locating units in highly productive areas. Hoxie Farm Site, Cook County, Illinois The Hoxie Farm site is a large, Upper Mississippian occupation located south of Chicago near Lansing, Illinois (Hargrave et al. 2004; Jackson 2003). An embankment associated with Interstate 80 bisects the site. From 2000 through 2003, ITARP excavated portions of the site on both sides of the highway that will be impacted by road improvements. One of the components, a fortified village dating to the fourteenth and early fifteenth centuries a.d., is spatially discrete and largely unmixed with other occupations. The paucity of artifacts and limited amount of feature superpositioning suggest that this village represents a relatively brief occupation. A defensive complex consisting of four shallow ditches and a single palisade defines the western limits of the village. On the east side, a natural ravine may have obviated the need for defensive ditches. The north edge of the site lies beneath the I-80 embankment, and an artificial (borrow pit) lake obscures the southern edge of the settlement. Excavations in the narrow east–west right-of-way revealed more than 80 small pithouses and more than 300 pits. The structures appear to be arranged in an orderly manner, with some suggestion that they occur in closely spaced rows or arcs. Structures were more common at the east end of the excavated area whereas pits were more common to the west (Hargrave et al. 2004).

Ground Truthing the Results of Geophysical Surveys ~ 289 I conducted an electrical resistance survey south of the excavated area in order to better document the extent and arrangement of features. The deepest of the defensive ditches was clearly detected at the west end of the survey area, and numerous resistance anomalies whose sizes and shapes were consistent with structures were detected at the east end. Surprisingly, very few anomalies were detected in the central area, even though nearby excavations had encountered features. The existence of a plaza-like open area was not consistent with archaeological expectations and it was suspected that the paucity of resistance anomalies could be related to seasonal differences in soil moisture. The resistance survey was conducted during three visits in March, June, and November, and the central area had been surveyed in June, when the soil was relatively dry. In an effort to minimize the effects of this possible bias, the resistance data collected during different seasons were reprocessed separately and lower thresholds were used in an effort to detect anomalies in the central area. This effort was successful, but relatively low amplitudes and asymmetrical shapes characterized the anomalies detected there. Because the resistance survey occurred in a portion of the site that lies in a county forest preserve, ground truthing had to be minimally invasive. Previous excavations at the site had demonstrated that pits and pithouses were typically characterized by very dark, organically enriched fill, whereas the surrounding soil matrix was a light-colored sandy loam. In some areas the soil proved to be extremely compact. Nevertheless, soil cores obtained using a 1-inch-diameter Oakfield soil sampler proved to be a very effective means of ground truthing. Ground truthing began with the use of paired soil cores, one excavated within a resistance anomaly and the other located just outside its apparent limits. Figure 12.3 illustrates this approach in a very small portion of the overall survey area. On the figure, red dots indicate that the soil core encountered feature fill whereas green dots indicate an absence of fill. The initial paired cores that are shown clearly suggested that the low-resistance (white) anomalies represent features whereas the high-resistance (black) anomalies do not. To better assess the reliability of the resistance survey results, ITARP ultimately excavated more than 2,000 soil cores at 2-m intervals within a number of the 20-×-20m survey blocks. The ground truthing demonstrated that the resistance survey results were more reliable in the areas surveyed during the spring and fall. However, a map dating to the time of the original highway construction indicated a plan to stockpile topsoil in the central area, and activities related to this may have diminished the potential for detecting features in the resistance survey there. Finally, the soil core results indicated that features in the central area tended to be slightly smaller, shallower, and less numerous, and these factors made them more difficult to detect. Ultimately, each of three factors (seasonal variation in soil moisture, possible impacts from road construction, and variation in feature characteristics) may have contributed to the spatial bias in reliability of the resistance results. The ground truthing excavations at Hoxie Farm played an essential role in investigating the site’s spatial extent and the possible causes of a bias in the resistance data.

290 ~ Michael L. Hargrave

Figure 12.3. Results of mechanized stripping, resistance survey, and soil cores at the Hoxie Farm site (color illustration shown on the CD) (excavation data courtesy of Illinois Transportation Archaeological Research Program, University of Illinois).

Ground Truthing the Results of Geophysical Surveys ~ 291 Used together, the resistance and ground truthing data provided substantially more information about site limits and the intrasite distribution of features than would either data set when used independently (Hargrave et al. 2004). Army City, Riley County, Kansas Army City was a civilian-owned entertainment center that provided goods and services to the soldiers training at Fort Riley during the First World War. It included several large movie theaters, hotels, restaurants, pool halls, barbershops, photography studios, and stores selling clothing, jewelry, and a wide variety of other merchandise. The army, like American society at the time, was segregated, and a portion of the complex south of the railroad tracks (“Army City South”) was set aside for the African American troops (Rion 1960). Army City was a planned community that was conceptualized, platted, and built over a very brief time to serve a particular group of customers. All of the buildings were required to conform to a Spanish Mission architectural style. Some buildings were constructed on concrete pads whereas others were set on concrete or stone piers. Army City’s wide streets were not paved, but the main business district boasted concrete sidewalks, electric streetlights, and telephones in many buildings. Water and sewer pipes ran down the alleys (Rion 1960). The Army City merchants flourished during the war years, but business declined precipitously after the armistice, when the influx of trainees ceased. A portion of the main business district burned in 1920 and Army City never recovered. Most of the buildings were sold as scrap, although several were moved intact to nearby Ogden (Hargrave, Somers et al. 2002; Rion 1960). As a result of Fort Riley’s expansion in 1942, the archaeological site of Army City is now located on the installation. Most of the site lies in a hay field, with few indications on the surface of the short-lived but boisterous community that was once there. In 1996, Fort Riley and CERL undertook a geophysical and archaeological investigation of the Army City site. Project goals were to assess the site’s eligibility for the NRHP, secure a detailed map of the archaeological deposits that could be used to minimize the impact of future infrastructure improvements, and evaluate the potential role of geophysical survey in the installation’s CRM program. A small-scale geophysical survey conducted by Geoscan Research USA in 1996 included electrical resistance, magnetic field gradiometry, and GPR. Resistance appeared to provide the most useful information, and in 1997 Geoscan returned to survey a large portion (9.2 ha) of the site. The large-area, low-data-density (1 reading per square meter) resistance survey provided an excellent map of the archaeological remains of those portions of Army City north of the railroad tracks. Sidewalks are clearly manifested by linear positive (dark) resistance anomalies, whereas utility trenches containing metal pipes appear as negative (white) linear anomalies (Figure 12.4). Some of the building locations are indicated by very strong positive anomalies associated with dense deposits of concrete and

292 ~ Michael L. Hargrave

Figure 12.4. Electrical resistance map of the Army City site, Fort Riley, Kansas.

other building materials. Other building locations are manifested by small anomalies associated with in situ piers or can be inferred on the basis of the arrangement of the anomalies associated with sidewalks and utility pipes. Also visible on the map are linear low-resistance anomalies related to vehicle ruts along the route used by trucks that hauled away debris from the 1920 fire and the meandering course of a filled-in stream bed. Other strong but amorphous anomalies appear to relate to geomorphic features of the Kansas River floodplain (Hargrave, Somers et al. 2002). Two consulting firms, Public Service Archaeology Program (PSAP) and LTA, Inc., conducted small-scale ground truthing excavations at Army City. The first season of ground truthing focused on what proved to be a mixed residential and hotel district. PSAP was requested to use shovel tests to investigate a number of the positive and negative anomalies. The shovel tests did encounter one in situ feature (a concrete slab) and demonstrated that the high-resistance anomalies were characterized by a relatively

Ground Truthing the Results of Geophysical Surveys ~ 293 greater (but in absolute terms, modest) occurrence of artifacts and disturbed soils than the negative anomalies. The archaeologists found, however, that shovel tests did not provide enough exposure to permit identification of the types of features or other deposits that were encountered (Kreisa and Walz 1997). The following year, LTA, Inc., was requested to use shovel tests, small test units, and trenches to ground truth anomalies in the main commercial district. Here, too, shovel tests were not fully satisfactory. In this part of the site, architectural debris and other artifacts were so abundant that the shovel tests did not always permit one to differentiate actual structure locations from dense deposits of secondary debris. Narrow trenches proved to be much more effective than alignments of closely spaced shovel tests, but at an increased cost. Small (1-×-1-m) test units were also less than ideal. In some cases, these encountered impenetrable concentrations of concrete debris. The LTA archaeologists suggested that, in future investigations, test units that exposed at least 2 m2 should be used to investigate large anomalies believed to be associated with architectural remains. While the LTA excavations were under way, I attempted to use an Oakfield soil sampler and a standard tile probe to quickly verify the presence or absence of architectural debris. Unfortunately, the soil was so dry and hard that summer that probes were not a viable ground truthing technique (Larson and Penny 1998). The sheer abundance of architectural debris in Army City’s main commercial district made all types of excavation a laborious undertaking. Techniques such as soil cores and shovel tests that represent low-cost options at some sites were problematic at Army City. Mechanized trenches were not used at Army City in 1996–1997 but, in retrospect, would have represented a very cost-effective approach to ground truthing. Army City also provided an opportunity to ground truth the geophysical results using contemporary maps and photographs. A map made by the U.S. Army in 1919, when many of the businesses in Army City were still operating, showed 13 structures north of the railroad but only 2 structures south of the tracks. In contrast, the resistance map indicated the presence of at least 17 structures north of the tracks (very little surveying was done south of the railroad). Several factors may account for the apparent inaccuracy of the 1919 map, including the construction of some buildings after the map was completed and less rigor on the part of army surveyors when dealing with infrastructure outside the installation boundary (Hargrave, Somers et al. 2002). A panoramic photograph of Army City taken from the bluffs north of the site about 1918 provides the best archival basis for evaluating the reliability of the geophysical map (Figure 12.5). At least 49 buildings are discernable in this photograph, and 11 of these are south of the railroad tracks. Some of the 38 structures north of the tracks were outside the survey area, but several others were not identified in the geophysical survey. Several of the buildings that are either barely detectable or “missing” from the geophysical map may be those that were moved intact rather than disassembled and sold as scrap (Hargrave, Somers et al. 2002).

294 ~ Michael L. Hargrave

Figure 12.5. Panoramic photograph of Army City, ca. 1918 (courtesy Kansas State Historical Society).

The panoramic photograph greatly enriches our understanding of Army City, providing detailed information on the appearance and design of building superstructures, as well as other facts. Useful as the panoramic photograph is, however, it too has many biases. Because it was taken at a highly oblique angle, many buildings are obscured by those in the foreground. Buildings that had not yet been built obviously do not appear in the photograph. In contrast, the geophysical data should hypothetically include some traces of all of the buildings that have ever stood in the surveyed area. In reality, of course, some buildings are manifested by exceedingly low-contrast resistance anomalies. Buildings not detected in the resistance data might well have been detected if other instruments had been used. For example, a multiple-instrument survey of the main commercial district conducted by Kvamme, Ernenwein, and others in 2002 provides a significantly finer level of detail than does the 1996–1997 resistance survey (see Kvamme, Johnson, and Haley, this volume, for a discussion of the multiple-instrument survey at Army City). Panoramic photographs such as the 1918 image of Army City are a valuable resource for historic archaeologists. Unfortunately, such photographs are not available for most projects. Archaeologists investigating historic sites, particularly those dating to the post-1900 era when panoramic photography was popular and many people owned and used simple box (e.g., “Brownie”) cameras, should always investigate the possibility of using contemporary photographs to ground truth or otherwise supplement the information provided by geophysical surveys.

Ground Truthing the Results of Geophysical Surveys ~ 295 Double Ditch Site, North Dakota Double Ditch is a large prehistoric (ancestral Mandan) fortified earthlodge village located north of Bismark, North Dakota. The site was occupied from approximately the late 1400s to the mid-1780s, when a smallpox epidemic probably led to site abandonment. The Double Ditch site is manifested on the surface by midden mounds and shallow house depressions. Ken Kvamme’s initial geophysical investigations in 1997 indicated the site was an excellent candidate for more extensive mapping (Kvamme 2000). Kvamme directed large-scale, multiple-instrument geophysical surveys at Double Ditch during the summers of 2001–2004. Results of geophysical surveys in 2002 included the discovery of two additional defensive ditches that had no surface manifestations and that dramatically increase the size of the site (Ahler et al. 2002). Ground truthing investigations were conducted by Stanley Ahler (Paleo Cultural Research Group), Fern Swenson (State Historical Society of North Dakota), and W. Raymond Wood (University of Missouri). Ground truthing at Double Ditch has included the use of soil cores, test units, and trenches. Figure 12.6 shows the results of a trench that was hand-excavated across one of the fortification ditches. Completely invisible on the surface, this ditch was readily detected in the magnetic survey on the basis of its content of relatively magnetic topsoil, midden, and occasional rocks. Although the ditch might have been verified using less intensive excavation units, the trench profile provides valuable information about the ditch’s morphology and the manner in which it was filled. The Double Ditch site exemplifies a situation in which a well-executed, large-area, multiple-instrument survey has provided imagery in which many elements of a complex site are interpretable without ground truthing. Readily interpretable features include fortification ditches, bastions, houses, storage pits, hearths, and previous excavation units (Kvamme 2002b, 2003a). Researchers can extract more information from these maps than they could just a few years ago as a result of surveys covering increasingly large areas and the synthesis of imagery from diverse instruments. Ground truthing continues to play an important role in these studies, not only in better interpreting the subtle nuances of variation in the imagery but also in effectively targeting excavations needed to recover the artifacts, food remains, and datable samples that will permit studies of the occupational history and social life of the settlement (Ahler and Swenson 2001; Ahler et al. 2002). Fort Clark Trading Post, North Dakota Fort Clark was an important trading center along the upper Missouri River between about 1830 and 1860. Founded by the American Fur Company, it was located near a Mandan village. Ken Kvamme, Jami Lockhart, and others conducted multiple-instrument geophysical surveys at Fort Clark in 2000 (Figure 12.7; Kvamme 2001b, 2002a, 2003b:444). A rich array of archival sources is available for this site, including paintings by Karl Bodmer and George Catlin, a plan drawn by Prince Maximilian, and a sketch by William Hays.

296 ~ Michael L. Hargrave

Figure 12.6. Trench excavated to ground truth a fortiďŹ cation ditch at the Double Ditch site (courtesy of Kenneth Kvamme, Fern Swenson, and Stanley Ahler).

Ground truthing investigations at Fort Clark were conducted by William J. Hunt, Jr. (National Park Service, Midwest Archeological Center). A block of contiguous test units in the northern bastion (Figure 12.8) revealed the presence of a large number of foundation stones approximately 50 cm below the surface. The presence of this highly magnetic sandstone helps explain the crisp deďŹ nition of architectural elements in the magnetic map (Kvamme 2003b:444). Excavation trenches dug by archaeologists in the 1970s add some confusion to the resistance maps, underscoring the destructive aspect of archaeological excavation.

Ground Truthing the Results of Geophysical Surveys ~ 297

Figure 12.7. Magnetic map of Fort Clark Trading Post, North Dakota (courtesy of Kenneth Kvamme).

Ellis Cemetery, Fort Bragg, North Carolina Ellis Cemetery is a small family cemetery located on what is now Fort Bragg. Ellis is one of four historic cemeteries mapped in 2002 by Archaeo-Physics LLC using a pulseEKKO 1000 GPR system with a 450-MHz antenna (Jones et al. 2003). The purpose of the study was to determine whether areas within some of the cemeteries that are devoid of grave markers are also devoid of graves and whether the existing fences are properly located to protect all graves from possible disturbance. Ground truthing by means of excavation was not appropriate in this project, given that there are no plans to move or otherwise disturb the graves. The locations of all grave markers (headstones and footstones) were mapped using an electronic distance measurement (EDM) instrument. Marker locations were overlain onto the geophysical maps in order to assess the correlation between markers and anomalies that could be

298 ~ Michael L. Hargrave

Figure 12.8. Map showing magnetic foundation stones documented in a block of contiguous test units at Fort Clark Trading Post (courtesy of William Hunt, Jr.).

related to graves. It was recognized that through the years some markers might have been removed from the cemetery whereas others may have been moved and then reinstalled in the wrong location. The GPR survey of Ellis Cemetery produced excellent results (Figure 12.9). The gravestones were found to be well aligned and closely spaced, and nearly all of the stones proved to be associated with a well-deďŹ ned anomaly. An open area in the northwest portion of the fenced area was devoid of both grave markers and anomalies suggestive of graves. Interestingly, however, at least one anomaly south of the fence and several anomalies east of the fence are likely to represent unmarked graves. This situation could reďŹ&#x201A;ect an improper relocation of the fence at some point in the past. It is also conceivable that the anomalies outside the fenced area are associated with the graves of servants. Careful mapping of the locations of extant grave markers is a useful noninvasive option for ground truthing the results of geophysical surveys of cemeteries. One must be aware that in cemeteries that have not seen continuous maintenance, gravestones may have been re-

Ground Truthing the Results of Geophysical Surveys ~ 299

Figure 12.9. GPR map of Ellis Cemetery showing the location of gravestones (color illustration shown on the CD).

moved, and some may have been reinstalled in the wrong location. The presence of a gravestone does not, therefore, represent an absolutely reliable basis for evaluating anomalies as possible graves. Where one does find a strong positive correlation between markers and anomalies, one can be more confident that the absence of an anomaly indicates the absence of a grave. One must still consider the synchronic and diachronic variation in burial practices (e.g., the presence and nature of coffins and vaults) and the implications this variation may have for grave detection. For example, graves with metal coffins and/or vaults should be much easier to detect than graves that lack those features.

Summary While ground truthing is not, strictly speaking, an essential component of a remote sensing study, it should be conducted whenever possible. Effective ground truthing can dramatically improve the reliability of interpretations of geophysical anomalies, greatly increasing the archaeological usefulness of a geophysical map. In many cases, the cost savings and improved information return that often represent the motivation for using remote sensing cannot be optimized without effective ground truthing. It has been emphasized here that the archaeological techniques used in ground truthing must be selected on the basis of site characteristics (Table 12.1), as well as according to the relative importance of five factors: (1) information return, (2) cost,

300 ~ Michael L. Hargrave (3) invasiveness, (4) social and political issues, and (5) risks to personnel. At many sites, a staged approach to ground truthing excavations will be effective. Stages that have proven effective in the Midwest include (1) visual inspection, (2) metal detection, (3) small-diameter soil coring, (4) shovel test pits, and (5) small test units. The effectiveness of soil cores and shovel tests depends on the archaeologist’s ability to recognize cultural deposits in these small units. Cultural features can be very difficult to recognize in small units in depositional situations where organic stains have disappeared or where discrete features are obfuscated by rich midden. Ground truthing programs that rely heavily on small excavation units (soil cores, shovel test pits, and so on) are best conducted by individuals with substantial excavation experience in the region. As remote sensing in the United States matures, we are beginning to appreciate the breadth of its potential contributions to archaeological inquiry. Those who wish to thoroughly integrate geophysics into their research and CRM programs may benefit from viewing the relationship between excavation and geophysical survey as a continuum. At one extreme, small-scale, single-sensor surveys can help the archaeologist increase the information return and reduce the costs and invasiveness of his or her excavations. At the other end of the spectrum, large-area, multisensor surveys such as those at Double Ditch, Fort Clark, and a number of other sites discussed in this volume do not simply help the archaeologist to dig more efficiently. Rather, these comprehensive remote sensing studies can play the lead role in multidisciplinary research programs, providing images of entire settlements that define the context for a wide array of specialized studies, most of which will require small-scale, carefully targeted excavations.

Acknowledgments Thanks go to Stanley Ahler, Susan Alt, Thomas Emerson, Michael Farkas, William Hunt, Jr., Doug Jackson, Kenneth Kvamme, Tim Pauketat, and Fern Swenson for permission to use graphics from their excavations. Steven Ahler has played a major role in developing the staged approach to ground truthing in his excavations conducted at Fort Leonard Wood sites. Thanks also go to Rinita Dalan, Jay Johnson, Kenneth Kvamme, Lewis Somers, and other authors of this volume for comments that improved this chapter. Finally, I thank Jay Johnson for immediately recognizing the appropriateness of including a chapter on ground truthing in a volume devoted to archaeological remote sensing.

References Cited Ahler, S. A., and F. E. Swenson 2001 Geophysics, Hand Coring, and Behavioral Organization at Double Ditch Village, North Dakota. Paper presented at the Plains Anthropological Conference, Lincoln, Nebraska.

Ground Truthing the Results of Geophysical Surveys ~ 301 Ahler, S. A., W. R. Wood, and F. E. Swenson 2002 A Century after Will and Swinden: Excavations at Double (Double) Ditch Village. Paper presented at the Plains Anthropological Conference, Oklahoma City, Oklahoma. Ahler, S. R., D. L. Asch, D. E. Harn, B. W. Styles, K. White, C. Diaz-Granados, and D. Ryckman 1999 National Register Eligibility Assessments of Seven Prehistoric Archaeological Sites at Fort Leonard Wood, Missouri. Submitted to U.S. Army Construction Engineering Research Laboratory, Champaign, Illinois, by Illinois State Museum Society, Quaternary Studies Program, Technical Report 98-1202-28, Springfield. Ahler, S. R., M. L. Hargrave, M. F. Kolb, and M. B. Schroeder 2003 National Register Evaluation of Five Prehistoric Stratified Archaeological Sites at Fort Leonard Wood, Missouri. Submitted to the Engineer Research and Development Center Construction Engineering Research Laboratory, Champaign, Illinois, by Illinois State Museum Society Landscape History Program, Technical Report No. 2003-1486-3, Springfield. Ahler, S. R., M. B. Schroeder, and K. White, with contributions by D. Johnson and C. A. Dobbs 1999 National Register Eligibility Assessment and Geophysical Investigation of Site 40MT28, Fort Campbell, Tennessee/Kentucky. Submitted to U.S. Army Construction Engineering Research Laboratory, Champaign, Illinois, by Illinois State Museum Society, Quaternary Studies Program, Technical Report 98-1247-28, Springfield. Alt, S. M. 2002 Identities, Traditions, and Diversity in Cahokia’s Uplands. Midcontinental Journal of Archaeology 27(2):217–236. Bareis, C., and J. W. Porter (editors) 1983 American Bottom Archaeology: A Summary of the FAI-270 Project Contribution to the Culture History of the Mississippi Valley Region. University of Illinois Press, Urbana-Champaign. Bevan, B. W. 1998 Geophysical Exploration for Archaeology: An Introduction to Geophysical Exploration. Special Report No. 1. U.S. Department of the Interior, National Park Service, Midwest Archeological Center, Lincoln, Nebraska. Clark, A. J. 1996 Seeing Beneath the Soil: Prospecting Methods in Archaeology, new ed. B. T. Batsford, London.

302 ~ Michael L. Hargrave Crawford, O. G. S. 1924 Air Survey and Archaeology. Ordnance Survey Professional Papers, New Series, 7. Ordnance Survey, London. Hargrave, M. L. 1999 A Comparison of Traditional and Geophysical Strategies for Assessing the National Register Status of Archaeological Sites at Fort Riley, Kansas. Special Report 99/22/ January 1999. U.S. Army Construction Engineering Research Laboratory, Champaign, Illinois. Hargrave, M. L. (editor) 1999 Geophysical and Archaeological Investigations of Historic Sites at Fort Riley, Kansas, by T. K. Larson, L. E. Somers, D. M. Penny, and M. L. Hargrave. Technical Report 99/47/June 1999. U.S. Army Construction Engineering Research Laboratory, Champaign, Illinois. Hargrave, M. L., S. M. Alt, D. Maki, and T. R. Pauketat 2002 Using Magnetic Gradiometry to Investigate an Early Mississippian Settlement. Poster presented at the 59th Annual Meeting of the Southeastern Archaeological Conference, Biloxi, Mississippi. Hargrave, M. L., D. K. Jackson, M. Farkas, and R. Dalan 2004 Geophysical Investigation of a FortiďŹ ed Late Prehistoric Settlement near Chicago. Poster presented at the 69th Annual Meeting of the Society for American Archaeology, Montreal, Canada. Hargrave, M. L., L. E. Somers, T. K. Larson, R. Shields, and J. Dendy 2002 The Role of Resistivity Survey in Historic Site Assessment and Management: An Example from Fort Riley, Kansas. Historical Archaeology 36(4):89â&#x20AC;&#x201C;110. Idol, B., with contributions by D. Maki and D. Leigh 2003 Phase II Archeological Evaluations of 15 Sites at Fort Bragg, Cumberland, Harnett, and Hoke Counties, North Carolina. Submitted to Engineer Research and Development Center Construction Engineering Research Laboratory, by TRC Garrow Associates, Inc., Durham, North Carolina. Idol, B., and S. Pullins 2001 Phase II Archaeological Evaluation of 25 Sites, Fort Bragg and Camp MacKall, Cumberland, Harnett, Hoke, and Moore Counties, North Carolina. Submitted to Engineer Research and Development Center Construction Engineering Research Laboratory, by TRC Garrow Associates, Inc., Durham, North Carolina.

Ground Truthing the Results of Geophysical Surveys ~ 303 Jackson, D. K. 2003 Introduction to the Hoxie Farm Site (11CK4) and the ITARP Investigations. Paper presented in the symposium “The ITARP Hoxie Farm Site Investigations: Preliminary Observations on a Complex, Late Prehistoric Site in the Chicago Area,” 49th Annual Midwest Archaeological Conference, Milwaukee. Jones, G., D. Maki, and M. L. Hargrave 2003 Ground Penetrating Radar Investigations of Four Historic Cemeteries on Fort Bragg, NC. Archaeo-Physics Report of Investigations No. 48. Submitted to Engineer Research and Development Center Construction Engineering Research Laboratory, by Archaeo-Physics, LLC, Minneapolis. Kreisa, P. P., and G. R. Walz 1997 Archaeological Test Excavations of Four Sites at Fort Riley, Riley and Geary Counties, Kansas. Public Service Archaeology Program Research Report No. 29. Submitted to U.S. Army Construction Engineering Research Laboratory, Champaign, Illinois, by University of Illinois at Urbana-Champaign. Kvamme, K. L. 2000 Geophysical Explorations at Double Ditch Indian Village State Park (32BL8), Burleigh County, North Dakota, 1997. Submitted to the State Historical Society of North Dakota, Bismark. 2001a Current Practices in Archaeogeophysics: Magnetics, Resistivity, Conductivity, and Ground-Penetrating Radar. In Earth Sciences and Archaeology, edited by P. Goldberg, V. Holliday, and R. Ferring, pp. 353–384. Kluwer/Plenum, New York. 2001b Final Report of Geophysical Investigations Conducted at the Mandan/Arikara Village, Fort Clark State Historic Site (32ME2), 2000. Submitted to the PaleoCultural Research Group, Flagstaff, Arizona, and the State Historical Society of North Dakota, Bismark. 2002a Final Report of Geophysical Investigations Conducted at the Fort Clark Trading Post, Fort Clark State Historic Site (32ME2), 2000–2001. Submitted to the PaleoCultural Research Group, Flagstaff, Arizona, and the State Historical Society of North Dakota, Bismark. 2002b Report of Geophysical Findings at the Double Ditch State Historic Site (32BL8): 2001 Investigations. Submitted to PaleoCultural Research Group, Flagstaff, Arizona, and the State Historical Society of North Dakota, Bismark. 2003a The Archaeological Remote Sensing Library of Geophysical Imagery. http:// WWW site maintained by Dr. Kenneth Kvamme, Archeo-Imaging Lab, Department of Anthropology and Center for Advanced Spatial Technologies, University of Arkansas, Fayetteville. 2003b Geophysical Surveys as Landscape Archaeology. American Antiquity 68(3):435–458.

304 ~ Michael L. Hargrave Larson, T. K., and D. M. Penny 1998 Results of Archaeological Ground Truthing Investigations at Historic Sites, Fort Riley, Kansas. Submitted to U.S. Army Construction Engineering Research Laboratory, Champaign, Illinois, by LTA, Inc., Laramie, Wyoming. Mathys, A., and D. L. Maki 1997 A Geophysical Survey at Fort Leonard Wood, Missouri. Reports of Investigation 474, IMA Consulting, Inc. Submitted to U.S. Army Construction Engineering Research Laboratory, Champaign, Illinois, by Institute for Minnesota Archaeology, Minneapolis. Pauketat, T. 2003 Resettled Farmers and the Making of a Mississippian Polity. American Antiquity 68(1):39–66. Rion, G. P. 1960 Army City, Kansas: The History of a World War I Camptown. Unpublished Master’s thesis, Department of History, Political Science, and Philosophy, Kansas State University of Agriculture and Applied Science, Manhattan. Scollar, I., A. Tabbagh, A. Hesse, and I. Herzog 1990 Archaeological Prospecting and Remote Sensing. Topics in Remote Sensing, No. 2, G. Hunt and M. Rycroft, series editors. Cambridge University Press, Cambridge. Somers, L. E., and M. L. Hargrave 2001 Magnetic and Resistivity Surveys of Four Sites. In Phase II Archaeological Evaluation of 25 Sites, Fort Bragg and Camp MacKall, Cumberland, Harnett, Hoke, and Moore Counties, North Carolina, by B. Idol and S. Pullins, pp. 348–365. Submitted to Engineer Research and Development Center Construction Engineering Research Laboratory, by TRC Garrow Associates, Inc., Durham, North Carolina.


A Comparative Guide to Applications Jay K. Johnson

In this concluding chapter, I would like to revisit a few of the central questions relating to the incorporation of remote sensing into the protocol for cultural resource management (CRM) archaeology in the United States. As was intended, most of these questions have been addressed in some detail in the preceding chapters. The first question is, of course, is there a place for remote sensing within the laws and regulations that guide CRM activities? Lockhart and Green (this volume) make it clear that while remote sensing is not explicitly mentioned, many of the CRM laws and standards set goals and guidelines that can best be satisfied using remote sensing techniques. In Great Britain, the change in regulations known as Planning Policy Guideline 16 that led to a florescence there in geophysical survey is only slightly more explicit, including the phrase “developers may wish to carry out geophysical surveys as part of their own initial archaeological assessment” (Gaffney and Gater 2003:22). According to Gaffney and Gater (2003:22), the more critical statement in the guideline is that “anything that is asked of the developer is ‘fair, reasonable and practicable.’” British developers have interpreted this statement in terms of cost effectiveness. This leads to the question of whether remote sensing site survey is more effective and less expensive than the more traditional techniques used by CRM-based archaeology in the United States. The simulation that Bryan Haley and I report in Chapter 3 makes a strong argument for the likelihood that better and less expensive data recovery would result from

306 ~ Jay K. Johnson the use of geophysical techniques on large, complex Mississippian sites. It is considerably more effective than shovel testing, a technique that has been adopted because there has been no other practical alternative at sites where ground visibility is limited. However, archaeologists have questioned shovel testing as a site discovery and evaluation technique in the United States for many years (Kintigh 1988; Krakker et al. 1983; Shott 1985, 1989). Kvamme (2003a:453) reaches similar conclusions about the advantages of geophysical survey over shovel testing. On Mississippian sites, remote sensing is likely to give a much better view of the structure and integrity of the site than controlled surface collections. And it is cheaper. It is also likely, given the level of response that we have gotten on these kinds of sites, that random or stratified random test pits and mechanical stripping of the plow zone could be dispensed with or radically reduced in scope and the fieldwork could move directly to block excavation. This would also save money and improve the quality of the data recovery. However, what about Woodland hamlets and Archaic camp sites? Would our simulation have worked as well on these classes of sites? At this stage in the development of remote sensing applications in North America, the answer would have to be no. When you download the gradiometer data from a Mississippian site and generate an image that includes a series of rectangular patterns, 3–4 m on a side, it doesn’t take much experience to conclude that the remains of burned, clay-plastered houses are preserved beneath the plow zone at that site. However, what if the structures at the site are much less substantial and the major intact cultural features are irregular pits dug into the subsoil? These pits are also likely to be detected using the geophysical techniques we have described. However, their identification will not rest primarily on pattern recognition. In order to distinguish them from similar, noncultural features at the site, a fundamental shift will be required in the way that remotely sensed data are interpreted. As was emphasized in the introduction to the chapter on multiple instrument applications (Chapter 11), rather than relying on shape characteristics as has been the case in most archaeological geophysical research in the past, we must start doing “spectral” analyses of features. That is, we need to compare the way in which the feature is measured using as many instruments as possible. To give a simple example, the pit features relating to the Woodland component at the Walford site in western Mississippi show up in both the gradiometer image and the susceptibility image (Figure 13.1), as do the burned Mississippian period houses at the Parchman Place Mounds. However, the ratio of remanent magnetism to magnetic susceptibility, which can be approximated by dividing the gradiometer values by the susceptibility values for each pixel in the images, is much different for the two kinds of features. Fortunately, the techniques for comparing differences in the way a portion of an image responds to different portions of the electromagnetic spectrum are well developed in digital remote sensing and those techniques may be readily borrowed using existing software. Granted all of the above, the question of which and how many instruments to use on a specific project remains to be answered. We are still learning, particularly on

A Comparative Guide to Applications ~ 307

Figure 13.1. Magnetic gradient (top) and magnetic susceptibility (bottom) images of the Walford site showing pit feature locations.

sites that do not contain well-defined structural remains. However, there is a considerable body of literature on the capabilities and limitations of the various instruments, most of which has been summarized in the applications chapters of this book. It all comes down to one fundamental factor: contrast. If the features that you are looking for contrast with the soil in which they are buried, you have a good chance of finding them. It is like looking for your SUV in a suburban mall parking garage—the job would be considerably easier if you were driving a ’59 Cadillac instead. Of course, we are presuming that there is sufficient light and that you are close enough to your car to see it. This introduces the concepts of sensor type and the depth below surface of the feature. Other factors, depth to bedrock and electromagnetic interference, must also be considered in evaluating the applicability of the instrument, but I won’t strain the analogy further. Although the chapters were organized by technique, the following discussion will emphasize site characteristics. While similar presentations have used tables (David 1995:table 2; Kvamme 2001:table 13.1), the question of which sensor to use under what conditions is complex enough that a narrative is more appropriate. David (1995: table 1) also presents a key in which a series of questions and responses guide the user through a flowchart to the appropriate instrument or combination of instruments. This suggests that decision support system software similar to that which Somers and

308 ~ Jay K. Johnson Hargrave (2003) developed to aid in geophysical survey design could be written. However, it does not seem that we have reached that stage of maturity in our application of the techniques, particularly on prehistoric sites.

Site Setting Bedrock The depth of the bedrock will obviously impact the utility of remote sensing, but it affects most sensors in the same way. However, when the bedrock is igneous and close to the surface, its magnetic signature will likely overwhelm those of cultural features and magnetometry will not be useful. Similarly, glacial soils with igneous gravels will be too “noisy” for magnetic survey. Soil Texture In general, fine-grained soils are better for resistivity, electromagnetic (EM) conductivity, and thermal infrared prospection. This is the direct result of the ability of such soils to retain moisture since all of these techniques, thermal infrared in particular, depend primarily on the differential distribution of moisture. On very dry sites, when the surface layer is too dry for the resistivity probes to make contact, it is sometimes still possible to use an EM conductivity meter because that instrument operates without the need to couple with the ground. Of course, if the texture and moisture-retaining characteristics of the cultural feature are the same as those of the surrounding soil, it will not likely be detected using these instruments. On the other hand, precisely because they do not drain as well, fine-grained soils are less likely to show crop marks that sometime reveal the location of buried features that might be detected using aerial photography or air- and satellite-borne multispectral sensors. Fine-grained soils are generally not appropriate for ground-penetrating radar (GPR) survey because the signal cannot penetrate very far. Likewise, saturated soils, particularly soils saturated with salt water, will attenuate the signal and yield poor GPR results. If there is a good deal of natural variation in soil texture—a complex series of filled channels or gravel cross bedded with silt or erosional remnants and bedrock outcrops, for example—the variation in the background may mask the differences between the cultural features and the background, particularly if the instrument that is used is sensitive to some aspect of the natural variation at the site. It is often possible to filter out the background noise, highlighting the cultural features. For example, resistance data that we collected at the Presidio de Santa Rosa in the sand dunes near Pensacola, Florida, show differences in elevation very nicely because of differences in soil moisture. Bryan Haley applied a high pass filter to enhance the edges of potential cultural features and a logarithmic compression to minimize broad trends in the data. As a result, several linear features that may relate to the Spanish colonial occupation of the site were brought out (Figure 13.2).

A Comparative Guide to Applications ~ 309 A Horizon Gaffney and Gater (2003:141) end their discussion of prehistoric case studies with the conclusion that “magnetometry is the preferred technique for identifying near surface cut, or negative, features that are commonly found on prehistoric sites.” In this they are referring to the fact that the natural processes of soil formation result in a higher magnetic susceptibility in the uppermost soil horizon. When this is combined with the remanent magnetism created by several decades of campfires, the humus layer in a relatively undisturbed location shows a magnetic signature that is higher than that of the soils below it. When a ditch is dug through this layer and refilled with a mixture of A and lower horizon soils, it will have a diluted magnetic response and show as a negative feature. We have seen this effect on Figure 13.2. A portion of the resistance imagery from the Presinineteenth-century cemeter- dio de Santa Rosa showing the utility of a filter. ies in northern Mississippi, where graves show up as oblong areas of low return in the gradiometer image (Figure 13.3). Therefore, on sites where the A horizon has not been eroded or plowed to oblivion, measurements of total field magnetism (gradiometer) or the magnetic susceptibility portion of the EM38 signal (EM measurements using the in-phase configuration) are likely to be informative. This same feature, the magnetic enhancement of the A horizon in terms of susceptibility, has been used by Dalan (2001; Dalan and Banerjee 1998; Dalan and Bevan 2002) to effectively map buried land surfaces and, sometimes, mound construction stages.

310 ~ Jay K. Johnson

Figure 13.3. Magnetic gradient image of the Confederate cemetery on campus at the University of Mississippi.

Ground Cover It is often said that the ideal setting for a geophysical survey would be a golf course or a park. However, it would need to be a golf course without a sprinkler system and a park without picnic tables or light poles. For gradiometer and EM surveys, metal artifacts that are not related to the cultural component being investigated can overwhelm the signals of interest. For example, the gradiometer and conductivity results from a recent survey of Old Mobile were rendered ambivalent on a portion of the site that had been bush hogged before the metal pin flags had been pulled (Clay 2002). On this site, resistivity or GPR is more likely to produce useful results. On the other hand, the scatter of metal debris is likely to allow a general definition of structure location and disposal patterns on historic sites. Incidently, allow me to take this occasion to point out the havoc that metal pin flags bring about for geophysical survey. As convenient as they are, if the site is likely to be explored using remote sensing techniques that are sensitive to metal, these flags should be scrupulously avoided. So, if the metal debris and structures likely in public areas like parks pose difficulties, perhaps agricultural fields would be better locations in which to conduct geophysical surveys. This is generally true, but surface conditions are always a consideration. If the field has just been turned, instruments that require a consistent orientation (gradiometers) or a consistent distance from the surface (gradiometers and conductivity

A Comparative Guide to Applications ~ 311 meters) are likely to show the plow scars because of regular deviation in the way this instrument is carried during the survey as a result of the rows and furrows. Although regular noise such as this can be removed using filters, subtle cultural features might also be lost. Of course, steep or rocky terrain will have a similar effect, slowing the survey or, in extreme cases, making it impossible. It is likewise possible to work around trees and bushes but, at best, such obstacles will slow the progress of the work. This is particularly true for resistivity surveys during which the cable to the remote probes will wrap around trees, making it necessary to backtrack often. Because trees and bushes have a local effect on the distribution of ground water, they can also impact resistivity and conductivity measurements. Dummy variables have to be inserted for locations where large trees or hedgerows make it impossible to take a reading, and this takes time. Tree roots on the surface of the ground can make it difficult to keep the GPR antenna in contact with the surface and the signal will be lost. Roots below the surface will create echoes in the GPR imagery that may mask the return from cultural features in those locations. Ground cover is particularly critical when airborne and satellite sensors are used. In our experiments with a blimp-mounted thermal infrared sensor (Haley et al. 2002), for example, bare earth produced the best results. If the features of interest are foundation trenches, ditches, or buried foundation stones that have a local effect on soil moisture, they might be expressed in differences in the way vegetation grows over the top of them. Likewise, we are beginning to recover evidence that the local enhancement of the A horizon resulting from the organic debris associated with human occupation can be detected in terms of subtle differences in susceptibility readings. This might also be detected in terms of crop vigor using airborne sensors. However, an earlier attempt to do just that proved to be unsuccessful (Johnson 1991). All of these differences are more likely to be detected if the site is covered in the same vegetation. Otherwise, differences in the way the different plants appear in the sensor image are likely to be stronger than differences caused by buried cultural features. Sometimes, however, differences in vegetation mark the location of local changes in soil characteristics that have resulted from human occupation, the shell mound–specific vegetation of the Louisiana marshes, for example, and such sites can be detected using airborne sensors (Giardino and Haley, this volume). Kinds of Targets Keeping in mind the general caveat that the feature must contrast with the soil matrix in order to be detectable, the nature of the cultural deposit that is anticipated at the site will dictate the class of instrument that is likely to be informative. For this reason, the first choice on many prehistoric and historic sites in the South is often the gradiometer. Because of a general lack of igneous rocks in much of the area, natural deposits are generally quiet in terms of disturbances in the magnetic field. Therefore, burning, a regular event at human settlements, is likely to be detected. This is particularly so when

312 ~ Jay K. Johnson the burning involves materials that contain a fair amount of iron, as do most clays, for example. Floors, hearths, and the daub rubble from late prehistoric burned houses are a fine example of this phenomenon. Brick foundations and brick burial crypts are likewise easy to find using a gradiometer. Ferrous metals also affect the magnetic field and can be detected using a gradiometer. This can sometimes be quite informative; for example, Kvamme (2003a, 2003b) was able to distinguish between contact period Plains pithouses that have a lot of iron debris and those that do not, thereby being able to begin to ask questions about the relative interaction with the nearby trading post without having to excavate. Of course, if you are looking for a large iron object, say the boiler from a steamboat, a magnetometer would be the instrument of choice. Features that are marked by the relative absence or presence of humus deposits can also be detected using a gradiometer. Pits or ditches filled with trash or topsoil can give a subtly higher reading using a magnetometer. A grave shaft in which a relatively intact A horizon deposit is diluted by mixing with the subsoil from deeper in the pit can sometimes be detected as a magnetic low. These same features can also be detected by measuring magnetic susceptibility. Pits or ditches in which a contrasting fill retains soil moisture can be detected using a resistivity or conductivity meter. Although the two instruments measure soil characteristics that are theoretically reciprocal, they do it in very different ways, often producing images with subtle differences. Moreover, on a site with a substantial amount of metal debris, soil moisture differences in the EM conductivity meter readings are likely to be hidden by the signals from the metal. Resistivity readings are not affected by metal. More than any other geophysical instrument, GPR relies entirely on the nature of the contrast between features and background. The EM signal that is generated by the antenna is reflected by a change in the way in which the soil conducts that energy. The strength of the return is determined by the magnitude of the change and the nature of the boundary. It is easy to imagine the boundary between the top of a buried stone wall and the overlying soil creating a reflection. It is somewhat counterintuitive to realize that the boundary between the bottom of the wall and the soil will also be reflected. Any change in the speed at which the signal passes through the soil, so long as it is substantial and relatively abrupt, will create an echo. Therefore, under ideal conditions it is sometimes possible to derive a three-dimensional image of a feature, top, bottom, and sides. Historic sites with walls, floors, pipes, and other features often produce the kinds of abrupt boundaries that are easy to detect using GPR. Prehistoric sites are more difficult. Nevertheless, GPR imagery is often useful on prehistoric sites, particularly those with structural remains. Overriding all considerations of contrast are the questions of size and depth. Of course, if the features at a site are large and shallow, they will be easier to detect than otherwise. Under normal conditions and using standard configurations, the effective depth for gradiometry, conductivity, and resistance readings is around 1–2 m. Given the right soils, GPR can record features at a substantially greater depth. However, for

A Comparative Guide to Applications ~ 313 deeper searches, a lower frequency antenna must be used and there is a consequent increase in the lower limit of detectable feature size. Depth can also be controlled with EM conductivity readings by using an instrument with greater coil separation. Likewise, probe spacing controls the depth of reading for resistivity. Multiple Instrument Applications As Hesse (1999) points out, the decision of which and how many instruments to use is a strategic one. As indicated in the preceding discussion, sometimes you have no choice; because of the nature of the location and the kinds of archaeological remains that you anticipate, only one instrument will work. This is generally not the case. But given the limitations of time and money that characterize all archaeological fieldwork, particularly in a CRM environment, there is a trade-off between the number of instruments you use and the extent of the survey. This is not an easy choice. When doing geophysical survey as part of a data-recovery project on highway projects, we regularly ask permission to extend the survey on either side of the right-ofway. This is because the recognition of patterns within the area to be excavated often depends on a broader view of the site. Gaffney and Gater (2003:91–92) provide a dramatic example of this problem in which a “Roman road” became a portion of an abandoned soccer field when the survey was expanded. Although we are unlikely to find either on North American sites, palisades, footpaths, large structures, and many other large linear features are much easier to recognize when seen in broad view. It is often useful to include a portion of the area surrounding the site in the imagery in order to be able to detect the contrast between where the occupation is and where it is not. So, covering a large area with one or two instruments can be a useful strategy. On the other hand, the recognition of cultural features, particularly on Woodland and Archaic sites, often depends on examining the differences between the ways the features are recorded using multiple techniques. And, of course, the multivariate techniques that are showing promise in the search for features on camp sites and small villages are based on the use of multiple instruments. However, the application of multivariate analyses involves more time to produce results than is usually allocated in the field. The ideal situation would be to cover all of the site with all of the instruments. Few of us have the time and money to operate in the ideal world. Therefore, a staged approach is useful. An initial review of the site’s characteristics, primarily the anticipated soil and feature types, will suggest the instruments that are likely to produce results. If you know enough about the site to be able to identify areas where features are likely to be found, trial applications should be run on those locations and the imagery evaluated in the field if possible. If the results match the expectations, the survey can then be expanded to include as much of the site as the budget allows, concentrating, of course, on the instruments that produced the best test results. If possible, the results from this broad-scale survey should be used to target areas of interest that can then be covered with other instruments or the same instrument using close-interval sampling.

314 ~ Jay K. Johnson

Toward an Integrated Application Ultimately, remote sensing should play a role in all phases of CRM archaeology, from survey to evaluation to mitigation. As the preceding chapters have demonstrated, remote sensing, particularly geophysical techniques, can contribute a great deal in terms of site assessment and excavation. The question is, how much use is it in site discovery? Aerial photography and geophysical prospection are routinely used in public archaeology in Great Britain (Clark 1996; Gaffney and Gater 2003). However, as has already been pointed out, the archaeological record in that country is full of site types that respond particularly well to these techniques. Many contain well-defined foundations, ditches, and roads that are revealed by obvious geometric patterns in the aerial photographs and geophysical images. Brick and pottery kilns are regular features of many settlements and, along with associated metal debris, they create a signal strong enough so that highway rights-of-way can be walked using a gradiometer with the goal of identifying “noisy” areas for more intensive survey. A similar use of the instrument in North America might help identify historic sites but would be more likely to pinpoint tractor parts and abandoned culverts. It seems to me that there is likely to be a role for remote sensing in Phase I surveys in North America, but this will rely on airborne digital multispectral sensors and require a considerable refinement in our techniques. My own efforts in this direction over the past 15 years provide a personal perspective on the difficulties and potential rewards. The first application relied on Landsat TM imagery in an attempt to refine site discovery and settlement pattern analysis as part of a large Phase I survey in north Mississippi. There were a number of difficulties to overcome, not the least of which was the mastery of what, by today’s standards, was difficult and cumbersome software. The largest problem was image resolution. Landsat TM images consisted of seven bands of data spread across the visible, near infrared, and thermal infrared spectrum. All but the thermal band had a 30-m resolution. Thermal infrared, because of its potential value in military applications, was degraded to a resolution of 80 m. At the time, we wished for better thermal resolution but, as it turns out, 30-m pixels are just too large to allow the detection of site-specific signatures. We were able to detect broad-scale environmental patterns that allowed the designation of high-probability areas (Johnson et al. 1988). In particular, on bare earth agricultural fields in the stream bottoms, we were able to map the terraces on the basis of the distribution of fragipan soils, which had a distinct signature in the satellite imagery. This designation was based on spectral characteristics, not spatial patterning, and was only possible when ground cover was absent. These are two important points. The next project involving remote sensing focused on protohistoric settlement in the Black Prairie of northeastern Mississippi (Johnson 1991). In order to overcome the problem of resolution, we switched to an airborne sensor. TIMS (Thermal Infrared Multispectral Sensor), a 6-band sensor with emphasis on the thermal infrared spectrum, was used to fly two transects across the survey area. Because the sensor is closer to the earth,

A Comparative Guide to Applications ~ 315 images with much higher resolution were recorded; pixel size was 5 m. In contrast with the general pattern for most of the survey area covered in the previous project, protohistoric settlement in the Prairie is confined to the upland. There were, therefore, large areas of pasture. Sites that were located in the terrestrial survey that was part of the project could be used to search for spectral differences in the ground. Two problems may have contributed to the failure of this experiment: the narrow range in the spectral coverage of the sensor and the relative variability in the ground cover. Different grasses with different levels of maintenance made up the pasture sample in the study area. Airborne remote sensing was a major component of our initial work at the Hollywood Mounds and represented our first attempt to apply remote sensing techniques on large Mississippian period sites in the Yazoo Basin of northeastern Mississippi. The advantage this time was that the site is owned by the Mississippi Department of Archives and History and we could control the ground-cover conditions. We chose bare earth and had the site thoroughly disced before it was surveyed using an ATLAS airborne sensor. Resolution was 2.5 m across 14 bands, including 4 visible, 2 near infrared, 2 mid-infrared, and 6 thermal infrared bands. In order to explore variation in the thermal bands, the site was surveyed at predawn, solar noon, and mid-afternoon. Although, in retrospect (Haley 2002), there is patterning in the data that can be related to the buried prehistoric structures at the site, it is subtle and we were unable to detect it using unsupervised analytical techniques (Johnson et al. 2000). Hollywood was the focus of another airborne sensor experiment in which we returned to an emphasis on thermal infrared. This time, however, we had a number of additional advantages. We owned the sensor, an Agema handheld broad-band thermal sensor, and we mounted it on a tethered blimp. This arrangement allowed us to record images whenever we wanted; the only expenses were time, travel costs, and helium. The other main advantage was that Berle Clay had already done EM conductivity and gradiometer surveys of most of the site. His images did two things. They convinced us that geophysical prospection was a direction that we wanted to take, and they pinpointed the buried structures, allowing us to test and confirm several in the field to the west and south of the big mound. We were therefore able to image specific structures in the thermal infrared tests. The features remained invisible for the first three sessions. Then it rained and one class of structures, the buried house locations, became evident. The second major structure class at Hollywood, the truncated mounds, have not yet been detected using the thermal sensor (Haley et al. 2002). Two additional sensors were used to image Hollywood from the air, an ADAR multiband sensor and a sensor built and operated by Air-O-Space. Once again, the resolution was sufficient, 0.7 m and 0.2 m, respectively. In these two instances, the site was surveyed after it had grown up in grasses. I didn’t have much hope for these experiments because of the variation in the kinds of grasses that volunteered at the site. However, Bryan Haley took it on as a thesis topic and was able to detect the arches and ovals that mark the locations of the truncated mounds (Haley 2002). The identification, however, was primarily visual and often difficult.

316 ~ Jay K. Johnson Haley and I revisited the airborne imagery from Hollywood recently (Johnson and Haley 2004, summarized in Chapter 11 of this volume). The major goal was to derive a spectral signature for the two structure classes using multivariate statistics. We were largely successful. The implications for site discovery are encouraging. As envisioned, the application would go something like this. Geophysical survey would be conducted on known sites in the survey area in order to locate features of interest. The survey area would then be surveyed using an airborne digital sensor with as many bands as possible. The feature locations from the known sites would serve as training fields for the multivariate analysis of the airborne imagery. The resultant spectral classes definitions would be reapplied to the sites in order to check the accuracy of the classification. The classification could then be applied to other portions of the survey area that had similar ground cover. In this way, high-probability areas could be designated. If I have made this sound easy, that was not my intent. I only mean to indicate that it is possible. Incidently, a recent attempt to apply similar statistics to a large Bronze Age site in Turkey as part of a thesis project was not successful (Aydin 2004). There were, however, some limitations imposed by the nature of the imagery that was used. The goal of this volume and this chapter is to provide information that will help CRM administrators make decisions about remote sensing applications. Toward that end, I would like to conclude that the use of remote sensing in site discovery, whether it is airborne or geophysical, is still in the developmental stages. There are, however, obvious implications for the future and we must continue to develop the techniques because the payoff will be substantial. Geophysical applications in site evaluation and excavation, on the other hand, are not something we have to look forward to; they are established, proven, and available. That is not to say that significant advances in these techniques are all in the past. We will continue to see improvements in instrumentation, computation, and field techniques. However, the area that holds the greatest potential for advancement in the geophysical exploration of cultural remains is multiple instrument integration. Once again, in order to be able to apply these techniques to as broad a range of sites as possible, we will need to make the fundamental shift in perspective, from the identification of features based on spatial patterning to the identification of features based on spectral patterning. This next step in the development of geophysical applications on archaeological sites will take more than sophisticated data-analysis techniques. It will take the kind of staged iteration that is described by Mike Hargrave in Chapter 12, with the remote sensing results guiding the excavation results and the excavation results being used to refine the remote sensing. The major contribution that CRM administrators can make to the future of remote sensing is to require this to take place. Much of the remote sensing that is undertaken in a CRM context today is done as a subcontract. After the images are delivered, the archaeologist who did the geophysical survey moves on to another project. There must be feedback between the excavator and the remote sensing specialist. The result will be a better understanding of the site as well as of the technology.

A Comparative Guide to Applications ~ 317 Regardless of how you assess the potential of remote sensing in terms of site discovery and of whether you feel that multivariate techniques will prove as useful as I think they will, in light of the images and discussions presented in the previous chapters of this book, it would be hard to deny the important contribution that remote sensing will make to the archaeology of the next decade. Because the potential payoff in terms of site boundary definition, feature detection, site structure, research design, and, in particular, the bottom line is likely to be substantial, the same factors that have made geophysical survey an essential part of contract archaeology in Great Britain will ensure a similar role for it in North American CRM. We are on the threshold of a new era in remote sensing applications in archaeology. It will change the way we dig in a fundamental way.

References Cited Aydin, N. 2004 Multi-Sensor Data Fusion Applications in Archaeology, Unpublished Master’s thesis, Department of Sociology and Anthropology, University of Mississippi. Clark, A. J. 1996 Seeing Beneath the Soil: Prospecting Methods in Archaeology, new ed. B. T. Batsford, London. Clay, R. B. 2002 Geophysical Survey of a Portion of Vieux Mobile, Alabama. Cultural Resource Analysts, Inc., Lexington, Kentucky. Report prepared for Dr. Gregory Waselkov, University of South Alabama, Contract Publication Series 01-198. Dalan, R. A. 2001 A Magnetic Susceptibility Logger for Archaeological Application. Geoarchaeology 16:263–273. Dalan, R. A., and S. K. Banerjee 1998 Solving Archaeological Problems Using Techniques of Soil Magnetism. Geoarchaeology 13:3–36. Dalan, R. A., and B. Bevan 2002 Geophysical Indicators of Culturally Emplaced Soils and Sediments. Geoarchaeology 17:779–810. David, A. 1995 Geophysical Survey in Archaeological Field Evaluation. Ancient Monuments Laboratory, English Heritage Society, London.

318 ~ Jay K. Johnson Gaffney, C., and J. Gater 2003 Revealing the Buried Past: Geophysics for Archaeologists. Tempus, Gloucestershire, Great Britain. Haley, B. S. 2002 Airborne Remote Sensing, Image Processing, and Multisensor Data Fusion at the Hollywood Site, a Large Late Mississippian Mound Center. Unpublished Master’s thesis, Department of Sociology and Anthropology, University of Mississippi, Oxford. Haley, B. S., J. K. Johnson, and R. Stallings 2002 The Utility of Low Cost Thermal Sensors in Archaeological Research. Center for Archaeological Research, University of Mississippi, Oxford. Report prepared for the Office of Naval Research, NASA grant NAG5-7671. Hesse, A. 1999 Multi-Parametric Survey for Archaeology: How and Why, or How and Why Not? Journal of Applied Geophysics 41:157–168. Johnson, J. K. 1991 Settlement Patterns, GIS, Remote Sensing, and the Late Prehistory of the Black Prairie in East Central Mississippi. In Applications of Space-Age Technology in Anthropology, edited by C. A. Behrens and T. L. Sever, pp. 111–119. NASA, John C. Stennis Space Center, Mississippi. Johnson, J. K., and B. S. Haley 2004 Multiple Sensor Applications in Archaeological Geophysics. In Sensors, Systems, and Next-Generation Satellites VII, edited by R. Meynart, S. P. Neeck, H. Simoda, J. B. Lurie, and M. L. Aten, pp. 688–697. Proceedings of SPIE, vol. 5234. SPIE, Bellingham, Washington. Johnson, J. K., T. L. Sever, S. L. H. Madry, and H. T. Hoff 1988 Remote Sensing and GIS Analysis in Large Scale Survey Design in North Mississippi. Southeastern Archaeology 7:24–131. Johnson, J. K., R. Stallings, N. Ross-Stallings, R. B. Clay, and V. S. Jones 2000 Remote Sensing and Ground Truth at the Hollywood Mounds Site in Tunica County, Mississippi. Center for Archaeological Research, University of Mississippi, Oxford. Submitted to the Mississippi Department of Archives and History.

A Comparative Guide to Applications ~ 319 Kintigh, K. W. 1988 The Effectiveness of Subsurface Testing: A Simulation Approach. American Antiquity 53:686–707. Krakker, J. J., M. J. Shott, and P. D. Welch 1983 Design and Evaluation of Shovel-Test Sampling in Regional Archaeological Survey. Journal of Field Archaeology 10:469–480. Kvamme, K. L. 2001 Current Practices in Archaeogeophysics: Magnetics, Resistivity, Conductivity, and Ground-Penetrating Radar. In Earth Sciences and Archaeology, edited by P. Goldberg, V. Holliday, and R. Ferring, pp. 353–384. Kluwer/Plenum, New York. 2003a Geophysical Surveys as Landscape Archaeology. American Antiquity 68(3):435– 458. 2003b Multidimensional Prospecting in North American Great Plains Village Sites. Archaeological Prospection 10:131–142. Shott, M. J. 1985 Shovel-Test Sampling as a Site Discovery Technique: A Case Study from Michigan. Journal of Field Archaeology 12:458–469. 1989 Shovel Test Sampling in Archaeological Survey: Comments on Nance and Ball, and Lightfoot. American Antiquity 54:396–404. Somers, L. E., and M. L. Hargrave 2003 Geophysical Surveys in Archaeology: Guidance for Surveyors and Sponsors. Construction Engineering Research Laboratory, U.S. Army Corps of Engineers, Champaign, Illinois.


R. Berle Clay is a Senior Project Archaeologist at Cultural Resource Analysts, Inc., having served as State Archaeologist and Director of the Office of State Archaeology at the University of Kentucky from 1976 to 1997. He received his Ph.D. in anthropology from Southern Illinois University Carbondale. Research specialties include ceramic analysis, quantitative methods, and geophysical survey. Lawrence B. Conyers is an Associate Professor of Anthropology at the University of Denver who specializes in geological and geophysical archaeological methods. He received his Ph.D. from the University of Colorado at Boulder, where he made major advances in the use of ground-penetrating radar methods for the discovery and mapping of buried archaeological sites. Rinita A. Dalan is an Associate Professor of Anthropology and Earth Science at Minnesota State University Moorhead. She received her Ph.D. in ancient studies from the University of Minnesota. Her research interests focus on the exploration of geophysical and soil magnetic methods as they apply to landscape research and studies of humanenvironment interactions. Marco Giardino is a scientist in the Earth Science Applications Directorate, NASA, Stennis Space Center. He has a Ph.D. in anthropology from Tulane University. Most of his fieldwork has taken place in the Southeast and his research interests include ground-penetrating radar, ceramic analysis, and digital airborne imagery applications. Thomas J. Green is the Director of the Arkansas Archeological Survey. A unit of the University of Arkansas System, the Survey is a statewide research, public service, and educational institution with 10 research stations in Arkansas. Green received a Bachelor’s degree in anthropology from the University of Southern California in 1968 and a Ph.D. in anthropology from Indiana University in 1977.

322 ~ Contributors Bryan S. Haley received his Master’s in anthropology from the University of Mississippi and is a research associate there. He is interested in Southeastern archaeology in general and remote sensing in particular. Michael L. Hargrave is a principal investigator at the Engineer Research and Development Center/Construction Engineering Research Laboratory, where he works extensively with remote sensing applications in archaeology. He has a Ph.D. in anthropology from Southern Illinois University Carbondale. Jay K. Johnson is a Professor of Anthropology and the Director of the Center for Archaeological Research at the University of Mississippi. He received his Ph.D. from Southern Illinois University Carbondale. Research interests include remote sensing, GIS, lithic analysis, and ethnohistory. Kenneth L. Kvamme received his Ph.D. in anthropology at the University of California at Santa Barbara. He is an Associate Professor of Anthropology at the University of Arkansas and the Director of the Archeo-Imaging Lab. Recent fieldwork has focused on the Middle Missouri River villagers of the Dakotas. He has published extensively on GIS, remote sensing, geophysical prospecting, quantitative methods, human spatial behavior, and lithic technology. J. J. Lockhart is Coordinator of the Computer Services Program for the Arkansas Archeological Survey. He holds a Master’s in geography and is a Ph.D. candidate in environmental dynamics at the University of Arkansas, Fayetteville. His research interests include integrated data management applications, geographic information systems, remote sensing, and cultural landscape analysis. Lewis Somers is the owner of Geoscan Research (USA) and a joint owner of ArchaeoPhysics LLC. A Ph.D. in physics and a great deal of experience in the archaeology of two continents have contributed to his success in the development of software and hardware specifically tailored for archaeological applications.  

Remote Sensing in Archaeology An Explicitly North American Perspective Edited by Jay K. Johnson The University of Alabama Press, Tuscaloosa...

Read more
Read more
Similar to
Popular now
Just for you